The receptive fields of cat retinal ganglion cells in physiological and pathological states: where we are after half a century of research

Transcription

1 Progress in Retinal and Eye Research 21 (2002) The receptive fields of cat retinal ganglion cells in physiological and pathological states: where we are after half a century of research J.B. Troy a,b, T. Shou c,d, * a Department of Biomedical Engineering & Neuroscience Institute, Northwestern University, Evanston, IL, USA b Laboratory of Neuropsychology, Section on Neural Coding & Computation, National Institute of Mental Health, Bethesda, MD, USA c Center for Brain Science Research, School of Life Sciences, Fudan University, Shanghai, People s Republic of China d Laboratory of Visual Information Processing, Chinese Academy of Sciences, Beijing, People s Republic of China Abstract Studies on the receptive field properties of cat retinal ganglion cells over the past half-century are reviewed within the context of the role played by the receptive field in visual information processing. Emphasis is placed on the work conducted within the past 20 years, but a summary of key contributions from the 1950s to 1970s is provided. We have sought to reviewaspects of the ganglion cell receptive field that have not been featured prominently in previous reviewarticles. Our reviewof the receptive field properties of X- and Y-cells focuses on quantitative studies and includes consideration of the function of the receptive field in visual signal processing. We discuss the non-classical as well as the classical receptive field. Attention is also given to the receptive field properties of the less well-studied cat ganglion cellsfthe W-cellsFand the effect of pathology on cat ganglion cell properties. Although work from our laboratories is highlighted, we hope that we have given a reasonably balanced view of the current state of the field. r 2002 Elsevier Science Ltd. All rights reserved. CONTENTS Abstract Introduction The 1950s The 1960s The 1970s Progress since Retinal ganglion cell-types Axonal projections X- and Y-cell axonal projections W-cell axonal projections Retinal ganglion cell morphologies The sluggish-sustained cell The sluggish-transient cell The color-coded cell The ON OFF cell The suppressed-by-contrast cell The ON OFF directionally selective cell The ON directionally selective cell Morphological types needing physiological counterparts Receptive field properties of X- and Y-cells The classical receptive field The non-classical receptive field *Corresponding author. Center for Brain Science Research, School of Life Sciences, Fudan University, Shanghai, People s Republic of China. Tel.: ; fax: address: tdshou@fudan.edu.cn (T. Shou) /02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S (02)

2 264 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) Receptive field properties of W-cells The sluggish-sustained cell The sluggish-transient cell The color-coded cell The ON OFF cell The suppressed-by-contrast cell The ON OFF directionally selective cell The ON directionally selective cell Receptive field properties under pathological conditions Effect of an abnormal visual environment during development Effect of elevated intraocular pressure Effect of reduced oxygen tension Future directions Acknowledgements References Introduction Sherrington is credited with coining the term receptive field, using it to refer to the region of a sensory surface from which a stimulus could evoke a response from neurons (Sherrington, 1906). The term was first applied to a group of neurons, but today it is reserved for the sensory region where a stimulus evokes a response from a single cell. In the visual system, one can consider the retina to be the sensory surface where receptive fields are located, but more often it is treated as the optical projection of the retina out into the world that is employed, the construct we call the visual field. Hence, for retinal ganglion cells, the receptive field could be either a small region of the retina or a small region of the visual field. Since the receptive field incorporates in the latter case an additional component due to optical blurring, which increases with descending light level as the pupil dilates, the two usages are not fully equivalent. In the first half of the 20th century, four of the most prominent people in the field of sensory physiology were Sherrington, two of his students, Adrian and Granit, and Hartline, all of whom became Nobel laureates. Adrian and Hartline, among other accomplishments, helped to establish the receptive field as the principal concept of sensory physiology. From Granit came the important insight that inhibition is a key component of the sensory response, and hence of the receptive field, as is excitation (Granit, 1943). Granit was also the first person to record from single cat retinal ganglion cells (Fig. 1). The modern era of sensory physiology has been dominated by the detailed study of receptive fields. Fifty years ago Stephen Kuffler (1952, 1953) initiated the division of the receptive fields of cat retinal ganglion cells into OFF-center/ON-surround and ON-center/ OFF-surround types (Fig. 2). This study might reasonably be considered the start of the modern era of sensory physiology. It forms a bridge between the work of the four giants of sensory physiology of the first half Fig. 1. The first recordings of cat ganglion cell discharges from Granit (1943). Case 1 shows the discharge of a ganglion cell in response to a diffuse light stimulus of 620 nm. Case 2 shows the ganglion cell s response to a diffuse light of the same wavelength but of lower intensity. Case 3 is the ganglion cell s response to a diffuse light of 460 nm and case 4 the response to a diffuse light of 500 nm. From Granit (1943). Reproduced by permission of Scandinavian Physiological Society. of the 20th century, mentioned above, and the work of two giants of the second half, Kuffler s own students Nobel laureates, David Hubel and Torsten Wiesel. Kuffler s description of the retinal ganglion cell receptive field has endured very well and continues to form the foundation of all textbook descriptions of ganglion cell functional properties. However, it has been known for a number of decades that this description fails to capture the full complexity of the ganglion cell receptive field. In particular, it has been demonstrated that the cat, like all vertebrates, has considerably more than two retinal ganglion cell types, and that some of these do not have a center surround receptive field. Furthermore, for even those ganglion cells with center surround receptive fields, the description provided by Kuffler is incomplete. We knownowthat stimulation of the retina with visual patterns extending beyond the region classically identified by Kuffler as the receptive

