10 Selak, I., Skaper, S. D. and Varon, S. (1985) J. Neurosci. 5, 23-28 11 Walicke, P., Varon, S. and Manthorpe, M. (1986)£ Neurosci. 6, 1114-1121 12 Gurney, M. E., Heinrich, S. P., Lee, M. R. and Yin, H-S. (1986) Science 234, 566-574 13 Gurney, M. E. (1984) Nature 307, 546-548 14 Gurney, M. E. et al. (1986) Science 234, 574-581 15 Gospodarowicz, D., Neufeld, G. and Schweiger, L. (1987) J. Cell. Physiol. 5, 12-26 16 Slack, J. M. W., Darlington, B. G., Heath, J. K. and Godsave, S. F. (1987) Nature 326, 197-200 17 Kirnelman, D. and Kirshner, M. (1987) Cell 51,869-877 18 Walicke, P., Cowan, W. M., Ueno, N., Baird, A. and Guillemin, R. (1986) Proc. Natl Acad. Sci. USA 83, 30123016 19 Morrison, R. S., Sharma, A., de Vellis, J. and Bradshaw, R. A. (1986) Proc. Natl Acad. Sci. USA 83, 7537-7541 20 Unsicker, K. etal. (1987) Proc. NatlAcad. Sci. USA 84, 54595463
21 Togari, A., Dickens, G., Kuzuya, H. and Guroff, G. (1985) J. Neurosci. 5, 307-316 22 Rydel, R. E. and Greene, L. A. (1987) J. Neurosci. 7, 36393653 23 Abraham, J. A., Whang, J. L., Tumulo, A., Mergia, A. and Fiddes, J. C. (1986) Cold Spring Harbor Symp. Quant. Biol. 51,657-668 24 Gospodarowicz,D. (1987)Methods Enzymol. 147, 106-119 25 Barbin, G., Manthorpe, M. and Varon, S. (1984) J. Neurochem. 43, 1468-1478 26 Varon, S., Nomura,J. and Shooter, E. M. (1967) Biochemistry 6, 2202-2209 27 Manthorpe, M., Skaper, S., Williams, L. R. and Varon, S. (1986) Brain Res. 367,282-286 28 Barde, Y-A., Edgar, D. and Thoenen, H. (1982) EMBO J. 1,549-553 29 Hofer, M. M. and Barde,Y-A. (1988) Nature 331,261-262 30 Chaput, N. et al. (1988) Nature 332,454-455 31 Faik, P., Walker, J. I. H., Redmill, A. A. M. and Morgan, M. J. (1988) Nature 332,455-457 m
Confocalmicroscopy:applicationsin neurobiology A. Fine, W. B. Amos, R. M. Durbin and P. A. M c N a u g h t o n A. Fineis at the Departmentof Physiolo~ and Biophysics,Dalhousie University,Halifax, NovaScotia,Canada, W. B.Arnosisat the Laboratoryof MolecularBiolog7, HillsRoad, Cambridge,UK,R.M. Durbinisat theKing's CollegeResearch Centre,King's College,Cambridge, UKandP.A. McNaughtonisat the Physiological Laboratory, Downing Street, Cambridge, UK.
New methods of confocal microscopypermit the observation at high resolution of structures deep within neural tissue. The resulting images are often striking and informative. Used in conjunction with fluorescent voltage-sensitive or ion.sensitive dyes, confocal microscopy may make possible a wide range of investigations of brain function and plasticity that were difficult or impossible with previous techniques.
