- Email: [email protected]
Displaced small amacrine cells in the retina of the marine teleost Callionymus lyra L
IWon Res. Vol. 27, No. 9, pp. 1431-1443, 1987 Printed in Great Britain. All rights rcservcd
Copyright 0
0042~6989/87 $3.00 + 0.00 1987 Pergamon Journals Ltd
DISPLACED SMALL AMACRINE CELLS IN THE RETINA OF THE MARINE TELEOST CAL;LIONYMU2? LYRA L. E.
VAN HAESENDONCK
and L. MISSOTTEN
Eye Research Laboratory, K. LJ. Leuven, Unive~itair
Ziekenhuis St. Rafa& B-3000 Leuven, Belgium
(Received 4 March 1986; in revised form 24 February 1987)
displaced small amacrine cells (DSA cells) in the dorsal pure cone part of the retina of the marine teleost Callionym~ lyra have been analysed in a combined light and electron microcopical study. These unistmtifi~ cells have their dendritic ar~~~tion at 70% of the depth of the inner plexiform layer (PS level). The DSA cells constitute a dense population and have variable dendritic field sizes. The bipolar input occurs in the P5,l and the P5,2 pattern layer. The short, central DSA dendrites make ribbon synapses with midget mixed di-cone bipolar cells and with two types of pure cone bipolar cells. The amacrine input and output occurs in the fibrous layer that separates both pattern layers. The dendritic arborization is most extensive and the dendrites of neighbouring DSA cells are interconnected. The thick, central DSA dendrites are presynaptic to adjacent DSA cells and possibly to large bistratified and diffuse ganglion cells. The fine, peripheral DSA dendrites receive input from nei~bo~ng DSA cells and probably from large uni- and bistratified and diffuse amacrine cells. A matching population of regularly placed small amacrine cells (RSA alls) has been observed. Their unistratified dendritic arborization is situated at 20% of the depth of the inner plexiform layer. The synaptic relations of RSA cells have not yet been completely analysed in detail. However, results up till now indicate that they most probably receive input from two bipolar cell types, one of which may be a pure cone type. In addition, the large bistratified amacrine and ganglion cells may be synaptically connected to the RSA cells as well as to the DSA cells. Abstract-The
Teleost retina
Displaced small amacrine cells
INTRODUCTION
Synaptic connections
ber of cone photore~ptor cells and a consequent thickening of the retina was observed. The retina of the marine teleost dragonet (CfzNionymus lyru L) consists of a ventral mixed Here, the inner plexiform layer has a clearly part and a dorsal pure cone part (Vilter, 1947; stratified appearance due to the alternation of Vrabec, 1966; Van Haesendonck and Missotten, six fibrous layers (Fl to F6) and five pattern 1979). The retinal neurons show a remarkable layers (Pl to P5) in which the synaptic buttons geometrical organization. The cone photo- of bipolar axons are grouped according to three receptor cells have a square pattern arrange- distinct square patterns: the a pattern measures ment and a similar regular disposition of neural 5 pm, the fl pattern and the y pattern measure 7 pm. The fl and y pattern show a 45” rotation components is observed in the outer plexiform layer. The square pattern arrangement also ex- with respect to the orientation of the a pattern tends to the neural ~m~nents of the inner (Van Haesendonck and Missotten, 1983a). The plexiform layer particularly in the dorsal pure synaptic buttons are interconnected and the cone part of the retina (Vrabec, 1966; Van morphological features of these contact sites suggest direct synaptic interactions between Haesendonck and Missotten, 1983a). similar as well as dissimilar bipolar axons. ReThe regular square pattern arrangement served as a useful guide for the detailed analysis ciprocal ribbon synapses and morphologically of the neural interconn~tions. Previously, we mixed synapses and synaptic complexes have described the synaptic contacts of the four been described (Van Haesendonck and Misphotoreceptor cell types with one rod and three sotten, 1983b). The interconnections between bipolar axons cone horizontal cell types (Van Haesendonck and Missotten, 1979) and with five mixed and in the inner plexiform layer are associated with four pure cone bipolar cell types (Van Hae- synapses from and onto various types of amasendonck and Missotten, 1984). In the dorsal crine and ganglion cells. We focused our attenpure cone retina a sixfold increase of the num- tion on a small, unistratified cell type with a cell “R
27p-c
1432
E.
VAN
HAESENDONCK and L. MISSO-ITEN
body situated in the ganglion cell layer and with a dentritic arborization at the most proximal pattern layer (PS-level) in the inner plexiform layer. We call this cell the displaced small amacrine cell (DSA cell) for reasons that will be demonstrated in this paper. This cell type has previously been reported in the dragonet retina by Vrabec (1966). Displaced amacrine cells were first mentioned by Cajal (1893; 1933). He assumed that displaced amacrine cells are present in all vertebrate classes. More recently, amacrine cells with cell body lying in the ganglion cell layer have been reported in the retinas of several species, such as the goldfish (Tumosa and Stell, 1986), the newt (Ball and Dickson, 1983), the pigeon (Bingelli and Paule, 1969; Hayes, 1984), the rat (Perry, 1979; Perry and Walker, 1980; Perry, 1981), the rabbit (Hughes and Vaney, 1980; Vaney et al., 1981; Famiglietti, 1983a, 1983b; Vaney, 1984; Famiglietti, 1985), the cat (Boycott and WBssle, 1974), the ground squirrel (West, 1976) and the cow (Osborne, 1984). The light microscopical studies reveal that displaced amacrine cells represent a substantial number of the neurons in the ganglion cell layer. Often they have a unistratified dendritic arborization located in the proximal half of the inner plexiform layer. Usually a matching amacrine cell type with cell body in the inner nuclear layer and unistratified dendritic arborization in the distal half of the inner plexiform layer has been observed, for example in the pigeon retina (Hayes, 1984) and in the rabbit retina (Vaney et al., 1981; Famiglietti, 1983a, 1983b). In the retina of the rat up to six different types of amacrine cells with a counterpart displaced to the ganglion cell layer have been described (Perry and Walker, 1980). The present paper reports on a combined light and electron microscopical study of the displaced small amacrine cell type in the ganglion cell layer, particularly in the dorsal half of the dragonet retina. The identification as amacrine cell is supported by the ultrastructural features of the synaptic contacts. MATERIAL
AND METHODS
teleost dragonet (Callionymus living predator fish. The animals for this study, were caught in the North Sea and killed by transection of the spinal cord. Immediately after enucleation the eyes were fixed in a freshly prepared ice-cold mixture of 2.5% glutaraldehyde in 0.05 M phosphate The marine
lyra L.) is a bottom
buffer (pH 7.3). Penetration of the fixative was facilitated by making a section at the equator of the eye ball. After isolation, the retinas were flattened on a glass slide and covered with a filter paper. They were postfixed in a 2% OsO,-solution in 0.05 M phosphate buffer (pH 7.4) for 2h at 4°C. Subsequently the retinas were processed according to a rapid Golgi method (Colonnier modification). After a quick rinse in distilled water, they were kept in a mixture of 3% K,Cr,O, (4 parts) and 2.5% glutaraldehyde (1 part) at room temperature in the dark for 72 hr. The excess dichromate was washed away in 2 rinses of a 0.75% AgNO,-solution and the retinas were then stored in a fresh 0.75% AgNOrsolution at room temperature in the dark for 96 hr. The retinas were dehydrated over a series of increasing alcohols and after passing through propylene oxide embedded in epoxyresin. DSA cells were selected from light microscopic examination of 60 pm radial and tangential sections made according to the technique of West (1972). For the electron microscopical investigation the combined Go&EM technique was used (Stell, 1965). Ultrathin tangential serial sections were stained with lead citrate for 8 min (Venable and Coggeshall, 1965). RESULTS
(1) General morphology of displaced small amacrine cells
The DSA cell in the inner plexiform layer is a very frequently impregnated cell type in the dorsal as well as in the ventral half of the dragonet retina. Figure 1 shows neighbouring Golgi-impregnated cells observed during light microscopical examination of 60 pm radial sections of the dorsal pure cone part. The oval cell body is situated in the upper row of the ganglion cell layer. The minor axes vary between 4 and 7 pm and the major axes between 5 and 10 pm. No axon running from the cell body into the optic fibre layer has been observed. From the distal pole of the cell body leaves a dendritic stem that mounts to the most proximal pattern layer (P5 level) at about 70% of the depth of the inner plexiform layer (Van Haesendonck and Missotten, 1983a). At this level, the dendritic trunk ramifies into a small unistratified but dense and brush-like dendritic arborization. The small displaced amacrine cells have elliptical dendritic fields. The minor axes range from 3 to 20 pm, the major axes range from 10 to 30 pm.
Fig. f . Photomontage of two light micrographs of Golgi impregnated small amacrine cells in the dragonet retina. The P levels with square pattern arrangement of bipolar synaptic buttons in the inner plexiform layer are indicated. The regular amacrine (RSA) has a oell body in the inner nuclear layer (INL) and a unistratified dendritic arborimtion at the PX level. Several displaced amacrines (DSA) are shown with their cell body in the ganglion cell layer (GCL) and a unistratified dendritic arborization at the P5 level. Bar: IO&m; section thickness: 6Opm.
