JOURNAL OF BIOLUMINESCENCE AND CHEMILUMINESCENCE VOL 5 123-130 (1990)
CCD Imaging of Luciferase Gene Expression in Single Mammalian Cells
Claire E. H o o p e r Robens Institute, University of Surrey, Guildford, Surrey GU2 5XH, UK
R. E. Ansorge Cavendish Laboratory, Department of Physics, University of Cambridge, Madingley Road, Cambridge CB3 OHE, and Image Research Ltd, St John's Innovation Centre, Cowley Road, Cambridge CB4 4WS. UK
Helena M. B r o w n e Division of Virology, Department of Pathology, University of Cambridge, Tennis Court Road, Cambridge CB2 IQP, UK
Patricia Tomkins Photonic Science, Old Station House, Robertsbridge, East Sussex TN32 5DG, UK
Quantitative and sensitive imaging of chemiluminescence, bioluminescence and fluorescence emissions is emerging as an increasingly important technique for a range of biomedical applications (Hooper e t a/.,1990). A brief review of low-light-level imaging is presented, w i t h particular reference t o charge-coupled devices (CCD). Detectors for sensitive imaging are described and compared, including various CCDs and photoncounting devices. Image analysis techniques based on digital image processing, may be applied t o quantify luminescent processes w i t h these detectors. Images of luciferase gene expression i n single mammalian cells have been obtained using a particular highsensitivity intensified CCD camera. The method is illustrated using cell monolayers infected w i t h recombinant vaccinia virus encoding the firefly luciferase, luc gene (Rodriguez e t a/., 1988). The CCD camera has been used t o detect luciferase expression in single, recombinant infected cells amongst over one million non-infected cells. The rapid detection of luciferase-expressing viruses may be used f o r the selection of virus deletion mutants into which the luciferase gene has been cloned at specific sites. This is particularly useful in the case of viruses such as cytomegalovirus which have slow replication cycles. This direct imaging technique is simple and versatile. It offers a rapid, non-invasive method for the sensitive detection of luciferase activity i n single, luciferase-expressing cells. One can envisage the use o f luciferase as a sensitive and convenient co-selection marker gene in the analysis of both gene expression and protein function. These methods offer tremendous potential i n the fields of molecular and cellular biology. Keywords: CCD; imaging; gene expression; single cells 0884-3996/90/020123-08$05.00 0 1990 by John Wiley & Sons, Ltd
Received 4 December 1989
124
C. E. HOOPER ET AL.
1 cm2.The sensitive area is divided into a rectangular matrix of light-sensitive elements (pixels) deMany processes of interest in the biomedical sci- fined by an array of closely spaced electrodes. ences are associated with low light-level emissions. Electrons liberated by incident photons migrate Such processes may be interesting in their own towards the electrodes within each pixel. These right (e.g. luminescent organisms) or increasingly pixels behave as a bucket or potential well in which as markers or labels (e.g. luciferase gene). A number electrons are accumulated. After a suitable integraof detectors are in use for low light measurement tion time the charges stored in each pixel may be (i.e. where the quantum nature of light is relevant) read out serially by a process called charge-couand we review them here, before discussing our pling (Janesick and Blouke, 1987). During the readresults. out process, the charge from each pixel may be used to create an analogue TV picture or digitized to create a digital image in a digital frame store. LOW LIGHT-LEVEL SENSORS There are three architectures for a CCD, ‘full There are two distinct physical processes used in frame’, ‘frame transfer’ and ‘interline’. The full frame CCD has the simplest structure with a single low light-level detectors, these are: area of pixels which integrate light for a period of (1) The Photoelectric effect. This is the process time. The exposure time may be controlled using a whereby a single electron is ejected from a shutter which then shields the light-sensitive pixels suitable material (the photocathode) by a sin- during the lengthy read-out process. Such a CCD is gle incident photon. Amplification of the not suitable for real-time applications. photoelectrons is usually necessary to produce The frame transfer CCD has two equal areas of useful signals. This process is the basis of detecpixels, the integration area which receives incident tion for photomultipliers, image intensifiers light and the storage area which is masked from and photon-counting image detectors. light. In operation an image is accumulated in the (2) Solid-state detectors. These use the effect that a integration area and rapidly shifted (typically single photon incident on a suitable (pnp) hundreds of microseconds) into the storage area. silicon junction may be absorbed and create a The image in the storage area is then read out single electron-hole pair. This process is the serially whilst the next image is being integrated. A basis for silicon photodiodes and chargeframe transfer CCD is suitable for real-time (TV coupled devices. frame rate) operation. The interline CCD is a variant of the frame Sensors based on these principles are the photomultiplier tube, the charge-coupled device and the transfer CCD, which has the integration area interleaved, line by line, with the storage area. image intensifier. INTRO D UCTlON
(a) The photomultiplier tube (PMT)
The PMT uses the photoelectric effect. The photoelectrons produced are amplified by a cascade electrode structure with a typical gain of lo6. Associated electronics yield a current proportional to the light flux for high rates. Alternatively a pulse train can be generated for photon counting at low light-levels. The PMT is used in a wide range of luminometers. (b) The charge-coupled device (CCD): full frame, frame transfer(vide0) and interline
