Documenta Ophthalmologica 69:175-186 (1988) 9 Kluwer Academic Publishers, Dordrecht - Printed in the Netherlands

Development of neurochemical separation of ON and OFF channels at retinal ganglion cells HISAKO IKEDA & J O N A T H A N ROBBINS Vision Research Unit of Sherrington School, The Rayne Institute, St. Thomas' Hospital, London SE1 7EH, UK

Key words: contrast vision, development, dopamine, gamma-aminobutyric acid, glycine, retinal ganglion cell Abstract. Anatomical and physiological segregation of neurons into ON (brightening detector) and O F F (darkening detector) channels in the retina and subsequent visual system ensure the high sensitivity required for contrast detection and spatial discrimination. This segregation 9is finest at the visual axis. Neurochemically, ON and OFF ganglion cells at the visual axis seem to be distinguished by different inhibitory transmitters but not excitatory transmitters. Microiontophoretic studies of inhibitory transmitters on the retinal ganglion cells in kittens and adult cats suggest that this neurochemical distinction is poor in immature ganglion cells at the visual axis. Initially both ON and OFF cells seem to be supplied by GABAergic, glycinergic, and catecholaminergic amacrine cells, but in adults, ON cells remain supplied only by GABAergic amacrines, while OFF cells are supplied by glycinergic amacrines. Postnatal elimination of multiple inputs and strengthening of the appropriate inputs, as seen in the central nervous system, also seem to occur at the retinal neurotransmitter synapses during development.

Introduction This paper is concerned with the relationship between the postnatal development of neurotransmitter actions which distinguish ON and OFF ganglion cells' receptive fields and the postnatal development of spatial resolution and contrast sensitivity at the visual axis. Electrophysiologically determined spatial resolution and contrast sensitivity of ganglion cells at the visual axis do not mature until 11-12 weeks of age in the cat [1]. This development is associated with a decrease in the receptive field center size and an increase in the strength of the inhibitory surround of ON and OFF cells, which respond to visual stimuli of opposite contrast in a given space [1,2,3,4]. In the central cone pathway of adult retinae, ON and OFF ganglion cells are distinguished morphologically by the different levels at which their

176 dendritic branching occurs in the inner plexiform layer [5]. Furthermore, under cone-driven conditions, the adult ON and OFF cells at the visual axis are pharmacologically distinguished by different inhibitory transmitters. Gamma-aminobutyric acid mediates inhibition of ON cells, whereas glycine mediates inhibition of OFF cells [6,7]. This chemical separation is eminently suitable for optimizing the local contrast since those visual stimuli which excite ON cells inhibit OFF cells by one transmitter (glycine), whereas those which excite OFF cells inhibit ON cells by another transmitter (GABA). If both ON and OFF cells are inhibited by the same transmitter at the same time, a reduction of local contrast could result. Our question is, then, can the postnatal development of the neurochemical separation of ON and OFF cells be partly responsible for the development of contrast sensitivity and spatial resolution at the visual axis?

Methods

Microiontophoresis by means of six-barrelled electrodes was performed on physiologically identified retinal ganglion cells in the optically intact eyes of barbiturate-anesthetized cats aged 18-22 weeks and kittens aged 7-9 weeks. The details of the methods, including the preparation of the eye and the selection and classification of retinal ganglion cells, have been described elsewhere [8]. All cells concerned in this paper were those located in the area centralis (a vessel-free zone approximately 15~ temporal and 5~ superior with respect to the disc). Drugs used were GABA (500 mM in water, pH 3.0, Sigma), bicuculline methobromide (15 mM in water, pH 3.0, Cambridge Research Biochemicals), glycine (500 mM in water, pH 3.0, Sigma), strychnine hydrochloride (5 mM in 165 mM NaC1, pH 3.0, Sigma), dopamine hydrochloride (500 mM in H20 in 1 mg ascorbic acid/ml, pH 3.5, Sigma), haloperidol (6mM in 165 mMNaC1, pH 3.5, Janssen), ~-flupenthixol dihydrochloride (5mM in 165 mM NaC1, pH 3.0, Merck, Sharp & Dohme) and L-sulpiride (20 mM in 165 mMNaC1, pH 4.0, Ravizza). One of the six barrels always contained sodium chloride (BDH 500mM in water, pH 7.0) providing a balance current at the tip of the electrodes. To prevent drugs leaking out, a retaining current of 35-45 nA of polarity opposite to that of the ejection current was always applied to each of the drug barrels except during the ejection period. The visual stimuli were either a bright or dark spot or annulus generated onto a TV screen (100cd/m2). The size of the spot was chosen so that it produced optimal firing from the cell, and that of the annulus was such that it produced optimal inhibition of firing from the cell. ON cells were sti-

