The ramification pattern of amacrine cells within the inner plexiform ...

Cell and Tissue Research

Cell Tissue Res (1981) 220:699-723

9 Springer-Verlag 1981

The ramification pattern of amacrine cells within the inner plexiform layer of the carp retina* Josef Ammermiiller and Reto Weiler Zoological Institute, University of Munich, Munich, Federal Republic of Germany

Summary. The morphology of amacrine cells in the retina of the carp is described using the Golgi technique. The ramification pattern of these cells was analyzed in flat-mounts of retinas. Based on these observations classification into five groups was made. Cells possessing one principal process leaving the soma were subdivided into starburst A-neurons and radiate neurons. Cells having two or more principal processes were subdivided into starburst Bneurons and spindle-shaped soma neurons. Small, diffuse amacrine cells form the fifth group. With respect to the shape of the field of arborization, the following cell types could be distinguished: (i) uniform cells, (ii) cells with a preferential direction, and (iii) cells with a marked edge, i.e., cells that lack processes in one direction. The latter form rarely occurs a m o n g starburst neurons; most of the spindle-shaped soma ceils possess processes with a preferred direction, and cells with a marked edge are mainly found a m o n g the radiate neurons. All five cell types are found throughout the retina. The size of the cells varies within each group, and there is no correlation between size and distance from the optic nerve. The radial arborization pattern of each cell was examined in serial transverse sections. Starburst A-neurons ramify in the middle of the inner plexiform layer (IPL), radiate neurons in the inner half, and spindle-shaped soma neurons without overlapping processes (type B) as well as starburst B-neurons in the outer half. The ramification can be monostratified (narrow or broad), bistratified or multistratified. Small, diffuse amacrine cells and spindle-shaped soma neurons with overlapping processes (type A) ramify throughout the entire IPL.

Key words: Retina (carp) - A m a c r i n e cells - Golgi impregnation - R a m i f i c a t i o n pattern Send offprint requests to: Dr. Reto Weiler, ZoologischesInstitut der Universit/itMiinehen, Luisenstr. 14,

D-8000 M/inchen, 2, Federal Republic of Germany * This work was supported by the Deutsche Forschungsgemeinschaft

0302-766X/81/0220/0699/$0.500

700

J. Ammermiiller and R. Weiler

In the vertebrate retina amacrine cells are neurons lacking an axon, the processes o f which arborize in the inner plexiforrn layer (IPL). Their somata are mostly located in the inner nuclear layer (INL), with some displaced cell bodies in the ganglion cell layer ( G C L ) or the IPL. Electron microscopic studies have revealed that these neurons make synaptic contacts with bipolar cells, ganglion cells and other amacrine cells (Dowling 1970; Dowling and Boycott 1965, 1966; Dubin 1970; Witkovsky and Dowling 1969; Witkovsky and Stell 1973). These results and physiological studies indicate an important role o f amacrine cells in the generation o f receptive fields, especially when these fields have a complex organization (Chan and N a k a 1976; Werblin 1972, 1977; Marchiafava and Torre 1978; Marchiafava 1979; Marchiafava and Weiler 1980; N a k a 1980). Cajal (1893) classified amacrine cells o f all vertebrate classes morphologically using radial sections o f the retina. The same was performed by Boycott and Dowling (1969) for primates and by W a g n e r (1973) for fish. There is evidence from recent data that the level o f stratification within the IPL is correlated to the " o n - o f f " characteristic o f carp amacrine cell photoresponses (Famiglietti et al. 1977). Furthermore, neuroactive peptides in the carp retina (Glickman et al. 1980; Karten and Brecha 1980), which are apparently restricted to amacrine cells, as well as the distribution o f glycinergic amacrine cells ( M a r c et al. 1979), are found to be level-specific. One characteristic morphological feature o f amacrine cells is their e n o r m o u s lateral extension and the great variety and complexity o f their horizontal branching pattern. This structural specificity is also reflected in their photoresponses, as shown by N a k a and Ohtsuka (1975) and M u r a k a m i and Shimoda (1977) for the fish retina and by Marchiafava and Weiler (1980) for the turtle retina using intracellular staining methods. N a k a and C a r r a w a y (1975) initially attempted a description o f the m o r p h o l o g y o f amacrine cells using the Golgi technique in fiat-mounts o f retinas from the catfish. Perry and Walker (1980) applied the same techniques in the rat retina. A l t h o u g h the carp is a favorite object in retinal research, the existing knowledge concerning the m o r p h o l o g y o f amacrine cells is still mainly based on Cajal's early w o r k o f 1893. The aim o f the present study is to provide a classification o f Golgistained amacrine cells in the carp retina based on their horizontal ramification pattern and to correlate this with the stratification level within the IPL. Materials and methods The retinas of adult carp (Cyprinus carpio) with a body weight of about 1000g were used for all studies. Preparation, Golgi-staining and flat-mount (FM)-embedding was carried out as described by Weiler (1978). Of the 82 retinas used, 52 were stained sufficiently. Only those cells that appeared to be completely stained were drawn and analyzed. As it is often very difficult to decide in Golgi-preparations whether a process represents an axon, only those cells where the somata were located in the INL and the arborizations of which extended into the IPL were classified as amacrine cells. In this way the possibility of confusing amacrine cells with displaced ganglion cells was very small, since Tachibana (1978) has shown that the proportion of displaced ganglion cells is only 1/4000 in the carp retina. Of the 120 cells that were drawn in FM-view, 75 cells were selected for radial sectioning. These cells were photographed, removed from the FM-preparation and re-embedded in Epon. Before sectioning, their position was ascertained, and the location of the somata was marked on the Epon block, which was then cut in 30-gin serial sections. By the marking of the soma, 61 cells were relocated, and the course of their processes in the IPL drawn as far as possible. If sectioning was not exactly perpendicular to the IPL, distortions were compensated for by focusing on the upper and lower part of the section and taking the

