Open Access
Add to BibSonomyAbstract
We introduce a fast spectral imaging system using an electron-multiplying charge-coupled device (EM-CCD) as a detector. Our system is combined with a custom-built two-photon excitation laser scanning microscope and has 80 detection channels, which allow for high spectral resolution and fast frame acquisition without any loss of spectral information. To demonstrate the efficiency of our approach, we applied this technology to monitor fluorescent proteins and quantum dot-labeled G protein-coupled receptors in living cells as well as autofluorescence in tissue samples.
©2010 Optical Society of America
1. Introduction
Since the advent of commercial confocal laser scanning microscopes, new fluorescence microscopy systems [1–3] and techniques derived therefrom [4,5] have been developed. One of them is two-photon excitation (2PE) laser scanning microscopy, which is well suited for imaging not only of single cells [6,7] but also and especially of thick tissue [8,9]. As two-photon microscopy provides deeper penetration depth, less photobleaching, less scattering and reduced noise level compared to single photon excitation confocal microscopes due to the use of an infrared excitation source far away from the visible emissions, it is a very powerful tool for numerous biological applications.
With the increasing use of fluorescence microscopy especially in biology [1–5], demand for the simultaneous acquisition of multiple fluorescent probes is growing. Especially for a quantitative systems biology approach to study molecular functions in living cells and tissues, the simultaneous readout of multiple parameters is essential. Among the spectral imaging point scanning systems that have been developed or are commercially available [10,11], typical implementations employ a grating or prism to disperse the confocal signal from each scanned point over a linear photomultiplier tube (PMT) array [7] or a charge-coupled device (CCD) chip [12–14] or use one or several large sensitive area PMTs combined with moving slits. In terms of temporal and spectral resolution, PMT array-based systems enable fast acquisition but can suffer either from loss of spectral information due to the dead zone between neighboring PMT elements or from low spectral resolution due to a small number of detection channels (Fig. 1 ) [11,15]. On the other hand, moving slit-based systems can provide very high spectral resolution, especially when reducing the slit width, but require relatively long recording times because of the inherent sequential image acquisition scheme. In addition, photobleaching might play a larger role due to the long exposure time. A promising approach is the use of CCD-based detection [12–14], but the read-out speed of CCD cameras still presents a limiting factor to the pixel dwell time. These drawbacks limit the broader use of spectral imaging for biological applications that require simultaneously high spectral resolution and fast acquisition times. Furthermore, the low quantum efficiency of PMT arrays especially towards the red end of the spectrum and the need for dedicated hardware for data acquisition and processing of PMT signals make these approaches challenging.
Fig. 1 Overview and comparison of spectral acquisition using an EM-CCD with common schemes using linear PMT arrays and moving slits. The effective detection areas are indicated in gray.
Recently, electron-multiplying charge-coupled device (EM-CCD) detectors have found their way into fluorescence microscopy and spectroscopy, not least due to their substantially higher sensitivity compared to conventional PMTs. As an EM-CCD is typically composed of several hundreds of pixels in each dimension arranged with a fill factor close to unity, it can be used as a spectrometer detector to provide higher spectral resolution than typical PMT array-based systems (Fig. 1). We could show previously that EM-CCDs can be used for spatially or spectrally resolved fluorescence correlation spectroscopy (FCS) based on single photon excitation laser scanning microscopy [16,17].
To overcome the drawbacks of current spectral imaging systems, we developed an EM-CCD-based spectral detection system and a 2PE laser scanning microscope for imaging with fast acquisition speed, high spectral resolution and high sensitivity. Here, we introduce the setup and show its application to biological systems such as cells expressing fluorescent proteins, tissue sample exhibiting autofluorescence and cells carrying quantum dot-labeled G protein-coupled receptors (GPCR).
2. Experimental setup
Figure 2A depicts schematically our custom-built 2PE laser scanning microscope setup. A tunable Ti-Sapphire laser (MIRA900, Coherent) is used for two-photon excitation in the range of 700-980 nm. The laser pulse width is set to ~130 fs and the repetition rate to 76 MHz. Using two convex lenses, the light is expanded to a full width at half maximum (FWHM) of 2.4 mm (laser and beam expander are omitted in Fig. 2A). The scan head is composed of two galvanometer mirrors (VM500, GSI). The light is expanded again by a scan and a tube lens to overfill the back aperture of an oil-immersion objective lens (HCX PL APO 63 × , NA 1.4, Leica) used to focus the light onto the sample and mounted on an inverted microscope (DM IRBE, Leica).
