Circuitry analysis of the outer plexiform layer in the rabbit retina

JOURNAL OF ULTRASTRUCTURE RESEARCH 62~ 54- 81

(1978)

Circuitry Analysis of the Outer Plexiform Layer in the Rabbit Retina I. Extension of Analysis of Circuits Interpreted to be Associated with Directional Sensitivity and Edge Detection FRITIOF S. SJOSTRAND

Department of Biology, University of California, Los Angeles, California Received August 31,1977 The earlier analysis of the circuitry of the outer plexiform layer in the rabbit retina [SjSstrand, F. S. (1969) in The Retina, p. 63. University of California Press, Los Angeles; (1974) J. Ultrastruct. Res. 49, 60; (1976) Vision Res. 16, 1; (1976) Japan. J. Ophthalmol.] has been extended with respect to the circuitry t h a t was proposed to be associated with directional sensitivity and edge detection. This circuitry involved one bipolar cell, here referred to as bipolar cell 1, which was connected to horizontal cells according to a characteristic pattern t h a t appeared suitable for a detector for directional sensitivity and for an edge. The same characteristics of the pattern of connections were found repeated at the contacts of this bipolar cell at a second photoreceptor cell of the fl type (a and E type cells are usually identified with rod and cone cells, respectively). The similarities involve the presence of collateral branches of bipolar cell 1 t h a t contact horizontal cell processes outside the subsynaptic neuropil, connections in the subsynaptic neuropil which involve both bipolar cell and horizontal cell processes, contacts with efferent bipolar cells in the subsynaptic neuropil, contact relations with horizontal cell processes at synaptic ribbon complexes, and contacts with the photoreceptor cells. Collateral branches are a unique feature of bipolar cell 1 among all bipolar cells contacting these two photoreceptor cells. As in the previous papers, the interpretation of the functional significance of the circuitry was based on distinguishing between two types of horizontal cell-bipolar cell connections. Horizontal cells were thus classified as contributing either processes t h a t showed a structural polarity or processes t h a t were lacking any structural polarity. In five synaptic ribbon complexes analyzed at photoreceptor cell E2, one of the two horizontal cell endings of each complex was contributed by a horizontal cell process with structural polarity, while the other was contributed by a process lacking structural polarity. This is in agreement with the most common a r r a n g e m e n t observed in earlier analysis (SjSstrand, 1974). It could also be confirmed t h a t the bipolar cell process to a synaptic ribbon complex was a side b r a n c h of the end b r a n c h which itself was received in a separate invagination of the plasma m e m b r a n e of the terminal. There were five horizontal cell processes with structural polarity t h a t contacted or were contacted by bipolar cell 1 in synaptic ribbon complexes, in the subsynaptic neuropil, and outside the subsynaptic neuropil. All these processes showed the same polarity which agreed with the polarity of corresponding horizontal cell processes' contacts at photoreceptor cell El. The systematic repetition of similar contact relations of bipolar cell 1 at both photoreceptor cells and the fact t h a t the similarities involved a characteristic, nonrandom but r a t h e r specific type of contact p a t t e r n support the concept t h a t the neurons are connected according to well-defined patterns designed for a particular type of information processing. As a consequence, circuit analysis of this kind together with electrophysiological analysis should form a concrete basis for revealing how information is processed in nervous centers. The subsynaptic neuropil showed an even more complex circuitry at this photoreceptor cell t h a n at photoreceptor cell El, supporting the earlier conclusion t h a t the subsynaptic neuropil is the site of a major part of retinal circuitry (SjSstrand, 1974, 1976, 1977). A pairing of contacts of endings of neurons was observed to be a general rule for contacts not only in the synaptic ribbon complexes but also in the subsynaptic neuropil and outside the subsynaptic neuropil. For horizontal cells this pairing involved one horizontal cell process with polarity paired with one lacking polarity like the most common combination at synaptic ribbon complexes. From a functional point of view the horizontal cell processes lacking structural polarity were interpreted to be involved in neural adaptation of the r e t i n a while the horizontal cell processes with structural polarity were interpreted to act 54

0022-5320/78/0621-0054502.00/0 Copyright © 1978 by Academic Press, Inc. All rights of reproduction in any form reserved.

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER

55

in an inhibitory way on bipolar cells on the basis of local, one-sidedconditions of illumination of the retina. This functional interpretation was the basis for proposing that the observed circuitry could be associated with directional sensitivity and edge detection. These proposals appear to have gained additional support from the fact that the same type of connections were found at both photoreceptorcell fll and/32. In previous papers (Sj6strand, 1969, 1974, 1976) the first steps in an analysis of the circuitry of the outer plexiform layer of the rabbit retina were described. The method that was used made it possible to approach the problem of circuitry analysis in a new way as compared to that used in classical neuroanatomy. Instead of describing the morphology of individual neurons based on the application of the Golgi technique in light microscopy or in electron microscopy, the present approach reveals as completely as possible the patterns according to which a group of neurons are connected in neuronal circuits. The justification for such an analysis is the assumption t h a t neurons are connected according to well-defined patterns, each pattern reflecting a particular type of information processing. A detailed knowledge of the circuitry should therefore make it possible to deduce what kind of processing is involved in different patterns of neuronal connections. In contrast to this type of analysis, the conventional approach in neuroanatomy furnishes information regarding the morphology of different types of neurons, each neuron being analyzed separately since the Golgi technique (which uses metal impregnation of neurons as a means of visualizing them) requires t ha t the metal impregnation be confined to individual neurons while the surrounding neurons are not impregnated. This description allows a morphological classification of neurons. In addition, certain information regarding types of contacts between neurons can be obtained by this technique particularly when combined with electron microscopic analysis (Stell, 1965, 1966; Blackstad, 1965). The Golgi technique has been very useful in analyzing patterns of connections

between nervous centers in the nervous system. It has also allowed drawing some general conclusions regarding types of relationships of neurons within nervous centers. This information has, however, not made it possible to reveal precise patterns of neuronal connections which form the circuitry designed for particular types of information processing. The reason for this is that the available information allows arranging neurons and interneuronal contacts according to an enormous number of arbitrarily chosen patterns. The actual patterns are much too complex to make it possible to deduce them on the basis of such general information. The lack of restrictions in trying to construct the circuitry from such information easily leads to oversimplified presentations of neuronal connections. Such schemes should not be confused with the type of circuit diagrams worked out on the basis of circuitry analysis referred to here. The method for analyzing the circuitry used in this work involves three-dimensional reconstructions of the circuitry from electron micrographs of serial sections. Since this is a most time-consuming procedure a major effort has been made to cut down the time involved. The present communication describes a simplified method for recording part of the threedimensional information by building a three-dimensional model using the building elements described by Sj6strand (1974). With this method the three-dimensional analysis reported in recent papers (SjSstrand, 1974, 1976) is being extended to include the connections of bipolar and horizontal cells with a large number of photoreceptor cells. The present report deals with the first stage in this extended analysis.

