Synapse formation and growth are tightly controlled processes. How synaptic growth is terminated after reaching proper size remains unclear. Here, we show that Leon, the Drosophila USP5 deubiquitinase, controls postsynaptic growth. In leon mutants, postsynaptic specializations of neuromuscular junctions are dramatically expanded, including the subsynaptic reticulum, the postsynaptic density, and the glutamate receptor cluster. Expansion of these postsynaptic features is caused by a disruption of ubiquitin homeostasis with accumulation of free ubiquitin chains and ubiquitinated substrates in the leon mutant. Accumulation of Ubiquilin (Ubqn), the ubiquitin receptor whose human homolog ubiquilin 2 is associated with familial amyotrophic lateral sclerosis, also contributes to defects in postsynaptic growth and ubiquitin homeostasis. Importantly, accumulations of postsynaptic proteins cause different aspects of postsynaptic overgrowth in leon mutants. Thus, the deubiquitinase Leon maintains ubiquitin homeostasis and proper Ubqn levels, preventing postsynaptic proteins from accumulation to confine postsynaptic growth.
A synapse is a specialized structure where signals are transmitted from a neuron to another neuron or other target cells such as muscles. Proper synapse formation is prerequisite to building functional synapses and constructing neuronal circuits. Synapse abnormalities are suggested to induce neurological and psychological disorders such as autism spectrum disorders and fragile X syndrome (van Spronsen and Hoogenraad, 2010). Formation of postsynapses requires coordinated formation of several specialized structures. One prominent postsynaptic feature at neuromuscular junctions (NMJs) is the extensively folded muscular membranes. Specialized folding of postjunctional membranes is thought to increase the area exposed to the synaptic cleft and ensure the effectiveness of neuromuscular transmission (Sanes and Lichtman, 1999; Wu et al., 2010). In addition to membrane specializations, the postsynaptic density (PSD) is also a common element whose size requires proper control. The PSD contains scaffolding proteins that recruit signaling protein complexes and neurotransmitter receptors, matching precisely the presynaptic active zones (Feng and Zhang, 2009; Sheng and Hoogenraad, 2007). Formations of postsynaptic membrane and PSD are tightly controlled and coordinated yet these processes remain elusive.
The Drosophila NMJ is a model to study synapse formation and activity-dependent synapse remodeling (Collins and DiAntonio, 2007; Ruiz-Cañada and Budnik, 2006). Synaptic boutons are swollen structures of axonal terminals embedded in highly folded muscular membranes called the subsynaptic reticulum (SSR) and each bouton contains tens of neurotransmitter release sites paired with PSDs. During larval development, the SSR and the PSD concomitantly form and gradually increase their sizes. Two crucial factors, postsynaptic density protein-95/Discs large (Dlg) localized at the SSR and Drosophila p21-activating kinase (dPak) localized at the PSD, regulate SSR formation (Albin and Davis, 2004; Budnik et al., 1996; Lahey et al., 1994). At the PSD, two types of localized glutamate receptors (GluRs), IIA and IIB, appear in distinct GluR clusters (Marrus et al., 2004). The abundance of GluRIIA at the PSD is regulated by PSD-localized dPak and the SSR-localized NF-κB complex, NF-κB/Dorsal (Dl), IκB/Cactus (Cact) and IRAK/Pelle (Pll) (Albin and Davis, 2004; Heckscher et al., 2007; Parnas et al., 2001; Zhou et al., 2015). Thus, the postsynaptic protein could localize at either SSR or PSD, and confer growth regulation on SSR, PSD or both.
Ubiquitination plays essential roles in various cellular processes including synaptic growth (DiAntonio et al., 2001). Ubiquitin species are dynamically balanced among free and substrate-conjugated forms of mono-ubiquitin and ubiquitin chains. Ubiquitin homeostasis, i.e. the maintenance of diverse ubiquitin species in proper proportions and levels, is regulated in cellular growth and differentiation (Hallengren et al., 2013; Kimura and Tanaka, 2010). Deubiquitinases (DUBs), a large superfamily of ubiquitin regulators, participate in the dynamic equilibrium of ubiquitin species. While some DUBs process newly synthesized ubiquitin precursors for ubiquitin supply, others recycle ubiquitin by cleaving ubiquitin chains from protein substrates prior to proteasomal degradation. USP5, the focus of this study, is dedicated to disassembly of free ubiquitin chains for recycling (Hochstrasser, 2009; Komander et al., 2009). Physiologically, heat shock stress in yeast causes a reduction of the mono-ubiquitin level. To compensate for ubiquitin depletion, the level of the DUB Doa4 is elevated, leading to an increase in the mono-ubiquitin level by cleaving free ubiquitin chains (Kimura et al., 2009). The ataxia mice axJ, carrying mutations in the DUB USP14, displayed nerve swelling and abnormal neurotransmission at NMJs. The defects are caused by a reduction in the ubiquitin level as lower ubiquitin levels were detected in the mutant mice and introducing an ubiquitin transgene suppressed the axJ phenotypes (Chen et al., 2011, 2009). Thus, regulation of the ubiquitin level is a critical step in synapse development and for preventing neurological disorders.
Drosophila USP5/Leon is essential to maintain ubiquitin homeostasis during tissue formation and controls activation of apoptosis and the JNK pathway during eye development (Fan et al., 2014; Wang et al., 2014). In this study, we characterized the role of Leon in postsynaptic growth after synapse formation. In leon mutants, while the presynapse maintains normal morphology, the postsynapse overelaborates, displaying expanded SSR, enlarged PSD and excess PSD-localized GluR clusters. Free ubiquitin chains and ubiquitinated substrates accumulate in leon postsynapses, revealing defects in ubiquitin homeostasis. Genetic analysis shows that accumulations of several postsynaptic proteins accounts for overelaborated postsynaptic structures. The ubiquitin receptor Ubqn recognizes and transfers ubiquitinated substrates to the proteasome for degradation (Finley, 2009; Lipinszki et al., 2011). The Ubqn level is elevated in leon postsynapses and reducing the Ubqn level suppresses leon mutant phenotypes. Importantly, co-overexpression of free ubiquitin chains and Ubqn promotes expansion of these postsynaptic features. Thus, ubiquitin homeostasis such as disassembly of free ubiquitin chains, timely degradation of proteins, and normal function of the ubiquitin receptor Ubqn are compromised in leon mutants, leading to postsynaptic overgrowth.
Abnormal NMJ morphology in leon mutants
We examined NMJs in third-instar larvae of leon2/19-2 mutants that died at mid-pupal stages and leon1/19-2 mutants that died at the late-third larval stages. Based on the viability and Western blot analysis, leon2/19-2 is considered a hypomorphic mutant and leon1/19-2 is a close-to-null mutant (Wang et al., 2014). Both NMJ phenotypes were compared to the control w1118 that had been used to backcross all leon alleles. At control NMJs, axonal terminals immunostained for presynaptic horseradish peroxidase (HRP) branched out extensively from initial targeting sites; synaptic boutons revealed by postsynaptic Dlg staining spread evenly along axonal tracks, displaying the beads-on-a-string pattern (Figure 1A). In leon mutants, axonal terminals failed to extend and synaptic boutons aggregated, making them larger in appearance (Figure 1A). To quantify the morphological defects, we scored the number of boutons and branch lengths of NMJs at muscles 6/7. In both leon1/19-2 and leon2/19-2 mutants, the bouton numbers were significantly decreased by about 25% (Figure 1B). Total branch length was also reduced in leon1/19-2, although the reduction was not significant in leon2/19-2 (Figure 1C). In both mutants, the muscle sizes were comparable to controls (Figure 1D). Reductions in both bouton number and branch length in leon1/19-2 were completely restored by GFP-leon-GR, a genomic rescue transgene of leon, suggesting that leon is required for normal NMJ formation (Figure 1A–C).
We then analyzed whether leon is required in pre- or post-synapses by performing tissue-specific rescue of leon mutants. Expression of UAS-leon by 24B-GAL4 in muscles of leon1/19-2 restored NMJ morphology and bouton number to the wild-type level (Figure 1E,F). In contrast, expression by D42-GAL4 in motor neurons failed to rescue leon mutant phenotypes (Figure 1G,H). The postsynaptic requirement of leon was further confirmed by driving the leonRNAi transgene in individual compartments. When driven by 24B-GAL4, the NMJ morphological defect was identical to that of leon mutants, with a significant reduction in the bouton number (Figure 1—figure supplement 1A,C). These phenotypes, however, were not detected by neuronal expression of leonRNAi (Figure 1—figure supplement 1B,C).
To analyze the requirement of Leon deubiquitinating activity for NMJ growth, the UAS-ED-leon transgene that expresses enzyme-dead Leon was introduced into leon mutants. Expression of UAS-ED-leon in muscles or motor neurons failed to rescue any of the leon mutant phenotypes (Figure 1E–H). Thus, in comparison to the effective rescue by wild-type UAS-leon, this result suggests that Leon functions as a DUB in regulating NMJ development.
