Supplementary Figure 3: Deconvolution of gated STED images.

Neurons were transfected with EGFP at DIV 0 and stained for endogenous PSD-95 at DIV 21-25. EGFP labeling was enhanced by staining with a GFP antibody (Online Methods) and visualized by an AlexaFluor-488-conjugated secondary antibody. Endogenous PSD-95 was simultaneously visualized in two channels by co-staining with primary antibodies to PSD-95 with AlexaFluor-594 and Atto-647N secondary antibodies (as in Fig. S1). After acquiring a confocal image using a Leica SP8 3x gated STED system a series of ten STED images were obtained using a pulsed 775 nm depletion beam on a single field of view to induce signal weakening due to bleaching. Gated detectors were used for both AlexaFluor 594 (0.3 to 6 nsec) and Atto-647N (0.2 to 6 nsec). Images were then deconvolved offline using Huygens deconvolution software (Online Methods). This software conducts deconvolution based on the laser intensities used, microscope PSF, background and signal-to-noise ratio in each channel independently. A maximum of 40 iterations was used. The inset demonstrates two spines with high PSD-95 signal that appear to contain a single PSD-95 cluster in both channels in confocal mode. Appearance of two discrete PSD-95 clusters is observed in raw STED images in both Atto-647N and AlexaFluor-594 channels in one spine (double arrow), while the neighboring spine has only one cluster (single arrow). Deconvolution further improved the resolution, resulting in two clearly separated clusters in a double clustered spine, but did not induce cluster artifacts (single cluster spine is still single clustered after deconvolution). The same clusters could be visualized using deconvolution even after bleaching reduced the signal-to-noise ratio (after the tenth STED image), indicating that deconvolution is unlikely to generate puncta resembling PSD-95 even in poor signal-to-noise conditions. Similar results were obtained from three independent experiments. Scale bars, 1 µm.