This is a review paper that includes original calculations and figures. The drone retina is composed of two essentially uniform populations of cells, the photoreceptors and the glial cells. The photoreceptors contain many mitochondria but no glycogen has been detected; the glial cells contain much glycogen and very few mitochondria. The oxygen consumption of the photoreceptors in the dark is 20 μl min−1 per g of retinal tissue and in response to a single flash of light there is an extra consumption that reaches a maximum of 40 μl min−1 per g. In addition, light stimulation of the photoreceptors leads to changes in the glycogen metabolism of the glial cells, and to movements of K+. Measurements with intracellular Insensitive micro-electrodes showed that during light stimulation with a series of flashes the K+ activity (aK) in the photoreceptors fell by an average of 27% while in the glial cells aK rose by an amount that is estimated to correspond to most of the quantity of K+ lost by the photoreceptors. The relative contributions to the clearance of extracellular K+ of extracellular diffusion, spatial buffering and possible net K+ uptake by glial cells are discussed.
Each of the various preparations that have been used for studying glial function has features appropriate for certain kinds of investigation. For example, in this volume, Orkand, Orkand & Tang (1981) describe a preparation containing only glial cells that can be obtained by axotomising Necturus optic nerve; and Treherne & Schofield (1981) report work on the well-developed blood-brain barrier formed by peripheral neuroglia in the cockroach central nervous system. The interest of the drone retina seems to us to lie in a histology that suggests a very marked compartmentation of energy metabolism between glial cells and sensory neurones; in the uniform and extensive structure that makes feasible quantitative studies of such functions as oxygen consumption ; and in the fact that a direct demonstration has been made of a large flux of potassium ions into the glial cells during quasi-physiological stimulation of the photoreceptor cells.
Nearly all the volume of the retina of the compound eye of the drone is made up of only two types of cells; the large retinula cells, which are photoreceptors, and the outer pigment cells, which, in the electron-microscope study of Perrelet (1970), are considered to be glial cells. The photoreceptors contain visual pigment and respond directly to light. They are about 400 μm long and are clustered in groups of six so that in cross-section each group (or retinula) appears as a rosette about 20 μm in diameter (Fig. 2). The visual pigment is located on tightly packed microvilli that constitute the rhabdom, a structure about 2 by 6 μm that runs down the centre of the retinula. Ranged close to the plasma membrane are numerous mitochondria, whose oxygen supply in the intact animal is provided by the tracheoles that run through the proximal half of the retina parallel to and between the retinulas. Apart from the extracellular clefts, which probably make up about 5 % of the total volume of the retina (Coles & Tsacopoulos, 1979 ; cf. Shaw, 1977, 1978), the rest of the space between the retinulas is filled by the pigment cells. About 30 of these surround each retinula and send finlike processes between the photoreceptors towards, but not reaching, the rhabdom; the whole functional unit is known as an ommatidium. In contrast to the photoreceptors, the outer pigment cells contain very few mitochondria, but large quantities of glycogen particles (Perrelet, 1970). Also, only the pigment cells stain with the periodic acid-Schiflf reaction, which normally detects glycogen (see Pearse, 1968; A. Perrelet & V. Evêquoz, personal communication). These observations suggest that most of the aerobic metabolism is confined to the photoreceptors, while carbohydrate stores appear to be mainly in the pigment cells. In some mammalian nervous tissue a similar, but less extreme, compartmentation is found (O’Connor, 1977) and there is evidence for compartmentation of certain other metabolic functions, such as amino acid synthesis (see Hertz, 1979).
III. BASIC ELECTROPHYSIOLOGY
Most experiments on the drone retina have been done on the cut head preparation (Baumann, 1968). A slice is removed from the caudal aspect of the isolated head to expose layers of ommatidia corresponding to the large dorsal facets (Van Praagh, Ribi, Wehrhahn & Wittmann, 1980). The head is then fixed in the floor of a chamber and superfused with an oxygenated Ringer solution (Fig. 1). Light stimulation is made perpendicular to the cut surface and, except near the cornea, attenuation is less than 0·3 log units at a depth of 300 μm. The visual pigment (unlike that of vertebrates) does not bleach when it is photo-isomerized and, provided that the spectral composition of the light is kept constant, the quantity of pigment in the active form does not change (Bertrand, Fuortes & Muri, 1979).