Fig. 2, Electron micrograph of a tangential section through a BPS,lp button. An impregnated DSA process occupies a postsynaptic position at a ribbon ’ synapse (arrow). Bar: 1 pm.
XUd
xdeu6s uoqqp e u! uollnq qod!q ‘uollnq PI‘SdB = @noJql uo!l=s
1 :lEa
‘(M0i.W)
+I s~~elum ssxad vsa e JO ydedoqu
[e!%'%w
pa~eu&xdtu! uv uoqaqz~ ‘p %J
1435
Fig. 7. Electron mkrograph
of a tangential section through the distal part of the fibrous layer between tbe PS.1 and the PQkvel. An amac&e cell with clear cytoplasm and low synaptic vesick density (AC) makes a conventional synapse to an impregnated DSA process (arrow). Bar: 1 pm.
Fig. 8. Ekctron micrograph of a tangential section through the proximal part of the fibrous layer. showing a large process whose siz. shape and situation allow its idcntification as a DSA dendrite (DSA, outlined by a dotted line). This process makes a conventional synapse to an imprrpnated process of a neighbooring DSA cell (arrow). The DSA synapses o&n show globular accmnulation of presynaptic ekctron dense material (arrow-heads). Bar: I pm.
Fig. 9. Ekctron micrograph of an impregnated DSA process with size and shape comparable to that of the DSA process in Fig. 8. The arrow indicates a contact site consisting of a regular, enlarged intercellular cleft, t&d with electron dense material. Such contact sites are considered being conventional synapses onto fine DSA processes. Bar: 1 pm. The inset shows a larger magnification of the contact site. 1436
1437
Displaced small amacrinc cells in the Cdionymus retina Table 1. Dendritic field size of the nine displaced small amacrine cells and review of the types and number of bipolar synaptic buttons contacted by each DSA cell No.
BP5.2b BP5.2s Dendritic field BP5,lp BPS,ld size in pm’ Sfirn square 7gm square Sgm square 7pm square
1
580
2
3
23 4 5 6
176 125 113 158 75
: ;
; 9
280 154 179
1
:
-1 1 -
43(4) 4 4 3
1 1 1
21 1
-
z 4
-1 1
pattern layer allows unambiguous determination of the buttons in spite of the somewhat poor preservation of the Golgi impregnated tissue. Bipolar input. The bipolar input to the DSA cells occurs in the P5,l and the P5,2 layer. The P5,l layer contains pale buttons with low synaptic vesicle density, organized according to the 5 pm square (BP5,lp buttons) and dark buttons with high synaptic vesicle density, arranged according to one of the 7pm squares (BP5,ld buttons). In the P5,2 layer, big buttons with a 5 pm square arrangement (BP5,2b buttons) and small buttons with a 7 pm square arrangement (BP5,2s buttons) are found. Table 1 is a review of the types and numbers of bipolar buttons that have been observed in contact with each of the nine analysed DSA cells. The corresponding size of the DSA dendritic field area is also given. The BP5,lp bipolar buttons each make one ribbon synapse with a DSA dendrite (Fig. 2). Occasionally, impregnated DSA processes that receive input from amacrine cells adjacent to the (II) Synaptic connectionsof displacedsmall ama- BP5,lp button have been observed (Fig. 3). crine cells These amacrine cells have a clear cytoplasm and Nine DSA cells of the dorsal retina have been a low synaptic vesicle content. analysed in detail on ultrathin tangential serial BP5,ld buttons have been observed making sections. Electron microscopical investigations synaptic contacts with three of the nine analysed reveal synaptic connections with bipolar, ama- DSA cells. Here too, the dendritic endbuds are crine and possibly ganglion cells. The short, connected to the bipolar buttons at only one central dendrites of the dendritic field contact ribbon synapse (Fig. 4). four types of bipolar synaptic buttons. The The BP5,2b bipolar buttons are the most longer dendrites make synapses with neigh- numerous to make synaptic contacts with the bouring DSA cells and other amacrine cells. dendrites of each analysed DSA cell. Figure 5 DSA synapses onto ganglion cells could not be shows an example. The DSA dendritic endbuds unequivocally identified at present, but they are postsynaptic in one to three synaptic ribbon may exist as well. These interconnections are complexes in each BP5,2b button. situated in three sublayers at the P5 level, BP5,2s bipolar buttons give input to six of the namely the P5,l and the P5,2 layer with geo- nine examinated cells. Figure 6 shows such a metrical pattern and a thin fibrous layer that synapse. The DSA dendritic endbuds occupy separates both pattern layers. The constant pos- postsynaptic positions in one or two ribbon ition of the bipolar synaptic buttons in the synapses of the BP5,2s buttons. Table 1 shows
Examination of tangential 60pm sections shows that in a given field all major axes are oriented in the same general direction. The dendritic field size of the analysed cells is ranging from 75 to 580 c(m’, with a mean of 204 pm*. Small and large fields occur at all locations in the dorsal retina. As is seen on Fig. 1, the spacing between the dendritic field centres is very narrow. This spacing was always found to be about 5 pm. Thus, the DSA cells show a constant arrangement according to the 5 pm square pattern and they have an approximative density of 40,000 cells/ mm*. Our observations indicate that most, if not all cell bodies in the outermost position in the ganglion cell layer belong to DSA cells. We estimate that these cells represent about one fourth of the total number of cell bodies in the ganglion cell layer. Consequently, there is a considerable amount of dendritic overlap. The mean factor of overlap for the DSA dendritic fields in the dorsal retina is 7 times.
1438
E.
VAN
HAXSENDONCK and L. MISSOTTEN
that, as a rule, DSA cells contact either BP5,2s buttons or BP&Id buttons. Only one of the analysed DSA cells contacts both types and one DSA cell did not show any synaptic connections with these bipolar axonal buttons. Light microscopy of Golgi imprecated retinas reveals three types of bipolar cells with endbuds at the PS level. One identification is certain, The BP5,2b button belongs to the midget d&cone type that, besides rods in the ventral part, contacts a pair of double cones and one additional member of a neighbou~ng pair. The contacts are wide cleft superficial junctions (Van Haesendonck and Missotten, 1984). This bipolar axon also has synaptic buttons situated at the P3 and the P4 level (respectively at 40 and 55% of the depth of the inner plexiform layer). The identification of two pure cone bipolar cell types from which the DSA cells receive synaptic input is still tentative. Results up till now indicate that they probably are a mono- and a di-cone bipolar. Our previous results show that pure cone bipolars contact cones exclusively through narrow cleft superficial junctions (Van Haesendonck and Missotten, 1984). The axon of one type has a single synaptic endbud situated in the P5,l layer. The axon of the other type has a synaptic button located at the P3 level and a synaptic button in the P5,l layer. The exact origin of the BP5,2s bipolar button has not yet been determined. Amacrine input and output. The most extensive part of the DSA dendritic arborization is found in the thin fibrous layer that separates the P5,l and the P5,2 sublayer, The large majority of the amacrine input to the DSA cells and the DSA output occurs in this layer. The fine DSA processes are found to be postsynaptic to large amacrine cells as well as to neighbouring DSA cells. The thick DSA processes are presynaptic to neighbouring DSA cells and possibly also to large ganglion cells. The longest branches of the DSA cells extent to the distal zone of the fibrous layer in which many large processes with clear cytoplasm and a relatively low to moderate synaptic vesicle content are found. Figure 7 shows an example of a fine impregnated DSA process in the postsynaptic position at a conventional synapse from such a large amacrine process. Light microscopy of Golgi-imp~~ated 60 pm cross sections reveals two large amacrine cell types that also have a dendritic arborisation at the P5 level. One is unistratified and branches only at this level, the other is bistratified and has a
second dendritic plexus at the PI level (at 20% of the depth of the inner plexiform layer). In addition, some diffuse amacrine cells have dendrites that end at the PS level. These three cell types may be connected to the DSA cells. In the proximal, most dense zone of the fibre layer the overlapping dendrites of neighbouring DSA cells are estimated to count for about 70% of the processes. The DSA cells are extensively interconnected in this zone. Numerous fine impregnated DSA processes are observed to be postsynaptic elements at conventional synapses from processes of variable size and shape, having a moderate to high synaptic vesicle content. Figure 8 shows an example. The synapses are often characterized by the globular arrangement of presynaptic electron dense material (arrow heads). The presynaptic processes probably represent unimpregnated DSA cell dendrites. Figure 9 shows an impregnated DSA process similar to the example given in Fig. 8. A regular, intercellular cleft, filled with electron dense material is seen at the contact site with a tine process (arrow). The presynaptic elements are obscured by the silver deposits in the large process. Contact sites like these most likely represent the output synapses of the large DSA processes. Because of the high density of DSA processes in this zone, most postsynaptic processes must be of DSA cell origin. However, the lack of synaptic vesicles in the majority of the small postsynaptic processes prevents unequivocal judgement as to whether they are amacrine or ganglion cell. Hence, we do not exclude the possibility that some of them belong to ganglion cell that branch at this level. Examination of 60 pm radial sections reveals a bistratified ganglion cells type with dendritic plexuses at the Pl and the P5 level (the mirror image of the bistratified amacrine cell). In addition, diffuse ganglion cells with branches ending at the P5 level are observed and may also be synaptically linked to the DSA dendrites. (III) Matching small amacrine ceils Light microscopy of Golgi stained retinas reveals regular small amacrine cells (RSA cells) which are more or less a mirror image of the DSA cells (Fig. 1). Their oval cell body lies in the most proximal row of the inner nuclear layer. The u~stratifi~ dendritic arborization is situated at 20% of the depth of the inner plexiforms layer (PI level). The elliptical dendritic field also have variable sizes and small and large dendritic fields have been observed at all
