Trans-Synaptic Teneurin Signalling in Neuromuscular Synapse Organization and Target Choice

LETTER doi:10.1038/nature10923

Trans-synaptic Teneurin signalling in neuromuscular synapse organization and target choice

Timothy J. Mosca1*, Weizhe Hong1*, Vardhan S. Dani1, Vincenzo Favaloro1 & Liqun Luo1

Synapse assembly requires trans-synaptic signals between the pre- The localization of Ten-a and Ten-m suggested their trans-synaptic and postsynapse1, but our understanding of the essential organiza- interaction. To examine this, we co-expressed Myc-tagged Ten-a in tional molecules involved in this process remains incomplete2. nerves using the Q system14 and haemagglutinin (HA)-tagged Teneurin proteins are conserved, epidermal growth factor (EGF)- Ten-m in muscles using GAL4. Muscle Ten-m was able to co- repeat-containing transmembrane proteins with large extracellular immunoprecipitate nerve Ten-a from larval synaptosomes (Fig. 1f), domains3.HereweshowthattwoDrosophila Teneurins, Ten-m and suggesting that the Teneurins form a heterophilic trans-synaptic Ten-a, are required for neuromuscular synapse organization and receptor pair at the NMJ. target selection. Ten-a is presynaptic whereas Ten-m is mostly post- To determine Teneurin function at the NMJ, we examined the ten-a synaptic; neuronal Ten-a and muscle Ten-m form a complex in vivo. null allele and larvae with neuron or muscleRNAi of ten-a and/or ten-m. Pre- or postsynaptic Teneurin perturbations cause severe synapse Following such perturbations, bouton number and size were altered: the loss and impair many facets of organization trans-synaptically and quantity was reduced by 55% (Fig. 2a–c, g and Supplementary Fig. 2) cell autonomously. These include defects in active zone apposition, release sites, membrane and vesicle organization, and synaptic transmission. Moreover, the presynaptic microtubule and postsynaptic a spectrin cytoskeletons are severely disrupted, suggesting a mechanism whereby Teneurins organize the cytoskeleton, which in turn Ten-a Ten-m Ten-a Ten-m HRP affects other aspects of synapse development. Supporting this, b Ten-m physically interacts with a-Spectrin. Genetic analyses of teneurin and neuroligin reveal that they have differential roles that Ten-a Ten-a Fas2 Fas2 HRP synergize to promote synapse assembly. Finally, at elevated c endogenous levels, Ten-m regulates target selection between specific motor neurons and muscles. Our study identifies the Teneurins as a Ten-a Ten-a Brp Brp HRP key bi-directional trans-synaptic signal involved in general synapse organization, and demonstrates that proteins such as these can also d regulate target selection. 4,5 Ten-m Ten-m Dlg Dlg HRP Vertebrate teneurins are enriched in the developing brain , localize to synapses in culture6, and pattern visual connections7. Both e Drosophila Teneurins, Ten-m and Ten-a, function in olfactory synaptic α α partner matching8 and were further identified in neuromuscular junc- Ten-m Ten-m -Spectrin -Spectrin HRP tion (NMJ) defect screens9,10, with Ten-m also affecting motor axon f IP: HA (Ten-m) guidance11. We examine their roles and the underlying mechanisms Ten-a–Myc: N – + –+ Ten-m–HA: M –– + + involved in synapse development. IB: Myc (Ten-a) Both Ten-m and Ten-a were enriched at the larval NMJ (Fig. 1a and Supplementary Fig. 1a). Ten-a was detected at neuronal membranes: IB: HA (Ten-m) this staining was undetectable beyond background in ten-a null IB: Brp (Input) mutants (Supplementary Fig. 1b) and barely detectable after neuronal ten-a RNA interference (RNAi; Supplementary Fig. 1c), indicating that Figure 1 | Teneurins are enriched at and interact across Drosophila Ten-a is predominantly presynaptic. Partial co-localization was neuromuscular synapses. a–e, Representative single confocal sections of synaptic boutons stained with antibodies against Ten-a (red) or Ten-m (green), observed between Ten-a and the periactive zone marker Fasciclin 2 13 horseradish peroxidase (HRP) to mark the neuronal membrane (blue), and a (ref. 12) as well as the active zone marker Bruchpilot (Fig. 1b, c), synaptic marker as indicated. a, Ten-a is associated with presynaptic suggesting that Ten-a is localized to the junction between the periactive membranes and Ten-m largely with the surrounding postsynapse (a). b, c,Ten- zone and the active zone. Ten-m appeared strongly postsynaptic and a shows limited co-localization with the periactive zone marker Fasciclin 2 surrounded each bouton (Fig. 1a and Supplementary Fig. 1a, d). (b), and Bruchpilot (Brp), an active zone marker (c). d, e, Ten-m co-localizes Muscle-specific ten-m RNAi eliminated the postsynaptic staining, with, and extends beyond, Dlg (d) and completely co-localizes with muscle but uncovered weak presynaptic staining (Supplementary Fig. 1e) that a-Spectrin (e). f, Immunoblots (IB) of larval synaptosomes expressing neuronal ubiquitous ten-m RNAi eliminated (Supplementary Fig. 1f). Thus, the Flag–Myc-tagged Ten-a (N) and muscle Flag–HA-tagged Ten-m (M) and immunoprecipitated (IP) using antibodies to HA. Ten-a is detected in the pull- Ten-m signal was specific and, although partly presynaptic, enriched down, indicating that nerve Ten-a and muscle Ten-m interact across the NMJ. postsynaptically. Consistently, muscle Ten-m colocalized extensively This is not seen in control lanes. Owing to low expression, neither transgene with Dlg (Fig. 1d) and completely with a-Spectrin (Fig. 1e) and is thus product is detectable in input lysates, which are enriched in Brp. probably coincident with all postsynaptic membranes. Scale bar, 5 mm.

1Department of Biology, Howard Hughes Medical Institute, Stanford University, Stanford, California 94305, USA. *These authors contributed equally to this work.

a b c d e f g 160 ** 120 NS ** 80 *** *** *** *** *** 40 *** Bouton number 0 k 60 *** *** ten-m ten-a + ten-a 40 *** Control M-IR ten-a Ten-a N nlg1 nlg1 *** NS *** 25 h i j 15 *** *** *** Control ten-m M-IR ten-a Percentage 10 unapposed AZs 5 *** ***

0 nlg1 ten-a ten-a + ten-a + ten-a + Brp DGluRIII HRP Control N Ten-a M Ten-a ten-a nlg1 r ten-m M-IR ten-m M-IR 100 l m n * Normal * * * 75 Misshapen Apposite Control ten-a ten-a 50 contractile tissue 25 Detached o * p q Double * 0 Percentage of T-bars Percentage +a ten-a ten-a ten-a ten-a M-IR N-IR Controlten-m+aten-m u s t 1 mV 150 QC EPSP amp. 10 mV mEPSP amp. 1 s mEPSP freq. 100 * * * 50 ms ** Control 50

Percentage of control Percentage 0 Control ten-a ten-a Control ten-a

Figure 2 | Teneurins affect the structure and function of the neuromuscular ten-a mutants (m–q) showing double T-bars (m), detached T-bars synapse. a–f, Representative NMJs stained with antibodies to HRP. Muscle- (n), misshapen T-bars (o), membrane ruffling (p, waved arrows) and T-bars specific ten-m RNAi (M-IR) and lots of ten-a decrease bouton number. facing contractile tissue (q). Some images show multiple defects. r, Distribution Neuronal (Ten-a N) but not muscle (Ten-a M) restoration of ten-a expression of T-bar defects as a percentage of the total T-bars. N-IR, neuronal RNAi. For rescues this phenotype. These defects resemble nlg1 mutants and are enhanced each genotype, n $ 3 larvae, 40 boutons. s, t, Representative evoked EPSP in ten-a nlg1 double mutants. g, Quantification of bouton number. (s) and mEPSP (t) traces from control and ten-a mutant genotypes. h–j, Representative NMJs stained with antibodies to Brp (green), the glutamate u, Quantification of mean EPSP amplitude (black), mEPSP amplitude (red), receptor subunit DGluRIII (red) and HRP (blue). In control larvae (h), Brp and mEPSP frequency (blue) and quantal content (QC, white) expressed as a DGluRIII puncta properly appose. teneurin perturbations (i, j) disrupt this percentage of the control average. For all genotypes, n $ 7 larvae, 8 muscles. active zone (yellow arrowhead) and glutamate receptor apposition (white Error bars represent s.e.m. Scale bars, 10 mm(a–f), 5 mm(h–j), 100 nm (l– arrowhead). k, Quantification of unapposed active zone/glutamate receptor q). ***P , 0.001, **P , 0.01, *P , 0.05, NS, not significant. Statistical pairs. For each quantification, n $ 8 larvae, 16 NMJs. l–q, Transmission comparisons are with control unless noted. electron microscopy of active zone T-bars (asterisks) in control larvae (l) and and the incidence of large boutons markedly increased (Supplemen- are marked by electron-dense membranes and single presynaptic tary Fig. 2k). Both changes indicate impaired synaptic morphogenesis. specializations called T-bars (Fig. 2l), which enable synapse assembly, The reduction in bouton number was probably cumulative through vesicle release and Ca21-channel clustering16. Teneurin disruption development, as it was visible in first instar ten-a mutants and persisted caused defects (Fig. 2m–r and Supplementary Fig. 3) in T-bar ultrastruc- (Supplementary Fig. 2k). In the ten-a mutant, bouton morphogenesis ture (Fig. 2m–o), membrane organization, and apposition to contractile was rescued by restoring Ten-a expression in neurons, but not muscles tissue (Fig. 2p, q). Teneurin perturbation also impaired postsynaptic (Fig. 2d, g and Supplementary Fig. 2). Neuronal Ten-m overexpres- densities while increasing membrane ruffling (Supplementary Table 1), sion could not substitute for the lack of Ten-a, revealing their non- further indicating organizational deficiency. These phenotypes equivalence (Supplementary Fig. 2e, l). Neuronal knockdown of Ten-a resemble mutants with adhesion and T-bar biogenesis defects17,18, or Ten-m resulted in fewer synaptic boutons (Supplementary Fig. 2f– suggesting a role for Teneurins in synaptic adhesion and stability. h, l), indicating that both have a presynaptic function, although pre- Synaptic vesicle populations similarly required Teneurins for cluster- synaptic Ten-a has a more predominant role (Supplementary Fig. 2l). ing at the bouton perimeter and proper density (Supplementary Fig. 4). Moreover, knocking down postsynaptic Ten-m in the ten-a mutant As these effects are not synonymous with active zone disruption19, did not enhance the phenotype (Fig. 2g). Thus, presynaptic Ten-a Teneurins are also required for synaptic vesicle organization. (and, to a lesser extent, Ten-m) and postsynaptic Ten-m are required Synapses lacking teneurin were also functionally impaired. The mean for synapse development. amplitude of evoked excitatorypostsynaptic potentials (EPSPs) inlarvae teneurin perturbation also caused defects in the apposition between was decreased by 28% in the ten-a mutant (Fig. 2s, u). Spontaneous presynaptic active zones (release sites) and postsynaptic glutamate miniature EPSPsshowed a 20% decrease in amplitude, a 46% decreasein receptor clusters15 (Fig. 2h and Supplementary Fig. 3): up to 15% of frequency (Fig. 2t, u), and an altered amplitude distribution compared the active zones/receptor clusters lacked their partner compared to with control (Supplementary Fig. 5a). These defects resulted in a 20% 1.8% in controls (Fig. 2h–k). Under electron microscopy, active zones reduction in quantal content (Fig. 2u), which could be partly due to