Fig. 2 (A) Schematics of the 2PE spectral fluorescence imaging setup using an EM-CCD camera. (B) Upon 2PE point-scanning excitation of the fluorescent sample, a spectral data stack comprised of 80 images is generated. B: blue, G: green, R: red. (C) Only 80 pixels in the bottom-most line of the detection area of the EM-CCD chip are illuminated with the spectrally dispersed fluorescence light at least for the time required to amplify and read out 80 pixels. Then a rapid single line shift is applied to move the photoelectrons to the first line of the storage area while the galvanometer scanner moves the focus to the next pixel. Eventually the photoelectrons from the 80 illuminated pixels arrive at the bottom-most line of the storage area. Only the 80 pixels of interest instead of the full line of 656 pixels are transferred to the charge amplifiers, the analog-to-digital converter and to the frame grabber. This is repeated for all pixels of an image frame and the recorded spectra are assigned appropriately to the respective pixels. (D) Line readout rate as a function of pixels per line that are read out. The pixel number/readout rate combination used in this study is highlighted.
Fluorescent light emitted from the sample passes objective, tube and scan lenses, and the galvanometer scanners. It is then reflected by a dichroic mirror, dispersed spectrally by a Pellin-Broca prism and focused onto a single line of an EM-CCD sensor (SamBa SE-34, Sensovation). As a result, we can obtain a full spectrum from any spatial point of the acquired image (Fig. 2B).
To enhance the image acquisition speed, we increased the readout frequency of the EM-CCD by restricting the recorded area to 1 × 80 pixels and operating it in fractional line-readout mode, i.e., pixel intensities were transferred line-by-line instead of frame-by-frame to the frame grabber (Fig. 2C). When setting the integration time of the camera to zero, ~84,000 spectra of 80 pixels could be recorded per second (Fig. 2D). Thus, the pixel dwell time was decreased to ~12 μs, which is an order of magnitude faster compared to using a full line of the chip [13]. Both the frame grabber and the galvanometer scanners were synchronized using the pixel clock of the EM-CCD, and the pixel dwell time was set to match the reciprocal of the readout frequency of the EM-CCD. For rapid assessment of the data, the measurements were monitored on a computer screen in real time.
Even though the flyback of the galvanometer scanning along the line is not used for data recording, the gain of speed is reflected in the acquisition time of full spectral stacks of 80 frames: our system operates in a low sampling mode (205 × 205 pixels) with an acquisition time of ~1 s per stack and in a high sampling mode (458 × 458 pixels) with an acquisition time of ~5 s per stack, allowing for full spectral time-lapse imaging and proving to be a fast yet spectrally well resolved alternative to other point scanning instruments: In [13], a setup with a higher spectral resolution of 512 channels but a longer pixel dwell time of 240 μs is presented with an anticipated acquisition time of ~5-10 s for a 208 × 208 pixel stack. A typical commercial confocal microscope employing a ‘moving slit’ spectral detection system (e.g. Leica TCS SP2 AOBS, Leica Microsystems) would require ~10 s for the acquisition of 40 spectral channels with a speed of 4 frames (256 × 256 pixels) per second.
In summary our spectral imaging system features a simple yet powerful and robust optical setup and is a good compromise between acquisition speed and spectral resolution adapted to imaging of living biological samples.
3. Sample preparation
3.1 Constructs and cell culture
The plasmids encoding for the cyan and yellow fluorescent proteins (CFP, YFP) derived from the Aequorea Victoria jellyfish and for the DsRed fluorescent protein originating from Discosoma Striata sea coral were obtained from Clontech. The streptavidin binding protein-tagged serotonin 5-HT2C receptor expression vector was generated upon insertion, after the signal sequence of the human 5-HT2C, of an SBP-tag sequence [18]. The SBP-5-HT2C was stably expressed in HEK-293 cell lines, and the fluorescent proteins (CFP, YFP, DsRed) were transiently expressed in HEK-293 cells.
