- Email: [email protected]
Fluorescence Intensity Decay Shape Analysis Microscopy (FIDSAM) for Quantitative and Sensitive Live-Cell Imaging: A Novel Technique for Fluorescence Microscopy of Endogenously Expressed Fusion-Proteins
Molecular Plant
•
Volume 3
•
Number 3
•
Pages 555–562
•
May 2010
RESEARCH ARTICLE
Fluorescence Intensity Decay Shape Analysis Microscopy (FIDSAM) for Quantitative and Sensitive Live-Cell Imaging: A Novel Technique for Fluorescence Microscopy of Endogenously Expressed Fusion-Proteins Frank Schleifenbauma,1, Kirstin Elgassb, Marcus Sackrowb, Katharina Caesara, Kenneth Berendzena, Alfred J. Meixnerb and Klaus Hartera,2 a Center for Plant Molecular Biology, Department of Plant Physiology, University of Tu¨bingen, Auf der Morgenstelle 1, 72076 Tu¨bingen, Germany b Institute of Physical and Theoretical Chemistry, University of Tu¨bingen, Auf der Morgenstelle 8, 72076 Tu¨bingen, Germany
ABSTRACT Fluorescent studies of living plant cells such as confocal microscopy and fluorescence lifetime imaging often suffer from a strong autofluorescent background contribution that significantly reduces the dynamic image contrast and the quantitative access to sub-cellular processes at high spatial resolution. Here, we present a novel technique—fluorescence intensity decay shape analysis microscopy (FIDSAM)—to enhance the dynamic contrast of a fluorescence image of at least one order of magnitude. The method is based on the analysis of the shape of the fluorescence intensity decay (fluorescence lifetime curve) and benefits from the fact that the decay patterns of typical fluorescence label dyes strongly differ from emission decay curves of autofluorescent sample areas. Using FIDSAM, we investigated Arabidopsis thaliana hypocotyl cells in their tissue environment, which accumulate an eGFP fusion of the plasma membrane marker protein LTI6b (LTI6b–eGFP) to low level. Whereas in conventional confocal fluorescence images, the membranes of neighboring cells can hardly be optically resolved due to the strong autofluorescence of the cell wall, FIDSAM allows for imaging of single, isolated membranes at high spatial resolution. Thus, FIDSAM will enable the sub-cellular analysis of even low-expressed fluorophoretagged proteins in living plant cells. Furthermore, the combination of FIDSAM with fluorescence lifetime imaging provides the basis to study the local physico-chemical environment of fluorophore-tagged biomolecules in living plant cells. Key words:
Cell structure; cell walls; membrane proteins; high-resolution fluorescence microscopy.
INTRODUCTION To achieve increased spatial resolution beyond Abbe’s diffraction limit is one of the most essential aspirations in optical microscopy to study molecular processes in living cells quantitatively and at high spatial resolution. Several approaches aiming to overcome this diffraction limit have been established within the last years (Betzig et al., 2006; Heilemann et al., 2008; Hell, 2003, 2007; Sauer, 2005; Willig et al., 2006). However, quantitative, high-resolution live-cell fluorescence microscopy and imaging are regularly confronted with the problem of background autofluorescence, which is inherent in many biological samples and represents a major problem for microscopy of living plants. Background autofluorescence does not only corrupt the image quality, but also non-identified background artifacts might be biologically misinterpreted (Billinton and
Knight, 2001). The autofluorescence background very often spectrally overlaps with the emission of commonly used fluorescent dyes such as GFP and its derivates (Schnell et al., 1999). Hence, a discrimination using color filters does not lead to satisfying results. Alternatively, fluorescence lifetime imaging microscopy (FLIM), a technique using different fluorescence 1 To whom correspondence should be addressed. E-mail frank.schleifenba [email protected]. 2 To whom correspondence can also be addressed. E-mail klaus.harter@zmbp. uni-tuebingen.de.
ª The Author 2009. Published by the Molecular Plant Shanghai Editorial Office in association with Oxford University Press on behalf of CSPP and IPPE, SIBS, CAS. doi: 10.1093/mp/ssp110, Advance Access publication 28 December 2009 Received 23 September 2009; accepted 25 November 2009
556
|
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
excited state lifetimes to obtain an enhanced image contrast (Esposito and Wouters, 2004; Munster and Gadella, 2005), was introduced. This technique is also limited in its applicability as soon as the autofluorescence background gets too dominant and the lifetime constant of the background signal is similar to that of the fluorescent dye (Billinton and Knight, 2001). The established ways to overcome this problem in plant cells utilize the overexpression of AFP fusion proteins to enhance the intensity signal and the use of protoplasts lacking the strongly autofluorescent cell wall. However, overexpression might affect the native function of the fusion protein. Similarly, the use of protoplasts, which have been taken out of their native tissue environment, might generate biological artifacts. Furthermore, molecular processes, such as cell wallrelated signaling, cell–cell communication, tissue and organ development, cannot be addressed easily by using protoplasts. Another possibility lies in the development of new microscopic techniques to enhance the image contrast by repressing the background autofluorescence signal and thus making the biological sample accessible to quantitative, high-resolution optical studies. However, despite the triumphal course of confocal laser scanning microscopy in biological applications, only marginal progress has been reported in the important field of contrast improvement. Techniques to separate autofluorescence established so far utilize different spectral characteristics of the autofluorescence emission and the specific marker fluorophore (spectral unmixing) (Dickinson et al., 2001). Although these methods are straightforward and often lead to satisfying results, they require a precise knowledge about the spectral emission properties of the utilized fluorophore. This can be challenging as soon as these properties change due to environmental influences such as an alteration in the pH. Moreover, no information about the fluorescence lifetime of the fluorophore can be obtained by using spectral unmixing. Here, we present a novel technique (FIDSAM) to enhance the contrast of fluorescence images by over at least one order of magnitude without the obsolete knowledge of the spectral emission properties of the fluorescence label. Our technique utilizes the shape characteristics of fluorescence lifetime curves to determine background emission. FIDSAM can be applied to living plant cells in their tissue environment and results in confocal and fluorescence lifetime images that are no longer adulterated by autofluorescence artifacts. Therefore, our method will enable the quantitative and highly sensitive imaging and the functional analysis of even low-expressed proteins in vivo.
RESULTS AND DISCUSSION Physical and Optical Principle of FIDSAM In plant cells, especially at the cytoplasm–plasmamembrane–cell wall interface, we observe a nearly quantitative overlap of both spectral emission and fluorescence lifetime distribution of background autofluorescence and eGFP (Figure 1). Accord-
Figure 1. Spectral and Fluorescence Lifetime Properties of eGFP and Autofluorescence of Plant Cells. (A) Spectral overlap of the emission spectrum of eGFP (green) and autofluorescent background of living plant cells (red). (B) Overlap of the fluorescence lifetimes of eGFP and autofluorescence background. For both eGFP and the autofluorescence, lifetimes are centered around 2.5 ns.
ingly, the autofluorescence cannot be separated from the target signal by spectral bandpass-filtering or by time-gating techniques, which utilize confined decay-time windows characteristic for the fluorescence lifetime of the marker dye to discriminate two or more emitters contributing to the signal. This makes high-resolution imaging and quantitative cell biology of especially low-expressed fusion proteins a major challenge in plants. To circumvent these restrictions, we developed a novel technique to discriminate background from target signal using a common confocal FLIM set-up. This was achieved by fluorescence intensity decay shape analysis microscopy (FIDSAM). Using a time-correlated single photon-counting (TCSPC) protocol, distinct spots on the sample were excited with a short laser pulse and the fluorescence intensity was recorded temporally resolved. The measured intensity decay profile was then compared to a reference function by a least square fit routine. The technique is based on the fact that the shape of the intensity decay originating from a fluorescence marker differs significantly from a background signal. Assuming a fluorescence marker to be without any background interference, only one distinct type of fluorescing species contributes to the recorded intensity decay. Accordingly, all recorded fluorescence photons follow the same statistics for the electronic transition. Consequently, one single decay time constant is observed, represented by a purely monoexponential intensity decay (Figure 2A) (Lakowicz, 1999). In contrast, background emission is highly anisotropic, as a multitude of different emitting species with individual decay constants contributes to the measured decay curve. Thus, the resulting intensity trajectory significantly deviates from a monoexponential shape (Figure 2B) and must be described by a multiexponential decay function. Accordingly, the more unspecific emission contributes to the measured fluorescence decay, the less accurately the recorded signal can be described by a monoexponential decay function (Figure 2C). To apply FIDSAM, the local fluorescence decay curve is measured at every spot of the sampled area, using a common FLIM protocol. Subsequently, the measured data are compared to
Schleifenbaum et al.