56

FRITIOF S. SJOSTRAND

In the earlier papers, a circuitry was described that was interpreted as possibly associated with directional sensitivity. The processing of the information which is the basis for directional sensitivity was shown by Barlow and Hill (1963) to be carried out in the retina of the rabbit. Certain ganglion cells respond with a train of spikes when an image moves in one particular direction, the preferred direction, over the receptive field of the ganglion cell, while movement in the opposite direction, the null direction, does not evoke any such response. This means that activation of a detector involved in this processing of the information must be inhibited when the movement of the image is in the null direction, while it is not inhibited when the image moves in the preferred direction. Since the detector must be activated by responses in photoreceptor cells within its receptive field, it must have been set in a state of inhibition before or at the same time the image is projected onto these photoreceptor cells when the image moves in the null direction. It is obvious that this inhibition must be provoked by the moving image and that therefore the inhibition must be imposed through lateral connections in the retina. It is also required that inhibition be confined to the image approaching from one particular side of the detector receptive field. As a consequence, the lateral inhibition must be unidirectional which requires lateral inhibition on the basis of local conditions of illumination on one side of the detector receptive field. It is furthermore characteristic that the detector mechanism involves units with receptive fields that are small in comparison to that of the ganglion cell and that therefore it is likely that there are several such units involved per ganglion cell. It seems obvious to ascribe an inhibitory function to the horizontal cells since these cells are associated with the mechanism of neural adaptation of the retina and

then function as inhibitory components. Neural adaptation is nondirectional. For unidirectional inhibition a particular arrangement of the horizontal cell processes would be required. In a previous paper (SjSstrand, 1974) it was proposed that certain bipolar cells furnished the basic units of the directional sensitivity detector circuit. The basis for this proposal was the particular way one type of bipolar cell was connected to horizontal cells. The pattern of connections satisfied the requirements mentioned above for a detector for directional sensitivity. Conditions for unidirectional inhibition through horizontal cells were fulfilled through connections with a number of horizontal cell processes that contacted photoreceptor cells on one side only of the receptive field of this bipolar cell. These horizontal cell processes would therefore inhibit the bipolar cell on the basis of the state of illumination on that side of the bipolar cell receptive field. It was characteristic that this type of bipolar cell was involved in a circuit with a large number of special features which seemed to rule out the possibility that these features were due to chance. For the next step in the analysis of the circuitry it was judged to be most important to establish whether the characteristic pattern of neuronal contacts of this bipolar cell would be consistent when including its contact with a second photoreceptor cell of the fl type that was available for analysis in the same series of sections. The analysis showed that the features that were the basis for associating the circuitry of this bipolar cell with directional sensitivity were even more pronounced in its contact relations at this second photoreceptor cell. A second type of circuit in which this bipolar cell was involved was proposed by SjSstrand (1976) as possibly associated with an edge detector mechanism. Also in this case consistency in the pattern of connections of the bipolar cell would be

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER required

as a necessary basis for this inter-

pretation. The extended analysis showed that the same pattern of contacts was repeated at the second photoreceptor cell. It was of basic importance to establish whether circuitry analysis pursued with this technique would reveal patterns of neuronal contacts that would be characteristic enough to allow identifying specific patterns that could be related to particular types of information processing. Or would t h e s e p a t t e r n s b e so d i f f u s e a n d g e n e r a l that such a conclusion would be excluded? The results show that the patterns of neuronal contacts are characteristic and that they therefore can furnish a basis for functional interpretations. MATERIALS AND METHODS For the present analysis the same material was used as t h a t described earlier (SjSstrand, 1974). Since the analysis is being extended to involve a r a t h e r large part of the outer plexiform layer eventually involving 70 to 80 photoreceptor cells it was important to build the three-dimensional model at not too large a scale. It was built at a magnification of 120,000. In order to speed up the work, a model was built which showed the paths along which neuronal processes extended and the contact relations of neurons but did not show the actual dimensions and shapes of the processes. For this purpose the simple building elements made of plastic t h a t were described earlier (SjSstrand, 1974) were used. Since the circuitry of the outer plexiform layer is very compact it was necessary to divide the region to be reconstructed into a n u m b e r of parts. The serial sections had been cut with the plane of the sections oriented close to parallel to a t a n g e n t to the surface of the retina. The area contained in each section was first divided into square-shaped regions with the side of the square measuring 43 cm. This corresponds to 3.6 /zm in the retina. With this dimension of the region, a single/3-type photoreceptor cell t e r m i n a l or a n u m b e r of c~-type photoreceptor cells would be included in one region. Due to the compactness of the circuitry and the r e q u i r e m e n t t h a t it be possible to observe the neuronal processes in the model in order to trace their paths, it was also necessary to subdivide each region in a radial direction; t h a t is, along planes parallel to the planes of the sections. These subdivisions include two to ten or more sections. The n u m b e r of sections in each subdivision was determined by how compact the circuitry was a r r a n g e d at t h a t level and by the requirement t h a t it should be possible to

57

see through all parts of the model. The positions of the profiles of neuronal processes in the electron micrographs were defined by placing the electron micrographs printed on t r a n s p a r e n t film on the top of a large viewing table with a square coordinate system drawn on its top surface. The coordinates of the region to be reconstructed were then determined, and a corresponding coordinate system at five times higher magnification was drawn on a square-shaped sheet of Plexiglas. The model was then built with the building elements positioned in relation to this coordinate system according to the coordinates read from the print on a viewing table top. For the subdivision of the region under reconstruction a new Plexiglas sheet with the coordinate system drawn was introduced, and the model building continued on this sheet as long as all processes could be easily observed t h r o u g h the entire subdivision. The fact t h a t the coordinate system was represented at many levels t h r o u g h the model made it easy at all levels to position the processes in the model in a proper way. The subdivisions belonging to one region were stacked on top of each other with proper spacers cut from Plexiglas rods as support between the sheets of plastic. The different reconstructed regions were then lined up to show the overall circuitry of the analyzed part of the r e t i n a (Fig. 1). To facilitate the building of the model and at the same time to make it easier to observe the intertwined process, the thickness of the individual sections was exaggerated by a factor of 2. Furthermore, the thickness of larger processes was indicated by making the processes extend in a vertical direction as far as their profiles appeared in the series of sections. This explains the regular lace-like patterns seen in the model. When tracing the neuronal processes in the electron micrographs, each profile of a process was numbered. Since the reconstruction started from the distal ends of the processes, there could be many profiles t h a t later were found to be branches from one and the same process. After the model had been finished and these relationships had been established, all branches originating from the same process were coded by painting in a color particular to t h a t process. Figures 1-3 illustrate the three-dimensional model. In Fig. 1 all parts of the model are assembled. The uppermost layers of the model are located in the i n n e r nuclear layer with some cell bodies of horizontal cells and bipolar cells indicated by an accumulation of multibranched building elements. The planes of the sections are oriented somewhat obliquely in relation to the boundary between the outer plexiform layer and the i n n e r nuclear layer. Figures 2 and 3 illustrate how the processes can be traced by stepwise removal of parts of the model.

