Seeing deep and wide with a microscope

Hank Hogan

Researchers can image the fluorescence of a point while rejecting other light using several methods, then zoom in on a location to improve the captured data. But to see everything requires scanning the sample. Now investigators from Boston-based Harvard Medical School and the Wellman Center for Photomedicine, both associated with Massachusetts General Hospital, have demonstrated a microscopy technique that eliminates scanning and offers a ranging depth of hundreds of microns and fields of view greater than 1 mm.

Alberto Bilenca, an instructor at Harvard Medical School and a member of the research team headed by professors Brett Bouma and Guillermo Tearney, noted that the method could prove useful in getting the big picture — in more ways than one. “The unique capabilities of this technique may open up new possibilities for molecular and dynamic investigations in life sciences in a more organismal context,” he said.

Bilenca explained that the approach, which the researchers dubbed spectral-domain fluorescence coherence tomography, employs fluorescence self-interference.

For spectral-domain fluorescence coherence tomography, the fluorescent sample is located between two matched, opposing low-numerical-aperture objectives (near z₀) and is illuminated with a line focus at the excitation wavelength (green). Self-interference fluorescence from the sample (orange) is imaged along the transversal dimension and spectrally resolved in the two-dimensional CCD array of an imaging spectrometer.

Fluorophores in a sample are excited by an appropriate light source, and their emission is captured by two matched and opposing low-numerical-aperture objectives. After being routed by mirrors, the two beams pass through an interferometer, where they interact. The self-interference is detected by an imaging spectrometer, and the depth information of each fluorophore is encoded in the interferometric signal of the emission spectrum. Thus, the depth, or ranging, profile of the fluorophore distribution is extractable from the captured emission. What is more, this information is provided over a wide area.

Spectral-domain fluorescence coherence tomography requires bright fluorophores and a detector with low noise and high sensitivity because, in part, the low-numerical-aperture optics provide a large field of focus but are not very efficient at collecting light.

Spectral-domain fluorescence coherence tomography was used to image a dual-layer fluorescent sample (top), and the tomogram (bottom) shows the signal detected across the dashed line in the top image.

Another contributing factor is that the camera must be operated such that shot noise is limited; thus, statistical fluctuations resulting from varying photon counts can be detected. Consequently, the fluorophores cannot be so bright as to saturate the camera.

Bilenca noted that a lower camera noise makes it more feasible to detect dim fluorescence and widens the camera’s dynamic range, increasing its ability to detect bright signals. Those parameters, however, must be balanced against the fluorophore emission, which must be enough but not too much.

In a demonstration of the technique, the researchers used an electron-multiplying CCD camera from Photometrics of Tucson, Ariz., and 100-nm-diameter fluorescent nanospheres from Duke Scientific Corp. of Fremont, Calif. The beads had an excitation wavelength of 540 nm and an emission wavelength of 610 nm.

The investigators placed the nanospheres in a solution, set a drop of the solution on a standard 170-μm-thick glass coverslip, then dried the drop to leave behind the nanospheres. They glued a second coverslip on top of the beads. For a light source, they used an Nd:YAG laser from Coherent Inc. of Santa Clara, Calif., that was frequency-doubled to operate at 532 nm.

They varied the length differences between the two arms of the interferometer from 50 to 250 μm. When they plotted the averaged signal from 10 measurements against axial position, they could clearly see spikes where the single layer of fluorescent beads was located. They estimated that the axial resolution was between 3.29 and 3.45 μm, with the value dependent upon the fluorophore. They calculated that GFP would have an axial resolution of 2.8 μm, while that of cyan fluorescent protein would be 1.8 μm.

As described in the Aug. 7 issue of Optics Express, they also constructed a sample with fluorophores at two distinct levels by fabricating beads on coverslips that were separated by a 120-μm-wide gasket, filling the gap with a transparent epoxy to minimize reflections. They averaged information from five consecutive images, each acquired in 0.1 s. They measured the mean distance to be 120.1 μm, with the beads visible over a transverse length of >1 mm.

Although the technique works, there are some limitations on the fluorophore distribution. For example, a continuous spread of fluorophores suffers from degraded spectral-domain fluorescence coherence tomography sensitivity because of emission overlap. The best situation is one where the fluorophore labeling follows specific guidelines.

“The labeling consists of thin — that is, an axial width smaller than one-half of the fluorophore’s center emission wavelength — fluorescent probes separated axially by greater than one-half of the fluorophore’s coherence length, which is typically a few microns,” Bilenca said.

Under these conditions, the fluorophores can be distributed across the usable ranging depth of spectral-domain fluorescence coherence tomography, which is hundreds of microns. This assumes, Bilenca added, that the samples are transparent enough for the fluorescent signal to emerge from that depth.

In one ongoing investigation, the researchers are looking at the technique’s performance in turbid media. In another, they are developing an extension that uses moderately high numerical aperture objectives for improved depth sectioning, higher lateral resolution and detection sensitivity in optically dense specimens. A second extension uses phase information in the spectral-domain fluorescence coherence tomography signal to enable nanometer-scale localization of fluorescent probes.

Both, Bilenca said, currently are being analyzed theoretically and tested experimentally.