The CCD is a sophisticated solid-state silicon detector, which has a photosensitive area of typically
(c) Image intensifier: first and second generation
Image intensifiers depend on the photoelectric effect to convert a visible image to an electron image. An image intensifier reproduces at the output an image identical with one presented at the input. A net gain in intensity is produced without drastically affecting the contrast of the image. Primary photoelectrons produced by the photocathode may be used in two ways. (1) First generation tubes. Here primary photoelectrons, which are accelerated by a high electric field, impinge on a phosphorescent screen. The energy gained by the electrons in the field is released as a flash of light in the screen. With
125
CCD IMAGING
modern screens, several hundred photons may be released for a single energetic electron impact. (2) Second generation tubes. Here the primary photoelectrons are multiplied by a microchannel plate, which consists of an array of microscopic channel electron multipliers. The spatial coherence of the image is preserved during this amplification process. Many thousands of electrons may be released for each electron which strikes the microchannel plate. These electrons are then accelerated onto a phosphor screen, where light is produced as in first generation devices. Each photoelectron entering the microchannel plate releases up to hundreds of thousands of photons from the phosphor screen. The electron multiplication enables far higher intensification to be achieved than with first generation devices. INSTRUMENTS COMMERCIALLY AVAl LAB LE
photons even though the CCD is operating at room temperature. In these systems the CCD can be operated at TV frame rates giving a true real-time imaging capability. The dynamic range for a single CCD frame is much less than for the cooled CCD (typically 10 to 100). However, by means of the frame averaging methods the dynamic range can be greatly extended to ultra-low light-levels. For example a dynamic range of 10,000:1 for an exposure time of one minute can be achieved. (c) Photon counting devices
Devices are available from, for example, Hamamatsu (Hamamatsu City, Japan) and Instrument Technology Ltd (St Leonards-on-Sea, Sussex, UK). These devices are similar to high-gain secondgeneration image intensifiers and have two or three microchannel plate stages. Single detected photons produce large output pulses whose positions are determined in real-time using a two-dimensional output stage which can be a variety of resistive or conductive anodes, a semiconductor array or even a CCD.
(a) Slow scan, cooled CCD detectors
Detectors are available from, for example Photometrics (Tucson, Arizona, USA), Astromed Ltd (Cambridge, UK) and Wright Instruments (London, UK). A full frame CCD, cooled to liquid nitrogen temperatures, can be used as an imaging photon integrator without adding any means of electron amplification. Cooling is essential to reduce or eliminate thermal noise during the integration time, to the level of a few electrons per pixel. The dynamic range of the signal within one pixel can be up to 65,000 (16 bits) because of the very low noise introduced. Typically the read-out time for a slow scan CCD is many seconds for a whole frame, thus limiting its use for real time applications. (b) Real-time, intensified CCD detectors
Detectors are available from, for example, Image Research Ltd (Cambridge, UK) and Photonic Science (Robertsbridge, Sussex, UK). A particularly interesting combination is an image intensifier coupled at its output to a CCD. The intensifier increases the brightness of the image at the CCD by several orders of magnitude. For this reason the intensified CCD detector can detect single incident
CHARACTERISTICS OF IMAGING SYSTEMS (a) Spectral response
All the detectors being discussed have sensitivities which are wavelength-dependent. For photocathode based detectors, the spectral response is dependent on the type of photocathode. Curves showing the response of various photocathodes are readily available and may be tailored to suit a particular application. CCDs have a spectral response somewhat different to photocathodes, tending to extend further into the red. The match between the spectral response of the detector and the spectral emission of the luminescent reaction being measured is an important consideration.
( b ) Sensitivity The sensitivity of a detector depends on the efficiency by which photons are converted to detectable electron. The peak efficiency for a photocathodebased device is typically 20% at about 400nm. Somewhat higher efficiencies can be achieved by
C.E. HOOPER E T A L .
126
CCDs, which are efficient at longer wavelengths than photocathodes. An additional consideration is the noise superimposed on a real signal by any detector. This includes noise from thermal effects and read-out noise from the associated amplification system. A cryogenic CCD had a noise equivalent to about 10 electrons per pixel, mainly due to read-out noise. Intensified CCDs and photon counting systems produce a large output signal for a detected signal photon, well above the local noise level. However, photocathode noise will produce similar random signals in the field of view. In the case of a steady but weak signal the importance of noise can be reduced by time averaging. Improvements in the signal/noise ratio are proportional to the square root of the time for which the averaging is performed.
(e) Resolution
(c) Image quality: geometric stability and uniformity
I M A G E ANALYSIS
CCD based detectors have excellent geometric stability. Photon-counting image devices are less good in this respect. Low-speed devices such as those using resistive anodes can suffer coincidence infilling and overload at high light-levels which reduces image quality. The response of a photocathode and the gain of an image intensifier can vary across the field of view. The importance of these effects on image quality can be minimized by calibration.