Immunostaining of larval tissues by anti-Leon antibodies showed that Leon was expressed ubiquitously. At NMJs, Leon was enriched within synaptic boutons (Figure 1—figure supplement 1D, arrowheads). Leon was also expressed in postsynapses with lower levels in the SSR marked by Cact (Figure 1—figure supplement 1E). These Leon expressions were almost diminished in the null 19–2 homozygous larvae, confirming that these signals represent Leon expression (Figure 1—figure supplement 1D, bottom panels). Residual puncta (arrowheads) within presynaptic boutons may represent background signals. The genomic rescue transgene GFP-leon-GR showed a similar expression pattern to endogenous Leon (Figure 1—figure supplement 1F), further confirming Leon expression at the NMJ.
Defective ubiquitin homeostasis in leon postsynapses
To examine whether ubiquitin homeostasis is disrupted in the leon mutant, we first performed Western blots to analyze ubiquitin profiles in dissected body wall muscles. By blotting with the anti-ubiquitin antibodies, we found that the levels of free ubiquitin chains were increased in leon mutants (Figure 2A). Whereas moderate increases were detected in hypomorphic leon2/19-2 (2.33 ± 0.64 folds, see Materials and methods), large increases were found in null 19-2/19-2 and close-to-null leon1/19-2 (4.25 ± 1.38 and 4.62 ± 1.61 folds, respectively). In addition, smearing signals at higher molecular weights representing ubiquitinated substrates were also increased in 19-2/19-2 and leon1/19-2 (1.88 ± 0.31 and 1.76 ± 0.37 folds, respectively). The higher molecular-weight smears, however, were increased slightly in leon2/19-2 (1.29 ± 0.78 folds). In contrast, the level of monoubiquitin was strongly increased in hypomorphic leon2/19-2 (1.9 ± 0.77 folds) but the increases were not prominent in 19-2/19-2 and leon1/19-2 (0.98 ± 0.17 and 1.25 ± 0.2 folds, respectively). Therefore, the severity in defective ubiquitin homeostasis, represented by the increased levels of free ubiquitin chains and ubiquitinated substrates, correlates with increasing leon mutant strengths.
To further confirm disrupted ubiquitin homeostasis in leon mutants, the FK2 antibody that recognizes mono- and poly-ubiquitinated substrates, and free ubiquitin chains was used to immunostain dissected larval body walls. In wild-type controls, FK2 signals were detected at NMJs and in nuclei and Z-bands of muscles (Figure 2B). In leon mutants, overall FK2 signals were enhanced, and the enhancement was very prominent at the postsynaptic sites surrounding presynaptic boutons (indicated by arrowheads, Figure 2B). Consistent with Western blot analysis, the enhancement in the FK2 signal level was more pronounced in leon1/19-2 than in leon2/19-2 (Figure 2B).
We then examined the requirement of postsynaptic Leon and its deubiquitinating activity for ubiquitin homeostasis. In leon1/19-2, muscle expression of UAS-leon, but not UAS-ED-leon, by 24B-GAL4 suppressed the enhanced FK2 signals (Figure 2B). Western blot analysis showed that 24B-GAL4-driven UAS-leon expression suppressed elevated free ubiquitin chains in leon1/19-2 (0.49 ± 0.02 folds in comparison to 24B-GAL4, leon1/19-2) and partially suppressed elevated ubiquitinated substrates (0.91 ± 0.04 folds). The partial suppression could be attributed to the limited muscle expression in the mutant. Interestingly, the expression of UAS-ED-Leon induced higher levels of free ubiquitin chains (2.17 ± 0.16 folds) and ubiquitinated substrates (2.19 ± 0.43 folds). It is possible that the enzymatic activity-deficient ED-Leon can associate with ubiquitin chains but cannot deubiquitinate them, and binding to ED-Leon sequesters ubiquitinated substrates from degradation. Taken together, these results indicate that Leon deubiquitinating activity is required for maintaining ubiquitin homeostasis in postsynapses.
Increases of postsynaptic proteins in leon mutants
Aberrant NMJ morphology and postsynaptic ubiquitin homeostasis defects in leon mutants prompted us to examine constituents of synaptic organization. Several pre- and post-synaptic proteins were analyzed for their expression patterns and levels in leon mutants. Immunostaining for Dlg that is enriched in the SSR revealed a striking phenotype. In controls, SSR-localized Dlg exhibited thin circular rings in bouton sections. In leon mutants, Dlg-localized rings were expanded, showing much greater thickness and higher protein levels (Figure 3A). The thickness of Dlg-positive zones increased by 66% in leon2/19-2 and by almost two folds in leon1/19-2 in comparison to controls (Figure 3C, left panel). We further examined SSR-localized Cact and Dl expressions, which show Dlg-resembling ring patterns (Heckscher et al., 2007). In leon mutants, circular Cact and Dl patterns were also expanded (Figure 3A). As Dlg, Cact and Dl present different aspects of postsynaptic functions, with Dlg promoting SSR formation, and Cact and Dl regulating GluRIIA abundance (Budnik et al., 1996; Heckscher et al., 2007; Lahey et al., 1994), expansions of SSR-localized Dlg, Cact and Dl suggest that the SSR is also likely expanded in leon mutants.
At NMJs, each bouton contains multiple release sites paired with discrete receptor clusters that can be revealed by the localizations of presynaptic ELKS/CAST family protein Bruchpilot (Brp), and postsynaptic dPak (Albin and Davis, 2004; Kittel et al., 2006; Wagh et al., 2006). Well-matched pairs of Brp and dPak were evenly distributed along the rim of and within bouton sections (Figure 3—figure supplement 1A). In leon mutants, the size, spacing and density of Brp puncta appeared normal (Figure 3—figure supplement 1A–B). Strikingly, the postsynaptic dPak patches were enormously enlarged, and the spacing among them was often diminished, appearing as a continuous structure in bouton sections, particularly in leon1/19-2 (Figure 3B). Quantification showed that the dPak levels were increased in leon postsynapses (Figure 3C). The normal one-to-one pairing between Brp and dPak could not be resolved in leon mutants because expanded and fused dPak patches could accommodate a few Brp puncta.
We further examined PSD-localized GluRs. GluRs are composed of four subunits, including essential subunits GluRIIC/GluRIII, GluRIID and GluRIIE and one of the two interchangeable subunits GluRIIA and GluRIIB (DiAntonio, 2006; Marrus et al., 2004). The GluRIIA and GluRIIB receptor clusters were distributed evenly in wild-type boutons. In leon mutants, GluRIIA and GluRIIB clusters were enlarged and overlapped, filling most of the space in bouton sections (Figure 3B, arrowheads). The wild-type GluRIIA cluster pattern, shown in the magnified image, was no longer present in leon mutants. Instead, GluRIIA clusters also appeared much like the dPak fusion pattern (Figure 3B, right panels). The average size of GluRIIA clusters was increased by 54% in leon2/19-2% and 89% in leon1/19-2 (Figure 3C). The GluRIIC clusters were also enlarged in leon mutants (Figure 3—figure supplement 1A). Statistically, the intensities of GluRIIA, GluRIIB and GluRIIC immunostaining were all elevated in both leon mutants (Figure 3C and Figure 3—figure supplement 1B). These results indicate that the levels of postsynaptic SSR- and PSD-localized proteins are increased in leon mutants.
GluR clustering appears after axonal terminals innervate muscles in mid embryonic stages and SSR is formed in the first instar stage (Guan et al., 1996; Harris and Littleton, 2015). The increases of SSR- and PSD-localized proteins in leon mutants could be a cumulative process from early larval stages. We compared expressions of SSR-localized Cact and PSD-localized GluRIIA in controls and leon mutants in the same stages. At leon NMJs, the thickness of Cact-positive zones and the intensity and size of GluRIIA clusters were larger than wild-type controls during 72 hr, 96 hr, and 120 hr AEL, with more severe defects in leon1/19-2 than in leon2/19-2 (Figure 3—figure supplement 2A–C). Therefore, the increases of SSR- and PSD-localized postsynaptic proteins in leon mutants are already prominent in early larval stages and progressively enhanced throughout later larval stages.
We then examined whether Leon deubiquitinating activity is required for proper postsynaptic growth. Expression of UAS-leon in muscles suppressed the expanded Dlg rings and the enlarged dPak and GluRIIA clusters in leon1/19-2. In contrast, enzyme-dead UAS-ED-leon failed to suppress these phenotypes (Figure 3D–E). Also, presynaptic expression of either UAS-leon or UAS-ED-leon had no effect on the increased size or intensity of the Dlg-positive zone, the dPak patch and the GluRIIA cluster in leon mutants (Figure 3—figure supplement 3A–B). To further confirm the requirement of postsynaptic leon for proper control of SSR- and PSD-localized protein levels, leonRNAi knockdown was performed in pre- or post-synaptic sites. When the leonRNAi transgene was driven by postsynaptic 24B-GAL4, the Dlg-positive zone, the dPak patch, and the GluRIIA cluster were increased (Figure 3—figure supplement 4A,C). These phenotypes, however, were not detected by presynaptic knockdown in neurons (Figure 3—figure supplement 4B,C). Taken together, these analyses suggest that the deubiquitinating activity of Leon is required in postsynapses to control postsynaptic protein levels.
