Multi-color quantum dot tracking using a high-speed hyperspectral line-scanning microscope

Abstract

Many cellular signaling processes are initiated by dimerization or oligomerization of membrane proteins. However, since the spatial scale of these interactions is below the diffraction limit of the light microscope, the dynamics of these interactions have been difficult to study on living cells. We have developed a novel high-speed hyperspectral microscope (HSM) to perform single particle tracking of up to 8 spectrally distinct species of quantum dots (QDs) at 27 frames per second. The distinct emission spectra of the QDs allows localization with ∼10 nm precision even when the probes are clustered at spatial scales below the diffraction limit. The capabilities of the HSM are demonstrated here by application of multi-color single particle tracking to observe membrane protein behavior, including: 1) dynamic formation and dissociation of Epidermal Growth Factor Receptor dimers; 2) resolving antigen induced aggregation of the high affinity IgE receptor, FcεR1; 3) four color QD tracking while simultaneously visualizing GFP-actin; and 4) high-density tracking for fast diffusion mapping.

Figure 1. Conceptual diagram of microscope.

The excitation beam (optics not shown) is reflected by a dichroic mirror and forms a laser line focused at the sample plane by the objective, concentrating the excitation light to a small volume of diffraction limited width. Here, the white spheres in the sample represent fluorophores that remain mostly in the ground state while the colored spheres denote those which are excited. The emitted light passes through the dichroic mirror and into a spectrometer, which distributes the light onto an EMCCD camera such that each exposure captures information of wavelength and position along the line. The entrance slit on the spectrometer also serves to reject out-of-focus light, providing a semi-confocal ability for imaging at any depth in the sample. A scanning mirror (not shown) advances the line position by one back-projected pixel length on the sample and another exposure is acquired. One hyperspectral “frame” is a reconstructed series of these steps (performed in post processing) to form an image containing x, y, and λ. A time series of these hyperspectral frames is acquired at 27 fps, providing spatial, spectral, and temporal resolution that enables localized single molecule tracking of multiple emitters within a given diffraction limited volume. For a more detailed description, see Figure 2 and Text S1.

Figure 2. Optical layout.

This schematic layout of the laser line scanning hyperspectral microscope displays the key components of the system. Focal lengths are shown in units of mm. A detailed description and parts reference is provided in Text S1.

Figure 3. SPT of individual QDs non-specifically adhered to a glass surface.

(A) Progressing from blue to red are sum projections of raw data (solid) and Gaussian fits (dashed) for single 525, 565, 585, 605, 625, 655, 705 and 800 nm QDs. (B) 3D representation of a single hyperspectral time frame with 3D boxes representing sub-volumes identified for further particle localization. Sum projections of the data onto each plane are used to represent the raw data (gray scale). Likewise, 2D projections of each 3D box onto each plane are used to highlight the sub-volumes. Fitting results for the colored sub-volumes (color corresponds to the spectral center of the box) in (B) are shown in (C), (D), and (E). In the top figures in (C), (D), and (E) red and green are sum projections of the raw data and fit respectively. Note that the color outlining (C), (D), and (E) correlate with the respective sub-volumes highlighted in (B) with the same color. The localized particles are represented by white ellipsoids in which the radius in each dimension shows 3 standard deviations in the estimated error in the fit using the Cramér Rao Bound and their projection onto each axis is represented by a gray ellipse. The bottom figures in (C), (D), and (E) show raw spectral features (black) and Gaussian fits (color corresponds to fit spectral emission peak). The vertical gray lines represent the spectral cutoffs for independently fit sub-volumes. See Video S1.

Figure 4. SPT of 8-colors (525, 565, 585, 605, 625, 655, 705, 800 nm) of QD-EGF on live A431 cells.

(A) 3D (x,y,t) representation of trajectories. Trajectories are represented by colored lines. Selected interacting trajectories are highlighted by thicker lines. Interactions are grouped by boxes in the xy-plane (identified numerically). (B) Red-green-blue (RGB) representation of raw data for the last time frame with overlaid trajectories and boxes corresponding to (A). Pairwise interaction distances for the selected pairs of QDs are shown in (C), (D), (E), and (F). Note the text identifying how the pairwise interaction distances correspond to the boxes in which the QDs are observed in (A) and (B). See Video S2.

Figure 5. SPT of 5-colors (565, 585, 605, 625, 655) of QD-IgE on live RBL cells.

All subfigures are derived from a single acquisition (27 fps) ∼7 minutes after crosslinking with DNP₃. (A) 3D (x, y, t) depiction of trajectories. The color scheme for the trajectories is dependent on the estimated emission peak of each QD as noted by the color bar. A trimer (565 QD cyan diamond; 605 QD yellow square; 655 QD red up triangle) and dimer (585 QD cyan circle; 605 QD yellow down triangle) are highlighted by their respective symbols. (B) and (C) show localizations for specified time frames. Coloration and symbols correspond to (A). Ellipses in the upper subfigures represent 3 times the localization accuracies over a gray scale projection of the raw data. Red ellipse in the upper left subfigure of (C) is a localized 655 QD that doesn't interact with the dimer. Raw spectra (solid black lines) and Gaussian fit (colored lines with symbols) for individual QDs are shown in the lower subfigures. See Video S3.

Figure 6. SPT of 4-colors (605, 625, 655, and 705) of QD-IgE-FcεRI on live RBL cells transiently transfected with GFP-actin acquired at 27 fps.

(left) Two representative frames are shown. The GFP-actin presented in normalized gray scale is the spatially deconvolved portion of the spectrum between 500–570 nm. It is overlaid with Gaussian blobs reconstructed for localized QDs and trajectories for QD-IgE-FcεRI spanning the previous 3.7 s (100 frames; coloration of blobs and trajectories according to emission peak; see color bar). Colored symbols highlight QD-IgE-FcεRI positions at the specified frame, and representative spectra are shown to the right. Text identifiers (D, M1, M2, and M3) are used to indicate monomers (M#) and dimer (D) trajectories. See Video S4.

Figure 7. High density single particle tracking (QD 525, 565, 585, 605, 625, 655, 705, and 800).

(A) Shows the instantaneous diffusion for a mcSPT experiment with an acquisition rate of 27 fps of QD-IgE. 1 µg/ml of DNP-BSA was added ∼20 seconds into the acquisition (vertical dashed line). (B) and (D) show spatial maps of diffusion coefficients (estimated diffusion coefficient for each spatial bin is defined by the color scheme indicated in the color bar; red faster, blue slower, black insufficient sampling) for the same experiment shown in (A) estimated using all squared displacements observed between ∼5–15 seconds (B) and ∼30–40 seconds (C). A maximum likelihood estimator is used to estimate the diffusion coefficient (manuscript in preparation Cutler P J, Relich P, Lidke KA). See Video S5.