Displaced small amacrinecc&
locations in the dorsal retina. Similar to the DSA cells, the RSA cells are closely packed and show extensive dendritic overlap (dendritic stem spacing about 5 pm factor of dendritic overlap is approximately 7 times). At present, the synaptic relations of the RSA cells with bipolar, amacrine and gangiion cells have not yet been analysed in detail. However, our previous studies in the dragonet retina (Van Haesendonck and Missotten, 1983a, 1983b) have shown that only two bipolar types have axonal buttons in the Pl layer. The bipolar synaptic buttons in the Pl, 1 layer have the 5 pm square arrangement, those in the P1,2 layer have a 7 pm square arrangement. Whereas the DSA cells receive input from at least three and probably four bipolar cell types, the RSA cells may receive input from only two bipolar cell types. Light microscopical observations of Golgi impregnated bipolar cells suggest that one of them is a pure cone type. It is important to note that no bipolar cell has been found having axonal synaptic buttons situated both at the Pl and the P5 level. (IV) Summary of results The overall picture shown in the scheme on Fig. 10 summarizes the observations of the present study. The inner plexiform layer is lined at both margins by a quasi continuous layer of small amacrine cells. The regular type (RSA cell) is situated at the distal margin and their mirror image, the displaced type (DSA cell), is found in equal number at the proximal margin. The DSA dendritic stem extends straight to the PS level in the inner plexiform layer. The main dendrite ramifies in short, central branches and thicker, more extensive branches. The short dendrites in the center of the dendritic field reach the P5,l and P5,2 layer, where they receive input at ribbon synapses from bipolar cell buttons that have a square pattern arrangement. In the P5,l layer the DSA cell is connected to an average of one sixth of the BPS,ip buttons in its dentritic field and to one or two BP5,ld buttons. These buttons belong to two types of pure cone bipolar cells, probably a mono- and a d&cone type. In the P5,2 layer, the DSA cell is in contact with an average of one fourth of the BP5,2b buttons, belonging to the midget mixed di-cone bipolar, and to one of two BPS,2s buttons. It is remarkable that, with one exception, the DSA arborization that receives synaptic input from the BP5,2d buttons have no synaptic contact with the BP5,2s buttons, and vice versa. How-
in the Culr~o~y~~ retina
1439
ever, the number of observations may be too low to be sure that this is a general rule. In the fibrous layer between the P5,l and the P5,2 pattern layer, neighbouring DSA cells are interconnected by conventional synapses. The thicker processes are always in presynaptic position, while the finest dendrites occupy the postsynaptic positions. Furthermore, the fine DSA processes receive input probably from large unior bistratified amacrine cells. The thick DSA processes may also give synaptic output to large bistratified ganglion cells. The RSA dendritic stem branches at the Pl level of the inner plexiform layer. The RSA cells may receive input from two bipolar cell types, one of which is a pure cone type. The large bist~tifi~ amacrine and ganglion cells that have a dendritic arborization at the PS level and that may be synaptically linked to the DSA cells have a second arborization at the PI level. They may also be connected to the RSA cells. DISCUSSION Cajal(1933) suggested that there is a distribution of retinal displaced amacrine cells throughout all vertebrate retinas. Recently, they have been described in teleost amphibian, avian and mammalian retinas. The present study confirms the occurrence of displaced amacrine cells in the fish retina. From the results of our morphological analyses it appears that in the dragonet retina, the small unistratified cell type with the cell body located in the outermost zone of the ganglion cell layer and the dendritic arborization at the P5 level in the inner plexiform layer is an amacrine cell. The lack of a clear axon and the synaptic relations made with other neurons in the inner plexiform layer indicate an amacrine cell function for this cell type. The dendritic branches receive input from bipolar cells and from neighbouring DSA cells as well as other amacrine cells. Their output occurs through conventional synapses to neighbouring DSA cells and possibly to ganglion cells. The small displaced amacrine cells represent a dense population of neurons with a 5pm square pattern arrangement. Their number thus corresponds to the number of double cone pairs and to the number of those bipolar cells that are arranged according to the 5 pm square pattern (Van Haesendonck and Missotten, 1983a). The DSA cells constitute about 20-25% of the total number of cells in the ganglion cell layer. This percentage agrees with estimates made for dis-
E. VAN HAESENDONCK and L. MISWITEN
1440
INL
Pi p2 p3
IPL
p, Ps
blstratifled GC
DSA
a
BPs.ts
b Fig. 10. Summary diagram of results. (a) Schema of bipolar amacrine and ganglion cells that may be synaptically linked to DSA and RSA cells. (b) Schema of synaptic connections of a DSA cell. Synapses onto ganglion cells are not included because they have not yet been unequivocally identified. Abbreviations: AC amacrine cell; CB cone bipolar cell; GC ganglion cell; GCL ganglion cell layer; INL inner nuclear layer; IPL inner plexiform layer; MB mixed bipolar cell; P pattern layer in IPL.
placed amacrine cell types in other vertebrate retinas: pigeon displaced amacrine cells 23% (Hayes, 1984); rabbit coronate cells 32%, increasing with excentricity from the visual streak (Vaney et al., 1981); rabbit starburst cells, 20-30% (Famiglietti, 1983a).
In the rat retina several types of displaced amacrine cells have been described (Perry and Waler, 1980). In the dragonet retina there is at present no evidence for more than one type of displaced amacrine cells, but this possibility should not be excluded. For instance, one may
Displaced small amacrinc cells in the Cullionynw retina
expect to find a large field displaced amacrine. Except in the pigeon retina where the majority of the displaced amacrine cells have dendritic field diameters measuring 50-80 pm, all types described in other species have large dendritic fields with diameters over 130 pm (Perry and Walker, 1980; Vaney et al., 1981; Famiglietti, 1983a; Hayes, 1984). In comparison, the now analysed DSA cells have remarkably narrow dendritic fields, with major axes no longer than 30pm. The size of the DSA dendritic fields varies. The dimensions of the minor as well as of the major axes range over 20 pm. An explicit correlation between the dendritic field size and the retinal location could not be established. In the retinas of the pigeon and the rabbit, on the other hand, the dendritic field size of the displaced amacrine cells is related to the eccentricity from the visual streak (Hayes, 1984; Famiglietti, 1985). In these species the retinal organisation is gradually changing from the central fovea1 part to the periphery. In contrast, the dragonet retina shows a uniform spacing of cones throughout the entire dorsal part. This homogenous distribution of the visual cells could explain the lack of a distinct correlation between size and excentricity in DSA cells. However, these cells show varying dimensions of their dendritic fields in any area of the retina. Some types of ganglion cells described in the retinas of the rat and the cat have also been found to lack any correlation between the dendritic field size and the retinal location. (Perry, 1979; Boycott and Wiissle, 1974). The unistratified dendritic arborization of the DSA cell is located at 70% of the depth of the inner plexiform layer, this is at the most proximal geometrically organized P5 level (Van Haesendonck and Missotten, 1983a). This level corresponds to Cajal’s sublamina 4. In the carp retina, Cajal(1893) reported a small type of cells with pyriform cell body in the ganglion cell layer and a fine, compact and unistratified dendritic arborization at the fourth level. He suggested the possibility that they are a displaced amacrine cell type. But Cajal’s limited description does not allow an adequate correlation between these cells and the small displaced amacrine cells in the dragonet retina. In the goldfish retina, a population of displaced amacrine cells that show choline acetyltransferase-like immunoreactivity is found to have their dendritic arborization in or near stratum 4 (Tumosa et al., 1984; Tumosa and Stell, 1986). A good correspondence, however, also exist with the den-
1441