Figure 2. The Principle of FIDSAM. (A) Fluorescence decay of a sample area accumulating high amounts of fluorophor-tagged protein (here, LTI6b–eGFP in hypocotyl cells of 5-day-old Arabidopsis seedlings) emission (circles). The measured data are well described by a monoexponential reference function (red line) as depicted by the residual plot (bottom). (B) Fluorescence decay of a sample area with strong background emission taken from the cell wall of wild-type Arabidopsis hypocotyl cells. The measured signal (circles) strongly deviates from a monoexponential reference function (red line) as demonstrated by the residual plot (bottom). (C) Schematic shape of the fluorescence decay for different ratios of marker to background contribution. For strong autofluorescence background, the decay (dotted line) significantly deviates from a monoexponential behavior (red line), resulting in high error values when a monoexponential reference function is fitted to the experimental data.
a reference function describing monoexponential intensity decay behavior (Equation 1) by fitting the free parameters of the amplitude A and the time constant s to the measured data: IðtÞ = A expð t=sÞ
ð1Þ
Given that fluorescence intensity decay curves with a major background contribution are expected to deviate strongly from the monoexponential fluorescence decay, we expect a poor fit quality of the reference function for autofluorescent sample areas. In contrast, if photons originating from the target fluorophore contribute dominantly to the recorded signal, the fluorescence decay can be well described by a monoexponential reference function (Kapusta et al., 2007). A convenient parameter to rate the fit quality is given by the error value v2, which describes the sum of the square deviations of the fitted values Nc(tk) from experimental values N(tk) for every data channel (Lakowicz, 1999): 2 =
n X ½Nðtk Þ Nc ðtk Þ2 Nðtk Þ k=1
ð2Þ
v2 approaches 1 for an optimal fit and increases for larger deviations. To obtain a FIDSAM image, the measured intensity sig-
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
|
557
nal is multiplied with the reciprocal v2-value or a proportional number, obtained from fitting the experimental data to a monoexponential function for every data point. Sample regions, comprising a large fraction of marker dye, remain widely unaffected, as they exhibit a fair fit quality and hence small v2 values. In contrast, sample areas with a high background contribution to the measured signal are suppressed, as the low fit quality results in high v2 values. Furthermore, a software package was established for data handling, which allows for loading and processing the recorded data (see Supplemental Figure 1 for a screenshot). Details with respect to the software are available upon request. The original data as derived from the FLIM set-up are loaded in binary format. A monoexponential reference function is used to fit the measured data subsequently to each data point via a simplex fitting routine. To obtain significant v2 values, it is crucial to normalize the measured data prior to the fitting routine, thus avoiding the intrinsic intensity dependence of the v2 numbers. To ensure that the fluorescence decay curves exhibit sufficient photon counting statistics for reasonable fitting results, pixels can be binned together. The resulting v2 values are stored in a separate 2D-error image, which is applied to the measured raw-data by multiplying its inverse values with the corresponding fluorescence intensities of the original image. Due to pixel-binning, the overall image resolution might appear to be reduced, which can be circumvented by a blurring algorithm. Here, the distinct pixels of the error image are convolved with a Gaussian of adjustable width, resulting in a smooth gradient of the error-values between adjacent pixels. The resolution achieved is then no longer limited by the pixel size of the error image, but represents the original optical resolution of the raw-image. FIDSAM can be applied to an image several times. Thus, the magnitude of the contrast enhancement can be adapted to every image individually. Moreover, this method is also applicable to fluorescent label dyes with an intensity decay pattern deviating from monoexponential behavior. In this case, the individual decay shape of the dye is taken as a reference function, and the locally recorded intensity decay is compared with this function. The method also benefits from the fact that except for the reference function, which can easily be defined by analyzing the decay shape of the pure marker dye, no further presumptions about the nature of the sample are required. Especially, the value of the fluorescence lifetime, which is the mean parameter in FLIM studies, is negligible in our method. Thus, local changes in the fluorescent lifetime do not affect the contrast-enhanced image. One restriction of the FIDSAM technique is the relatively long measurement times inherent in FLIM-based microscopy. This might hamper dynamic studies, which require high image frame rates. However, we consider the FIDSAM technique as a powerful tool for non-time critical studies. In addition to an improved image quality, the temporal evolution of fluorescence emission, such as the fluorescence lifetime, is recorded. This opens the field for highly sensitive studies of, for instance,
558
|
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
the physico-chemical environment of a fluorophore combined with an outstanding image contrast (see Application section for further details).
Application of FIDSAM in High-Resolution Confocal Microscopy To demonstrate the feasibility of FIDSAM, we investigated hypocotyl cells of 10-day-old Arabidopsis thaliana seedlings, expressing an eGFP fusion of the plasma membrane marker protein LTI6b at low level (LTI6b–eGFP) (Cutler et al., 2000) using a customized confocal laser microscope (Blum et al., 2006) with a common FLIM extension (see Methods section for details). The raster image of the LTI6b–eGFP-labeled plasmalemmatas of two neighboring cells, separated by the non-marked autofluorescent cell wall, showed that the two membranes can hardly be spatially resolved (Figure 3A and 3B). Accordingly, a quantitative spatial investigation of the LTI6b–eGFP location was not possible. To improve the image contrast by suppressing the autofluorescence contribution, monoexponential decay curves (Equation 1), which were reconvolved with the instrument response function, were fitted to the measured data and the obtained v2 values, representing a measure for the fit quality, were arranged in an error image (Figure 3F). Subsequently, the intensity image was multiplied with the inverse error image. In accordance with the theoretical model, this procedure significantly reduced the background interference. The areas with a strong background signal appeared as dominant contributions in the error image due to high v2 values, whereas the sample areas with a high contribution of the eGFP remained basically unaffected (Figure 3C–3E). After application of the process to the raw image for several rounds (Figure 3B), the dominant cell wall autofluorescence was effectively suppressed and the LTI6b–eGFP-labeled plasma membranes of the two adjacent cells were spatially resolved (Figure 3C–3E). The recorded intensity cross-sections (Figure 3B–3E, dashed white lines) were used to calculate the achieved contrast enhancement (Table 1) and yielded an enhancement factor for the dynamic contrast of up to E9=14.1 for a nine-fold correction process. The raw image showed inhomogeneities of the fluorescence signal at distinct areas of the wall, which appeared to resemble optically resolved plasmalemmata prior to FIDSAM correction. However, these ‘plasmalemmata’ appeared to be less separated (D 0.6 lm) than in the processed images (D 1.6 lm). The origin of this initially contradictory finding was clarified by mathematically analyzing the raw and the processed data, fitting the measured intensity cross-sections to well defined distribution curves. As a starting point, we described the raw image profile by two Lorentzian distributions (green lines in the insets of Figure 3B–3E) and kept parameters for the distinct peak widths and the center positions constant for further analysis. As a next step, we initially fitted the ninefold corrected data to a double Gaussian, describing the intensity profiles of the two plasmalemmata (red lines in the insets
Figure 3. Application of FIDSAM to Tissue-Embedded Plant Cells. (A–F) Series of FIDSAM images of two adjacent hypocotyl cells from 10-day-old Arabidopsis seedlings expressing low levels of plasmalemma-associated LTI6B–eGFP. The insets in the respective images refer to the intensity profiles taken along the white lines. (A) Overview image showing the analyzed hypocotyl tissue. Images (B) to (E) are zoomed as indicated by the white square. (B) Uncorrected fluorescence intensity raw-image. (C–E) Fluorescence images presented in (B) after a threefold (C), five-time (D), and nine-time (E) decay shape analysis correction using the error image displayed in (F) (10-fold Gaussian blurring). (F) Error image as obtained from decay shape analysis. Bright areas represent regions with strong deviation of the fluorescence decay curve from the mono-exponential reference function. The insets in (B) to (E) represent the intensity cross-section (white dots) along the white line. The experimental data were fitted with two Gaussian distributions (red curves) to describe the plasma-membranelocated LTI6b–eGFP emission (red curves) and two Lorentzian distributions to describe the cell wall autofluorescence (green curves). (G) Original confocal image recorded with a commercial state-ofthe-art confocal microscope. (H) Fluorescence image presented in (G) after application of a spectral unmixing routine.