177 mulated by a bright (197cd/m 2) spot or annulus, O F F cells by a dark (20cd/m 2) spot or annulus. In some experiments sinusoidally modulated gratings of different spatial frequencies (mean luminance 100cd/m 2) were generated onto the TV screen and drifted at a temporal frequency of 1 Hz. In order to compare the effects of drugs on adult and kitten cells, an effective current was determined for each drug in each cell, as described in detail by Ikeda and Robbins [8]. The effective current for G A B A or glycine was defined as a current level at which 95% of the cell firing was abolished. If G A B A reached the effective current, glycine was applied to the same cell at that current level for l min to see if it caused any effect. However, if glycine reached the effective current level, GABA was applied on the same cell at that current level for 1 min to see if it had any effect. The effective currents for bicuculline and strychnine were defined as those which completely abolished the visually induced inhibition of cells. The transport numbers of GABA, glycine and dopamine are 0.3, 0.5, and 0.4. respectively [9], and thus allow us to compare their effectiveness by microiontophoresis. Unfortunately, there are no data available so far on transport numbers for the antagonists used. Since the effect of dopamine was so weak that it could never abolish firing by 95%, the effects of dopaminergic drugs were compared between kitten cells and adult cells by measuring the percentage of inhibition obtained by the drug applied at 50nA to 100nA for 1 minute. All cells that showed no recovery from drug effects or revealed a change in spike height or waveform during drug application were rejected from this study. Any change in spike waveform or size would indicate a non-specific membrane effect rather than post-synaptic action of the drug. All cells studied were those that produced a large spike with a notch, suggesting that the position of the recording electrode was close to the soma [10]. The drug receptors with which we were concerned, therefore, were probably those present on the soma or on the dendritic field 10-20/~m from it.

Results

Electrophysiological evidence for postnatal development of spatial selectivity and resolution of retinal ganglion cells in the visual axis ON cells give excitatory firing to stimuli brighter than background and show inhibition to stimuli darker than background at the receptive field center. These cells therefore give rhythmically modulated firing, showing alternating excitatory and inhibitory response, when a sinusoidal grating is moved

178 Age of Animal

Optimal

B

Visual Acuity

7~ weeks

"5 ..Q

E z

18-22 weeks

d

d

~

g

N

N

~

4

4

Spatial frequency (cycles/degree) Fig. 1. Comparisons of optimal spatial frequency (A) and visual acuity (B) obtained from adult cat and kitten cells at the visual axis. The optimal frequency is the grating spatial frequency at which cells gave their best modulated firing; visual acuity is the highest grating spatial frequncy to which cells gave a modulated firing. Gratings: sinusoidally mean luminance, 100 cd/m2; contrast, 0.8; drift rate, 1 Hz. All cells were classified as sustained (X) type.

at a constant speed at their receptive field center. Similarly, OFF cells also give modulated firing, since they give excitatory firing to a dark phase and show inhibition to a brighter phase of gratings. The modulated firing of a cell indicates that the cell is responding to the contrast modulation of the grating pattern. Thus with gratings of different spatial frequencies one can measure the optimal spatial frequency or visual acuity for each cell. The optimal spatial frequency is the spatial frequency of the grating to which the cell responds best, whereas the visual acuity is the highest spatial frequency to which the cell can give modulated firing. Figs. 1A and B show the distribution of the optimal spatial frequency and the visual acuity of ganglion cells from the visual axis of adult and kitten retinae respectively. The receptive field of a kitten cell is tuned to a lower spatial frequency and provides poorer visual acuity than that of an adult. Thus the spatial resolving power of the ganglion cells in the visual axis is still not fully developed at the age of 7-9 weeks in the cat. We need to establish whether inhibitory transmitter action, which separates ON and OFF ganglion cells at the visual axis, contributes at all to spatial resolution.