2 | NATURE | VOL 000 | 00 MONTH 2012 ©2012 Macmillan Publishers Limited. All rights reserved LETTER RESEARCH fewer boutons and release sites. However, release probability may alsobe (Supplementary Fig. 9a–d). Thus, Teneurins are involved in the organ- reduced, as suggested by an increased paired pulse ratio in ten-a mutants ization of the pre- and postsynaptic cytoskeletons and postsynaptic (Supplementary Fig. 5d, e). The decay kinetics of responses were faster membranes. Further, endogenous a-Spectrin co-immunoprecipitated in ten-a mutants, suggesting additional postsynaptic effects on glutam- with muscle-expressed, Flag-tagged Ten-m (Fig. 3i), suggesting that ate receptors and/or intrinsic membrane properties (Supplementary Ten-m physically links the synaptic membrane to the cytoskeleton. Fig. 5b, c). Further, FM1-43 dye loading revealed markedly defective Because the most severe defects following teneurin perturbation vesicle cycling in ten-a mutants (Supplementary Fig. 5f, h). Consistent were cytoskeletal, we propose that Teneurins primarily organize the with physiological impairment, teneurin-perturbed larvae exhibited presynaptic microtubule and postsynaptic spectrin-based cytoskeletons profound locomotor defects (Supplementary Fig. 5i). In summary, (Fig. 3j), which then organize additional synaptic aspects20,21.However, Teneurins are required for multiple aspects of NMJ organization and such a solitary role cannot fully explain the observed phenotypes. The function. reduction in bouton number associated with cytoskeletal disruption is As a potential mechanism for synaptic disorganization following milder than that following teneurin disruption20,21,24. Also, although teneurin perturbation, we examined the pre- and postsynaptic active zone dynamics are affected by cytoskeletal perturbation21,defects cytoskeletons. In the presynaptic terminal, organized microtubules in apposition are not21,25. Moreover, the T-bar structural defects more contain Futsch (a microtubule-binding protein)-positive ‘loops’, closely resemble synapse adhesion and active zone formation whereas disorganized microtubules possess punctate, ‘unbundled’ defects17,18. Thus, Teneurins may regulate release site organization Futsch20. Each classification normally represented ,10% (often distal) and synaptic adhesion independent of the cytoskeleton (Fig. 3j). of boutons (Fig. 3a, d and Supplementary Fig. 6). Upon teneurin per- Our data also indicate that Teneurins act bi-directionally across the turbation, many more boutons had unbundled Futsch (Fig. 3b, c and synaptic cleft. Ten-a acts predominantly in neurons, as evidenced by Supplementary Fig. 6) whereas those with looped microtubules were localization, phenotypes caused by neuronal (but not muscle) knock- decreased by 62–95% (Fig. 3d). Therefore, proper microtubule organ- down, and mutant rescue by neuronal (but not muscle) expression ization requires pre- and postsynaptic Teneurins. In contrast to mild (Figs 2 and 3 and Supplementary Figs 2–4, 6, 7 and 9). Yet, in addition active zone/glutamate receptor apposition defects, most boutons dis- to the presynaptic phenotypes, many others were postsynaptic, includ- played microtubule organizational defects. ing reduced muscle spectrin, SSR, and membrane apposition (Fig. 3 teneurin perturbation also severely disrupted the postsynaptic spectrin and Supplementary Figs 7–9). Similarly, although Ten-m is present cytoskeleton, with which Ten-m colocalized (Fig. 1e). Postsynaptic both pre- and postsynaptically, muscle knockdown resulted in pre- a-Spectrin normally surrounds the bouton (Fig. 3e). Perturbing neuronal synaptic defects, including microtubule and vesicle disorganization, or muscle Teneurins markedly reduced postsynaptic a-Spectrin without reduced active zone apposition, and T-bar defects (Figs 2 and 3 and affecting Dlg (Fig. 3f–h and Supplementary Fig. 7). Postsynaptic Supplementary Figs 3, 4, 6 and 7). Thus, Teneurins function in b-Spectrin21, Adducin22 and Wsp were similarly affected (Supplemen- bi-directional trans-synaptic signalling to organize neuromuscular tary Fig. 8). In muscle, a-Spectrin is coincident with and essential for synapses. This may involve downstream pathways or simply establish the integrity of the membranous subsynaptic reticulum (SSR)21,23. an organizational framework by the receptors themselves. Moreover, Consistent with this, teneurin disruption reduced SSR width up to as the results of single disruptions of neuronal ten-a or muscle ten-m 70% (Supplementary Fig. 9d–g) and increased the frequency of ‘ghost’ were similarly severe and not enhanced by combination (Figs 2g and boutons, which are failures of postsynaptic membrane organization23 3d, h and Supplementary Fig. 2k), both Ten-a and Ten-m probably

d 15 a b c *** 80 Unbundled (%) *** *** 60 *** 10 *** *** *** ** 40 *** 5 *** Loops (%) 20 *** *** ***

Futsch HRP h 0 0 Control ten-m M-IR ten-a 15 *** 20

NS Dlg (A.U.) *** 15 e f g 10 *** 10 *** ***

5 *** *** 5

α -Spectrin (A.U.) 0 0 nlg1 ten-a ten-a + ten-a + ten-a + Control Control Ten-a N Ten-a Ten-a M Ten-a α -Spectrin Dlg HRP

Control ten-m M-IR ten-a ten-m M-IR ten-m M-IR i Input IP: Flag j Vesicle Nlg1 Spectrin Ten-m – ++– organization Nrx cytoskeleton (Flag) GluRII 250 kDa Active MT loop IB: α-Spectrin zone Membrane Bouton Organization 300 kDa Ten-a Presynaptic number Ten-m Ten-m IB: Flag motor neuron Postsynaptic muscle Figure 3 | Teneurin perturbation results in marked cytoskeletal and Dlg (red) fluorescence. A.U., arbitrary units. For all genotypes, n $ 6 larvae, disorganization. a–c, Representative NMJs stained with antibodies to Futsch 12 NMJs. i, Immunoblots (IB) showing that a-Spectrin is detected in the Flag (green) and HRP (magenta). Arrowheads indicate looped organization. Arrows immunoprecipitates (IP) of larvae expressing muscle Flag–HA-tagged Ten-m indicate unbundled Futsch. d, Quantification of the percentage of total boutons but not in control larvae. Owing to low expression, Flag–HA-Ten-m is only with looped or unbundled microtubules. e–g, Representative NMJs stained detectable after enrichment by immunoprecipitation. j, Model showing the with antibodies to a-Spectrin (green), Dlg (red) and HRP (blue). Following roles of Teneurins, Neurexin and Neuroligin at the NMJ. Arrow size represents teneurin perturbation, a-Spectrin staining is largely lost. Axonal a-Spectrin is the relative contribution of each pathway to the cellular process as inferred from unaffected by muscle teneurin RNAi (f). h, Quantification of a-Spectrin (green) mutant phenotypic severity. Scale bars, 5 mm. ***P , 0.001, NS, not significant.