3.2 Quantum-dot cell labeling
Cells stably expressing the SBP-5-HT2C construct were seeded 72 hours before imaging in 4 well LabTek chambered coverglasses (Nunc). 24 hours post cell seeding, the cell culture medium was replaced with DMEM supplemented with 1% FBS, which promotes the 5-HT2C translocation at the plasmalemma cellular compartment. Three different types of streptavidin-conjugated quantum dots (QD 525, QD 585 and QD 655, Invitrogen) were incubated sequentially for 30 minutes at 24, 6 and 1 hour(s), respectively, prior to spectral image acquisition (see Fig. 7A ). One hour after each QD treatment, the cells were carefully washed with phenol red-free DMEM.
Fig. 7 Multiple QD pulse-chase staining of 5-HT2C receptors undergoing endocytosis and spectral images. (A) SBP-tagged 5-HT2C receptors were incubated sequentially, washed and kept in the chamber at 37°C until the next pulse-chase staining. This took place at 24 hours, 6 hours and 1 hour, and used three different streptavidin-conjugated QDs for 30 minutes. Image stacks were acquired 30 minutes after the last staining (B). The images representing (C) QD 525, (D) QD 585 and (E) QD 655 were obtained from the channels specific to each peak, ranging from 526 to 535 nm for QD 525, 576-588 nm for QD 585 and 635-647 nm for QD 655. The scale bar represents 20 μm. In the overlay image (B), spectrally different QD-labeled 5-HT2C can be found in intracellular compartments (F and G) and in the plasmalemma (H) based on the receptor internalization rate and depending on the pulse chase time point.
4. Characterization of the setup
4.1 Spatial resolution
In order to measure the point spread function (PSF) of our two-photon spectral imaging system we performed a 3D scanning of 0.17 μm diameter yellow-green fluorescent beads (P7220, Molecular Probes), with an excitation wavelength of 900 nm. The pixel interval was set to 0.14 μm and sequential images were taken by moving the objective lens with steps of 0.05 μm. Figure 3A and 3B show a lateral section of a fluorescent bead and its fluorescence spectrum, respectively. Figure 3C and 3D show the intensity profiles of axial and lateral line scans obtained by averaging 10 beads. The FWHMs of axial and lateral line profiles are 0.51 μm and 0.36 μm, respectively, i.e., our setup provides close-to diffraction-limited axial and lateral resolution.
Fig. 3 Point spread function measurement. (A) Lateral section of a 0.17 μm diameter fluorescent bead excited at 900 nm. (B) Emission spectrum of the bead. The FWHM values of the beads obtained from axial (C) and lateral (D) line scans are 0.51 and 0.36 μm, respectively. The red curves indicate Gaussian fits.
4.2 Spectral resolution and calibration
The optical system was designed to distribute fluorescent light, ranging from 400 to 750 nm, with an almost uniform diffraction-limited spot size on the detector throughout the entire detection range of 1.2 pixels (FWHM). The spectral detection was calibrated using Ar ion and He-Ne laser lines (458, 488 and 633 nm) and quantum dots. Because of the nonlinear dispersion of the prism, the spectral sampling given in nm per channel was found to be nonuniform (Fig. 4A ), resulting in a corresponding nonuniform spectral resolution. Figure 4B and 4C show the acquired spectra of three laser lines using a commercial spectrometer (USB4000, Ocean Optics) and our spectral imaging system, respectively. The measured bandwidths of the three laser lines at 458, 488 and 633 nm were 1.35, 1.32 and 1.80 nm, respectively (Fig. 4B), i.e., they were narrow enough to evaluate the spectral resolution of our system. The wavelength verification of the three laser lines (Fig. 4B) was performed using the commercial spectrometer. We calibrated our spectral imaging system based on the wavelength values obtained (Fig. 4C). The FWHMs of the spectra were 1.8, 1.5 and 1.6 pixels corresponding to 6, 5.7 and 13 nm (Fig. 4D–4F), i.e., the spot sizes on the chip were a bit larger than expected but relatively constant throughout the spectral range. We found that the spectral resolution of the imaging system decreases with increasing wavelength as expected due to the nonlinear dispersion and the close-to diffraction-limited spot size. Fortunately, this effect is partially compensated by the observation that the bandwidth of emission spectra is broader for typical reddish compared to typical bluish fluorophores. The spectral calibration was additionally confirmed using quantum dots (Invitrogen).