179 OFF-cell

ON-cell ~o ,plk~,l=

1

Kitten control I

I

Adult control

Adult defocused (-8 D)

Adult

Bicuculline

Adult

Strychnine

Fig. 2. Effects of age, stimulus defocus by - 8 D and iontophoretically applied inhibitory transmitter antagonists on the response of ON and OFF sustained retinal ganglion cells at the visual axis. These cells are responding to sinusoidally modulated gratings of 3 cycles per degree drifted in the cell's receptive field at 1 Hz (contrast, 0.8; mean luminance, 100cd/mZ). Bicuculline (GABA receptor antagonist) abolished the modulation of firing of the adult ON cell and strychnine (glycine receptor antagonist) abolished that of the adult cat OFF cells. The kitten cell responses are not modulated and mimic the responses of the adult cells under optical defocus or blockade of inhibitory transmitter inputs.

Role of inhibitory transmitter action on spatial selectivity of the retinal ganglion cells Fig. 2 shows the post stimulus histograms constructed f r o m responses o f O N (left) and O F F (right) ganglion cells to a sinusoidal grating o f 3 cycles per degree. T h e top responses were obtained f r o m an 8-week-old kitten. These show no m o d u l a t i o n in the firing, i.e. they fail to respond to dark/light b o r d e r o f the gratings. As in Fig. 1, the highest spatial frequency o f grating to which kitten cells give a m o d u l a t e d response is m o s t l y lower t h a n 3 cycles per degree.

180 ADULT

KITTEN

IIi.._,._.,llilm ON CELLS

GABA 1OnA

GABA 70nA

IIIIIIIlllIlllIIillll GLYCINE 20nA

GLYCINE 70nA

~~

No. of spikes

25 0

DOPAMINE 5OnA

OFF CELLS

GABA 20nA

DOPAMINE 5OnA

GABA 60nA

llll llllllllilltlI I HI I I ,,IHIIHIhlHII n...1~ n.~,Jl~..

GLYCINE 1OnA

....

.......

GLYCINE 60nA

lil[ll],,i,l,,tl[llitIl/lIIlfll DOPAMINE 5OnA

DOPAMINE 5OnA

181 The cells from the adult cat (Fig. 2, second line), on the other hand, show good modulation of firing, but the modulation disappears when the grating is defocused by - 8 D lens (Fig. 2, third line). This suggests that the modulation of firing is an indication of the cell's property of contrast-dependent spatial selectivity. Thus optical defocus makes the cell amblyopic, mimicking the immature cell's response (Fig. 2, top line). We could also abolish the modulation of firing of the adult ON cell by an iontophoretic application of bicuculline (GABA receptor antagonist), and that of the OFF cell by strychnine (glycine receptor antagonist). Strychnine has no effect on the ON cells and bicuculline no effect on the OFF cells (Fig. 2, fourth and fifth lines). The results were confirmed in nine ON cells and 12 OFF cells in the adult area centralis. Thus naturally-released GABA on ON cells and naturally-released glycine on OFF-cells seem to play an important role for the contrast-dependent spatial resolution of ganglion cells.