00 MONTH 2012 | VOL 000 | NATURE | 3 ©2012 Macmillan Publishers Limited. All rights reserved RESEARCH LETTER function in the same pathway. Our finding that Ten-a and Ten-m ten-m-GAL4>mCD8GFP Muscle 3 Muscle 4 60 co-immunoprecipitate from different cells in vitro8 and across the a b c d Ten-m Syt1 *** NMJ in vivo (Fig. 1f) further suggests a signal via a trans-synaptic 30 complex. Teneurin function, however, may not be solely trans- Nerve Signal (A.U.)

( ten-m M-IR) 0 synaptic. In some cases (vesicle density, SSR width), cell-autonomous MN 4 MN 3 80 knockdown resulted in stronger phenotypes than knocking down in e f g *** synaptic partners (Supplementary Figs 3, 4, 9 and Supplementary 40 GFP Phalloidin Dlg Ten-m Synaptotagmin 1 Ten-m

Table 1). This suggests additional cell-autonomous roles in addition Muscle Signal (A.U.)

( ten-m N-IR) 0 to trans-synaptic Teneurin signalling. Muscle 4 Muscle 3 Signalling involving the transmembrane proteins Neurexin and h m4 m3 i m4 m3 j 20 NS Muscle 3 26 ** Muscle 4 Neuroligin also mediates synapse development .InDrosophila, Neurexin 15 ** ** Muscle 2 (nrx)andNeuroligin1(nlg1) mutations cause phenotypes similar to 10 NS NS teneurin perturbation: reductions in bouton number, active zone organ- Percentage 5 27,28 errors targeting ization, transmission, and SSR width . nlg1 and nrx mutations do not 0 IR IR IR IR IR IR

enhance each other, suggesting that they function in the same path- HRP Dlg Phalloidin

28 27,28 Control Control Control way . Consistently , we found that nrx and nlg1 mutants exhibited + ten-a + ten-a + ten-a + ten-m + ten-m + ten-m + elav-G80 + elav-G80 + elav-G80

Control ten-m-GAL4>ten-m IR + Mhc-G80 + Mhc-G80 + Mhc-G80 largely similar phenotypes (data not shown). To investigate the rela- k l m 100 tionship between the teneurin genes and nrx and nlg1, we focused on m6 m7 the nlg1 null mutant. Both Nlg1 tagged with enhanced green fluor- m7 m6 m7 m6 75 escent protein (Nlg1–eGFP) and endogenous Ten-m occupied a 50 similar postsynaptic space (Supplementary Fig. 10a). teneurin and NS NS NS total boutons Percentage of Percentage nlg1 loss-of-function mutations also displayed similar bouton number 25 *** reductions (Fig. 2e, g), vesicle disorganization (Supplementary Fig. 4), Dlg HRP and ghost bouton frequencies (Supplementary Fig. 9). Other pheno- 0 types showed notable differences in severity. In nlg1 mutants, there was H94-GAL4 + ten-a + ten-m Control >ten-m + ten-m, + ten-m, elav-G80 a 29% failure of active zone/glutamate receptor apposition (Fig. 2k and Mhc-G80 Supplementary Fig. 10d), compared to 15% for the strongest teneurin perturbation. The cytoskeleton of nlg1 mutants, however, was only Figure 4 | High-level Ten-m expression regulates muscle target selection. mildly impaired compared to that seen with teneurin perturbations a, Representative images of hemisegment A3 stained with antibodies to Dlg (Fig. 3d, h and Supplementary Figs 6 and 7). (blue), phalloidin (red), and expressing GFP via ten-m-GAL4 (green). High-level expression is observed in muscles 3 and 8 and basally in all muscles. b, c, Muscle 3 To examine further the interplay of teneurin and nlg1, we analysed (b)and4(c) NMJs show differential Ten-m (red) but similar Synaptotagmin 1 ten-a nlg1 double mutants. Both single mutants were viable, despite (Syt1; green) expression (from a ten-m muscle knockdown animal). their synaptic defects. Double mutants, however, were larval lethal. We d, Quantification of presynaptic Ten-m (red) and Syt1 (green) fluorescence at obtained rare escapers, which showed a 72% reduction in boutons, muscle 3 and 4 NMJs. MN, motor neuron. e, f,NMJsatmuscles3(e)and4 compared to a 50–55% decrease in single mutants (Fig. 2e). Active (f) show differential Ten-m (red) but similar Syt1 (green) expression in muscles zone apposition in double mutants was enhanced synergistically over (from a ten-m nerve knockdown). g, Quantification of postsynaptic Ten-m (red) either single mutant (Fig. 2k and Supplementary Fig. 10e). Cytoskeletal and Syt1 (green) fluorescence at muscle 3 and 4 NMJs. h, i, Representative defects in the double mutant resembled the ten-a mutant (Fig. 3 and images stained with phalloidin (blue) and antibodies to HRP (green) and Dlg (red) to visualize motor neurons and muscles in control (h)orten-m- Supplementary Figs 6 and 7). These data suggest that teneurin genes GAL4.ten-m RNAi larvae (i). m, muscle. j, Quantification of the hemisegment and nrx and nlg1 act in partially overlapping pathways, cooperating to percentage with failed muscle 3 (red), 4 (black) or 2 (blue) innervation. IR, organize synapses properly, with Teneurins contributing more to interfering RNA. k, l, Representative images of the muscle 6/7 NMJ labelled with cytoskeletal organization and Neurexin and Neuroligin to active zone antibodies to Dlg (green) and HRP (magenta). The characteristic wild-type apposition (Fig. 3j). arrangement of boutons (k) is shifted towards muscle 6 when Ten-m is In the accompanying manuscript8, we showed that although the basal overexpressed in that muscle and the innervating motor neurons Teneurins are broadly expressed in the Drosophila antennal lobe, ele- (l). m, Quantification of the total bouton percentage on muscles 6 (blue) and 7 vated expression in select glomeruli mediates olfactory neuron partner (red). All genotypes contain H94-GAL4, additional transgenes are indicated (for matching. At the NMJ, this basal level mediates synapse organization. details, see Methods). The Ten-m-mediated shift is abolished by neuronal or muscle GAL80 transgenes. Scale bars, 100 mm(a), 5 mm(b–i), 10 mm(k, l). In all Analogous to the antennal lobe, we found elevated ten-m expression at cases, n $ 12 larvae. ***P , 0.001, **P , 0.01, NS, not significant. muscles 3 and 8 using the ten-m-GAL4 enhancer trap (Fig. 4a). We confirmed this for endogenous ten-m, and determined that it was con- was specific to muscle 3, as innervation onto the immediately proximal tributed by elevated Ten-m expression in both nerves and muscles or distal muscle was unchanged (Fig. 4j). The low penetrance is probably (Fig.4b–g).Indeed,ten-m-GAL4 was highly expressed in select motor due to redundant target selection mechanisms30. Where innervation neurons, including MN3-Ib, which innervates muscle 3 (ref. 29; did occur, the terminal displayed similarly severe phenotypes to other Supplementary Fig. 11c). This elevated larval expression also varied NMJs (not shown). Thus, in addition to generally mediating synaptic along the anterior–posterior axis (Supplementary Fig. 12), and was organization, Ten-m also contributes to correct target selection at a specific for Ten-m, as Ten-a expression did not differ within or between specific NMJ. segments (data not shown). To determine whether Ten-m overexpression could alter connec- To test whether elevated Ten-m expression in muscle 3 and MN3-Ib tivity, we expressed Ten-m in muscle 6 (but not the adjacent muscle 7), affects neuromuscular connectivity, we expressed ten-m RNAi using and the motor neurons innervating both muscles using H94-GAL4. ten-m-GAL4. Wild-type muscle 3 was almost always innervated Normally, 60% of the boutons at muscles 6/7 are present on muscle 6 (Fig. 4h). However, after ten-m knockdown, muscle 3 innervation with 40% on muscle 7 (Fig. 4k, m). Ten-m overexpression caused a failed in 11% of hemisegments (Fig. 4i, j). This required Ten-m on shift whereby 81% of boutons synapsed onto muscle 6 and only 19% both sides of the synapse, as the targeting phenotype persisted follow- onto muscle 7 (Fig. 4l, m). This shift also required both neuronal and ing neuronal or muscle RNAi suppression using tissue-specific GAL80 muscle Ten-m, as neuronal or muscle GAL80 abrogated it (Fig. 4m). transgenes (Fig. 4j). ten-a RNAi did not show this phenotype (Fig. 4j), The effect was specific because Ten-a overexpression did not alter this consistent with homophilic target selection via Ten-m. The phenotype synaptic balance (Fig. 4m), nor was it secondary to altered bouton

