Open Access
Add to BibSonomyAbstract
We demonstrate simultaneous imaging of multiple fluorophores using wide-field epi-fluorescence microscopy with a monochrome camera. The intensities of the three lasers are modulated by a sinusoidal waveform in order to excite each fluorophore with the same modulation frequency and a different time-delay. Then, the modulated fluorescence emissions are simultaneously detected by a camera operating at four times the excitation frequency. We show that two different fluorescence beads having crosstalk can be clearly separated using digital processing based on the phase information. In addition, multiple organelles within multi-stained single cells are shown with the phase mapping method, demonstrating an improved dynamic range and contrast compared to the conventional fluorescence image. These findings suggest that wide-field epi-fluorescence microscopy with four-bucket detection could be utilized for high-contrast multicolor imaging applications such as drug delivery and fluorescence in situ hybridization.
© 2016 Optical Society of America
1. Introduction
Epi-fluorescence microscopy is a very simple method that can identify the locations of various organelles of single cells using staining of a selectively targeted organelle [1–5]. However, it is difficult to recognize the complete shape and appearance of cells from a partial image or to analyze single cells. For a better understanding of single cells, each organelle in the cells should be stained by individual fluorescent dyes having different emission spectra, and the stained organelles should be imaged with individual filter sets. Then, a multicolor image can be produced using digital image processing. Because the composite image can show the whole shape of single cells, multiple information sets can be determined, enabling the thorough examination of single cells.
Several kinds of epi-fluorescence microscopy have been developed for multicolor imaging using different filter sets (i.e., full-multiband filter set, multiband Pinkel filter set, multiband Sedat filter set). Each configuration has advantages and disadvantages. Imaging with a full-multiband filter set can achieve multicolor images with one imaging pass, but this configuration requires a color camera, which is relatively expensive and noisier than a monochrome camera. Configurations using the multiband Pinkel filter and Sedat filter sets can achieve higher quality images than the full-multiband filter set, but these methods are time-consuming because filters need to be exchanged with a motorized wheel in order to obtain a multicolor image.
Furthermore, these three filter set approaches in epi-fluorescence microscopy cannot avoid contrast decrease due to crosstalk, which occurs when emission spectra of individual fluorophores partially overlap each other [6]. Crosstalk is difficult to avoid because most fluorophores have a wideband emission spectrum [7], and thus, researchers working with fluorescence microscopes must carefully choose fluorescence filters and dyes. There are two methods for handling crosstalk. The degree of crosstalk can be decreased by the use of an emission filter with a narrow bandwidth, which reduces the signal-to-noise ratio (SNR) of fluorescence microscopy due to the decrease in the fluorescence emission arriving at the camera. On the other hand, the crosstalk problem can be overcome using linear spectral unmixing. Many studies have verified that the fluorescent signals from each fluorophore can be almost completely discriminated by this method, even when a substantial spectral overlap occurs in the emission spectrum [8–12].
Meanwhile, fast multicolor imaging of biological cells has become more important, especially after the development of fluorescence in situ hybridization (FISH) in cytogenetic research [13]. In particular, FISH is a method for observing deoxyribonucleic acid (DNA) sequences of chromosomes or gene mutation, confirming the presence or location of targeted DNA hybridized with multiple probes. For fast multicolor imaging of biological cells, imaging has been performed by the switching the laser and LED sources in time series [14], dual-view or quad view systems that multiplex different spectral channels on a single detector plane [15] and a host of snapshot spectral imaging [16]. Fast multicolor imaging with high contrast is also required in studies on drug delivery. In addition, because many drugs are inherently auto-fluorescent, it is important to enhance the contrast of the fluorescence image [17].
In this paper, we demonstrated epi-fluorescence microscopy for multicolor imaging by employing a digital lock-in technique known as four-bucket detection. This technique enables simultaneous multicolor imaging for fluorescent beads and multi-stained cells without significant contrast decrease due to a narrow bandpass emission filter or crosstalk.
