LS-3-O-2434 Single-molecule correlative microscopy using optical excitations in low energy-loss spectra of quantum dots

Correlative light and electron microscopy allows the combination of dynamic imaging in the light microscope (LM) with visualization of cellular ultrastructures in the electron microscope (EM). The use of fluorescent, semiconductor nanocrystallites (quantum dots, QDs) offers a way to apply only one marker for both modalities. Since they are more electron opaque than biological materials, QDs can be identified in the EM (Fig. 1).
Multi-colour labelling is feasible with QDs in LM but separation in EM is hindered since conventional contrast of different QD particles is similar. Use of different sizes or shapes would allow for segmentation. To prevent varying labelling efficiencies and penetration depths into plastic sections [1], we use two different commercially available QDs: QD-655 and QD-705 (Molecular Probes, USA). The only difference of these core-shell particles is a minimal variation in core composition to alter optical absorption and emission characteristics.
We prepared ultrathin sections of resin (HM20) embedded muscle tissue containing neuromuscular junctions. Acetylcholine receptors (AChRs) of the synapse were labelled by antibody-conjugated QDs-655 and QDs-705. As shown in the light micrographs in Fig. 2A,B, the fluorescent signal of both types is concentrated at synapses. This is verified by the signal of an Alexa Fluor 555 dye conjugated to bungarotoxin, which was used as control to label AChRs (Fig. 2C). Synapses can be relocated in the EM with single QDs at the postsynaptic membrane (Fig. 2D). To separate the randomly distributed QDs we applied scanning transmission electron microscopy (STEM) in combination with electron energy-loss spectroscopy (EELS) at both high spatial and energy resolution. This allows acquisition of EEL signals in the low-loss regime including optical excitations. To identify the spectral features we used samples that were labelled with either only QD-655 or only QD-705 and acquired EELS data sets for single particles. Multivariate statistical analysis and machine learning was applied to learn on the known data sets followed by classification of the unknown data sets from the sample with combined labelling. An ensemble of particles at a synapse is shown in Fig. 3A,B. Selected QDs were classified and highlighted. EEL spectra from the learning and classified sets (Fig. 3C) indicate the separating features, which are attributed to low-energy interband excitations.
Our results show that minor differences of spectral properties of correlative markers on biological samples can be detected in the analytical STEM. This opens the door to multi-colour EM imaging at the single molecule level.
[1] Giepmans, B.N.G. et al.: Nature methods (2005), 2, 743.
[2] Microscopes: FEI Titan G2, Libra 200 (Carl Zeiss Microscopy GmbH).