Three classes of electrical response can be recorded with a glass micropipette in response to intense flashes of light (in the order of 10mj/cm2). The extracellular potentials are predominantly negative-going, have a complex waveform and an amplitude of as much as 15 mV (see Baumann, 1974, and Gardner-Medwin, Coles & Tsacopoulos, 1981). The response recorded from within a photoreceptor is depolarizing and has an amplitude of at least 50 mV, and it arises because light causes an increase in the conductance of the membrane to sodium ions (Fulpius & Baumann, 1969). Concomitant with the response of the photoreceptors, the outer pigment cells also undergo a transient depolarization whose maximum amplitude is usually 10–20 mV. This depolarization reaches its peak more slowly than does the receptor potential of the photoreceptors, and has a slower decay (Bertrand, 1974). It seems to be analogous to the response recorded by Orkand, Nicholls & Kuffler (1966) in the glial cells of Necturus optic nerve and which they attributed to the increase in extracellular K+ concentration caused by nerve impulses. This explanation can account reasonably for the time course of the response of the drone pigment cell to a single light flash, and also for the effect on its amplitude of changing the K+ concentration in the superfusate (Bertrand, 1974). Hence, on morphological, histological and electrophysiological grounds, the outer pigment cells can be identified as glial cells.
The glial cells would also depolarize after a light flash if there were electrical junctions between the glial cells and the photoreceptors and, since we will later be discussing the possibility of a transfer of substances between the glial cells and the photoreceptors, it seems appropriate to list the observations concerning the absence of junctions between the two classes of cell : (1) When the fluorescent dye Lucifer Yellow is injected into a photoreceptor, it passes into thé other large photoreceptors of the retinula, which are known to be electrically coupled (Shaw, 1969), but the dye is not seen in glial cells (A. Perrelet, unpublished). (2) Although gap junctions have been observed histologically between cells of the same type (between photoreceptors, or between outer pigment cells) they have not been observed between the different classes of cell (Perrelet, 1970, and personal communication). (3) As described in Section V, the intracellular K+ activities measured by Coles & Tsacopoulos (1979) were significantly different in photoreceptors and glial cells.
We conclude that there are few, if any, junctions between photoreceptors and glial cells. A general review of this question for other tissues can be found in Varon & Somjen, (1979).
(1) Oxygen consumption
The drone retina cannot function without oxygen. If the Ringer solution superfusing the cut head preparation is replaced by one lacking oxygen, then the anaerobic metabolism of the photoreceptors is insufficient to maintain the membrane potential in the dark (M. Tsacopoulos & A. Mauro, in preparation); with light stimulation, anoxia causes complete depolarization within 3 min (Baumann & Mauro, 1973). These observations suggest that the energy metabolism of the photoreceptors is predominantly aerobic, and since nearly all the mitochondria of the retina are in the photoreceptors, the oxygen consumption should be a measure of the rate of the photoreceptor energy metabolism. Knowledge of this should provide an indication of the necessary size of stores of metabolic substrate (Section IV. 2) and of the energy available for sustaining ion fluxes (Section VI).
Tsacopoulos, Poitry & Borsellino (1980, 1981) have shown that (provided the superfusion is well arranged) the diffusion of oxygen into the retina of the cut head occurs essentially in one dimension, perpendicular to the plane of the cut. They further showed that oxygen consumption was independent of down to a small value (less than 20 torr). With this information they could calculate the steady-state oxygen consumption from measurements of made with an oxygen microelectrode (tip diameter less than 2 μm) at different depths in the tissue and in the bath. The oxygen consumption they found in the dark at 23 °C was 20 μl min−1 per g of tissue, which may be compared to the value of 40 μl min−1 per g of tissue found for the isolated retina of the worker bee (with its smaller cells) by Autrum & Hamdorf (1964) (see Tsacopoulos et al. (1981) for the conversion of the units). In the drone cut head preparation the oxygen entering through the cut surface is sufficient to maintain a constant oxygen consumption down to a depth of about 460 μm in the dark, and about half of this when the photoreceptors are stimulated by a series of light flashes.
