Table S1. Fly stocks and antibodies used in this study.

Fig. S1. Molecular characterisation of wah. (A) Genomic structure of CG4699/wah showing introns as dashed lines (not to scale) and numbered exons (http://flybase.org/). Non-coding exon regions are in black and coding exons in alternating light and dark grey. The three predicted splice versions (wah-RA,-RB,-RC) differ only in the length of exon 1 and encode therefore the same predicted protein of 1570 amino acid residues. Two lethal P-element insertions in the 5′ untranslated region of CG4699, P{lacW}l(3)S009413 (wah^P1; sequence accession FM986317) and P{PZ}l(3)06536 (wah^P2; sequence accession AQ034070), fail to complement each others lethality (not shown). (B) Schematic representations of wah-RB transcript (1), Wah protein (2, 5), human KIAA1267 protein (3) and human AAI26156 protein (4). For the Wah protein (2, 5) positions of the PEHE domain (Marin, 2003), the histone fold domain (according to Ensembl; www.ensembl.org) and predicted nuclear localisation signals (*; according to PSORT; http://psort.nibb.ac.jp/form2.html) are indicated; connecting dashed lines show regions of sequence homology (alternative predictions for domain delineations indicated by slash) for the following pairings: Wah versus human KIAA1267 protein (dotted and dashed boxes), human KIAA1267 versus human AAI26156 (white boxes), human AAI26156 versus Wah (black bars); percentages of homology (identity/positives) are listed in boxes between protein pairs. The wah^IR fragment and probes for in situ hybridisations are indicated. (C-F) In situ hybridisation (based on alkaline phosphatase staining following standard protocols) (Plickert et al., 1997; Tautz and Pfeifle, 1989) to wild type (wt) and wah^P1 mutant embryos at stages 12 and 16 (e12, e16) using the 3′ probe as indicated in B; the central nervous system (asterisks) is strongly labelled in wild type at both stages, whereas staining outside the nervous system is visible at stage 12 but no longer at stage 16 (white arrows indicate the ventral musculature) − in agreement with the fact that no ubiquitin phenotypes are observed in late embryonic muscles (not shown); all staining is abolished in wah^P1 embryos stained in the same batch as wild-type specimens, indicating staining to be specific. Scale bar (in A) represents 30 mm in all images.

Fig. S2. Effects of Wah loss-of-function on NMJ morphology. A-C) Images of NMJs (arrowheads) on ventral longitudinal muscles 3 and 4 (VL3/4; muscle nomenclature according to Bate, 1993) in late stage 17 embryos (e17) with different genotypes: wildtype, wah^P1 mutant, or with targeted expression of wah^IR (driven in embryonic motorneurons by Hb9-Gal4). (D) Quantification of the relative lengths of embryonic NMJs on VL3/4 (NMJ lengths divided by respective muscle lengths; mean ± SEM; n, number of terminals analysed; asterisks, p≤0.002 assessed by t-test); the wah^P1 mutant phenotype is reproduced by motorneuronal expression of wah^IR in wildtype embryos and rescued by motorneuronal expression of HA-wah in wah^P1 mutant embryos. E-P) Images of late larval (L3) NMJs (arrowheads) on different muscles (VL3/4, SBM, VO6); motorneurons at these NMJs are either wildtype (left), are wah^P1 mutant (induced by MARCM mosaic strategy; middle), or display targeted expression of wah^IR (synchronously driven by the two larval motorneuronal drivers OK6-Gal4 and D42-Gal4; right); the wah^P1 mutation and knock-down of Wah cause comparable phenotypes: longer NMJs with normal bouton size on VL3/4 (E-G), smaller boutons but normal NMJ size on SBM (H-J), and no detectable phenotypes on muscle VO6 (K-P; one example of a small and one of a large NMJ are given, respectively). MARCM analysis was carried out as described previously (Lee and Luo, 1999; Vogler and Urban, 2008): two hour egg lays obtained from hs-FLP¹²²; FRT82B, wah^P1/TM6b females crossed to elav-Gal4, hs-FLP¹/Y; UAS-mCD8-GFP; FRT82B, tub-Gal80 males were kept for 165 minutes at 25°C, heat shocked for 90 min at 37°C, and raised at 25°C into late third instar larvae. Female larvae (hs-FLP¹²²/ elav-Gal4, hsFLP¹; UAS-mCD8-GFP/+; FRT82B, tub-Gal80/FRT82B, wah^P1) showing green fluorescent patches in the CNS were selected for dissection and stained. Stainings: act, phalloidin-stained F-actin; CD8, mCD8-GFP; Syn, Synapsin. Scale bar (in A) represents 15 mm in A-C, E-G, K-P, 50 mm in H-J.