With the normal distribution, size and intensity of Brp puncta in leon mutants, we examined expression patterns of other presynaptic proteins. Immunostaining for synaptic vesicle-associated Synapsin and cysteine string protein (CSP) that display a pattern of small spreading puncta in boutons (Fuentes-Medel et al., 2009), revealed no abnormality in leon mutants. The signals of microtubule-associated Futsch staining reach the terminal boutons in wild-type larvae, but stopped short without reaching terminals in leon mutants, suggesting a possible defect in the microtubule structure or stability in presynapses (Figure 3—figure supplement 1C).
Enlarged SSRs and synaptic membranes in leon mutants
We further examined the ultrastructures of leon mutant boutons by transmission electron microscopy. Folded SSR surrounding boutons (colored in pink in Figure 4A), electron-dense membranes (within the pairs of red arrows in Figure 4B) and T-bars (indicated by blue arrows) were analyzed in 19-2/19-2 and leon1/19-2. In both mutants, SSR areas were expanded and membrane folds were tightly packed, resulting in compact membrane layers surrounding the sectioned bouton (Figure 4A,A’). While the bouton sizes in both leon mutants were comparable to controls, the SSR areas were dramatically increased by as much as two-fold (Figure 4C). Thus, the ultrastructural analysis confirms the expansion of SSR suggested by analyzing the SSR-localized Dlg, Cact and Dl in leon mutants (Figure 3A).
The electron-dense synaptic membranes along the bouton circumference were also prominently increased in leon mutants (Figure 4B). The average length of synaptic membranes was significantly increased in both leon mutants (Figure 4C). Therefore, synaptic membranes account for more than 60% of the bouton circumference in leon mutants, as compared to about 35% in controls. This phenotype is consistent with the enlargement of the PSD that is suggested by immunostaining of dPak and GluRs (Figure 3B). Presynaptically, the number of T-bars and vesicle numbers showed no significant difference to controls (Figure 4C). Thus, the ultrastructural analysis confirms the expansion of SSR and PSD in leon mutants.
Impaired electrophysiological properties at leon mutant NMJs
We then examined the electrophysiological properties in leon mutants because of abnormal NMJ morphology in leon mutants. In both 19-2/19-2 and leon1/19-2 mutants, the mEJC frequencies were comparable to the +/19–2 control, showing no significant differences (Figure 5A). mEJC amplitudes in 19-2/19-2 and leon1/19-2 were slightly larger than that in +/19–2 control, although no statistical significance was detected (p=0.07 and p=0.13, respectively). Another independent mEJC amplitudes recording also suggested leon1/19-2 had slightly larger mEJC amplitudes than +/19–2 and w1118 controls (Figure 5—figure supplement 1A). Interestingly, the EJC showed a dramatic reduction in leon mutants compared to the +/19–2 control. The reduction in EJC leads to a reduction in the quantal content, determined by the ratio of the EJC amplitude to the mEJC amplitude (Figure 5A, bottom panels). Failure analysis also suggests that the quantal content was significantly reduced in the leon1/19-2 mutant (Figure 5—figure supplement 1B). Both EJC amplitude and quantal content were restored by introducing GFP-leon-GR into leon mutants (Figure 5A).
We then examined whether Leon is required in pre- or post-synapses for restoring the EJC and quantal content in leon1/19-2. Presynaptic Leon expression by D42-GAL4 failed to restore both EJC amplitude and quantal content in leon mutants (Figure 5B). Postsynaptic Leon expression by 24B-GAL4 partially restored the EJC amplitude and the quantal content in leon1/19-2 (Figure 5C). The mEJC in leon mutants carrying D42-GAL4 or 24B-GAL4 was significantly larger than that in respective GAL4 driver control, suggesting the mEJC increase in leon mutants might be sensitive to the variation in genetic backgrounds (Figure 5B,C). As postsynaptic expression of Leon also suppressed bouton reduction in the leon mutant, the reduction of the bouton and hence the total release sites in leon mutants could contribute partly to the electrophysiological defects. Indeed, there were about 20% reduction in the bouton number (Figure 1A,B,E and F), and about 15% reduction in the release sites at the leon mutant NMJ (Brp number per NMJ: w1118, 731.6 ± 31.6; leon2/19-2, 631.9 ± 28.1; leon1/19-2, 619.5 ± 31.1; n = 10 for all genotypes). With the 50% or more reduction in the EJC amplitude and the quantal content, other processes are also likely defective in the leon mutants.
Suppression of leon mutant phenotypes by reducing postsynaptic proteins
Accumulations of postsynaptic proteins in the SSR or the PSD of leon mutants might cause some aspects of the mutant phenotypes. To test this idea, we examined whether reductions of the gene dosage for these postsynaptic proteins would alleviate leon mutant phenotypes. We first examined SSR-localized Dlg that is required for SSR formation (Lahey et al., 1994). Replacing the wild-type allele by the null dlgX1-2 allele significantly suppressed the SSR-localized Cact expansion in leon1/19-2 (Figure 6A,B). However, the intensity and size of PSD-localized GluRIIA clusters was unaltered. Thus, this result suggests that the Dlg might mediate SSR but not PSD expansion in the leon mutant.
The PSD-localized dPak is required for GluRIIA cluster localization and SSR formation (Albin and Davis, 2004). Replacing the wild-type copy of dpak by the null dpak6 allele in leon1/19-2 suppressed the GluRIIA cluster size and intensity. The Cact-positive zone, however, was not affected (Figure 6A,B). We also examined the SSR-localized complex of Pll, Dl and Cact that regulate GluRIIA abundance at the PSD (Heckscher et al., 2007). Replacing the wild-type allele with respective pll25, dlH, or cact13 in leon1/19-2 suppressed the intensity and size of GluRIIA clusters but had no effect on the thickness of Cact- or Dlg-positive zones (Figure 6A–D). These results suggest that dPak and the Dl/Pll/Cact complex may mediate more specifically the expansion of GlRIIA clusters in the leon mutant.
To further confirm that reduced expressions of these postsynaptic proteins would suppress SSR or PSD enlargement, dlgRNAi, dpakRNAi, dlRNAi, cactRNAi, and pllRNAi that have been used previously (Dent et al., 2015; Sun and Irvine, 2011; Zhou et al., 2015) or shown in this study (Figure 6—figure supplement 1A) to effectively knockdown respective gene expression were introduced into leon1/19-2. To show the effect on the membranous SSR, muscle-expressed mCD8GFP that is enriched in the SSR membrane was used (Figure 6—figure supplement 1B). Expansion of mCD8GFP areas and enhancement of the signals were also detected in leon1/19-2, consistent with the increase of SSR (Figure 6—figure supplement 1C,D). Similarly, dlg knockdown suppressed mCD8GFP-enriched areas in leon1/19-2. However, the expanded PSD-localized GulRIIA clusters remained the same. Instead, pll, dl or cact knockdown suppressed the intensity and size of GluRIIA clusters in leon1/19-2 but had no effect on the expanded mCD8GFP area (Figure 6—figure supplement 1C,D). Interestingly, dpak knockdown in leon1/19-2 suppressed both mCD8GFP areas and GluRIIA clusters (Figure 6—figure supplement 1C,D), consistent with its role in the formation of both SSR and GluRIIA clusters (Albin and Davis, 2004). Thus, these genetic suppressions support that accumulations of postsynaptic proteins at the SSR or PSD could mediate the expansion of postsynaptic specializations in leon mutants.
We then addressed whether the accumulations of postsynaptic proteins at the SSR or PSD are accompanied with increases in the total protein levels. Western blots for examining Dlg, dPak and Cact protein levels were performed in isolated body-wall muscles of wild-type and leon mutants. Quantification of the protein levels showed that Dlg, dPak and Cact were increased in leon1/19-2 (1.19 ± 0.01, 1.3 ± 0.06, and 1.57 ± 0.28 folds, respectively), which were not significantly altered in leon2/19-2 (0.97 ± 0.06, 0.93 ± 0.09, and 1.13 ± 0.13 folds, respectively). Thus, the increases in the protein levels, as well as other mechanisms, could contribute to postsynaptic accumulations of Dlg, dPak and Cact in leon mutants.