Postnatal development of inhibitory transmitter action The next question was whether the inhibitory transmitter action developed postnatally and whether that development played a part in establishing the adult level of contrast-dependent spatial resolution of these cells. Fig. 3 compares the responses of adult ON and OFF cells with those of kitten ON and OFF cells to iontophoretically applied different inhibitory transmitter agonists: GABA, glycine and dopamine. All cells were classified as sustained-X and were encountered within 5~ of the visual axis. Each upward deflection registered the number of spikes generated by the cell in response to an optimal spot at the receptive field center: bright spot to ON and dark spot to OFF cells. ON and OFF cells of the adult visual axis seemed to have only one type of receptor. ON cells had a GABA receptor and OFF-cells a glycine receptor. The kitten cells were receptive to all three substances, GABA, glycine, and dopamine, although ON-cells were more sensitive to GABA than to glycine or dopamine while OFF-cells were more sensitive to glycine than to GABA or dopamine. Fig. 4 shows the mean iontophoretic current levels of each drug that produced a 50% reduction in the receptive field center response of cells in the area centralis. The values were derived from doseFig. 3. Comparison of iontophorectically-applied GABA, glycine, and dopamine on ON and O F F retinal ganglion cells at the visual axis in a cat aged 20 weeks (left) and a kitten aged 8 weeks (right). The pen recorder tracings show the count of spikes from cells responding to an optimal spot flashing repeatedly at the receptive field center. Length of each trace, 1.5 minutes.

182 CO

CO

GO

O0

g

m

o~

AduR

Kitten

= o -lo o

.9

._.9.

ON-cells

OFF-cells

GABA

ON-cells

OFF-cells

glycine

ON-cells

OFF-cells

dopamine

Fig. 4. Mean iontophoretic currents of GABA, glycine, and dopamine needed to produce 50%

inhibition of response in ON and OFF retinal ganglion cells responding to optimal spot presented in the receptive field center. All cells were encountered in the visual axis of adult cats and kittens. Note that GABA is the correct transmitter for ON and glycine for OFF cells. The kitten cells required a higher current than adult cat cells. The transport numbers of each drug shown are taken from Stone [9]. response curves o f each drug obtained f r o m kitten ON, adult ON, kitten O F F , and adult O F F cells. The numbers o f O N and O F F cells studied were 22 and 13 for G A B A , 7 and 25 for glycine, and 10 and 8 for dopamine, whereas those for kitten cells were 16 and 10 for G A B A , 8 and 25 for glycine, and 8 and 9 for dopamine. In addition, the sensitivity o f kitten O N cells to the correct transmitter, G A B A , was much lower (effective current (95% inhibition) 81.9 + 12.3 nA, n = 16) than that o f the adult O N cells (effective current 16.0 + 2.5nA, n = 22). Similarly the sensitivity o f kitten O F F cells to their correct transmitter, glycine, was m u c h lower (effective current 76.6 + 12.0 nA, n = 25) than that o f adult O F F cells (effective current 21.8 + 2.3 nA, n = 25). The receptive field center response o f the adult O N cell was virtually abolished by G A B A 10nA and that o f the kitten O N cell by G A B A 7 0 n A (Fig. 3). Similarly, the adult O F F cell's response was virtually abolished by 1 0 n A and the kitten O F F cell's response by 60 nA o f glycine. The actions o f the antagonists for G A B A , glycine, and d o p a m i n e D1 and D2 receptors, furthermore, distinguished kitten and adult cells at the visual axis. As reported by Ikeda and Robbins [8], although both G A B A and

183 glycine influenced both kitten ON and O F F cells, their antagonists' actions were already adult-like at the age of 7-9 weeks: bicuculline only blocked the inhibition of ON but not of O F F cells, and strychnine only blocked the inhibition of O F F but not ON cells in kittens. Haloperidol (the antagonist for D1 and D2 receptors), ~-flupenthixol (the potent antagonist for D1 receptors), and L-sulpiride (the selective antagonist for D2 receptors), on the other hand, could have had an excitatory effect on both ON and O F F kitten cells, suggesting that naturally-released dopamine was present on D1 as well as D2 receptors on kitten ganglion cells. However, no antagonism of endogenous dopamine was found in adult ganglion cells at the visual axis. The full accounts of dopaminergic actions on the ganglion cells in the adult cat and kitten cells have been described elsewhere [11,12].