Fig. 4 Calibration and characterization of the EM-CCD spectral imaging system. (A) Spectral sampling of our setup as calculated with Zemax (Zemax Development Corp.). The three wavelengths used for calibration are highlighted, anticipating a FWHM of 3.9, 4.8 and 9.6 nm when taking a diffraction-limited spot size of 1.2 pixels into consideration. The three laser lines (458, 488 and 633 nm) were measured using (B) a commercial spectrometer under the same conditions and (C) our setup. Based on (B), (C) was calibrated, and a wavelength range was assigned to each channel. A Gaussian function was fitted to the spectrum of each laser line to measure its bandwidth. The emission bandwidths of the three laser lines were sufficiently narrow (~1 nm) for a proper estimation of the spectral resolution of our system. The full widths at half maximum were 6, 5.7 and 13 nm at 458, 488 and 633 nm, respectively. In (D)-(F), the circles indicate the measured spectra from (A), and the red curves show Gaussian fits.
With the combination of sampling and resolution chosen for our setup, ~5-10 data points per emission bandwidth of a typical organic fluorophore or fluorescent protein and still ~3-6 data points for quantum dots are obtained, i.e., the spectra are sufficiently oversampled to allow for advanced spectral unmixing/deconvolution approaches [19].
4.3 Sensitivity
We could show previously [16] the photon-counting capabilities of our spectrometer setup, which features a mean quantum yield of ~30% in the visible range and an effective readout noise of 0.6 photoelectrons (RMS). Therefore, the recorded intensities values obey mainly Poisson statistics for single pixels as well as for rebinned spectral windows, i.e., the spectra can be compared immediately to a distribution one would obtain with linear dispersion. Thus, the signal-to-noise ratio does not show any spectral dependence except for the impact of the spectral dependence of the quantum yield.
5. Spectrally resolved imaging of cells and tissue
Since spectral imaging allows for the simultaneous recording of both spectral and spatial information, it enables to identify multiple fluorescence and autofluorescence signatures from labeled and unlabeled cells as well as tissues in a single acquisition step. To demonstrate the usability of this technology, we conducted experiments on various biological samples: first, we studied the spectral signature of fluorescent proteins covering the visible spectrum, i.e., the cyan fluorescent protein CFP, the yellow fluorescent protein YFP, and the red fluorescent protein DsRed (Fig. 5A –5C). Reduced single-channel images, acquired from HEK-293 cells transiently expressing one of the fluorescent molecules, were reconstructed by binning several channels around the peak of the spectrum. The main peaks for CFP, YFP and DsRed were found at 513, 531 and 588 nm, respectively, in good agreement with previously reported values (Fig. 5D) [20].
Fig. 5 Spectral images of HEK-293 cells expressing (A) CFP, (B) YFP and (C) DsRed were acquired at excitation wavelengths of (A) 800, (B) 900, and (C) 940 nm, and detected at (A) 486-560, (B) 508-588 and (C) 560-654 nm, respectively. The spectra in the regions of interest are shown in (D). The scale bar represents 20 μm.
As a second application, we acquired an autofluorescence spectral image stack from the rhizome of Convallaria Majalis tissue. Upon screening at various excitation wavelengths within the 700-950 nm range, we found a specific excitation wavelength at 758 nm that showed distinctive autofluorescence spectra in two subcellular compartments. As shown in the overlay image (Fig. 6A ) composed of distinct single spectral images (Fig. 6B–6E), the cell wall (Fig. 6B) could be easily distinguished from the plasma membrane (Fig. 6E). Here, the images (Fig. 6B–6E) were taken from single channels with the spectral width of 8-9 nm. We found that cell wall (a) and plasma membrane (b) have their own distinctive autofluorescence spectra as shown in Fig. 6F.
Fig. 6 Spectral autofluorescence images of Convallaria Majalis tissue. Multiple spectral signatures reveal subcellular compartments (B-F) as seen in the overlay image (A) taken from single channels at (B) 639-647, (C) 661-669, (D) 693-701 and (E) 719-728 nm. The spectra of cell wall (a) and plasma membrane (b) are shown in (F). The scale bar represents 20 μm.
