fluctuations are most commonly considered and certainly most easily described according to their analog waveform characteristics, i.e., amplitude and frequency. This also reflects the fact that signaling functions are widely believed to depend on distinct waveform encoded signal patterns controlling, for example, up- and down-regulation of selected gene expression.1, 2 Such hypotheses have proved technically challenging to test, in as much as functional (nonpathological) signal fluctuations can display durations between milliseconds to minutes, and resonate in oscillatory states where frequency can vary widely.3, 4 Furthermore, distinct ordered patterns of functional signaling activity can occur at unpredictable intervals (minutes, to hours, to days) during otherwise long periods of quiescence.5 For all these reasons, the major criterion for a method to monitor signals is that it should allow one to follow a broad range of dynamic phenotypes continuously, noninvasively, and during unlimited periods of time. Currently, the most widely used methods for monitoring depend on fluorescence imaging methods that do not fulfill these requirements.
The discovery of photoproteins in organisms like Aequorea sp. and Photinus pyralis revolutionized the field of optical imaging. In the Aequorea jellyfish, two photoproteins were isolated and later cloned, including green fluorescent protein (GFP) and the sensitive bioluminescent protein, aequorin. 6, 7, 8, 9 When isolated, binding to aequorin causes an intramolecular oxidation of its bound chromophore substrate, coelenterazine, resulting in the emission of blue light ( ). In contrast, the isolated GFP emits green light ( ) on excitation , making it a useful gene expression reporter. In the jellyfish, the two proteins are located in close enough proximity so that the energy transition from the -bound aequorin oxidation of coelenterazine is transferred nonradiatively to GFP, resulting in the emission of green light, a process known as bioluminescence resonance energy transfer (BRET).10
As an optical reporter probe with a low binding affinity, aequorin provided one of the first opportunities to address questions concerning signal transduction by high domains.11 Since apoaequorin was cloned, it has also been used extensively to selectively measure in subcellular compartments like the mitochondrial matrix and the endoplasmic reticulum.12, 13 However, detection of the very weak luminescence of aequorin required high-sensitivity photon detection devices, limiting the possible applications for living intact cell work, and the method was largely abandoned (where it concerned single-cell measurements) in favor of sensitive fluorescent dyes,14 and more recently, genetically encoded calcium sensitive fluorescent proteins. 15, 16, 17, 18 Here we demonstrate how very recent cutting-edge advances in photon detector cameras, optical microscopy, and molecular probe technologies promise to greatly simplify and improve the use of semi quantitative bioluminescence imaging, and are promising a renaissance for the use of aequorin to measure intracellular calcium concentrations .
Materials and Methods
Cell Culture and Transfection
Neuro2A and HEK-293 cells were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% Fetal Calf Serum (FCS) (Invitrogen, Life Technologies). Cultures were incubated at 37°C in a humidified atmosphere of 5% . Cells were transfected using FuGENE transfection reagent (FuGENE®6, Roche, France) when below 50% confluency, and left in culture for 24 to before the beginning of the experiments. Cells were then incubated with either native or coelenterazine ( , see http://www.interchim.com/) for 1 to prior to experiments. For experiments on neuro2A cells, cultures were serum-starved prior to the beginning of experiments to induce apoptosis.19 In other experiments, HEK-293 cells were stimulated with different concentrations of ATP (0.03 to ) and were then recorded with an electron-multiplying charge-coupled detector (EMCCD) on the Luminoview system or on a conventional microscope (details discussed later). ATP disodium salt (see http://www.sigmaaldrich.com/) was dissolved in deionized to a final concentration of and diluted further with phosphate buffered saline (PBS) prior to cell application. For stimulation of cells, of ATP solution was added by pipette to the MatTek dish ( ; see http://www.glass-bottom-dishes.com/) containing monolayers of HEK-293 cells and of medium. The dark box was then closed and acquisition started immediately.
Transgenic Drosophila Melanogaster
The generation of transgenic D. melanogaster expressing GFP-apoaequorin was previously described.19 Briefly, pG5A20 was inserted into the pUAST vector and transformants were then crossed with the P[elav-GAL4]C155 fly line from the Bloomington Stock Center (Bloomington, IN). Elav is expressed exclusively in neurons from midembryogenesis up to and including adulthood.
