Supplementary Notes Downregulation of atlastin does not affect secretory traffic We investigated whether Atlastin might play a role in secretory traffic. Traffic impairment results in disruption of Golgi morphology with dispersal of Golgi markers 1, 2. Accordingly, RNAi-mediated inactivation of α-cop and Sar-1, components of the COP I and COP II coat complexes involved in secretory system transport, resulted in a diffuse GalT-GFP signal partially redistributed to the ER (Fig. S3). In contrast, EM analysis demonstrated that Golgi morphology was not perturbed in neurons with downregulated Atlastin levels (Fig. 2A), confirming our immunofluorescence observations (Fig. S7). Because maintenance of Golgi structure depends on the integrity of ER export 2, we examined the morphology of exit sites in tubulin-gal4/+;uas-atlastin-rnai/+ larva muscles using a Sar1-GFP reporter 3. Despite ER fragmentation, the integrity of ER exit sites was preserved although they became enriched near the nucleus mirroring the increased perinuclear distribution of ER punctae in atlastin RNAi muscle (Fig. S4). To further investigate the potential involvement of Atlastin in traffic through the Golgi complex, we analyzed the distribution of the transmembrane protein mcd8-gfp 4, 5 in atlastin RNAi larva neurons and found that it did not differ from that of controls (Fig. S5a). Moreover, transport of mcd8-gfp to the pre-synaptic membrane of the third instar larva neuromuscular junction was evaluated by quantitative analysis of GFP fluorescence intensity. Normalized GFP fluorescence intensity at wild type and atlastin RNAi synaptic terminals was statistically identical (Fig. S5c, d), demonstrating that mcd8-gfp delivery to the membrane is not impaired in atlastin RNAi animals. The findings that loss of Atlastin induced ER network fragmentation with loss of continuity but had no effects on Golgi architecture, did not perturb ER exit sites and did not alter mcd8-gfp membrane transport indicate that, in agreement with results for human atlastins 6, Atlastin is unlikely to participate in secretory traffic. Rather, these data strongly suggest that Atlastin function is required to maintain ER integrity. Although elimination of Atlastin did not produce obvious transport impairment, subtle transport defects due to architectural www.nature.com/nature 1
disorganization of the ER following fragmentation cannot be ruled out completely. Reduced membrane traffic in Drosophila results in cell growth defects 7, indicating that subtle transport impairment secondary to ER disorganization may explain the small size of atlastin mutant cells and individuals. Validation of in vivo RNA interference Several experiments were carried out to validate the efficiency and specificity of RNAi. First, ubiquitous expression of UAS-atlastin-RNAi using the tubulin-gal4 driver was semi-lethal and escapers displayed small body size (Figure S6) and shortened life span, accurately phenocopying atlastin null mutants 8. Second, western blot analysis of lysates prepared from tubulin-gal4/+;uas-atlastin-rnai/+ escapers showed a drastic reduction in Atlastin protein levels (Figure S6). Third, the small eye produced by overexpression of UAS-atlastin using the eye specific driver GMR-Gal4 was completely suppressed by coexpression of UAS-atlastin-RNAi under the control of the same driver (data not shown). Collectively these data demonstrate that in vivo downregulation of Atlastin is specific and effective. Maternal contribution is likely account for the apparently mild loss of Atlastin function phenotypes If Atlastin is crucial to promote membrane fusion and thus plays a fundamental role in ER biogenesis, it would be reasonable to expect that loss of protein function would result in embryonic lethality. However, both atlastin null mutant and knockdown flies partially make it to adulthood. Immunohistochemistry experiments showed that Atlastin expression levels are elevated during embryonic development especially in syncytial and www.nature.com/nature 2
