Hyperspectral Confocal Fluorescence Microscope: A New Look into the Cell
M. Cristina Pedroso,1 * Michael B. Sinclair,2 Howland D.T. Jones,2 and David M. Haaland2 1 Monsanto Company, 800 N Lindbergh Blvd, Crop Analytics, U106N, St. Louis, MO 63117 2 Sandia National Laboratories, Albuquerque, NM 87185-0895
*
cristina.ubach@monsanto.com
Introduction Confocal microscopy is widely used in cell biology. Like
other filter-based systems, traditional confocal microscopes are limited by the spectral bands established by each optical filter. As a result, emission spectra from labels and/or autofluo- rescence can be overlapped leading to spectral crosstalk and inability to quantify the amount of signal originating from each individual fluorescent species. Te need for accurate quantification of in vivo cellular processes and in-depth knowledge of organelle development and microstructure led Monsanto to search for non-commercial microscopes that could achieve those goals. Trough a cooperative research and development agreement (CRADA) established between Monsanto and Sandia Corporation in August 2006, we built a new 3D-hyperspectral confocal fluorescence imaging system, specifically designed to meet the analytical requirements of plant specimens.
Instrumentation Te hyperspectral confocal fluorescence microscope is a
non-filter-based imaging system developed for the quantitative analysis of fluorescence (primary or secondary) from biological and non-biological specimens. Te instrument simultaneously records the entire emission spectrum from 500 nm to 800 nm (512 wavelengths) for every voxel within the imaged volume with a spectral resolution of at least 3 nm [1]. Te current optical design employs a solid state 488-nm laser (Coherent Inc.) as the excitation source and a confocal pinhole to maximize spatial resolution and optical sectioning performance (Figure 1). Photon detection efficiency was optimized using a high-throughput prism spectrometer and a high quantum-efficiency electron multiplying charge coupled device (EMCCD) detector (Andor Technologies). A novel readout mode allows data to be acquired at a rate of ~8300 spectra per second, so that a two-dimensional hyperspectral image of a 25 µm by 25 µm field of view, sampled at 0.12-µm intervals, can be acquired in 5 seconds. Te readout mode also eliminates keystone and curvature distortions that oſten plague spectral imaging systems. Epi and Trans light emitting diodes (LEDs) and an Andor Luca camera were added to the microscope for conferring wide-field epi-fluorescence capability for quick selection of regions of interest on the specimen prior to hyperspectral image capture (Figure 2). For a detailed description of the instrument see Sinclair et al. 2006 [1]. Te information-rich and very large hyperspectral images
collected by our microscope require fast and efficient analysis algorithms to provide qualitative and quantitative information about the images in a timely manner. Sandia National Laboratories has developed such algorithms using multivariate curve resolution (MCR) and have incorporated these algorithms into their MCR analysis soſtware package (runAxsia)
14
customized for pre-processing, analyzing, and viewing these types of hyperspectral fluorescence images. Tis hyperspectral image analysis soſtware was written in Matlab (Te MathWorks, Inc. Natick, MA) and calls C++ code (AXSIA) [2], which is the engine for all MCR analyses. Te MCR algorithms use a constrained alternating least-squares implementation to extract the pure spectral components (the qualitative information) and the corresponding intensities (the quantitative information) of each spectral component for each voxel in the image. Te advantage of using this MCR soſtware is that it can discover and quantify all independently emitting species within the image that are above the noise without the need for standards or any a priori information about the sample. Tis makes MCR an ideal analysis technique when using our hyperspectral microscope for exploring unknown biological samples [4–7]. Tere are other commercial soſtware packages that utilize MCR to analyze spectroscopic data (for example, soſtware from CAMO, Inc.), however Sandia’s implementation of MCR is very fast and memory efficient [2]. A single hyperspectral image containing over 40,000 spectra can be analyzed in less than 5 seconds. To improve our MCR results, it is necessary to employ
pre-processing techniques to remove unwanted features from the hyperspectral image datasets. Tese pre-processing steps include removing: 1) spikes in the spectral data that are the result of cosmic rays hitting the detector, 2) offsets that were added to each spectrum from the EMCCD detector, and 3) a
Figure 1: Optical layout of the hyperspectral confocal fluorescence microscope.
doi:10.1017/S1551929510000854
www.microscopy-today.com • 2010 September
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76