4 | NATURE | VOL 000 | 00 MONTH 2012 ©2012 Macmillan Publishers Limited. All rights reserved LETTER RESEARCH number, which was unchanged (data not shown). Therefore, elevated 13. Wagh, D. A. et al. Bruchpilot, a protein with homology to ELKS/CAST, is required for structural integrity and function of synaptic active zones in Drosophila. Neuron 49, Ten-m on both sides of the NMJ can bias target choice. This, combined 833–844 (2006). with evidence that Ten-m can mediate homophilic interaction in vitro8, 14. Potter, C. J., Tasic, B., Russler, E. V., Liang, L. & Luo, L. The Q system: a repressible supports a trans-synaptic homophilic attraction model at the NMJ as binary system for transgene expression, lineage tracing, and mosaic analysis. Cell 141, 536–548 (2010). in the olfactory system. 15. Marrus, S. B., Portman, S. L., Allen, M. J., Moffat, K. G. & DiAntonio, A. Differential We identified a two-tier mechanism for Teneurin function in synapse localization of glutamate receptor subunits at the Drosophila neuromuscular development at the Drosophila NMJ. At the basal level, Teneurins are junction. J. Neurosci. 24, 1406–1415 (2004). 16. Wichmann, C. & Sigrist, S. J. The active zone T-bar—a plasticity module? expressed at all synapses and engage in hetero- and homophilic J. Neurogenet. 24, 133–145 (2010). bi-directional trans-synaptic signalling to organize synapses properly 17. Aberle, H. et al. wishful thinking encodes a BMP type II receptor that regulates (Fig. 3j). Supporting this, Teneurins can mediate homo- and hetero- synaptic growth in Drosophila. Neuron 33, 545–558 (2002). 8 18. Owald, D. et al. A Syd-1 homologue regulates pre- and postsynaptic maturation in philic interactions in vitro and heterophilic interactions in vivo Drosophila. J. Cell Biol. 188, 565–579 (2010). (Fig. 1f). At the synapse, Teneurins organize the cytoskeleton, interact 19. Kittel, R. J. et al. Bruchpilot promotes active zone assembly, Ca21 channel with a-Spectrin, and enable proper adhesion and release site forma- clustering, and vesicle release. Science 312, 1051–1054 (2006). tion. Further, elevated Ten-m expression regulates target selection in 20. Roos, J., Hummel, T., Ng, N., Klambt, C. & Davis, G. W. Drosophila Futsch regulates synaptic microtubule organization and is necessary for synaptic growth. Neuron specific motor neurons and muscles via homophilic matching and 26, 371–382 (2000). functions with additional molecules30 to mediate precise neuromuscular 21. Pielage, J., Fetter, R. D. & Davis, G. W. A postsynaptic spectrin scaffold defines active connectivity. Teneurin-mediated target selection at the NMJ is ana- zone size, spacing, and efficacy at the Drosophila neuromuscular junction. J. Cell 8 Biol. 175, 491–503 (2006). logous to its role in olfactory synaptic partner matching . As Teneurins 22. Pielage, J., Bulat, V., Zuchero, J. B., Fetter, R. D. & Davis, G. W. Hts/Adducin controls are expressed broadly throughout the antennal lobe, it remains an synaptic elaboration and elimination. Neuron 69, 1114–1131 (2011). attractive possibility that they also regulate synapse organization in 23. Mosca, T. J. & Schwarz, T. L. The nuclear import of Frizzled2-C by Importins-b11 and a2 promotes postsynaptic development. Nature Neurosci. 13, 935–943 the central nervous system. (2010). 24. Featherstone, D. E., Davis, W. S., Dubreuil, R. R. & Broadie, K. Drosophila a-and METHODS SUMMARY b-spectrin mutations disrupt presynaptic neurotransmitter release. J. Neurosci. 21, 4215–4224 (2001). Details of Drosophila stocks, immunostaining, electron microscopy, functional 25. Miech, C., Pauer, H. U., He, X. & Schwarz, T. L. Presynaptic local signaling by a assays, construction of epitope-tagged Teneurin constructs, immunoprecipitation, canonical wingless pathway regulates development of the Drosophila imaging and statistical analysis can be found in Methods. neuromuscular junction. J. Neurosci. 28, 10875–10884 (2008). 26. Craig, A. M. & Kang, Y. Neurexin–neuroligin signaling in synapse development. Full Methods and any associated references are available in the online version of Curr. Opin. Neurobiol. 17, 43–52 (2007). the paper at www.nature.com/nature. 27. Li, J., Ashley, J., Budnik, V. & Bhat, M. A. Crucial role of Drosophila Neurexin in proper active zone apposition to postsynaptic densities, synaptic growth, and synaptic Received 14 June 2011; accepted 7 February 2012. transmission. Neuron 55, 741–755 (2007). 28. Banovic, D. et al. Drosophila Neuroligin 1 promotes growth and postsynaptic Published online 18 March 2012. differentiation at glutamatergic neuromuscular junctions. Neuron 66, 724–738 (2010). 1. Williams, M. E., de Wit, J. & Ghosh, A. Molecular mechanisms of synaptic specificity 29. Landgraf, M., Roy, S., Prokop, A., VijayRaghavan, K. & Bate, M. even-skipped in developing neural circuits. Neuron 68, 9–18 (2010). determines the dorsal growth of motor axons in Drosophila. Neuron 22, 43–52 2. Giagtzoglou, N., Ly, C. V. & Bellen, H. J. Cell adhesion, the backbone of the synapse: (1999). ‘‘vertebrate’’ and ‘‘invertebrate’’ perspectives. Cold Spring Harb. Perspect. Biol. 1, 30. Sanes, J. R. & Yamagata, M. Many paths to synaptic specificity. Annu. Rev. Cell Dev. a003079 (2009). Biol. 25, 161–195 (2009). 3. Young, T. R. & Leamey, C. A. Teneurins: important regulators of neural circuitry. Int. J. Biochem. Cell Biol. 41, 990–993 (2009). Supplementary Information is linked to the online version of the paper at 4. Kenzelmann, D., Chiquet-Ehrismann, R., Leachman, N. T. & Tucker, R. P. Teneurin- www.nature.com/nature. 1 is expressed in interconnected regions of the developing brain and is processed in vivo. BMC Dev. Biol. 8, 30 (2008). Acknowledgements We thank H. Aberle, V. Budnik, A. DiAntonio, R. Dubreuil, 5. Li, H., Bishop, K. M. & O’Leary, D. D. Potential target genes of EMX2 include Odz/ D. Featherstone, N. Reist, T. Schwarz, S. Stowers, D. Van Vactor, R. Wides, the Ten-M and other gene families with implications for cortical patterning. Mol. Cell. Bloomington Stock Center and the Developmental Studies Hybridoma Bank for fly Neurosci. 33, 136–149 (2006). stocks, antibodies and reagents; J. Perrino and D. Luginbuhl for technical assistance; 6. Silva, J. P. et al. Latrophilin 1 and its endogenous ligand Lasso/teneurin-2 form a K. Shen, K. Zinn, D. Banovic, D. Berns, Y. Chou, C. A. Frank, X. Gao, S. Hippenmeyer, high-affinity transsynaptic receptor pair with signaling capabilities. Proc. Natl K. Miyamichi, K. Sillar, B. Tasic, X. Yu and S. Zosimus for critiques. Supported by a Acad. Sci. USA 108, 12113–12118 (2011). National Institutes of Health (NIH) grant (R01 DC-005982 to L.L.), and Epilepsy, 7. Leamey, C. A. et al. Ten_m3 regulates eye-specific patterning in the mammalian Neonatology and Developmental Biology Training Grants (NIH 5T32 NS007280 and visual pathway and is required for binocular vision. PLoS Biol. 5, e241 (2007) HD007249 to T.J.M.). L.L. is an investigator of the Howard Hughes Medical Institute. CrossRef. Author Contributions T.J.M. designed and performed all experiments (apart from 8. Hong, W., Mosca, T. J. & Luo, L. Teneurins instruct synaptic partner matching in an electrophysiology). W.H. characterized and provided new reagents, and assisted in olfactory map. Nature http://dx.doi.org/10.1038/nature10926 (this issue). some experiments. V.S.D. and T.J.M. designed and V.S.D. performed electrophysiology 9. Liebl, F. L. et al. Genome-wide P-element screen for Drosophila synaptogenesis experiments with assistance from T.J.M. V.F. provided new reagents. L.L. supervised the mutants. J. Neurobiol. 66, 332–347 (2006). project. T.J.M. wrote the manuscript with feedback from all authors. 10. Kurusu, M. et al. A screen of cell-surface molecules identifies leucine-rich repeat proteins as key mediators of synaptic target selection. Neuron 59, 972–985 (2008). Author Information Reprints and permissions information is available at 11. Zheng, L. et al. Drosophila Ten-m and Filamin affect motor neuron growth cone www.nature.com/reprints. The authors declare no competing financial interests. guidance. PLoS ONE 6, e22956 (2011). Readers are welcome to comment on the online version of this article at 12. Sone, M. et al. Synaptic development is controlled in the periactive zones of www.nature.com/nature. Correspondence and requests for materials should be Drosophila synapses. Development 127, 4157–4168 (2000). addressed to T.J.M. ([email protected]).