Suppression of leon mutant phenotypes by reducing Ubqn levels
In leon mutants, ubiquitinated substrates accumulate (Figure 2A) while the enzymatic activities of the proteasome remain intact (Wang et al., 2014), suggesting that ubiquitinated substrates fail to be transported for proteasomal degradation. We examined the involvement of the ubiquitin-like (UBL) and ubiquitin-association (UBA) domain proteins (UBL-UBA) that function as ubiquitin receptors to bind and shuttle ubiquitinated substrates to the proteasome for degradation. The Drosophila genome encodes three UBA-UBL ubiquitin receptors, Rad23, Ddi1 and Ubqn/Dsk2. We then tested whether any of the UBL-UBA proteins contribute to postsynaptic defects in leon mutants. RNAi transgenes for knocking down Ubqn, Rad23 or Ddi1 were effective in suppressing respective gene expression (Figure 7—figure supplement 1A), and were introduced into leon mutants to test their suppression of leon mutant phenotypes. Interestingly, postsynaptic Ubqn knockdown suppressed the size or intensity of Dlg-positive zones, dPak patches and GluRIIA clusters in both leon2/19-2 and leon1/19-2 (Figure 7A,B and Figure 7—figure supplement 1B,C). In contrast, reductions in Rad23 or Ddi1 expression showed no obvious alteration of leon mutant phenotypes, except for a slight enhancement of Dlg-positive zones in leon2/19-2 by Rad23RNAi (Figure 7—figure supplement 1B,C). These results suggest that Ubqn plays a prominent role in mediating postsynaptic phenotypes in leon mutants. We then investigated whether Ubqn knockdown has any effect on the defective ubiquitin homeostasis in leon mutant postsynapses. The FK2 immunostaining intensity, while increased in leon mutants, was dramatically reduced in muscles and synapses in Ubqn knockdown (Figure 7A,B). Therefore, the UBL-UBA protein Ubqn mediates ubiquitin homeostasis defects and postsynaptic defects in leon mutants.
Given the reduction in Ubqn suppressed leon mutant phenotypes, we examined whether Ubqn expression is altered in leon mutants. As expected, Ubqn protein levels on Western blots showed increases to more than two folds in both leon mutants (Figure 7C, leon2/19-2: 2.2 ± 0.54 folds and leon1/19-2: 2.03 ± 0.35 folds). Immunostaining for Ubqn revealed ubiquitous expression at NMJs and in muscles, with enrichment in nuclei (Figure 7D). In leon mutants, however, postsynaptic Ubqn immunostaining signals were highly elevated at the SSR (>1.5 folds) and only weakly in other regions (1.2 folds in muscle). These results are consistent with that the increase in the Ubqn level in leon mutants could mediate expansions of Dlg-positive zones, dPak patches and GluRIIA clusters in leon mutants.
Ubqn induces and associates with ubiquitinated postsynaptic proteins
We then examined the role of Ubqn in postsynaptic development. The Flag-Ubqn transgene expressed by 24B-GAL4 in postsynapses caused expansions of Dlg-positive zones and GluRIIA clusters (Figure 8A,B), suggesting that elevated Ubqn levels could promote postsynaptic protein accumulation. However, Ubqn depletion by UbqnRNAi knockdown had no effect on these postsynaptic proteins (Figure 8A,B). As a ubiquitin receptor, Ubqn could associate with free ubiquitin chains, ubiquitinated substrates, or both in leon mutants. Immunoprecipitates of Flag-Ubqn probed by ubiquitin antibodies in Western blots displayed only smearing ubiquitin signals in high molecular weights, which were more prominent in leon mutants (Figure 8C, left panels). The ubiquitin signals represent Ubqn-associated, ubiquitinated substrates rather than ubiquitinated Ubqn, as SDS treatment of the Flag-Ubqn immunoprecipitates largely depleted the associated signals (Figure 8C, right panels). Taken together, these results indicate that Ubqn associates with ubiquitinated substrates but not free ubiquitin chains.
We then tested whether elevation of Ubqn could associate and stabilize ubiquitinated postsynaptic proteins. Myc-tagged dPak, Dlg, Cact and GFP were separately co-transfected with HA-Ub into Drosophila S2 cells. Myc-immunoprecipitates probed by HA antibodies in Western blots displayed smearing signals (Figure 8D, lanes 1–4), indicating they are ubiquitinated proteins. Interestingly, upon further co-transfection of Flag-Ubqn, the Myc-precipitates present stronger ubiquitination signals, suggesting that Ubqn enhances ubiquitination of postsynaptic proteins (Figure 8D, lanes 5–8). Ubiquitination of GFP was also enhanced, indicating that Ubqn likely recognizes the conjugated ubiquitin chains for binding, rather than protein substrates. In summary, we propose that Ubqn is elevated in leon postsynaptic sites to bind and stabilize ubiquitinated proteins, a mechanism that could account for the accumulation of postsynaptic proteins in leon mutants.
Induction of leon mutant phenotypes by free ubiquitin chains
In the absence of Leon deubiquitinating activity, free ubiquitin chains also accumulated in the postsynaptic area (Figure 2C). Thus, we tested whether free ubiquitin chains could induce phenotypes observed in leon mutants. To increase the level of free ubiquitin chains in vivo, the conserved C-terminal residues Gly75 and Gly76 that are essential for substrate conjugation were mutated to Ala. UbAA can be conjugated by the endogenous ubiquitin C-terminus onto the Lys48 residue, forming free ubiquitin chains, but is unable to conjugate onto substrates including endogenous ubiquitin due to the C-terminal mutations. Also, UbAA-induced free ubiquitin chains are resistant to Leon enzymatic activity as the C-terminal AA motif would prevent recognition by USP5 (Dayal et al., 2009). As controls, we also generated wild-type UAS-HA-UbGG that can conjugate to substrates, and UAS-HA-UbAA-K48R in which Lys48 was replaced by Arg, preventing ubiquitin chain formation on Lys48 of UbAA (Figure 9A). Larval lysates of UAS-HA-UbGG, UAS-HA-UbAA and UAS-HA-UbAA-K48R driven by 24B-GAL4 were probed with anti-HA antibodies in Western blot analyses (Figure 9B). As expected, UbGG produced smearing signals at higher-molecular weights, representing substrate-conjugated ubiquitin chains (lane 2). As expected, expression of UbAA formed ladders of free ubiquitin chains (lane 3). Finally, expression of UbAA-K48R failed to induce higher-molecular weight, smearing signals and free ubiquitin chains signals, except ubiquitin dimers, likely through conjugation of non-K48 residues (lane 4).
To test which types of ubiquitins have impact on postsynaptic proteins, we performed immunostaining for Dlg, dPak and GluRIIA when these Ub transgenes were expressed in muscles. Quantitative analyses showed that postsynaptic expression of UbAA induced the expansion of Dlg-positive zones and GluRIIA clusters, and enhanced the intensity of dPak and GluRIIA clusters at postsynaptic sites (Figure 9C,D). These phenotypes, however, were not detected in expression of UbGG or UbAA-K48R. Given that UbAA, but not UbGG or UbAA-K48R, induced the formation of free ubiquitin chains (Figure 9B). Thus, these results are consistent with that K48-linked free ubiquitin chains induce higher-levels of postsynaptic proteins at postsynaptic sites.
However, UbAA overexpression failed to fully recapitulate the extreme severity of leon phenotypes, partly due to the difficulty of expressing high levels of free ubiquitin chains in vivo. To further correlate the level of free ubiquitin chains and the phenotypic severity, UbAA was overexpressed in leon2/19-2 and leon1/19-2 in which lower or higher levels of free ubiquitin chains accumulated (Figure 2A). Quantification of Dlg-positive zones and GluRIIA cluster intensity and size indicated that overexpression of UbAA further enhanced the hypomorphic leon2/19-2 phenotypes, but was unable to exacerbate leon1/19-2 defects (Figure 9E). These analyses suggest that the amounts of free ubiquitin chains correlate to the protein levels at postsynaptic sites.
Recapitulation of leon mutant phenotypes by coexpression of free ubiquitin chains and Ubqn
Both levels of free ubiquitin chains and Ubqn were highly elevated in leon mutants. Also, when overexpressed, both were able to induce protein accumulation at postsynaptic sites. We then investigated the relationship between free ubiquitin chains and Ubqn in these processes. Postsynaptic expression of UbAA, but not UbGG or UbAA-K48R, slightly induced the Ubqn levels, including synaptic sites and muscles (Figure 10A,B). Conversely, overexpression of Ubqn could not induce free ubiquitin chains (Figure 10C). As UbAA only induced mild postsynaptic phenotypes accompanying with slight Ubqn elevation, we therefore tested whether co-expression of UbAA and Ubqn could further enhance these phenotypes. This experiment was performed in a leon heterozygous background, +/19–2, in which postsynapses are normal but more sensitive to the induction of phenotypes. Expression of either UbAA or Ubqn in the sensitive background induced mild enhancement of FK2 intensity in postsynaptic sites. However, combined overexpression of UbAA and Ubqn in the same background caused a large increase in FK2 immunostaining intensity (Figure 10D,E). This co-expression also resulted in larger expansion of Cact-positive zones and GluRIIA clusters than expression of either one alone (Figure 10D,E). Thus, these data suggest that Ubqn and free ubiquitin chains in leon mutants could function together to induce defective ubiquitin homeostasis, and protein accumulations at the postsynaptic site.