3 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) time, howthe receptive field properties of cat ganglion cells are affected by pathological conditions. We begin with a brief summary of the landmark contributions from the 1950s, 1960s and 1970s. For those interested in a more comprehensive coverage of this work and that which preceded it, the following are recommended as supplemental resources: Rodieck (1973), Stone (1983) and Enroth-Cugell (1993) The 1950s Fig. 2. An ON-center/OFF-surround receptive field from Kuffler (1953). The cell s discharge rate also increases in response to increments of light presented to the central region (crosses). Its discharge rate increases in response to decrements of light presented to the peripheral region (open circles). In the intervening zone, the cell increases its firing rate both for light increments and decrements. The location of the recording electrode is shown. Based on center size and the ON OFF responses resulting from stimulation in the intervening zone, one supposes that this was an Y-cell. Reproduced by permission of: The American Physiological Society. field can affect the discharge of a ganglion cell. But these deficiencies in Kuffler s description of the receptive field should not diminish one s appreciation of his substantial contribution. He laid down the basis for all subsequent descriptions of ganglion cell receptive fields. Over the past 50 years, detailed investigation of ganglion cell receptive fields has been carried out in a number of species. Here we choose to focus on the receptive field properties of cat retinal ganglion cells and emphasize results gathered during the past 20 years. The receptive fields of cat ganglion cells remain the most thoroughly investigated and continue to serve as the yardstick against which receptive field properties of ganglion cells from other species are compared. In addition, most of the properties reported for the receptive fields of cat retinal ganglion cells have been found later for ganglion cells of other species. It is little surprising then that when modeling of the retina s contribution to visual information processing is undertaken it often rests heavily on descriptions of the receptive field properties of cat retinal ganglion cells. There have been many fine reviews of ganglion cell properties over the past quarter-century or so (e.g., Levick, 1975; Levick and Thibos, 1983; Shapley and Perry, 1986; Meister and Berry, 1999), and, rather than duplicate these contributions to the literature, we emphasize here recent work, mostly not covered in these earlier reviews. By recent work we mean that published since We also review, for maybe the first Besides Kuffler s initial ground-breaking description of the ganglion cell receptive field, there were other notable contributions to our understanding of cat ganglion cell properties in the 1950s. From Kuffler s laboratory came the first investigation of howthe receptive field properties of ganglion cells are affected by adaptation to mean light level (Barlowet al., 1957a, b). The same laboratory also gave us the first characterization of the statistical properties of the maintained discharge of cat retinal ganglion cells (Kuffler et al., 1957). Such a description is important when one seeks to discriminate the change in firing resulting from a visual stimulus from changes in discharge that occur randomly even when no stimulus is present. In other words, looking from the perspective of the brain, what is it in a ganglion cell s discharge that justifies registering a visual stimulus as being present? In Kuffler s laboratory, FitzHugh (1957, 1958) tackled this question with considerable mathematical sophistication, providing an early strong contribution to the still ongoing debate about what constitutes the neural code. It was also in Kuffler s laboratory that the first intracellular recordings from cat retinal ganglion cells were made (Fig. 3, Brown and Wiesel, 1959; Wiesel, 1959). Outside Kuffler s laboratory, Bornschein (1958) had started to investigate the effect of ischemia on retinal ganglion cell discharge, and studies employing electrical stimulation were adding support to Bishop and O Leary s (1942) observation that the cat optic nerve Fig. 3. Intracellular recording from a cat ganglion cell located in the area centralis from Brown and Wiesel (1959). Steady membrane potential is 60 mv. The action potentials have been truncated. Note that the stimulus, which was a 3 mm spot 32 times brighter than the background, generates a depolarization of only a fewmv. Excluding the transient excursions of the action potential, the operational range for the ganglion cell membrane potential is surprisingly small (see also Fig. 11). Reproduced by permission of: The Physiological Society.

4 266 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) contains groups of axons that can be partitioned according to different conduction velocity groups (Bishop et al., 1953; Bishop and Clare, 1955; Granit, 1955; Chang, 1956; Lennox, 1957). This latter work made it seem plausible that the message carried by retinal ganglion cells to the brain might be partitioned between fast and slowly transmitted components, much as the message carried by peripheral nerves to the central nervous system is partitioned (e.g., Gasser and Erlanger, 1929). Erlanger and Gasser were awarded a Nobel prize for their work on peripheral nerve The 1960s Fig. 4. X-cells sum light linearly within their receptive fields, while Y-cells do not (from Enroth-Cugell and Robson, 1966). Panel A shows the responses of an OFF-center X-cell to the introduction and withdrawal of a sinusoidal grating, positioned in four phases relative to a diameter of the receptive field center (right-hand cartoons). Two phases, 901 and 2701, can be found for which the X-cell evokes no response. This is the expectation for a receptive field that sums light linearly, since the integral of light over the receptive field center is equal to the mean light level for these two phases. No phases exist for the OFF-center Y-cell of panel B where this cell fails to signal the presence of a