cellularizing blastoderm when the embryo develops under the control of maternally provided proteins 9. It is therefore likely that the robust maternal contribution, possibly coupled to slow protein turnover, allows progression through development. Drosophila eggs inherit a vast deposit of maternal gene products, which can mask the effect of zygotic loss of gene products until late in embryogenesis or well into larval development depending of the longevity of the gene product. The observation that surviving atlastin mutant females are sterile due to an inability to produce mature eggs 8, further underscores the importance of Atlastin in development. Deposition of maternal Atlastin would therefore ensure that ER biogenesis occurs unscathed throughout development. ER fragmentation set off by atlastin RNAi-mediated downregulation becomes manifest during larva development probably arising when maternally provided Atlastin becomes limiting and zygotic transcripts are eliminated by RNAi. Therefore, the observed fragmentation likely reflects a deterioration of ER maintenance rather than loss of de novo ER biogenesis. Overexpression of Atlastin severely impairs membrane traffic In neurons overexpressing Atlastin normal Golgi complexes were absent and Golgi markers relocated to the ER. Therefore, we tested whether Atlastin overexpression might compromise endocellular traffic by monitoring ER to-plasma membrane transport of mcd8-gfp. Overexpression of Atlastin produced an accumulation of mcd8-gfp in cytoplasmic agglomerates positive for the ER marker BiP, indicating that mcd8-gfp was not properly delivered to the plasma membrane but was instead trapped within structures with ER identity (Fig. S5b). Moreover, we found that transport of mcd8-gfp to the neuromuscular junction synapse was largely diminished in motor neurons overexpressing Atlastin (Fig. S5c, d), indicating that excessive fusion of ER membranes brought about by Atlastin severely impairs secretory traffic. www.nature.com/nature 3
Supplementary Figures Figure S1. Schematic comparison between Drosophila and human Atlastins. All Atlastins share elevated sequence homology with precise conservation of domain organization. TM, transmembrane domain. www.nature.com/nature 4
Figure S2. Endogenous Atlastin is localized on the ER. Atlastin was localized in Drosophila S2 cells by immunogold labelling. Representative electron micrograph shows that gold particles localize principally on ER membranes. Scale bar 0.5 µm (m, mitochondrion; n, nucleus; G, Golgi apparatus). The bar graph illustrates the quantitative analysis of gold particles distribution in different subcellular compartments. ER, endoplasmic reticulum; Mito, mitochondria; PM, plasma membrane; Cyt, describes gold particles that label the cytosol and do not associate with any membrane compartment; n.d. describes association with a membrane compartment not readily identifiable. www.nature.com/nature 5
Figure S3. Disruption of COP I and COP II complexes function results in disorganization and re-distribution of the Golgi apparatus to the ER. Transgenic RNAi lines targeting α-cop and Sar1 were employed to inactivate COP I and COP II, respectively. Muscle-specific downregulation of α-cop and Sar1 driven by Mef2-GAL4 causes the normally punctuate Golgi marker GalT-GFP (green) to become diffuse and overlap with ER membranes labeled with anti-bip antibody (red). Scale bar 10 µm. www.nature.com/nature 6
Figure S4. Endoplasmic reticulum exit sites are morphologically normal in Atlastin depleted tissue. ER exit sites marked with Sar1-GFP form small foci distributed throughout the larval muscle fibre (control) are present and essentially unchanged in tubulin-gal4/+;uas-atlastin-rnai/+ muscle (RNAi). In RNAi muscle, the ER visualized with anti-bip antibody displays a characteristic perinuclear enrichment. Because of the weak fluorescence of the Sar1-GFP reporter in muscles, an anti-gfp antibody was used for better visualization. Scale bar 10 µm (n, nucleus). www.nature.com/nature 7