As shown in Fig. 3, a single light flash evokes a change in within the superfused retina that is readily measurable with an oxygen microelectrode (Tsacopoulos & Lehmenkühler, 1977; Tsacopoulos, 1978). In a preparation oxygenated all the way through (in practice, a slice of retina 300 μm thick) the light-induced change in oxygen consumption can be calculated from the by the method of discrete Fourier transforms (Poitry & Tsacopoulos, 1981). The extra consumption reaches a peak very rapidly (2–3 s) after the flash and the total extra consumption is about 12 μl per g of tissue for a flash of 40 mJ cm−2. Comparable measurements exist only for frog striated muscle (Mahler, 1978) where the kinetics are found to be much slower.
(2) What is the source of metabolic substrate?
The cut-head preparation gives electrical responses to light stimulation for at least 12 h while superfused with a Ringer solution that contains no glucose or other metabolic substrate (F. Baumann & M. Tsacopoulos, unpublished). The oxygen consumption in the dark does not change detectably for about 6 h of this superfusion (Tsacopoulos et al. 1981) and light flashes continue to evoke rapid increases in consumption (Tsacopoulos, 1978). It seems clear that the retina is drawing on an internal store of substrate. We estimate that during 12 h of superfusion with intermittent stimulation the average oxygen consumption will be about 40 μl min−1 per g of tissue. If the substrate were carbohydrate, as is the case for the flight muscles of the bee (Jongbloed & Wiersma, 1935), then 54 μl g min−1 per g of tissue would be oxidized, or about 38 mg per g of tissue in 12 h. The tempting hypothesis that the photoreceptors draw on the glycogen store in the glial cells has not been rigorously tested experimentally, but there is clear evidence that when the photoreceptors are stimulated by light the glycogen metabolism is modified. The experiment is the following one (Evêquoz, Deshusses & Tsacopoulos, 1978; Tsacopoulos & Evêquoz, 1980). One eye of a drone is occluded with opaque paint and the other eye is stimulated with bright flashes of light for 30 min. A bolus of radioactive glucose is then injected into the heart, and the light stimulation is continued. At a chosen time (5–60 min) after the injection the head is frozen and it is found that the glycogen of the stimulated eye has a higher radioactivity than that of the unstimulated eye, although the total quantity of glycogen does not change significantly. Hence, in vivo, stimulation of the photoreceptors can lead to an increase in the turnover of the glycogen in the eye, presumably in the glial cells. Some signal, in the broadest sense of the word, must pass from the stimulated photoreceptor to the glial cell. We do not know what the relevant signal is, but a mechanism involving potassium seems to be a possibility. Salem, Hammerschlag, Bracho & Orkand (1975) have shown that the metabolism of glucose in glial cells of Necturus optic nerve can be stimulated by increased extracellular K+ concentration and, as we describe in Section V, in the drone retina there do appear to be substantial movements of K+ ions at the onset of photostimulation.
V. MEASUREMENT OF POTASSIUM ACTIVITY
Double-barrelled potassium-sensitive microelectrodes of the liquid-membrane type (Walker, 1971) can be made fine enough to give stable recordings from both the photoreceptors and the glial cells of the drone retina (Coles & Tsacopoulos, 1979). The electrodes measure K+ activity (aK) rather than concentration and to estimate the latter we assume that the ratio of the two (the activity coefficient) is the same as in the bee Ringer solution (0 ·70). We consider particularly the transition from darkness to the quasi-steady state produced by stimulation with a train of light flashes each 20 ms long and presented at 1 Hz.