The ramification pattern of carp amacrine cells

701

resulting displacement into consideration when drawing. The microscope used was a Zeiss Ultraphot III with a camera lucida device. The FM-drawn ceils as well as the radially drawn ceils were ana/yzed quantitatively. A transparent foil with concentric circles of 33.33 ~tmseparation on retinal level was centered on the somata of the FMdrawn amacrine ceils. The circles were subdivided into 12 sectors, each of which measured 30 degrees. Three parameters were measured: (1) The number of branching-points between two successivecircles (givingthe number of branchingpoints in relation to the distance). (2) The number of intersections of the processes with every circle (givingthe number of processes in relation to the distance). (3) The number of these intersections within each sector of 30 degrees (giving the number of processes in relation to the direction). The pattern of ramification from the radially sectioned cells was analyzed by measuring the level and the vertical extension in serial sections from a cell. These values are given as percentage of the IPL, with 0 % being the outer border (between IPL and INL) and 100 % the inner border (between IPL and GCB).

Results L Classification based on fiat-mounts By examining flat m o u n t preparations one m a y get the initial impression that amacrine cells exist in a confusing multiplicity o f form. However, u p o n closer examination several gross types can be recognized, which can be distinguished with certainty when a sufficiently large sample o f cells is available. The amacrine cells were first classified subjectively on the basis o f their shape in flat-mounts o f retinas. Thereafter the cells were analyzed as described under Methods, and the resulting numerical data for each subjectively defined class compared. This c o m p a r i s o n supported the classification o f gross types and was an additional aid when d o u b t s arose concerning the categorization o f some cells. Initially, rough subdivision o f all cells leads to three distinct, large groups. The main criterion for the first two groups o f cells is the n u m b e r o f principal processes: the first group possesses only one, the second two or more principal processes. In the third g r o u p it is often difficult to determine the n u m b e r o f principal processes in flat-mount preparations. This group, however, c a n n o t be confused with the other two classes, due to the extension o f the cell and the pattern o f arborization. The criterion o f the n u m b e r o f principal processes was not only taken for its ease in recognition but also for its importance for the distribution o f the cell processes within the IPL. Cells having one principal process invading the I P L usually ramify at a m o r e inner level o f the IPL than cells having more principal processes.

1. Cells with one principal process. Fig. 1 shows camera lucida-drawings o f six examples o f cells from flat-mounts o f retinas with a single principal process. D u e to the low magnification, this process c a n n o t be seen in cells I / l c and III/6(16); however, it is clearly recognizable at higher magnification. All cells send a single, rather thick process vitread into the IPL, where it gives off secondary processes that extend horizontally. Based on the shape o f arborization, this large group o f cells was subdivided into two further classes. The first class is called starburst A-neurons (Fig. 1 a). These cells, with very long, fine branches that extend straight for long distances, have also been described by N a k a and C a r r a w a y (1975) in the catfish retina. They introduced

702

J. Ammerm/iller and R, Weiler

'r/321

/

/ ]/16(17)

i

I]I/6(16)

II/11(1

1/20

P b

Fig. I a, b. Camera lucida drawings of six examples of amacrine cells with one principal process, showing variability in size, shape and branching pattern, a Starburst A-cells; b radiate cells. Scale: 100 gm

--L_

CELL re/SO6) ,o'J

20

20

IL\ t~

10-

9 soma

~soo

N 30"

d [.m]

o~

~o ~

16o*

r

~o ~

36o ~

N

CELL 1/20

30-

20-

20

-7 TO-

]

F~

~___J

9

c.1 F - 7

soma

__FI

d [~m]

500

00

,~o o

2Vo ~

CELL 11/110/.,) 30

20

20

I0-

10-

--J

\

L ::

soma

360*

_5-

&l

N 30

9

~O*

;l 500

d [.m]

0~

,o o

,~o ~

270"

Fig. 2. Quantitative analysis of three cells shown in Fig. 1. The diagrams on the left include the data of the number of processes (solidline) and the n u m b e r of branching-points (dashedline) plotted against the distance from the soma for a cell. The data were obtained by centering a transparent foil with concentric circles (331/3 w n distance) onto the soma. On the right hand side the n u m b e r o fintersections of processes with all circles within each sector o f 30 ~ is plotted against the sector