2. Multiplexing in phase domain
Figure 1 schematically presents multiplexing of modulated fluorescence emissions. Several kinds of fluorophores are excited by excitation sources that are sinusoidally modulated at same frequency with time delays and are simultaneously detected. Excited fluorescence emissions having different time delays can be separated in the phase domain. Fluorescence emission detected on the camera can be expressed as
where is the intensity of the unmodulated signal, is the magnitude of the modulated fluorescence signal, is the modulation frequency, and is the delay of the modulated signal. Then, the four frames captured by the camera at a 4 frequency are expressed as
Fig. 1 Multiplexing scheme for multicolor imaging. Intensities of each laser are modulated with the same frequency but different phase. Fluorescence signals generated by excitation sources with different phase delays are multiplexed in the phase domain.
From , and , the magnitude of the modulated signal and the relative phase can be respectively extracted using following relations:
As a result, multi-fluorescence emissions with time delay can be detected by a single monochrome camera and reconstructed using the four-bucket technique. The magnitude of the modulated signal is the fluorescence intensity image including the fluorescence emissions from all fluorophores. Each fluorophore can be separated by applying the relative phase , which is defined by different time delays.3. Digital processing
All digital image processing was performed using functions in Labview software. The image processing procedure for the discrimination of multi-fluorophores is schematically represented in Fig. 2. First, the fluorescence intensity image is transformed into a binary image by applying a threshold value that extract fluorescence signals from the image. Then, the phase image is multiplied by the binary image obtained from the fluorescence intensity image. This process is performed to reject random phase values in the phase image. From the phase image in which random phase regions are removed, different phase images are separated according to its phase values by applying threshold values. Then, these masks are utilized for the extraction of fluorescence emissions having different time delays in the fluorescence intensity image. Finally, by multiplying each mask by the fluorescence intensity image, each fluorescence emission is discriminated from the fluorescence intensity image obtained using the four-bucket technique.
Fig. 2 Digital image processing procedure for multiplexing the individual fluorophores using the phase image.
4. Experiments
The experimental set-up is similar to a typical wide-field epi-fluorescence microscope. Three lasers were used as light sources as shown in Fig. 3. A 488-nm laser diode (OZ Optics, OZ-1000-488), a 532-nm laser diode (OZ Optics, OZ-1000-532), and 635-nm laser diode (OZ Optics, OZ-1000-635) were used for the excitation of multi-fluorophores. Laser beams from each laser were combined using a red-green-blue (RGB) combiner (OZ Optics, WDM) to simultaneously illuminate the sample. The light intensity for each laser was modulated using the sinusoidal waveform signals generated by a data acquisition (DAQ) board (National Instruments, PCIe-6353). The modulated frequency for the excitation laser was 6 Hz, and each modulation signal had different time delays. These signals were synchronized with the trigger signal that operates a camera (Hamamatsu, ORCA-Flash4.0, 2048 × 2048 pixels, 16 bits) under the control of Labview software. The light source from the RGB combiner passes through two lenses and a dichroic mirror (Seomrock, Di01-R405/488/532/635), which reflects the illumination light and transmits fluorescence emissions from the sample. Then, the light source illuminates the sample via a microscope objective. The excited fluorescence emissions, which are also sinusoidally modulated, sequentially pass through the microscope objective, the dichroic mirror, and a notch filter (Semrock, NF03-405/488/532/635). The notch filter was positioned in front of the camera and used for rejecting the residual illumination light passing through the dichroic mirror.