dritic arborization level of the unistratified displaced amacrine cells in the pigeon retina and the unistratified starburst amacrine cells in the rabbit retina. In the pigeon retina, the dendritic plexus is situated at 64% proximal to the inner nuclear layer (Hayes, 1984). In the rabbit retina, the type b starburst amacrine cells ramify at 68% (Famiglietti, 1983a). Our observations indicate that DSA cells are mutually linked. In addition, at least three bipolar and possibly two other amacrine and one ganglion cell type make synapses with the DSA cells. The bipolar input is clearly segregated from the amacrine input and output. The ribbon synapses in bipolar cells occur in two distinct, geometrically arranged layers and are predominantly made by the short, central dendrites of the DSA arborization. The conventional synapses from and onto other amacrine cells or ganglion cells occur in the thin fibrous layer separating the two pattern layers. These synapses are made by the longer dendrites of the arborization. The segregation of synaptic functions in narrow, consecutive layers results in the brush-like appearance of the dendritic tree. The DSA cells are synaptically linked to mixed as well as pure cone bipolar cells. These bipolar cells contact photoreceptor cells with either wide cleft or narrow cleft superficial junctions (Van Haesendonck and Missotten, 1984). Their axons all end at the P5 level (sublamina b) in the inner plexiform layer. Thus, our results do not allow to make any correlation between the type of superficial junction at the photoreceptor-bipolar interconnection and the level in the inner plexiform layer at which the bipolar axon terminal is located. In the dorsal pure cone part of the dragonet retina these bipolars obviously receive only cone input, but in the ventral mixed part they also get rod input. In this respect the displaced small amacrine cells differ from the starburst displaced amacrine cells in the rabbit retina. Famiglietti (1983b) observed interconnections with cone bipolar cells alone. The number of synapsing bipolar cells, ranging from five to eleven, also reflects the varying dimensions of the displaced amacrine cell dendritic fields. Remarkably, seven analysed cells contact either BP5,ld buttons or BP5,2s buttons. Both types of synaptic buttons have a 7pm square pattern arrangement (Van Haesendonck and Missotten, 1983a). Maybe the BP5,2s buttons correspond to a fourth type of
1442
E. VAN
HABENDONCK
bipolar axon ending at this level. However, light microscopy of Golgi impregnated bipolar cells suggest that they may be appendages of the BPS, Id buttons. The mean dendritic overlap factor, calculated to be 7, indicates that at a single point in the dorsal retina, the dendritic fields of seven neighbouring DSA cells interfere. In the dense network of the fibrous layer their processes make numerous conventional synapses upon each other. Centrally in the dendritic field where the branching DSA processes are thick, they give synaptic output to the fine peripheral dendritic branches of neighbouring DSA cells. Thus, the DSA output and the bipolar input occur in the center of the dendritic field. They are surrounded by an annular zone of input from DSA cells and other amacrine cells. Light microscopical investigations revealed two additional types of amacrine cells that have a dendritic plexus at the level of the DSA arborization. Both the unistratified and the bistratifi~ type are potential candidates to synapse on the DSA cells. The large processes that are presynaptic to the DSA cells in the distal part of the fibrous layer all had the same electron microscopical appearance. We have not yet observed any ultrastructural details that allow us to distinguish different types. Fine terminal dendrites of a bistratified ganglion cell type which is the mirror image of the large bistratified amacrine cell may occupy some of the postsynaptic positions in the DSA conventional synapses. Finally, it is also possible that branches of diffuse amacrine and ganglion cells that pass the PS level are synaptically linked to the DSA cells. Further detailed analysis of Golgi-impregnated or immunocytochemically stained examples of these stratified and diffuse amacrine and ganglion cells has to reveal whether they are really connected to the DSA cells. Matching regular and displaced amacrine cells, have been reported in the retinas of several species. In the carp retina, Cajal (1893) described small regular amacrine cells which are very similar to the dragonet RSA cells. They have a thin descendant process, that gives rise to a small compact arborization with varicose branches at the second level. In the goldfish retina, Tumosa et al. (1984) observed choline acetyltransferase-like immunoreactive regular amacrine cells that have their dendritic arborization in or near stratum 2. The second stratum corresponds with the PI level in the dragonet
and L. MISSOTIXN
retina. The level of the RSA dendritic arborization also agrees well with the branching levels observed for the comparable amacrine cell populations in the pigeon retina: 24% (Hayes, 1984) and the rabbit retina: 22% (Famiglietti, 1983a). The existence of two matching amacrine cell populations is generally related to the On-Off lamination of the inner plexiform layer (Perry, 1980; Perry and Walker, 1980; Vaney et al., 1981; Famiglietti 1983a, 1983b; Hayes 1984). The displaced amacrine cells are the type b cells that are involved in the transmission of information through the On pathway, while the matching regular amacrines represent the type a cells that participate in a similar Off pathway. The resembling inte~onn~tions observed for type a and type b starburst amacrine cells in the rabbit retina seem to support this hypothesis (Famiglietti, 1983b). Our results indicate that RSA and DSA cells receive direct input from different bipolar cells. The DSA cells seem to collect a wider range of bipolar information than the RSA cells. The DSA and the RSA cells, however, may have comparable functions. The large bistratified amacrine and ganglion cells that co-stratify with the DSA cells also co-stratify to the RSA cells. This implies that they may be connected to the RSA cells at the Pl level and to the DSA cells at the P5 level. Thus, it is possible that one and the same large amacrine cell type would modify the signals passing through RSA and DSA cells. In addition, these signals perhaps also converge into the same large ganglion cell type. The question remains to what extent the RSA and the DSA cells are truly functionally matching cell types. Further detailed electron microscopical inv~tigation of the RSA cells and the large bistratified amacrine and ganglion cells has to reveal their interconnections in the inner plexiform layer and allow us to answer this question,
REFERENCES Ball A. K. and Dickson D. M. (1983) Displaced amacrine and ganglion cells in the newt retina. Expl Eye Res. 36, 199-213. Bingelli R. L. and Pauie W. J. (1969) The p$eon retina: quantitative aspects of the optic nerve and ganglion cell layer J. camp. Neurol. 137, i-18. Boycott B. B. and Wassle H. (1974) The morphological types of ganglion cells of the domestic cat’s retina. 1. Physiol., Lond. 240, 397-419.
Disphtced small amacrine cells in the C~liony~ Cajal S. R. (1893) La r&tine des vertcbrcs. Lo Celhle 9, 17-257. Translation as appendix in Rodrieck R. W. (1973) The Vertebrate Retina, pp. 775-904. W. H. Freeman, San Francisco, Calif. Famiglietti E. V. Jr (1983a) Starburst amacrinc cells and cholinergic neurons: mirror-symmetric ON and OFF amacrine cells of rabbit retina. Brain Res. 261, 138-144. Famiglietti E. V. Jr (1983b) On and off pathways through amacrine cells in mammalian retina: the synaptic connections of starburst amacrine cells. Vision Res. 23, 1265-1279. Famiglietti E. V. Jr (1985) Starburst amacrine cells: rno~holo~~l constancy and systematic variation in the anisotropic field of rabbit retinal neurons. J. Netrrosci. 5, 562-577. Hayes B. P. (1984) Cell populations of the ganglion cell layer: displaced amacrine and matching amacrine cells in the pigeon retina. Expl Brain Res. S6, 565-573. Hughes A. and Vaney D. I. (1980) Coronate cells: displaced amacrines of the rabbit retina? 3. camp. Neural. 11)9, 169-189. Osborne N. N. (1984) Substance P in the bovine retina: localization, identification, release, uptake and receptor analysis. J. Physiol., Lond. 349, 83-93. Perry V. H. (1979) The ganglion cell layer of the retina of the rat: a Golgi study. Proc. R. Eoc. B 204, 363-375. Perry V. H. (1981) Evidence for an amacrine cell system in the ganglion cell layer of the rat retina. Neuroscience a, 931-944. Perry V. H. and Walker M. (1980) Amacrine cells, displaced amacrine cells and interplexiform cells in the retina of the rat. Proc. R. Sot. B. 208, 415-431. Stell W. K. (1965) Correlation of retinal cytoarchitecture and uIt~st~cture in Golgi p~arations. Amt. Rec. 153, 389-398. Tumosa N., Eckmstein F. and Stell W. K. (1984) Immunocytochemical localization of putative cholinergic neurons in the goldfish retina. Neurosci. Lat. 48, 255-275. Tumosa N. and Stell W. K. (1986) Choline Acetyltransferase immuno~acti~ty suggests that ganglion cells
retina
1443
in the goldfish retina are not cholinergic. J. camp. Neurol. 244, 267-275. Vaney D. I. (1984) ‘Coronate’ amacrine cells in the rabbit retina have the starburst dendritic morphology. Proc. R. Sot. B 228, 501-508. Vaney D. I., Peichl L, and Boycott 8. B. (1981) Matching ~pulations of amacrine cells in the inner nuclear and ganglion cell layers of the rabbit retina. J. camp. Neurol. 199, 373-391. Van Haesendonck E. and Missotten L. (1979) Synaptic contacts of the horizontal cells in the retina of the marine teleost, &aliionymuslyra L. J. romp. Neurol. l&I, 167-191. Van Haesendonck E. and Missotten L. (1983a) Stratification and square pattern arrangements in the dorsal inner plexiform layer in the retina of CuBionymus lyra L. J. Ultrastruct. Res. 83, 296302. Van Haesendonck E. and Missotten L. (1983b) Interbipolar contacts in the dorsal inner plexiform layer in the retina of CaIiionym~ lyra L. .I. Ultr~t~ct. Res. 83, 303-311. Van Haesendonck E. and Missotten L. (1984) Synaptic contacts between bipolar and photoreceptor cells in the retina of Calionymus lyra L. J. camp. Neurol. 223, 387-399. Venable J. H. and Coggeshall R. (1965) A simplified lead citrate stain for use in electron microscopy. J. cetf Biof. 25,407-408. Vilter V. (1947) Dissociation spatiale des champs photosensoriels a cBnes et a batonnets chez un poisson marin, le Callionymuslyra. C.r. Sot. Biol., Paris 141, 344-346. Vrabec F. (1966) A new finding in the retina of a marine teleost, Callionymus lyra f... Foha Morphoi. XW, 143-147. Vrabec F. (1974) Inverted growth of dendrites of the retinal ganglion cells in Caflionymus MyraL (dragonet). Acta anat. 87, 93-96. West R. W. (1976) Light and electron microscopy of the ground squirrel retina: functional considerations. J. camp. Neural. laS, 355-378.