Fig. S3. Selective co-accumulation of Hrs with differently induced ubiquitin puncta. Confocal images show horizontal views of late larval muscles expressing different constructs (indicated bottom left), stained against ubiquitin (Ubi, green) and Hrs (magenta). Hrs accumulated upon wah^IR but not tau::myc or dts5 expression. Scale bar (in A) represents 50mm in all images; insets are 200% enlarged.

Fig. S4. Wah loss-of-function affects SMAD complex localisation in muscle nuclei. Late larval muscles of different genotypes (wt, wildtype; GFP, UAS-eIF4AIII-GFP; wah^IR, UAS-wah^IR; mad-GFP, UAS-mad-GFP; all constructs expressed with the muscle-specific driver BG57-Gal4), stained with antibodies against Medea (Med), phosphorylated Mothers against Dpp (pMad), ubiquitin (ubi) or visualised for GFP fluorescence. (A-D) Medea and p-MAD reliably accumulate in nuclei of wild-type muscles when dissected in PBS containing 500 nm Ca²⁺ (A,C); upon knock-down of Wah, nuclear levels of Medea or pMAD are suppressed (B,D); instead extranuclear patches of Medea can occasionally be seen that do not colocalise with ubiquitin puncta (close-up in B). (E,F) Nuclear Medea levels are rescued if wah^IR is co-expressed with HA-wah, but not with a GFP control construct. (G,H) Upon knock-down of Wah, Mad::GFP forms extracellular patches which do not overlap with ubiquitin puncta, confirming findings with anti-Medea; Mad::GFP enters nuclei even if Wah is knocked down, suggesting that the effect of wah^IR on SMAD complex localisation is not absolute. White arrowheads point at stained, open arrowheads at non-stained muscle nuclei. A,B,G,H represent confocal images, C-F were obtained on a conventional fluorescent microscope. Scale bar (in A) represents 50mm in A-F and 18 mm in G, H; insets are 200% enlarged.

Fig. S5. Sequences of cDNAs associated with different ESTs of wah. Schematic representation of wah-RB transcript, and sequences of cDNAs associated with different wah ESTs obtained for the cloning of the full-length wah gene. cDNAs were obtained from DGRC and sequenced multiple times. All sequences and annotations have been deposited with the following accession numbers: FM867599 (SD06860), FM867600 (AT07776), FM867601 (AT22722), FM867602 (LD46639) and FM867603 (LD39557). Different inconsistencies were found in most cDNAs, particularly in or around intron 5, and none represents a full-length cDNA. SD06860 contains 50 bp from the first intron, part of the 59 UTR and all introns until it terminates at the XhoI site. LP09056 starts at 1639 bp and covers the rest of the coding sequence and 3′ UTR; it has a complete intron 5-6 sequence and a premature stop codon at 3859 bp. AT07776/AT22722 contains 50 bp of the 5′ UTR and covers 5 exons before terminating 60 bp into intron 5-6, followed by a poly A tail. LD46639 contains the complete 5′ UTR of isoform RA and terminates at the XhoI site; sequence contains 10 bp from intron 5-6. LD39557 represents isoform RB including the complete 5′ UTR, and terminates at the XhoI site; it lacks about 63 bp of exon 5.

Movie 1. Tomogram of sausage body fragment. Tomogram calculated from a tilt series of a 120 nm thick section through a sausage body; see Fig. 2 for still images and detailed explanations.

Movie 2. 3D reconstruction of a sausage body fragment. 3D reconstruction of the tomogram shown in movie 1; tubules forming continuous entities within the analysed fragment are colour coded distinctly; see Fig. 2 for still images and detailed explanations.