Fig. 3 A schematic of the experimental set-up. All excitation lasers are sinusoidally modulated with the same frequency and different time delays. The imaging camera is phase-locked to the laser-intensity modulation. B: blue laser; G: green laser; R: red laser; DAQ: data acquisition board; MO: microscope objective; Multi-DM: Multi-dichroic mirror; Multi-NF: Multi-notch filter; L1, L2, and L3: lenses
4.1 Fluorescent bead imaging
Two kinds of fluorescent polystyrene beads were prepared to validate the discrimination capability of the proposed method. In this experiment, two lasers (488-nm and 532-nm laser diodes) were used to illuminate the sample. The two prepared different beads were mixed and dispersed on a slide glass for simultaneous imaging. The fluorescence excitation and emission spectra of the fluorescent beads are shown in Fig. 4. Since most of the fluorophores have a wide spectral bandwidth, we chose fluorescent beads whose fluorescence excitation/emission wavelength partially overlapped. Figure 5 shows fluorescence images obtained using our proposed method. Figure 5(a) and 5(b) were the four-bucket fluorescence images obtained under illumination from the 488-nm laser and 532-nm laser, respectively. As estimated in Fig. 4, both the yellow-green fluorescent beads and orange fluorescent beads were shown in the image when we only illuminate the sample with 488-nm laser light, and only the orange fluorescent beads were shown in sample illuminated with a 532-nm laser. This difference occurs because of bleed-through, which limits the use of fluorescent dye.
Fig. 4 Fluorescence excitation and emission spectra of two different polystyrene beads. Bars indicate the spectra of the two excitation sources used in the experiment.
Fig. 5 Fluorescence images of two different fluorescent polystyrene beads (yellow-green and orange) mixed on a slide glass. (a) Four-bucket image taken with 488-nm laser illumination. (b) Four-bucket image taken with 532-nm laser illumination. (c) Four-bucket image taken with simultaneous illumination of both the 488-nm laser and 532-nm laser. (d) Orange and (e) yellow-green polystyrene bead images separated from (c) by applying the phase image. (f) Merged image of (d) and (e). The white bar indicates a scale of 100 µm.
For simultaneous multicolor imaging, light from the two lasers simultaneously illuminated the sample, while the intensity the two lasers was sinusoidally modulated at the frequency of 6 Hz with different time delays with an average incident power of 96 µW. Figure 5(c) shows the results. Two fluorescent beads were not discriminated in the image in Fig. 5(c) because the two types of fluorescence emissions were detected together by the monochrome camera. However, we can separate the two kinds of fluorescent beads using the phase image obtained together with the fluorescence intensity image as explained in the digital image processing section. Orange and yellow-green fluorescent beads were separated as shown in Fig. 5(d) and Fig. 5(e), respectively. Then, these two images are merged into one image after creating pseudo-colored images as shown in Fig. 5(f).
4.2 Cell imaging
After the bead experiments, we tested on single cells. The used cells were Hela cells stained with Alexa Fluo 488 Phalloidin (Thermo Fisher Scientific), MitoTracker Red CMXRos (Thermo Fisher Scientific), and To-Pro-3 Iodide (Thermo Fisher Scientific) for actin, mitochondria, and the nucleus of the cells, respectively. Figures 6(a)–6(c) are fluorescence images of the nucleus, mitochondria, and actin in the triple-labeled cells taken using single-band filters in sequence, respectively. The merged image of Figs. 6(a)–6(c) is shown in Fig. 6(d). Figures 6(e)–6(g) are fluorescence images acquired using a multi-notch filter with four-bucket detection and Fig. 6(h) is the merged image of Fig. 6(e)–6(g). Figure 6(i) is the normalized line intensity profiles indicated at the white dotted line in Fig. 6(a) and 6(e). Figure 6(j) shows the phase image obtained with the four-bucket method. At first, we followed the image-processing algorithm explained above. However, there was a problem in which the organelles that were located in the same region could not be separated. In other words, two fluorescence emissions arriving at the same pixel could not discriminated because we cannot divide one signal from the other fluorescence signal using phase information. Thus, we compared the phase image with the conventional image. Though the two images looked similar, they had several differences. Even though one fluorescence emission had a low emission level compared to the other emissions, it could clearly be observed in the phase image. The innermost nucleus was clearly observed, and the regions containing mitochondria and actin were easily discriminated regardless if the mitochondria and actin were overlapped.
