Two-color STED microscopy of living synapses using a single laser-beam pair

Abstract

The advent of superresolution microscopy has opened up new research opportunities into dynamic processes at the nanoscale inside living biological specimens. This is particularly true for synapses, which are very small, highly dynamic, and embedded in brain tissue. Stimulated emission depletion (STED) microscopy, a recently developed laser-scanning technique, has been shown to be well suited for imaging living synapses in brain slices using yellow fluorescent protein as a single label. However, it would be highly desirable to be able to image presynaptic boutons and postsynaptic spines, which together form synapses, using two different fluorophores. As STED microscopy uses separate laser beams for fluorescence excitation and quenching, incorporation of multicolor imaging for STED is more difficult than for conventional light microscopy. Although two-color schemes exist for STED microscopy, these approaches have several drawbacks due to their complexity, cost, and incompatibility with common labeling strategies and fluorophores. Therefore, we set out to develop a straightforward method for two-color STED microscopy that permits the use of popular green-yellow fluorescent labels such as green fluorescent protein, yellow fluorescent protein, Alexa Fluor 488, and calcein green. Our new (to our knowledge) method is based on a single-excitation/STED laser-beam pair to simultaneously excite and quench pairs of these fluorophores, whose signals can be separated by spectral detection and linear unmixing. We illustrate the potential of this approach by two-color superresolution time-lapse imaging of axonal boutons and dendritic spines in living organotypic brain slices.

Figure 1

(A) Scheme of custom-built STED microscope. The lasers for excitation (485 nm) and depletion (595 nm) are synchronized (small arrow) and their beams are brought into coincidence and scanned across the brain-slice cultures in the recording chamber. A vortex phase plate imposes the doughnut shape on the depletion beam. Emitted fluorescence from the sample is descanned and split into the two detection channels on either side of the 514-nm long-pass dichroic mirror placed in front of channel 2. Removing this long-pass filter reverts the STED setup to single-channel detection. (B) Excitation and excited-state depletion is separated spectrally by >100 nm, leaving open a wide spectral range for fluorescence detection. The emission spectra for GFP and YFP are shown, illustrating their differing relative distribution on either side of the 514-nm long-pass filter (dotted line) that allows linear spectral unmixing of the two detection channels. The emission spectra for Alexa Fluor 488 and Calcein green are very similar to that of GFP (not shown). (C) Imaging the same frame of fluorescent beads by confocal and STED microscopy reveals the gain in spatial resolution by STED. Plots of the bead intensity profiles corresponding to the dotted lines in the images at left are depicted in the graph at right. The gain in resolution achieved by STED (red line, right axis), compared to confocal images (black line, left axis), is evident. FWHMs (gray dotted lines) of the three beads are calculated from Gaussian fits of the intensity profile (black dotted and dashed lines) and presented in the graph.

Figure 2

(A) Depletion of YFP, GFP, Alexa Fluor 488, and Calcein green is largely independent of the depletion wavelength over the range 580–600 nm. The spectra were recorded at constant depletion power (∼42 mW into the aperture of the objective, with phase mask in place) and normalized to the maximum depletion achieved for each fluorophore. (B) Fluorescence depletion of the fluorophores measured inside living neurons in brain slices as a function of STED laser power at the back aperture of the objective. All values are normalized to the maximum depletion of GFP to facilitate comparison across fluorophores. The curves were recorded at 595 nm and under identical conditions. A maximal depletion of 89% was achieved for Alexa Fluor 488 at ∼50 mW of STED laser power (without phase mask).

Figure 3

(A–D) Examples of spine necks imaged using YFP (A), GFP (B), calcein green (C), and Alexa Fluor 488 (D) as volume labels of spine morphology. The profile plots of the blue bars across the spine necks (insets) illustrate that superresolved images can be acquired with all four fluorophores using the same excitation (485 nm) and depletion (595 nm) wavelengths. (E) Cumulative probability plot of spine neck widths of all mushrooms and thin spines in focus in the analyzed images, obtained with the four respective fluorophores. It is clear that the majority of spine necks are thinner than what can be resolved with conventional diffraction-limited techniques (diffraction limit indicated by yellow bar). Student unpaired t-tests returned no significant differences between the neck widths reported by the four fluorophores, suggesting that they are equally well suited for imaging dendritic spine neck morphology under the given settings.

Figure 4

(A) YFP and GFP imaged together in two channels but merged using a gray lookup table. (B) The two-channel raw data merged in red and green colors, respectively. It is noteworthy that from merging the two mixed channels in different colors, structures expressing either YFP or GFP can already be readily discriminated due to their differing relative contribution in each of these. (C) Linear unmixing of the two fluorophores results in clear separation of these structures, leaving <10% residual cross talk in the channels. (D and E) Zoom-ins of the outlined regions in C, depicting examples of superresolved structures labeled by GFP (D) and YFP (E), along with their corresponding line intensity plots and FWHM values. Scale bars, 1 μm.

Figure 5

Time-lapse z-stack imaging of YFP (red) and GFP (green) synaptic structures did not result in any apparent photodynamic damage. Fluorophore bleaching was not an issue, presumably because the fluorophores are freely diffusible, allowing for replenishment even after bleaching. All frames depict maximal intensity projections of z-stacks. (E and F) Magnifications of the outlined structures in A, illustrating that both fluorophores are superresolved in the images. Scale bars, 1 μm.

Figure 6

(A) Two detection channels merged when imaging YFP and Alexa Fluor 488. (B) As described, by assigning red to one channel and green to the other, structures labeled by either of the two fluorophores can be readily discriminated. (C) The linearly unmixed fluorophores are unambiguously separated, revealing axonal structures genetically labeled by YFP and two crossing dendrites whole-cell labeled with Alexa Fluor 488. The FWHMs of the intensity profiles of the two white lines illustrate that both fluorophores are superresolved in the image. (D–F) Images of slices colabeled with YFP and calcein green. The two fluorophores are easily distinguishable simply by assigning different colors to the two detection channels (E), and linear unmixing can clearly separate the fluorophore signals (F). Among the dense network of calcein-green-labeled axons in F, a few YFP (red) axons can be seen, which may originate from the same cell as the genetically YFP-labeled dendritic segment displayed or from other YFP-expressing cells in the slice culture. Again, as seen from the FWHMs of the depicted line profiles in F, both fluorophores are superresolved. The relatively high signal background in the calcein-green channel (D–F) stems from the ester-loading procedure, where many cells are loaded around the bolus injection site. Images are maximum-intensity projections of z-stacks. Scale bars, 1 μm.

Figure 7

Channel cross talk before and after unmixing of the fluorophores, expressed as the ratio of signal intensity for the same region in the two detection channels. Before unmixing, the channel ratios of the four respective fluorophores differ, in agreement with their emission spectra and the 514-nm long-pass emission filter. Unmixing effectively separates the fluorophores into separate channels, leaving ∼10% residual cross talk regardless of the initial signal ratios.