Where indicated, the -dependent luminescence of GFP-aequorin was detected using an imaging photon detector (IPD 3, Photek Limited, East Sussex, United Kingdom) on the baseport of a widefield inverted microscope (Axiovert 200M, Zeiss, Germany).21 The IPD is a position-sensing high gain photomultiplier tube using a cascaded stack of four multichannel plates [for a more detailed description, see http://www.photek.co.uk/phoprodf3.htm and, Ref. 22]. It consisted a bialkali photocathode and fiber optic input window. According to the manufacturer’s data, the sensitivity of the detector was highest in the blue to green wavelength range with quantum efficiencies ranging from 12.7% at to 5.2% at , and a dark noise of with a spatial resolution of at the photocathode surface. A oil objective lens with a numerical aperture (NA) of 1.3 was used in Neuro 2A studies, and a objective lens with a NA of 0.6 was used for studies on the fly brain. The detector, microscope, antivibration table, and fiber optics were housed inside a light-tight dark box (see, http://www.sciencewares.com/). Both the epifluorescence and halogen lamps were mounted outside the black box and connected via light guides to the microscope. The software enables automation of acquisition by the IPD or the CCD when the blackbox is closed (IPD for Windows 95, see http://www.sciencewares.com/). -induced bioluminescent signals were monitored continuously over long periods (hours) using the IPD (see Refs. 19, 21 for more details). For correlation to cell morphology, luminescence imaging was interrupted briefly so that brightfield or epifluorescence images could be periodically taken with a CCD connected to the microscope C-port. In studies on neuro2A cells undergoing apoptosis, brightfield images were taken once every hour.
Single-cell -induced luminescence of GFP-aequorin was also monitored with an electron multiplier CCD camera (Hamamatsu, Back-Thinned Frame Transfer, EM-CCD, C9100-13, image area, pixels, cooled to ) mounted on the baseport of a microscope optimized to detect luminescence (LV200, Luminoview Olympus Corporation, Japan). The luminescence imaging microscope is optimized for greater light collection by using high NA objectives and through improvements to the transmission efficiency of each optical element. In addition, the brightness of the image is further enhanced by a magnification lens behind the objective, which enables a larger field of view (FOV) to be projected onto a smaller chip area . In experiments here, we used a oil objective , which gave a final image magnification of and FOV of approximately .
Single-cell imaging was also undertaken with an EMCCD (iXon DV887 back-illuminated, array of pixels cooled to , Andor Technology, Belfast, UK) mounted on the baseport of a conventional microscope (200M, Zeiss, Germany). According to spectral response data for both cameras, the quantum efficiency of each EMCCD ranged from 90% at 500 and , to 87% at and 70% at . In both cases, the imaging chamber and objectives were housed inside a light-tight dark box.
Electron-Multiplying Charge-Coupled Detector and Image Photon Detector Low-Light Imaging Studies
In studies investigating the performance of the IPD and the EMCCD, both cameras were mounted on a Zeiss Axiovert 200M microscope, and light from light-emitting diodes (LEDs) was detected simultaneously using a beamsplitter. Specifically, the IPD (IPD425, bialkali or S20 photocathode, Photek, United Kingdom) was mounted on the sideport with a reducing lens (Zeiss to C-mount adapter), and the EMCCD (DV897, Andor Technology, array of pixels) was mounted on the baseport. The DV897 was operated at ( time resolution) and cooled to . Light was then projected through a pinhole onto an 1.2-mm area using a 430- or 555-nm LED. A triangle wave voltage in series with a load-limiting resistor was applied to the photodiode, which produced light with a softened peak.
Results and Discussion
Calcium-Dependent Green Fluorescent Protein Aequorin Bioluminescence
Recently it has been demonstrated that aequorin fusion to fluorescent proteins like GFP, 19, 20, 21, 23, 24, 25 enhances and facilitates its utility as an in-situ ion probe [Fig. 1a ]. These chimerical proteins undergo a so-called bioluminescent resonance energy transfer (BRET) in a -dependent manner. Thus, while gene expression of the reporter can be monitored via GFP fluorescence, signals are measured (in the absence of excitation light) simply by detecting green light emission because the light excited energy derived from the reaction of aequorin, , and coelenterazine is transferred nonradiatively to the GFP moiety [Fig. 1b]. That signals can be detected without need for light excitation arguably provides what may be considered one of the most significant advantages of bioluminescence experimental design over fluorescence-based methods: the absence of excitation light-induced phototoxicity. This is illustrated in a simple experiment based on the paradigm of spontaneous neuronal cell death (Fig. 2 ).