58

FRITIOF S. SJOSTRAND

Fro. 1. The linear model assembled. The compact and complex a r r a n g e m e n t of the processes in the subsynaptic neuropil is furthermore illustrated in these two pictures. The detailed analysis of the circuitry was pursued by observing the paths along which the processes with all branches extended through the model and simultaneously tracing the processes in the electron micrographs. In this way it was possible to add to the information contained in the linear model detailed information regarding dimensions of the processes, precise contact relations between

neurons, presence of specialized structures at these contacts like synaptic vesicles, and specialized plasma membrane relations. The contact relations in the subsynaptic neuropil were frequently so complex that it was considered necessary for certain parts of the subsynaptic neuropil to build an additional anatomical model in which the profiles of the processes were cut out in sheets of plastic and piled up using spacers of the proper dimensions. Figures 4 and 5 show such a model involving some of the connections of bipolar

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER

59

FIa. 2. Part of the model removed to expose the subsynaptic neuropil of photoreceptor cell ~2 (RC~2). FIG. 3. Further removal of parts of the model exposes the subsynaptic neuropil at a level closer to the base of the terminal than in Fig. 2. Two a photoreceptor cell terminals are seen at lower right corner (RC a).

60

F R I T I O F S. S J O S T R A N D

iii~iiiii ii~i iiiiiiiiiiii i ii~iiill

~

i :: !ill¸ i ~ iii ~i~i~i ~ ~

Q

FIGS. 4 AND 5. A n a t o m i c a l model of a p a r t of t h e s u b s y n a p t i c neuropil o f p h o t o r e c e p t o r cell f~2 i n v o l v i n g s o m e of t h e c o m p o n e n t s a s s o c i a t e d w i t h t h e b i p o l a r cell 1 circuitry. T h e c o n n e c t i o n s of one collateral b r a n c h of bipolar cell 1 a r e also included in t h e model. T h i s collateral e x t e n d s to t h e left in t h e p i c t u r e s , w h e r e it c o n t a c t s one h o r i z o n t a l cell process b r a n c h i n g off from a l a r g e h o r i z o n t a l cell b r a n c h (18). A t t h i s c o n t a c t t h e bipolar cell 1 collateral is contacted by a h o r i z o n t a l cell process w i t h polarity (72).

61

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER

FIo. 6. Schematic presentation of a horizontal cell process with structural polarity.

II

J

(7>

i

,U

FIG. 7. Schematic presentation of a horizontal cell process lacking structural polarity.

62

FRITIOF S. SJOSTRAND

72

15

31

22

77 i

,

i I

....

i

_

.-...............~m i i i i

~31

Ii i I ]

il I

,

72

~

k

41 "~31

7

._i.

'I .........T

i

-

I

1 i i

72 1

v

Fia. 8. Diagram summarizing the contact relations of bipolar cell 1 at photoreceptor cell f12. The roman numbers mark the four collateral branches of bipolar cell 1. cell 1 in the subsynaptic neuropil that were of a particularly complex nature. This model represents only a part of the circuitry of the subsynaptic neuropil. In addition, the connection of one collateral of bipolar cell i is included in the model. This model was built at a magnification of 91,000. Due to the complex and compact arrangement of the neuronal branches in the subsynaptic neuropil it had to be subdivided into several parts. Figure 5 shows the model after one part has been removed exposing the branchings of the processes at a level closer to the receptor terminal than the level exposed in Fig. 4. The next step in the analysis involved drawing graphs showing the contact relations first of individual neurons and then of a group of neurons that were mutually connected. RESULTS

Some Definitions T h e e a r l i e r a n a l y s i s b y S j S s t r a n d (1974)

s h o w e d t h a t t h e h o r i z o n t a l cell processes t h a t w e r e c o n t a c t e d b y or c o n t a c t e d t h e e n d b r a n c h e s of o n e b i p o l a r cell app r o a c h e d t h e site of c o n t a c t f r o m p a r t i c u lar directions and not randomly. The latter p o s s i b i l i t y w a s r u l e d o u t b y t h e fact t h a t w h e n a b i p o l a r cell m a d e m u l t i p l e c o n t a c t s w i t h h o r i z o n t a l cells t h r o u g h s e v e r a l e n d b r a n c h e s t h e s a m e c o m b i n a t i o n of app r o a c h d i r e c t i o n s of t h e h o r i z o n t a l cell processes w e r e r e p e a t e d e v e n if t h e horiz o n t a l cell processes b e l o n g e d to d i f f e r e n t h o r i z o n t a l cells. S i n c e t h i s a r r a n g e m e n t of t h e c o n t a c t r e l a t i o n s is h i g h l y l i k e l y to be of f u n c t i o n a l s i g n i f i c a n c e it is c o n s i d e r e d i m p o r t a n t to specify t h e a p p r o a c h d i r e c t i o n s of h o r i z o n t a l cell processes.

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER

A

B

C

K

63

D

FIGURE 8 (continued)

The directions will be indicated by the four quarters, and the orientation of the four quarters with respect to the present model is the same as in the earlier model. The approach direction appears to be of particular importance when the end branch of a horizontal cell process that makes the contact is the most peripheral part of a process of the horizontal cell, meaning that its contacts with photoreceptor cells are confined to photoreceptor cells located only in the approach direction of the process as illustrated in Fig. 6. This means that photoreceptor cells located beyond the site of contact with the bipolar cell in the approach direction will not affect this horizontal cell process. The input from photoreceptor cells to such a process therefore shows a polarity with respect to its contact with the bipolar cell.

Such horizontal cell processes are referred to as processes with structural polarity or processes with polarity (SjSstrand, 1976). Other horizontal cell processes which usually were considerably larger in size made contacts with photoreceptor cells which were located on both sides of the site of contact with a bipolar cell; t hat is, both in the approach direction and beyond as illustrated in Fig. 7. Such horizontal cell processes are referred to as processes lacking structural polarity or processes lacking polarity (SjSstrand, 1976). Processes with polarity were interpreted by Sj5strand (1974, 1976) to be functionally different from processes lacking polarity. The subsynaptic neuropil was defined by SjSstrand (1974) as a region located immediately vitread of the photoreceptor cell terminals where branches from hori-

64

FRITIOF S. SJOSTRAND Horizontal Cell Processes Lacking Polarity

Horizontal Cell Processes With Polarity

Bipolar Cell Processes

Efferent Bipolar Cell Processes of the RS.Type

~

>

>

Process with Synaptic Vesicles

A v

Ending of Process

Widened Part of Process

Widened Part of Process Containing Synaptic Vesicles

Specialized Contact Relations Between Widened Part of One Process and Another Process

-//////////.