stimulus. Reproduced by permission of: The Physiological Society. Clearer evidence that there were different functional classes of retinal ganglion cell in cat was forthcoming in the 1960s. Enroth-Cugell and Robson (1966) demonstrated that among the ganglion cells with center surround receptive fields there existed two varieties that they termed X- and Y-cells. Light was found to sum linearly within the receptive fields of X-cells to evoke their responses, while, for Y-cells, the summation of light was non-linear (Fig. 4). The first reports of cat ganglion cells with other unusual receptive field properties also appeared in the 1960s (Stone and Fabian, 1966; Rodieck, 1967a). These cells had receptive fields lacking Kuffler s center surround organization. Other significant contributions to our understanding of cat retinal ganglion cells from the 1960s were as follows: * The first recordings of optic tract fibers in behaving cats (Hubel, 1960). * The introduction of a quantitative model of the center surround ganglion cell receptive field as the difference of two Gaussian functions (Fig. 5; Rodieck, 1965; Enroth-Cugell and Robson, 1966). * The application of sinusoidal grating patterns to the study of ganglion cell receptive fields (Fig. 5; Enroth- Cugell and Robson, 1966). * The first report of the simultaneous recording of more than one retinal ganglion cell (Gestri et al., 1966). * The employment of pharmacological agents to investigate ganglion cell properties (Bornschein and Heiss, 1966; Straschill, 1968). * The first evidence that the discharges of groups of cat retinal ganglion cells could be correlated (Rodieck, 1967b). * Evidence that the mechanism for light adaptation of center-evoked responses integrates over a field equal in size to the ganglion cell s receptive field center (Cleland and Enroth-Cugell, 1968). * Evidence that light adaptation in cat retinal ganglion cells obeys Weber s Law(Sakmann and Creutzfeldt, 1969). * Evidence that signals from the center and surround of the receptive field are combined additively (Maffei and Cervetto, 1968). * Evidence that the discharges of ganglion cells can be influenced by visual patterns moved across parts of the visual field not considered a part of Kuffler s classical receptive field (Fig. 6; McIlwain, 1964, 1966; Levick et al., 1965). * The classical description of howthe maintained discharge of cat retinal ganglion cells changes with mean light level (Barlowand Levick, 1969b), and hownoise in this discharge affects signal detection (Barlowand Levick, 1969a). Thus, by the end of the 1960s, studies of receptive fields had moved firmly into the quantitative realm. The retina seemed nowto represent the visual scene in a more complex fashion than one might have expected from Kuffler s early study. There were now more ganglion cell types with which to contend. Their receptive fields were more varied and not as circumscribed as Kuffler had presented them. The existence of correlations in the firing of ganglion cells raised the possibility that there may be some redundancy in their messages. Alternatively, it might be that what we probe when we look only at the discharge of a single ganglion cell is meaningless to the brain, since the brain has access to and perhaps makes sense from the related discharges

5 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) Fig. 5. Rodieck s (1965) difference of Gaussians receptive field model (panel C). The center surround receptive field is modeled as resulting from two components. Panel B shows that the center mechanism sums light with a Gaussian weighting that has a narrow spatial extent and the surround mechanism, which is concentric and overlapping the center, sums light with a Gaussian weighting of broader spatial extent (i.e., larger standard deviation). Rodieck (1965) modeled the temporal response of the cell to a step of light as shown in panel A. He assumed that the form of the temporal response was invariant with the spatial nature of the visual stimulus (i.e., space and time were separable). We now know that this is not the case (e.g., Enroth-Cugell et al., 1983; Troy, 1983b; Dawis et al., 1984). Panel D shows fits of the difference of Gaussians model to measurements of contrast sensitivity as a function of spatial frequency for an ON-center X-cell from Enroth-Cugell and Robson (1966). Measurements were made for vertical gratings moving from left to right in the cat s visual field (open circles) and for horizontal gratings moving from above downwards (filled circles). The same model was used to fit both sets of data (i.e., the receptive field is assumed to have circular symmetry). For more details on the relationship between the difference of Gaussians receptive field model represented in spatial and spatial frequency domains, the reader is referred to Enroth- Cugell and Robson (1966). Panel D reproduced by permission of: The Physiological Society. Fig. 6. Ganglion cells respond to stimulation outside Kuffler s classical receptive field. An OFF-center ganglion cell, with a very low resting discharge, fires extra action potentials when a 51 disk is moved 81 from the receptive field center. Panel A is for movement of the disk away from the area centralis and panel B for movement towards it. From McIlwain (1964). Reproduced by permission of: The American Physiological Society.

6 268 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) Fig. 7. One of the most innovative proposals for neural coding by ganglion cells of recent years is Meister s (1996) suggestion that the visual signal they carry is multiplexed across the discharges of an ensemble of them. Panel A shows the cross-correlation between the discharges recorded from two salamander ganglion cells in darkness. Clearly, many spikes discharged by cell 1 are paired to within 725 ms with a spike from cell 2. The closer positioned are cell pairs, the higher the degree of correlation (panel B). If the brain were to monitor the correlated output of cells 1 and 2, by predicating a spike in a neuron fed by cells 1 and 2 on the arrival of two spikes within 25 ms, the receptive field center of this hypothetical higher-level neuron could occupy the region demarcated by the heavy circle in panel C. The finer lined ellipses are the receptive field centers of cells 1 and 2. In this case the message encoded by the hypothetical higher-level neuron is buried (multiplexed) in the discharges of cells 1 and 2. The spatial regions of the receptive field centers of