METHODS C-terminal TEV recognition site and 33Flag, 63Myc and 103His tags. The Drosophila stocks. All Drosophila strains and controls were raised at 29 uCto resulting constructs were verified by restriction digest and sequencing and inte- grated into the attP24 or 86Fb landing sites on the second and third chromo- maximize GAL4 expression. All mutants and transgenes were maintained over 57 GFP balancer chromosomes to enable larval selection. Mhc-GAL4 or Mef2- somes . Transgenic flies were verified by immunoprecipitation on western blot GAL4 (ref. 31) was used to drive expression in all somatic muscles. Nrv2-GAL4 and overexpression experiments. (ref. 32) and elav-GAL4 (ref. 33) were used to drive expression in all neurons. H94- Immunoprecipitation, western blots and SDS–PAGE analysis. For Ten-m and GAL4 was used to drive expression in muscles 6, 13 and 4 and their corresponding Ten-a, QUAS-Ten-a-Flag-Myc was expressed in nerves using Synj-QF and UAS- 34 35 Ten-m-Flag-HA in muscles using mhc-GAL4. Larval synaptosomes were prepared motor neurons . daughterless-GAL4 was used to drive expression ubiquitously . 58 Synj-QF36 was used to drive expression in all nerves. NP6658-GAL4 (ten-m-GAL4) from larval body wall fillets as described . For Ten-m and a-Spectrin, control was used to drive expression in the pattern of endogenous Ten-m expression8. The larvae consisted of Mef2-GAL4 without UAS-Ten-m-Flag-HA whereas experi- Df(X)ten-a deletion was used as a ten-a null mutant8. For nlg1 mutants, the I960 mental flies combined the two. Immunoprecipitation was conducted as described 28 using M2-anti-Flag-conjugated agarose (Sigma) or Affi-Prep Protein A beads and ex2.3 alleles were used in trans and double mutant larvae with ten-a and nlg1 23 mutations were obtained using optimized rearing conditions37. Because of the (Bio-Rad) and rat antibodies to HA (Roche) . Proteins were separated on early lethality of the ten-m mutant11, and to assess independently ten-a, tissue- NuPAGE 3–8% Tris-Acetate Gels (Invitrogen) and transferred to nitrocellulose. specific RNAi was used to examine teneurin perturbation using the following Primary antibodies were applied overnight at 4 uC and secondary antibodies at RNAi transgenic strains: for ten-m,UAS-ten-mRNAi-V51173 and for ten-a (ref. 8), 21 uC for 1 h. The following primary antibodies were used: mouse antibody to UAS-ten-aRNAi-V32482. The following transgenic strains were used: UAS-Dcr2 a-Spectrin (mAb3A9, 1:2,000), mouse antibody to Brp (mAbnc82, 1:100), mouse (ref. 38), UAS-Fas2 (ref. 34), UAS-mCD8GFP (ref. 39), UAS-Nlg1-eGFP antibody to Flag (M2, 1:5,000, Sigma-Aldrich), mouse antibody to Myc (3E10, (ref. 28), UAS-Ten-a (ref. 8), P{GS}9267 for ten-m overexpression8. In all cases, 1:1,500, Santa Cruz Biotechnology), rat antibody to HA (3F10, 1:1,500, Roche). the efficacy of RNAi transgenes, overexpression transgenes and the ten-a deletion HRP-conjugated secondary antibodies (Jackson ImmunoResearch) were used at mutant were assessed and verified by the alteration of antibody staining (loss, 1:10,000. Blots were developed using the SuperSignal West Femto Maximum reduction or increase) using tissue-specific GAL4 drivers. For all cases, N-IR Sensitivity Substrate (ThermoScientific). indicates neuronal RNAi, M-IR indicates muscle RNAi, U-IR indicates ubiquitous Imaging analysis. Larvae were imaged with a Zeiss LSM 510 Meta laser-scanning RNAi. For rescuing ten-a mutants, Ten-a N indicates neuronal overexpression confocal microscope (Carl Zeiss) using either a 363 1.4 NA or a 340 1.0 NA with elav-GAL4 of UAS-ten-a and Ten-a M indicates muscle overexpression of objective. NMJ images were taken as confocal z-stacks with the upper and lower UAS-ten-a with Mef2-GAL4. bounds defined by HRP staining unless otherwise noted. For all metrics, boutons Immunostaining. Wandering third instar larvae were processed as previously were assessed in segment A3 at muscle 6/7 and muscle 4 on both the left and right described23. The following primary antibodies were used: mouse antibody to sides. Fluorescence intensity measurements were taken from terminals on muscle Ten-m (mAb20, 1:500)40, guinea pig antibody to Ten-a (1:100)41, mouse antibody 4. All phenotypes, however, were observed at all synapses regardless of muscle to Brp (mAbnc82, 1:250)42, rabbit antibody to Synaptotagmin 1 (1:4,000)43, mouse fibre or segment. For membrane organization, vesicle distribution and Teneurin antibody to Cysteine String Protein (mAb6D6, 1:100)44, mouse antibody to Dlg colocalization, NMJ images were taken as single optical sections at the precise (mAb4F3, 1:500)45, rabbit antibody to Dlg (1:40,000)46, mouse antibody to centre of the bouton as determined by HRP staining. Images were processed with a-Spectrin (mAb3A9, 1:50)47, mouse antibody to Fasciclin 2 (mAb1D4, 1:20)48, the LSM software and Adobe Photoshop CS4. Bouton number, active zone/ 46 glutamate receptor apposition, fluorescent intensity and microtubule organization rabbit antibody to Fasciclin 2 (1:5,000) , mouse antibody to Futsch (mAb22C10, 23 1:50)20, rabbit antibody to DGluRIII (1:2,500)15, rat antibody to Elav were quantified as previously described . Targeting errors for each larva were quan- (mAb7E8A10, 1:25)49, mouse antibody to Even-skipped (mAb3C10, 1:100)50, tified as the percentage of hemisegments from A1 to A7 in a single animal with a b 51 22 failure of target innervation. There was no difference in targeting errors based on rabbit antibody to -Spectrin (1:1,000) , mouse antibody to Hts (1:500) , guinea 34 pig antibody to Wsp (1:1,000)52. Alexa488-, Alexa546- or Alexa647-conjugated body wall segment. Experiments using H94-GAL4 were conducted as described , secondary antibodies were used at 1:250 (Invitrogen). Texas-Red-conjugated and their effects confirmed using Fasciclin 2 overexpression (control 5 58.1% of boutons on muscle 6, 41.9% on muscle 7; Fas 2 overexpression 5 73.0% on muscle Phalloidin was used at 1:300. FITC-, Cy3- or Cy5-conjugated antibodies to HRP 34 were used at 1:100 (Jackson ImmunoResearch). 6, 27.0% on muscle 7; n 5 8animalsforeach,P , 0.0001) . In electron micrographs, parameters were quantified as previously described Electron microscopy. Wandering third instar larvae were processed and sec- 23 tioned as described23. Sections were imaged on a JEM-1400 (JEOL) transmission using ImageJ (NIH) . T-bar defects were classified into one of five categories: electron microscope at 33,000 to 320,000 magnification. normal (no discernible defect), double (two T-bars were observed in the same, Electrophysiology. Larvae were dissected in HL3 saline53 containing 0 mM Ca21 continuous active zone), detached (where the T-bar was clearly visible but was not and 4 mM Mg21. They were then transferred to saline containing 0.6 mM Ca21 explicitly connected to the membrane associated with the nearest PSD), apposite and recordings conducted by impaling larval muscle 6 in body wall segments A3 contractile tissue (where the T-bar was not apposed to the SSR, but rather, the and A4 using sharp intracellular electrodes (10–20 MV), fabricated from contractile tissue of the muscle), misshapen (where an electron-dense T-bar was borosilicate glass capillaries and filled with 3 M KCl solution. For evoked EPSPs, visible but did not conform to the ‘T’ shape. Often, the T-bars were ‘X’ shaped). For severed nerve bundles were stimulated using a suction electrode connected to a Fig. 2r, each defect is expressed as a percentage of the total number of T-bars linear stimulus isolated (A395, World Precision Instruments). Data, acquired observed in a particular genotype. using Multiclamp 700B amplifiers (Molecular Devices), were low-pass filtered at For Ten-m gradient calculation, single optical sections were taken through the 3 kHz and digitized at 10 kHz. Recordings were acquired and analysed using Igor centre of the NMJ on muscle 3 or muscle 4, as determined by HRP immunoreacti- Pro software (Wavemetrics) and custom-written programs. All recordings in vity. The GFP signal (ten-m-GAL4) or antibody signal was then measured using which the resting membrane potential was higher than 260 mV and/or whose ImageJ (NIH). For each larva, measurements were taken on the right and left sides resting potential, input resistance or access resistance changed by more than 20% of each indicated segment. The fluorescence for each segment was expressed as a during the duration of data acquisition were excluded from analysis. All recordings percentage of the fluorescence from segment A1 in the same animal, on the same and data analyses were performed blind to the genotype. Quantal content was side of the larvae. For all larvae, segment A1 represented the maximal fluorescence. corrected for nonlinear summation54. Statistical analysis. Statistical analysis used GraphPad Prism 5 (Graphpad Larval locomotion. Crawling assays were conducted as described55. Software). In all cases involving more than two samples, significance was calcu- FM1-43 dye loading. FM1-43 (Invitrogen) dye loading was conducted as lated using ANOVA followed by a Dunnett post-hoc test to the control sample and described56 with the following modifications: loading was conducted in 1.5 mM a Bonferroni post-hoc test among all samples. For two-sample cases, an unpaired Ca21,90mMK1 saline for 1 min followed by six 2-min washes in 0 mM Ca21 Student’s t-test was used to assess significance, unless otherwise indicated. In all saline. Imaging was conducted on a Zeiss LSM 510 Meta Confocal (Carl Zeiss) cases, both methods provided similar significance measurements. In all figures, with a 340, PlanApo NA 1.0 water immersion lens (Carl Zeiss). significance is with respect to control genotypes unless otherwise noted. Construction of epitope-tagged teneurin transgenes. The ten-m and ten-a 31. Lilly, B. et al. Requirement of MADS domain transcription factor D-MEF2 for coding sequences lacking the stop codon were cloned into pENTR-D/TOPO muscle formation in Drosophila. Science 267, 688–693 (1995). (Invitrogen) from pENTR-ten-m and pENTR-ten-a8. pENTR-ten-m(-stop) and 32. Sun, B., Xu, P. & Salvaterra, P. M. Dynamic visualization of nervous system in live pENTR-ten-a(-stop) were recombined into the destination vectors pUASTattB- Drosophila. Proc. Natl Acad. Sci. USA 96, 10438–10443 (1999). gtw-tFHAH and pQUASTattB-gtw-tFMH, respectively, using LR Clonase II 33. Luo, L., Liao, Y. J., Jan, L. Y. & Jan, Y. N. Distinct morphogenetic functions of similar small GTPases: Drosophila Drac1 is involved in axonal outgrowth and myoblast (Invitrogen). pUASTattB-gtw-tFHAH is a pUASt-Gateway-attB based vector with fusion. Genes Dev. 8, 1787–1802 (1994). a C-terminal TEV recognition site and 33Flag, 33HA and 103His tags. 34. Davis, G. W. & Goodman, C. S. Synapse-specific control of synaptic efficacy at the pQUASTattB-gtw-tFMH is a pQUASt-Gateway-attB8 based vector with a terminals of a single neuron. Nature 392, 82–86 (1998).