Copyright 0
0042~6989/87 $3.00 + 0.00 1987 Pergamon Journals Ltd
DISPLACED SMALL AMACRINE CELLS IN THE RETINA OF THE MARINE TELEOST CAL;LIONYMU2? LYRA L. E.
VAN HAESENDONCK
and L. MISSOTTEN
Eye Research Laboratory, K. LJ. Leuven, Unive~itair
Ziekenhuis St. Rafa& B-3000 Leuven, Belgium
(Received 4 March 1986; in revised form 24 February 1987)
displaced small amacrine cells (DSA cells) in the dorsal pure cone part of the retina of the marine teleost Callionym~ lyra have been analysed in a combined light and electron microcopical study. These unistmtifi~ cells have their dendritic ar~~~tion at 70% of the depth of the inner plexiform layer (PS level). The DSA cells constitute a dense population and have variable dendritic field sizes. The bipolar input occurs in the P5,l and the P5,2 pattern layer. The short, central DSA dendrites make ribbon synapses with midget mixed di-cone bipolar cells and with two types of pure cone bipolar cells. The amacrine input and output occurs in the fibrous layer that separates both pattern layers. The dendritic arborization is most extensive and the dendrites of neighbouring DSA cells are interconnected. The thick, central DSA dendrites are presynaptic to adjacent DSA cells and possibly to large bistratified and diffuse ganglion cells. The fine, peripheral DSA dendrites receive input from nei~bo~ng DSA cells and probably from large uni- and bistratified and diffuse amacrine cells. A matching population of regularly placed small amacrine cells (RSA alls) has been observed. Their unistratified dendritic arborization is situated at 20% of the depth of the inner plexiform layer. The synaptic relations of RSA cells have not yet been completely analysed in detail. However, results up till now indicate that they most probably receive input from two bipolar cell types, one of which may be a pure cone type. In addition, the large bistratified amacrine and ganglion cells may be synaptically connected to the RSA cells as well as to the DSA cells. Abstract-The
Teleost retina
Displaced small amacrine cells
INTRODUCTION
Synaptic connections
ber of cone photore~ptor cells and a consequent thickening of the retina was observed. The retina of the marine teleost dragonet (CfzNionymus lyru L) consists of a ventral mixed Here, the inner plexiform layer has a clearly part and a dorsal pure cone part (Vilter, 1947; stratified appearance due to the alternation of Vrabec, 1966; Van Haesendonck and Missotten, six fibrous layers (Fl to F6) and five pattern 1979). The retinal neurons show a remarkable layers (Pl to P5) in which the synaptic buttons geometrical organization. The cone photo- of bipolar axons are grouped according to three receptor cells have a square pattern arrange- distinct square patterns: the a pattern measures ment and a similar regular disposition of neural 5 pm, the fl pattern and the y pattern measure 7 pm. The fl and y pattern show a 45” rotation components is observed in the outer plexiform layer. The square pattern arrangement also ex- with respect to the orientation of the a pattern tends to the neural ~m~nents of the inner (Van Haesendonck and Missotten, 1983a). The plexiform layer particularly in the dorsal pure synaptic buttons are interconnected and the cone part of the retina (Vrabec, 1966; Van morphological features of these contact sites suggest direct synaptic interactions between Haesendonck and Missotten, 1983a). similar as well as dissimilar bipolar axons. ReThe regular square pattern arrangement served as a useful guide for the detailed analysis ciprocal ribbon synapses and morphologically of the neural interconn~tions. Previously, we mixed synapses and synaptic complexes have described the synaptic contacts of the four been described (Van Haesendonck and Misphotoreceptor cell types with one rod and three sotten, 1983b). The interconnections between bipolar axons cone horizontal cell types (Van Haesendonck and Missotten, 1979) and with five mixed and in the inner plexiform layer are associated with four pure cone bipolar cell types (Van Hae- synapses from and onto various types of amasendonck and Missotten, 1984). In the dorsal crine and ganglion cells. We focused our attenpure cone retina a sixfold increase of the num- tion on a small, unistratified cell type with a cell “R
27p-c
1432
E.
VAN
HAESENDONCK and L. MISSO-ITEN
body situated in the ganglion cell layer and with a dentritic arborization at the most proximal pattern layer (PS-level) in the inner plexiform layer. We call this cell the displaced small amacrine cell (DSA cell) for reasons that will be demonstrated in this paper. This cell type has previously been reported in the dragonet retina by Vrabec (1966). Displaced amacrine cells were first mentioned by Cajal (1893; 1933). He assumed that displaced amacrine cells are present in all vertebrate classes. More recently, amacrine cells with cell body lying in the ganglion cell layer have been reported in the retinas of several species, such as the goldfish (Tumosa and Stell, 1986), the newt (Ball and Dickson, 1983), the pigeon (Bingelli and Paule, 1969; Hayes, 1984), the rat (Perry, 1979; Perry and Walker, 1980; Perry, 1981), the rabbit (Hughes and Vaney, 1980; Vaney et al., 1981; Famiglietti, 1983a, 1983b; Vaney, 1984; Famiglietti, 1985), the cat (Boycott and WBssle, 1974), the ground squirrel (West, 1976) and the cow (Osborne, 1984). The light microscopical studies reveal that displaced amacrine cells represent a substantial number of the neurons in the ganglion cell layer. Often they have a unistratified dendritic arborization located in the proximal half of the inner plexiform layer. Usually a matching amacrine cell type with cell body in the inner nuclear layer and unistratified dendritic arborization in the distal half of the inner plexiform layer has been observed, for example in the pigeon retina (Hayes, 1984) and in the rabbit retina (Vaney et al., 1981; Famiglietti, 1983a, 1983b). In the retina of the rat up to six different types of amacrine cells with a counterpart displaced to the ganglion cell layer have been described (Perry and Walker, 1980). The present paper reports on a combined light and electron microscopical study of the displaced small amacrine cell type in the ganglion cell layer, particularly in the dorsal half of the dragonet retina. The identification as amacrine cell is supported by the ultrastructural features of the synaptic contacts. MATERIAL
AND METHODS
teleost dragonet (Callionymus living predator fish. The animals for this study, were caught in the North Sea and killed by transection of the spinal cord. Immediately after enucleation the eyes were fixed in a freshly prepared ice-cold mixture of 2.5% glutaraldehyde in 0.05 M phosphate The marine
lyra L.) is a bottom
buffer (pH 7.3). Penetration of the fixative was facilitated by making a section at the equator of the eye ball. After isolation, the retinas were flattened on a glass slide and covered with a filter paper. They were postfixed in a 2% OsO,-solution in 0.05 M phosphate buffer (pH 7.4) for 2h at 4°C. Subsequently the retinas were processed according to a rapid Golgi method (Colonnier modification). After a quick rinse in distilled water, they were kept in a mixture of 3% K,Cr,O, (4 parts) and 2.5% glutaraldehyde (1 part) at room temperature in the dark for 72 hr. The excess dichromate was washed away in 2 rinses of a 0.75% AgNO,-solution and the retinas were then stored in a fresh 0.75% AgNOrsolution at room temperature in the dark for 96 hr. The retinas were dehydrated over a series of increasing alcohols and after passing through propylene oxide embedded in epoxyresin. DSA cells were selected from light microscopic examination of 60 pm radial and tangential sections made according to the technique of West (1972). For the electron microscopical investigation the combined Go&EM technique was used (Stell, 1965). Ultrathin tangential serial sections were stained with lead citrate for 8 min (Venable and Coggeshall, 1965). RESULTS
(1) General morphology of displaced small amacrine cells
The DSA cell in the inner plexiform layer is a very frequently impregnated cell type in the dorsal as well as in the ventral half of the dragonet retina. Figure 1 shows neighbouring Golgi-impregnated cells observed during light microscopical examination of 60 pm radial sections of the dorsal pure cone part. The oval cell body is situated in the upper row of the ganglion cell layer. The minor axes vary between 4 and 7 pm and the major axes between 5 and 10 pm. No axon running from the cell body into the optic fibre layer has been observed. From the distal pole of the cell body leaves a dendritic stem that mounts to the most proximal pattern layer (P5 level) at about 70% of the depth of the inner plexiform layer (Van Haesendonck and Missotten, 1983a). At this level, the dendritic trunk ramifies into a small unistratified but dense and brush-like dendritic arborization. The small displaced amacrine cells have elliptical dendritic fields. The minor axes range from 3 to 20 pm, the major axes range from 10 to 30 pm.
Fig. f . Photomontage of two light micrographs of Golgi impregnated small amacrine cells in the dragonet retina. The P levels with square pattern arrangement of bipolar synaptic buttons in the inner plexiform layer are indicated. The regular amacrine (RSA) has a oell body in the inner nuclear layer (INL) and a unistratified dendritic arborimtion at the PX level. Several displaced amacrines (DSA) are shown with their cell body in the ganglion cell layer (GCL) and a unistratified dendritic arborization at the P5 level. Bar: IO&m; section thickness: 6Opm.
Fig. 2, Electron micrograph of a tangential section through a BPS,lp button. An impregnated DSA process occupies a postsynaptic position at a ribbon ’ synapse (arrow). Bar: 1 pm.