A typical record is shown in Fig. 4 and averages of results obtained by Coles & Tsacopoulos (1979) are given in Table 1. At the onset of the photostimulation aK fell by an average of 27%, with a half time of 30 s. A priori this change might be due to an efflux of K+ ions from the cell, an increase in the proportion of intracellular K+ that is bound, or to an influx of water. The first of these hypotheses seems to be consistent with all the available experimental data and it is the only one we consider further. It is supported by the demonstration of a light-induced efflux of K+ from invertebrate photoreceptors with radioactive tracers, in Limulus ventral nerve (Holt & Brown, 1972) and in crayfish (Stieve & Hartung, 1977). Further, in Limulus ventral photoreceptor there is a voltage-dependent K+ conductance (Millecchia & Mauro, 1969; Pepose & Lisman, 1978) that would tend to facilitate such an efflux. In the drone photoreceptors, the change in aK seems strikingly large, but if allowance is made for the large surface-to-volume ratio of the cells (1·8 μm−1, due mainly to the microvilli) and for the long duration (50–100 ms) of the receptor potentials induced by the light flashes, then the estimated flux of ions per unit area of surface membrane is not very different from that measured for the squid giant axon during a nerve impulse.
(2) Extracellular space: the need for a K+ sink
At depths of 20–150 µm from the surface of the cut head preparation Coles & Tsacopoulos (1979) found that at the onset of the photo-stimulation the extracellular aK increased to a piaximum and then declined to a plateau that was, on average, 3·5 mm above the value in the dark (Table 1). Fig. 5 A shows an extracellular recording made under different, but better defined, conditions: a precise depth in a fully oxygenated slice of retina. During the photostimulation the K+ voltage signal increased and at the end it transiently undershot the dark level. With the aid of the calibration curve of the electrode the K+ concentration was calculated for the first 80 s of the stimulation and this is shown as the solid line in Fig. 5 B : the maximum concentration is about 20 mm. Since we know the efflux of K+ from the photoreceptors, we can make assumptions about its fate in the extracellular space, calculate the predicted change in extracellular K+ concentration, and test the assumptions by comparing the predictions with the experimental measurement.
(3) Glial cells
The magnitudes of the resting potentials of the glial cells, measured with the reference barrel of a K+ electrode, were about 50 mV, considerably lower than those of glial cells in most other preparations (see, for example, Somjen, 1975; Bracho, Orkand & Orkand, 1975). However, neither Bertrand (1974) nor Coles & Tsacopoulos (1979) were able to record larger resting potentials with fine, single-barrelled electrodes either in the cut-head preparation or in the retina in situ. As found by Nicholls & Kuffler (1964, 1965) for glial cells in the leech, and predicted by Kuffler, Nicholls & Orkand (1966) for those in Necturus optic nerve, aK in the drone glial cells in the dark coincided (within experimental variability) with that corresponding to Nernst equilibrium (Table 1). During photostimulation aK rises, as shown in the recording in Fig. 4B. The mean increase was 14 mm, from a dark level of 52 mm; the half time of the rise was variable, but not obviously different from that of the corresponding fall in the photoreceptors.
Mori, Miller & Tomita (1976) recorded an increase in intracellular aK in the glia cell (Müller cell) of the vertebrate retina during spreading depression, but Coles & Tsacopoulos, 1979, appears to be the first report of an increase in aK in a glial cell during quasi-physiological stimulation of a tissue, so it is appropriate to consider the validity of these results. (1) Cells gave repeated, closely reproducible responses to light stimulation for as long as the dark membrane potential remained constant (a maximum of 25 min). (2) The change in the glial cells was an increase of aK, while that in the photoreceptors was a decrease. If both responses are artifacts of some kind of cell damage, then the difference in sign seems somewhat surprising. (3) A drawback of the K+ sensor used in the intracellular electrodes (Corning 477317) is that it is highly sensitive also to quaternary ammonium ions, including acetylcholine. However, in the extracellular space, at least, no differences have been observed between the light-induced ion signals recorded with this sensor or with other sensors more highly selective for K+ (see Fig. 5 A). (4) But the strongest argument for the validity of the results in the cut-head preparation is the self-consistency of the results : this is described in the next section.