Using Neuro2A cells we aimed to record fluctuations during spontaneous cell death occurring in culture. Enhanced probability of spontaneous cell death came in the absence of serum, and the presence of millimolar quantities of extracellular calcium. In this case, our prior knowledge of the experimental system allowed us to estimate the approximate time window (several hours) over which cell death might occur, but not the exact moment. Further, while the role of fluctuations is strongly suggested, the exact “signature” remained completely unknown. Using GFP-aequorin, we were able to continuously record in a small field of cells during several hours with only brief interruptions, so that transmission snapshots could be monitored for spontaneous cell-death-associated changes (e.g., membrane blebbing and nuclear chromatin condensation). Cells displaying changes corresponding to spontaneous death were eventually identified easily from transmission light images. In all cases, analysis of related cellular changes was accompanied by short repeat bursts of spiking patterns [Figs. 2a, 2b, 2c]. Occasional individual spiking bursts of activity comprised prolonged (tens of seconds) plateaus superimposed by low- to high-frequency low-amplitude oscillations [e.g., Fig. 2c]. Initially, such bursts occurred occasionally, but eventually became more frequent as cell death progressed further during some [Fig. 2e] until cataclysmic cell death occurred [Figs. 2d and 2e]. These data indicate that at least for Neuro2A cells cultured under these conditions, spontaneous cell death events occur accompanied by distinct fluctuation patterns. Such a conclusion would be difficult to draw based on measurements using calcium-sensitive fluorescent dyes (e.g., Fura 2)14 or genetically encoded proteins16 because continuous excitation illumination itself can lead to phototoxicity and cell death.26
Monitoring by Bioluminescence In Vivo
In addition to frequency and amplitude, the spatial aspect of signaling at the level of cell-cell networks is widely believed an important aspect of information coding in vivo. This again presents a technical challenge, wherein ideally methods for detecting fluctuations should maintain high temporal resolution, over long durations, and be capable to detect simultaneously cell-cell network propagation where spatially heterogeneous signals cross many hundreds of microns to millimeters. Along these lines, GFP-aequorin bioluminescence-based methods for measuring calcium are singularly powerful.
Like other genetically encoded probes, GFP-apoaequorin can be targeted to subcellular compartments,21 or to specific cell types by transgenesis.19, 24 However, the advantage of using GFP-aequorin and bioluminescence is especially apparent considering the latter paradigm (Fig. 3 and Video 1 ). -dependent bioluminescent signals were detected in vivo in transgenic flies expressing GFP-aequorin exclusively in neurons of the brain. Prolonged recordings during more than ten hours allowed “spontaneous” signals with highly variable kinetic properties to be visualized in different areas of the fly brain [Figs. 3a and 3b]. These results highlight the capacity of this approach to resolve dynamic processes like signaling across a wide temporal range and simultaneously across wide areas containing multiple regions of interest. This is particularly useful for exploratory type studies where the characteristics of signals are unknown or highly variable, and/or unpredictable.10.1117/1.2937236.1
Electron-Multiplying Charge-Coupled Detector Detection of Ultralow Light Signals
As illustrated by previous examples with GFP-aequorin, IPD-based photon-counting detector technologies have to date provided the means to measure bioluminescence in single cells.21, 27 and intact organs/tissues.19 However, IPD technology has a number of disadvantages. Notably, IPD detectors are expensive and sophisticated; they are highly vulnerable to irreversible light damage, and quantum efficiencies are low with respect to genetically encoded fluorescent protein tags in common use, including GFP. These facts have strongly discouraged the routine use of IPD detectors and therefore single-cell bioluminescence in laboratories lacking the technical expertise to host this type of equipment. Fortunately, dependence on IPD detectors as the unique recourse to achieve single-cell bioluminescence detection has very recently been overturned by the development of EMCCD cameras,28, 29 a new generation of highly sensitive detectors measuring ultralow light levels with modest integration times (seconds). Further, in complete contrast to IPDs, EMCCDs are extremely resistant to light damage, and provide robust quantum efficiency in excess of 70% throughout most of the visible spectrum (wavelengths 450 to ), and as high as 90% from 500 to . We simultaneously compared the performance of each camera (IPD on the sideport versus EMCCD on the baseport) by detecting the light output from a source LED (in series with a limiting resistor and driven by a triangle wave voltage source) through a beamsplitter (see methods for details). At two wavelengths (430 versus ), the IPD (bialkali photocathode) detected less photons overall, but the ratio of the detected signal over the noise of the IPD was far superior to that obtained on the EMCCD [Fig. 4a ]. This means that the IPD should provide a much better temporal resolution than the EMCCD. On the other hand, the QE of the EMCCD is significantly higher than the IPD, which means that overall the EMCCD detects a lot more photons and will give better counting statistics, providing that the dark noise inside an ROI is kept below 10% of the statistical noise limit [Fig. 4a]. In addition, our results showed that the spatial resolution of the EMCCD was at least twice as good as that obtained on the IPD [Fig. 4b].