Schleifenbaum et al.
Figure 3B–3E). Again, the values of the center positions and the distribution widths were kept constant. Then, we fitted the raw and the processed images to an analytical function, which contained a double Gaussian (plasmalemmata) and two Lorentzian distributions (cell wall autofluorescence), keeping the ‘peak width’ and ‘center position’ parameters constant for every single distribution. We found that all images were well described by the predefined fit functions (gray lines in the inset Figure 3B–3E) with the only free parameter being the intensity ratio between the Gaussian and the Lorentzian distributions. Hence, we conclude that the intensity profiles of the recorded images did not change in their composition, but only in their relative weighting. This analysis proves that our FIDSAM data processing does not generate artifacts, but exclusively discriminates signals with different fluorescence decay shapes. In the case of the Arabidopsis cells, this finding has distinct biological relevance, as it clearly demonstrates that the intensity profile observed in the raw unprocessed data is mainly due to inhomogeneities of the cell wall autofluorescence and does not necessarily reflect the emission of the plasmalemma-located LTI6b–eGFP. To judge the performance of the FIDSAM technique, we again imaged LTI6b–eGFP-expressing hypocotyl cells using a commercially available state-of-the-art confocal microscope. Consistent with the raw data recorded using the FLIM setup (Figure 3A), we observed strong cell wall autofluorescence, which significantly interfered with the plasma membranelocated LTI6b–eGFP (Figure 3G). To increase the image quality, we now applied a spectral unmixing routine to the raw data. Remarkably, using this method, the autofluorescence could not efficiently be suppressed, but remained dominant (Figure 3H). A detailed analysis of these data revealed that the overall image contrast between the cell wall and the cytoplasm was increased by a factor of four. Accordingly, spectral unmixing is an appropriate method to reduce a moderate contribution of autofluorescent background. However, as soon as the autofluorescence gets too dominant, a discrimination between the specific signal of the fluorophore and the background is no longer feasible by using spectral unmixing. Compared to the FIDSAM method, another major disadvantage lies in the fact that the spectral unmixing routine cannot be adjusted with respect to the degree of background suppression. In contrast, the FIDSAM method can be applied several times to a recorded image and, thus, the magnitude of autofluorescence suppression may be precisely selected. Table 1. Calculated Intensity Ratios m w of the LTI6b–eGFP-Labeled Plasma Membranes to the Wall of Arabidopsis Hypocotyl Cells and the Corresponding Contrast Enhancement Factors E for Different Decay Shape Analysis Cycles. Correction cycles
0
3
5
9
Intensity ratio plasma membrane to wall m w
0.61
1.39
2.25
8.6
Enhancement factor E
1
2.3
3.6
14.1
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
|
559
For further validation of our FIDSAM method, a series of control tests were performed. First, we applied FIDSAM analysis to hypocotyl cells of wild-type Arabidopsis seedlings, where exclusively autofluorescence could be observed (Figure 4). Already, after three correction cycles, the samples did not show any intensity contrast and the background autofluorescence signal disappeared (Figure 4). As a further control test, we imaged hypocotyl cells of 5-day-old Arabidopsis seedlings. At this developmental stage, LTI6b–eGFP accumulation is significantly higher than in 10-day-old seedlings. Here, we did not observe a contrast enhancement after FIDSAM analysis (Figure 5A–5D), indicating a low level of background contribution to the overall fluorescence signal. Furthermore, we applied the blurring algorithm to the error image, namely the error image is convolved with a Gaussian distribution of adjustable width (see FIDSAM principle section for details). We also did not observe significant changes in the corrected
Figure 4. Series of FIDSAM Images of Three Adjacent Hypocotyl Cells from 10-Day-Old, Wild-Type Arabidopsis Seedlings. (A) Overview image showing the analyzed hypocotyl tissue. (B) Zoomed-in raw image. Applying the decay shape analysis-based contrast enhancement methods one (C), two (D), and three times (E), the image contrast diminishes. (F) Error image.
560
|
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
Figure 5. Series of FIDSAM Images of Adjacent Hypocotyl Cells from 5-Day-Old Arabidopsis Seedlings Expressing Enhanced Level of Plasmalemma-Associated LTI6B–eGFP. Different FIDSAM cycles were applied (A–D) and the images were blurred by a smoothing algorithm convoluting the error image with Gaussian distributions of different widths (i–iii).
images in comparison to the raw data (Figure 5i–iii). These results substantiate our observation that the presented method exclusively reduces autofluorescence background, whereas emission from specific fluorophores remains largely unaffected.
Application of FIDSAM in Fluorescence Lifetime Imaging Microscopy (FLIM) The presented FIDSAM technique can also be used to significantly improve FLIM images of living plant cells. This is demonstrated on the same sample area shown in Figure 3 for two adjacent hypocotyl cells from 10-day-old Arabidopsis seedlings expressing low levels of LTI6b–eGFP. Comparable to the raw intensity image (Figure 3A and 3B), we predominantly observed the fluorescence lifetime of the autofluorescence from the cell wall in the raw fluorescence lifetime image (Figure 6A). No information about the fluorescence lifetime of the membrane-associated LTI6b–eGFP could be extracted. However, the application of the FIDSAM technique
Figure 6. Application of FIDSAM to Fluorescence Lifetime Imaging. (A) Raw FILM image of two adjacent hypocotyl cells from 10-dayold Arabidopsis seedlings expressing low levels of plasmalemmaassociated LTI6B–eGFP. (B) FIDSAM-FLIM image of (A) (nine correction cycles, blurring width 10).
to the dataset by multiplying the lifetime values with the FIDSAM-intensity image (nine correction cycles, blurring width: 10 pixels) resulted in a FIDSAM-FLIM image in which only
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
|
561
the fluorescence lifetime information of the LTI6b–eGFP remained visible (Figure 6B). Remarkably, the eGFP fluorescence lifetimes were homogeneous for the individual cell membranes but significantly differed between the membranes of the two neighboring cells by a factor of about 1.5. This indicates significant physico-chemical differences in the membrane or at the membrane–cytoplasm interface of the individual cells such as local pH or membrane potential, which is reflected in the different fluorescence lifetime of LTI6B–eGFP (Billinton and Knight, 2001; Nakabayashi et al., 2008; Pepperkok et al., 1999).