35. Wodarz, A., Hinz, U., Engelbert, M. & Knust, E. Expression of crumbs confers apical 48. Van Vactor, D., Sink, H., Fambrough, D., Tsoo, R. & Goodman, C. S. Genes that character on plasma membrane domains of ectodermal epithelia of Drosophila. control neuromuscular specificity in Drosophila. Cell 73, 1137–1153 (1993). Cell 82, 67–76 (1995). 49. O’Neill, E. M., Rebay, I., Tjian, R. & Rubin, G. M. The activities of two Ets-related 36. Petersen, L. K. & Stowers, R. S. A Gateway MultiSite recombination cloning toolkit. transcription factors required for Drosophila eye development are modulated by PLoS ONE 6, e24531 (2011). the Ras/MAPK pathway. Cell 78, 137–147 (1994). 37. Loewen, C. A., Mackler, J. M. & Reist, N. E. Drosophila synaptotagmin I null mutants 50. Patel, N. H., Schafer, B., Goodman, C. S. & Holmgren, R. The role of segment polarity survive to early adulthood. Genesis 31, 30–36 (2001). genes during Drosophila neurogenesis. Genes Dev. 3, 890–904 (1989). 38. Dietzl, G. et al. A genome-wide transgenic RNAi library for conditional gene 51. Byers, T. J., Husain-Chishti, A., Dubreuil, R. R., Branton, D. & Goldstein, L. S. inactivation in Drosophila. Nature 448, 151–156 (2007). Sequence similarity of the amino-terminal domain of Drosophila beta spectrin to 39. Lee, T. & Luo, L. Mosaic analysis with a repressible cell marker for studies of gene alpha actinin and dystrophin. J. Cell Biol. 109, 1633–1641 (1989). function in neuronal morphogenesis. Neuron 22, 451–461 (1999). 52. Ben-Yaacov, S., Le Borgne, R., Abramson, I., Schweisguth, F. & Schejter, E. D. Wasp, 40. Levine, A. et al. odd Oz: a novel Drosophila pair rule gene. Cell 77, 587–598 (1994). the Drosophila Wiskott-Aldrich syndrome gene homologue, is required for cell fate 41. Rakovitsky, N. et al. Drosophila Ten-a is a maternal pair-rule and patterning gene. decisions mediated by Notch signaling. J. Cell Biol. 152, 1–13 (2001). 53. Stewart, B. A., Atwood, H. L., Renger, J. J., Wang, J. & Wu, C. F. Improved stability of Mech. Dev. 124, 911–924 (2007). Drosophila larval neuromuscular preparations in haemolymph-like physiological 42. Laissue, P. P. et al. Three-dimensional reconstruction of the antennal lobe in solutions. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 175, 179–191 Drosophila melanogaster. J. Comp. Neurol. 405, 543–552 (1999). (1994). 43. Mackler, J. M., Drummond, J. A., Loewen, C. A., Robinson, I. M. & Reist, N. E. The C2B 21 54. Martin, A. R. A further study of the statistical composition on the end-plate Ca -binding motif of synaptotagmin is required for synaptic transmission in vivo. potential. J. Physiol. (Lond.) 130, 114–122 (1955). Nature 418, 340–344 (2002). 55. Lnenicka, G. A., Spencer, G. M. & Keshishian, H. Effect of reduced impulse activity 44. Zinsmaier, K. E., Eberle, K. K., Buchner, E., Walter, N. & Benzer, S. Paralysis and early on the development of identified motor terminals in Drosophila larvae. J. Neurobiol. death in cysteine string protein mutants of Drosophila. Science 263, 977–980 54, 337–345 (2003). (1994). 56. Verstreken, P., Ohyama, T. & Bellen, H. J. FM 1-43 labeling of synaptic vesicle pools 45. Parnas, D., Haghighi, A. P., Fetter, R. D., Kim, S. W. & Goodman, C. S. Regulation of at the Drosophila neuromuscular junction. Methods Mol. Biol. 440, 349–369 postsynaptic structure and protein localization by the Rho-type guanine (2008). nucleotide exchange factor dPix. Neuron 32, 415–424 (2001). 57. Markstein, M., Pitsouli, C., Villalta, C., Celniker, S. E. & Perrimon, N. Exploiting 46. Koh, Y. H., Popova, E., Thomas, U., Griffith, L. C. & Budnik, V. Regulation of DLG position effects and the gypsy retrovirus insulator to engineer precisely expressed localization at synapses by CaMKII-dependent phosphorylation. Cell 98, 353–363 transgenes. Nature Genet. 40, 476–483 (2008). (1999). 58. Higashi-Kovtun, M. E., Mosca, T. J., Dickman, D. K., Meinertzhagen, I. A. & Schwarz, 47. Byers, T. J., Dubreuil, R., Branton, D., Kiehart, D. P. & Goldstein, L. S. Drosophila T. L. Importin-b11 regulates synaptic phosphorylated mothers against spectrin. II. Conserved features of the alpha-subunit are revealed by analysis of decapentaplegic, and thereby influences synaptic development and function at cDNA clones and fusion proteins. J. Cell Biol. 105, 2103–2110 (1987). the Drosophila neuromuscular junction. J. Neurosci. 30, 5253–5268 (2010).

Supplementary Figure 1  Ten-a and Ten-m antibody and cell type specificities. a – c, Representative single confocal sections of a Drosophila NMJ stained with antibodies to Ten-a (red), Ten-m (green) and HRP to recognize neuronal membranes (blue). a, Ten-a and Ten-m are both expressed at the larval NMJ. b, In ten-a mutant larvae, the Ten-a signal is no longer detected, demonstrating antibody specificity. The Ten-m signal is still detected, but is manifest in a thinner ring surrounding the bouton due to the reduction of the SSR in the ten-a mutant. c, In larvae expressing interfering RNA (IR, RNAi) against ten- a in the nervous system using nrv2-GAL4 (ten-a N-IR), Ten-a immunoreactivity is nearly eliminated, suggesting that Ten-a is neuronal. Ten-m is unaffected, but is collapsed similarly as in the ten-a mutant. d – f, Representative confocal single optical sections of NMJs stained with antibodies to Ten-m (green) and HRP (blue) to recognize neuronal membranes. d, Ten-m at control synapses extends beyond the neuronal membrane and into the muscle. e, In larvae expressing RNAi against ten-m in the musculature using Mef2- GAL4, the muscle staining is eliminated, leaving a population of Ten-m immunoreactivity contained with the presynaptic neuron. f, In larvae expressing the same RNAi ubiquitously using daughterless-GAL4, the immunoreactivity is completely eliminated. Thus, the antibody is specific to Ten-m and Ten-m itself is endogenously expressed both pre- and postsynaptically. Scale bar, 5 μm.

WWW.NATURE.COM/NATURE | 1 RESEARCH SUPPLEMENTARY INFORMATION

Supplementary Figure 2  Additional genotypes demonstrating teneurin perturbations on synaptic arborization. a – j, Representative confocal images of NMJs at muscle 6/7 in various genotypes. (a) Control NMJ. Loss of ten-a (b) decreases boutons by 55%, which can be rescued by neuronal (c) but not muscle (d) restoration of expression using a UAS-ten-a transgene or neuronal overexpression of Ten-m using the P{GS}9267 line (e). Neuronal knockdown of ten-a (f), ten-m (g) or both (h) all reduce synaptic bouton number. Thus, the presynaptic pool of Ten-m also functions in mediating proper bouton number at the NMJ, though its contribution is less than that of ten-a. Concomitant knockdown of both ten-m and ten-a in neurons is equivalent to the sum of each individual knockdown phenotype. Muscle knockdown of ten-a alone has no effect on synaptic arborization (i) but knockdown of ten-m and ten-a (j) has marked effects and is indistinguishable from expressing RNAi against ten-m alone (Fig. 2). Scale bar, 20 μm. k, Developmental analysis of bouton number in control (black) and ten-a mutant (red) larvae. At each stage observed, there was a significant reduction in bouton number in ten-a mutants, suggesting that the phenotype is cumulative throughout development. l, Quantification of synaptic boutons at the NMJ (black) and of boutons with a diameter greater than 5 μm (red) at muscle 6/7 in segment A3 of wandering third instar larvae. In addition to reductions in the overall number of synaptic boutons, teneurin perturbations also increase the number of improperly sized boutons. Both of these phenotypes are hallmarks of impaired synaptic development. Error bars represent s.e.m. *** P < 0.001. Statistical comparisons are with control values unless otherwise noted.