Fig. 6 Each organelle of a triple-labeled sample was captured using three separate single-band filter sets in sequence (Brightline 680/42 nm, 600/37 nm, and 514/30 nm from Semrock Inc.). The nucleus, mitochondria, and actin of the Hela cells were stained with TO-PRO-3 Iodide, MitoTracker Red CMXRos, and Alexa Fluo 488 Phalloidin, respectively. (a) Fluorescence image taken using a single bandpass filter with 635-nm laser illumination. (b) Fluorescence image taken with 532-nm laser illumination. (c) Fluorescence imag with 488-nm laser illumination. (d) Merged image of (a), (b) and (c). (a-c) are acquired using single bandpass filters and (e-g) are acquired using a multi-notch filter with four-bucket detection. (h) Merged image of (e), (f) and (g). (i) Normalized line intensity profiles indicated at the white dashed line in (a) and (e). (j) Phase image obtained under simultaneous irradiation of the 488-, 532-, 635-nm lasers. The white bar indicateds a scale of 40 µm.
In order to calculate the SNR, we took 10 × 10 pixels in a region of interest (ROI) and a background region of Fig. 6(d) and 6(h). As shown in Fig. 7, the SNR of the four-bucket image was about 10 times higher than that of the conventional image. After normalizing each image, we also calculated the contrast ratio of ROI indicated as a white square in Fig. 6(d), 6(h) and 6(j). The contrast ratio of the conventional image and four-bucket image nearly made no difference. Because the fluorescent dyes used in experiments have spectral crosstalk, resulting in the row contrast ratio between two regions. But, the phase image showed a good contrast ratio of 2.3 compared to those two images. This occurs because the phase values between the two ROIs are changed into opposite directions to show a contrast between two different fluorescence emissions even if there is crosstalk.
Fig. 7 Calculated signal-to-noise ratio and contrast ratio of the conventional, four-bucket and phase images.
5. Discussion
This work demonstrated simultaneous multicolor epi-fluorescence microscopy based on the four-bucket technique and proposed a novel concept in which the phase information obtained from the four-bucket calculation could be used to discriminate two fluorophores that have partial overlap of the emission spectra. We performed two experiments. 1) In the experiment with fluorescent beads, two kinds of fluorescent beads, whose emission spectra partially overlapped, were mixed on a slide glass and imaged using the proposed method. As a result, two kinds of fluorescent beads were successfully separated with high brightness by means of multiplying a binary image obtained from the phase image to the intensity image. This method provides differentiated performance compared to that of the intensity thresholding method where the thresholding values are defined fully based on the decision of the operator [18]. On the other hand, the thresholding value for the binary image (our case) is determined based on the time delay of the fluorescence emissions. Thus, no relevant pixels are eliminated. 2) The second experiment was aimed at multiply stained cells. The difference between the fluorescent bead sample and cells is whether the two fluorescent probes overlap in the z-dimension. In cells, because the actin covers the nucleus, two fluorescence emissions exist in the region of the nucleus. This problem has a bigger effect on wide-field fluorescence microscopy because the depth resolving power of wide-field epi-fluorescence microscopy is lower than that of confocal fluorescence microscopy. Therefore, if the fluorescence emissions from regions where the nucleus and actin are spatially overlapped are separated through digital processing similar to that used in the fluorescent bead experiments, this resulted in “stamped out” nuclei in the actin channel. Thus, scientific images cannot be provided by this method. Nevertheless, we found two improvements in the phase images. First, the contrast of the nucleus, actin, and mitochondria were improved in the phase images compared to the conventional fluorescence image (Fig. 7). In the merged image formed by conventional fluorescence microscopy, the nuclei and actin were not clearly observed because of crosstalk and the low fluorescence signal level, respectively. Second, the dynamic range of the image was improved with four-bucket detection and mapping phases. In Fig. 6(i), the image with four-bucket showed lower background level than the conventional image, extending the dynamic range. Besides, the phase mapping of the four-bucket image showed a good contrast in weak signal regions. In the conventional merged image, the region with weak fluorescence emission in the field of view had low contrast. In contrast, the same region was clearly observed by means of the phase contrast (Fig. 6(j)). This improvement can be useful when the fluorescence signal of interest has a relatively low signal level compared to the surrounding fluorescence signal or when autofluorescence occurs.
