Exp Brain Res (1991) 83:643 655
Experimental BrainResearch 9 Springer-Verlag1991
Mammalian neurons in dissociated cultures form clusters in the presence of retinal pigment epithelium J.F. MacDonald, L. Brandes, M. Deverill, I. Mody, M.W. Salter, and E. Theriault Playfair Neuroscience Unit, Departments of Physiologyand Surgery, University of Toronto, and Toronto Western Hospital, 399 Bathurst Street, Toronto, Ont M5T 2S8, Canada Received June 21, 1990 / Accepted August 31, 1990
Summary. The objective of this study was to investigate
Key words: Retina - Cytoarchitecture
the cellular processes involved in the formation of the cytoarchitectonics of the retina. Neurons derived from the retina, spinal cord, cerebral cortex and hippocampus were grown in dissociated monolayer tissue culture using standard techniques. The cultures of retina were unique in that the neurons actively formed into cell clusters. On the other hand, cultures of neurons from the other regions of the CNS grew without forming any obvious histotypical pattern. Cell clusters consisted of an apparent monolayer of neurons above a population of flat cells and clusters were observed in retinal cultures derived from all species studied (mouse, cat and guinea pig). Each cluster was surrounded by whorls of fibroblasts; astrocytes (GFAP-positive cells) were often closely associated with clusters. Formation of clusters appeared to depend strongly upon the presence of cells derived from the retinal pigment epithelium (RPE) because the ability of retinal cells to form clusters was markedly impaired when the RPE was omitted from the cultures. Interestingly, monolayer cultures of neurons from other regions of the CNS could be induced to form clusters, but only when cells of the RPE layer were included at the time of plating. In cultures grown without the RPE layer, clusters did not form when media taken from cultures expressing clusters was used, indicating that the formation of clusters was not caused by a mediabourne factor. On the other hand, clusters did form when neurons without RPE were grown on feeder plates in which clusters had previously been expressed and the neurons subsequently killed by prolonged culturing or by treatment with kainic acid. Hence, physical contact between neurons and cells derived from the RPE appears critical for the formation of clusters. Our results suggest that the cellular processes underlying the formation of clusters may reflect those in the development of the retina in vivo. Thus, cluster formation may be a useful model for investigating the initial stages in the development of retinal cytoarchitecture.
Retinal pigment epithelium - Mouse
Offprint requests to." E. Theriault, Playfair Neuroscience Unit,
Room 11-416, Toronto Western Hospital (address see above)
Development Cats Guinea pig
Introduction The organization of neuronal somata into distinct nuclei or laminae is a characteristic feature of many regions of the CNS. Such organization is considered functionally important because neurons in particular nuclei or laminae often have unique physiological properties. The complexities of the intact CNS make it difficult to determine the factors responsible for establishing neuronal cytoarchitecture, although the interactions of developing neurons with neuroepithelial tissue are thought to play a major role (Goffinet 1985). Simplified in vitro tissue culture systems have been used to study the mechanisms responsible for the development of cytoarchitechtonic organization. In this regard, the retina has been particularly useful because unlike cultures derived from other regions of the CNS, re-aggregate cultures of retinal cells demonstrate a remarkable ability to spontaneously reconstruct major cellular laminations (Akagawa and Barnstable 1986; Liu et al. 1988; Vollmer and Layer 1986A, B; Vollmer et al. 1984). A key element in this reconstruction process may be the participation of retinal pigment epithelial (RPE) cells (Vollmer and Layer 1986A, B; Vollmer et al. 1984). Monolayer cultures of CNS tissue provide a model system which is even more simplified than that of reaggregate cultures. However, monolayer cultures have thus far provided limited information of the development of cytoarchitectural organization because primary cultures of neurons from brain and spinal cord do not demonstrate overt signs of cytoarchitectonics. Here we present evidence that monolayer cultures of the mammalian retina differ significantly in their architectonics from dissociated cultures of other regions of the CNS in that retinal neurons actively sort themselves into discrete clusters of neurons. Formation of the clusters appears to be dependent upon physical contact with cells derived from the RPE cell layer.