XUd
xdeu6s uoqqp e u! uollnq qod!q ‘uollnq PI‘SdB = @noJql uo!l=s
1 :lEa
‘(M0i.W)
+I s~~elum ssxad vsa e JO ydedoqu
[e!%'%w
pa~eu&xdtu! uv uoqaqz~ ‘p %J
1435
Fig. 7. Electron mkrograph
of a tangential section through the distal part of the fibrous layer between tbe PS.1 and the PQkvel. An amac&e cell with clear cytoplasm and low synaptic vesick density (AC) makes a conventional synapse to an impregnated DSA process (arrow). Bar: 1 pm.
Fig. 8. Ekctron micrograph of a tangential section through the proximal part of the fibrous layer. showing a large process whose siz. shape and situation allow its idcntification as a DSA dendrite (DSA, outlined by a dotted line). This process makes a conventional synapse to an imprrpnated process of a neighbooring DSA cell (arrow). The DSA synapses o&n show globular accmnulation of presynaptic ekctron dense material (arrow-heads). Bar: I pm.
Fig. 9. Ekctron micrograph of an impregnated DSA process with size and shape comparable to that of the DSA process in Fig. 8. The arrow indicates a contact site consisting of a regular, enlarged intercellular cleft, t&d with electron dense material. Such contact sites are considered being conventional synapses onto fine DSA processes. Bar: 1 pm. The inset shows a larger magnification of the contact site. 1436
1437
Displaced small amacrinc cells in the Cdionymus retina Table 1. Dendritic field size of the nine displaced small amacrine cells and review of the types and number of bipolar synaptic buttons contacted by each DSA cell No.
BP5.2b BP5.2s Dendritic field BP5,lp BPS,ld size in pm’ Sfirn square 7gm square Sgm square 7pm square
1
580
2
3
23 4 5 6
176 125 113 158 75
: ;
; 9
280 154 179
1
:
-1 1 -
43(4) 4 4 3
1 1 1
21 1
-
z 4
-1 1
pattern layer allows unambiguous determination of the buttons in spite of the somewhat poor preservation of the Golgi impregnated tissue. Bipolar input. The bipolar input to the DSA cells occurs in the P5,l and the P5,2 layer. The P5,l layer contains pale buttons with low synaptic vesicle density, organized according to the 5 pm square (BP5,lp buttons) and dark buttons with high synaptic vesicle density, arranged according to one of the 7pm squares (BP5,ld buttons). In the P5,2 layer, big buttons with a 5 pm square arrangement (BP5,2b buttons) and small buttons with a 7 pm square arrangement (BP5,2s buttons) are found. Table 1 is a review of the types and numbers of bipolar buttons that have been observed in contact with each of the nine analysed DSA cells. The corresponding size of the DSA dendritic field area is also given. The BP5,lp bipolar buttons each make one ribbon synapse with a DSA dendrite (Fig. 2). Occasionally, impregnated DSA processes that receive input from amacrine cells adjacent to the (II) Synaptic connectionsof displacedsmall ama- BP5,lp button have been observed (Fig. 3). crine cells These amacrine cells have a clear cytoplasm and Nine DSA cells of the dorsal retina have been a low synaptic vesicle content. analysed in detail on ultrathin tangential serial BP5,ld buttons have been observed making sections. Electron microscopical investigations synaptic contacts with three of the nine analysed reveal synaptic connections with bipolar, ama- DSA cells. Here too, the dendritic endbuds are crine and possibly ganglion cells. The short, connected to the bipolar buttons at only one central dendrites of the dendritic field contact ribbon synapse (Fig. 4). four types of bipolar synaptic buttons. The The BP5,2b bipolar buttons are the most longer dendrites make synapses with neigh- numerous to make synaptic contacts with the bouring DSA cells and other amacrine cells. dendrites of each analysed DSA cell. Figure 5 DSA synapses onto ganglion cells could not be shows an example. The DSA dendritic endbuds unequivocally identified at present, but they are postsynaptic in one to three synaptic ribbon may exist as well. These interconnections are complexes in each BP5,2b button. situated in three sublayers at the P5 level, BP5,2s bipolar buttons give input to six of the namely the P5,l and the P5,2 layer with geo- nine examinated cells. Figure 6 shows such a metrical pattern and a thin fibrous layer that synapse. The DSA dendritic endbuds occupy separates both pattern layers. The constant pos- postsynaptic positions in one or two ribbon ition of the bipolar synaptic buttons in the synapses of the BP5,2s buttons. Table 1 shows
Examination of tangential 60pm sections shows that in a given field all major axes are oriented in the same general direction. The dendritic field size of the analysed cells is ranging from 75 to 580 c(m’, with a mean of 204 pm*. Small and large fields occur at all locations in the dorsal retina. As is seen on Fig. 1, the spacing between the dendritic field centres is very narrow. This spacing was always found to be about 5 pm. Thus, the DSA cells show a constant arrangement according to the 5 pm square pattern and they have an approximative density of 40,000 cells/ mm*. Our observations indicate that most, if not all cell bodies in the outermost position in the ganglion cell layer belong to DSA cells. We estimate that these cells represent about one fourth of the total number of cell bodies in the ganglion cell layer. Consequently, there is a considerable amount of dendritic overlap. The mean factor of overlap for the DSA dendritic fields in the dorsal retina is 7 times.
1438
E.
VAN
HAXSENDONCK and L. MISSOTTEN
that, as a rule, DSA cells contact either BP5,2s buttons or BP&Id buttons. Only one of the analysed DSA cells contacts both types and one DSA cell did not show any synaptic connections with these bipolar axonal buttons. Light microscopy of Golgi imprecated retinas reveals three types of bipolar cells with endbuds at the PS level. One identification is certain, The BP5,2b button belongs to the midget d&cone type that, besides rods in the ventral part, contacts a pair of double cones and one additional member of a neighbou~ng pair. The contacts are wide cleft superficial junctions (Van Haesendonck and Missotten, 1984). This bipolar axon also has synaptic buttons situated at the P3 and the P4 level (respectively at 40 and 55% of the depth of the inner plexiform layer). The identification of two pure cone bipolar cell types from which the DSA cells receive synaptic input is still tentative. Results up till now indicate that they probably are a mono- and a di-cone bipolar. Our previous results show that pure cone bipolars contact cones exclusively through narrow cleft superficial junctions (Van Haesendonck and Missotten, 1984). The axon of one type has a single synaptic endbud situated in the P5,l layer. The axon of the other type has a synaptic button located at the P3 level and a synaptic button in the P5,l layer. The exact origin of the BP5,2s bipolar button has not yet been determined. Amacrine input and output. The most extensive part of the DSA dendritic arborization is found in the thin fibrous layer that separates the P5,l and the P5,2 sublayer, The large majority of the amacrine input to the DSA cells and the DSA output occurs in this layer. The fine DSA processes are found to be postsynaptic to large amacrine cells as well as to neighbouring DSA cells. The thick DSA processes are presynaptic to neighbouring DSA cells and possibly also to large ganglion cells. The longest branches of the DSA cells extent to the distal zone of the fibrous layer in which many large processes with clear cytoplasm and a relatively low to moderate synaptic vesicle content are found. Figure 7 shows an example of a fine impregnated DSA process in the postsynaptic position at a conventional synapse from such a large amacrine process. Light microscopy of Golgi-imp~~ated 60 pm cross sections reveals two large amacrine cell types that also have a dendritic arborisation at the P5 level. One is unistratified and branches only at this level, the other is bistratified and has a
second dendritic plexus at the PI level (at 20% of the depth of the inner plexiform layer). In addition, some diffuse amacrine cells have dendrites that end at the PS level. These three cell types may be connected to the DSA cells. In the proximal, most dense zone of the fibre layer the overlapping dendrites of neighbouring DSA cells are estimated to count for about 70% of the processes. The DSA cells are extensively interconnected in this zone. Numerous fine impregnated DSA processes are observed to be postsynaptic elements at conventional synapses from processes of variable size and shape, having a moderate to high synaptic vesicle content. Figure 8 shows an example. The synapses are often characterized by the globular arrangement of presynaptic electron dense material (arrow heads). The presynaptic processes probably represent unimpregnated DSA cell dendrites. Figure 9 shows an impregnated DSA process similar to the example given in Fig. 8. A regular, intercellular cleft, filled with electron dense material is seen at the contact site with a tine process (arrow). The presynaptic elements are obscured by the silver deposits in the large process. Contact sites like these most likely represent the output synapses of the large DSA processes. Because of the high density of DSA processes in this zone, most postsynaptic processes must be of DSA cell origin. However, the lack of synaptic vesicles in the majority of the small postsynaptic processes prevents unequivocal judgement as to whether they are amacrine or ganglion cell. Hence, we do not exclude the possibility that some of them belong to ganglion cell that branch at this level. Examination of 60 pm radial sections reveals a bistratified ganglion cells type with dendritic plexuses at the Pl and the P5 level (the mirror image of the bistratified amacrine cell). In addition, diffuse ganglion cells with branches ending at the P5 level are observed and may also be synaptically linked to the DSA dendrites. (III) Matching small amacrine ceils Light microscopy of Golgi stained retinas reveals regular small amacrine cells (RSA cells) which are more or less a mirror image of the DSA cells (Fig. 1). Their oval cell body lies in the most proximal row of the inner nuclear layer. The u~stratifi~ dendritic arborization is situated at 20% of the depth of the inner plexiforms layer (PI level). The elliptical dendritic field also have variable sizes and small and large dendritic fields have been observed at all