2 | WWW.NATURE.COM/NATURE SUPPLEMENTARY INFORMATION RESEARCH

Supplementary Figure 3  Additional EM micrographs showing synaptic structural defects following teneurin perturbations. Representative electron micrographs of individual active zones in control larvae (a), ten-a mutant larvae (b, e, h, k, n), or in larvae following neuronal knockdown of ten-m and ten-a (c, f, i, l, o) or muscle knockdown of ten-m (d, g, j, m, p). Examples are of five major active zone defects including double T-bars (b – d), detached T-bars (e – g), misshapen T-bars (h – j), membrane ruffling at active zones (k – m) and active zones improperly opposite to contractile muscle tissue (n – p). Some examples show multiple defects. For example, in (c), there is a double T-bar with membrane ruffling at the active zone. In (p), the active zone is ruffled as well as incorrectly opposite to contractile tissue. Asterisks mark T-bars; waved arrows mark membrane ruffles. Scale bar, 100 nm.

WWW.NATURE.COM/NATURE | 3 RESEARCH SUPPLEMENTARY INFORMATION

Supplementary Figure 4  Synaptic vesicle organization and density defects following teneurin perturbations. a – f, Representative single optical sections through the bouton center. a, In control larvae, synaptic vesicle markers adorn the membrane, forming a “donut”. Following teneurin perturbation using muscle ten-m RNAi (b), neuronal ten-m and ten-a RNAi (c) or the ten-a mutant (d), this organization is abrogated. Instead, immunostaining for the synaptic vesicle markers Cysteine String Protein (CSP) and Synaptotagmin 1 (Syt1) fills ~80% of synaptic boutons without a discernible pattern. The ten-a mutant phenotype can be rescued by neuronal (f) but not muscle (e) Ten-a expression. g, Quantification of the bouton proportion with “organized” (discernible membrane association, gray) and “disorganized” (no obvious membrane association, red) synaptic vesicles. h – k, Representative electron micrographs of boutons in control (h), muscle ten-m RNAi (i), neuronal ten-m and ten-a RNAi (j) and ten-a mutant (k) larvae. Normally, vesicles are tightly packed and clustered near the membrane with the center nearly devoid of vesicles. Following teneurin perturbation, density is decreased and organization is impaired: more vesicles are in the bouton center and are less tightly packed. l, Quantification of vesicle density. m, Quantification of vesicle diameter. Despite impaired organization, individual vesicle diameter is unaffected. n, Quantification of large diameter vesicles (> 150 nm). teneurin perturbation does not affect the distribution of other vesicular structures is unchanged. Scale bars, 5 μm (a-f), 200 nm (h-k). In all graphs, error bars are s.e.m. *** P < 0.001, ** P < 0.01, NS, not significant. Statistical comparisons are with control values unless otherwise noted.

4 | WWW.NATURE.COM/NATURE SUPPLEMENTARY INFORMATION RESEARCH

Supplementary Figure 5  Physiological and behavioral defects following teneurin perturbations. a, Cumulative frequency histogram (from 50 events per recorded muscle) of spontaneous ‘miniature’ EPSP (mEPSP) amplitude in control (black) and ten-a mutant (red) larvae. The distribution is significantly shifted towards lower amplitudes in the ten-a mutant (P < 0.0001, Kolmogorov-Smirnov test). b, Average mEPSP kinetics. While the rise time (black) is not significantly different, individual mEPSPs decay (white) with a faster time constant and have a smaller time integral (area, red) in ten-a mutants. c, Evoked response (EPSP) kinetics. Both the EPSP rise time (black) and latency (blue) remain unchanged, but the decay time (white) is significantly reduced in ten-a mutants. The altered decay kinetics could owe to altered postsynaptic receptor kinetics or membrane properties, indicating additional postsynaptic effects, consistent with a transsynaptic Teneurin function. d – e, Paired pulse ratios (PPR: the 2nd EPSP amplitude divided by the 1st) were obtained by delivering two precisely timed pulses at an interval of 25 ms (40 Hz). Representative traces (d) and mean PPR quantification (e) in control and ten-a mutant larvae. ten-a mutants show reduced short-term depression over controls, suggesting their reduced initial release probability. For all genotypes, n ≥ 8 muscles, 7 animals. f – g, Representative confocal stacks of FM1-43 fluorescence at muscle 12 terminals in control (f) and ten-a mutant (g) larvae. There is a stark reduction in dye loaded in ten-a mutants. Scale bar, 10 μm. h, Quantification of fluorescence at control and ten-a mutant synapses reveals a 70% reduction in intensity, suggesting impaired vesicle cycling in the absence of ten-a. Here, n ≥ 16 NMJs, 4 animals. i, Quantification of larval crawling in control and teneurin-perturbed larvae. There is a 90% reduction in the number of grids crossed in 30 seconds following teneurin disruption. This can be partially rescued by neuronal, but not muscle expression of Ten-a. For all genotypes, n ≥ 30 larvae. Error bars represent s.e.m. * P < 0.05, ** P < 0.01, *** P < 0.001, NS, not significant.

WWW.NATURE.COM/NATURE | 5 RESEARCH SUPPLEMENTARY INFORMATION

Supplementary Figure 6  Presynaptic microtubule organization is altered by Teneurin loss. a,c,e,g,i,k, Representative NMJs stained with antibodies to Futsch (green) and HRP (magenta). Looped, organized microtubules (arrowheads) are seen in control larvae (a) but not following neuronal (c) or muscle (e) teneurin RNAi, nor in the ten-a mutant (g). Instead, most boutons show unbundled (arrows), disorganized microtubules. This phenotype can be partially rescued by restoring Ten-a expression in neurons (i), but not muscles (k). b,d,f,h,j,l, Representative high magnification images of single boutons as above. m, Quantification of the bouton fraction with looped (black) and unbundled microtubules (red). Neuronal teneurin RNAi results in fewer looped and more unbundled microtubules: both presynaptic Ten-a and Ten-m are involved in cytoskeletal organization, but ten-a contributes more strongly. In the muscle, ten- a RNAi has no effect on microtubules, but ten-m knockdown results in severe disorganization. There is very mild impairment in the nlg1 mutant. Double ten-a nlg1 mutants resemble the ten-a mutant. Scale bar, 5 μm. Error bars represent s.e.m. *** P < 0.001. Statistical comparisons are with control values.

6 | WWW.NATURE.COM/NATURE SUPPLEMENTARY INFORMATION RESEARCH

Supplementary Figure 7  teneurin perturbation severely alters the α-Spectrin cytoskeleton. a – f, Representative confocal images of larval NMJs stained with antibodies to α-Spectrin (green), Discs Large (red) and HRP (blue) to mark neuronal membranes. a, In control larvae, α-Spectrin is observed in the presynaptic axon and postsynaptically at terminals. b, In ten-a mutant larvae, the postsynaptic staining is severely reduced, though the presynaptic axonal staining is still evident. This can be rescued by restoring neuronal (c) but not muscle (d) expression of Ten-a. A similar, but much less severe phenotype is evident in nlg1 mutants (e) and nlg1 mutations do not enhance the phenotype of ten-a (f). Scale bar, 5 μm. g, Quantification of α-Spectrin and Dlg fluorescence in various genotypes. The changes in α-Spectrin fluorescence are specific as synapses do not show a significant change in Discs Large fluorescence (P > 0.6 in all cases). Error bars represent s.e.m. *** P < 0.001, NS, not significant. Statistical comparisons are with control values unless otherwise noted.

WWW.NATURE.COM/NATURE | 7 RESEARCH SUPPLEMENTARY INFORMATION

Supplementary Figure 8  Ten-a also regulates the distribution of other postsynaptic proteins. a – b, Representative confocal images of larval NMJs stained with antibodies to β-Spectrin (green), Dlg (red) and HRP (blue) to mark neuronal membranes. a, In control larvae, β-Spectrin is observed in the presynaptic axon and postsynaptically at terminals. b, In ten-a mutant larvae, the postsynaptic staining is severely reduced, though the presynaptic axonal staining is still evident. c – d, Representative confocal stacks as above stained with antibodies to Hts / Adducin (green), Dlg (red) and HRP (blue) to mark neuronal membranes. c, In control larvae, Hts strongly stains the postsynaptic NMJ as well as the presynaptic axon. d, In ten-a mutant larvae, the postsynaptic staining is reduced but not absent. e – f, Representative confocal stacks as above stained with antibodies to Wsp (green), Dlg (red) and HRP (blue) to mark neuronal membranes. e, Wsp also stains the postsynaptic membranes in control larvae. f, This staining is largely absent in ten-a mutant larvae. All of these proteins are associated with postsynaptic cytoskeletons and reside in or around the SSR1. The reduction of β-Spectrin demonstrates that the effects of teneurin perturbation are not limited to one subunit of the spectrin cytoskeleton. Further, the concomitant reduction of Wsp2, an actin-regulatory protein and Hts / adducin, which links the actin and spectrin cytoskeletons3, demonstrate that the effects of teneurin perturbation extend throughout the SSR to spectrin- interacting proteins and perhaps the actin cytoskeleton. Scale bar, 5 μm.