cells 1 and 2, not covered by the heavy circle, could then be the receptive field centers of two other hypothetical higher-level neurons whose discharges mirror, respectively, the discharges of cell 1 and cell 2 minus their synchronous discharges. In this way visual stimulation over three regions of space are encoded in the discharges of just two ganglion cells. Extrapolating this coding scheme to the population of all ganglion cells, it is clear that a neural image with a higher spatial resolution than one might expect simply from the number of optic nerve fibers could be transmitted to the brain. Whether or not ganglion cells encode neural images in this manner remains to be determined. From Meister (1996). r 1996 National Academy of Sciences, U.S.A. of the full cohort of ganglion cells (e.g., Fig. 7; Meister, 1996) The 1970s One of the main points of focus for ganglion cell studies in the 1970s was the characterization of the different cell-types (e.g., Stone and Hoffman, 1972; Cleland and Levick, 1974a, b; Fukuda and Stone, 1974; Stone and Fukuda, 1974). There was interest in how many varieties of retinal ganglion cell there were, to where in the brain these cells transmit their information (e.g., Cleland et al., 1971) and what this information might be. Although the discussion about howto subtype cells became somewhat contentious (Rowe and Stone, 1977; Hughes, 1979; Rodieck and Brening, 1983), by the 1980s certain cellular characteristics had reached axiomatic status in the art of ganglion cell typing. It was by then generally accepted that, for a cell-type to be considered distinct, the following criteria should be satisfied: * All ganglion cells of a particular type must have physiological properties that render them distinct from all other ganglion cells. * All cells of one type have a similar morphology that is distinctive, including the pattern of stratification of their dendritic arbors within the retina s inner plexiform layer. * The projection of the axons of one cell-type followa standard pattern (e.g., they all terminate in the A laminae of the dorsal lateral geniculate nucleus). * The array of cells of one type form a pseudo-regular and well-organized mosaic when viewed in the plane of the retina. In spite of the emergence of chemical markers for specific cell-types over the past years (e.g., White and Chalupa, 1991; Hutsler et al., 1993), these four criteria remain the principal means of sub-classifying retinal ganglion cells. A major boost to the study of ganglion cell-types was provided by Boycott and W.assle s (1974) investigation of the morphology of cat retinal ganglion cells in the whole-mount preparation. In this classic study they were able to differentiate four morphological varieties, which they termed a-, b-, g- and d-cells (Fig. 8). It was postulated immediately that the a-cells corresponded to Enroth-Cugell and Robson s (1966) physiological Y-cells and that the b-cells corresponded to their X-cells. Both hypotheses have since been proven correct (Cleland et al., 1975; Peichl and W.assle, 1981; Saito, 1983a, b; Stanford and Sherman, 1984; Fukuda et al., 1984, 1985; Stanford, 1987a, b). Boycott and W.assle s (1974) elegant study had strengthened the case for distinct retinal ganglion cell classes greatly and laid the ground-work for the anatomical investigation of ganglion cell diversity. Avoiding the sampling bias of microelectrode recording and the limitation on cell sample number attainable through an electrophysiological experiment, anatomy seemed to provide a means to determine precisely howmany there were of a particular ganglion cell-type, where in the brain that the cell-type sent its information and howthe representatives of that cell-type were distributed across the retina. Early success was made in addressing some of these questions for Y-cells (e.g., Fig. 10, W.assle et al., 1975). However, for most ganglion cell-types these questions remain unresolved even today in spite of considerable study.

7 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) Fig. 9. Hochstein and Shapley s (1976b) spatial model for the receptive field of a Y-cell. Like Rodieck (1965) the center and surround mechanisms are represented as summing light with Gaussian weighting. The nonlinear component of Y-cell responses results from the summed activity of a set of nonlinear subunits that individually sum light over a region smaller in area than the linear center. The nonlinear subunits are assumed to overlap the classical receptive field and extend beyond this region. Reproduced by permission of: The Physiological Society. Fig. 8. Morphologies of three a-cells (A, B and C) and three b-cells (D, E and F) from Boycott and W.assle (1974). Eccentricities: (A) 1.2 mm; (B) 8.3 mm; (C) B10.0 mm; (D) 1.2 mm; (E) 2.9 mm; (F) 10.0 mm. Reproduced by permission of: The Physiological Society. Other notable contributions from the 1970s were as follows: * Evidence that a single photon can evoke a measurable signal from a cat retinal ganglion cell (Barlow et al., 1971). * The first simultaneous recordings of functionally connected retinal ganglion cells and their target neurons in the brain (Cleland et al., 1971; Levick et al., 1972; Dubin and Cleland, 1977). * Evidence that stimulation outside Kuffler s classical receptive field can evoke large responses from ganglion cells (Kr.uger and Fischer, 1973). * A comprehensive investigation of the adaptational properties of cat retinal ganglion cells (Enroth-Cugell and Shapley, 1973a, b; Enroth-Cugell et al., 1977a, b). * The use of Fourier analysis to decompose ganglion cell discharges into its frequency components (Hochstein and Shapley, 1976a, b). Fig. 10. The retinal distribution of aðyþ-cells from W.assle et al. (1975). Solid lines are of isodensity. r 1975 Wiley-Liss, Inc. Reproduced with permission of Wiley Periodicals Inc. * A quantitative non-linear model for the Y-cell s receptive field (Fig. 9; Hochstein and Shapley, 1976b). * Evidence that retinal ganglion cell responses are influenced by the contrast in a visual scene (Victor et al., 1977; Shapley and Victor, 1978, 1979a, b). * The application of non-linear system identification to the study of retinal ganglion cell receptive fields (Victor et al., 1977; Shapley and Victor, 1978, 1979a, b). * The use of pharmacological agents to tease apart components of the receptive field of cat retinal ganglion cells (Kirby and Enroth-Cugell, 1976). * The first isodensity map of a retinal ganglion cell type (Fig. 10; W.assle et al., 1975). * Evidence that the dendrites of OFF- and ON-center ganglion cells stratify within different sublaminae of