Summary and Conclusion In summary, we introduced a robust and reliable method, FIDSAM, to significantly increase the dynamic contrast of fluorescence images by effectively reducing the autofluorescence contribution. The presented method allows for contrast enhancements of at least one order of magnitude, as demonstrated for LTI6b–eGFP-expressing Arabidopsis hypocotyl cells. Hence, our method now enables quantitative and sensitive, high-resolution live-imaging of even low-expressed proteins in cells and tissues with high autofluorescence background. Moreover, the FIDSAM technique can be combined with FLIM information. This way, the fluorescence lifetime of solely the fluorophore-tagged marker signal is accessible, opening the field for high-resolution fluorescence lifetime imaging without background interference and for the elucidation of the physico-chemical environment of fluorophore-tagged biomolecules in living plant cells. We consider our FIDSAM technique as a powerful tool for future quantitative life-cell fluorescence studies. For example, time-resolved fluorescence resonance energy transfer (FRET) investigations can benefit from the method, when FIDSAM is applied to the acceptor emission in a confined spectral range. Therefore, quantitative FRET data with very high accuracy can be recorded. Moreover, we suggest that FIDSAM enables highly sensitive fluorescence correlation spectroscopic (FCS) studies in living, tissue-embedded plant cells. Here, the uncorrelated background emission, which intrinsically corrupts the FCS correlation curve, can be significantly suppressed.
METHODS All confocal measurements have been carried out using a custom-built confocal laser scanning microscope based on a Zeiss Axiovert 135 TV (Blum et al., 2006) with FLIM extension (Picoharp 300, Picoquant GmbH, Berlin, Germany) as sketched in Figure 7. The focusing of the excitation light and the detection of the fluorescence emission were accomplished through the objective of the microscope (Plan-Neofluar, 1003/1,30 oil, Zeiss) and a dichroic beamsplitter (Zeiss FT380). For sample scanning, a feedback-controlled sample stage (PI, E-710.3CD) with nanometer precision was utilized. A 473-nm pulsed laser diode was used as an excitation source (LDH-P-C-470, Picoquant GmbH, Berlin, Germany) with an excitation power
Figure 7. Scheme of a Confocal FLIM Microscope Capable of FIDSAM. The main requirements comprise a pulsed laser excitation source, a confocal microscope equipped with sample or beam scanning device, and a sensitive detector unit such as an avalanche photodiode (APD) or a photomultiplier tube (PMT). Measurement control and data analysis are accomplished by a standard personal computer system. The data collection is linked to a fluorescence lifetime imaging (FLIM) extension based on a time-correlated single photon-counting (TCSPC) system, which time-correlates detected photons to the laser pulses.
chosen to be well below insetting pile-up effects (Lakowicz, 1999). Back-scattered excitation light was blocked by a steep edge filter (LP02-473RU-25 Semrock) and the fluorescence light was focused onto the active area of a spectral integrating avalanche photo diode (APD) (SPCM 200, Perkin Elmer). Data acquisition was accomplished by a commercial available software package (SymPhoTime, Picoquant GmbH, Berlin, Germany). Note that the FIDSAM method is not restricted to this custom-built set-up but can be used with any standard confocal fluorescence microscope equipped with a FLIM-extension. For data analysis, a mono-exponential decay curve, which was re-convolved with the instrument response function for every pixel, was fitted to the normalized lifetime curves using a simplex routine. The obtained v2 values as a measure for the deviation of the measured data from the monoexponential reference function are stored spatially resolved in a separate error image. The home-built software package provides the opportunity to bin several image pixels, which results in a reliable photon counting statistics and can be obtained on license basis upon request. Control measurements using spectral unmixing algorithms have been carried out on a commercial Zeiss LSM 700 confocal microscope (LSM 700, Zeiss). Arabidopsis thaliana seedlings were grown and prepared for microscopic analysis as described in Elgass et al. (2009).
SUPPLEMENTARY DATA Supplementary Data are available at Molecular Plant Online.
562
|
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
FUNDING This work was funded by grants from the State of BadenWu¨rttemberg and the University of Tu¨bingen to F.S., K.E., and K.C., and a DFG grant to K.H. (HA2146/10-1).
Esposito, A., and Wouters, F.S. (2004). Fluorescence lifetime imaging microscopy. Current Protocols in Cell Biology. 4, 4.14. Heilemann, M. (2008). Subdiffraction-resolution fluorescence imaging with conventional fluorescent probes. Angew. Chem. Int. Ed. 47, 6172–6176.
ACKNOWLEDGMENTS
Hell, S.W. (2003). Toward fluorescence nanoscopy. Nature Biotechnology. 21, 1347–1355.
We thank Felicity de Courcy for proofreading the manuscript. No conflict of interest declared.
Hell, S.W. (2007). Far-field optical nanoscopy. Science. 316, 1153–1158.
REFERENCES
Kapusta, P., Wahl, M., Benda, A., Hof, M., and Enderlein, J. (2007). Fluorescence lifetime correlation spectroscopy. J. Fluoresc. 17, 43–48.
Betzig, E., et al. (2006). Imaging intracellular fluorescent proteins at nanometer resolution. Science. 313, 1642–1645.
Lakowicz, J.R. (1999). Principles of Fluorescence Spectroscopy. (New York: Kluwer Academic).
Billinton, N., and Knight, A.W. (2001). Seeing the wood through the trees: a review of techniques for distinguishing green fluorescent protein from endogenous autofluorescence. Anal. Biochem. 291, 175–197.
Munster, E.Bv., and Gadella, T.W. (2005). Fluorescence lifetime imaging microscopy (FLIM). Adv. Biochem. Eng. Biotechnol. 95, 143–175.
Blum, C., Meixner, A.J., and Subramaniam, V. (2006). Single oligomer spectra probe chromophore nanoenvironments of tetrameric fluorescent proteins. J. Am. Chem. Soc. 128, 8664–8670.
Nakabayashi, T., Wang, H.-P., Kinjo, M., and Ohta, N. (2008). Application of fluorescence lifetime imaging of enhanced green fluorescent protein to intracellular pH measurements. Photochem. Photobiol. Sci. 7, 668–670.
Cutler, S.R., Ehrhard, D.W., Griffitts, J.S., and Somerville, C.R. (2000). Random GFP:cDNA fusions enable visualization of subcellular structures in cells of Arabidopsis at high frequency. Proc. Nat. Am. Soc. 97, 3718–3723.
Pepperkok, R., Squire, A., Geley, S., and Bastiaens, P. (1999). Simultaneous detection of multiple green fluorescent proteins in live cells by fluorescence lifetime imaging microscopy. Current Biology. 9, 269–272.
Dickinson, M.E., Bearman, G., Tille, S., Lansford, R., and Fraser, S.E. (2001). Multi-spectral imaging and linear unmixing add a whole new dimension to laser scanning fluorescence microscopy. BioTechniques. 31, 1272–1278.
Sauer, M. (2005). Reversible molecular photoswitches: a key technology for nanoscience and fluorescence microscopy. Proc. Natl Acad. Sci. U S A. 102, 9433–9434.
Elgass, K., Caesar, K., Schleifenbaum, F., Stierhof, Y.D., Meixner, A.J., and Harter, K. (2009). Novel application of fluorescence lifetime and fluorescence microscopy enables quantitative access to subcellular dynamics in plant cells. PLoS One. 4, e5716.
Schnell, S.A., Staines, W.A., and Wessendor, M.W. (1999). Reduction of lipofuscin-like autofluorescence in fluorescently labeled tissue. J. Histochem. Cytochem. 47, 719–730. Willig, K., Rizzoli, S.O., Westphal, V., Jahn, R., and Hell, S.W. (2006). STED micorscopy reveals that synaptotagmin remains clustered after synaptic vesicle exocytosis. Nature. 440, 935–939.