704

J. Ammermiillerand R. Weiler

the term "starburst neurons" but included in this group another type of cell resembling radiate neurons and starburst B-neurons of the present study. Based on the following data, we suggest that these cells be classified separately from the starburst A-neurons. Cajal (1893) described neurons similar to the starburst A-type as stellate cells in perch and lizards, as did de Castro (1966) in the chick retina. Fig. 1 b shows the second subclass of amacrine cells with one principal process, called "radiate neurons". Ramon-Moliner (1962) named neurons of the CNS with a similar appearance "radiate". The processes are also long but slightly less straight and are thicker than those of starburst neurons. Fig. 2 shows the analysis of three cells from Fig. 1. The diagrams on the left include the data on the number of processes (solid line) and the number of branching-points (dashed line) plotted against the distance from the soma. Cell III/6(16) reaches its maximal number of processes very close to the soma, this value being high. In contrast, cells II/11(14) and 1/20 have their peaks, which are lower, further from the soma. Correlated to this, all branching-points of the first cell are also very near to the soma, while in the latter cases branching-points occur distant from the soma. The pattern ofarborization in the horizontal plane is shown in the diagrams on the right hand side of Fig. 2. Here, the total number of intersections of the cell with those arcs lying in each 30 degree sector was plotted against the sector. The center of these arcs was again determined by the soma (see Methods). The resulting diagram is a quantitative description of the distribution of cell processes within an assumed circular field with the soma at the center, the diameter being given by the length of the longest process. This quantitative analysis of the projection of the cell processes in the horizontal plane is called the pattern of the anatomical receptive field (RF-pattern). It includes information about the number of cell processes in relation to the direction from the cell body. To describe the RFpattern three subclasses were introduced. A "uniform" pattern means that none of the 30~ lacks cell processes and that the number of processes within each sector is almost the same. An RF-pattern called "preferential direction" implies that also in all sectors cell processes are found, but that their number in one or more sectors far exceeds the values of the other sectors. Finally, the term "marked-edge" characterizes a RF-pattern where two or more sectors lack cell processes. Cell III/6(16) has a rather uniform distribution o f processes, as all sectors contain processes. The density of the processes is slightly higher around 270 ~ Cell 1/20, on the other hand, does not show a uniform distribution. There is a clear concentration of processes between 120 ~ and 270 ~ however, again all sectors contain processes. Thus, this cell has a preferential direction with respect to the density of its processes. In contrast, the cell II/11(14) not only has a preferential direction, but the RFpattern forms a "marked-edge" since the sectors from 0 ~ to 150 ~ lack processes. The description "uniform", "preferential direction" and "marked-edge" of RFpattern of a cell is later used in Table 1, where the pattern of all cells is summarized. The data for all cells with one principal process are summarized in Fig. 3. The striped columns represent the starburst A-cells, and the stippled columns the radiate type. Fig. 3a is a histogram, in which the number of cells is plotted against the distance of the last branching-point from the soma. Fig. 3b is the same for the distance of the maximal number of processes from the soma, and Fig. 3c shows the distribution of the maximal number of processes.


10-

Starburst A cells

/

{3

"6 n

705

Radiatecells

5-

E Z

I

I

I

I

I

~um

5OO soma

ul

Distance from some

10-

"S JD

5-

E

i

I

t

I

I

I

500

some

1

~m

Distance from soma

uo

t0-

o

x~ E

5-

Z

c

I

I

I

10

20

30

Number of processes

Fig. 3a-e. Histograms of(a) the distance of the last branching-point from the soma, (b) the distance of the maximal number of processes from the soma, and (e) the maximal number of processes of all amacrine cells with one principal process. Stippled columns represent the data of radiate neurons and striped columns those of starburst A-neurons These figures s h o w t h a t the subjectively classified ceils are i n d e e d q u a n t i t a t i v e l y different. All s t a r b u r s t A - n e u r o n s (striped c o l u m n s in Fig. 3) have their b r a n c h i n g p o i n t s very close to the s o m a (Fig. 3 a), r e a c h i n g the m a x i m a l n u m b e r o f processes very close to the s o m a (Fig. 3b), this m a x i m a l n u m b e r being very high (Fig. 3c). In c o n t r a s t to this, r a d i a t e a m a c r i n e cells usually possess less processes (Fig. 3 c),

706

J. Ammermfillerand R. Weiler

with the maximal number of processes distant from the soma (Fig. 3b), and terminal branching-points occurring farther from the soma (Fig. 3a). Because of the numerous intermediate types, the values for starburst- and radiate-cells overlap. However, it can be seen that these cells are always intermediate only for one numerical parameter, and are clearly defined by the other two. For example, cell I(20) has maximally 19processes, and thus for this parameter is an intermediate cell. Because this m a x i m u m represents processes distant from the soma (230 lam), which is also true for the last branching point (430 ~tm), this cell is quantitatively clearly defined as a radiate type. 2. Cells with two or more principal processes. Fig. 4 shows five camera lucida

drawings of amacrine cells with more than one principal process. The cells shown in Fig. 4a, I I I / l l and 1/12(2), resemble very closely those cells in Fig. 1 that were classified as starburst A-neurons. Thus, we call these cells starburst B-neurons, which also have very straight, thin processes. In contrast to starburst A-neurons, however, more than four principal processes usually emerge from the soma, thus giving the soma an irregular shape. Cells such as those shown in the Fig. 4b all have a more or less spindle-shaped soma. Typical examples are cells IV/B and 1/28 in Fig. 4b, while the soma of cell 1/12(1) is slightly distorted. Due to the shape of the soma, we call these cells amacrine cells with spindle-shaped soma. They are similar to the cells called "spindle-type neurons" by N a k a and Carraway (1975) in the catfish retina. Their processes are thicker than corresponding elements of starburst cells and the course of the processes is more curved. One special feature of amacrine cells with spindleshaped somata is the fact that m a n y of the cells have overlapping processes, and therefore m a n y intersections of cell processes can be seen in the flat-mount view. This is not the case, to the same extent, in the other classes of cells mentioned above. Figure 5 presents the numerical analysis of two neurons shown in Fig. 4. Cell 1/12(2) displays m a n y processes near the soma (Fig. 5, left half, solid line), and all branching-points close to the soma (Fig. 5, left half, dashed line). The RF-pattern is uniform (right half of Fig. 5). Cell IV/8, in contrast, shows few processes, with branching-points far distant from the soma. The RF-pattern reveals a markededge, as there are no processes between 300 ~ and 360 ~. As the maximal number of processes of cell IV/8 is rather close to the soma, this neuron is classified intermediate for this parameter. The data for the starburst B-neuron 1/12(2) confirms the similarity with starburst A-neurons (Fig. 2, III/6(16)). The main difference is the number of principal processes leaving the soma and therefore a different distribution within the sublayers of the IPL as will be shown below. A compilation for all amacrine cells with more than one principal process is given in Fig. 6. Starburst B-neurons are indicated by striped columns, amacrine cells with spindle-shaped soma by stippled columns. For all three measured parameters there is a clear difference between the two classes. All branching-points of starburst Bcells (Fig. 6a), and also the maximal number of processes (Fig. 6b) are close to the soma. The maximal number of processes is very high (Fig. 6c). Amacrine cells with spindle-shaped somata have branching-points distant from the soma (Fig. 6a); the maximal number of processes is not as close to the soma as in