( d ) Dynamic range The dynamic range may be regarded as the range of light-levels over which the system is useful. The lower end of the range is set by the sensitivity and the upper end by the saturation of response or overload characteristics of the detector. The cryogenic CCD has the greatest dynamic range, typically 5 x lo4: 1 per pixel. An intensified CCD has a dynamic range of about 100: 1 per pixel per frame. However, real-time frame averaging can extend this to i04:i per pixel. Photon-counting systems have a dynamic range which is limited by complicated effect of overload and various versions show different limiting effects. Generally, they have a local saturation limit which may produce problems with photometry should the image contain bright highlights.
Nearly all systems have resolutions of the order of 500 x 500 pixel. Systems with 1000 x 1000 pixels are starting to become available but currently these are very expensive. ( f ) Real-time capability
True real-time performance is only available for systems operating at T V frame rates. It is important to bear this in mind if a study of fast kinetics is required. Moreover, a fairly sophisticated computing system is needed to handle data at these high rates. With commercially available high-speed frame grabbers such systems may be implemented relatively easily and economically.
The digital images obtained by these systems can be manipulated and enhanced by a wide range of well known digital image-processing methods (Gonzalez and Wintz, 1987). Some examples of these methods, including frame averaging, background subtraction, smoothing and edge finding, are shown in Fig. l(a) to l(f). QUANTITATIVE I M A G E ANALYSIS
A digital image can be used for quantitative as well as qualitative work. A number of corrections to the values in each pixel are typically required. These corrections include dark field subtraction, bright field correction (i.e. correction for detector response variation across the field of view), and possible geometrical corrections. APPLICATION OF INTENSIFIED CCD IM A G IN G
The potential application of intensified CCD imaging to the study of luciferase gene expression in mammalian cells has been demonstrated using a vaccinia virus recombinant. This virus, constructed by Rodriquez et al. (1988) contains the gene for firefly luciferase cloned into the thymidine kinase locus of the vaccinia genome by means of homologous recombination. Luciferase expression in cells
CCD IMAGING
127
(a)
Figure 1. (a) Single TV frame of 2 x 2 mm grid engraved on base of petri dish, illuminated by low light-level source and directly viewed by a sensitive high-gain CCD image intensifier, the Biomedical Image Quantifier. (b) Average of 256 frames like (a). (c) Average of 256 frames without petri dish (bright field). (d) As (b). n o w corrected for gain variation. (e) AS (b). with smoothing filter applied. (f) As (c), with edge finding filter applied
(b) Cursor a t 237
191 Lop LH c o r n e r o f zone
1 8 x 18 p i x e l readout
1
1
2
2
1
1
1
1
0
0
0
0
0
0
0
0
2
1
2
1
2
1
3
3
1
2
1
3
1
3
3
1
1
0
0
0
5
6
1
2
1
1
1
2
2
2
2
2
1
1
0
1
0
1
1
1
0
1
2
3
3
4
2
1
2
2
1
2
6
3
2
2
4
1
0
3
2
2
7
4
4
3
5
3
4
2
1
0
0
2
4
0
1
0
3
5
2
2
2
1
3
3
2
4
4
3
1
0
6
1
3
1
2
5
6
4
4
1
2
3
3
3
3
6
6
4
3
1
3
3 1 0 1 2 1 5 1 0
3
3
3
1 9 1 3 1 2
7
4
4
1
4
2
1 1 1 2 3 2 7 1 6
7
5
5
7
9
9
8
1
3
2
3
12
15
22
26
22
10
2
1
9
13
17
20
24
11
4
I
2
4
11
22
37
44
29
12
j
5
9
17
29
31
17
1
2
9 1 1
3
2
1
9
20
32
36
25
8
4
6
13
18
34
42
29
7
1 1 11
2
8
I4
34
35
20
10
3
6
6
20
34
33
17
6
3
3
3
4
13
19
I6
12
5
4
4
4
10
20
25
13
2
1
1
1
2
2
1
1
3
3
6
3
4
6
3
7
6
6
4
5
3
2
3
1
1
2
2
4
1
3
2
1
1
1
3
2
3
3
2
1
1
2
1
1
3
3
2
3
3
3
2
0
2
4
3
2
1
1
0
0
2
2
0
0
3
2
3
?
0
0
0
0
1
0
1
Figure 2. Analysis of luciferase activity in vaccinia recombinant virus infected cell monolayer. viewed using the Biomedical Image Quantifier. The petri dish is demagnified onto the camera by a factor of 9 5 : l . (a) 20-cm2 petri dish, viewed 1 2 hours post-infection (average of 1000 frames). (b) Digital pixel values for t w o bright spots in (a). & (c)-(f) Direct imaging of small area (1.4 cm2) of infected cells (c) 36 hours post-infection (average of 256 frames). (d) 9 hours post-infection (average of 1000 frames). (e) 6 hours post-infection (average of 1000frames). (f) Zoom display of (e) (top left) showing individual lightemittrng cells.