8 270 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) Fig. 11. The dendrites of OFF-center ganglion cells stratify within lamina a of the inner plexiform layer and the dendrites of ON-center ganglion cells stratify within lamina b: Lamina a is closer to the photoreceptors and lamina b is closer to the vitreous. Panels A, B and C showcells whose membrane potential is hyperpolarized by light. Their dendrites all stratify in lamina a: Panels D and E showcells whose membrane potential is depolarized by light. Their dendrites stratify in lamina b: From Nelson et al. (1978). Reproduced by permission of: The American Physiological Society. the retina s inner plexiform layer (Fig. 11; Famiglietti and Kolb, 1976; Nelson et al., 1978). 2. Progress since 1980 Over the period from 1980 until the present day we have seen a gradual increase in the depth of our knowledge of ganglion cell receptive field properties. It might be said that we have moved from the general descriptions that dominated early studies to more complete descriptions. We are edging towards a precise understanding of the role of ganglion cells in vision. To those outside the field, the areas of investigation may at times seem arcane. We hope that in the following sections of this review we will be able to help redress this assessment by presenting some of the results of recent studies within the framework of visual information processing Retinal ganglion cell-types By the early 1980s it was known that the adult cat retina contained between 150,000 and 200,000 ganglion cells (Hughes and W.assle, 1976; Hughes, 1981; Williams and Chalupa, 1983). It was also clear by then, as noted earlier, that the cat contained a number of ganglion cell-types other than the X- and Y-cells (also called brisk-sustained and brisk-transient cells, respectively, by Cleland and Levick, 1974a). In spite of the fact that these cells that were neither X- nor Y-cells have diverse physiological properties (Cleland and Levick, 1974a, b; Stone and Fukuda, 1974), the term W-cell (Stone and Hoffmann, 1972) has become somewhat unfortunately the standard way of referring to them, suggesting a commonality of function that is unjustified. It is now estimated that X-cells constitute 35 55%, Y-cells 3 5% and W-cells 45 60% of the cat s ganglion cell complement (W.assle et al., 1975; Hughes and W.assle, 1976; Stone, 1978; Hughes, 1981; W.assle et al., 1981a, b; Williams and Chalupa, 1983; Fukuda et al., 1985; Leventhal et al., 1985; Stein et al., 1996). Good evidence exists for at least seven types of W- cell. Three of the seven have center surround receptive fields. One of these will be referred to as the color-coded cell. Its receptive field has a spectrally opponent ONcenter/OFF-surround arrangement, with the center favoring short-wavelength (blue) light and the surround favoring middle-wavelength (green) light (Rowe and Cox, 1993). A number of investigators have reported encountering this cell-type (Cleland and Levick, 1974b; Rowe and Stone, 1976; Wienrich and Zrenner, 1984;

9 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) Rowe and Cox, 1993). The other two W-cell types with center surround receptive fields come with both OFFand ON-center sub-varieties. These two were named sluggish-sustained and sluggish-transient cells by Cleland and Levick (1974a) and we will use this terminology for them. They were named tonic and phasic W-cells, respectively, by Stone and Fukuda (1974) and Enroth- Cugell et al. (1983) and Troy et al. (1995) have called the sluggish-sustained variety Q-cells. Many investigators have recorded from sluggish-sustained cells (Cleland and Levick, 1974a; Stone and Fukuda, 1974; Rowe and Stone, 1976; Enroth-Cugell et al., 1983; Thibos and Levick, 1983a, b; Fukuda et al., 1984; Stanford and Sherman, 1984; Stanford, 1987a,b; Rowe and Cox, 1993; Troy et al., 1995). There are also numerous reports of recordings from the sluggish-transient cells (Stone and Hoffmann, 1972; Cleland and Levick, 1974a; Stone and Fukuda, 1974; Rowe and Stone, 1976; Levick and Thibos, 1983a; Stanford, 1987a,b; Rowe and Cox, 1993; Rowe and Palmer, 1995). Two W-cell types have been found with receptive fields that exhibit strong directional selectivity. While both cell-types prefer motion of a stimulus across the receptive field in one particular direction, their responses differ to light flashed within the central region of the receptive field. One responds with a burst of action potentials to light flashed in this region and we will refer to this cell as the ON directionally selective cell. The other responds with a burst of action potentials both to the onset and the offset of light in the receptive field center. These cells will be referred to as ON OFF directionally selective cells. Many investigators have encountered the ON OFF directionally selective cells (Stone and Fabian, 1966; Stone and Hoffmann, 1972; Hoffmann, 1973; Cleland and Levick, 1974a; Stone and Fukuda, 1974; Rowe and Stone, 1976; Rowe and Cox, 1993; Rowe and Palmer, 1995). There are not so many reports of the ON directionally selective cells (Stone and Fukuda, 1974; Hoffmann and Stone, 1985; Rowe and Palmer, 1995), but enough to consider their existence beyond doubt. There are two other varieties of the W-cells that respond to both the presentation and removal of light from their receptive fields. One of these provides a burst of action potentials to both phases. We will call this the ON OFF cell. It has been variously termed excited-by-contrast cell (Stone and Hoffmann, 1972; Hoffmann, 1973), local edge detector (Cleland and Levick, 1974a), ON OFF phasic W-cell (Stone and Fukuda, 1974), and impressed-by-contrast cell (Troy et al., 1989). The other variety is known as the suppressed-by-contrast cell. This cell has a healthy resting rate of action potential discharge that is interrupted when a visual stimulus is presented to or withdrawn from its receptive field. There have been many investigators who have reported encountering both the ON OFF cell (Stone and Hoffmann, 1972; Hoffmann, 1973; Cleland and Levick, 1974b; Stone and Fukuda, 1974; Rowe and Stone, 1976; Stein et al., 1983; Fukuda et al., 1984, 1985; Stanford, 1987a,b; Troy et al., 1989; Rowe and Cox, 1993; Rowe and Palmer, 1995) and the suppressed-by-contrast cell (Rodieck, 1967a; Stone and Hoffmann, 