At NMJs, postsynaptic SSR membranes are highly convoluted, showing many layers of folded membranes that surround presynaptic boutons. Each bouton contains multiple release sites where neurotransmitters are released from synaptic vesicles and received by receptors localized at the PSD (Broadie and Richmond, 2002; Collins and DiAntonio, 2007). Leon negatively regulates these unique postsynaptic specializations through control of postsynaptic protein levels. The Leon function in postsynapses is specific as presynaptic defects were not obvious, and leon mutant defects could be rescued mostly by postsynaptic Leon expression. Leon tightly controls postsynaptic specializations throughout larval development. As such dramatic and concomitant expansion of SSRs and PSDs is rarely detected in other reported mutants, our study suggests that Leon likely executes a coordinated program in confining postsynaptic specializations.
The requirement of Leon deubiquitinating activity indicates that Leon functions as a conserved member of the USP5 family to disassemble free ubiquitin chains, thereby maintaining ubiquitin homeostasis in vivo. In leon mutants, accumulations of free ubiquitin chains and ubiquitinated substrates were detected (Figure 2A). The immunostaining signals by FK2 antibodies were also highly intensified, in particular, at the postsynaptic site (Figure 2B). Thus, free ubiquitin chains and ubiquitinated substrates are two major targets that are suppressed by USP5/Leon activity in postsynapses. Postsynaptic expression of UbAA that forms free ubiquitin chains mildly enhanced FK2 intensity and induced higher-levels of SSR- and PSD-localized proteins, similar to but not as dramatic as what was observed in leon mutants. In addition to free ubiquitin chains, ubiquitinated substrates also contribute to SSR and PSD expansion. In leon mutants, accumulation of ubiquitinated substrates could be attributed to disruption of proteasomal degradation, although the enzymatic activity of the proteasome remains active (Wang et al., 2014). We therefore hypothesized that the delivery of ubiquitinated substrates to proteasomal degradation is hindered, leading to their accumulation. In supporting of this idea, we showed that reducing the level of the UBA-UBL ubiquitin shuttling protein Ubqn could suppress SSR- and PSD-localized protein levels in leon mutants. Overexpression of Ubqn also caused accumulations of ubiquitinated substrates in postsynapses. The Ubqn protein itself also accumulated at the SSR region, and bound ubiquitinated substrates. Thus, in leon mutant postsynapses, higher levels of Ubqn could stall proteasomal degradation and cause accumulation of ubiquitinated substrates in postsynapses (Figure 11). Taken together, we propose that accumulated free ubiquitin chains and Ubqn-bound ubiquitinated substrates are two primary causes of leon mutant defects.
Interestingly, Ubqn level was also upregulated in UbAA overexpression. Conversely, free ubiquitin chains maintained constant levels when Ubqn was overexpressed. These experiments would suggest that accumulation of free ubiquitin chains directly induces higher levels of Ubqn in leon mutants, which could be mediated through stabilization of Ubqn or other pathways. The effect of postsynaptic overexpression of UbAA or Ubqn only induced partial leon mutant phenotypes, such as in the levels of FK2 and postsynaptic proteins, implying tight control on ubiquitin homeostasis and protein substrate levels. Co-overexpression of UbAA and Ubqn together induced much higher FK2 intensity, and higher postsynaptic protein levels than overexpression of either one alone. Thus, Ubqn and UbAA are unlikely to function in a simple linear pathway, as the overexpression of UbAA would have suggested. Other factors such as ubiquitinated substrates have to be taken into account for the full expressivity of leon mutant phenotypes.
With the dramatic expansions of SSR and PSD in leon mutants, we assumed many postsynaptic proteins would be elevated in leon postsynapses. By examining some representative proteins, we showed that those proteins contribute to leon postsynaptic expansions. The SSR-localized Dlg was elevated and its reduction specifically alleviated SSR expansion. PSD-localized dPak was highly elevated near postsynaptic membranes, and dPak accumulation contributes SSR and PSD expansion in leon mutants. Two of SSR-localized NF-κB complex, Cact and Dl, were found to accumulate in leon mutant SSR. Eliminating one wild-type allele or RNAi knockdown of cact, dl, or pll in leon mutants also suppressed the enlargement of GluRIIA clusters, consistent with their roles in regulating GluRIIA cluster abundance. As dramatic increases of protein levels were not detected in Western blots for leon mutants, as would be expected from immunostaining of postsynaptic proteins, other mechanisms like subcellular recruitment to postsynapses could be also involved.
How these postsynaptic proteins accumulated at leon mutant postsynapses? Ubqn could be the key factor in this process. Ubqn associated with ubiquitinated substrates in leon mutants, and could stabilize ubiquitinated dPak, Dlg and Cact (Figure 8D). These results suggest that Ubqn could associate with and stabilize ubiquitinated postsynaptic proteins, which contribute to leon mutant phenotypes. In particular, Ubqn localized at postsynaptic sites and was enriched locally when leon was inactivated, providing the specificity of Leon regulation to postsynaptic specializations (Figure 11). Interestingly, while reduction of Ubqn suppressed SSR- and PSD-localized protein levels in leon mutants, it enhanced leon mutant lethality, indicating the distinct role of Ubqn in postsynapses. Our current model suggests sequential events in leon mutant postsynapses, in which elevated free ubiquitin chains induced Ubqn upregulation, Ubqn further stabilized postsynaptic proteins, and accumulated postsynaptic proteins promote SSR and PSD expansion (Figure 11). The inability to fully recapitulate leon mutant phenotypes could be due to limitation in overexpression or involvement of other factors. Thus, Leon maintains ubiquitin homeostasis through regulation of free ubiquitin chains, Ubqn levels and ubiquitinated substrates for proper development of postsynaptic specializations.
Our model that hinges on postsynaptic Ubqn has resemblances to the pathogenesis of ubiquilin 2-associated ALS (Deng et al., 2011; Ferraiuolo et al., 2011). Ubiquitin and ubiquilin 2 accumulations are common features in ALS patients, reminiscent of Ubqn and free ubiquitin chain accumulations in leon mutants. In addition to ubiquitin homeostatic imbalance and postsynaptic differentiation in leon mutants, overexpression of Ubqn causes pupal lethality and modifies TDP-43 toxicity in an ALS model (Hanson et al., 2010; Lipinszki et al., 2011). Whether ubiquilin 2 accumulation could recruit free ubiquitin chains or ubiquitinated substrates to ubiquitin inclusions and enhance disease progression will be pivotal to understand the pathological mechanisms. Thus, Leon/USP5 holds Ubqn in check and prevents ubiquitin homeostatic imbalance, making USP5 as a potential candidate disease gene in ubiquitin homeostasis-related diseases.
Materials and methods
leon1, leon2, 19–2 mutant alleles, UAS-Flag-leon and UAS-Flag-ED-leon are described in our previous study (Wang et al., 2014). All flies were reared at 25°C.GFP-leon-GR was constructed by fusing GFP to the ATG codon of leon cDNA driven by the genomic sequence between BtbVII ATG and leon ATG. Transgenic flies carrying UAS-HA-UbGG, UAS-HA-UbAA, UAS-HA-UbAA-K48R and UAS-Flag-Ubqn were generated in this study. dlgX1-2 (Zhang et al., 2007), dlH, cact13and pll25 (Heckscher et al., 2007) and UAS-UbqnRNAi#2 (Ganguly et al., 2008) have been described in respective studies. 24B-GAL4 (RRID:BDSC_1767), D42-GAL4 (RRID:BDSC_8816), da-GAL4 (RRID:BDSC_55851), UAS-mCD8GFP (RRID:BDSC_5137) and dpak6 (RRID:BDSC_8809) were obtained from Bloomington Drosophila Stock Center, Bloomington, IN. UAS-leonRNAi (ID 17567), UAS-UbqnRNAi#1 (ID 47447), UAS-Rad23RNAi (ID 30497), UAS-Ddi1RNAi (ID 40512), UAS-pllRNAi (ID 103774), UAS-dlgRNAi (ID 41136) and UAS-dpakRNAi (ID 108937) were from Vienna Drosophila Resource Center, Austria. UAS-cactRNAi (5848 R-3) was from NIG-FLY, Japan. UAS-dl-BRNAi was from Steven A. Wasserman (Zhou et al., 2015).