Facile Long-Term Continuous Measurement of Single-Cell -Dependent Bioluminescence
The temporal resolution of the EMCCD can be improved by at least two means. First, by increasing the total amount of light that can be collected, and second, by enhancing the light emission of the bioluminescence reaction. In the first case, using optics with a high numerical aperture and transmission efficiency can optimize light collection.30 Very recently, one microscope manufacturer released a dedicated bioluminescence imaging microscope (Luminoview, LV-200, Olympus Corporation, Japan) designed to yield significantly greater amounts of light collection compared to a conventional light (epifluorescence) microscope. The improvement in photon collection performance comes from a proprietary adaptation of the internal light relay optics, which we predicted in combination with an EMCCD device should provide for overall bioluminescence imaging performance at least as well as conventional IPD methods, and with more practical (robust) characteristics. Along these lines, we tested the combination of a highly sensitive EMCCD (Hamamatsu C9100-13), see Materials and Methods in Sec. 2 porting a so-called back-thinned CCD chip that provides the very highest photon sensitivity, combined with very low noise. As a target, we measured ATP stimulated purinergic receptor-driven fluctuations in HEK-293 cells (transfected to express GFP-apoaequorin) (Fig. 5 and Video 2 ). The addition of ATP to culture medium resulted immediately in heterogeneous transients, detected in nearly all cells [Fig. 5a and Video 2]. In this representative example, compared to nonstimulated conditions, ATP stimulated large increases in the light intensity of individual cells [e.g., Fig. 5b]. Depending on the concentration of ATP used, many cells displayed single isolated transients of short duration [5 to ; Figs. 5b and 5d]. Further analysis of longer duration recordings revealed that some of these transients comprised oscillations [Fig. 5c].10.1117/1.2937236.2
Our studies and those of others suggest that the BRET reaction of GFP-aequorin produces a significant enhancement in the overall light output of aequorin alone. 19, 20, 21, 23, 24, 25, 31 The amount of light emission produced by bioluminescent reactions can be further enhanced by increasing the level of reporter expression (e.g., by viral mediated gene transfer) or by using substrates with modified light emission properties. Synthetic analogs of coelenterazine are commercially available that confer different affinities and spectral properties on aequorin,32 and can be used instead of native coelenterazine for lower or higher sensitivity. HEK-293 cells endogenously express low levels of P2Y receptors and induce low amplitude responses,33 and we therefore loaded cells with the ultrasensitive analog, coelenterazine . Coelenterazine is estimated to have a relative luminescence intensity at least ten times higher than aequorin reconstituted with native coelenterazine.32 In this condition, high concentrations [Fig. 6a and Video 3 ] or low concentrations of ATP [Fig. 6b and Video 4 ] induced oscillations that could be readily detected using an EMCCD camera (iXon DV887, Andor Technology) attached to the baseport of a conventional microscope (Axiovert 200M, Zeiss, Germany). Using the luminescence microscope (LV200, Luminoview, Olympus Corporation, Japan), together with coelenterazine , should therefore deliver even higher sensitivity for EMCCD detection of signals with GFP-aequorin and further improve temporal resolution.10.1117/1.2937236.310.1117/1.2937236.4
Conclusions and Perspectives
Live-cell imaging using fluorescence has a number of limitations because of phototoxicity; photobleaching and the amount of excitation light used must be balanced between the quality of the image and cell viability. Bioluminescence imaging does not require light excitation, but presents the difficulty of detecting sufficient luminescence so that an image can be formed. Previous studies demonstrating the improved light emission properties of the BRET based-approach helped to make possible single-cell measurements using an intensified CCD (ICCD),20 or IPD-based technologies.21 We reproduce this approach and compare IPD performance with EMCCD. Image detection by EMCCD is facile compared with intensified devices, and allows us to easily detect receptor-driven oscillations in single cells. In as much as these are functional fluctuations representing changes between 50 and (estimated from fura-2 measurements in cell suspensions),33 this suggests EMCCD detection is much more sensitive than ICCD where comparable integration times are only sufficient to detect immense, nonphysiological calcium fluxes driven by application of A21387 ionophore drug in the presence of high levels of extracellular calcium.20 ICCDs or cameras, such as the Electron Bombardment CCD, can also have other drawbacks such as resolution artifacts, higher noise, and lower QEs.29 EMCCD performance is also more comparable to IPD methods that are mostly the only recourse for single-cell detection by bioluminescence. Overall, improvements in the sensitivity provided by the sensitive probe, GFP-aequorin, and camera technologies, like the EMCCD, will extend current applications with this approach and promises to deliver high sensitivity and readout speeds for other dynamic reporter systems based on BRET.
The authors thank Q. Trinh-xuan (Hamamatsu, France), H. Gautier (Olympus, France), and C. Machu (PFID). This work was funded by grants from the Région Ile de France (program SESAME), Project MODEXA du pole de compétitivité Medicen Paris-Région (coordinator Escaich), European Commission (FP6 NEST program; project AUTOMATION), and the CNRS, ANR, and the Institut Pasteur.