Neuroscientists dream of being able to watch the brain work and change. The difficulty in seeing these things is one reason why many important questions about CNS function and plasticity remain unanswered. This difficulty is diminishing with the introduction of new optical methods, many involving fluorescence or reflectance microscopy. Intracellular or membranebound fluorescent dyes, for example, have been used to follow the growth and long-term changes of nerve terminals I-3, and retrogradely transported fluorescent markers have been used to identify living neurones with particular projections for subsequent electrophysiological or structural study 4. Voltageand Ca2+-sensitive indicator dyes have been used to observe patterns of electrical activity 5-I° without the constraints that microelectrodes impose on the size and number of structures that can be monitored. Such voltage-sensitive dyes have already helped to reveal aspects of the functional organization of vertebrate forebrain structures 1I-I 3. Unfortunately, fluorescence and reflectance images are often severely degraded by scattered, emitted or reflected light from tissue structures outside the plane of focus. The usefulness of the new optical methods has been further limited by the poor depth discrimination of ordinary light microscopy. These limitations have been only partly overcome by video image processing 1 4 - 1 6 and deconvolution 17 ,18 techniques. -
increases in resolution by physical means. The principle of confocal microscopy was described by Minsky as early as 195719 and first used successfully by Egger and Petran in 19672° to view unstained neural tissue. The technique uses light from a point source, combining focal illumination of a single point in the specimen with imaging of the illuminated specimen point on a detector pinhole. Light from out-of-focus elements can thereby be virtually eliminated (Fig. IA). There is also a theoretical 1.4fold improvement in resolution in the plane of focus compared with normal microscopy. A confocal optical system can be realized in a number of ways. For example, a pinhole in the incident light path can be made to correspond with a second pinhole in the image plane; a complete image of a specimen at the highly restricted focal depth can then be built up either by scanning the specimen under the beam or, as in Egger and Petran's design 2°, by coordinated (tandem) scanning of the pinholes across the specimen and image planes. Such tandem scanning confocal microscopes can produce extremely high resolution images, scanning fast enough for the full image to be visible 21. But because the apertures exclude such a high proportion of light, only extremely bright light sources and highly reflective or fluorescent specimens can be used, This difficulty was overcome by the use of laser illumination with a mechanically scanned specimen, as pioneered by Brakenhoff and colleagues 22. Mechanical scanning of the specimen, however, imposes limits on the mass and stability of the preparation and on the speed of the scan. These limitations have in turn been largely overcome through the development in Sweden 23 and the UK 24 of beam-steering methods to scan the illuminating spot within the fixed microscope optics.
Principle of confocal microscopy An alternative solution to this problem is offered by confocal microscopy, which achieves dramatic 346
A practical confocal microscope The
© 1988,ElsevierPublications,Cambridge 0378-5912/88/$0200
TINS, Vol. 11, No. 8,1988
Research Council (MRC) Centre in Cambridge is shown schematically in Fig. 1B, together with apparatus for electrical recording and stimulation of neurobiological preparations. The instrument uses a high numerical aperture (NA) objective to focus the 488 nm or 514 nm output of an argon ion laser to a diffraction-limited spot on or in the specimen. An ordinary epifluorescence microscope is used for this purpose, with a semi-silvered mirror (for reflectance) or a dichroic mirror (for fluorescence) serving to introduce the illuminating beam into the microscope optical axis. A pair of microcomputer-controlled galvanometer mirrors beyond the dichroic or semisilvered mirror steer the spot in a raster pattern over the object being viewed; reflected or fluorescent light from the illuminated spot is returned and descanned by the same galvanometer mirrors, passed by the semi-silvered or dichroic mirror, and focused on a pinhole in front of the detector - a low-noise photomultiplier tube. A beam splitter and second photomultiplier permit dual-wavelength fluorescence imaging and ratio measurements. Images are built up from the digitized photomultiplier output at each point in the raster scan by a frame store in the microcomputer. A full 768 x 512 pixel image can be collected in 1 s, and images can be collected more rapidly by reducing the number of pixels. The true magnification of the image can be increased by compressing the raster into a smaller region about the optical axis. Images can be summed, averaged, subtracted from or added to stored images, and a wide range of digital image enhancement methods can be applied, including contrast-stretching, false-color intensity coding, and various image convolutions leading to edgeenhancement or other forms of spatial frequency filtering. By integrating successive images, weakly fluorescent structures can be recorded with a sensitivity comparable to that of a silicon intensified target (SIT) camera. The beam-steering and photodetector apparatus is attached to the microscope in much the same way as an ordinary videa camera. The microscope is also equipped with a routine mercury-arc epifluorescence system so that the observer cam easily switch between direct visual observation of a fluorescence image (useful for rapidly locating areas of interest) and video observation of the confocal image. A light collector can be mounted below the substage condenser, directing transmitted laser light to the photomultiplier tube via a fibre optic conduit. In this way, ordinary brightfield, phase contrast or differential interference contrast images of thin specimens can be collected for direct comparison with confocal fluorescence or reflectance images.