1972; Hoffmann, 1973; Cleland and Levick, 1974b; Stone and Fukuda, 1974; Rowe and Stone, 1976; Stein et al., 1983; Hoffmann and Stone, 1985; Mastronarde, 1985; Troy et al., 1989; Rowe and Cox, 1993). To many the array of W-cell types represents a confusing mass of detail. It is not surprising therefore that some investigators have attempted to simplify our description of W-cells by grouping sub-varieties. So far this effort has met with limited success. While all W-cells have smaller caliber axons than X- or Y-cells and, hence, slower rate of action potential conduction, it was found that one-third of W-cell bodies are as large as those of X-cells. Leventhal et al. (1985) exploited this difference to divide W-cells into two groups. W-cells with soma diameters of 18 mm or greater were lumped into a group called g 1 cells, and W-cells with somata smaller than 18 mm in diameter into another group called g 2 cells. One-third of W-cells fell in the g 1 cell group and twothirds into the g 2 cell group. Others have separated W-cells, based on physiological properties, into tonic and phasic sub-types (e.g., Stone and Hoffmann, 1972; Fukuda and Stone, 1974; Rowe and Palmer, 1995), with the tonic W-cells being the color-coded cell, the sluggishsustained cell and the suppressed-by-contrast cell, and the phasic cells being the sluggish-transient cell, the ON OFF cell and the two varieties of directionally selective cell. There is a tendency for g 1 cells to equate with the tonic W-cells and g 2 cells with the phasic W-cells, and for the former to send their axons primarily to the lateral geniculate nucleus with the latter sending theirs primarily to the superior colliculus, lending support to the idea of grouping W-cells in the manner proposed. However, there are numerous results that argue against so simplistic a form of W-cell grouping. The prevailing dogma among those who study cat ganglion cells is that each W-cell variety serves a distinct function in visual information processing, with its message segregated from other ganglion cell varieties for independent processing within the brain Axonal projections The axons of cat retinal ganglion cells are known to project to the lateral geniculate nucleus, the superior colliculus, the pretectum, the suprachiasmatic nucleus and the nuclei of the accessory optic system, with some cells innervating more than one site. It is believed that it is the projection through the lateral geniculate nucleus that underlies what most of us would consider visual perception, while the projections to the other sites serve

10 272 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) basic functions like the control of pupil size, retinal image stabilization and photic input to the circadian clock. The lateral geniculate nucleus has a dorsal and a ventral portion. The dorsal lateral geniculate nucleus has three parts. Its largest component is a clearly laminated structure that has six layers, nowcalled A, A 1, C, C 1, C 2 and C 3 (Guillery, 1970). Three layers are visible in Nissl stained tissue, the two A laminae and a third ventral lamina that used to be called lamina B. This lamina B was shown subsequently to comprise four distinct layers that have been renamed laminae C C 3. Laminae A, C, and C 2 receive input from axons of the contralateral eye, laminae A 1 and C 1 from axons of the ipsilateral eye, while lamina C 3 receives no retinal input (Hickey and Guillery, 1974). The other parts of the dorsal lateral geniculate nucleus are the medial interlaminar nucleus and the geniculate wing (Guillery et al., 1980). The geniculate wing has also been considered part of the cat s pulvinar (Berman and Jones, 1977; Kawamura et al., 1979; Leventhal et al., 1980). The cat s superior colliculus was determined cytoarchitectonically to have seven layers by Kanaseki and Sprague (1974), who also suggested that the lateral nucleus of the periaqueductal gray might be considered an eighth layer. The seven layers were termed in order, starting from the top: (I) stratum zonale, (II) stratum griseum superficiale, (III) stratum opticum, (IV) stratum griseum intermediale, (V) stratum album intermediale, (VI) stratum griseum profundum, and (VII) stratum album profundum. The stratum griseum superficiale, which has three sublayers, is the site of retinal input (Sterling, 1971, 1973; Kanaseki and Sprague, 1974; Graybiel, 1975, 1976; Harting and Guillery, 1976; McIlwain, 1978; Behan, 1981, 1982; Freeman and Singer, 1983; Mize, 1983a, b; Berson, 1987). Kanaseki and Sprague (1974) also defined seven nuclei of the cat s pretectum, only two of which receive input from the retina. These are the nucleus of the optic tract and the olivary pretectal nucleus (Berman, 1977). The olivary pretectal nucleus is known to be involved in the pupillary response to light, while the nucleus of the optic tract plays an important role in generating optokinetic nystagmus (Collewijn, 1975; Schoppmann and Hoffmann, 1979; Cazin et al., 1980). Some have suggested that the nucleus of the optic tract might better be considered functionally part of the accessory optic system. Hayhow(1959) divided the accessory optic system of cat into three nuclei, named the dorsal terminal nucleus, the lateral terminal nucleus and the medial terminal nucleus. In mammals, the nucleus of the optic tract has a common border with the dorsal terminal nucleus (Simpson, 1984), and, given that the accessory optic system plays a role in the reflex stabilization of gaze, associating the nucleus of the optic tract with this system makes sense from a functional perspective. The suprachiasmatic nucleus plays a key role in the generation of circadian rhythms. Leventhal et al. (1985) estimated that 40% of the cat s ganglion cells innervate the A laminae of the dorsal lateral geniculate nucleus, 20% the C laminae, 3% the geniculate wing, 10% the ventral lateral geniculate nucleus, 20% the pretectum and 40% the superior colliculus. The numbers of ganglion cells that innervate the accessory optic system and the suprachiasmatic nucleus are very low(farmer and Rodieck, 1982; Murakami et al., 1989; Pu, 1999, 2000). The percentages Leventhal et al. (1985) provide for ganglion cells innervating the pretectum and the ventral lateral geniculate nucleus seem high, but we are not aware of any data that contradict these numbers X- and Y-cell axonal projections It seems probable that all of the retina s X- and Y-cells provide input to the A laminae of the dorsal lateral geniculate