Discussion

Contrast-dependent and spatially selective electrical responses of the visual system originate in the center/surround type of receptive fields of depolarizing and hyperpolarizing bipolar cells. ON and O F F center ganglion cells are fed by the depolarizing and hyperpolarizing bipolar cells respectively. At the retinal ganglion cell level, the concept of depolarizing/hyperpolarizing cells or that of O N / O F F cells needs to be questioned. OFF center ganglion cells in fact depolarize and generate spikes to a stimulus darker than background at the receptive field center. Similarly, ON center ganglion cells hyperpolarize and produce reduced spikes to a stimulus darker than background at the receptive field centre. At the receptive field surround, the opposite response occurs for each type. Thus at the ganglion cell level a bright flashing light spot on a dark background is no longer an ideal stimulus. In this study we first demonstrated that such contrast-dependent and spatially-selective receptive fields of ON and O F F ganglion cells of the visual axis developed postnatatly to reach the adult level of spatial resolution (Figs. 1 and 2). Adult cells can distinguish between a light and dark period of grating of higher spatial frequency by modulated firing much better than kitten cells can. However, the adult cell's fine spatial contrast discriminating property diminishes (i.e. becomes amblyopic) when an inhibitory transmitter antagonist is iontophoretically applied to the cell (Fig. 2). In other words, these antagonists mimic the kitten cell's response or the response of adult cells to stimulus defocus (Fig. 2). More importantly, the antagonist which produces amblyopia in ON cells and that which produces amblyopia in O F F cells is different. Bicuculline (GABA receptor antagonist) produces amblyopia in ON cells while strych-

184 Immature cell presyraptlc

postsynaptic

Mature cell presynaptic

postsynaptic

ON-Cell ~

Q

OFF-Ceil( Q

Dopamine~

GABA , ~ Dopamine~

Fig. 5. Diagrams showing postnatal development of inhibitory transmitter synapse distinguishing ON and OFF retinal ganglion cells at the visual axis: multiple innervation, weak transmitter release, and lower receptor efficiencycharacterize immature cells, while elimination of excess terminals with an increase in release and in receptor efficiencyof appropriate transmitter characterize mature cells.

nine (glycine receptor antagonist) produces amblyopia in O F F cells. Thus, G A B A inputs from GABAergic amacrine ceils on ON and glycine inputs from glycinergic amacrines on O F F cells seem to play an important role in the production of the contrast-dependent spatially selective response of these cells. We have found, furthermore, that kitten retinal ganglion cells show poor contrast sensitivity and spatial resolution and their selectivity and sensitivity to inhibitory transmitters which are used to distinguish between ON and O F F ganglion cells are not fully developed (Figs. 3 and 4). Adult O N cells respond only to G A B A but with extremely high sensitivity. Kitten O N cells, on the other hand, respond to G A B A with lower sensitivity than in the adult, but are also sensitive to glycine as well as dopamine. Similarly, adult O F F cells respond only to glycine with a very high sensitivity, while kitten O F F cells are influenced by all three substances, although their sensitivity to glycine is higher than to the other two agents. The sensitivity of kitten ON cells to G A B A (correct transmitter) and O F F cells to glycine (correct transmitter) is also still poorer than that of adult cells. It may be appropriate at this point to refer to the observation that the responses of adult O N and O F F cells to G A B A and glycine in the peripheral retina resemble those of kitten cells in the area centralis. Peripheral retinal