.

t i

/

1

i

•

i~l \ b 1/28

, ~

1/12(I) IK/8

Fig. 4a, b. Camera lucida drawings of amacrine ceils with more than one principal process, a Starburst B-neurons; b amacrine cells with spindle-shaped soma. Cells [/12(i) and 1/28 have overlapping processes. Scale: 100 pm

708

J. AmmermiJllerand R. Weiler

CELL 1/12(2)

N

.1

N

30-

2O-

ii i i

9 soma

N 30.

~)o

d

CELL ZI/8

~.]

o~

~o ~

1~o ~

2Vo ~

3

N

30

20

20,

10"

Lq 9 soma

5()0

d

~ml

O*

~o"

1;o ~

2~0 ~

360*

Fig. 5. Quantitative analysis of two cells, shown in Fig. 4. For further details, see legend Fig. 2 starburst B-neurons (Fig. 6b), and there are fewer processes than in starburst Bcells (Fig. 6c). Comparing the data of Fig. 6 with data of Fig. 3, it can be seen that not only the described similarity between starburst A- and B-neurons is evident, but also between radiate and spindle-shaped soma neurons. The main criterion for a separation is again the number of principal processes.

3. Small, diffuse amacrine cells. Small amacrine cells, which occur frequently, have either one or m a n y short principal processes leaving the soma. Their shape, however, is uniform, and they cannot be classified into the groups given above, since their pattern of ramification differs qualitatively and quantitatively (Fig. 7). They are rather small in size, compared with the other classes of amacrines, and show an extreme overlapping of processes in the flat-mounted retina, which is due to their diffuse distribution throughout the IPL. The course of the branches, which are very thin, is much more sinuous than that of radiate- or spindle-shaped cells.


709

10u

Starburst B ceils r~

~

5-

Spindle shaped soma cells

E Z

I

I

I

I

soma

I

I

IR

L

I

500 Distance from soma

I

~

pm

_~ 10~5 E

5-

z

t soma

b

i

50O

jJm

Distance from soma

10O r

"6 E

5-

Z

I

I

I

10 20 30 c Number of processes Fig. 6a-c. See legend Fig. 3. Stippled columns represent amacrme ceils with spindle-shaped soma, striped columns starburst-B neurons An analysis of cell V/1 shows m a n y branching-points near the s o m a (Fig. 8, dashed line), but fewer processes (Fig. 8, solid line). This results from the very short branches, which do not intersect the circles. The R F - p a t t e r n is fairly uniform, with few intersections per sector due to the small size o f the cell. Fig. 9 summarizes the results o f the cells examined. The m a x i m a l n u m b e r of processes lies between 10 and 20 (Fig. 9c), with this m a x i m u m very close to the s o m a (Fig. 9b). The terminal branching-points occur farther away from the s o m a (Fig. 9 a). The s o m a t a o f small

710

J. Ammermfiller and R. Weiler

v/~

v/3

Fig. 7. Camera lucida drawings of two small, diffuse amacrine cells. Scale: 100 ~tm N

30.

CELL

30"

~./1

20

rn i , i i ] i

20'

10" !

1084 i

9 soma

5~o

] cl [ u m ]

0~

~o o

1~o ~

~yo ~

360 ~

Fig. 8. Quantitative analysis of cell V/I, shown in Fig. 7. See legend Fig. 2 diffuse cells are usually smaller (8-12 lam) than those o f the other amacrine cells, the size o f the somata varying between 10 and 40 ~tm.

4. The RF-pattern of cells of the different classes. N o t only the n u m b e r o f processes and branching-points in relation to the distance was analyzed, but also the R F pattern as described above. Examples can be seen in the right half o f Figs. 2, 5 and 8. Based on the quantitative analysis three kinds o f RF-patterns o f a cell can be distinguished, which have been introduced as uniform, with preferential direction


711

10U

~

"5 5E Z

f

{

t

Smatt diffuse cel{s

I

1

1

500

soma

~m

Distance from soma

cn

10-

U

..a

E Z

t

I

I

I

500

somQ

pm

D i s t a n c e f r o m soma

10 O

"5 E

:3 Z

I

I

I

10

20

30

Number of processes

Fig. 9a--c. Summary of all small, diffuse amacrine cells. The histograms show the distance of the last branching-points from the soma (a), the distance of the maximal number of processes from the soma (b), and the maximal number of processes (c)

and with a marked-edge. This distinction was supported by consulting the camera iucida drawings o f the cells. Table 1 gives the classification of all cells with respect to their RF-pattern. The cells are grouped into the five main classes described above. Cells with a marked-edge are usually radiate amacrine cells or cells with a

712

J. Ammermtiller and R. Weiler

Table 1. Distribution of the three kinds (uniform, preferential direction, marked-edge) of RF-pattern among the amacrine cell classes. The values for each class are given in percentage of the total number of cells investigated from this class. The total number is given in the bottom row. RF-pattern

Type: Starburst A

Radiate

Spindleshaped

Starburst B

Small diffuse

Uniform

52 %

26 %

20 %

50 %

43 %

Preferential direction

43 %

43 %

64 %

50 %

57

5%

31%

16 %

53

25

6

7

Marked-edge Number of cells

19

,-ff

9

.E2-

~

9

o o

.