Displaced small amacrinecc&
locations in the dorsal retina. Similar to the DSA cells, the RSA cells are closely packed and show extensive dendritic overlap (dendritic stem spacing about 5 pm factor of dendritic overlap is approximately 7 times). At present, the synaptic relations of the RSA cells with bipolar, amacrine and gangiion cells have not yet been analysed in detail. However, our previous studies in the dragonet retina (Van Haesendonck and Missotten, 1983a, 1983b) have shown that only two bipolar types have axonal buttons in the Pl layer. The bipolar synaptic buttons in the Pl, 1 layer have the 5 pm square arrangement, those in the P1,2 layer have a 7 pm square arrangement. Whereas the DSA cells receive input from at least three and probably four bipolar cell types, the RSA cells may receive input from only two bipolar cell types. Light microscopical observations of Golgi impregnated bipolar cells suggest that one of them is a pure cone type. It is important to note that no bipolar cell has been found having axonal synaptic buttons situated both at the Pl and the P5 level. (IV) Summary of results The overall picture shown in the scheme on Fig. 10 summarizes the observations of the present study. The inner plexiform layer is lined at both margins by a quasi continuous layer of small amacrine cells. The regular type (RSA cell) is situated at the distal margin and their mirror image, the displaced type (DSA cell), is found in equal number at the proximal margin. The DSA dendritic stem extends straight to the PS level in the inner plexiform layer. The main dendrite ramifies in short, central branches and thicker, more extensive branches. The short dendrites in the center of the dendritic field reach the P5,l and P5,2 layer, where they receive input at ribbon synapses from bipolar cell buttons that have a square pattern arrangement. In the P5,l layer the DSA cell is connected to an average of one sixth of the BPS,ip buttons in its dentritic field and to one or two BP5,ld buttons. These buttons belong to two types of pure cone bipolar cells, probably a mono- and a d&cone type. In the P5,2 layer, the DSA cell is in contact with an average of one fourth of the BP5,2b buttons, belonging to the midget mixed di-cone bipolar, and to one of two BPS,2s buttons. It is remarkable that, with one exception, the DSA arborization that receives synaptic input from the BP5,2d buttons have no synaptic contact with the BP5,2s buttons, and vice versa. How-
in the Culr~o~y~~ retina
1439
ever, the number of observations may be too low to be sure that this is a general rule. In the fibrous layer between the P5,l and the P5,2 pattern layer, neighbouring DSA cells are interconnected by conventional synapses. The thicker processes are always in presynaptic position, while the finest dendrites occupy the postsynaptic positions. Furthermore, the fine DSA processes receive input probably from large unior bistratified amacrine cells. The thick DSA processes may also give synaptic output to large bistratified ganglion cells. The RSA dendritic stem branches at the Pl level of the inner plexiform layer. The RSA cells may receive input from two bipolar cell types, one of which is a pure cone type. The large bist~tifi~ amacrine and ganglion cells that have a dendritic arborization at the PS level and that may be synaptically linked to the DSA cells have a second arborization at the PI level. They may also be connected to the RSA cells. DISCUSSION Cajal(1933) suggested that there is a distribution of retinal displaced amacrine cells throughout all vertebrate retinas. Recently, they have been described in teleost amphibian, avian and mammalian retinas. The present study confirms the occurrence of displaced amacrine cells in the fish retina. From the results of our morphological analyses it appears that in the dragonet retina, the small unistratified cell type with the cell body located in the outermost zone of the ganglion cell layer and the dendritic arborization at the P5 level in the inner plexiform layer is an amacrine cell. The lack of a clear axon and the synaptic relations made with other neurons in the inner plexiform layer indicate an amacrine cell function for this cell type. The dendritic branches receive input from bipolar cells and from neighbouring DSA cells as well as other amacrine cells. Their output occurs through conventional synapses to neighbouring DSA cells and possibly to ganglion cells. The small displaced amacrine cells represent a dense population of neurons with a 5pm square pattern arrangement. Their number thus corresponds to the number of double cone pairs and to the number of those bipolar cells that are arranged according to the 5 pm square pattern (Van Haesendonck and Missotten, 1983a). The DSA cells constitute about 20-25% of the total number of cells in the ganglion cell layer. This percentage agrees with estimates made for dis-
E. VAN HAESENDONCK and L. MISWITEN
1440
INL
Pi p2 p3
IPL
p, Ps
blstratifled GC
DSA
a
BPs.ts
b Fig. 10. Summary diagram of results. (a) Schema of bipolar amacrine and ganglion cells that may be synaptically linked to DSA and RSA cells. (b) Schema of synaptic connections of a DSA cell. Synapses onto ganglion cells are not included because they have not yet been unequivocally identified. Abbreviations: AC amacrine cell; CB cone bipolar cell; GC ganglion cell; GCL ganglion cell layer; INL inner nuclear layer; IPL inner plexiform layer; MB mixed bipolar cell; P pattern layer in IPL.
placed amacrine cell types in other vertebrate retinas: pigeon displaced amacrine cells 23% (Hayes, 1984); rabbit coronate cells 32%, increasing with excentricity from the visual streak (Vaney et al., 1981); rabbit starburst cells, 20-30% (Famiglietti, 1983a).
In the rat retina several types of displaced amacrine cells have been described (Perry and Waler, 1980). In the dragonet retina there is at present no evidence for more than one type of displaced amacrine cells, but this possibility should not be excluded. For instance, one may
Displaced small amacrinc cells in the Cullionynw retina
expect to find a large field displaced amacrine. Except in the pigeon retina where the majority of the displaced amacrine cells have dendritic field diameters measuring 50-80 pm, all types described in other species have large dendritic fields with diameters over 130 pm (Perry and Walker, 1980; Vaney et al., 1981; Famiglietti, 1983a; Hayes, 1984). In comparison, the now analysed DSA cells have remarkably narrow dendritic fields, with major axes no longer than 30pm. The size of the DSA dendritic fields varies. The dimensions of the minor as well as of the major axes range over 20 pm. An explicit correlation between the dendritic field size and the retinal location could not be established. In the retinas of the pigeon and the rabbit, on the other hand, the dendritic field size of the displaced amacrine cells is related to the eccentricity from the visual streak (Hayes, 1984; Famiglietti, 1985). In these species the retinal organisation is gradually changing from the central fovea1 part to the periphery. In contrast, the dragonet retina shows a uniform spacing of cones throughout the entire dorsal part. This homogenous distribution of the visual cells could explain the lack of a distinct correlation between size and excentricity in DSA cells. However, these cells show varying dimensions of their dendritic fields in any area of the retina. Some types of ganglion cells described in the retinas of the rat and the cat have also been found to lack any correlation between the dendritic field size and the retinal location. (Perry, 1979; Boycott and Wiissle, 1974). The unistratified dendritic arborization of the DSA cell is located at 70% of the depth of the inner plexiform layer, this is at the most proximal geometrically organized P5 level (Van Haesendonck and Missotten, 1983a). This level corresponds to Cajal’s sublamina 4. In the carp retina, Cajal(1893) reported a small type of cells with pyriform cell body in the ganglion cell layer and a fine, compact and unistratified dendritic arborization at the fourth level. He suggested the possibility that they are a displaced amacrine cell type. But Cajal’s limited description does not allow an adequate correlation between these cells and the small displaced amacrine cells in the dragonet retina. In the goldfish retina, a population of displaced amacrine cells that show choline acetyltransferase-like immunoreactivity is found to have their dendritic arborization in or near stratum 4 (Tumosa et al., 1984; Tumosa and Stell, 1986). A good correspondence, however, also exist with the den-
1441