644
Material and methods
Tissue culture CNS tissue was taken from foetal mice (Swiss white), cats and guinea pigs. All foetal animals were killed by cervical transection. Prior to surgical removal of foetuses pregnant mice were also killed by cervical dislocation. Pregnant cats were anaesthetized using halothane (Somnothane, Hoechst) and euthanized with an overdose of sodium pentobarbital. Guinea pigs were overdosed with ether anaesthesia. Retinae were dissected from foetuses of mice of embryonic day 18 (El8) and near-term guinea pigs (about E40). The eyes were removed and each retina was dissected away from the sclera and other tissue. Retinae were dissociated by trituration with a Pasteur pipette. Tissues were cooled during the dissection and dissociation of cells by placing containers on crushed ice. In a separate series of experiments cat retinae were cultured from approximately E40 foetuses (7.0 to 9.0 cm in length). Eleven pregnant cats were used and dissections were performed with a total of 52 foetuses. Cat retinae were treated with trypsin prior to dissociation. Cells were dissociated and plated in minimum essential media (MEM) supplemented with 10 % foetal bovine and 10 % horse serum and standard culture procedures followed (MacDonald and Wojtowicz 1982). Cultures from mouse and guinea pig were maintained for approximately 3 to 5 weeks while cat retinae were cultured for as long as 4 months. In order to enrich cultures of mouse retina with background cells approximately one millimeter length of tissue behind the retina was separately dissected and dissociated by treatment with trypsin followed by trituration. This tissue included RPE, sclera, mesenchyme and muscular choroid. With the mouse foetal retina we were unable to dissect out the RPE layer itself because it was not pigmented and did not easily separate into a distinct layer during the dissection. The cells dissociated from the non-neuronal tissue were combined with cells from 10 full retinae and plated on collagencoated 35 mm plastic culture dishes. In other dishes, retinae were also cultured without this additional tissue. The concentrations of dissociated cells was kept low ( < 106 cells per ml) to reduce aggregation and clumping of cells during and following dissociation. This also helped to minimize the clumping of developing neurons in culture in later weeks and favored the formation of a monolayer of neurons lying on top of a confluent multilayer of background cells. All cultures were treated with fluorodeoxyuridine (Hoffmann-La Roche) after one week in culture in order to terminate active mitosis of background cells. Cultures were grown for two to three weeks prior to analysis. A total of 68 dissections of timed pregnant mice were performed (approximately 500 foetal retinae) giving the 200 cultured plates used in these experiments. Some experiments employed guinea pig retinae (24 culture plates) because it was possible to separate the RPE from other ocular tissue. Guinea pig retina was cultured with or without retinal pigment epithelium. Cultures were also prepared from hippocampus, spinal cord and cerebral cortex of foetal mice using standard techniques previously described (MacDonald et al. 1987; MacDonald and Wojtowicz 1982). In the vast majority of experiments hippocampal cultures (250 plates) were used for comparison with retinal cultures. To investigate whether a media-bourne factor might be involved in cluster formation,"conditioned" media was withdrawn from the cultured mouse retina plates which had formed cell clusters. This conditioned media was used to support the growth of hippocampal neurons. Conditioned media from hippocampal cultures was fed to retinal cultures and vice versa. For example, media conditioned for 3 days in hippocampal cultures was diluted 2:1 (conditioned to fresh media) and fed for two weeks to retinal cultures from the time of the initial plating. Similarly, conditioned media from retina was fed to hippocampal cultures. In addition, a simple "split-plate" method was developed to permit growth media to be shared chronically between cultures originating from two different CNS regions or retina with or with-
out non-neuronal retina. This method involved scoring a line across the middle of the bottom of a plastic culture dish which effectively isolated the halves of the dish. A drop of dissociated retinal cells was then placed on one side of the dish and hippocampal cells on the other. Surface tension prevented the drops from mixing. Once the cells had attached, the dish was filled with media and cultured as previously described. Although both sides of the dish became confluent with background cells it was clear that cultured cells did not appreciably migrate across the score mark.