Figure S5. Atlastin overexpression, but not its downregulation, leads to transport defects. (a, b) Confocal images of individual ventral ganglion neuron bodies revealed with plasma membrane localized mcd8-gfp. (a) mcd8-gfp distribution in controls and elav-gal4/+;uas-atlastin-rnai/uas-mcd8-gfp larva brains is essentially unchanged. (b) In motor neurons overexpressing Atlastin, mcd8-gfp fluorescence www.nature.com/nature 8
(green) displays an aberrant distribution and accumulates in globular cytoplasmic structures labeled by ER specific anti-bip antibody (red), absent in control neurons. (c) Confocal images showing the distribution of mcd8-gfp at third instar larva neuromuscular junction. mcd8-gfp fluorescence in controls and D42-Gal4/+;UASatlastin-RNAi/UAS-mCD8-GFP neuromuscular junction is essentially unchanged. In contrast, overexpression of Atlastin leads to a strong decrease of fluorescence intensity, indicating that mcd8-gfp is not properly delivered to the pre-synaptic membrane. (d) Histogram displaying normalized fluorescence intensity of mcd8-gfp at the third instar larva neuromuscular junction. Error bars represent s.d.. n>8; *p<0.00001. n, nucleus. Scale bars 10 µm. www.nature.com/nature 9
Figure S6. Efficiency of Atlastin depletion by in vivo RNAi. (a) tubulin-gal4/+;uasatlastin-rnai/+ escapers phenocopy the reduced size phenotype of atlastin null mutants. (b) Western blot analysis of lysates prepared from tubulin-gal4/+;uas-atlastin-rnai/+ flies demonstrates that endogenous Atlastin levels are greatly decreased. www.nature.com/nature 10
Figure S7. Immunofluorescence analysis of tissues depleted of Atlastin shows mild morphological defects of the ER. (a) tubulin-gal4/uas-gfp-kdel;uas-atlastin- RNAi/+ larval ventral ganglion neurons and (b) body wall muscles were analyzed by fluorescence confocal microscopy. Loss of Atlastin in these tissues results in enrichment of ER punctae in the vicinity of nuclei as visualized using the ER reporter GFP-KDEL (green). Simultaneous labelling with the p120 marker (red) reveals that Golgi morphology is preserved. Scale bar 10 µm. www.nature.com/nature 11
a G m m m n n UAS-atl/+; arm-gal4,atl1/atl1 b UAS-atl(K51A)/+; arm-gal4,atl1/atl1 * 1400 1200 1000 800 600 400 200 0 Control c atl1 Atl Atl(K51A) Distribution of ER length (%) 60 Control atl1/atl1 Atlastin Atlastin(K51A) 50 40 30 20 10 0 0-200 200-400 400-600 600-800 800-1000 1000-1200 1200-1400 >1400 Length (nm) Figure S8. Expression of wild type but not K51A Atlastin rescues ER fragmentation in the atl1 null background. Wild type and mutant Atlastin were expressed at 22 C using the ubiquitous driver armadillo-gal4. a, representative EM images of third instar larva brains. n, nucleus; m, mitochondria; G, Golgi apparatus; white arrows indicate ER. Scale bar 0.5 µm. Average ER profile length (b) and ER profile length distribution (c). Error bars represent s.d. n>100; * p<0,000001. www.nature.com/nature 12
Figure S9. Atlastin-HA is not immunoprecipitated by anti-myc antibody. In a control experiment HeLa cells were transfected with (a) Atlastin-HA and (b) Atlastin(K51A)- HA. Membrane vesicles were prepared as indicated and immunoprecipitated with anti- Myc antibody. www.nature.com/nature 13