"o

o

~oO -~-.

~ ~

o

,

~ o 19 o

o

o

~'~

9 o

9 i

o 9

o o

o

o o

I

s OEm., ] Fig. 10. Correlation between the field size, measured as the greatest extension of the cell, and the distance from the optic nerve. -~, Starburst A- and starburst B- cells, o radiate cells; 9 radiate cells with a marked edge; 9 amacrine cells with spindle-shaped soma

spindle-shaped soma, while starburst A, starburst B and small, diffuse amacrine cells tend to be more uniform in the radiation of their processes.

5. The size and distribution of the different cell types. It is of interest whether the existence of a certain cell type is limited to a particular region within the retina, and whether a correlation exists between the size of the cell and its location. When the retina remained intact during preparation, it was possible to identify the position of single cells with respect to the optic nerve, which is the only point in carp retina that can be used for orientation. Carps do not have a fovea centralis, but an "area" that is thought to be the site of best visual resolution. This "area" is situated near the site where the optic nerve leaves the retina. In Fig. 10 the largest extension of the cells, called field size, is plotted against its distance from the optic nerve. The symbols represent the different types ofamacrine cells. Radiate cells and cells with spindle-shaped somata are found at all distances from the optic nerve, whereas starburst A- and B-neurons were only found up to 4.3 m m distant from the optic nerve. Unfortunately, no well-stained small, diffuse


~J O_

713

50-

100

starburst A cells

radiate cetLs

A 0

III ,00 /l

9 9

starbur t

spindle shaped

B ceLlsl

celts type A

soma

type B

~al(

Fuse cells

B Fig. 11A, B. Distribution of the cellprocesses of different amacrine cell types across the IPL 0 ~ refers to the scleral and 100 ~ to the vitreal border of the IPL. Each vertical line represents one cell and its extension marks the region within the IPL where processes of the cell were found. Lines with a symbol represent stratified cells. They are either mono- (one symbol)or bistratified (two symbols).Lines without symbol represent diffuse cells. The position of the symbol indicates the main level of stratification. The type of symbol includes information concerning the RF-patteru of the cell revealed in the flat-mount analysis (e uniform; [] preferential direction; /~ marked edge) a m a c r i n e cells were f o u n d in u n d a m a g e d retinas. A n o t h e r c o n c l u s i o n that can be d r a w n from Fig. 10 is that n o correlation exists between the size o f the cells a n d their position relative to the optic nerve. Thus, the great variety o f sizes in each class (starburst A - n e u r o n s : 300 ~ t m - 3,000 ~tm ~ ; radiate n e u r o n s : 400 ~ t m - 2,500 ~tm ~ ; starburst B - n e u r o n s : 500 lain - 1600 Ixm ~ ; spindle-shaped n e u r o n s : 600 Ixm - 1,500 txna ~ ; small diffuse n e u r o n s : 100 ~tm - 500 p m ~ ) c a n n o t be correlated to location within the retina.

II. The radial distribution o f the arborizations across the I P L Sixty one cells, classified in F M , were s u b s e q u e n t l y sectioned radially. The processes o f almost all cells could be followed for 100 lxm to 500 I,tm d i s t a n t from the soma, some even further. Over this distance the p a t t e r n o f r a m i f i c a t i o n within the IPL was d e t e r m i n e d a n d the level of the stratification m e a s u r e d as

Fig. 12a, b. Photomontage of serial, radial sections o f a starburst A amacrine cell (cell I/1 c in Fig. 1 a). a Main part of the cell with the soma. h More distant course of the processes. Arrows at the border indicate the connection of the figures. Brackets mark the IPL. Scale: 50 ~tm

=..

Fig. 13a, b. Photomontage of a radiate amacrine cell (cell 1/20 in Fig. 1 b). See legend Fig. 12 ~h

O

~q

Fig. 15. Photomontage of a small, diffuse amacrine cell (cell V/3 in Fig. 7). Arrow points at the soma. Brackets mark IPL

Fig. 14a, b. Photomontage of an amacrine cell with spindle-shaped soma (cell IV/8 in Fig. 4b). See legend Fig. 12

Fig. 16a, b. Photomontage of an amacrine cell with spindle-shaped soma and overlapping processes (cell 1/28 in Fig. 4b). See legend Fig. 12