Invaginated Ending of Process

Synaptic Ribbon Complex

IN FIG, 14 IN FIG.8

Lateral Contact of Synaptic Ribbon Complex

~

Soma of Neuron

m

FIGURE 8 (continued)

zontal cells and bipolar cells are closely packed without any Mfiller's cells processes being interposed between these branches. A few branches passed through this region to make contact with the photoreceptor cell terminal without making any special contacts with other neurons within the subsynaptic neuropil. Most branches, however, either ended in this

region or sent off side branches ending in contact with other neurons in the subsynaptic neuropil. Some examples are shown in Fig. 8. Through such contacts a considerable part of the retinal circuitry is contained in the subsynaptic neuropil (Sj6strand, 1976). The earlier analysis revealed the presence of processes contacting the photore-

Fro. 9. Section number 84 in the series through the terminal ofphotoreceptor cell f12 showing several of the synaptic ribbon complexes (indicated by letters) and a part of the subsynaptic neuropil. A long arrow points to area where bipolar cell process 41 sends a short side branch that contacts bipolar cell process 1. Notice the dense aggregation of synaptic vesicles in process 41 except for the side branch to process 1. Synaptic vesicles are also present in two end branches of efferent bipolar cell 15. × 36,000.

65

66

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER ceptor cell terminal with deeply invaginated endings containing synaptic vesicles arranged just as densely as those in the photoreceptor cell terminal. The dense arrangement of synaptic vesicles in these endings was used as one criterion to classify these processes as belonging to efferent neurons. Since the processes appeared identical to bipolar cell processes in other respects (see discussion in SjSstrand, 1974), it was assumed t ha t these processes belonged to efferent bipolar cells. This naming is used but it is pointed out t hat this interpretation is tentative. These efferent bipolar cells could contain densely arranged synaptic vesicles in their end branches contacting the photoreceptor cell terminal. Bipolar cell 15 was of this type, and two of its endings contacting the photoreceptor cell terminal is shown in Fig. 9. Other bipolar cells contained densely arranged synaptic vesicles in a large number of end branches to the subsynaptic neuropil, while the end branches contacting the photoreceptor cell terminal did not contain any synaptic vesicles. These bipolar cells can be interpreted to be efferent with respect to the neurons contacted in the subsynaptic neuropil. Bipolar cell 41 was of this latter type. Figures 9 and 10 show one branch of this bipolar cell ending in the subsynaptic neuropil and containing a dense aggregation of synaptic vesicles. This type of bipolar cell is also tentatively referred to as efferent bipolar cells. It seems practical to distinguish these two types of efferent bipolar cells as efferent bipolar cell of the RS type (R for receptor cell, S for subsynaptic neuropil) and of the S type (S for subsynaptic neuropil).

67

General Description of Bipolar Cell i Processes Bipolar cell 1 entered the analyzed region from the west as a 0.54-t~m-wide process and split into two branches, one branch ending at the photoreceptor cell analyzed earlier, photoreceptor cell ill, and the other branch ending at the photoreceptor cell around which the present analysis has been centered which will be referred to as photoreceptor cell f12. At the branching of the bipolar cell 1 process, this process widened to 0.71 t~m, and it extended a thick short side branch which contacted horizontal cell process 4. The branch to photoreceptor cell /~2 was 0.23 t~m thick. It widened when entering the subsynaptic neuropil to 0.33 tLm. After sending off one branch to one synaptic ribbon complex, it split up into two 0.33t~m-thick main branches t hat passed through the subsynaptic neuropil. Both main branches in t urn split up into two branches. Of the four branches formed this way, three eventually left the subsynaptic neuropil and made contacts with other neurons outside the subsynaptic neuropil (Fig. 8). The parts of these branches t hat extended outside the subsynaptic neuropil will be referred to as collaterals. The thickness of these collaterals was 0.30 to 0.33 tLm. Figure 11 shows three collateral branches of bipolar cell 1, labeled C1, C2, and C3, as they emerge from the subsynaptic neuropil. Two of the branches are also seen located in the subsynaptic neuropil. A fifth branch was traced within the subsynaptic neuropil (Figs. 9 and 10). Its end appeared to make contact with the terminal, but the most distal end part of

FIG. 10. The section followingthat pictured in Fig. 9 (section 85) through the terminal ofphotoreceptor cell f12, A long arrow points to the area that correspondsto that indicated by a long arrow in Fig. 9. In this picture it is shown that the branch of the bipolar cell 1 process that was contacted by a short side branch of bipolar process 41 in Fig. 9 is passing by this latter branch over its sclerad surface (section 85 is located sclerad in relation to section 84). There is an accumulation of synaptic vesicles in this branch of the bipolar cell 1 process, x 33,000.

68

FRITIOF S. SJOSTRAND

this b r a n c h extended out of the region covered by this series of sections. In addition, t h e r e were two short b r a n c h e s e x t e n d i n g from these m a i n b r a n c h e s and e n d i n g in the s u b s y n a p t i c neuropil. Bipolar cell 1 was the only bipol a r cell contacting photoreceptor cell f12 t h a t g a v e off collaterals. The s a m e applied to bipolar cell 1 at photoreceptor cell ill. For comparison, the dimensions of the b r a n c h to photoreceptor cell fll will be given. It m e a s u r e d 0.34 tLm in t h i c k n e s s close to the b r a n c h i n g point, and l a t e r the d i a m e t e r was reduced to 0.25 t~m. It m a i n tained this w i d t h for some distance b u t before e n t e r i n g the s u b s y n a p t i c neuropil of photoreceptor cell fll the d i a m e t e r increased to 0.42 tLm, a n d in the s u b s y n a p t i c neuropil it increased f u r t h e r to 0.50 t~m. The width of the collaterals was 0.21 tzm.

Horizontal Cell Processes Associated with Bipolar Cell Process 1 One thick horizontal cell process (31) lacking p o l a r i t y w a s p a r t i c u l a r l y intim a t e l y associated w i t h the t e r m i n a l of photoreceptor cell f12. It passed j u s t vit r e a d of the s u b s y n a p t i c neuropil (Fig. 11) and sent off t h r e e b r a n c h e s to t h r e e synaptic ribbon complexes. It was contacted in all t h r e e synaptic ribbon complexes by t h i n b r a n c h e s f r o m bipolar cell 1. In addition, it sent off several t h i n b r a n c h e s to the s u b s y n a p t i c neuropil which showed p a r t i c u l a r preference for contacts w i t h bipolar cell 1 branches. It t h u s contacted bipolar cell 1 b r a n c h e s a t t h r e e different sites in the s u b s y n a p t i c neuropil. In addition, it contacted two of the collaterals of bipolar cell 1. Horizontal cell process 31