(4) Different routes of K+ clearance from the extracellular space
Some of the potassium that is released by the photoreceptors diffuses through the extracellular clefts to the bath or, in the cut-head preparation, to the deep inactive layers of the retina. We feel that, at present, the uncertainties in the values of the parameters make an elaborate calculation unjustified, but a rough estimate can be made on the basis of an extracellular K+ concentration gradient such as that in Fig. 6. We calculate that at a distance of 150 μm from the bath, probably less than 20 % of the K+ that leaves the photoreceptors is cleared by diffusion through the extracellular space. The total quantity of additional K+ that remains in the extracellular space is also comparatively small (because both the K+ concentration and the volume are small). Is the quantity that enters the glial cells sufficient to account for the rest? Coles & Tsacopoulos (1979) estimated that the total volume of the outer pigment cells in the retina was about 1·5 times the volumes of the photoreceptors. If the increase in aK in the glial cells (14 mm) is multiplied by this factor the value obtained is close to the decrease in aK measured in the photoreceptors (21·5 mm, see Table 1). Hence, within the uncertainties of the values, all the potassium released by the photoreceptors is reasonably accounted for.
(5) The entry of potassium into the glial cells: spatial buffering
The different mechanisms by which potassium might enter cells have been reviewed by Gardner-Medwin (1980). One such mechanism, the ‘spatial buffering’ proposed by Orkand et al. (1966), appears to make a major contribution to the entry of potassium into the glial cells in the experiments just described on the drone retina, and we discuss this first.
Spatial buffering is a process by which an increase in the quantity of K+ in the extracellular space in a limited region of tissue is spread over a larger volume by passive movement of K+ through cells; it operates when an unexcited cell, or an electrically coupled syncytium of such cells, extends from the region of high extracellular K+ concentration to regions of lower concentration. The mechanism works best if the surface membranes are selectively permeable to K+ and hence glial cells are likely to be more effective than neurones, but for generality we call them ‘transfer cells’ (Gardner-Medwin, 1981 b). The net K+ flux across any part of the surface membrane of a transfer cell is zero when the membrane potential is at the Nemst equilibrium potential for K+. If the extracellular K+ concentration rises, as normally occurs in a region of tissue where neurones are active, then K+ will enter the transfer cells. If all parts of the cell membrane were confronted by the same increased K+ concentration then the entry of only a very small quantity, enough to increase the intracellular concentration by a few μmoles/l, would suffice to depolarize the membrane capacitance to the new Nernst equilibrium potential. However, parts of the transfer cell (or the syncytium) extend to regions where the extracellular K+ concentration has not increased and, by electrical coupling, the membrane potential in the active region is prevented from reaching the new Nernst equilibrium. Consequently, K+ continues to enter the cells in the active region. Conversely, the membranes in the inactive region are depolarized by the electrical coupling and here there is an efflux of K+. Electrical current, carried by unspecified ions, flows intracellularly through the transfer cell syncytium from the active region to the inactive region to maintain the transmembrane K+ fluxes and there are necessarily return current loops in the extracellular space. These extracellular currents are associated with extracellular potential gradients, and measurement of these gradients can provide a basis for estimating the magnitude of the spatial buffering of K+ (Gardner-Medwin, Gibson & Willshaw, 1979). The circumstances in which spatial buffering makes an appreciable contribution to clearance of raised extracellular K+ concentration are being investigated in a number of laboratories (e.g. Gardner-Medwin, 1977, 1980, 1981a; Nicholson et al. 1979; Dietzel, Heinemann, Hofmeier & Lux, 1980).
An estimate of the contribution of spatial buffering to the K+ movements in the Irone cut head preparation has been made by Gardner-Medwin et al. (1981). Extra-llular K+ concentrations and also extracellular potentials were measured at different distances from the superfused surface of the retina. During light stimulation the photoreceptors released K+ into the extracellular space in an active layer some 200 μm thick (Fig. 6A).The extracellular potential evoked by a light flash in this region was resolved into two components, a rapid one, corresponding to current that entered the photoreceptors and left through their axons, and a slower one corresponding to a current that was deduced to enter the glial cells. Measurement of the slow component at different depths (Fig. 6B) showed that this current flowed in two loops. It entered the glial cells in the active region and left either close to the bath or in the deep anoxic layers where the photoreceptors did not respond to light. The current loops were completed through the extracellular space as shown in the diagram in Fig. 6C. The changes in intracellular aK measured in the glial cells (Section V. 3) fit with the idea that the current entering the glial cells is carried by K+ ions.