185 ganglion cells with poor spatial resolution do not show the high selectivity and sensitivity to GABA and glycine, as demonstrated for GABA by Bolz et al. [13] and for both GABA and glycine by Priest et al [14]. Thus it appears that the postnatal development of GABA and glycine synapses at the ganglion cells is restricted to the visual axis where the image analysis of an object of interest is habitually made. The development of chemical synapses which separate ON and OFF ganglion cells in the visual axis is described diagrammatically in Fig. 5. Both ON and OFF cells initially receive inputs from glycinergic, GABAergic, and dopaminergic amacrines, but the amount of release and the synaptic efficiency are still poor. In the course of development, however, only GABA inputs to ON and glycine inputs to OFF cells remain. Furthermore, while the efficiency of the appropriate synapses (i.e. GABA for ON and glycine for OFF cells) increases, inappropriate transmitter inputs are eliminated. Thus the principles of development are a) to eliminate excess terminals and inappropriate transmitters and b) to increase the efficiency of synapse involving a functionally important transmitter in order to achieve the adult level of specificity and sensitivity of the neuron. Although such events have never been demonstrated in the retina, nor for inhibitory transmitter synapses in any other part of the central nervous system, similar developmental principles have been described for the postnatal development of cholinergic synapses in sympathetic ganglia and in the neuromuscular junction [15,16]. We would like to stress two clinically relevant points. First, since the postnatal development of transmitter selectivity and efficiency of the retinal cells parallel the maturation of contrast sensitivity and spatial resolution, the possibility of a neurochemical component in developmental amblyopia cannot be ruled out. Second, the finding that immature retinal cells contain multiple transmitter terminals and receptors implies that any drug actions on the retina are expected to be different in neonates from those in adults.

Acknowledgement This work was supported by grants from the Medical Research Council, Special Trustees of St. Thomas' Hospital, University of London and the Smith-Kline Foundation.

References 1. Ikeda H. Physiological basis of visual acuity and its development in kittens. Child Care Health Devel 1979; 5:375-383

186 2. Hamasaki DI, Flynn JT. Physiological properties of retinal ganglion cells of 3-week-old kittens. Vision Res 1977; 17:275-284 3. RusoffAC, Dubin MW. Development of receptive field properties of retinal ganglion cells in kittens. J Neurophysiol 1977; 40:1188-1198 4. Hamasaki DI, Sutija VG. Development of X- and Y-cells in kittens. Exp Brain Res 1979; 35:9-23 5. Nelson R, Famiglietti EV, Kolb H. Intracellular staining reveals different levels of stratification for on-centre and off-centre ganglion cells in the cat retina. J Neurophysiol 1978; 41:472-483 6. Ikeda H, Sheardown MJ. Transmitters mediating inhibition of ganglion cells in the cat retina: iontophoretic studies in vivo. Neuroscience 1983; 8:837-853 7. Ikeda H, Sheardown MJ. Functional transmitters at retinal ganglion cells in cats. Vision Res 1983; 23:1161-1175 8. Ikeda H, Robbins J. Postnatal development of GABA- and glycine-mediated inhibition of feline retinal ganglion cells in the area centralis. Devel Brain Res 1985; 23:1-17 9. Stone TW. Microiontophoresis and pressure ejection. IBRO Handbook Series Methods in the neuroscinces, vol 8, New York: John Wiley & Sons, 1985:50-51 10. Fussey IF, Kidd C, Whitman JG. The differentiation of axonal and soma-dendritic spike activity. Pfliiger's Arch 1970; 321:283-292 11. Ikeda H, Priest TD, Robbins J, Wakakuwa K. Silent dopaminergic synapses at feline retinal ganglion cells? Clin Vision Sci 1986; 1:25-38 12. Ikeda H, Robbins J, Wakakuwa K Evidence of dopaminergic innervationon kitten retinal ganglion cells. Devel Brain Res 35:83-89 13. Bolz J, Frumkes T, Voigt T, W~issleH Action and localisation of y-aminobutyric acid in the cat retina. J Physiol 1985; 362:369-393 14. Priest TD, Robbins J, Ikeda H. Theaction of inhibitory neurotransitters, y-aminobutyric acid and glycine may distinguish between the area centralis and the peripheral retina in cats. Vision Res 1985; 25:1761-1770 15. O'Brien RAD, Ostberg AJC, V~bova G. Observations on the elimination of polyneuronal innervation in developing mammalian skeletal muscle. J Physiol 1978; 282:571-582 16. Hirst GDS, McLachlan EM. Post-natal development of ganglia in the lower lumbar sympathetic chain of the rat. J Physiol 1984; 349:119-134

Address for offprints." Vision Research Unit of Sherrington School, The Rayne Institute, St. Thomas' Hospital, London SE1 7EH, UK.