Optical sections of living tissue The dramatic improvement in resolution achieved by the use of confocal microscopy is demonstrated by the fluorescence images of a living kidney glomerulus stained with a fluorescent styryl dye25 in Fig. 2. The figure shows the same field viewed using laser-scanning confocal microscopy (A) and ordinary epifluorescence microscopy (B). The confocal image TINS, Vol. 11, No. 8,1988
:, . B
PLANE OF FOCUS PHOTOI~,ILTIPLtERS
Fig. 1. (A) Principle of confocal microscopy. A specimen point in the focal plane is illuminated (in this illustration, via a dichroic or semi-silvered mirror) with a diffraction-limited spot of li&ht from a laser or from a non-coherent light source (L) and pinhole (the illumination aperture). Emitted or reflected light from the illuminated point retraces the incident light path (solid fine) throu&h the objective lens, passing the dichroic (for fluorescence) or semi-silvered (for reflectance) mirror, and is then imaged on a second pinhole (the detector aperture). The confocal effect depends both on focused illumination of a single point in the specimen, so that the illuminatin& intensity falls off as r-2, and on the use of a pinhole in the ima&e plane to exclude the majority of residual out-ofplane emission (dotted and broken fines). The thickness of the optical section is reduced by the use of a high numerical aperture objective, and can be increased by opening the detector aperture. (B) Schematic diagram of the MRC confocal scanning laser microscope (Bio-Rad Lasersharp), as used for imaging activitydependent optical changes. Beam steering and focus are con~,olled by microprocessors in synchrony with the stimulator. DiEitized photomultipfier output is relayed to a framestore in the computer, and the processed image is displayed on the color monitor and photomonitor. In addition to confocal epifluorescence or reflectance images, scanned phase contrast or differential interference contrast transmission images can be obtained with the fibre-optic light conduit from condenser to photomultiplier. is an 'optical section' through the specimen. The thickness of this section is inversely related to the NA of the objective, and may be as little as 0.5 t~m with a NA 1.4 lens. Useful optical sections can be obtained even when using low magnification, long 347
Fig. 2. Fluorescence imagesof a&tomerulus at the transverselycut surfaceof a livingrat kidney,stainedwitha voltage-sensitive styryldye25.The samefieldof view observedwith(A) confocalopticshasfar greaterdefinition than when(B) ordinary epifluorescenceoptics are used.Spacebaris l O011m.
Fig. 3. (A) Confocal fluorescence image of the photoreceptor layer of the parafoveal region of an excised marmoset retina stained with the styryl pyridinium dye Di24-SP-BS. Outer segments of both rods and cones are clearly visible, though some were broken in the process of removal of the retina from the pigment epithelium. The large cone inner segments and the more slender rod inner segments are brightly stained, probably because of the abundance of mitochondrial membrane in this region of the cell. Photoreceptor cell bodies (dark) and the outer plexiform layer are visible to the right of the picture. The dark lines show the approximate planes of optical section of the views in (B) and (C). Lens x60, NA 1.4, oil immersion. Scale bar is lO#m. (B) Confocal tangential optical section through the same retina as in (A) at the level of the inner segments. The mosaic of rods (small circles) and cones (large circles) is clearly evident in this preparation. Scale bar is l Ol~m. (C) Section parallel to (B), at the level of the photoreceptor cell bodies. Scale bar is 25#m. (D) Section through the ganglion cell layer, viewed from the vitreal surface. Scale bar is 25#m.