Copyright of Molecular Plant is the property of Oxford University Press / UK and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
•
Volume 3
•
Number 3
•
Pages 555–562
•
May 2010
RESEARCH ARTICLE
Fluorescence Intensity Decay Shape Analysis Microscopy (FIDSAM) for Quantitative and Sensitive Live-Cell Imaging: A Novel Technique for Fluorescence Microscopy of Endogenously Expressed Fusion-Proteins Frank Schleifenbauma,1, Kirstin Elgassb, Marcus Sackrowb, Katharina Caesara, Kenneth Berendzena, Alfred J. Meixnerb and Klaus Hartera,2 a Center for Plant Molecular Biology, Department of Plant Physiology, University of Tu¨bingen, Auf der Morgenstelle 1, 72076 Tu¨bingen, Germany b Institute of Physical and Theoretical Chemistry, University of Tu¨bingen, Auf der Morgenstelle 8, 72076 Tu¨bingen, Germany
ABSTRACT Fluorescent studies of living plant cells such as confocal microscopy and fluorescence lifetime imaging often suffer from a strong autofluorescent background contribution that significantly reduces the dynamic image contrast and the quantitative access to sub-cellular processes at high spatial resolution. Here, we present a novel technique—fluorescence intensity decay shape analysis microscopy (FIDSAM)—to enhance the dynamic contrast of a fluorescence image of at least one order of magnitude. The method is based on the analysis of the shape of the fluorescence intensity decay (fluorescence lifetime curve) and benefits from the fact that the decay patterns of typical fluorescence label dyes strongly differ from emission decay curves of autofluorescent sample areas. Using FIDSAM, we investigated Arabidopsis thaliana hypocotyl cells in their tissue environment, which accumulate an eGFP fusion of the plasma membrane marker protein LTI6b (LTI6b–eGFP) to low level. Whereas in conventional confocal fluorescence images, the membranes of neighboring cells can hardly be optically resolved due to the strong autofluorescence of the cell wall, FIDSAM allows for imaging of single, isolated membranes at high spatial resolution. Thus, FIDSAM will enable the sub-cellular analysis of even low-expressed fluorophoretagged proteins in living plant cells. Furthermore, the combination of FIDSAM with fluorescence lifetime imaging provides the basis to study the local physico-chemical environment of fluorophore-tagged biomolecules in living plant cells. Key words:
Cell structure; cell walls; membrane proteins; high-resolution fluorescence microscopy.
INTRODUCTION To achieve increased spatial resolution beyond Abbe’s diffraction limit is one of the most essential aspirations in optical microscopy to study molecular processes in living cells quantitatively and at high spatial resolution. Several approaches aiming to overcome this diffraction limit have been established within the last years (Betzig et al., 2006; Heilemann et al., 2008; Hell, 2003, 2007; Sauer, 2005; Willig et al., 2006). However, quantitative, high-resolution live-cell fluorescence microscopy and imaging are regularly confronted with the problem of background autofluorescence, which is inherent in many biological samples and represents a major problem for microscopy of living plants. Background autofluorescence does not only corrupt the image quality, but also non-identified background artifacts might be biologically misinterpreted (Billinton and
Knight, 2001). The autofluorescence background very often spectrally overlaps with the emission of commonly used fluorescent dyes such as GFP and its derivates (Schnell et al., 1999). Hence, a discrimination using color filters does not lead to satisfying results. Alternatively, fluorescence lifetime imaging microscopy (FLIM), a technique using different fluorescence 1 To whom correspondence should be addressed. E-mail frank.schleifenba [email protected]. 2 To whom correspondence can also be addressed. E-mail klaus.harter@zmbp. uni-tuebingen.de.
ª The Author 2009. Published by the Molecular Plant Shanghai Editorial Office in association with Oxford University Press on behalf of CSPP and IPPE, SIBS, CAS. doi: 10.1093/mp/ssp110, Advance Access publication 28 December 2009 Received 23 September 2009; accepted 25 November 2009
556
|
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
excited state lifetimes to obtain an enhanced image contrast (Esposito and Wouters, 2004; Munster and Gadella, 2005), was introduced. This technique is also limited in its applicability as soon as the autofluorescence background gets too dominant and the lifetime constant of the background signal is similar to that of the fluorescent dye (Billinton and Knight, 2001). The established ways to overcome this problem in plant cells utilize the overexpression of AFP fusion proteins to enhance the intensity signal and the use of protoplasts lacking the strongly autofluorescent cell wall. However, overexpression might affect the native function of the fusion protein. Similarly, the use of protoplasts, which have been taken out of their native tissue environment, might generate biological artifacts. Furthermore, molecular processes, such as cell wallrelated signaling, cell–cell communication, tissue and organ development, cannot be addressed easily by using protoplasts. Another possibility lies in the development of new microscopic techniques to enhance the image contrast by repressing the background autofluorescence signal and thus making the biological sample accessible to quantitative, high-resolution optical studies. However, despite the triumphal course of confocal laser scanning microscopy in biological applications, only marginal progress has been reported in the important field of contrast improvement. Techniques to separate autofluorescence established so far utilize different spectral characteristics of the autofluorescence emission and the specific marker fluorophore (spectral unmixing) (Dickinson et al., 2001). Although these methods are straightforward and often lead to satisfying results, they require a precise knowledge about the spectral emission properties of the utilized fluorophore. This can be challenging as soon as these properties change due to environmental influences such as an alteration in the pH. Moreover, no information about the fluorescence lifetime of the fluorophore can be obtained by using spectral unmixing. Here, we present a novel technique (FIDSAM) to enhance the contrast of fluorescence images by over at least one order of magnitude without the obsolete knowledge of the spectral emission properties of the fluorescence label. Our technique utilizes the shape characteristics of fluorescence lifetime curves to determine background emission. FIDSAM can be applied to living plant cells in their tissue environment and results in confocal and fluorescence lifetime images that are no longer adulterated by autofluorescence artifacts. Therefore, our method will enable the quantitative and highly sensitive imaging and the functional analysis of even low-expressed proteins in vivo.
RESULTS AND DISCUSSION Physical and Optical Principle of FIDSAM In plant cells, especially at the cytoplasm–plasmamembrane–cell wall interface, we observe a nearly quantitative overlap of both spectral emission and fluorescence lifetime distribution of background autofluorescence and eGFP (Figure 1). Accord-
Figure 1. Spectral and Fluorescence Lifetime Properties of eGFP and Autofluorescence of Plant Cells. (A) Spectral overlap of the emission spectrum of eGFP (green) and autofluorescent background of living plant cells (red). (B) Overlap of the fluorescence lifetimes of eGFP and autofluorescence background. For both eGFP and the autofluorescence, lifetimes are centered around 2.5 ns.
ingly, the autofluorescence cannot be separated from the target signal by spectral bandpass-filtering or by time-gating techniques, which utilize confined decay-time windows characteristic for the fluorescence lifetime of the marker dye to discriminate two or more emitters contributing to the signal. This makes high-resolution imaging and quantitative cell biology of especially low-expressed fusion proteins a major challenge in plants. To circumvent these restrictions, we developed a novel technique to discriminate background from target signal using a common confocal FLIM set-up. This was achieved by fluorescence intensity decay shape analysis microscopy (FIDSAM). Using a time-correlated single photon-counting (TCSPC) protocol, distinct spots on the sample were excited with a short laser pulse and the fluorescence intensity was recorded temporally resolved. The measured intensity decay profile was then compared to a reference function by a least square fit routine. The technique is based on the fact that the shape of the intensity decay originating from a fluorescence marker differs significantly from a background signal. Assuming a fluorescence marker to be without any background interference, only one distinct type of fluorescing species contributes to the recorded intensity decay. Accordingly, all recorded fluorescence photons follow the same statistics for the electronic transition. Consequently, one single decay time constant is observed, represented by a purely monoexponential intensity decay (Figure 2A) (Lakowicz, 1999). In contrast, background emission is highly anisotropic, as a multitude of different emitting species with individual decay constants contributes to the measured decay curve. Thus, the resulting intensity trajectory significantly deviates from a monoexponential shape (Figure 2B) and must be described by a multiexponential decay function. Accordingly, the more unspecific emission contributes to the measured fluorescence decay, the less accurately the recorded signal can be described by a monoexponential decay function (Figure 2C). To apply FIDSAM, the local fluorescence decay curve is measured at every spot of the sampled area, using a common FLIM protocol. Subsequently, the measured data are compared to
Schleifenbaum et al.