nucleus (e.g., Clare et al., 1969; Cleland et al., 1971; Levick et al., 1972; Hoffmann et al., 1972; Singer and Bedworth, 1973; Cleland and Levick, 1974a; Fukuda and Stone, 1974; Dubin and Cleland, 1977; Dreher and Sefton, 1979; Troy, 1983a, b), with the X-cell projection being located more dorsally in the A and A 1 laminae (Mitzdorf and Singer, 1977). All contralaterally projecting Y-cells probably also innervate the geniculate lamina C too and there is a substantial Y-cell projection to the medial interlaminar nucleus (Mason, 1975; Kratz et al., 1978; Dreher and Sefton, 1979; Leventhal et al., 1985). A small fraction of the X-cells may innervate the C laminae and the medial interlaminar nucleus (Mason, 1975; Dreher and Sefton, 1979; Leventhal et al., 1985), and there is some evidence that a fraction of X-cells innervate the pretectum (Leventhal et al., 1985; Cleland and Levick, 1974a; Schoppmann, 1985; Hada and Hayashi, 1990) and the superior colliculus (Cleland and Levick, 1974a; W.assle and Illing, 1980; Hada and Hayashi, 1990), with 4% of X-cells reported to innervate both structures (Hada and Hayashi, 1990). Cleland and Levick (1974a) have reported also that the X-cell projection to the pretectum is biased in favor of ON-center X-cells while the X-cell projection to the superior colliculus favors OFF-center X-cells. The projection of Y-cells to the superior colliculus is substantial (e.g., Clare et al., 1969; Hoffmann, 1972, 1973; Fukuda and Stone, 1974; Schoppmann and Hoffmann, 1979; W.assle and Illing, 1980; Leventhal et al., 1985). There is also a Y-cell projection to the pretectum (Schoppmann and Hoffmann, 1979; Leventhal et al., 1985; Schoppmann, 1985). The pattern of decussation differs between X- and Y-cells also. Some Y-cells with receptive fields 151 eccentric in the temporal retina send their axons to the

11 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) contralateral side of the brain, while no X-cells with eccentricities greater than 21 into the temporal retina do so (Kirk et al., 1976b). Levick (1977) provides an interesting hypothesis for howthis difference is relevant functionally. It is prudent to be suspicious of some of the reports, especially of the axonal projections of X-cells. In some cases the identification of XðbÞ-cells has relied only on the sustained nature of their responses to flashed stimuli and, in other cases, only on cell body size. Neither characteristic fully separates X-cells from some varieties of W-cell. Hence, where not all X-cells are believed to innervate a structure, one should consider the possibility that a W-cell projection has been credited in error to X-cells. Even allowing for this possibility, it is clear that the axons of X- and Y-cells have characteristically different patterns of innervation of the brain. This would seem to support the hypothesis that X- and Y-cells serve distinct functions in vision. One supposes therefore that they constitute independent information channels W-cell axonal projections Taken as a whole, W-cell axons innervate the C- laminae of the dorsal lateral geniculate nucleus (Cleland et al., 1976; Wilson et al., 1976; Rowe and Dreher, 1982; Sur and Sherman, 1982; Leventhal et al., 1985), the medial interlaminar nucleus (Rowe and Dreher, 1982), the geniculate wing (Kawamura et al., 1979; Leventhal et al., 1980, 1985; Pu et al., 1994), the ventral lateral geniculate nucleus (Spear et al., 1977; Rowe and Dreher, 1982; Leventhal et al., 1985), the pretectal complex (Cleland and Levick, 1974a, b; Schoppmann and Hoffmann, 1979; Koontz et al., 1985; Leventhal et al., 1985; Schoppmann, 1985), the superior colliculus (Clare et al., 1969; Hoffmann, 1972, 1973; Cleland and Levick, 1974a, b; Fukuda and Stone, 1974; Kelly and Gilbert, 1975; Magalhaes-Castro et al., 1976; McIlwain, 1978; Schoppmann and Hoffmann, 1979; W.assle and Illing, 1980; Itoh et al., 1981; Leventhal et al., 1985; Berson, 1987, 1988), the accessory optic system (Farmer and Rodieck, 1982) and the suprachiasmatic nucleus (Murakami et al., 1989; Pu, 1999, 2000). However, different varieties of W-cell innervate different sites. What is known about the axonal projections of specific W-cell types will be presented below when we discuss their morphologies and receptive field properties. There is also variation between W-cells in their pattern of decussation. For most types of W-cell, with perhaps the color-coded cell excepted, the fraction of their complement that project axons to the contralateral side of the brain from temporal retina is reported to be significantly greater than for X- or Y-cells. This is particularly so for the sluggish-transient, ON OFF and ON OFF directionally selective cells (Kirk et al., 1976a) Retinal ganglion cell morphologies It was noted earlier that the morphological forms of X- and Y-cells are the b- and a-cells, respectively. There are two varieties of a-cell, one whose dendrites arborize within the outer lamina a of the retina s inner plexiform layer and another whose dendrites arborize within its inner lamina b (Fig. 11). These correspond to OFF- and ON-center Y-cells, respectively. The dendrites of Y-cells tend to stratify in a narrowband in the inner plexiform layer close to the border between laminae a and b (Freed and Sterling, 1988; Berson et al., 1998). There are also two corresponding varieties of b-cell for the OFF- and ON-center X-cells. The dendritic arbors of OFF- and ON-center b-cells tend to stratify, respectively, throughout most of either lamina a or b of the inner plexiform layer (Berson et al., 1998). Fig. 11D is probably an ON-center YðaÞ-cell, while Fig. 11E is probably an ONcenter XðbÞ-cell. By the beginning of the 1980s there were known to be many morphological varieties of cat retinal ganglion cell (Stone and Clarke, 1980; Kolb et al., 1981), suggesting that, if morphological diversity implied physiological diversity, there was even more diversity within the W-cell grouping than had been found physiologically. Hence, with the morphological forms of X- and Y-cells determined, a number of groups set about the task of linking the physiological types of W-cell to distinct dendritic morphologies. An early approach came to be known as the stick and stain technique, because morphology is revealed through a chemical marker injected through a microelectrode into the cell. The technique involves first characterizing the receptive field properties of a ganglion cell, often through extracellular recording, and then