129
CCD IMAGING
infected with this virus has been previously demonstrated using conventional photographic detection methods at 24 h after initial infection. We have examined the sensitivity and precision of intensified CCD imaging for the analysis of luciferase activity in mammalian cells infected with this vaccinia recombinant at a range of times after infection. The imaging system used for this study was the Biomedical Image Quantifier as supplied by Image Research Ltd, and Photonic Science. For all experiments, monolayers of CV-1 cells (derived from African Green Monkey kidney) were sealed in 20cm2 plastic petridishes. Each dish of 2 x lo6 cells was infected with approximately 500 plaque-forming units of the luciferase-expressing vaccinia virus. Light detection was performed at 6 h, 9 h, 12 h and 36 h post-infection, after removal of the growth medium and addition of luciferase substrate solution (De Wet et al., 1987). Fig. 2(a) shows the image obtained from a 20-cm2 dish of cells infected with recombinant virus for 9 hours. The dish was viewed using a f1.8, 25 mm lens. At this time after infection the virus infection is confined to single cells in which the virus genome is replicating before the onset of new virus particle production. The non-synchronous nature of the infection is indicated by the variation in intensity between individual ‘glowing’ cells. This method of analysis is clearly valuable for visualizing a large area of cells in which low numbers may be expressing the luciferase gene. Direct imaging of a small1 area of the infected cell monolayer (1.4 cm2) with the intensified CCD is shown in Figs 2(c)-2(f). Fig. 2(c) illustrates the sensitivity of this method when applied to recombinant infected cells which have been incubated for 36 hours. Although a few cells can be distinguished, the image is mostly a result of plaque formation, where new virus particles are produced and released to infect surrounding cells. This is the time post-infection when luciferase can be detected using photographic methods. Fig. 2(d) shows cells infected with the recombinant virus after 9 hours. Individual cells are easily detected. Variation in intensity between cells is the result of the non-synchronous virus infection and correlates with the levels of virus DNA replication and luciferase expression within these cells. Selection of individual light-emitting cells from a large population, without the problem of contamination from adjacent infected cells, would be relatively
straightforward using this imaging detection method. Fig. 2(e) illustrates the extreme sensitivity of this method for the early detection of single foci of infection. Luciferase activity was analysed 6 hours after infection. Even at this early time,.when virus DNA replication has only just begun, distinct lightemitting cells could be visualized. The single lightemitting cells are enlarged in Fig. 2(f) using a zoom display function. CONCLUSIONS
The detection of single mammalian cells, infected with recombinant vaccinia virus expressing luciferase, demonstrates that intensified CCD imaging offers a wide range of potential applications to the study of biological systems. We hope to use this method to select virus deletion mutants into which the luciferase gene will be introduced at specific sites to disrupt desired genes. Recombinant viruses, which are likely to constitute as little as 0.1% of the transfection progeny, must be distinguished from wild type and subsequently picked and cloned. A rapid and sensitive method detection system is particularly desirable for selecting mutants of viruses which grow slowly in vitro in tissue culture, such as human cytomegalovirus. This non-invasive imaging technique would enable single recombinant infected cells to be picked and plaque purified in a considerably reduced time. The ability to see single cells at early times of infection would also minimize contamination with wildtype parental virus. Virus deletion mutants generated and selected in this way would provide valuable tools for investigating the biological function of specific virus gene products. Wider application of this approach to molecular and cellular biology can be envisaged. Rapid screening of cells co-transfected with luciferase as a selectable marker gene would allow easy selection of stable eukaryotic cell lines. This offers advantages both in uitro and in uivo for the analysis of gene expression and protein function. REFER ENCES De Wet, J. R., Wood, K. V., DeLuca, M., Helinski, D. R. and Subramani, S. (1987). Cloning and expression of the firefly luciferase gene in mammalian cells. In: Bioluminescence and
130 Chemiluminescence: New Perspectives, Scholmerich, J., Andreesen, R., Kapp, A,, Ernest, M., Woods, W. G., (Eds), John Wiley, Chichester, pp. 369-372. Gonzalez, R. C. and Wintz, P. (1987). DigitaI Image Processing, Addison-Wesley, Reading, MA USA. Hooper, C. E., Ansorge, R. E. (1990). Trends in Analytical Chemisfry (in preparation).
C. E. HOOPER ET AL.
Janesick, J. R. and Blouke, M. (1987). Sky on a chip: the fabulous CCD. S k y and Telescope, 238-242. Rodriguez, J. F., Rodriquez, D., Rodriguez, J. R., McGowan, E. B. and Esteban, M. (1988). Expression of the firefly luciferase gene in vaccinia virus: a highly sensitive gene marker to follow virus dissemination in tissues of infected animals. Proc. Nail. Acad. Sci. U S A . , 85, 1667-1671.