NMJ 6/7 phenotypes were analyzed at the A3 segment of late third instar larvae as previously described (Tsai et al., 2012a, 2012b). Larvae were dissected in cold calcium free HL3 saline (70 mM NaCl, 5 mM KCl, 20 mM MgCl2, 10 mM NaHCO3, 5 mM trehalose, 115 mM sucrose, and 5 mM HEPES, pH 7.2) and larval body fillets were fixed in 4% paraformaldehyde for 20 min and washed in PBT (0.01% triton-X-100) for 15 min three times. Fixed fillets were incubated with primary antibodies overnight at 4°C, rinsed in PBT three times, and incubated with secondary antibodies for 2 hr at room temperature. Primary antibodies: mouse anti-Dlg (4F3, 1:100; Developmental Studies Hybridoma Bank, DSHB, Iowa City, IA; RRID:AB_528203), mouse anti-Brp (nc82, 1:100; DSHB; RRID:AB_2314868), mouse anti-Synapsin (3C11/SYNORF1, 1:50; DSHB; RRID:AB_528479), mouse anti-CSP (ab49, 1:100; DSHB; RRID:AB_2307345), mouse anti-Futsch (22C10, 1:200; DSHB; RRID:AB_528403), rabbit anti-Cact (1:500; RRID:AB_2314056; [Reach et al., 1996]), rabbit anti-Dl rabbit (1:500; RRID:AB_2570310), anti-dPak (1:1000; RRID:AB_2567913; [Vlachos and Harden, 2011]), rabbit anti-GluRIIB (1:1000; RRID:AB_2568753; [Marrus et al., 2004]), rabbit anti-GluRIIC (1:2500; RRID:AB_2568754; [Marrus et al., 2004]), mouse anti-FK2 (1:500; RRID:AB_10541840; Enzo Life Sciences, Farmingdale, NY.) chicken anti-GFP (1:500; RRID:AB_300798; Abcam, UK.), mouse anti-UBQLN2 (1:100; RRID:AB_565683; Abnova, Taiwan), goat anti-HRP conjugated FITC, rabbit anti-HRP conjugated TRITC or Cy5 (RRID:AB_2314647; RRID:AB_2340257; Jackson ImmunoResearch, West Grove, PA) and mouse anti-Leon (1:100) generated against GST-Leon fusion protein (LTK BioLaboratories, Taiwan). Muscles were visualized by FITC-conjugated phalloidin (1:2000; Sigma-Aldrich, St. Louis, Mo), which are not shown in figures. For GluRIIA (mouse, 1:100, RRID:AB_528269; DSHB) immunocytochemistry, larval fillets were fixed in Bouins fixative (Sigma-Aldrich.) for 5 min, followed by the protocol described above. Larval fillets were mounted onto slides with PBS containing 87.5% glycerol and 0.22M 1, 4-diaza-byciclo (2.2.2) octane (Dabco, Sigma-Aldrich.). NMJ images were acquired by confocal Z-stack scanning (Zeiss LSM510, Germany) using 40x water and 100x oil objectives and processed by LSM 5 image examiner and Adobe Photoshop.
Western blots, Immunoprecipitation and S2 cell transfection
Third instar larva fillets were homogenized in lysis buffer (150 mM NaCl, 5 mM EDTA, 0.5% Triton-X-100, 1% NP-40 and 50 mM Tris-HCl, pH, 7.4) supplemented with protease inhibitor cocktails (Roche, Swiss), 1 mM PMSF, 1 mM DTT and 2 mM Na3VO4 and separated in Nupage 4% ~ 12% Bis-Tris gel (Thermo Fisher Scientific, Waltham, MA) for Western blots. For immunoprecipitation, larvae were lysed in lysis buffer or lysis buffer with 0.5% SDS. Lysates with 0.5% SDS buffer were further diluted to final 0.2% SDS for immunoprecipitation. Lysates after incubation with Flag M2 beads (AB_2637089; Sigma-Aldrich) overnight were washed by lysis buffer three times and separated in Nupage 4% ~ 12% Bis-Tris gel for Western blots. Drosophila S2 cell line was purchased from Invitrogen (RRID:CVCL_Z232; Invitrogen, Thermo Fisher Scientific) and Drosophila S2 cells were cultured at 25°C in Schneider’s Drosophila medium with 10% heat-inactivated fetal bovine serum (Gibco, Thermo Fisher Scientific). S2 cells were transfected with plasmids by Effectene Transfection Reagent (QIAGEN, Germany). Plasmids include Ub-GAL4, Myc-dPak, Myc-Dlg, Myc-Cact, Myc-GFP, Flag-Ubqn, and HA-Ub. After 72 hr transfection, S2 cells were lysed in lysis buffer for immunoprecipitation by Myc agarose beads (9E10; Santa Cruz Biotechnology, Dallas, TX). Immunoprecipitated lysates were washed by lysis buffer three times and separated in 8% acrylamide gels. Antibodies against Dlg (1:1000), dPak (1:5000), Cact (1:5000), ubiquitin (P4D1, 1:1000; RRID; AB_628423; Santa Cruz Biotechnology), mouse anti-UBQLN2 (1:100; Abnova), α-Tubulin (1:10000; Sigma-Aldrich), anti-Myc (9E10, 1:1000; RRID:AB_627268; Santa Cruz Biotechnology) and HA-HRP (1:2000; Sigma-Aldrich) were used in Western blots.
Processing and analysis of ultrastructures of synaptic boutons by electron microscope were as described previously (Tsai et al., 2012b), with some modification. Larval fillets were dissected in cold calcium-free HL3 saline and fixed in modified Trump’s fixative (4% paraformaldehyde/1% glutaraldehyde/0.1 M sodium cacodylate buffer, pH 7.2) and postfixed with 1% aqueous osmium tetroxide/0.1M sodium cacodylate buffer (pH7.2). The dissected muscles 6/7 of A3 segments were stained with 2% uranyl acetate, dehydrated in a graded ethanol series and infiltrated with graded Spurr’s series in a microwave (PELCO BioWave Laboratory Microwave System, Ted Pella, Redding, CA). Thin sections (100 nm) were sectioned by ultramicrotome (Leica, Germany.) and further stained with uranyl acetate and lead citrate. Images were acquired by Tecnai G2 Spirit TWIN (FEI Co, Hillsboro, OR.) and a Gatan CCD Camera (794.10 .BP2 MultiScan). TEM data were quantified by MetaMorph V6.3r7 (Molecular Devices, Sunnyvale, CA).
The thickness of Dlg-, Cact-, or mCD8GFP-positive zone was analyzed in the original confocal Z-stack images and averaged for lengths across the zone defined by eight straight lines radiating from the bouton center. GluRIIA cluster diameters were analyzed in the original confocal Z-stack images, which were scanned with 0.5 μm intervals and covered overall boutons. Isolated GluRIIA cluster with circular shapes in different Z sections were selected to avoid choosing super-imposed ones for scoring their diameters. When elliptic cluster were chosen, the long and short diameters were averaged to reach the average diameters. Brp number and HRP-labeled NMJ areas were analyzed in the original confocal Z-stack images and Brp density was the ratio of Brp number to HRP-labeled NMJ area. Branch length was calculated from HRP labeling. GluRIIA/IIB/IIC and dPak intensities were the summation from all pixels of NMJs, which was normalized to the total HRP intensity at NMJs by Image J. Synaptic FK2 and Ubqn intensities were calculated by mean pixels in the synaptic area normalized to mean HRP pixels in the synaptic area by Image J. Muscular FK2 and Ubqn intensities were averaged from mean pixels of four muscle areas and normalized to mean HRP pixels in the postsynaptic sites by Image J. The intensities were normalized again to the control genotype for presentation in bar graphs.
To estimate expression levels from Western blots, each set of experiments was performed three times independently, and intensities for bands or areas of interest were processed by Image J with normalization to the internal control α-Tub. The areas of interest for mono-ubiquitin, free ubiquitin chains and ubiquitinated substrates are indicated in Figure 2A.
Larvae for knockdowns of UAS-UbqnRNAi (#1 and #2), UAS-Rad23RNAi, UAS-Ddi1RNAi, UAS-SyndRNAi and UAS-pllRNAi driven by da-GAL4 were reared at 25°C. mRNA of indicated genotypes was extracted from third-instar larvae by RNAzol RT (Molecular Research Center, Cincinnati, OH), followed by the reverse transcription of cDNAs by ImProm-II Reverse Transcription System through oligo-dT (Promega, Madison, WI). The mRNA expressions were amplified by PCR using Ubqn, Rad23, Ddi1, pll and Rpl19 specific primers.
For sample preparation, larvae were dissected with the segmental nerve cut near the ventral ganglion in cold modified Ca2+-free HL3.1 saline (70 mM NaCl, 5 mM KCl, 10 mM MgCl2, 10 mM NaHCO3, 5 mM trehalose, 115 mM sucrose, 5 mM HEPES, pH 7.2). Preparations were then incubated in modified HL3.1 saline containing 0.6 mM CaCl2 for stimulation, and recordings were taken at room temperature. The two-electrode voltage-clamp was filled with 3 M KCl and impaled in muscle 6 of the A3 segment. One microelectrode (15 ~ 20 MΩ) monitored the muscle membrane potential while the other (5 ~ 8 MΩ) passed electric currents. The muscle membrane potential was clamped at a command value of −60 mV. mEJCs occurring in the background within 100 s were obtained without any stimulation on the segmental nerve. To record an EJC, the segmental nerve was stimulated every 10 s from the cut end that was connected by a suction electrode with 0.1 msec of pulse duration at the voltage two times that of the threshold. Signals were digitized at 50 kHz by a DigiData 1440 interface (Molecular Devices), low-pass filtered at 10 kHz, and saved on an IBM-compatible PC for analysis. For failure analysis, EJC was evoked in 0.2 mM [Ca2+], and the failure rate was calculated by ln(n/N), with n the number of failure events, and N the total number of stimuli.