In other tissues where spatial buffering has been studied, both neurones and glial syncytia extend over distances of several hundred microns and it has not been entirely certain which cell type subserved the spatial buffering: in the drone retina it is unambiguously the glia, at least in the direction perpendicular to the axes of the ommatidia, since the photoreceptors in adjacent retinulas are isolated from each other.
To estimate the magnitude of the spatial buffering current in the drone retina Gardner-Medwin et al. (1981) made calculations that assumed a value of the extracellular space fraction, α, of 5%, and of the tortuosity factor, λ, of 1 · 6 (see Section V. 2). It was concluded that within the uncertainties of the calculation, the ionic current that entered the glial cells was sufficient to account for the dispersal of the K+ released by the photoreceptors. (Alternatively, we note that this is independent evidence, not depending on intracellular recordings, that the photoreceptors lose a substantial fraction of their potassium during stimulation.)
In most cells, K+ is thought to be the principle diffusible ion in the cytoplasm so that within the glial syncytium the current will be carried mainly by K+, and the quantity of K+ that accumulates in the glial cells in the active region should be much less than the quantity lost from the excited cells. In the drone retina the two quantities appear to be of similar magnitudes. A possible explanation for this is that the free K+ (74 mg ion/1) represents only a small fraction of the total osmotically active particles (587 mg species/1) and might have a transport number roughly in this proportion (74/587) in the cytoplasm and in the intercellular junctions. This hypothesis requires that most of the other particles be electrically charged, and their nature is not obvious; it is desirable that the composition of the cytoplasm be studied further.
(6) Are there additional mechanisms by which K+ enters the glial cells?
In those tissues in which extracellular potassium dispersal has been analysed most fully, the most satisfactory explanation of the data seems to be that over distances of more than a few hundred microns potassium is cleared mainly by spatial buffering and partly by net uptake into cells (Gardner-Medwin, 1981 a). The analysis just described shows that in the drone cut-head preparation spatial buffering apparently plays a major role, but it is not precise enough to decide whether or not there are significant contributions from other processes, such as, for example, entry of K+ together with Cl−, or Na+/K+ exchange. There, are indeed, observations, which we now summarize, that suggest that perhaps the latter process might occur.
Firstly, the photoreceptors in the cut drone head continue to respond to light stimulation for up to 12 h when the Ringer solution contains no sodium (Fulpius & Baumann, 1969) and the removal of calcium as well does not abolish the response (J. A. Coles, unpublished). Fulpius & Baumann (1969) suggested that the glia might contain a store of sodium which could be released to prevent depletion in the extracellular space. Secondly, spatial buffering in mammalian brain is associated with changes in the volume of extracellular space (Dietzel et al. 1980), but in the drone retina the volume changes appear to be smaller than would be predicted (Coles et al. 1981). Thirdly, Tsacopoulos & Coles (1978) found that after superfusing the retina with Ringer solution that contained strophanthidin-k, a specific inhibitor of (Na+,K+)-ATPase, (2 × 10−5 M for 2 min) the light-induced changes in glial aK were smaller than in the control. This treatment with strophanthidin caused no detectable modification of the light-induced changes in aK in the photoreceptors nor, paradoxically, in the extra cellular space.
These observations are inconclusive, but they do suggest that it would be worthwhile to continue to investigate the mechanism by which K+ enters the glial cells. In doing this, it may prove necessary to take a less simplistic view of the extracellular space, since there is evidence in other species that the activity coefficients for cations may be much less than in the Ringer solution (Treherne & Schofield, 1981) and that the osmolarity may vary (Kraig & Nicholson, 1978).