CCD Imaging of Luciferase Gene Expression in Single Mammalian Cells
Claire E. H o o p e r Robens Institute, University of Surrey, Guildford, Surrey GU2 5XH, UK
R. E. Ansorge Cavendish Laboratory, Department of Physics, University of Cambridge, Madingley Road, Cambridge CB3 OHE, and Image Research Ltd, St John's Innovation Centre, Cowley Road, Cambridge CB4 4WS. UK
Helena M. B r o w n e Division of Virology, Department of Pathology, University of Cambridge, Tennis Court Road, Cambridge CB2 IQP, UK
Patricia Tomkins Photonic Science, Old Station House, Robertsbridge, East Sussex TN32 5DG, UK
Quantitative and sensitive imaging of chemiluminescence, bioluminescence and fluorescence emissions is emerging as an increasingly important technique for a range of biomedical applications (Hooper e t a/.,1990). A brief review of low-light-level imaging is presented, w i t h particular reference t o charge-coupled devices (CCD). Detectors for sensitive imaging are described and compared, including various CCDs and photoncounting devices. Image analysis techniques based on digital image processing, may be applied t o quantify luminescent processes w i t h these detectors. Images of luciferase gene expression i n single mammalian cells have been obtained using a particular highsensitivity intensified CCD camera. The method is illustrated using cell monolayers infected w i t h recombinant vaccinia virus encoding the firefly luciferase, luc gene (Rodriguez e t a/., 1988). The CCD camera has been used t o detect luciferase expression in single, recombinant infected cells amongst over one million non-infected cells. The rapid detection of luciferase-expressing viruses may be used f o r the selection of virus deletion mutants into which the luciferase gene has been cloned at specific sites. This is particularly useful in the case of viruses such as cytomegalovirus which have slow replication cycles. This direct imaging technique is simple and versatile. It offers a rapid, non-invasive method for the sensitive detection of luciferase activity i n single, luciferase-expressing cells. One can envisage the use o f luciferase as a sensitive and convenient co-selection marker gene in the analysis of both gene expression and protein function. These methods offer tremendous potential i n the fields of molecular and cellular biology. Keywords: CCD; imaging; gene expression; single cells 0884-3996/90/020123-08$05.00 0 1990 by John Wiley & Sons, Ltd
Received 4 December 1989
124
C. E. HOOPER ET AL.
1 cm2.The sensitive area is divided into a rectangular matrix of light-sensitive elements (pixels) deMany processes of interest in the biomedical sci- fined by an array of closely spaced electrodes. ences are associated with low light-level emissions. Electrons liberated by incident photons migrate Such processes may be interesting in their own towards the electrodes within each pixel. These right (e.g. luminescent organisms) or increasingly pixels behave as a bucket or potential well in which as markers or labels (e.g. luciferase gene). A number electrons are accumulated. After a suitable integraof detectors are in use for low light measurement tion time the charges stored in each pixel may be (i.e. where the quantum nature of light is relevant) read out serially by a process called charge-couand we review them here, before discussing our pling (Janesick and Blouke, 1987). During the readresults. out process, the charge from each pixel may be used to create an analogue TV picture or digitized to create a digital image in a digital frame store. LOW LIGHT-LEVEL SENSORS There are three architectures for a CCD, ‘full There are two distinct physical processes used in frame’, ‘frame transfer’ and ‘interline’. The full frame CCD has the simplest structure with a single low light-level detectors, these are: area of pixels which integrate light for a period of (1) The Photoelectric effect. This is the process time. The exposure time may be controlled using a whereby a single electron is ejected from a shutter which then shields the light-sensitive pixels suitable material (the photocathode) by a sin- during the lengthy read-out process. Such a CCD is gle incident photon. Amplification of the not suitable for real-time applications. photoelectrons is usually necessary to produce The frame transfer CCD has two equal areas of useful signals. This process is the basis of detecpixels, the integration area which receives incident tion for photomultipliers, image intensifiers light and the storage area which is masked from and photon-counting image detectors. light. In operation an image is accumulated in the (2) Solid-state detectors. These use the effect that a integration area and rapidly shifted (typically single photon incident on a suitable (pnp) hundreds of microseconds) into the storage area. silicon junction may be absorbed and create a The image in the storage area is then read out single electron-hole pair. This process is the serially whilst the next image is being integrated. A basis for silicon photodiodes and chargeframe transfer CCD is suitable for real-time (TV coupled devices. frame rate) operation. The interline CCD is a variant of the frame Sensors based on these principles are the photomultiplier tube, the charge-coupled device and the transfer CCD, which has the integration area interleaved, line by line, with the storage area. image intensifier. INTRO D UCTlON
(a) The photomultiplier tube (PMT)
The PMT uses the photoelectric effect. The photoelectrons produced are amplified by a cascade electrode structure with a typical gain of lo6. Associated electronics yield a current proportional to the light flux for high rates. Alternatively a pulse train can be generated for photon counting at low light-levels. The PMT is used in a wide range of luminometers. (b) The charge-coupled device (CCD): full frame, frame transfer(vide0) and interline
The CCD is a sophisticated solid-state silicon detector, which has a photosensitive area of typically