m e a s u r e d 0.88 tLm in d i a m e t e r p r o x i m a l to the t e r m i n a l of photoreceptor cell fi2 and 0.67 t~m distal to this t e r m i n a l . T h e d i a m e t e r was reduced to 0.42 tLm before it contacted photoreceptor cell fll a n d s e n t one ending to one synaptic ribbon complex (C) of t h a t t e r m i n a l . It did not m a k e a n y contacts w i t h bipolar cell 1 at the t e r m i n a l of photoreceptor cell ill. A second thick horizontal cell process lacking polarity (18) a p p r o a c h e d the term i n a l from the west and m e a s u r e d 2.1 t~m in d i a m e t e r (Fig. 12). A t h i r d horizontal cell process lacking polarity (7) approached from n o r t h and m e a s u r e d 0.92 ~ m in d i a m e t e r . It sent a t h i n b r a n c h only 0.17 to 0.29 t~m in a sclerad direction to the t e r m i n a l . A fourth horizontal cell process lacking polarity (4) w a s a b r a n c h from one of the large horizontal cell process (4 in earlier analysis) associated w i t h photoreceptor cell fll t h a t a p p r o a c h e d f r o m the north. The horizontal cell processes with polarity m e a s u r e d 0.42 to 0.46 t~m in d i a m e t e r , while the bipolar processes, w i t h t h e exception of bipolar cell 1, m e a s u r e d 0.25 to 0.29 t~m in d i a m e t e r . The s o m a of one horizontal cell (3) t h a t sent a b r a n c h t h a t contacted bipolar cell 1 in t h r e e synaptic ribbon complexes in photoreceptor cell fll was located close to photoreceptor cell f12 a n d west of this receptor cell. The process to photoreceptor cell fll was one with polarity, which sent off a side b r a n c h t h a t ended in contact w i t h horizontal cell process 4 as described e a r l i e r (SjSstrand, 1974). A n o t h e r t h i n n e r b r a n c h e x t e n d e d from the s o m a of this horizontal cell in an

Fro. 11. Section number 67 through the terminal of photoreceptor cell f12 at a level located considerably further in a vitread direction than that of the sections in Figs. 9 and 10. The terminal occupies only small areas (T) in this section next to the subsynaptic neuropil. Three of the collateral branches of bipolar cell 1, labeled C1, C2, and C3, are shown as they leave the subsynaptic neuropil. Notice the accumulation of vesicles and granules in process 29. Slightly left to the middle of the picture a widened part of horizontal cell process 4 is in contact with one bipolar cell 1 branch with synaptic vesicles in process 4. Efferent bipolar cell 15 also contacts bipolar cell 1 at the same region. The three-dimensional reconstruction showed that the area of contact of bipolar cell 1 to efferent bipolar cell 15 and horizontal cell 4 is considerably more extensive than shown in this micrograph and that the synaptic vesicles in processes 15 and 4 were concentrated at the area of contact. × 33,000.

69

70

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER

eastern direction and ended in contact with one of the bipolar cell i collaterals at photoreceptor cell ~1. The contact was located where the ending of this collateral contacted horizontal cell process 4 approaching from the north. In this way the horizontal cell contacts with the collateral were paired. Another thin branch contacted photoreceptor cell f12 outside any synaptic ribbon complex and sent several branches contacting neurons other than bipolar cell 1 in the subsynaptic neuropil. A wider process extended from the soma in a northern direction. The soma of horizontal cell 3 was in close apposition to that of another horizontal cell that contacted a number of photoreceptor cells of the a type. The contact with this horizontal cell soma was along the southern surface of the soma of horizontal cell 3. Although the entire soma of this horizontal cell was not included in this series of sections, it was possible to extend the tracing far enough into the inner nuclear layer to exclude the possibility that this horizontal cell sent off any processes extending in a southern direction. In addition to the processes just mentioned which originated from the same region of the soma, one process extended from the soma westward. This horizontal cell was characterized by the fact that all these five processes extending from the soma were r a t he r thin. In addition, three of them t ha t were traced to their ends were short with two contacting one single photoreceptor cell each, while the third did not contact any photoreceptor cell.

Contact Relations of Bipolar Cell 1 at Photoreceptor Cell f12 Bipolar cell 1 contacted photoreceptor cell f12 with one 0.25-t~m-thick end branch

71

which was received in a deep invagination of the plasma membrane of the receptor cell terminal. In addition, it sent off three side branches t hat contacted the photoreceptor cell at three synaptic ribbon complexes. The contacts of bipolar cell 1 with horizontal and bipolar cell processes were established at three levels: at synaptic ribbon complexes, in the subsynaptic neuropil and in the neuropil of the outer plexiform layer involving the collaterals. All contacts are accounted for in Fig. 8. Most contacts that were considered to be of functional significance were characterized by end branches that extended from horizontal cell processes or the bipolar cell branches that ended at the area of contact, end contacts. Most of the endings of the horizontal cells contained synaptic vesicles, and frequently the plasma membranes showed a specialized structure at the area of contact. In addition, contacts of branches passing by were considered functional when (1) the contact involved a widened part of this branch, (2) the plasma membranes at the area of contact showed special features forming a thin boundary due to the minute width of the space separating the two plasma membranes, and (3) the branch passing by contained synaptic vesicles at the contact region. These contacts are referred to as passing by contacts. It was characteristic t hat the contacts of horizontal cells and bipolar cells to bipolar cell i were paired; t hat is, pairs of neurons contacted bipolar cell 1 while mutually in contact like the triad arrangement at the synaptic ribbon complexes. One of the two contacts was always an end contact while the second contact could be either an end contact or a passing by contact. Figure 8 illustrates that there were a

FIG. 12. Section 83 t h r o u g h the t e r m i n a l of photoreceptor cell f12 showing the end of collateral 3 of bipolar cell 1 in contact with a b r a n c h of horizontal cell 18. Notice the a c c u m u l a t i o n of synaptic vesicles in process 18 at area of contact. Horizontal cell process 72 ends in contact with the bipolar cell 1 collateral and horizontal cell 30. x 33 000.

72

FRITIOF S. SJOSTRAND

total of nine horizontal cells and five bipolar cells that were involved in contact relations with bipolar cell 1. Five of the horizontal cells contributed processes with polarity which all approached the photoreceptor cell f12 terminal from the north. The processes of the four other horizontal cells were lacking polarity and approached either from the north or west. It was characteristic that the contacts with bipolar cell 1 were paired at all levels of contact.

Contacts at synaptic ribbon complexes. The three end branches that bipolar cell 1 sent off to three synaptic ribbon complexes contacted both horizontal cell endings in each complex simultaneously. The contact was confined to the narrow part of the horizontal cell end branches proximal to the widened end vacuole. This corresponds to the description presented earlier for the most common type of bipolar cell contacts at synaptic ribbon complexes (SjSstrand, 1958, 1974). The thickness of the end branches of bipolar cell 1 varied between 0.17 and 0.19 t~m. In the subsynaptic neuropil, five of the nine horizontal cell processes mentioned above were involved in contacts with bipolar cell 1. There were a total of 23 contacts between bipolar cell 1 and other neurons in the subsynaptic neuropil that were considered likely to be functional contacts. Twelve of these contacts were established by end branches ending in contact with bipolar cell 1. There were a total of 11 sites of contact with two different neurons contacting bipolar cell 1; the neurons were closely apposed at nine of these sites. At two sites, three different neurons were involved. It was characteristic that at all 11 of the 11 sites of contact, bipolar cell 1 was contated by other bipolar cells. The bipolar cell contacts were predominantly paired with contacts of horizontal cell processes with polarity. Of the five bipolar cells that contacted bipolar cell 1 in the sub-

