Supplementary Figure 2: Procedure for image thresholding and deconvolution.
From: Synaptic nanomodules underlie the organization and plasticity of spine synapses

Raw confocal and STED images from the control condition from which the spine in Fig. 5a was chosen. Dendritic arbors were labeled by transfecting neurons with tdTomato at DIV0. Images are of transfected neurons stained at DIV21-25 with primary antibodies to mRFP, PSD-95 and vGlut1 as indicated. a) Images for tdTomato were acquired in conventional confocal mode using a Cy3 conjugated secondary antibody and a 561 nm excitation laser line (AOTF = 7%). Imaging was performed using an SP5 Leica instrument. Anti-PSD-95 was visualized by using an Atto-425-conjugated secondary antibody that was excited by a 442 nm line (AOTF = 35%). Raw STED PSD-95 images were obtained by depleting with a 592 nm line (AOTF = 90%). Anti-vGlut1 was visualized by labeling with an AlexaFluor-488-conjugated secondary antibody excited with a 488 nm line (AOTF = 12%). Raw STED images were obtained by depleting with a 592 nm line (AOTF = 100%). Clear PSD-95 (green) and vGlut1 (red) puncta showing a high degree of apposition (merge) were obtained using this method. b) Raw STED images from panel a were subjected to background subtraction (mean intensity of all pixels in the image) followed by a Gaussian blur (2 pixel size). c) Deconvolution of raw STED images in panel a in SP5 Leica Application Suite Advance Fluorescence software using 80 nm microscope resolution. Scale bar for a, b and c, 2µm. d) Line profiles (dotted line) through the puncta in a control spine (also shown in Fig. 5a) demonstrated sub-diffraction size PSD-95 (FWHM = 210 nm) and vGlut1 puncta (FWHM = 190 nm) in unprocessed STED images. e) Gaussian blur/background subtraction performed on raw images resulted in visibly brighter puncta with FWHM = 220 nm (PSD-95) and FWHM = 200 nm (vGlut1). f) Deconvolution resulted in puncta fluorescence intensity that was comparable to the Gaussian blur/background subtraction method of analysis used in panels b and e. Deconvolution improved resolution of both PSD-95 puncta (FWHM = 190 nm) and vGlut1 puncta (FWHM = 150 nm), but it did not result in artifactual PSD-95 or vGlut1 clusters. Scale bar for d, e and f, 1µm. Similar results were obtained from three independent experiments (see also Fig. 5).