-a

---.I

o

o=

g~

g~

718

J. Ammermiillerand R. Weiler

described in Methods. We cannot exclude the possibility that terminal branches of a very large cell, which could not be fully reconstructed in the serial sections, may terminate in layers not covered by the branches within the examined region. However, we do not expect a fundamental difference in pattern of such terminals, since for every cell type it was possible to analyze cells of smaller size up to their finest terminals. In Fig. 11 the data of the ramification within the IPL of the investigated 61 cells are graphically presented. Each vertical line represents one cell and the extent of the line marks the region within IPL where processes of the cell were found in the serial transverse sections. Lines with a symbol represent stratified cells. They are either mono- (one symbol) or bistratified (two symbols). No symbol represents a diffuse cell. The position of the symbol indicates the level of IPL where the main part of the processes was found. The type of symbol includes information concerning the pattern of the anatomical receptive field (RF-pattern). Photomontages of the radial view of a few examples of cells are presented in Figs. 12-16. The starburst A-cell of Fig. 12 is the cell I/1 c of Fig. 1 a. In Fig. 13a part of the radiate amacrine cell 1/20 of Fig. 1 b is presented and in Fig. 14 the spindle-shaped soma cell IV/8 of Fig. 4b. The diffuse distribution across the IPL of the small, diffuse amacrine cell V/3 of Fig. 7 is seen in Fig. 15. Fig. 16 shows the ramification of the spindle-shaped soma cell 1/28 with coverlapping processes of Fig. 4b. The following conclusions can be drawn (see Fig. 11): a) Starburst A-cells arborize in the center of the IPL, the main part of the processes is always found around 50 ~o. Single processes may vary between 20 and 70 ~ of the IPL. b) Radiate amacrine cells have their processes near the ganglion cell bodies, in the inner half of the IPL, between 45 ~o and 100 ~ of the IPL. There is no relation between the RF-pattern and the position within the IPL. c) Cells with a spindle-shaped soma appear to constitute two forms. One group is diffuse throughout the IPL, the other arborizes in the sclerad part of the IPL, between 0 ~o and 30 ~. A re-examination of the cell drawings from the flat-mounted retina revealed that the diffuse cells are those showing an overlapping of processes in the fiat-mount view. Cells arborizing in the sclerad 30 ~o of the IPL do not show intersections of their processes in the flat-mount view. Based on this clear correlation between the fiat-mount view and the ramification pattern throughout the IPL, the spindle-shaped soma cells are subdivided into typeA (overlapping of processes ) and type B (without overlapping of processes). d) The sectioned starburst B-cells arborize in the outermost part of the IPL, at levels between 0 ~ and 40 ~ of the IPL. e) The small, diffuse cells send their processes throughout the complete IPL, between 0 ~o and 100 ~ . Some of the sectioned cells appear to be bistratified, with two main levels of arborization. However, since it is very difficult to decide between bi- or multistratified and diffuse cells in the carp retina, this was indicated but not used for further classification.


719

Table 2. Classificationof amacrine cells, including the data of the flat-mounted retina and of transverse section analysis Ramification pattern seen in the flatmounted retina

Amacrinecell type as described in Results

Main arborization level within IPL

Most frequent RF-pattern (see Table 1)

Starburst

Starburst A Starburst B

50 % 0-30 %

uniform and preferred direction

Radiate

Radiate Spindle-shaped soma ceils type B small, diffuse Spindle-shaped soma cells type A

50-100 %

marked-edge and preferred direction

Diffuse

0-30 0-100 % 0-100 %

uniform and preferred direction

Discussion In the present study amacrine cells of the carp retina were investigated by means of the Golgi-technique. In Golgi-preparations the problem of confusing amacrine cells with ganglion cells always exists, since the branching pattern of both types of cells m a y be similar and axons may often be unstained; therefore only perikarya located in the inner nuclear layer were considered. Since Tachibana (1978) has shown by use of the HRP-technique that only 1/4000 of the cells in the amacrine cell layer in the carp retina are displaced ganglion cells, and that there are no amacrine cells in the layer of the ganglion-cell perikarya, the probability of confusing ganglion cells with amacrine cells is very low. Thus, the error in this classification, produced by displaced ganglion cells, is very small. Five main classes of amacrine cells are introduced with respect to their appearance in flat-mounts of retinas. The numerical analysis of the ramification pattern shows that this distinction is based on quantitative data and not only on empirical observations. However, the arborization o f starburst A- and starburst B-cells displays a similar pattern; this holds true also for the pattern of radiate and spindle-shaped soma cells. These neurons differ only in having one or several principal processes. Since this arrangement influences the radial distribution but obviously not the horizontal pattern, one can simplify the classification. Hence, the ramification pattern observed in the flat-mounted retina can be described as "starburst" (TypeA + B), "radiate" (radiate and spindle-shaped soma type) and "diffuse" (small, diffuse cells). The analysis of the transverse sections, on the other hand, leads to the following classification: In the outer half of the IPL the processes of starburst B-cells and the spindleshaped soma cells of type B are found. The center of the IPL displays the processes of starburst A-cells, and the inner half o f the IPL contains the processes o f radiate neurons. Small, diffuse cells and spindle-shaped soma cells of t y p e A are scattered throughout the IPL.