8 | WWW.NATURE.COM/NATURE SUPPLEMENTARY INFORMATION RESEARCH

Supplementary Figure 9  teneurin perturbation results in an increased occurrence of ghost boutons and a reduction in the subsynaptic reticulum. a-c, Representative confocal images of NMJs stained with antibodies to Dlg (green) and HRP (magenta). a, In control larvae, all presynaptic boutons also have postsynaptic Dlg staining. When teneurin function is impaired in ten-a mutant larvae (b) and in muscle ten-m RNAi (c), ghost boutons that are HRP-positive but Dlg-negative are evident (arrowheads). d, Quantification of ghost boutons. In ten-a mutant larvae, the ghost phenotype can be rescued by presynaptic, but not postsynaptic, Ten-a expression. e-g, Representative electron micrographs of the subsynaptic reticulum (SSR), a folded postsynaptic membrane structure present at all boutons. SSR width is markedly reduced following teneurin disruption. h, Quantification of the mean SSR width (black) and the ratio of the SSR area to the bouton area (red). Both parameters are similarly decreased by teneurin perturbation. Error bars represent s.e.m. Scale bars, 5 µm (a-c), 500 nm (e- g). *** P<0.001, NS, not significant. Statistical comparisons are with control values unless otherwise noted.

WWW.NATURE.COM/NATURE | 9 RESEARCH SUPPLEMENTARY INFORMATION

Supplementary Figure 10  Neuroligin 1 and Ten-a co-regulate active zone apposition. a, Representative high magnification single optical sections images of NMJs in animals expressing UAS- Nlg1-eGFP (red) under the control of the muscle-specific Mef2-GAL4 driver and stained with antibodies to Ten-m (green) and HRP (blue) to mark neuronal membranes. Both Nlg1-eGFP and Ten-m occupy similar synaptic space. b – e, Representative high magnification confocal images of NMJs stained for the active zone protein Brp (green), the glutamate receptor subunit DGluRIII (red) and HRP to recognize neuronal membranes (blue). b, Control larva showing normal apposition between active zone and glutamate receptor puncta (see Fig. 2h). One unapposed active zone (yellow arrowhead) is visible. c, When ten-a is removed, the proportion of unapposed glutamate receptor puncta (white arrowheads) and unapposed active zones (yellow arrowheads) increases to 11%. d, In the nlg1 mutant, nearly 30% of active zones / glutamate receptors are unapposed by their obligate partner (marked by arrowheads as above). Additionally, changes in the size and distribution of puncta are evident. e, In the ten-a nlg1 double mutant, a larger proportion of active zones (50%) are unopposed by glutamate receptors (quantified in Fig. 2k), revealing synergy between the Teneurin and Nrx/Nlg1 pathways. In (e), arrowheads are omitted for clarity. Scale bar, 5 μm.

10 | WWW.NATURE.COM/NATURE SUPPLEMENTARY INFORMATION RESEARCH

Supplementary Figure 11  ten-m-GAL4 expression in the muscle and the ventral nerve cord. a, Representative image of larvae expressing the reporter mCD8GFP under the control of the ten-m-GAL4 driver and stained with antibodies to Dlg (blue) and with Texas Red-conjugated phalloidin (red) to mark the muscle fibers. ten-m-GAL4 shows expression at all neuromuscular terminals, consistent with the antibody expression. b, Representative high magnification single optical sections of an NMJ as above and stained with antibodies to Syt1 (blue) and HRP (red) to mark neuronal membranes. The GFP signal is coincident with the membrane and within the incoming axon but also extends beyond the membrane (green fluorescence beyond the red) as a postsynaptic signal. This corresponds to both pre- and postsynaptic expression, similar to endogenous Ten-m (see Fig. 1, Supplementary Fig. 1). c, Representative confocal images of a larval ventral nerve cord as above and stained with antibodies to Even-skipped (red) and Elav (blue). The ten-m-GAL4 positive cells are neurons and a subset is positive for Even-skipped, which is expressed in dorsal motor neurons and some interneurons. Arrowheads indicate likely position of DA3 / MN3-Ib, the type Ib motor neuron that innervates muscle 3. Scale bars, 20 μm (a), 10 μm (b), 5 μm (c).

WWW.NATURE.COM/NATURE | 11 RESEARCH SUPPLEMENTARY INFORMATION

Supplementary Figure 12  Ten-m is expressed in a segmental gradient at the larval NMJ. a, Representative whole-mount images of Drosophila larval fillets stained with phalloidin (red) and expressing GFP under the ten-m-GAL4 enhancer trap (green). Specific expression is observed in two muscle fibers per hemisegment and in a graded fashion. b – h, Single confocal slices of endogenous Ten-m staining at individual muscle 3 synapses in the indicated body wall segment. Endogenous staining demonstrates a similar gradient to that observed with the ten-m-GAL4 enhancer trap. i – o, Single optical sections of Ten-m staining at individual muscle 4 synapses in the indicated body wall segment. No gradient is observed. The synapses are indicated by the dashed lines. p, Quantification of endogenous Ten-m at muscle 3 (red) and muscle 4 (black) and ten-m-GAL4 driving GFP fluorescence at muscle 3 (green) and muscle 4 (blue) expressed as a function of the fluorescence in segment A1 of the observed animal. The gradient at muscle 3 is similarly present endogenously as in the enhancer trap, validating the enhancer trap expression. In contrast, no gradient is observed at muscle 4. Segmentally graded expression of known genes is uncommon and, in the case of Ten-m could potentially contribute to the mechanism whereby the same muscle fibers of progressively descending segments have differing bouton numbers4. Error bars are s.e.m. Scale bar, 500 μm (a), 5 μm (b-o).

12 | WWW.NATURE.COM/NATURE SUPPLEMENTARY INFORMATION RESEARCH

Supplementary Table 1 Quantification of bouton parameters by electron microscopy.

Control ten-m+a N-IR ten-m M-IR ten-a n (3 larvae, 44 boutons) n (5 larvae, 45 boutons) n (4 larvae, 40 boutons) n (4 larvae, 86 boutons)

Perimeter 8.1 ± 0.68 7.1 ± 0.56 7.0 ± 0.61 7.1 ± 0.44 (μm) n/a P > 0.4 P > 0.4 P > 0.4

Area 4.3 ± 0.60 3.4 ± 0.57 3.3 ± 0.65 3.1 ± 0.32 (μm2) n/a P > 0.3 P > 0.3 P > 0.3

PSD Length 560.3 ± 14.40 424.6 ± 20.69 599.6 ± 22.67 437.2 ± 13.41 (nm) n/a P < 0.0001 P > 0.4 P < 0.0001

T-Bar Profiles 1.5 ± 0.23 1.0 ± 0.13 1.1 ± 0.13 1.0 ± 0.12 (per section) n/a P < 0.05 P < 0.05 P < 0.05

3.8 ± 0.20 2.2 ± 0.18 2.9 ± 0.17 2.6 ± 0.18 PSD Number n/a P < 0.0001 P < 0.001 P < 0.0001

SSR Crossings 12 ± 4.5 11 ± 4.4 12 ± 1.0 14 ± 9.1 (cross/μm) n/a P > 0.1 P > 0.1 P > 0.1

SSR Width 1002 ± 44.41 701.1 ± 46.27 411.5 ± 33.56 327.6 ± 29.93 (nm) n/a P < 0.0001 P < 0.0001 P < 0.0001

SSR / Bouton 4.1 ± 0.43 2.7 ± 0.38 1.7 ± 0.18 0.97 ± 0.093 Area n/a P < 0.0001 P < 0.0001 P < 0.0001

Vesicle Density 154 ± 9.90 92.9 ± 9.64 110 ± 9.38 68.7 ± 3.94 (ves / μm2) n/a P < 0.0001 P < 0.001 P < 0.0001

7.9 ± 0.74 6.6 ± 0.66 7.0 ± 0.77 6.8 ± 0.52 % Mito. Area n/a P > 0.5 P > 0.5 P > 0.5

Large Vesicles 3.5 ± 0.37 3.8 ± 0.48 4.4 ± 0.44 3.4 ± 0.26 (diam > 150 nm) n/a P > 0.2 P > 0.2 P > 0.2

27.4 ± 1.39 13.8 ± 0.933 26.5 ± 1.77 16.9 ± 0.977 % AZ Length n/a P < 0.0001 P > 0.5 P < 0.0001

Membrane Ruffles 0.9 ± 0.16 5.1 ± 0.51 3.2 ± 0.46 6.7 ± 0.48 (per AZ) n/a P < 0.0001 P < 0.05 P < 0.0001

WWW.NATURE.COM/NATURE | 13 RESEARCH SUPPLEMENTARY INFORMATION

Supplemental References

1 Kumar, V. et al. Syndapin promotes formation of a postsynaptic membrane system in Drosophila. Mol Biol Cell 20, 2254-2264, (2009).

2 Coyle, I. P. et al. Nervous wreck, an SH3 adaptor protein that interacts with Wsp, regulates synaptic growth in Drosophila. Neuron 41, 521-534, (2004).

3 Pielage, J., Bulat, V., Zuchero, J. B., Fetter, R. D. & Davis, G. W. Hts/Adducin controls synaptic elaboration and elimination. Neuron 69, 1114-1131, (2011).

4 Johansen, J., Halpern, M. E., Johansen, K. M. & Keshishian, H. Stereotypic morphology of glutamatergic synapses on identified muscle cells of Drosophila larvae. J Neurosci 9, 710-725., (1989).

14 | WWW.NATURE.COM/NATURE