Quantification of spatial distributions of neurons Specific groups of cultures were subjected to tests of the randomness of the spatial arrangement of neuronal somata. Photographs were taken of the cultures under a phase contrast microscope. A standard square grid (85 gm x 85 gm, calibrated with respect to the magnification) was then placed over each photograph and the number of neurons in each square recorded. Initially, the randomness of the spatial arrangement of neurons was determined by fitting the observed data to a Poisson distribution possessing the same mean. Deviations from the predicted Poisson distribution would indicate some degree of spatial orientation of the cell bodies with each other. To further quantify the spatial arrangement cell density was measured; crowding and clustering indices were then calculated according to previously described methods (Pielou 1977). The crowding index was taken as the mean number per neuron of other neurons in a given square. The clustering index, or "patchiness" index of Lloyd (cf Pielou 1977), served to quantify the degree of spatial patterning of the neuronal cell bodies and was calculated as the ratio of crowding index to mean density. The clustering index is a measure which is a property of the spatial pattern itself independent of density. Theoretically, this measurement should also be independent of random losses of the population due to die off, which is particularly appropriate for the cultured neurons examined in the present study. Note that a value of 1 for the clustering index indicates that no clustering is present, whereas values greater than 1 indicate increasing degrees of clustering.
Immunocytochemistry In some experiments murine retinal neurons were cultured on plastic coverslips that were collagen-coated, sterilized and placed in plastic tissue culture dishes. Cells were grown for 10 to 14 days as described above, then removed from the incubator, washed in serum-free M E M and fixed at room temperature for 30 to 40 rain in a fixative solution consisting of either 2% or 4% paraformaldehyde in 0.1 M phosphate-buffered saline. Cells were then washed for 30 min in buffer, incubated in a 10% blocking serum and incubated for a further 60 min at room temperature in primary antiserum. Commercially available antisera to neuron-specific enolase (NSE; Dakopatts, Inc.) and glial fibrillary acidic protein (GFAP; Dakopatts, Inc.) were used in a 1:50 or 1:100 dilution.
Fig. 1. A Dissociated monolayer culture of mouse hippocampus (phase-contrast microscopy). Note that the cells are dispersed over the surface of the culture plate and show no evidence of cluster formation. Individual neurons, however, do show clear evidence of morphological differentiation after 10 days in culture, as evidenced by extensive process growth and by synaptic contacts (as determined electrophysiologically). There is no apparent organization either of the neurons nor of the surrounding support cells. Compare this with a similar field of mouse retinal neurons B. The figure shows three separate clusters of small phase-bright retinal neurons. Each cluster is separated and surrounded by whorls of phase-dark background cells. Occasionally individual neurons can be found outside the clusters. Calibration bars for A, B, 100 pm
645
Fig. 1
646 Table I. Comparison of the frequency of clustered and non-clustered dishes under various culturing conditions. All results from mouse cultures except for those indicated otherwise. The (*) indicates the presence of non-neuronal retinal tissue in the dissociation Culture conditions
No. of dissections
No. of plates (clustered/unclustered)
Retina* Hippocampus Hippocampus with conditioned retinal medium* Retina* with conditioned hippocampat medium Retina* split-plate with hippocampus Hippocampus split-plate with retina* Retina* on feeder plates from hippocampus Retina* on feeder plates from cerebral cortex Retina* on feeder plates from spinal cord Guinea pig retina with RPE Guinea pig retina no RPE
20 25 2
141/0 0/200 0/4
1
2/0
5
13/0
Cultures were then stained for immunoreactivity using the avidinbiotin system (Vectastain Kits, Vector Labs) with diaminobenzidine as the chromagen. Following a final wash in buffer, the coverslips were inverted onto glass slides and photographed in brightfield. Controls included omission of the primary antiserum and replacement with blocking serum.