synaptic neuropil, two were classified as efferent bipolar cells: one of the RS type (15) and the other of the S type (41). Bipolar cell 41 showed a characteristic relationship with respect to bipolar cell 1. First, it contacted bipolar cell 1 at four different sites in the subsynaptic neuropil: at two sites with end branches ending in contact with bipolar cell 1 and at two sites with end branches to the photoreceptor cell passing by bipolar cell 1. In both the latter cases it was a widened part of the process of bipolar cell 41 that made the contact. Second, processes of bipolar cell 41 contained numerous synaptic vesicles in the subsynaptic neuropil but no vesicles at the areas of contact with bipolar cell 1 or with the photoreceptor cell. This is illustrated in Figs. 9 and 10. Third, the bipolar cell 1 processes contained synaptic vesicles at the areas of contact with bipolar cell 41 (Figs. 9 and 10). Fourth, the plasma membranes showed a specialized arrangement at the sites of contact with narrow separation of the two plasma membranes. Furthermore, at three of the contacts of bipolar cell 41 with bipolar cell 1, the contacts were paired with one and the same horizontal cell with polarity (72). The collaterals established contacts with eight of the ten horizontal cell processes. Five of these eight processes were processes with polarity. Each end of the four collaterals was in contact with pairs or, in one case, a triplet of horizontal cell processes. With only one exception, one horizontal cell process with polarity was paired with one lacking polarity. At the contact that was an exception, two horizontal cell processes with polarity were paired. However, at a second adjacent site of contact with this collateral, one of these horizontal cell processes was paired with a horizontal cell process lacking polarity. Figures 12 and 13 illustrate the contact relations at the end of one of the collaterals. The end of this collateral is in contact with a branch from a large horizontal cell process (18) lacking polarity. An end

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER

73

@ FIG. 13. Drawing from three-dimensional anatomical model showing contact relations at end of collateral b r a n c h 3 of bipolar cell 1.

branch of horizontal cell 72 with polarity Horizontal Cell-Horizontal Cell Contacts contacts bipolar cell I as well as horizontal In agreement with the observations cell 30. made in the earlier analysis (SjSstrand, Figure 13 is a drawing of this part of 1974), horizontal cells contacted each other the model shown in Figs. 4 and 5 and through special branches. This type of illustrates the relationships between the connection was, however, confined to horend branches of bipolar cell 1 and hori- izontal cells with polarity contacting or zontal cell 72 at the site of contact. The being contacted by horizontal cells lacking drawing shows that horizontal cell process polarity. 72 contacts the bipolar cell 1 collateral at Such connections also involved two horits end immediately adjacent to the area izontal cell processes with polarity apwhere this collateral contacts horizontal proaching from the north that contacted cell 18. It furthermore shows that this two horizontal cell processes approaching process of horizontal cell 72 contacts horifrom the sourth. zontal cell process 30 with its very end. Process 30 extends a small process which Confirmation of Some Other Observations Made Earlier is in contact with the end of the process Some observations described in earlier of horizontal cell 72. Process 72 is a process with polarity, while horizontal cell process papers (Sj5strand, 1974, 1976, 1977)were 18 is one lacking polarity. This is an ex- confirmed in this extended analysis. The topographical arrangement of horizontal ample of a paired contact. Notice in Fig. 12 the accumulation of cell and bipolar cell endings at synaptic synaptic vesicles in horizontal cell process ribbon complexes described earlier could 18 at its contact with bipolar cell 1. Hori- be confirmed in all eight complexes reconzontal cell process 30 which approached structed in photoreceptor cell B2. The horthis area from south also contained syn- izontal cell processes contributing endings aptic vesicles at the area where it was to five of these complexes were traced contacted by horizontal cell process 72. extensively enough to make it possible to

74

FRITIOF S. SJOSTRAND

i

4

CONTACTS OUTSIDE SUBSYNAPTIC NEUROPIL

CONTACTS WITH PHOTORECEPTOR CELLS CONTACTS IN SUBSYNAPTIC NEUROPIL

® FIG. 14. Diagram showing contact relations in a simplified version of bipolar cell 1 at photoreceptor cells fll and /~2. The diagram illustrates the basic similarities in contact patterns at both photoreceptor cells. See Fig. 8 for explanation of symbols. establish t h a t one ending of a horizontal cell process with polarity was paired w i t h one horizontal cell process lacking polarity. It was also established t h a t all bipolar cell end branches to synaptic ribbon complexes were small side branches of a major end b r a n c h received in a separate invagin a t i o n of the receptor cell plasma membrane. DISCUSSION

Specificity of neuronal contact patterns. This analysis has shown t h a t bipolar cell 1 was contacted by or contacted a n u m b e r of horizontal cell processes at synaptic ribbon complexes in the sybsynaptic neu-

ropil and outside the subsynaptic neuropil by m e a n s of end contacts. T h e r e were a total of 36 contacts. The contacts involved both horizontal cell processes with polarity and processes lacking polarity. It was characteristic that

all processes with polarity approached from the north. These processes contained synaptic vesicles at the area of contact. This is i n t e r p r e t e d to indicate t h a t these horizontal cells are presynaptic components. The connections of bipolar cell 1 with horizontal cell processes with polarity follows a p a t t e r n at photoreceptor cell f12 which is the same as t h a t observed at photoreceptor cell fll in t h a t all these

CIRCUITRY

OF RABBIT

RETINA

PLEXIFORM

\

75

LAYER

5¸

p Rcl

H

ib l i i i

J H %

- - _ ~

:

L~

. . . .

.--.

iI! !::::::::-iiiiii:[ Ill ......

:,

,L

•

L

l : .... : :

F I a U R E 14

processes approached from the north. The sending off of a number of collaterals extending outside the subsynaptic neuropil is a special feature of this bipolar cell both at photoreceptor cells fll and f12. On the whole, it is obvious from Fig. 14 that the circuitry that was characteristic for bipolar cell 1 in its connections with photoreceptor cell ~1 is duplicated at photoreceptor cell f12. There are, however, quantitative differences. At photoreceptor cell fi2 there were eight horizontal cell processes approaching from the north which were involved in contact relations with bipolar cell 1, wbJle there were four at photoreceptor cell/31. Furthermore, bipolar cell 1 sent offfour collateral branches at photoreceptor cell f12 and two at photoreceptor cell/~1.

........... ®

(continued)

There are other similarities. For instance, it is characteristic that at both photoreceptor cells fll and f12 the contacts of bipolar cell 1 involved other bipolar cells and horizontal cells in the subsynaptic neuropil but predominantly or entirely horizontal cells at the collaterals. Bipolar cell ! furthermore was contacted by or contacted processes of other bipolar cells, and at both receptor cells these contacts involved two efferent bipolars, one of the RS type and one of the S type. The fact that the particular pattern of horizontal cell and bipolar cell contacts that was characteristic for bipolar cell I at photoreceptor cell fil is repeated in an even more pronounced way at photoreceptor cell f12 further supports the conclusion that this pattern is a specific pattern.