Fig. 4. Stereo pair reconstructed from a series of confocal images taken at 0.5#m intervals through a chick retina. The retina was cryosectioned at 601~m, labelled with antibodies to a Ca2+-binding protein (calretinin) and visualized with a fluoresceinconjugated second antibody. The figure shows (from top) the unstained photoreceptor inner segments, the outer synaptic layer, horizontal, bipolar and amacrine cells in the inner nuclear layer, the thick inner synaptic layer and the ganglion cells (bottom). (Preparation kindly provided by Dr John Rogers, Physiological Laboratory, Cambridge.)
Fig. 5. Confocal images of neocortical neurones of an intact rat brain. After craniotomy and reflection of the dura, the fluorescent voltage-sensitive dye RH414 (Ref. 25) was added for30 min to the artificial CSFbathing the exposed cortex, and was then washed off before viewing. (A) Cells close to the pial surface can be seen adjacent to an erythrocyte-filled capillary. Cell bodies appear dark, surrounded by the brightly fluorescent, stained cell membranes and processes within the neuropil. (B) Cell bodies and processes can be seen at least 2501~m below the pial surface.
Fig. 6. (A) Low magnification confocal image of living rat hippocampal dentate gyrus, in a 4001~m slice preparation maintained in vitro, stained with the voltage-sensitive dye RH414. (B) Higher magnification confocal image of dentaJte granule cells and processes from the same field of view as in Fig. 6A. The optical section is 150t~m below the surface of the hippocampa/ slice.
working distance objectives. For example, details of individual cells can be observed up to 0.5 mm below the surface of a fluorescently stained chick embryo using a x6.3, NA 0.2 objective 24. With high magnification, high NA objectives, resolution of 200 nm structures can be achieved within the plane of the optical section, though it is possible to detect considerably smaller isolated sources of fluorescence or reflection, such as fluorescent microspheres or colloidal gold. The reduction of out-of-focus light in the confocal image makes it possible to see details of thick biological specimens with a resolution unobtainable by other means. Fig. 3A shows the photoreceptors of an excised monkey retina stained with voltagesensitive dye; the outer segments, less than 1.5 l~m across, can be seen clearly. With ordinary epifluorescence microscopy these structures are obscured by other outer segments out of the focal plane. When the surface of the retina is viewed from the photoreceptor side, the optical sectioning accomplished by the confocal microscope permits views of the mosaic of rods and cones at various depths (see Fig. 3B and 3C; the approximate planes of section are shown in Fig. 3A). The ganglion cell layer in the same preparation, viewed from the vitreal surface, is shown in Fig. 3D.
Three-dimensional image reconstruction Thin optical sections generated by the confocal microscope can provide an efficient alternative to tedious microtome sectioning or camera lucida reconstructions of three-dimensional neuronal structures. By changing the plane of focus with a microcomputer-controlled stepper motor (see Fig. 1B), serial optical sections can be collected with arbitrary depth increments through the specimen. The stored images contain a full three-dimensional representation of the specimen, and can be combined by a simple algorithm to generate high resolution left and right stereo images of the full three-dimensional structure. The example shown in Fig. 4 is of a thick section of chick retina stained with a fluorescent antibody to a Ca2+-binding protein. The three-dimensional locations of the stained cells within the section can be clearly seen when the two images are brought into coincidence.
Tracing of neuronal pathways The high resolution of thick structures obtained using confocal microscopy makes possible a range of otherwise impractical neurobiological experiments. For example, if a retrogradely transported fluorescent marker is microinjected into rat brain, neurones that project to the injection site can subsequently be seen in the intact cortex, and could be impaled, under visual guidance, with a microelectrode. In the past, intracellular recording from neurones with such defined projections could only be achieved by random impalement and subsequent antidromic stimulation.
