(Aa) Surface model of the assembly from the principal docked vesicle shown in Figure 3F–I. (Ab) Surface model of an assembly from another docked vesicle. (Ba,b) Surface models of assemblies from undocked vesicles. The assemblies in (Ab,Ba,Bb) were rotated until their shape matched that of the principal vesicle. The NMJ’s used for the assemblies in (A) and (B) were fixed by rapid freezing. In all cases, the assemblies are bilateral with irregular arms radiating from near the center of the vesicle. Nubs arise from the arms to connect at their end to the vesicle membrane; the terminal 3 voxels of each nub were made blue to mark its connection site on the membrane. (Ca–c) Alignment model generated by gray-scale density alignment of the 12 docked and undocked vesicles in the two primary reconstructions, which included those shown in (Aa,Ba). In (Ca) the alignment model is oriented according to the orientation of the surface model of the assembly from the principal vesicle as shown in (Aa). In (Cb) the alignment model is rotated 90 degrees to the right around the vertical axis in (Ca). In (Cc) the alignment model is rotated 180 degrees around the vertical axis in (Ca). The alignment model shows a bilateral arrangement of 4 radiating arms as do the models of individual assemblies in (Aa,b,Ba,b). (Da–c) The alignment model (red) generated from gray-scale density alignment of the 12 luminal assemblies in (Ca–c) aligned together with two alignment models (blue and green) generated from the same 12 assemblies represented according to their surface models. The different orientations of the superimposed alignment models are the same as for that in (Ca–c). The alignment models, which were calculated to have 95% 3D overlap, were used to measure the similarity in the shape of the assemblies from vesicle to vesicle and for establishing the orientation of the shape from vesicle to vesicle. (Ea–c) The surface model in (Aa) inserted into the alignment model in (Ca–c) according to its position in the alignment model. The different orientations of the combined models are the same as in (Ca–c). The portions of the surface model least included in the alignment model are the nubs, probably due to the nubs being smaller and/or having a somewhat greater variability in positioning than the arms among the surface models used for generating the alignment model.