76

FRITIOF S. SJOSTRAND

This is important since it means that the contacts between neurons follow particular patterns that vary with different neurons belonging to a morphologically identical group of neurons. The neurons therefore now can be classified with respect to the type of circuit or circuits in which they participate. This will eventually lead to a more specific and functionally more precise classification than the present classification based on morphological features observed in Golgi preparations. It will be of interest to see to what extent this classification can be correlated to the latter morphological classifications. That specific circuitry patterns exist can be interpreted to mean that a particular type of information processing requires a certain precise pattern of interneuronal contacts. It speaks against a circuitry that predominantly would be a diffuse neuronal network. As a consequence, it appears obvious that a sufficiently detailed knowledge of the circuitry reflecting a particular type of information processing should make it possible to deduce how this processing is carried out by the circuit. The technical limitations with respect to direct recording of the electrical events associated with information processing make it furthermore obvious that the knowledge of the circuitry will become a prerequisite for analyzing information processing in the nervous system. The circuitry associated with bipolar cell 1 at photoreceptor cell 1 was interpreted to be associated with directional sensitivity (SjSstrand, 1974). In addition, the combination of horizontal cell processes contacted by bipolar cell 1 at synaptic ribbon complexes was considered a suitable arrangement for the circuit of an edge detector (SjSstrand, 1976). These interpretations of the circuitry are based on the assumption that the horizontal cells are inhibitory components in the circuits, that the level of inhibition is proportional to the input from photorecep-

tor cells, and that the horizontal cells inhibit bipolar cells. The input to a horizontal cell from photoreceptor cells is furthermore assumed to be graded according to the intensity of the light falling on the photoreceptor cells. It is also assumed that there is considerable attenuation of the response within the horizontal cell processes, particularly in thin processes. Due to this attenuation or to the horizontal cells belonging to two systems of horizontal cells which are responding to a great extent independently of each other (SjSstrand, 1976), the bipolar cells will be exposed to inhibition based on two different conditions of illumination of the retina. While one type of horizontal cell processes, those lacking polarity, inhibit on the basis of the level of illumination over a large area of the retina around the receptive field of a bipolar cell, the other type of horizontal cell processes, those with polarity, would inhibit on the basis of local conditions of illumination. As pointed out earlier (SjSstrand, 1974), the constant and structurally well-defined simultaneous double contact of bipolar cell endings with horizontal cells in the synaptic ribbon complexes makes sense if the information transferred by the two horizontal cells is different. The edge detector. If we now consider the case of an edge detector, then bipolar cell 1 would qualify for the following reasons. At the synaptic ribbon complexes contacted by bipolar cell 1, one of the two horizontal cells was one with polarity. It would therefore inhibit bipolar cell 1 on the basis of the level of illumination over an area located in one particular direction from photoreceptor cells fll and f12. The other horizontal cell process contacted by bipolar cell 1 in the synaptic ribbon complexes was always one lacking polarity. These latter connections would inhibit bipolar cell 1 on the basis of the general background level of illumination. The responses of bipolar cell 1 would thus always be adjusted to that level of illumination.

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER In this way neural adaptation will influence the horizontal cell-bipolar cell interaction at all these contacts. For inhibition through horizontal cell processes with polarity to be of special significance it is required that the same polarity be systematically repeated at all contacts of bipolar cell 1 at synaptic ribbon complexes. Bipolar cell 1 sent endings to a total of eight synaptic ribbon complexes at the terminals of photoreceptor cell fll (5) and f12 (3). In four of these, the horizontal cell processes with polarity approached from the north and in two, from the west (photoreceptor cell riD. In one synaptic ribbon complex two horizontal cell processes lacking polarity which approached from the north and west were involved, and in one the approach direction of one of the horizontal cells could not be established with certainty. North is the predominant approach direction, and only one horizontal cell process with polarity was an exception with its approach direction being west. However, this was furnished by horizontal cell 3 which had no connections with photoreceptor cells in a southern direction. Instead, the process to photoreceptor cell 1 originated from the soma next to a process extending north. It is therefore conceivable that functionally this process should be considered identical to those with polarity that approached from north. There is thus a striking consistency with respect to the contact relations of bipolar cell 1 with photoreceptor cells fll and f12. Such a systematic structural arrangement of the circuitry is likely to have a functional significance. One simple interpretation of this functional significance appears to be to relate this circuitry to that of an edge detector mechanism. An edge detector is required to sense a difference in the level of illumination on one side of a boundary from that on the opposite side. When placing the detector close to the boundary, it would be required

77

that the detector receives information with respect to the state of illumination locally on one side of the boundary in relation to the overall background level of illumination. One way this could be achieved is by inhibitory influence suppressing the detector response when the illumination is uniform on both sides of the boundary, while the inhibition is released when there is a difference in the levels of illumination. Uniform illumination on all sides of photoreceptor cells fll and fi2 will mean that bipolar cell 1 will be exposed to inhibition by all horizontal cells it is in contact with. A lower level of illumination over the area located north will reduce the inhibition on bipolar cells 1 by reducing the inhibition imposed by horizontal cell processes with polarity approaching from the north. Bipolar cell 1 could then respond when the level of illumination is lower than the background level north of the photoreceptor cells it contacted. Such a difference corresponds to projecting an image of an edge with the shade cast by the edge located north of these photoreceptor cells. Figure 15 illustrates schematically the circuitry that is proposed to be associated with edge detection. The actual position of the image of the edge on the retina in relation to photoreceptor cells contacted by bipolar cell 1 would be determined by the distance between this photoreceptor cell and the nearest located photoreceptor cells contacted by the horizontal cell processes with polarity. For bipolar cell 1 to code for the presence of an edge as well as the location of this edge on the retina it would be required that bipolar cell 1 only contacted photoreceptor cells located along a line extending from west to east. This is highly unlikely. If we, however, reduce the requirements with respect to precision in locating the edge, it would be possible for bipolar cell 1 to function as an edge detector. We then have to assume that at all

78

FRITIOF S. SJOSTRAND

LJ ® Fro. 15. Schematic presentation of circuit for an edge detector.

contacts of bipolar cell 1 with photoreceptor cells the same pattern with respect to horizontal cell contacts should be repeated. This would secure coding for an edge with one particular orientation located anywhere over the entire receptive field of bipolar cell 1. In the visual center in the brain, the precise location of the edge could be established by the combined information from several edge detectors with overlapping receptive fields, or by combining the informarion regarding the edge with some other information t h a t would allow a precise positioning of the edge.