In a third application, we visualized quantum dots (QDs), which are promising labels for 2PE spectral microscopy. QDs are co-excitable with a single excitation source and emit narrow emission spectra throughout the visible and near-infrared spectral ranges. Using multiple QD staining at various time points, we tracked the redistribution of GPCR along the endocytotic internalization process. To accomplish this, we incorporated an SBP-tag (a short sequence of 38 amino acids that recognizes the biotin-binding site of streptavidin, SA) at the N-terminus of the 5-HT2C serotonin receptor (Fig. 7A). SA-linked labels are ideal markers for receptors because they do not cross the cell membrane and can access more sterically restricted spaces than antibodies (8-53 kDa versus 150 kDa). We labeled selectively the surface-associated pools of 5-HT2C receptors at different times (24 h, 6 h and 1 h prior to imaging) by successively using three SA-conjugated, spectrally distinctive fluorescent QDs (Fig. 7A). Due to their broad absorption spectra, all three QDs (quantum dots 525, 585 and 655) were simultaneously excited at a wavelength of 800 nm. Figure 7B shows the overlay image of the SBP-5-HT2C expressing cell line labeled with three QDs (Fig. 7C–7E). Figure 7F–7H shows the spectra of the regions indicated in Fig. 7B and representing different subcellular locations. The 5-HT2C receptors that had been labeled at 24 and 6 h before imaging with QDs 525 and 585 were located mostly in the cellular interior (Fig. 7C and 7D). In contrast, the QD 655 nm spectral signature remained exclusively at the plasmalemma (Fig. 7E). This result proves the feasibility to study the life cycle of receptors using GPCR staining with multiple QDs and 2PE spectral imaging. Figure 7 corroborates the usability of our system for imaging samples labeled with multiple probes or showing complex autofluorescence.
6. Conclusion
We have developed a simple and useful method and setup that enables the acquisition of full spectral fluorescence images of biological samples with high spectral and temporal resolution and high sensitivity. We could show that coupled with a 2PE laser scanning microscope, our spectral detection system provides a high signal-to-noise ratio and good spectral distinction of the emitted light from the excitation source and of different fluorophores from each other over the full detection range of the visible spectrum. As an outlook, our setup should allow to obtain information about diffusion and interaction properties of biomolecules in living cells or tissue using FCS [16]. Moreover, advanced unmixing methods [19] will enable an even better decomposition of multiple fluorophores especially when their emission spectra overlap. We expect that our approach, coupled with robust and fast image analysis, will advance high content microscopy of complex samples beyond single cells.
Acknowledgments
This research was supported by the Basic Science Research Program of the National Research Foundation of Korea (NRF), funded by the Ministry of Education, Science and Technology (2010-0013312).
References and links
1. J. A. Conchello and J. W. Lichtman, “Optical sectioning microscopy,” Nat. Methods 2(12), 920–931 (2005). [CrossRef] [PubMed]
2. E. Wang, C. M. Babbey, and K. W. Dunn, “Performance comparison between the high-speed Yokogawa spinning disc confocal system and single-point scanning confocal systems,” J. Microsc. 218(2), 148–159 (2005). [CrossRef] [PubMed]
3. F. Helmchen and W. Denk, “Deep tissue two-photon microscopy,” Nat. Methods 2(12), 932–940 (2005). [CrossRef] [PubMed]
4. M. Wachsmuth, and K. Weisshart, “Fluorescence photobleaching and fluorescence correlation spectroscopy: two complementary technologies to study molecular dynamics in living cells,” in Imaging Cellular and Molecular Biological Functions (Springer Verlag, Heidelberg, 2007).