We thank JT Wu, Aaron DiAntonio, Steven A Wasserman, Bloomington Drosophila Stock Center, DGRC, Vienna Drosophila RNAi Center, and DSHB for providing reagents, and SP Lee and SP Tsai for technical support. CTC is supported by grants from Ministry of Science and Technology and Academia Sinica of Taiwan.
In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.
[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]
Thank you for choosing to send your work, "USP5/Leon deubiquitinase confines postsynaptic differentiation by maintaining ubiquitin homeostasis through Ubiquilin", for consideration at eLife. Your article has been reviewed by K VijayRaghavan as Senior editor and three reviewers, one of whom, Hugo Bellen, is a member of the Board of Reviewing Editors. Although the work is of interest, we regret to inform you that the findings at this stage are too preliminary for further consideration at eLife. However, we are willing to entertain a resubmission that addresses the reviewers’ comments.
We think that the mechanistic angle including the interpretation of the electrophysiological phenotype will be hard to fix within two months. However, we believe that you should be able to revise the manuscript in a satisfying manner and it should then hopefully be a candidate for eLife. We would be happy to re-review such a thoroughly revised version.
In the current manuscript, Wang et al. assess the function of the deubiquitinase Leon during Drosophila neuromuscular junction (NMJ) development and document that it controls proper postsynaptic differentiation. Leon is the homolog of USP5, a deubiquitinase that functions specifically to disassemble free, substrate-unconjugated polyubiquitin chains to replenish the pool of mono-ubiquitin. Previously, the same group described that loss of Drosophila leon is lethal and leads to the accumulation of both free poly-ubiquitin chains, as well as polyubiquitinated proteins. In the current study, the authors describe a role for leon in regulating ubiquitin homeostasis and restricting the growth and development of the postsynaptic density (PSD) and subsynaptic reticulum (SSR): whereas presynaptic components are unaffected, the levels of numerous postsynaptic components (Dlg, GluRIIA/B, dPak, Cact and Dl) are strongly increased. Surprisingly, the observed phenotypes can be significantly rescued by reducing the levels of Ubiquilin, a protein that transports ubiquitinated substrates to the proteasome. In addition, leon mutants display increased Ubiquilin levels in leon mutants and overexpressing ubiquilin in a wildtype background mimics the phenotypes found in leon mutants, albeit to a lesser extent. These data argue for a role of Ubiquilin in mediating PSD and SSR growth. However, free polyubiquitin chains seem to also play a role as overexpressing these chains also shows a milder version of the phenotypes observed in leon mutants. Finally, combined overexpression of Ubiquilin and free ubiquitin chains in a wildtype background completely mimics leon mutants. Hence, based on the presented data, Leon appears to restrict postsynaptic development by regulating the levels of both Ubiquilin and free polyubiquitin chains.
This is a carefully executed, well-controlled study that consists of an extensive and detailed phenotypic characterization of leon mutants. The figures are nicely organized and in most cases very convincingly bring the message across. It is counterintuitive that reducing the levels of a protein that removes polyubiquitinated proteins in a background in which the levels of polyubiquitinated substrates are elevated. However, the genetic experiments that were performed are clean and clearly show that free polyubiquitin chains, in combination with elevated Ubiquilin levels can affect postsynaptic development. How this occurs remains elusive and is not well discussed by the authors. The experiments to tease out the details of the actual mechanism underlying NMJ overgrowth in leon mutants are not trivial and may represent a continuation / expansion on this paper. I enjoyed reading this paper and would like to recommend it for publication in eLife, if the authors can expand / hypothesize on the consequences of free ubiquitin chains / elevated Ubiquilin levels in the muscle.
– My major comment is a lack of interpretation of the results in the Discussion. At the moment, the Discussion mostly recapitulates the data. How do the authors think Ubiquilin and free ubiquitin chains affect NMJ development?
As the levels of Dlg, Cac, etc are elevated at the earliest stages of NMJ development, it appears that either (a) the activity and/or levels of the master regulator of postsynaptic density formation, which remains unknown, are increased in leon mutants or (b) that the individual proteins that are found at the PSD and SSR are directly regulated by Ubiquilin and/or free ubiquitin chains. This is not discussed at the moment, but is essential for the reader to help interpret the consequences of these findings.
How could free ubiquitin chains alter protein levels / activity? Does overexpressing free ubiquitin chains lead to a similar increase in Ubiquilin levels or are these two events independent?
In addition, is the increase in Ubiquilin due to a lack of free ubiquitin monomers? For instance, if the authors overexpress ubiquitin in leon mutants, can this rescue the increase in Ubiquilin levels? Does this affect NMJ development?
The manuscript reports a very pronounced synaptic phenotype for mutations in Leon, the fly homolog of USP5. Studying the neuromuscular junction, they provide strong evidence that there is an extreme expansion of the folds of the subsynaptic reticulum and a concomittent increase in the postsynaptic presence of proteins associated with that structure. In addition, they show that some proteins found in the postsynaptic density of the NMJ, most notably the receptor subunit GluRIIA, the scaffold protein dlg, and the signaling molecule dPAK are also expanded as is the density by EM. These postsynaptic changes are accompanied by a large change in the gross morpology of the synapse, including a reduction in bouton number. The changes in the nerve structure, however, seem to be an indirect consequence of the role of Leon in the muscle.
The strength of the paper is in the basic description of these phenotypes. The weaknesses of the paper are in the characterization of the electrophysiology and, above all, in the attempt to formulate a mechanistic explanation of the phenotype. There are small technical issues throughout, such as the failure to quantify Western blots and imprecisions in the description (such as equating Cactus or dlg staining with SSR), but where the manuscript really falls apart is from Figure 5 on.
Figure 5: it is not clear to me why they use the heterozygote as the control in panel A. The values for EJC amplitude and quantal content are much lower than the control values in panels B and C. Also, the values for miniEJC amplitude vary greatly from one genetic background to another, hovering around 700 in the D42 and 24B Gal4 lines and more like 900 in the +/19-2 "control". Because this value is the denominator by which they judge quantal content it is possible that the apparent decrease in quantal content is largely due to an increase in mini amplitude that they missed because the +/19-2 value is abnormally high. I would strongly recommend they include w1118 controls, recorded on the same days, alongside all the measurements for genotypes presented so that one can distinguish effects of the genetic background from those of the mutations. They should use the method of failures to determine quantal content independently since there is good reason to suspect they are misled at present. The increase in quantal size cannot alone explain the decrease in EJC amplitude but might explain a homeostatic decrease in probability of release. In any case, in its current state, the electrophysiology is not interpretable and the final line of that section, which says something vague about correlating with the expansion phenotype, is not really an interpretation and it isn't clear what "correlating" refers to.
Nevertheless, the obvious explanation for the decreased EJC amplitude is that there are far fewer boutons present and therefore fewer release sites and less release. I don't understand why they do not make this point. That miniEJC frequency is normal may reflect a homeostatic compensation, but it seems inadequate to compensate for the sharp drop in bouton #.
Subsection “Impaired electrophysiological properties at leon mutant NMJs” paragraph two: do the authors intend mEJC amplitude or EJC amplitude. Did they mean the driver by itself or the driver driving the transgene? This sentence was unclear.
The same paragraph overstates the result. EJC amplitude are not restored to wt levels (and the significance of the change is not tested statistically). What they mean to say is "partially restores" of "improves relative to the mutant" or something like that.
Mechanism: the manuscript does not describe the ubiquitination or a change in ubiquitination for any of the proteins that they claim are misregulated as a result of the leon mutations. They have only overexposed Western blots that are likely to be saturated and no accurate quantification of the levels of the postsynaptic proteins in Figure 6.
Re: Figure 6, one cannot equate Cactus staining with the SSR (see point re Figure 3 also). It might label the entire SSR or it might not, but it would be best just to refer to the width of the cactus immunoreactive zone unless one also is going to do EM to look at the SSR or syndapin staining which might be OK as a surrogate for SSR.
The authors say that knockdown of syndapin has no effect on cactus or GluRIIA expansion but offer no evidence that their RNAi altered syndapin expression. Similarly, they base all the arguments in Figure 6 on the effects of heterozygotes but do not indicate if those heterozygotes changed levels of the encoded protein and over-interpret a failure to alter a parameter as a lack of involvement of that protein in the expansions they observed. Since they do not know quantitatively if leon mutations have altered the levels of that protein, they cannot know if the heterozygous allele they use (a seemingly arbitrary decision) has had any impact on the level of the protein, decreased it back to control levels or not, or rendered it lower than controls. These experiments are not interpretable.