Figure 2. The Principle of FIDSAM. (A) Fluorescence decay of a sample area accumulating high amounts of fluorophor-tagged protein (here, LTI6b–eGFP in hypocotyl cells of 5-day-old Arabidopsis seedlings) emission (circles). The measured data are well described by a monoexponential reference function (red line) as depicted by the residual plot (bottom). (B) Fluorescence decay of a sample area with strong background emission taken from the cell wall of wild-type Arabidopsis hypocotyl cells. The measured signal (circles) strongly deviates from a monoexponential reference function (red line) as demonstrated by the residual plot (bottom). (C) Schematic shape of the fluorescence decay for different ratios of marker to background contribution. For strong autofluorescence background, the decay (dotted line) significantly deviates from a monoexponential behavior (red line), resulting in high error values when a monoexponential reference function is fitted to the experimental data.
a reference function describing monoexponential intensity decay behavior (Equation 1) by fitting the free parameters of the amplitude A and the time constant s to the measured data: IðtÞ = A expð t=sÞ
ð1Þ
Given that fluorescence intensity decay curves with a major background contribution are expected to deviate strongly from the monoexponential fluorescence decay, we expect a poor fit quality of the reference function for autofluorescent sample areas. In contrast, if photons originating from the target fluorophore contribute dominantly to the recorded signal, the fluorescence decay can be well described by a monoexponential reference function (Kapusta et al., 2007). A convenient parameter to rate the fit quality is given by the error value v2, which describes the sum of the square deviations of the fitted values Nc(tk) from experimental values N(tk) for every data channel (Lakowicz, 1999): 2 =
n X ½Nðtk Þ Nc ðtk Þ2 Nðtk Þ k=1
ð2Þ
v2 approaches 1 for an optimal fit and increases for larger deviations. To obtain a FIDSAM image, the measured intensity sig-
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
|
557
nal is multiplied with the reciprocal v2-value or a proportional number, obtained from fitting the experimental data to a monoexponential function for every data point. Sample regions, comprising a large fraction of marker dye, remain widely unaffected, as they exhibit a fair fit quality and hence small v2 values. In contrast, sample areas with a high background contribution to the measured signal are suppressed, as the low fit quality results in high v2 values. Furthermore, a software package was established for data handling, which allows for loading and processing the recorded data (see Supplemental Figure 1 for a screenshot). Details with respect to the software are available upon request. The original data as derived from the FLIM set-up are loaded in binary format. A monoexponential reference function is used to fit the measured data subsequently to each data point via a simplex fitting routine. To obtain significant v2 values, it is crucial to normalize the measured data prior to the fitting routine, thus avoiding the intrinsic intensity dependence of the v2 numbers. To ensure that the fluorescence decay curves exhibit sufficient photon counting statistics for reasonable fitting results, pixels can be binned together. The resulting v2 values are stored in a separate 2D-error image, which is applied to the measured raw-data by multiplying its inverse values with the corresponding fluorescence intensities of the original image. Due to pixel-binning, the overall image resolution might appear to be reduced, which can be circumvented by a blurring algorithm. Here, the distinct pixels of the error image are convolved with a Gaussian of adjustable width, resulting in a smooth gradient of the error-values between adjacent pixels. The resolution achieved is then no longer limited by the pixel size of the error image, but represents the original optical resolution of the raw-image. FIDSAM can be applied to an image several times. Thus, the magnitude of the contrast enhancement can be adapted to every image individually. Moreover, this method is also applicable to fluorescent label dyes with an intensity decay pattern deviating from monoexponential behavior. In this case, the individual decay shape of the dye is taken as a reference function, and the locally recorded intensity decay is compared with this function. The method also benefits from the fact that except for the reference function, which can easily be defined by analyzing the decay shape of the pure marker dye, no further presumptions about the nature of the sample are required. Especially, the value of the fluorescence lifetime, which is the mean parameter in FLIM studies, is negligible in our method. Thus, local changes in the fluorescent lifetime do not affect the contrast-enhanced image. One restriction of the FIDSAM technique is the relatively long measurement times inherent in FLIM-based microscopy. This might hamper dynamic studies, which require high image frame rates. However, we consider the FIDSAM technique as a powerful tool for non-time critical studies. In addition to an improved image quality, the temporal evolution of fluorescence emission, such as the fluorescence lifetime, is recorded. This opens the field for highly sensitive studies of, for instance,
558
|
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
the physico-chemical environment of a fluorophore combined with an outstanding image contrast (see Application section for further details).
Application of FIDSAM in High-Resolution Confocal Microscopy To demonstrate the feasibility of FIDSAM, we investigated hypocotyl cells of 10-day-old Arabidopsis thaliana seedlings, expressing an eGFP fusion of the plasma membrane marker protein LTI6b at low level (LTI6b–eGFP) (Cutler et al., 2000) using a customized confocal laser microscope (Blum et al., 2006) with a common FLIM extension (see Methods section for details). The raster image of the LTI6b–eGFP-labeled plasmalemmatas of two neighboring cells, separated by the non-marked autofluorescent cell wall, showed that the two membranes can hardly be spatially resolved (Figure 3A and 3B). Accordingly, a quantitative spatial investigation of the LTI6b–eGFP location was not possible. To improve the image contrast by suppressing the autofluorescence contribution, monoexponential decay curves (Equation 1), which were reconvolved with the instrument response function, were fitted to the measured data and the obtained v2 values, representing a measure for the fit quality, were arranged in an error image (Figure 3F). Subsequently, the intensity image was multiplied with the inverse error image. In accordance with the theoretical model, this procedure significantly reduced the background interference. The areas with a strong background signal appeared as dominant contributions in the error image due to high v2 values, whereas the sample areas with a high contribution of the eGFP remained basically unaffected (Figure 3C–3E). After application of the process to the raw image for several rounds (Figure 3B), the dominant cell wall autofluorescence was effectively suppressed and the LTI6b–eGFP-labeled plasma membranes of the two adjacent cells were spatially resolved (Figure 3C–3E). The recorded intensity cross-sections (Figure 3B–3E, dashed white lines) were used to calculate the achieved contrast enhancement (Table 1) and yielded an enhancement factor for the dynamic contrast of up to E9=14.1 for a nine-fold correction process. The raw image showed inhomogeneities of the fluorescence signal at distinct areas of the wall, which appeared to resemble optically resolved plasmalemmata prior to FIDSAM correction. However, these ‘plasmalemmata’ appeared to be less separated (D 0.6 lm) than in the processed images (D 1.6 lm). The origin of this initially contradictory finding was clarified by mathematically analyzing the raw and the processed data, fitting the measured intensity cross-sections to well defined distribution curves. As a starting point, we described the raw image profile by two Lorentzian distributions (green lines in the insets of Figure 3B–3E) and kept parameters for the distinct peak widths and the center positions constant for further analysis. As a next step, we initially fitted the ninefold corrected data to a double Gaussian, describing the intensity profiles of the two plasmalemmata (red lines in the insets
Figure 3. Application of FIDSAM to Tissue-Embedded Plant Cells. (A–F) Series of FIDSAM images of two adjacent hypocotyl cells from 10-day-old Arabidopsis seedlings expressing low levels of plasmalemma-associated LTI6B–eGFP. The insets in the respective images refer to the intensity profiles taken along the white lines. (A) Overview image showing the analyzed hypocotyl tissue. Images (B) to (E) are zoomed as indicated by the white square. (B) Uncorrected fluorescence intensity raw-image. (C–E) Fluorescence images presented in (B) after a threefold (C), five-time (D), and nine-time (E) decay shape analysis correction using the error image displayed in (F) (10-fold Gaussian blurring). (F) Error image as obtained from decay shape analysis. Bright areas represent regions with strong deviation of the fluorescence decay curve from the mono-exponential reference function. The insets in (B) to (E) represent the intensity cross-section (white dots) along the white line. The experimental data were fitted with two Gaussian distributions (red curves) to describe the plasma-membranelocated LTI6b–eGFP emission (red curves) and two Lorentzian distributions to describe the cell wall autofluorescence (green curves). (G) Original confocal image recorded with a commercial state-ofthe-art confocal microscope. (H) Fluorescence image presented in (G) after application of a spectral unmixing routine.