(c) Image intensifier: first and second generation
Image intensifiers depend on the photoelectric effect to convert a visible image to an electron image. An image intensifier reproduces at the output an image identical with one presented at the input. A net gain in intensity is produced without drastically affecting the contrast of the image. Primary photoelectrons produced by the photocathode may be used in two ways. (1) First generation tubes. Here primary photoelectrons, which are accelerated by a high electric field, impinge on a phosphorescent screen. The energy gained by the electrons in the field is released as a flash of light in the screen. With
125
CCD IMAGING
modern screens, several hundred photons may be released for a single energetic electron impact. (2) Second generation tubes. Here the primary photoelectrons are multiplied by a microchannel plate, which consists of an array of microscopic channel electron multipliers. The spatial coherence of the image is preserved during this amplification process. Many thousands of electrons may be released for each electron which strikes the microchannel plate. These electrons are then accelerated onto a phosphor screen, where light is produced as in first generation devices. Each photoelectron entering the microchannel plate releases up to hundreds of thousands of photons from the phosphor screen. The electron multiplication enables far higher intensification to be achieved than with first generation devices. INSTRUMENTS COMMERCIALLY AVAl LAB LE
photons even though the CCD is operating at room temperature. In these systems the CCD can be operated at TV frame rates giving a true real-time imaging capability. The dynamic range for a single CCD frame is much less than for the cooled CCD (typically 10 to 100). However, by means of the frame averaging methods the dynamic range can be greatly extended to ultra-low light-levels. For example a dynamic range of 10,000:1 for an exposure time of one minute can be achieved. (c) Photon counting devices
Devices are available from, for example, Hamamatsu (Hamamatsu City, Japan) and Instrument Technology Ltd (St Leonards-on-Sea, Sussex, UK). These devices are similar to high-gain secondgeneration image intensifiers and have two or three microchannel plate stages. Single detected photons produce large output pulses whose positions are determined in real-time using a two-dimensional output stage which can be a variety of resistive or conductive anodes, a semiconductor array or even a CCD.
(a) Slow scan, cooled CCD detectors
Detectors are available from, for example Photometrics (Tucson, Arizona, USA), Astromed Ltd (Cambridge, UK) and Wright Instruments (London, UK). A full frame CCD, cooled to liquid nitrogen temperatures, can be used as an imaging photon integrator without adding any means of electron amplification. Cooling is essential to reduce or eliminate thermal noise during the integration time, to the level of a few electrons per pixel. The dynamic range of the signal within one pixel can be up to 65,000 (16 bits) because of the very low noise introduced. Typically the read-out time for a slow scan CCD is many seconds for a whole frame, thus limiting its use for real time applications. (b) Real-time, intensified CCD detectors
Detectors are available from, for example, Image Research Ltd (Cambridge, UK) and Photonic Science (Robertsbridge, Sussex, UK). A particularly interesting combination is an image intensifier coupled at its output to a CCD. The intensifier increases the brightness of the image at the CCD by several orders of magnitude. For this reason the intensified CCD detector can detect single incident
CHARACTERISTICS OF IMAGING SYSTEMS (a) Spectral response
All the detectors being discussed have sensitivities which are wavelength-dependent. For photocathode based detectors, the spectral response is dependent on the type of photocathode. Curves showing the response of various photocathodes are readily available and may be tailored to suit a particular application. CCDs have a spectral response somewhat different to photocathodes, tending to extend further into the red. The match between the spectral response of the detector and the spectral emission of the luminescent reaction being measured is an important consideration.
( b ) Sensitivity The sensitivity of a detector depends on the efficiency by which photons are converted to detectable electron. The peak efficiency for a photocathodebased device is typically 20% at about 400nm. Somewhat higher efficiencies can be achieved by
C.E. HOOPER E T A L .
126
CCDs, which are efficient at longer wavelengths than photocathodes. An additional consideration is the noise superimposed on a real signal by any detector. This includes noise from thermal effects and read-out noise from the associated amplification system. A cryogenic CCD had a noise equivalent to about 10 electrons per pixel, mainly due to read-out noise. Intensified CCDs and photon counting systems produce a large output signal for a detected signal photon, well above the local noise level. However, photocathode noise will produce similar random signals in the field of view. In the case of a steady but weak signal the importance of noise can be reduced by time averaging. Improvements in the signal/noise ratio are proportional to the square root of the time for which the averaging is performed.
(e) Resolution
(c) Image quality: geometric stability and uniformity
I M A G E ANALYSIS
CCD based detectors have excellent geometric stability. Photon-counting image devices are less good in this respect. Low-speed devices such as those using resistive anodes can suffer coincidence infilling and overload at high light-levels which reduces image quality. The response of a photocathode and the gain of an image intensifier can vary across the field of view. The importance of these effects on image quality can be minimized by calibration.
