Figure 4. HVC synapse sizes are variable and weakly affected by tutoring.

(A) We acquired close to isotropic imagery of an (8 μm)³ cube of HVC tissue using focused ion beam electron microscopy (FIBSEM). (B) Example image of excitatory (red) and inhibitory (blue) synapses. (C) Example reconstructed excitatory (red) and inhibitory (blue) synapses. Shown are the synapses in the original FIBSEM dataset (left), 3D orthoslices intersected at the synapse centers with the segmented synapse voxels shown in color (middle), and 3D reconstruction of the segmented voxels with arrows pointing to the postsynaptic side (right). Scale bars: 0.5 μm. (D) Excitatory (asymmetric) synapses in tutored (SHORT) birds were larger than in untutored (ISO) birds, no significant change in synapse size was associated with extensive tutoring (LONG birds). (E) Neither SHORT nor LONG tutoring was associated with significant changes in inhibitory synapse sizes. C-D White/gray bars represent group means and the error bars represent the means in individual birds ± the standard deviations.

Figure 4—figure supplement 1. Histograms of synapse sizes per bird group and synapse type.

Sizes (X-axis) were log-transformed. Gaussian fits are shown in red.

Figure 4—figure supplement 2. Segmenting synapses in FIBSEM imagery using Ilastik.

(a) This sample image served as a reference for panels (b) – (d). (b) Objects (synapses in red, membranes in green, vesicles in purple, mitochondria in blue, and remainder in yellow) were manually labeled and served as ground truth for automated segmentation. (c) Ilastik-predicted pixel identities in corresponding colors. Note that despite the small mistake in b of labeling parts of a mitochondrion as ‘remainder’, the predicted pixel classification is robust and mitochondria are mostly correctly identified. (d) Binary image of segmented synapses. Pixels segmented as synapses are shown in red, superimposed on the original EM image.