The proposed method does not use emission filters for discriminating each fluorophore, and thus, the fluorescence sensitivity is intrinsically higher than that when the narrow bandpass filter is used to reduce crosstalk. In our previous research, we demonstrated that applying the four-bucket technique to fluorescence imaging improves the signal-to-background ratio of the fluorescence image [19]. The current transmission rate of the individual filters has been improving. However, exchanging filters with a motorized wheel results in a low imaging speed, and employing acousto-optic tunable filters (AOTF) or liquid crystal tunable filters (LCTF), which have fast wavelength switching times, is the main cause of the increase in the imaging implement cost [9].
There have been several multicolor wide-field fluorescence imaging approaches used thus far, such as a single color camera and fluorophores that are spectrally well separated [2], imaging with switching the laser and LED sources in time series [14], dual-view or quad-view systems that multiplex different spectral channels on a single detector plane [15], and a host of snapshot spectral imaging [16]. The single color camera approach has the advantage of direct multicolor detection where sequential imaging is unnecessary, while it has the disadvantage of relatively high background noise compared to a monochrome camera. Imaging with laser or LED sources switching provides a fast imaging speed because it is not necessary to change the individual excitation filters. However, this method is vulnerable to spectral crosstalk. Dual-view or quad-view Optical Insights systems can simultaneously provide fluorescence images of multiple fluorophores without any delay between two or four images, respectively. However, this approach also requires the selection of appropriate filter sets for the efficient selectivity of multiple fluorophores, which hampers its use to wide applications, and the crosstalk problem still remains. A host of snapshot spectral imaging shows simple and fast imaging capabilities, but it potentially has difficult hardware configuration and does not resolve the crosstalk problem. On the other hand, four-bucket-based wide-field fluorescence imaging provides robustness to crosstalk and requires no emission filter to separate each fluorophore, allowing high SNR in the system.
Unlike other lock-in techniques (e.g., fast Fourier transform (FFT)), because only four images are required for reconstructing the intensity and phase images, the four-bucket technique provides not only a small data load to the computer but also short image acquisition times for achieving the final separated image, which is less than 1 s in our experiment. Of course, the capability to selectively detect signals of interest might decrease compared with the use of FFT because reconstruction of the signal is conducted using few images [20]. However, based on the results of our experiments on the discrimination of multi-fluorescence emissions, the four-bucket method had sufficient capability to discriminate each multi-fluorescence emission. Furthermore, imaging speed of our system can be significantly improved by binning of camera pixels and use of the fast (kHz) cameras. With the proposed method, we believe that dynamic reactions between cells, which occur within a few seconds, could be detected because of its short imaging time and simultaneous imaging of multiple fluorophores. Therefore, the four-bucket-based multi-color imaging could be a good option for studying single cell mechanisms.
7. Summary
We have utilized the four-bucket technique to introduce simultaneous multicolor wide-field epi-fluorescence microscopy. With this approach, multiple fluorescent beads were simultaneously imaged and separated by digital processing. Because a narrow bandwidth emission filter was not used to reduce the crosstalk in this technique, simultaneous multicolor imaging could be achieved without brightness loss by multiplexing the multi-fluorescence emissions. Moreover, the phase mapping approach showed that a relatively weak fluorescence signal compared to the signal of than surrounding area can be clearly observed in the phase image. Therefore, we believe that this method can contribute to the study of drug delivery to cells and FISH applications because it has the potential to observe dynamic reactions in cells with high contrast within a few seconds.
Acknowledgments
This work was supported by the Korea Basic Science Institute grant (D36500).
References and links
1. D. B. Murphy, Fundamentals of Light Microscopy and Electronic Imaging (Wiley-Liss, 2001).
2. D. J. Webb and C. M. Brown, “Epi-Fluorescence Microscopy,” in Cell Imaging Techniques, D. J. Taatjes and J. Roth, eds. (Humana Press, 2012), Vol. 931, pp. 29–59.