( d ) Dynamic range The dynamic range may be regarded as the range of light-levels over which the system is useful. The lower end of the range is set by the sensitivity and the upper end by the saturation of response or overload characteristics of the detector. The cryogenic CCD has the greatest dynamic range, typically 5 x lo4: 1 per pixel. An intensified CCD has a dynamic range of about 100: 1 per pixel per frame. However, real-time frame averaging can extend this to i04:i per pixel. Photon-counting systems have a dynamic range which is limited by complicated effect of overload and various versions show different limiting effects. Generally, they have a local saturation limit which may produce problems with photometry should the image contain bright highlights.
Nearly all systems have resolutions of the order of 500 x 500 pixel. Systems with 1000 x 1000 pixels are starting to become available but currently these are very expensive. ( f ) Real-time capability
True real-time performance is only available for systems operating at T V frame rates. It is important to bear this in mind if a study of fast kinetics is required. Moreover, a fairly sophisticated computing system is needed to handle data at these high rates. With commercially available high-speed frame grabbers such systems may be implemented relatively easily and economically.
The digital images obtained by these systems can be manipulated and enhanced by a wide range of well known digital image-processing methods (Gonzalez and Wintz, 1987). Some examples of these methods, including frame averaging, background subtraction, smoothing and edge finding, are shown in Fig. l(a) to l(f). QUANTITATIVE I M A G E ANALYSIS
A digital image can be used for quantitative as well as qualitative work. A number of corrections to the values in each pixel are typically required. These corrections include dark field subtraction, bright field correction (i.e. correction for detector response variation across the field of view), and possible geometrical corrections. APPLICATION OF INTENSIFIED CCD IM A G IN G
The potential application of intensified CCD imaging to the study of luciferase gene expression in mammalian cells has been demonstrated using a vaccinia virus recombinant. This virus, constructed by Rodriquez et al. (1988) contains the gene for firefly luciferase cloned into the thymidine kinase locus of the vaccinia genome by means of homologous recombination. Luciferase expression in cells
CCD IMAGING
127
(a)
Figure 1. (a) Single TV frame of 2 x 2 mm grid engraved on base of petri dish, illuminated by low light-level source and directly viewed by a sensitive high-gain CCD image intensifier, the Biomedical Image Quantifier. (b) Average of 256 frames like (a). (c) Average of 256 frames without petri dish (bright field). (d) As (b). n o w corrected for gain variation. (e) AS (b). with smoothing filter applied. (f) As (c), with edge finding filter applied
(b) Cursor a t 237
191 Lop LH c o r n e r o f zone
1 8 x 18 p i x e l readout
1
1
2
2
1
1
1
1
0
0
0
0
0
0
0
0
2
1
2
1
2
1
3
3
1
2
1
3
1
3
3
1
1
0
0
0
5
6
1
2
1
1
1
2
2
2
2
2
1
1
0
1
0
1
1
1
0
1
2
3
3
4
2
1
2
2
1
2
6
3
2
2
4
1
0
3
2
2
7
4
4
3
5
3
4
2
1
0
0
2
4
0
1
0
3
5
2
2
2
1
3
3
2
4
4
3
1
0
6
1
3
1
2
5
6
4
4
1
2
3
3
3
3
6
6
4
3
1
3
3 1 0 1 2 1 5 1 0
3
3
3
1 9 1 3 1 2
7
4
4
1
4
2
1 1 1 2 3 2 7 1 6
7
5
5
7
9
9
8
1
3
2
3
12
15
22
26
22
10
2
1
9
13
17
20
24
11
4
I
2
4
11
22
37
44
29
12
j
5
9
17
29
31
17
1
2
9 1 1
3
2
1
9
20
32
36
25
8
4
6
13
18
34
42
29
7
1 1 11
2
8
I4
34
35
20
10
3
6
6
20
34
33
17
6
3
3
3
4
13
19
I6
12
5
4
4
4
10
20
25
13
2
1
1
1
2
2
1
1
3
3
6
3
4
6
3
7
6
6
4
5
3
2
3
1
1
2
2
4
1
3
2
1
1
1
3
2
3
3
2
1
1
2
1
1
3
3
2
3
3
3
2
0
2
4
3
2
1
1
0
0
2
2
0
0
3
2
3
?
0
0
0
0
1
0
1
Figure 2. Analysis of luciferase activity in vaccinia recombinant virus infected cell monolayer. viewed using the Biomedical Image Quantifier. The petri dish is demagnified onto the camera by a factor of 9 5 : l . (a) 20-cm2 petri dish, viewed 1 2 hours post-infection (average of 1000 frames). (b) Digital pixel values for t w o bright spots in (a). & (c)-(f) Direct imaging of small area (1.4 cm2) of infected cells (c) 36 hours post-infection (average of 256 frames). (d) 9 hours post-infection (average of 1000 frames). (e) 6 hours post-infection (average of 1000frames). (f) Zoom display of (e) (top left) showing individual lightemittrng cells.