5. T. Misteli and R. D. Phair, “High mobility of proteins in the mammalian cell nucleus,” Nature 404(6778), 604–609 (2000). [CrossRef] [PubMed]
6. P. Indovina, M. Collini, G. Chirico, and M. T. Santini, “Three-dimensional cell organization leads to almost immediate HRE activity as demonstrated by molecular imaging of MG-63 spheroids using two-photon excitation microscopy,” FEBS Lett. 581(4), 719–726 (2007). [CrossRef] [PubMed]
7. A. Hayek, A. Grichine, T. Huault, C. Ricard, F. Bolze, B. Van Der Sanden, J. C. Vial, Y. Mély, A. Duperray, P. L. Baldeck, and J. F. Nicoud, “Cell-permeant cytoplasmic blue fluorophores optimized for in vivo two-photon microscopy with low-power excitation,” Microsc. Res. Tech. 70(10), 880–885 (2007). [CrossRef] [PubMed]
8. C. Li, R. K. Pastila, C. Pitsillides, J. M. Runnels, M. Puoris’haag, D. Côté, and C. P. Lin, “Imaging leukocyte trafficking in vivo with two-photon-excited endogenous tryptophan fluorescence,” Opt. Express 18(2), 988–999 (2010). [CrossRef] [PubMed]
9. R. Cicchi, A. Crisci, A. Cosci, G. Nesi, D. Kapsokalyvas, S. Giancane, M. Carini, and F. S. Pavone, “Time- and Spectral-resolved two-photon imaging of healthy bladder mucosa and carcinoma in situ,” Opt. Express 18(4), 3840–3849 (2010). [CrossRef] [PubMed]
10. Y. Hiraoka, T. Shimi, and T. Haraguchi, “Multispectral imaging fluorescence microscopy for living cells,” Cell Struct. Funct. 27(5), 367–374 (2002). [CrossRef] [PubMed]
11. R. H. Berg, “Evaluation of spectral imaging for plant cell analysis,” J. Microsc. 214(2), 174–181 (2004). [CrossRef] [PubMed]
12. P. D. Beule, D. M. Owen, H. B. Manning, C. B. Talbot, J. Requejo-Isidro, C. Dunsby, J. Mcginty, R. K. P. Benninger, D. S. Elson, I. Munro, M. John Lever, P. Anand, M. A. A. Neil, and P. M. W. French, “Rapid hyperspectral fluorescence lifetime imaging,” Microsc. Res. Tech. 70(5), 481–484 (2007). [CrossRef] [PubMed]
13. M. B. Sinclair, D. M. Haaland, J. A. Timlin, and H. D. T. Jones, “Hyperspectral confocal microscope,” Appl. Opt. 45(24), 6283–6291 (2006). [CrossRef] [PubMed]
14. S. Kumazaki, M. Hasegawa, M. Ghoneim, Y. Shimizu, K. Okamoto, M. Nishiyama, H. Oh-Oka, and M. Terazima, “A line-scanning semi-confocal multi-photon fluorescence microscope with a simultaneous broadband spectral acquisition and its application to the study of the thylakoid membrane of a cyanobacterium Anabaena PCC7120,” J. Microsc. 228(2), 240–254 (2007). [CrossRef] [PubMed]
15. J. Rietdorf, and E. H. K. Stelzer, “Special optical elements,” in Handbook of biological confocal microscopy, J. B. Pawley, 3ed. (Springer, 2006), pp. 43–58.
16. F. Bestvater, Z. Seghiri, M. S. Kang, N. Gröner, J. Y. Lee, K. B. Im, and M. Wachsmuth, “EMCCD-based spectrally resolved fluorescence correlation spectroscopy,” Opt. Express 18(23), 23818–23828 (2010). [CrossRef] [PubMed]
17. G. Heuvelman, F. Erdel, M. Wachsmuth, and K. Rippe, “Analysis of protein mobilities and interactions in living cells by multifocal fluorescence fluctuation microscopy,” Eur. Biophys. J. 38(6), 813–828 (2009). [CrossRef] [PubMed]
18. A. D. Keefe, D. S. Wilson, B. Seelig, and J. W. Szostak, “One-step purification of recombinant proteins using a nanomolar-affinity streptavidin-binding peptide, the SBP-Tag,” Protein Expr. Purif. 23(3), 440–446 (2001). [CrossRef] [PubMed]
19. S. Schlachter, S. Schwedler, A. Esposito, G. S. Kaminski Schierle, G. D. Moggridge, and C. F. Kaminski, “A method to unmix multiple fluorophores in microscopy images with minimal a priori information,” Opt. Express 17(25), 22747–22760 (2009). [CrossRef]
20. K. Svoboda and R. Yasuda, “Principles of two-photon excitation microscopy and its applications to neuroscience,” Neuron 50(6), 823–839 (2006). [CrossRef] [PubMed]