Schleifenbaum et al.
Figure 3B–3E). Again, the values of the center positions and the distribution widths were kept constant. Then, we fitted the raw and the processed images to an analytical function, which contained a double Gaussian (plasmalemmata) and two Lorentzian distributions (cell wall autofluorescence), keeping the ‘peak width’ and ‘center position’ parameters constant for every single distribution. We found that all images were well described by the predefined fit functions (gray lines in the inset Figure 3B–3E) with the only free parameter being the intensity ratio between the Gaussian and the Lorentzian distributions. Hence, we conclude that the intensity profiles of the recorded images did not change in their composition, but only in their relative weighting. This analysis proves that our FIDSAM data processing does not generate artifacts, but exclusively discriminates signals with different fluorescence decay shapes. In the case of the Arabidopsis cells, this finding has distinct biological relevance, as it clearly demonstrates that the intensity profile observed in the raw unprocessed data is mainly due to inhomogeneities of the cell wall autofluorescence and does not necessarily reflect the emission of the plasmalemma-located LTI6b–eGFP. To judge the performance of the FIDSAM technique, we again imaged LTI6b–eGFP-expressing hypocotyl cells using a commercially available state-of-the-art confocal microscope. Consistent with the raw data recorded using the FLIM setup (Figure 3A), we observed strong cell wall autofluorescence, which significantly interfered with the plasma membranelocated LTI6b–eGFP (Figure 3G). To increase the image quality, we now applied a spectral unmixing routine to the raw data. Remarkably, using this method, the autofluorescence could not efficiently be suppressed, but remained dominant (Figure 3H). A detailed analysis of these data revealed that the overall image contrast between the cell wall and the cytoplasm was increased by a factor of four. Accordingly, spectral unmixing is an appropriate method to reduce a moderate contribution of autofluorescent background. However, as soon as the autofluorescence gets too dominant, a discrimination between the specific signal of the fluorophore and the background is no longer feasible by using spectral unmixing. Compared to the FIDSAM method, another major disadvantage lies in the fact that the spectral unmixing routine cannot be adjusted with respect to the degree of background suppression. In contrast, the FIDSAM method can be applied several times to a recorded image and, thus, the magnitude of autofluorescence suppression may be precisely selected. Table 1. Calculated Intensity Ratios m w of the LTI6b–eGFP-Labeled Plasma Membranes to the Wall of Arabidopsis Hypocotyl Cells and the Corresponding Contrast Enhancement Factors E for Different Decay Shape Analysis Cycles. Correction cycles
0
3
5
9
Intensity ratio plasma membrane to wall m w
0.61
1.39
2.25
8.6
Enhancement factor E
1
2.3
3.6
14.1
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
|
559
For further validation of our FIDSAM method, a series of control tests were performed. First, we applied FIDSAM analysis to hypocotyl cells of wild-type Arabidopsis seedlings, where exclusively autofluorescence could be observed (Figure 4). Already, after three correction cycles, the samples did not show any intensity contrast and the background autofluorescence signal disappeared (Figure 4). As a further control test, we imaged hypocotyl cells of 5-day-old Arabidopsis seedlings. At this developmental stage, LTI6b–eGFP accumulation is significantly higher than in 10-day-old seedlings. Here, we did not observe a contrast enhancement after FIDSAM analysis (Figure 5A–5D), indicating a low level of background contribution to the overall fluorescence signal. Furthermore, we applied the blurring algorithm to the error image, namely the error image is convolved with a Gaussian distribution of adjustable width (see FIDSAM principle section for details). We also did not observe significant changes in the corrected
Figure 4. Series of FIDSAM Images of Three Adjacent Hypocotyl Cells from 10-Day-Old, Wild-Type Arabidopsis Seedlings. (A) Overview image showing the analyzed hypocotyl tissue. (B) Zoomed-in raw image. Applying the decay shape analysis-based contrast enhancement methods one (C), two (D), and three times (E), the image contrast diminishes. (F) Error image.
560
|
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
Figure 5. Series of FIDSAM Images of Adjacent Hypocotyl Cells from 5-Day-Old Arabidopsis Seedlings Expressing Enhanced Level of Plasmalemma-Associated LTI6B–eGFP. Different FIDSAM cycles were applied (A–D) and the images were blurred by a smoothing algorithm convoluting the error image with Gaussian distributions of different widths (i–iii).
images in comparison to the raw data (Figure 5i–iii). These results substantiate our observation that the presented method exclusively reduces autofluorescence background, whereas emission from specific fluorophores remains largely unaffected.
Application of FIDSAM in Fluorescence Lifetime Imaging Microscopy (FLIM) The presented FIDSAM technique can also be used to significantly improve FLIM images of living plant cells. This is demonstrated on the same sample area shown in Figure 3 for two adjacent hypocotyl cells from 10-day-old Arabidopsis seedlings expressing low levels of LTI6b–eGFP. Comparable to the raw intensity image (Figure 3A and 3B), we predominantly observed the fluorescence lifetime of the autofluorescence from the cell wall in the raw fluorescence lifetime image (Figure 6A). No information about the fluorescence lifetime of the membrane-associated LTI6b–eGFP could be extracted. However, the application of the FIDSAM technique
Figure 6. Application of FIDSAM to Fluorescence Lifetime Imaging. (A) Raw FILM image of two adjacent hypocotyl cells from 10-dayold Arabidopsis seedlings expressing low levels of plasmalemmaassociated LTI6B–eGFP. (B) FIDSAM-FLIM image of (A) (nine correction cycles, blurring width 10).
to the dataset by multiplying the lifetime values with the FIDSAM-intensity image (nine correction cycles, blurring width: 10 pixels) resulted in a FIDSAM-FLIM image in which only
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
|
561
the fluorescence lifetime information of the LTI6b–eGFP remained visible (Figure 6B). Remarkably, the eGFP fluorescence lifetimes were homogeneous for the individual cell membranes but significantly differed between the membranes of the two neighboring cells by a factor of about 1.5. This indicates significant physico-chemical differences in the membrane or at the membrane–cytoplasm interface of the individual cells such as local pH or membrane potential, which is reflected in the different fluorescence lifetime of LTI6B–eGFP (Billinton and Knight, 2001; Nakabayashi et al., 2008; Pepperkok et al., 1999).