The detector for directional sensitivity. Compared to an edge detector, the situation is different in the case of a detector for the movement of an image in one preferred direction. In that case we can consider the image of an edge moving over the retina. It can be an edge delimiting

an area with either a higher or lower brightness than the background illumination. The directional sensitivity can be explained by suppression of the response of a detector when an image moves in the null direction. Such directionality requires that the horizontal cells, as possible inhibitory components of the circuit, must exert inhibition on the basis of local conditions of illumination. The arrangement of the horizontal cell processes with polarity appears to satisfy this condition and it therefore seems reasonable to assume that the null direction is north to south. There seem to be different requirements with respect to the timing of the inhibition in the case of a detector for directional sensitivity as compared to an edge detector. In the circuit of a detector for directional sensitivity, it seems advantageous for the inhibition to become efficient before

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER the moving image passes over the receptor cells furnishing the input to the circuit. Contacts of bipolar cell i with horizontal cell processes proximal to the synaptic ribbon complexes would then be advantageous because they could impose a certain level of inhibition at an earlier time than the contacts at synaptic ribbon complexes at photoreceptor cells fll and f12. As a consequence, connections with horizontal cell processes with polarity that do not contact the terminals of photoreceptor cells/~1 and f12 would be particularly efficient from this point of view. For those reasons the contacts made by the collaterals of bipolar cell 1 were considered particularly important for qualifying bipolar cell 1 as part of a circuit associated with directional sensitivity. The contacts bipolar cell 1 made through its collaterals involved horizontal cell processes with polarity that approached from the north. Of these horizontal cell processes, there were three processes altogether with polarity, one at photoreceptor cell fll and two at photoreceptor cell f12, that contacted bipolar cell 1 without contacting the terminals of photoreceptor cells/~1 and ~2. There is thus an elaborate and systematic arrangement of contacts between bipolar cell 1 and horizontal cell processes with polarity approaching from the north at the collaterals of bipolar cell 1. It is of importance to consider how efficiently the effects of these contacts of bipolar cell 1 with horizontal cells can affect the state of polarization in the main trunk of the dendritic arborization of bipolar cell 1. The effect on the main trunk will be determined by the attenuation of any state of polarization imposed at these contacts. One factor that will affect attenuation is the diameter of the processes. It was then characteristic that the dendritic branch to photoreceptor cells E1 and fi2 was considerably thicker than those of other bipolar cells and that the collaterals as well as

79

the branches in the subsynaptic neuropil were rather thick (0.33 t~m), while the end branches to the synaptic ribbon complexes and to the photoreceptor cells were considerably thinner (0.17 to 0.19 ~m and 0.25 tLm, respectively). These dimensions would indicate that the effects on the collaterals and on the branches in the subsynaptic neuropil could spread along the bipolar cell 1 process with limited attenuation. The pairing of the contacts of bipolar cell 1 and horizontal and bipolar cell processes is proposed to potentiate the effect of transmitter release by allowing the transmitter molecules released from both endings of a pair to act on the same receptor sites on the postsynaptic membrane. Under such conditions a response stronger than that corresponding to a mere summation of transmitter release from the two endings could be conceived of as a consequence of a cooperative effect. Such a mechanism was proposed for the horizontal cell-bipolar cell interaction at the synaptic ribbon complexes (SjSstrand, 1974). It is a simple explanation for the fact that the horizontal cell contacts with bipolar cells at synaptic ribbon complexes are always paired. The present study has shown that such pairing occurs in a general way. In conclusion, the contacts of bipolar cell 1 with horizontal cells with polarity are considered to furnish a basis for associating bipolar cell 1 with a circuit for directional sensitivity on the following grounds. These connections allow modulating the state of polarization of bipolar cell 1 on the basis of the level of illumination in one particular direction (north). The contact relations expose bipolar cell 1 to this influence before any change in the level of illumination occurs at the photoreceptor cells fll and fi2. The large number of contacts, the pairing of the contacts, and the large size of the bipolar cell branches and collaterals

80

FRITIOF S. SJOSTRAND

contacted by horizontal cells favor efficiency of the modulating influence exerted by those horizontal cell processes (for further discussion, see SjSstrand, 1974). It is characteristic that at both photoreceptor cells fll and fi2 bipolar cell 1 was contacted by other bipolar cells and in both cases by two efferent bipolar cells. The functional significance of these connections is not obvious because we do not as yet know enough about the overall connections of these efferent bipolar cells. One possible function of these connections could be that of connecting several bipolar cells associated with a detector mechanism for directional sensitivity. Bipolar cell 1 can be considered as a basic component in a circuit coding for directional sensitivity. Over the receptive field

i

l

of a ganglion cell in this circuit, there must be a number of such components. The efferent bipolar cells could connect such components to synchronize their responses to moving images through facilitation. That such a mechanism might be involved was indicated by the contacts between efferent bipolar cell 41 and bipolar cell 1. There were four such contacts, and it was characteristic that at the areas of contact there were synaptic vesicles in bipolar cell 1 but no synaptic vesicles in bipolar cell 41, although other parts of the processes from this bipolar cell contained dense aggregates of synaptic vesicles. This could be interpreted to indicate that bipolar cell ] is presynaptic in relation to bipolar cell 41.

l

1

® FIG. 16. Schematic presentation of basic features of horizontal cell connections of bipolar cell 1. The roman figures indicate the three levels of such contacts: I, at synaptic ribbon complexes; II, in subsynaptic neuropil; and III, outside the subsynaptic neuropil. The latter contacts were established by collateral branches of bipolar cell 1. Small arrows indicate input to horizontal cells at their connections with photoreceptor cells. Large arrows indicate the polarity of the horizontal cell influence on bipolar cell 1 due to the structural polarity of certain horizontal cell processes. Horizontal cell processes lacking structural polarity are indicated by thin lines. The pairing of contacts of connections to bipolar cell 1 of horizontal cell processes with structural polarity and horizontal cell processes lacking structural polarity is illustrated.

CIRCUITRY OF RABBIT RETINA PLEXIFORM LAYER

While bipolar cell 1 could exert a modulating influence on bipolar cell 41, it is possible that the efferent bipolar cell 15 could exert a similar effect on bipolar cell 1. Bipolar cell 15 contained synaptic vesicles at the areas of contact with bipolar cell 1 and could therefore be conceived of as being presynaptic to bipolar cell 1. REFERENCES BARLOW, H. B., AND HILL, R. M. (1963) Science 139, 412. BLACKSTAD, J. W. (1965) Z. Zellforsch. Mikrosk. Anat. 67, 819.

81

SJOSTRAND,F. S. (1958) J. Ultrastruct. Res. 2, 122. SJOSTRAND,F. S. (1969) in STRAATSMA,B. R., HALL, M. 0., ALLEN, R. A., AND CRESCITELLI,F. (Eds.), The Retina: Morphology, Function and Clinical Characteristics, p. 63, University of California Press, Los Angeles. SJOSTRAND,F. S. (1974)J.Ultrastruct. Res. 49, 60. SJOSTRAND,f . S. (1976) Vision Res. 16, 1. SJ6STRAND, F. S. (1976) in YAMADA, E. AND MISHIMA, S. (eds.), Structure of the Eye III, p. 281. Japan. J. Ophthalmol. STELL,W. K. (1965)Anat. Rec. 153, 389. STELL, W. K. (1966) The Structure of Horizontal Cells and Synaptic Relations in the Outer Plexiform Layer of the Goldfish Retina as Revealed by the Golgi Method and Electron Microscopy, Thesis, University of Chicago.