penetrating the cell s plasma membrane to inject the chemical marker. In those cases where this is done in vivo (e.g., Fukuda et al., 1984, 1985), the penetration into the cell is done blindly and some have questioned whether the labeled cell is always the one whose physiology had been characterized. Another approach to linking morphological with physiological cell-type has been to retrogradely label cells from the sites of their axonal projections, relying on the hypothesis that W-cells with different physiological properties connect differentially to targets in the brain. Initially, the retrograde tracer horseradish peroxidase was a popular means to tag cells, but its use is rare now since it provides poor staining of a cell s dendritic tree. Today, fluorescent markers are typically used to tag cells. The marker is transported back to the cell body where, in an in vitro preparation, the fluorescing cells are observed microscopically and injected with a second marker that can be used to visualize the full dendritic tree. In the in vitro preparation the receptive field properties may be studied either extracellularly before

12 274 J.B. Troy, T. Shou / Progress in Retinal and Eye Research 21 (2002) penetrating the cell or intracellularly. In the latter case, there can be little dispute that the cell whose morphology was revealed is the one whose physiology was studied. Even in the case of extracellular recording, there is much less likelihood of a false stick than when one penetrates blindly, as in the in vivo studies. This is not to say that the approach of intracellular labeling of fluorescently tagged cells is free of drawbacks. The technique relies on the uptake of the fluorescent tag at synapses in the target site and the retrograde transport of enough of this to visualize a tagged cell microscopically. One can imagine that some axons transport the fluorescent tag faster or better than others. It is also possible that cells with low firing rates might pick up less of the tag, since it is believed that the tracer enters the cell during endocytosis following neurotransmitter release. Finally, intracellular recording is more stable with larger cell bodies, presenting the potential problem of biasing one s cell population towards large cells. All these technical factors should be kept in mind in the following discussion of W-cell morphologies The sluggish-sustained cell The OFF-center sluggish-sustained cell is nowtentatively associated (Stanford (1987a); W.assle et al. (1987); Dacey (1989); but see Fukuda et al., 1984) with the morphological cell-type called d-cell by Boycott and W.assle (1974). For the ON-center sluggish-sustained cell the association with morphology is more complicated. There is the e-cell (Fig. 12; Leventhal et al., 1980; Pu et al., 1994) that seems to have some of the physiological properties associated with the ON-center sluggishsustained cell. However, its maintained discharge was found by Pu et al. (1994) to be more irregular than has been reported for ON-center sluggish-sustained cells (Robson and Troy, 1987; Troy et al., 1995). Dendritic morphologies other than that of the e-cell have also been reported for some ON-center sluggish-sustained cells (Saito, 1983a, b; Stanford, 1987a), raising the possibility that the ON-center sluggish-sustained cell grouping may comprise more than one cell-type. There is also diversity in the physiological properties of ON-center sluggishsustained cells. Notably, Barlowand Levick (1969b) reported a fewganglion cells that they called luminance units, because their mean discharge rate increased monotonically with light level. The luminance unit was initially associated with the ON-center sluggish-sustained cell (Cleland and Levick, 1974a; Stone and Fukuda, 1974). Subsequent work has indicated, however, that luminance units are merely a subset of this cell-type (Rowe and Stone, 1976; Wilson et al., 1976; Troy et al., 1995). Troy et al. (1995) found that only ON-center sluggish-sustained cells of the nasal inferior retina behave as luminance units, a result seemingly consistent with their presumed role in driving the cat s pupillary light reflex. Distler and Hoffmann (1989) Fig. 12. Different morphological forms of cat W-cell, as seen in wholemounts. The y-cell, the e-cell, the Z-cell and the z-cell. From Pu et al. (1994), Reproduced by permission of: The Society of Neuroscience. Berson et al. (1998, 1999) and Isayama et al. (2000). r 1998, 1999, 2000 Wiley-Liss, Inc. Reproduced by permission of Wiley Periodicals Inc. The e-cell is believed to be an ON-center sluggish-sustained cell. The y- and z-cells are presumed to be ON OFF cells. The Z-cell is presumed to be an OFF-center sluggish-transient cell. To the novice, the morphologies of the y-, z- and Z-cell seem rather similar in this view, but their dendritic trees ramify within different strata of the inner plexiform layer. The dendritic tree of the z-cell straddles the border between laminae a and b: The dendritic tree of the y-cell is bistratified, branching in both laminae a and b: The dendritic tree of the Z-cell ramifies only in lamina a: At the same retinal location, the dendritic tree of the e-cell has a spread across the retina roughly like that of the a-cell. The dendritic trees of the y-, z- and Z-cells have spreads that are intermediate between a- and b-cells. report that a fewganglion cells of the nasal inferior retina provide the main projection to the olivary pretectal nucleus, and illumination of the inferior nasal retina in humans has been shown to produce more pupillary constriction than illumination of the other retinal quadrants (Wyatt and Musselman, 1981; Pickworth et al., 1993). Hence, d-cells are OFF-center sluggish-sustained cells and e-cells are ON-center sluggish-sustained cells, but we do not know whether other morphologies fit the sluggish-sustained physiological cell-type also. Certainly, the ganglion cell-type that innervates the suprachiasmatic nucleus has an ON-center receptive field center and gives a sustained response to light stimulation (Pu, 2000). However, these cells have very large receptive field centers and dendritic trees (Pu, 1999, 2000) which are clearly different from e- cells. Given their large receptive field centers it seems improbable that these cells would be confused with the typical ON-center sluggish-sustained cell encountered in electrophysiological experiments. Nonetheless, even ignoring this cell-type, there is sufficient variability in