Summary and Conclusion In summary, we introduced a robust and reliable method, FIDSAM, to significantly increase the dynamic contrast of fluorescence images by effectively reducing the autofluorescence contribution. The presented method allows for contrast enhancements of at least one order of magnitude, as demonstrated for LTI6b–eGFP-expressing Arabidopsis hypocotyl cells. Hence, our method now enables quantitative and sensitive, high-resolution live-imaging of even low-expressed proteins in cells and tissues with high autofluorescence background. Moreover, the FIDSAM technique can be combined with FLIM information. This way, the fluorescence lifetime of solely the fluorophore-tagged marker signal is accessible, opening the field for high-resolution fluorescence lifetime imaging without background interference and for the elucidation of the physico-chemical environment of fluorophore-tagged biomolecules in living plant cells. We consider our FIDSAM technique as a powerful tool for future quantitative life-cell fluorescence studies. For example, time-resolved fluorescence resonance energy transfer (FRET) investigations can benefit from the method, when FIDSAM is applied to the acceptor emission in a confined spectral range. Therefore, quantitative FRET data with very high accuracy can be recorded. Moreover, we suggest that FIDSAM enables highly sensitive fluorescence correlation spectroscopic (FCS) studies in living, tissue-embedded plant cells. Here, the uncorrelated background emission, which intrinsically corrupts the FCS correlation curve, can be significantly suppressed.
METHODS All confocal measurements have been carried out using a custom-built confocal laser scanning microscope based on a Zeiss Axiovert 135 TV (Blum et al., 2006) with FLIM extension (Picoharp 300, Picoquant GmbH, Berlin, Germany) as sketched in Figure 7. The focusing of the excitation light and the detection of the fluorescence emission were accomplished through the objective of the microscope (Plan-Neofluar, 1003/1,30 oil, Zeiss) and a dichroic beamsplitter (Zeiss FT380). For sample scanning, a feedback-controlled sample stage (PI, E-710.3CD) with nanometer precision was utilized. A 473-nm pulsed laser diode was used as an excitation source (LDH-P-C-470, Picoquant GmbH, Berlin, Germany) with an excitation power
Figure 7. Scheme of a Confocal FLIM Microscope Capable of FIDSAM. The main requirements comprise a pulsed laser excitation source, a confocal microscope equipped with sample or beam scanning device, and a sensitive detector unit such as an avalanche photodiode (APD) or a photomultiplier tube (PMT). Measurement control and data analysis are accomplished by a standard personal computer system. The data collection is linked to a fluorescence lifetime imaging (FLIM) extension based on a time-correlated single photon-counting (TCSPC) system, which time-correlates detected photons to the laser pulses.
chosen to be well below insetting pile-up effects (Lakowicz, 1999). Back-scattered excitation light was blocked by a steep edge filter (LP02-473RU-25 Semrock) and the fluorescence light was focused onto the active area of a spectral integrating avalanche photo diode (APD) (SPCM 200, Perkin Elmer). Data acquisition was accomplished by a commercial available software package (SymPhoTime, Picoquant GmbH, Berlin, Germany). Note that the FIDSAM method is not restricted to this custom-built set-up but can be used with any standard confocal fluorescence microscope equipped with a FLIM-extension. For data analysis, a mono-exponential decay curve, which was re-convolved with the instrument response function for every pixel, was fitted to the normalized lifetime curves using a simplex routine. The obtained v2 values as a measure for the deviation of the measured data from the monoexponential reference function are stored spatially resolved in a separate error image. The home-built software package provides the opportunity to bin several image pixels, which results in a reliable photon counting statistics and can be obtained on license basis upon request. Control measurements using spectral unmixing algorithms have been carried out on a commercial Zeiss LSM 700 confocal microscope (LSM 700, Zeiss). Arabidopsis thaliana seedlings were grown and prepared for microscopic analysis as described in Elgass et al. (2009).
SUPPLEMENTARY DATA Supplementary Data are available at Molecular Plant Online.
562
|
Schleifenbaum et al.
d
FIDSAM for Quantitative and Sensitive Live-Cell Imaging
FUNDING This work was funded by grants from the State of BadenWu¨rttemberg and the University of Tu¨bingen to F.S., K.E., and K.C., and a DFG grant to K.H. (HA2146/10-1).
Esposito, A., and Wouters, F.S. (2004). Fluorescence lifetime imaging microscopy. Current Protocols in Cell Biology. 4, 4.14. Heilemann, M. (2008). Subdiffraction-resolution fluorescence imaging with conventional fluorescent probes. Angew. Chem. Int. Ed. 47, 6172–6176.
ACKNOWLEDGMENTS
Hell, S.W. (2003). Toward fluorescence nanoscopy. Nature Biotechnology. 21, 1347–1355.
We thank Felicity de Courcy for proofreading the manuscript. No conflict of interest declared.
Hell, S.W. (2007). Far-field optical nanoscopy. Science. 316, 1153–1158.
REFERENCES
Kapusta, P., Wahl, M., Benda, A., Hof, M., and Enderlein, J. (2007). Fluorescence lifetime correlation spectroscopy. J. Fluoresc. 17, 43–48.
Betzig, E., et al. (2006). Imaging intracellular fluorescent proteins at nanometer resolution. Science. 313, 1642–1645.
Lakowicz, J.R. (1999). Principles of Fluorescence Spectroscopy. (New York: Kluwer Academic).
Billinton, N., and Knight, A.W. (2001). Seeing the wood through the trees: a review of techniques for distinguishing green fluorescent protein from endogenous autofluorescence. Anal. Biochem. 291, 175–197.
Munster, E.Bv., and Gadella, T.W. (2005). Fluorescence lifetime imaging microscopy (FLIM). Adv. Biochem. Eng. Biotechnol. 95, 143–175.
Blum, C., Meixner, A.J., and Subramaniam, V. (2006). Single oligomer spectra probe chromophore nanoenvironments of tetrameric fluorescent proteins. J. Am. Chem. Soc. 128, 8664–8670.
Nakabayashi, T., Wang, H.-P., Kinjo, M., and Ohta, N. (2008). Application of fluorescence lifetime imaging of enhanced green fluorescent protein to intracellular pH measurements. Photochem. Photobiol. Sci. 7, 668–670.
Cutler, S.R., Ehrhard, D.W., Griffitts, J.S., and Somerville, C.R. (2000). Random GFP:cDNA fusions enable visualization of subcellular structures in cells of Arabidopsis at high frequency. Proc. Nat. Am. Soc. 97, 3718–3723.
Pepperkok, R., Squire, A., Geley, S., and Bastiaens, P. (1999). Simultaneous detection of multiple green fluorescent proteins in live cells by fluorescence lifetime imaging microscopy. Current Biology. 9, 269–272.
Dickinson, M.E., Bearman, G., Tille, S., Lansford, R., and Fraser, S.E. (2001). Multi-spectral imaging and linear unmixing add a whole new dimension to laser scanning fluorescence microscopy. BioTechniques. 31, 1272–1278.
Sauer, M. (2005). Reversible molecular photoswitches: a key technology for nanoscience and fluorescence microscopy. Proc. Natl Acad. Sci. U S A. 102, 9433–9434.
Elgass, K., Caesar, K., Schleifenbaum, F., Stierhof, Y.D., Meixner, A.J., and Harter, K. (2009). Novel application of fluorescence lifetime and fluorescence microscopy enables quantitative access to subcellular dynamics in plant cells. PLoS One. 4, e5716.
Schnell, S.A., Staines, W.A., and Wessendor, M.W. (1999). Reduction of lipofuscin-like autofluorescence in fluorescently labeled tissue. J. Histochem. Cytochem. 47, 719–730. Willig, K., Rizzoli, S.O., Westphal, V., Jahn, R., and Hell, S.W. (2006). STED micorscopy reveals that synaptotagmin remains clustered after synaptic vesicle exocytosis. Nature. 440, 935–939.
Copyright of Molecular Plant is the property of Oxford University Press / UK and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.