Figure S10. GTP concentration dependence of Atlastin mediate proteoliposome fusion. Kinetic fusion graph of unlabeled acceptor Atlastin proteoliposomes (300µM total lipid) fused with fluorescently labeled donor Atlastin proteoliposomes (300µM total lipid) with decreasing concentrations of GTP in the reaction. Proteoliposomes were mixed in a 96 well fluoronunc polysorp plate and warmed to 37 C in TECAN Infinite M200 fluorescence plate reader. NBD fluorescence was measured at 1 minute intervals and detergent (2.5% dodecylmaltoside) was added at 60 minutes to determine maximum fluorescence. Data are represented as percent (%) of maximum fluorescence versus time. All fusion reactions were brought up to a final volume of 50 µl with buffer (25mM HEPES-KOH, ph7,4, 10% Glycerol 100mM KCl, 2mM BME 1 mm EDTA), 5 mm MgCl 2 and the indicated concentration of GTP. At lower GTP concentration, fusion abruptly stops when GTP supplies are exhausted. www.nature.com/nature 14
Figure S11. Atlastin mediated fusion of liposomes containing ER-like lipids. Kinetic fusion graph of unlabeled acceptor Atlastin proteoliposomes (300 µm total lipid) fused with fluorescently labeled donor Atlastin proteoliposomes (300 µm total lipid) made with two different lipid compositions. Proteoliposomes were mixed in a 96 well fluoronunc polysorp plate and warmed to 37 C in TECAN Infinite M200 fluorescence plate reader. NBD fluorescence was measured at 1 minute intervals and detergent (2.5% dodecylmaltoside) was added at 60 minutes to determine maximum fluorescence. Data are represented as percent (%) of maximum fluorescence versus time. All fusion reactions were brought up to a final volume of 50 µl with buffer (25 mm HEPES-KOH, ph7.4, 10% Glycerol 100 mm KCl, 2 mm BME, 1 mm EDTA), 5 mm MgCl 2 and 1 mm GTP where indicated. Atlastin fuses liposomes made of standard lipid mix (POPC:DOPS in a 85:15 mole ratio) or a lipid mix that resembles the lipid composition of the ER membrane 10 ; POPC:POPE:DOPS:cholesterol in a 55:20:15:10 mole ratio. www.nature.com/nature 15
Figure S12. Nucleotide dependence of Atlastin mediate proteoliposome fusion. Kinetic fusion graph of unlabeled acceptor Atlastin proteoliposomes (300µM total lipid) fused with fluorescently labeled donor Atlastin proteoliposomes (300µM total lipid) with different nucleotides in the reaction. Proteoliposomes were mixed in a 96 well fluoronunc polysorp plate and warmed to 37 C in TECAN Infinite M200 fluorescence plate reader. NBD fluorescence was measured at 1 minute intervals and detergent (2.5% dodecylmaltoside) was added at 60 minutes to determine maximum fluorescence. Data are represented as percent (%) of maximum fluorescence versus time. All fusion reactions were brought up to a final volume of 50 µl with buffer (25mM HEPES-KOH, ph7,4, 10% Glycerol 100mM KCl, 2mM BME, 1 mm EDTA), 5 mm MgCl 2 (or 1 mm GTPγS) and 1 mm of the indicated nucleotide. No fusion is seen unless GTP and magnesium are both present. www.nature.com/nature 16
Figure S13. Inner leaflet mixing driven by Atlastin. Inner-leaflet mixing is measured by reducing 7-nitro-2,1,3-benzoxadiazol (NBD) to the non-fluorescent 7-amino-2,1,3- benzoxadiazol (ABD) with a membrane impermeable compound, sodium dithionite. (a) Kinetic graph of unlabeled acceptor Atlastin proteoliposomes (300 µm total lipid) fused with fluorescently labelled donor Atlastin proteoliposomes (300 µm total lipid) treated with dithionite. Proteoliposomes were mixed in a 96 well fluoronunc polysorp plate and warmed to 37 C in TECAN Infinite M200 fluorescence plate reader. 100 mm dithionite in buffer (25 mm HEPES-KOH ph10, 10% Glycerol 100 mm KCl) was added to wells www.nature.com/nature 17
at -13.5 and -8.5 minutes. NBD fluorescence was measured at 30 second to 1 minute intervals and detergent (2.5% dodecylmaltoside) was added at 60 minutes to determine maximum fluorescence. (b) Data are represented as percent (%) of maximum fluorescence versus time. All fusion reactions were brought up to a final volume of 50 µl with buffer (25 mm HEPES-KOH, ph7.4, 10% Glycerol 100 mm KCl, 2 mm BME, 1 mm EDTA), 5 mm MgCl 2 and 1 mm GTP where indicated. Atlastin at 1:300 or 1:500 protein:lipid ratio promotes full fusion. www.nature.com/nature 18