Activity-dependent changes in structure Confocal microscopy offers the prospect of moni350
toring structural changes that may occur as a result of electrical activity in identified cells deep within intact cortex. Figure 5A shows a view of the pial surface of an intact rat brain stained by topical application of a fluorescent dye; a blood-filled capillary runs diagonally across the field of view. Figure 5B shows a section 250 t~m beneath the pial surface of the same brain; the network of nerve fibres and the cell bodies can clearly be seen at this depth. Similar images can be obtained from subcortical structures in slice preparations in vitro. Figure 6A shows a low-magnification image of the hippocampal dentate gyrus stained with a voltagesensitive dye. Granule cell bodies within the dentate gyrus appear black, while the fluorescent membranes of fibres appear white. A highermagnification image of the granule cells 150~m below the surface of the slice is shown in Fig. 6B. Fine processes close to the cell bodies can clearly be seen. Identified cell bodies such as those shown in Figs 5 and 6 might be impaled under visual control and injected with a fluorescent dye emitting at a different wavelength. It would then be possible to determine the morphology of the cell and its processes, as well as the presence and morphology of any electrically coupled cells, while stimulating or recording from the injection pipette. In a similar manner it may be possible to investigate structural changes associated with development, aging and such phenomena as long-term potentiation.
Imaging of electrical activity One of the most exciting possibilities opened up by confocal microscopy may be the ability to image the activity of single cells within an intact, functioning brain or tissue slice preparation. Voltage-sensitive fluorescent dyes generate bright and clear images of single neurones using the confocal microscope (see Figs 3, 5 and 6), and in principle two images of the same preparation, one obtained with stimulation and the other without, could be subtracted to reveal the activity of single neurones. In practice the difference signal in experiments to date has been below the signal-to-noise limit of the microscope, but technical improvements, such as the ability to steer the beam so that it scans only areas of interest, may enhance both the temporal resolution and the signal-to-noise ratio to the point where useful images of nervous activity could be obtained. An alternative to using voltage-sensitive dyes to observe electrical activity may be to use indirect monitors of cellular activity such as the intracellular Ca2+- or pH-sensitive dyes developed by Tsien and his colleagues 9'26. The use of optical methods for monitoring electrical activity may open the way to a detailed analysis of the interactions of large numbers of neurons in the complex structures of an intact functioning CNS. The application of confocal microscopy to overcome problems of resolution in depth should lead to significant advances in our understanding of the function of brain and other complex tissues. At the same time, confocal microscopy promises to produce images of the living, functioning brain of a sort never TINS, Vol. 11, No. 8,1988
seen before- images that delight us as much for their beauty as for their informative power.
Selected references 10'Rourke, N. A. and Fraser, S. E. (1986) Dev. BioL 114, 265276 2 Purves, D. and Voyvodic, J. T. (1987) Trends NeuroscL 10, 398-404 3 Lichtmann, J. W., Magrassi, L. and Purves, D. (1987) J. Neurosci. 7, 1215-1222 4 Katz, L. C. (1987) J. Neurosci. 7, 1223-1249 5 Cohen, L. B. and Salzberg, B.M. (1978) Rev. Physiol. Biochem. Pharm. 83, 85-88 6 CJrinvald, A., Fine, A., Father, i. C. and Hildesheim, R. (1983) Biophys. J. 42, 195-198
7 Grinvald,A. (1985) Annu. Rev. Neurosci. 8, 263-305 8 London, J. A., Zecevic, D., Loew, L. M., Orbach, H. S. and Cohen, L. B. (1986) in Applications of Fluorescence in the Biomedical Sciences (Taylor, D. L. et al., eds), pp, 423-437, Alan R. Liss 9 Williams, D. A., Fogarty, K.E., Tsien, R.Y. and Fay, F.S. (1985) Nature 318, 558-561 10 Ross, W. N., Lewenstein-Stockbridge, L. and Stockbridge,
N. L. (1986)J. Neurosci. 6, 1148-1159 11 Blasdel, G. G. and Salama, G. (1986) Nature 321,579-585 12 Grinvald, A., Lieke, E., Frosti, R. D., Gilbert, C. D. and Wiesel, T. N. (1986) Nature 324, 361-364 13 Kauer,J. S. (1988) Nature 331, 166-168 14 Breuer, A. C., Allen, R. D. and Lewis, L. J. (1981) Neurology 31, 118 15 Allen, R. D. (1985) Annu. Rev. Biophys. Chem. 14, 265-290 16 Inou6, S. (1986) Video Microscopy, Plenum Press 17 Agard, D. A. and Sedat, J. W. (1983) Nature 302,676-681 18 Hiraoka, Y., Sedat, J.W. and Agard, D.A. (1987) Science 238, 36-41 19 Minsky, M. (1957) US Patent No. 3013467 20 Egger, M. D. and Petran, M. (1967) Science 157, 305-307 21 Boyde,A. (1985) Science 230, 1270-1272 22 Brakenhoff, G. J., van der Voort, H. T. M., van Spronsen, E. A., Linnemans, W. A. M. and Nanninga, N. (1985) Nature 317,748--749 23 Carlsson, K. etal. (1985) Opt. Left. 10, 53-55 24 White, J. G., Amos, W. B. and Fordham, M. (1987) J. Cell Biol. 105, 41-48 25 Grinvald, A., Anglister, L., Freeman, J. R., Hildesheim, R. and Manker, A. (1984) Nature 308, 848-850 26 Tsien, R. Y. (1981) Nature 290, 527-528
". . . . . . . . .