VI. A COMPARISON OF OXYGEN CONSUMPTION AND THE ESTIMATED ENERGY COST OF THE LIGHT-lNDUCED ION MOVEMENTS
In the photoreceptors, light stimulation activates channels in the surface membrane that admit a current carried mainly by sodium ions (Section III). This sodium must be pumped out of the cell again and the work required is one evident use of the ATP that will be produced by the aerobic metabolism. We now use the results obtained with K+-sensitive microelectrodes to make a tentative estimate of the quantity of ATP consumed by the extra work of the pump. If the cell is considered before and after a receptor potential, when the membrane has repolarized, it is clear that the electric charge carried into the cell by the Na+ must be balanced by other charge movements. A small contribution is made by an inward movement of Cl- ions (Fulpius & Baumann, 1966) and some current passes down the photoreceptor axon (Gardner-Medwin et al. 1981), But, presumably, most of the charge is balanced by the outward movement of K+. Hence, we infer that the fluxes of K+ across the surface membranes of the photoreceptors are rather smaller than the fluxes of Na+. However, in order to make an approximate calculation, we assume that there is a one-for-one exchange of Na+ for K+.
During stimulation with a series of light flashes, the photoreceptor reaches a quasi-steady state in which the Na+ that enters the cell after a flash is pumped out again before the next one. Concomitantly there must be fluctuations in the intracellular K+ concentration. Although corresponding fluctuations can be detected in the extracellular space with an ion-selective microelectrode, fine, intracellular electrodes, such as that used for the recording in Fig. 4A, cannot follow the small, rapid (1 Hz) fluctuations in intracellular aK; so in order to estimate the Na+ influx per flash we consider the initial rate of change of intracellular K+ at the onset of photostimulation. At the very beginning of the stimulation there will be no additional Na+ in the cell to be pumped out, i.e. the only significant Na+ flux will be into the cell, and, hence, the rate of change of the intracellular K+ concentration will be a measure of this flux. The value given by equation (1) is 0 · 7 mm s−1. Coles & Tsacopoulos (1979) estimated that the photoreceptors occupy about 38% of the volume of the retina, so this concentration change is equivalent to 0 · 27 μ mol per flash per g of tissue. In excitable tissues the sodium pump usually hydrolyses one ATP molecule for roughly three Na+ ions transported (see Glynn & Karlish, 1975), so the estimated requirement for the drone retina is 0 · 27/3 = 0 · 09 μ moles ATP per flash per g of tissue. The extra oxygen consumption in the steady state during stimulation with a series of flashes about the same intensity is 0 · 040 ml min− 1 per g of tissue (Tsacopoulos et al. 1980). This is equivalent to 0 · 030 μ mole of oxygen per g of tissue per flash. If the substrate of the photoreceptor metabolism were glucose, then consumption of one molecule of oxygen would correspond to the production of six molecules of ATP (see, for example, Stryer, 1975), i.e. 6 × 0 · 03 = 0 · 18 μ mole of ATP per g of tissue. Hence, there seems to be ample ATP to sustain the observed ion fluxes in the photoreceptors.
In the glial cells the K+ fluxes associated with spatial buffering will be driven entirely by the changes in extracellular K+ concentration. In addition, the anaerobic breakdown of glycogen into a substrate that can be transferred to the photoreceptors (glucose or pyruvate are a priori possibilities) would be associated with the production of ATP that could be used to pump K+ if, in addition to spatial buffering, there is a net uptake during activity (Section V. 6).
The results demonstrate the intimate dependence of the photoreceptors on the glial cells, firstly, for preventing large changes in extracellular K+ concentration, and, secondly, for the possible supply of metabolic substrate. This is summarized in Fig. 7. The fractional changes in intracellular aK in the photoreceptors and the glial cells sight be large enough to affect the functioning of these cells and it is perhaps note worthy that the surface-to-volume ratio of the photoreceptors is less than that of cylinder 2 μ in in diameter : dendrites and axons in vertebrate nervous tissue are often finer than this and it seems possible that significant changes in intracellular aK may occur in them too.
Two lines of enquiry seem evident: (1) By what mechanism, in addition to spatial buffering, does potassium enter the glial cells? (2) What is the identity of the metabolic substrate for the photoreceptors? How is it transported from the glial cells? And by what signal does photoreceptor activity affect glycogenolysis in the glia?
We thank Dr F. Baumann and other colleagues in Geneva for helpful discussion, and also Drs A. R. Gardner-Medwin & C. J. Karwoski for commenting on an early draft of the manuscript. This work was supported in part by the Swiss National Fund for Scientific Research, grant no. 3.399.78, and the Kernen Foundation.