dritic arborization level of the unistratified displaced amacrine cells in the pigeon retina and the unistratified starburst amacrine cells in the rabbit retina. In the pigeon retina, the dendritic plexus is situated at 64% proximal to the inner nuclear layer (Hayes, 1984). In the rabbit retina, the type b starburst amacrine cells ramify at 68% (Famiglietti, 1983a). Our observations indicate that DSA cells are mutually linked. In addition, at least three bipolar and possibly two other amacrine and one ganglion cell type make synapses with the DSA cells. The bipolar input is clearly segregated from the amacrine input and output. The ribbon synapses in bipolar cells occur in two distinct, geometrically arranged layers and are predominantly made by the short, central dendrites of the DSA arborization. The conventional synapses from and onto other amacrine cells or ganglion cells occur in the thin fibrous layer separating the two pattern layers. These synapses are made by the longer dendrites of the arborization. The segregation of synaptic functions in narrow, consecutive layers results in the brush-like appearance of the dendritic tree. The DSA cells are synaptically linked to mixed as well as pure cone bipolar cells. These bipolar cells contact photoreceptor cells with either wide cleft or narrow cleft superficial junctions (Van Haesendonck and Missotten, 1984). Their axons all end at the P5 level (sublamina b) in the inner plexiform layer. Thus, our results do not allow to make any correlation between the type of superficial junction at the photoreceptor-bipolar interconnection and the level in the inner plexiform layer at which the bipolar axon terminal is located. In the dorsal pure cone part of the dragonet retina these bipolars obviously receive only cone input, but in the ventral mixed part they also get rod input. In this respect the displaced small amacrine cells differ from the starburst displaced amacrine cells in the rabbit retina. Famiglietti (1983b) observed interconnections with cone bipolar cells alone. The number of synapsing bipolar cells, ranging from five to eleven, also reflects the varying dimensions of the displaced amacrine cell dendritic fields. Remarkably, seven analysed cells contact either BP5,ld buttons or BP5,2s buttons. Both types of synaptic buttons have a 7pm square pattern arrangement (Van Haesendonck and Missotten, 1983a). Maybe the BP5,2s buttons correspond to a fourth type of
1442
E. VAN
HABENDONCK
bipolar axon ending at this level. However, light microscopy of Golgi impregnated bipolar cells suggest that they may be appendages of the BPS, Id buttons. The mean dendritic overlap factor, calculated to be 7, indicates that at a single point in the dorsal retina, the dendritic fields of seven neighbouring DSA cells interfere. In the dense network of the fibrous layer their processes make numerous conventional synapses upon each other. Centrally in the dendritic field where the branching DSA processes are thick, they give synaptic output to the fine peripheral dendritic branches of neighbouring DSA cells. Thus, the DSA output and the bipolar input occur in the center of the dendritic field. They are surrounded by an annular zone of input from DSA cells and other amacrine cells. Light microscopical investigations revealed two additional types of amacrine cells that have a dendritic plexus at the level of the DSA arborization. Both the unistratified and the bistratifi~ type are potential candidates to synapse on the DSA cells. The large processes that are presynaptic to the DSA cells in the distal part of the fibrous layer all had the same electron microscopical appearance. We have not yet observed any ultrastructural details that allow us to distinguish different types. Fine terminal dendrites of a bistratified ganglion cell type which is the mirror image of the large bistratified amacrine cell may occupy some of the postsynaptic positions in the DSA conventional synapses. Finally, it is also possible that branches of diffuse amacrine and ganglion cells that pass the PS level are synaptically linked to the DSA cells. Further detailed analysis of Golgi-impregnated or immunocytochemically stained examples of these stratified and diffuse amacrine and ganglion cells has to reveal whether they are really connected to the DSA cells. Matching regular and displaced amacrine cells, have been reported in the retinas of several species. In the carp retina, Cajal (1893) described small regular amacrine cells which are very similar to the dragonet RSA cells. They have a thin descendant process, that gives rise to a small compact arborization with varicose branches at the second level. In the goldfish retina, Tumosa et al. (1984) observed choline acetyltransferase-like immunoreactive regular amacrine cells that have their dendritic arborization in or near stratum 2. The second stratum corresponds with the PI level in the dragonet
and L. MISSOTIXN
retina. The level of the RSA dendritic arborization also agrees well with the branching levels observed for the comparable amacrine cell populations in the pigeon retina: 24% (Hayes, 1984) and the rabbit retina: 22% (Famiglietti, 1983a). The existence of two matching amacrine cell populations is generally related to the On-Off lamination of the inner plexiform layer (Perry, 1980; Perry and Walker, 1980; Vaney et al., 1981; Famiglietti 1983a, 1983b; Hayes 1984). The displaced amacrine cells are the type b cells that are involved in the transmission of information through the On pathway, while the matching regular amacrines represent the type a cells that participate in a similar Off pathway. The resembling inte~onn~tions observed for type a and type b starburst amacrine cells in the rabbit retina seem to support this hypothesis (Famiglietti, 1983b). Our results indicate that RSA and DSA cells receive direct input from different bipolar cells. The DSA cells seem to collect a wider range of bipolar information than the RSA cells. The DSA and the RSA cells, however, may have comparable functions. The large bistratified amacrine and ganglion cells that co-stratify with the DSA cells also co-stratify to the RSA cells. This implies that they may be connected to the RSA cells at the Pl level and to the DSA cells at the P5 level. Thus, it is possible that one and the same large amacrine cell type would modify the signals passing through RSA and DSA cells. In addition, these signals perhaps also converge into the same large ganglion cell type. The question remains to what extent the RSA and the DSA cells are truly functionally matching cell types. Further detailed electron microscopical inv~tigation of the RSA cells and the large bistratified amacrine and ganglion cells has to reveal their interconnections in the inner plexiform layer and allow us to answer this question,
REFERENCES Ball A. K. and Dickson D. M. (1983) Displaced amacrine and ganglion cells in the newt retina. Expl Eye Res. 36, 199-213. Bingelli R. L. and Pauie W. J. (1969) The p$eon retina: quantitative aspects of the optic nerve and ganglion cell layer J. camp. Neurol. 137, i-18. Boycott B. B. and Wassle H. (1974) The morphological types of ganglion cells of the domestic cat’s retina. 1. Physiol., Lond. 240, 397-419.
Disphtced small amacrine cells in the C~liony~ Cajal S. R. (1893) La r&tine des vertcbrcs. Lo Celhle 9, 17-257. Translation as appendix in Rodrieck R. W. (1973) The Vertebrate Retina, pp. 775-904. W. H. Freeman, San Francisco, Calif. Famiglietti E. V. Jr (1983a) Starburst amacrinc cells and cholinergic neurons: mirror-symmetric ON and OFF amacrine cells of rabbit retina. Brain Res. 261, 138-144. Famiglietti E. V. Jr (1983b) On and off pathways through amacrine cells in mammalian retina: the synaptic connections of starburst amacrine cells. Vision Res. 23, 1265-1279. Famiglietti E. V. Jr (1985) Starburst amacrine cells: rno~holo~~l constancy and systematic variation in the anisotropic field of rabbit retinal neurons. J. Netrrosci. 5, 562-577. Hayes B. P. (1984) Cell populations of the ganglion cell layer: displaced amacrine and matching amacrine cells in the pigeon retina. Expl Brain Res. S6, 565-573. Hughes A. and Vaney D. I. (1980) Coronate cells: displaced amacrines of the rabbit retina? 3. camp. Neural. 11)9, 169-189. Osborne N. N. (1984) Substance P in the bovine retina: localization, identification, release, uptake and receptor analysis. J. Physiol., Lond. 349, 83-93. Perry V. H. (1979) The ganglion cell layer of the retina of the rat: a Golgi study. Proc. R. Eoc. B 204, 363-375. Perry V. H. (1981) Evidence for an amacrine cell system in the ganglion cell layer of the rat retina. Neuroscience a, 931-944. Perry V. H. and Walker M. (1980) Amacrine cells, displaced amacrine cells and interplexiform cells in the retina of the rat. Proc. R. Sot. B. 208, 415-431. Stell W. K. (1965) Correlation of retinal cytoarchitecture and uIt~st~cture in Golgi p~arations. Amt. Rec. 153, 389-398. Tumosa N., Eckmstein F. and Stell W. K. (1984) Immunocytochemical localization of putative cholinergic neurons in the goldfish retina. Neurosci. Lat. 48, 255-275. Tumosa N. and Stell W. K. (1986) Choline Acetyltransferase immuno~acti~ty suggests that ganglion cells
retina
1443
in the goldfish retina are not cholinergic. J. camp. Neurol. 244, 267-275. Vaney D. I. (1984) ‘Coronate’ amacrine cells in the rabbit retina have the starburst dendritic morphology. Proc. R. Sot. B 228, 501-508. Vaney D. I., Peichl L, and Boycott 8. B. (1981) Matching ~pulations of amacrine cells in the inner nuclear and ganglion cell layers of the rabbit retina. J. camp. Neurol. 199, 373-391. Van Haesendonck E. and Missotten L. (1979) Synaptic contacts of the horizontal cells in the retina of the marine teleost, &aliionymuslyra L. J. romp. Neurol. l&I, 167-191. Van Haesendonck E. and Missotten L. (1983a) Stratification and square pattern arrangements in the dorsal inner plexiform layer in the retina of CuBionymus lyra L. J. Ultrastruct. Res. 83, 296302. Van Haesendonck E. and Missotten L. (1983b) Interbipolar contacts in the dorsal inner plexiform layer in the retina of CaIiionym~ lyra L. .I. Ultr~t~ct. Res. 83, 303-311. Van Haesendonck E. and Missotten L. (1984) Synaptic contacts between bipolar and photoreceptor cells in the retina of Calionymus lyra L. J. camp. Neurol. 223, 387-399. Venable J. H. and Coggeshall R. (1965) A simplified lead citrate stain for use in electron microscopy. J. cetf Biof. 25,407-408. Vilter V. (1947) Dissociation spatiale des champs photosensoriels a cBnes et a batonnets chez un poisson marin, le Callionymuslyra. C.r. Sot. Biol., Paris 141, 344-346. Vrabec F. (1966) A new finding in the retina of a marine teleost, Callionymus lyra f... Foha Morphoi. XW, 143-147. Vrabec F. (1974) Inverted growth of dendrites of the retinal ganglion cells in Caflionymus MyraL (dragonet). Acta anat. 87, 93-96. West R. W. (1976) Light and electron microscopy of the ground squirrel retina: functional considerations. J. camp. Neural. laS, 355-378.