Figure S14. Atlastin proteoliposome size change measured by dynamic light scattering (DLS). Light scatter was measured over time for a reaction containing unlabeled acceptor Atlastin proteoliposomes (300µM total lipid) and labeled donor Atlastin proteoliposomes (300µM total lipid). Nucleotides (1mM, GTP or GTPγS) was added at six minutes and fusion was triggered by the addition of magnesium at 15 minutes. A rapid and persistent Z-average size increase is seen with the sample containing GTP and magnesium, but not the GTPγS sample. At 25 minutes, EDTA is added to the reaction to stop fusion and resolve the size contribution of aggregated liposomes and fused liposomes. www.nature.com/nature 19
Figure S15. Transmission electron microscopy of negative stained Atlastin proteoliposomes. Atlastin proteoliposomes (1 mm total lipid) were incubated with or without 2 mm GTP and 5 mm Mg2+ for 15 min at 37 C. EDTA was added to 10 mm concentration to stop liposome fusion following incubation. Samples were absorbed to bacitracin-treated carbon/formvar coated copper grids (Ted Pella), incubated for 3-5 min at room temp, and washed and with sterile filtered 1% (w/v) ammonium molybdate. Samples were stained with 1% (w/v) ammonium molybdate for one minute at room temperature and dried for at least 30 min. Stained samples were imaged on a JOEL 1230 high contrast transmission electron microscope operated at 80 kev. (a) Unfused Atlastin proteoliposomes incubated without GTP. (b and c) Fused Atlastin proteoliposomes incubated with 2 mm GTP. Scale bars = 200 nm. www.nature.com/nature 20
Figure S16. Protein concentration in Atlastin proteoliposomes. Coomassie blue stained gels show the relative amount of protein from various proteoliposomes. (a) Proteoliposomes used in the fusion reactions shown in Figure 5b are shown. Six pmol of total lipid is loaded in each lane. Lane 1, 2.42 µg GST-Atlastin, lane 2, 1.67 µg GST- Atlastin, lane 3, 0.94 µg GST-Atlastin. (b) Proteoliposomes containing wild type and mutant Atlastin are shown. Lane 1 1.93 µg GST-Atlastin, lane 2, 2.61 µg GST- Atlastin(K51A). Note that more GST-Atlastin(K51A) is present in the proteoliposomes, yet fusion does not occur. www.nature.com/nature 21
References 1. Hatsuzawa, K. et al. Syntaxin 18, a SNAP receptor that functions in the endoplasmic reticulum, intermediate compartment, and cis-golgi vesicle trafficking. J Biol Chem 275, 13713-20 (2000). 2. Ward, T. H., Polishchuk, R. S., Caplan, S., Hirschberg, K. & Lippincott-Schwartz, J. Maintenance of Golgi structure and function depends on the integrity of ER export. J Cell Biol 155, 557-70 (2001). 3. Wilhelm, J. E., Buszczak, M. & Sayles, S. Efficient protein trafficking requires trailer hitch, a component of a ribonucleoprotein complex localized to the ER in Drosophila. Dev Cell 9, 675-85 (2005). 4. Friggi-Grelin, F., Rabouille, C. & Therond, P. The cis-golgi Drosophila GMAP has a role in anterograde transport and Golgi organization in vivo, similar to its mammalian ortholog in tissue culture cells. Eur J Cell Biol 85, 1155-66 (2006). 5. Xu, H. et al. Syntaxin 5 is required for cytokinesis and spermatid differentiation in Drosophila. Dev Biol 251, 294-306 (2002). 6. Rismanchi, N., Soderblom, C., Stadler, J., Zhu, P. P. & Blackstone, C. Atlastin GTPases are required for Golgi apparatus and ER morphogenesis. Hum Mol Genet (2008). 7. Lee, S. & Cooley, L. Jagunal is required for reorganizing the endoplasmic reticulum during Drosophila oogenesis. J Cell Biol 176, 941-52 (2007). 8. Lee, Y. et al. Loss of spastic paraplegia gene atlastin induces age-dependent death of dopaminergic neurons in Drosophila. Neurobiol Aging (2006). 9. De Renzis, S., Elemento, O., Tavazoie, S. & Wieschaus, E. F. Unmasking activation of the zygotic genome using chromosomal deletions in the Drosophila embryo. PLoS Biol 5, e117 (2007). 10. Cuzner, M. L. & Davison, A. N. The lipid composition of rat brain myelin and subcellular fractions during development. Biochem J 106, 29-34 (1968). www.nature.com/nature 22