Figure 7 comes the closest to a mechanistic understanding by implicating Ubqn in the pathway. The data are imperfect however since there is no quantification of the claimed increase in Ubqn in the leon mutants and the ability of Ubqn loss of function to prevent the expansion of the postsynaptic markers does not really prove that it was directly in the path of the changes. Presumably any protein that was required for the SSR or the PSD would be necessary regardless of whether it was regulated by Leon. Other specific issues for Figure 7 include: In Figure 7D, FK2 intensity is measured; is this with background subtraction? Given the big change in overall levels this seems to be necessary, especially since the authors are emphasizing a synaptic decrease, but it is not mentioned in the text/methods. From Figure 7E the authors claim that Ubqn levels are upregulated in leon mutants. The change in band size is too subtle to tell, especially with the saturation of both the Ubqn and loading control bands. Either a quantification with several replicates or a shorter exposure blot would be necessary. From Figure 7F the authors claim that Ubqn immunostaining is increased in leon mutants, "highly elevated at SSR rather than other regions"; the magnification is too low to tell, and a quantification is necessary to demonstrate the robustness of the phenotype. If quantification cannot be done, please at least show multiple examples. Figure 7—figure supplement 2C: no loading control shown on the blots.
Ubiquitin-dependent proteolysis is a complex pathway of protein metabolism implicated in such diverse cellular functions. A late step of the process involves disassembly of the polyubiquitin chains on degraded proteins into ubiquitin monomers. USP5 disassembles branched polyubiquitin chains by a sequential exo mechanism, starting at the proximal end of the chain. Particularly, the enzyme is meant to disassemble free, substrate-unconjugated polyubiquitin chains to replenish the pool of mono-ubiquitin, and thus maintain cellular ubiquitin homeostasis.
The Chien lab recently published a first genetic analysis for USP5 in Drosophila, showing that Leon is essential for viability, and that leon mutants accumulate both free and substrate-conjugated polyubiquitin chains Proteasomal enzymatic activities were elevated in leon mutants, while proteasomal degradation of ubiquitinated substrates was impaired.
Here, the Chien lab executed an analysis of Leon function at the larval NMJ system. Core findings/experiments they present are 1) an expansion of the SSR muscle membrane infoldings and the size of postsynaptic glutamate receptor fields/PSD area due to a lack of muscle Leon function; 2) Western blots and immunostainings of dissected body wall muscles pointing towards increased levels of free and substrate-conjugated polyubiquitin chains; 3) based on TEVC recordings, an impairment of evoked release EJC amplitude and quantal content due to lack of postsynaptic muscle Leon; 4) heterozygous suppression of SSR outgrowth/GluRIIA receptor field outgrowth via dlg /dpak, creating arguments for accumulation of postsynaptic proteins to promote SSR expansion; 5) muscle knockdown of UBA-UBL ubiquitin receptor Ubqn suppressed expansions of Dlg rings and GluRIIA.
Per se, I am impressed by the severity of the phenotypes (SSR expansion, GluR fields) they report, and I think if appropriately revised this manuscript might become a strong candidate for eLife. My major concerns are the following:
1) Can they exclude that they are looking at a "heterochronic" phenotype? In the case of dlg mutants, SSR formation is impaired throughout larval development when comparing precisely timed animals. However, as pupa formation is impaired or completely suppressed in dlg mutants, SSR ultimately growths to over a value found in wild type controls. Are their leon mutants undergoing pupa formation timely? They should compare precisely timed cohorts to make sure that really the pace of SSR expansion is increased here.
2) Ubqn knockdown they apparently only tested within leon mutant background. How does the phenotype look like in wt background? Are they maybe just seeing the addition of two phenotypes here without any specific functional interaction? Anyway, I must confess that I cannot fully follow the mechanistic connection in between Leon and Ubqn.
3) They should try to directly monitor a poly-U species of some relevant postsynaptic protein. This would make their case more convincing.
4) Concerning the diameter of the GluRIIA fields they report: values appear very high. Can they exclude that they are imaging two or more super-imposed receptor fields as one?
5) Somewhat surprisingly but not without example, they find mini amplitudes to be unchanged despite a severe increase of particularly GluRIIA. However, I was surprised to see a down-regulation of evoked release, means that quantal content is severely down here. Concerning the mechanistic basis of their electrophysiological phenotypes, they say: "which correlates with its requirement in postsynaptic differentiation" I am afraid this is insufficient.
6) They state that both free polyubiquitin chains and ubiquitinated substrate proteins were changed in leon mutants. Is the availability of monomeric Ubiquitin determining their phenotype? Could they test for this?
7) Why in Figure 2A much more poly-U signal when overexpressing the kinase dead ED Leon?
General comment: the manuscript clearly needs language editing. Moreover, the format of the figures is very diverse, and could be improved by e.g. showing blow ups of the respective GluRIIA phenotypes.
[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]
Thank you for resubmitting your work entitled "USP5/Leon deubiquitinase confines postsynaptic growth by maintaining ubiquitin homeostasis through Ubiquilin" for further consideration at eLife. Your revised article has been favorably evaluated by K VijayRaghavan (Senior editor), a Reviewing editor, and one reviewer.
The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below. Please make the textual changes suggested by his reviewer..
The manuscript is stronger, the electrophysiology more rigorous, and I also think it should be published. The phenotype is big, clear, and the genetics of rescue and RNAi are all very robust. The supplemental table with full statistics is very valuable, as is the inclusion of data from the cd8GFP reporter of the ssr. Some issues can (and should) be improved in the text. My chief concern is that the mechanism behind the phenotypes is not nearly as clear as the authors state. It is fine with me that some things remain poorly explained; but the authors need not try to make it appear as if the very modest changes in certain protein levels are sufficient to explain the phenoptyes or that the changes in bouton number can adequately explain what is happening to quantal content. Ubiquitin homeostasis and protein turn over are very complicated and broadly acting processes and not easily pinned down to the few proteins one looks at. By the end of the Results, a very strong case has been made for the importance of ubiquitination and Ubqn in regulating the development of the postsynapse. The remarks below are to help the authors avoid certain overstatements and premature conclusions in what is otherwise a very interesting and complete manuscript.
1. In the subsection “Increases of postsynaptic proteins in leon mutants” and in the Discussion, it is a little awkward to say that each bouton houses tens of synapses (and it isn't always that many) because the term synapse is used sometimes to refer to the whole junction, sometimes to a bouton, and sometimes to each release site. If they say each bouton contains multiple release sites that face discrete receptor clusters the sentence will be less awkward.
2. At the end of the subsection “Impaired electrophysiological properties at leon mutant NMJs”: the change in bouton number (which this reviewer suggested as an explanation) doesn't really completely account for the change in the ejc amplitude and quantal content. It may account for some of it, but the ephys parameters are decreased by 50% or more (perhaps 65% for quantal content) and bouton # is only decreased by 20-25%. If the density of release sites per bouton and size of boutons is really unchanged, there remains a mismatch that should still be acknowledged. It is tedious, but it would also help if the authors counted brp puncta in an entire NMJ to get a real value for release sites per NMJ. There may be changes in density, for example, that don't reach statistical significance on their own but, coupled with changes in bouton diameter and number might add up to a larger difference in release sites per NMJ. The bottom line is that it is ok if some of the ephys remains mysterious to explain - but this reviewer did not mean to imply that bouton number alone can actually explain the whole phenotype.
3. There is obviously a disconnect between the very large changes in the SSR and dlg and gluR distributions, and the very small (and in the case of the weaker allele, insignificant) changes in the levels of Pak, dlg, and cact. I think it is fair to guess that many other proteins are increased in order to produce the large changes in the structure of the synapse. In addition, the fact that mutations in those proteins and others can prevent to increase in SSR or GluR markers does not in any way indicate that those proteins are part of the pathway through which leon is acting – it only confirms those proteins are important for normal SSR and GluR development, i.e. they could as easily be part of parallel pathways regulating development as part of a linear pathway.
4. At the end of the subsection “Suppression of leon mutant phenotypes by reducing Ubqn levels”: the conclusion regarding the Ubqn increases seen in leon mutants is also overstated at this point in the paper. It is certainly likely that this contributes to the other phenotypes and the finding of a large increase in Ubqn in leon mutants and the finding that reducing Ubqn can prevent some or all of the phenotypes is important. But to really prove that the change in Ubqn is causative and not merely part of the phenotype, it needs the data in Figure 8 that shows that overexpression of Ubqn in the muscle has the same phenotype as leon loss of function. (Ideally, they would also show that simply reducing Ubqn back to wildtype levels with RNAi or a heterozygote would be sufficient to reverse the leon phenotype. Instead they have used RNAi that nearly entirely remove the Ubqn.) The text for Figure 7 should therefore be more circumspect or the big summary of the data for Ubqn in the pathway should wait till after the overexpression of Ubqn is presented in 8A,B.