129
CCD IMAGING
infected with this virus has been previously demonstrated using conventional photographic detection methods at 24 h after initial infection. We have examined the sensitivity and precision of intensified CCD imaging for the analysis of luciferase activity in mammalian cells infected with this vaccinia recombinant at a range of times after infection. The imaging system used for this study was the Biomedical Image Quantifier as supplied by Image Research Ltd, and Photonic Science. For all experiments, monolayers of CV-1 cells (derived from African Green Monkey kidney) were sealed in 20cm2 plastic petridishes. Each dish of 2 x lo6 cells was infected with approximately 500 plaque-forming units of the luciferase-expressing vaccinia virus. Light detection was performed at 6 h, 9 h, 12 h and 36 h post-infection, after removal of the growth medium and addition of luciferase substrate solution (De Wet et al., 1987). Fig. 2(a) shows the image obtained from a 20-cm2 dish of cells infected with recombinant virus for 9 hours. The dish was viewed using a f1.8, 25 mm lens. At this time after infection the virus infection is confined to single cells in which the virus genome is replicating before the onset of new virus particle production. The non-synchronous nature of the infection is indicated by the variation in intensity between individual ‘glowing’ cells. This method of analysis is clearly valuable for visualizing a large area of cells in which low numbers may be expressing the luciferase gene. Direct imaging of a small1 area of the infected cell monolayer (1.4 cm2) with the intensified CCD is shown in Figs 2(c)-2(f). Fig. 2(c) illustrates the sensitivity of this method when applied to recombinant infected cells which have been incubated for 36 hours. Although a few cells can be distinguished, the image is mostly a result of plaque formation, where new virus particles are produced and released to infect surrounding cells. This is the time post-infection when luciferase can be detected using photographic methods. Fig. 2(d) shows cells infected with the recombinant virus after 9 hours. Individual cells are easily detected. Variation in intensity between cells is the result of the non-synchronous virus infection and correlates with the levels of virus DNA replication and luciferase expression within these cells. Selection of individual light-emitting cells from a large population, without the problem of contamination from adjacent infected cells, would be relatively
straightforward using this imaging detection method. Fig. 2(e) illustrates the extreme sensitivity of this method for the early detection of single foci of infection. Luciferase activity was analysed 6 hours after infection. Even at this early time,.when virus DNA replication has only just begun, distinct lightemitting cells could be visualized. The single lightemitting cells are enlarged in Fig. 2(f) using a zoom display function. CONCLUSIONS
The detection of single mammalian cells, infected with recombinant vaccinia virus expressing luciferase, demonstrates that intensified CCD imaging offers a wide range of potential applications to the study of biological systems. We hope to use this method to select virus deletion mutants into which the luciferase gene will be introduced at specific sites to disrupt desired genes. Recombinant viruses, which are likely to constitute as little as 0.1% of the transfection progeny, must be distinguished from wild type and subsequently picked and cloned. A rapid and sensitive method detection system is particularly desirable for selecting mutants of viruses which grow slowly in vitro in tissue culture, such as human cytomegalovirus. This non-invasive imaging technique would enable single recombinant infected cells to be picked and plaque purified in a considerably reduced time. The ability to see single cells at early times of infection would also minimize contamination with wildtype parental virus. Virus deletion mutants generated and selected in this way would provide valuable tools for investigating the biological function of specific virus gene products. Wider application of this approach to molecular and cellular biology can be envisaged. Rapid screening of cells co-transfected with luciferase as a selectable marker gene would allow easy selection of stable eukaryotic cell lines. This offers advantages both in uitro and in uivo for the analysis of gene expression and protein function. REFER ENCES De Wet, J. R., Wood, K. V., DeLuca, M., Helinski, D. R. and Subramani, S. (1987). Cloning and expression of the firefly luciferase gene in mammalian cells. In: Bioluminescence and
130 Chemiluminescence: New Perspectives, Scholmerich, J., Andreesen, R., Kapp, A,, Ernest, M., Woods, W. G., (Eds), John Wiley, Chichester, pp. 369-372. Gonzalez, R. C. and Wintz, P. (1987). DigitaI Image Processing, Addison-Wesley, Reading, MA USA. Hooper, C. E., Ansorge, R. E. (1990). Trends in Analytical Chemisfry (in preparation).
C. E. HOOPER ET AL.
Janesick, J. R. and Blouke, M. (1987). Sky on a chip: the fabulous CCD. S k y and Telescope, 238-242. Rodriguez, J. F., Rodriquez, D., Rodriguez, J. R., McGowan, E. B. and Esteban, M. (1988). Expression of the firefly luciferase gene in vaccinia virus: a highly sensitive gene marker to follow virus dissemination in tissues of infected animals. Proc. Nail. Acad. Sci. U S A . , 85, 1667-1671.