3. P. J. Shaw, “Comparison of widefield/deconvolution and confocal microscopy for three-dimensional imaging,” in Handbook of Biological Confocal Microscopy (Springer, 2006), pp. 453–467.
4. D. Tseng and A. Ozcan with Hongying Zhu and O. Yaglidere, Ting-Wei Su, “Wide-field fluorescent microscopy on a cell-phone,” in (IEEE, 2011), pp. 6801–6804.
5. J. C. Waters, “Accuracy and precision in quantitative fluorescence microscopy,” J. Cell Biol. 185(7), 1135–1148 (2009). [CrossRef] [PubMed]
6. J. W. Lichtman and J.-A. Conchello, “Fluorescence microscopy,” Nat. Methods 2(12), 910–919 (2005). [CrossRef] [PubMed]
7. C. Yang, V. Hou, L. Y. Nelson, and E. J. Seibel, “Mitigating fluorescence spectral overlap in wide-field endoscopic imaging,” J. Biomed. Opt. 18(8), 086012 (2013). [CrossRef] [PubMed]
8. T. Zimmermann, J. Rietdorf, and R. Pepperkok, “Spectral imaging and its applications in live cell microscopy,” FEBS Lett. 546(1), 87–92 (2003). [CrossRef] [PubMed]
9. B. Kraus, M. Ziegler, and H. Wolff, “Linear fluorescence unmixing in cell biological research,” Mod. Res. Educ. Top. Microsc. 2, 863–873 (2007).
10. T. Zimmermann, “Spectral Imaging and Linear Unmixing in Light Microscopy,” in Microscopy Techniques, J. Rietdorf, ed. (Springer Berlin Heidelberg, 2005), Vol. 95, pp. 245–265.
11. D. Chorvat Jr, J. Kirchnerova, M. Cagalinec, J. Smolka, A. Mateasik, and A. Chorvatova, “Spectral unmixing of flavin autofluorescence components in cardiac myocytes,” Biophys. J. 89(6), L55–L57 (2005). [CrossRef] [PubMed]
12. Y. Garini, I. T. Young, and G. McNamara, “Spectral imaging: Principles and applications,” Cytometry A 69(8), 735–747 (2006). [CrossRef] [PubMed]
13. R. Bishop, “Applications of fluorescence in situ hybridization (FISH) in detecting genetic aberrations of medical significance,” Biosci. Horiz. 3(1), 85–95 (2010). [CrossRef]
14. K. Aswani, T. Jinadasa, and C. M. Brown, “Fluorescence Microscopy Light Sources,” Micros. Today 20(04), 22–28 (2012). [CrossRef]
15. L. Liu, O. Yermolaieva, W. A. Johnson, F. M. Abboud, and M. J. Welsh, “Identification and function of thermosensory neurons in Drosophila larvae,” Nat. Neurosci. 6(3), 267–273 (2003). [CrossRef] [PubMed]
16. N. Hagen and M. W. Kudenov, “Review of snapshot spectral imaging technologies,” Opt. Eng. 52(9), 090901 (2013). [CrossRef]
17. N. S. White and R. J. Errington, “Fluorescence techniques for drug delivery research: theory and practice,” Adv. Drug Deliv. Rev. 57(1), 17–42 (2005). [CrossRef] [PubMed]
18. K. W. Dunn, M. M. Kamocka, and J. H. McDonald, “A practical guide to evaluating colocalization in biological microscopy,” Am. J. Physiol. Cell Physiol. 300(4), C723–C742 (2011). [CrossRef] [PubMed]
19. K. S. Park, W. J. Choi, J. B. Eom, K. S. Chang, and B. H. Lee, “High-contrast epi-fluorescence wide-field imaging of biological cells using integrating-bucket method,” Opt. Commun. 355, 427–432 (2015). [CrossRef]
20. W. J. Choi, S. Y. Ryu, J. K. Kim, D. U. Kim, G. H. Kim, and K. S. Chang, “High-speed thermoreflectance microscopy using charge-coupled device-based Fourier-domain filtering,” Opt. Lett. 38(18), 3581–3584 (2013). [CrossRef] [PubMed]