720

J. Ammermfillerand R. Weiler

Table 2 is an attempt to summarize all these data. Included is the description of the most frequent pattern of the anatomical receptive field (RF-pattern) found in each group. The classical study of Cajal (1893) is still one of the most important morphological contributions to this field of neuroanatomy, especially with reference to the IPL of the carp retina. A comparison of the present results with his findings can only be carried out with the radially sectioned cells, since the classification of Cajal is based on such sections. The two main types of amacrine cells, diffuse and stratified as described by Cajal, were also found in the present study. Cajal's small, diffuse amacrine cells, also called pyriform cells, appear also small and diffuse in our preparations (see Fig. 15), whereas his large, diffuse amacrine cells, or ordinary multipolar cells, are the spindle-shaped amacrine cells of the present study, characterized by overlapping processes (see Fig. 16). Some of the cells of this type, however, also resemble Cajal's large two-layered amacrine cells. A further correlation between the results of Cajal and our flat-mount classification is more difficult. The most probable correlation appears to exist between Cajal's stratified amacrine ceils of the first sublayer and the starburst Bcells or the spindle-shaped soma cells free of overlapping processes, respectively (see Fig. 14). It is very probable that the different types of stratified cells of the fourth and fifth sublayer are all radiate cells of different sizes (see Fig. 13). Furthermore, the stratified cells of the second and third sublayer may be starburst A-cells (see Fig. 12). If we compare our main types of amacrine cells with the flat-mount classification of Naka and Carraway (1975) for the catfish retina, two main differences are evident. Firstly, the latter authors did not observe the small, diffuse amacrine cells, and secondly, our starburst A-cells, starburst B-cells and radiate cells appear to correspond only to a particular cell class of Naka and C a r r a w a y - the starburst neurons. The quantitative analysis (see Fig. 3) shows that these cells indeed form different classes in the carp retina. In the study of Murakami and Shimoda (1977) four Procion Yellow-stained cells are presented, which we can readily identify. They show two cells, which we would call radiate cells, branching in the middle and the inner half of the IPL, respectively, and two cells with spindleshaped soma and overlapping processes, which are diffuse. These two groups of cells differed physiologically in their photoresponses. The first two cells produced sustained photoresponses, the third and fourth cells were transient type amacrine cells. Since the input organization forming a receptive field of a cell may not only depend on its size, but also on the distribution of its processes around the soma, we have studied the anatomical pattern of the cellular receptive field. Table 1 shows that all patterns are present in each class of cells, but that the frequency differs for every class. It is very tempting to speculate that cells with asymmetrical shape may be involved in direction selectivity. Recently, Naka (1980) demonstrated that this is the case in catfish. The spindle-shaped amacrine cells belonging to his NB-amacrine cell class give rise to a different response pattern depending on the movement direction of a small bar across the visual field. Figs. 1 and 4 show that one cell type can exist in different size classes. The question arises whether there is a correlation between the size of a cell and its


721

location within the retina, as has been shown for ganglion cells by Boycott and W~issle (1974) in the cat and turtle retinas (Peterson and Ulinski 1979). This is not true for amacrine cells in the carp retina (see Fig. 10). Perry and Walker (1980) also did not find such a correlation in the rat retina. Thus, the size of a cell appears to be determined by other factors than its location within the retina. Moreover, Fig. 10 shows that the different classes of amacrine cells can be found throughout the retina. Therefore, in contrast to the radial distribution (Fig. 11), in the horizontal plane, there is no specific location of a particular cell type. The IPL of most vertebrate retinas shows a stratified pattern in transverse section. This pattern results from the endings of the entering neurons, which are often restricted to distinct levels within the IPL. Such a pattern is also visible in the teleost retina: Cajal (1893) described five levels of stratification. Unfortunately, these levels are not very distinct in Golgi-stained preparations of carp retinas. Therefore, we have not used this five-level scheme for the analysis of the radial distribution, but expressed the level of a stratification in percentage of the total IPL, starting at the outer border with 0 ~ . Recent results have demonstrated that the arborizations of bipolar, amacrine and ganglion cells giving either on- or offphotoresponses are found at different levels (Famiglietti et al. 1977). Cells producing an on-response arborize in the inner half, cells with an off-response in the outer half of the IPL. A similar distribution is described for the cat retina (Nelson et al. 1978) and for some neurons of the turtle retina (Marchiafava and Weiler 1980). Our analysis of the main ramification levels of the six classes of amacrine cells within the IPL leads to the distinction of three levels (Fig. 11). A scleral level covers approximately the outer 30 ~ of the IPL and contains the processes of starburst B and spindle-shaped soma amacrine cells free of overlapping processes. A narrow layer at the center contains the processes of the starburst A-neurons, while the remaining part of the IPL contains the radiate neurons. Small, diffuse cells and amacrine cells with spindle-shaped soma and overlapping processes ramify throughout the entire IPL. In these two groups it is very difficult to distinguish between diffuse cells and multistratified cells. Therefore, some neurons, which were called "diffuse", may be multistratified or bistratified. The difficulty results from the fact that the levels of ramification of many cells are very broad. Thus, the different levels of a multistratified cell probably overlap to some extent; however, the main levels form a distinct stratification scheme. This difficulty of classification is also true for monostratified cells, where some neurons show very narrow levels of arborization, while other cells may be called "broad monostratified cells" (see Fig. 11 .) In the turtle retina, Weiler and Marchiafava (1981), using intracellular staining with HRP, distinguished narrow monolayered from broad monolayered amacrine cells and also described bistratified and diffuse amacrine cells. Assuming that the on-off distinction holds for all cell types ending within the IPL in the carp retina, the radiate cells are on-type neurons, whereas starburst Bcells and spindle-shaped cells without overlapping processes are off-type neurons. Starburst A-cells as well as the diffuse cells may be candidates for on-off neurons. This correlation is confirmed for the four cells shown by Murakami and Shimoda (1977). Two diffuse cells, which correspond to spindle-shaped cells with overlapping processes, show on-off photoresponses. Their on-type cells are radiate

722

J. Ammermfiller and R. Weiler

cells a n d r a m i f y in the i n n e r h a l f o f the I P L . A s yet t h e r e is n o f u n c t i o n a l e x p l a n a t i o n for the d i f f e r e n t b r a n c h i n g p a t t e r n s (Figs. 1, 4, 7) o r f o r the d i f f e r e n t R F - p a t t e r n o f a cell ( T a b l e t ) . I t is o b v i o u s t h a t in t h e I P L t h e a m o u n t o f cells h a v i n g a s y m m e t r i c a l l y o r i e n t e d b r a n c h i n g p a t t e r n s far exceeds t h a t in the O P L , a n d t h u s makes the IPL capable of motion and pattern recognition.