Visual course control in flies relies on neuronal computation of object and backgroundmotion Martin Egelhaaf, Klaus Hausen, Werner Reichardt, and Christian Wehrhahn
The spatial distribution of light intensity received by the eyes changes continually when an animal moves around in its environment. These retinal activity patterns contain a wealth of information on the structure of the environment, the direction and speed of selfmotion, and on the independent motion of objectsl"e. I f evaluated properly by the nervous system this information can be used in visual orientation. In a combination of both behavioural and electrophysiological analysis and modelling, this article establishes the neural mechanisms by which the visual system of the fly evaluates two types of basic retinal motion patterns: coherent retinal large-field motion as induced by selfmotion of the animal, and relative motion between objects and their background. Separate neuronal networks are specifically tuned to each of these motion patterns and make use of them in two different visual orientation tasks. Visual orientation greatly relies on the evaluation of the global visual motion patterns received by the eyes when an animal moves around. These motion patterns depend in a characteristic way on the trajectory described by the moving animal as well as on the particular three-dimensional structure of the visual environment z'2. Consider, for instance, two simple commonplace situations. In the first, a moving animal unintentionally deviates from its course. This results in a displacement of the entire visual scene, which contains a strong rotational component. When this rotational component is extracted from the retinal motion pattern, it can be used to control the compensatory optomotor responses of the animal. In this way, the course may be stabilized against internal and external disturbances. A different situation is TINS, Vol. 11, No. 8,1988
encountered when the animal passes nearby objects located in front of a more distant background. The retinal images of these objects and their background then move relative to each other leading to discontinuities in the motion field. This relative motion may indicate the existence of nearby stationary or moving objects. This information can be used to discriminate objects from their background and might serve as the basic cue in various visual orientation tasks, such as fixation of stationary objects or pursuit of moving targets. These types of global retinal motion patterns do not only occur when the animal moves around bodily. Similar motion patterns may also arise during head and eye movements. This review concentrates on recent studies on the visual system of the fly, which has proved, during the past few decades, to be a suitable model system for the elucidation of the neuronal computations underlying various behavioural motion-dependent tasks 3'4. We analyse the basic mechanisms by which the nervous system of the fly processes coherent largefield motion, and relative motion between objects and background, and how these motion patterns are exploited in mediating optomotor course stabilization and object-induced orientation. Whether related mechanisms play a role in evaluating global retinal motion patterns in other species has yet to be established, although it is not unreasonable to expect that this will be the case. This has already been shown for the mechanisms underlying other motion information processing tasks. The basic mechanism of local movement detection for instance (see below), which was initially discovered in the insect visual system, was later also found in vertebrates, including
Martin Egelhaaf, KlausHausen, WernerReichardtand ChristianWehrhahn areat the MaxPlanck-lnsb'tutfiJr biologische Kybemetik, 5pemannstrasse38, D-7400 TObingen, FRG.
© 1988,ElsevierPublications.Cambridge 0378-5912/88/$0200