Acknowledgements. The authors wish to thank HJ. Wagner, F. Zettler and R. Douglas for critical reading of the manuscript, I. Dragun and B. Scheyerer for providing some of their unpublished material and I. Menzel for her continuous help during the preparation of the manuscript.

References Boycott BB, Dowling J E (1969) Organisation of the primate retina: Light microscopy. Phil Trans R Soc (London) 255:109-I 84 Boycott BB, W/issle H (1974) The morphological types of ganglion cells of the domestic cat's retina. J Physiol 240:397-419 Cajal S, Ram6n Y (1893) La r6tine des Vert6br6s. Cellule 9:119-225 Chan RZ, Naka KI (1976) The amacrine cell. Vision Res 16:1119-1129 DoMing JE (1970) Organisation of vertebrate retinas. Invest Ophthal 9:655-680 Dowling JE, Boycott BB (1965) Neural connections of the retina: Fine structure of the inner plexiform layer. Cold Spring Harbor Syrup Quant BiN 30:393-402 DoMing JE, Boycott BB (1966) Organisation of the primate retina: electron microscopy. Proc R Soc London Ser B 166:80-111 Dubin MW (1970) The inner plexiform layer of the vertebrate retina. J Comp Neurol 140:479-506 Famiglietti EV Jr, Kaneko A, Tachibana M (1977) Neuronal architecture of ON and OFF pathways to ganglion cells in carp retina. Science 198:1267-1269 Glickman RD, Adolph AR, DoMing JE (1980) Does Substance P have a physiological role in the carp retina? Invest Ophtb Vis Sci (Suppl) 19:281 Karten HI, Brecha N (1980) Localisation of Substance P immunoreactivity in amacrine cells of the retina. Nature (Lond) 283:87-88 Marc RE, Lain DMK, Stell WK (1979) Glycinergic pathways in the goldfish retina. Invest Ophth Vis Sci (Suppl.) 18 Marchiafava PL (1979) The responses of retinal ganglion cells to stationary and moving visual stimuli. Vision Res 19:1203-1211 Marchiafava PL, Torre V (1978) The responses of amacrine cells to light and intracellularly applied currents. J Physiol 276:83-102 Marchiafava PL, Weiler R (I 980) Intracellular analysis and structural correlates of the organisation of inputs to ganglion cells in the retina of the turtle. Proc R Soc London Ser B 208;103-113 Murakami M, Shimoda Y (1977) Identification of amacrine and ganglion cells in the carp retina. J Physiol 264: 801-818 Naka KI (1980) A class of catfish amacrine cells responds preferentially to objects which move vertically. Vision Res 20:961-966 Naka KI, Carraway NRG (1975) Morphological and functional identification of catfish retinal neurons. I.) Classical morphology. J Neurophysiol 38:53-71 Naka KI, Ohtsuka T (1975) Morphological and functional identification of catfish retinal neurons. II.) Morphological identification. J Neurophysiol 38:72-91 Nelson R, Famiglietti EV Jr, Kolb H (1978) Intracellular staining reveals different levels of stratification for on- and off-center ganglion cells in cat retina. J Neurophysiol 41:472-483 Perry VH, Walker M (1980) Amacrine cells, displaced amacrine cells and interplexiform ceils in the retina of the rat. Proc R Soc London 208:415-431 Peterson EH, Ulinski PS (1979) Quantitative studies of retinal ganglion cells in a turtle, Pseudemys seripta elegans I.) Number and distribution of ganglion cell. J Comp Neurol 186:17-47 Ramon-Moliner E (1962) An attempt at classifying nerve cells on the basis of their dendritic patterns. J Comp Neurol 119:211-227 Tachibana M (1978) Displaced ganglion cells in carp retina revealed by the horseradish peroxidase technique. Neuroscience letters 9:153-157


723

Wagner HJ (1973) Die nerv6sen Netzhautelemente von Nannacara anornala (Cichlidae, Teleostei). I.) Darstellung durch Silberimprggnation. Z Zellforsch 137:63-86 Weiler R (1978) Horizontal cells of the carp retina: Golgi impregnation and Procion Yellow injection. Cell Tissue Res 195:515-526 Weiler R, Marchiafava PL (1981) Photoresponses and morphology of amacrine cells in the turtle retina. Invest Ophth Vis Sci (Suppl) 20:183 Werblin FS (1972) Lateral interactions at inner plexiform layer of vertebrate retina: antagonistic responses to change. Science 175:1008 Werblin FS (1977) Regenerative amacrine cell depolarisation and formation of on-off ganglion cell response. J Physiol 264:767-785 Witkovsky P, Dowling JE (1969) Synaptic relationship in the plexiform layers of the carp retina. Z Zellforsch 100: 60-82 Witkovsky P, Stell WK (1973) Retinal structure in the smooth dogfish Mustelus canis: Electron microscopy of serially sectioned bipolar cell synaptic terminals. J Comp Neur 150:147-168 Accepted June 19, 1981