Results Cultures, regardless of the tissue of origin, became confluent with background cells consisting of fibroblasts, glia and other, unidentified, cells after about one week in culture. Provided the initial plating density was kept low, neurons formed a monolayer upon this diverse non-neuronal background. Unlike the background cells, the monolayer of neurons did not become confluent. Instead, neurons grown from hippocampus, spinal cord and cerebral cortex were distributed over the surface layer of cells with no obvious relationship of one neuronal cell body to another (i.e. Fig. 1A). Some subtle architectonic relationships m a y have developed in such cultures but these were not immediately obvious to us as a pattern.
5
0/13
12
0/21
1
0/2
2
0/3
3 3
12/0 0/12
pigs (Fig. 2B) each cluster was oval in shape and separated from other clusters by regions of background cells either containing no neurons, or sparsely populated by small numbers of individual neurons. The diameter of clusters varied from about 100 to 1000 ~tm. In mouse cultures the border of each cluster was associated with a circular whorl of flat background cells (Fig. 2A). H o w ever, such whorls were not a distinctive feature of clusters of the cat (not shown) or guinea pig retina (Fig. 2B). The neurons within each cluster formed a monolayer but particularly in the guinea pig, with longer periods in culture, they tended to pack more tightly and occasionally form multiple neuronal layers (Fig. 2B).
Identification of neurons and 9lia in cell clusters Cells considered to be neurons on the basis of the appearance of the s o m a under phase-contrast microscopy were confirmed as neurons using electrophysiological and immunocytochemical techniques. These cells showed voltage-dependent sodium and calcium currents and were excited or inhibited with appropriate changes in conductance by excitatory amino acids or G A B A , re-
Retinal cultures >
When retinae from foetal mice, cats or guinea pigs were cultured using standard techniques a highly distinctive pattern of neuronal architectonics was observed. Following 1 to 2 weeks in culture, the majority of neurons were found in loosely-packed monolayer groups which we have called "clusters" (Fig. 1B). Clusters were never observed in hippocampal, spinal cord or cerebral cortical cultures from the mouse (i.e. Fig. 1A; see also Table 1). Retinal cultures could be identified with 100% accuracy from other CNS cultures under phase contrast microscopy by this distinct pattern of clustering, In retinal cultures grown from foetal mice (Fig. 2A) and guinea
Fig. 2. A Dissociated monolayer of mouse retina prepared with non-neuronal retinal tissues and photographed with phase-contrast microscopy. One cluster is shown clearly surrounded by regions of background cells largely devoid of other neurons. Note the small circular or spherical phase-bright neurons within the cluster, some of which are out of focus due to the multilayer composition of the cluster. B Phase-contrast photomicrograph of dissociated monolayers of guinea pig retina which were cultured with the identified retinal pigment epithelial cell layer. A typical cluster of retinal neurons is presented, with neuronal cell bodies appearing phasebright. The cluster appears to be surrounded by a clearly defined phase-bright membrane. The neurons within the cluster appear tightly packed and seem to form multiple layers. Calibration bars for A, B, 100 lam
647
Fig. 2
648
Fig. 3. A This micrograph has been taken using brightfield photomicroscopy. The mouse retinal culture has been fixed with aldehydes and stained for NSE. Note that the spherical or circularshaped neurons are positively stained and they are located primarily within the clusters (small white arrows indicate several neuronal somata within a cluster). The flat background cells underlying the clusters do not show specific staining for NSE in these preparations
(compare with Fig. 5B). A single neuron with bipolar processes is indicated in the top left corner (arrowhead). B In contrast to the neurons, immunoreactivity for GFAP is found in a population of large thin background cells. These cells are often found associated with clusters (as shown in this photomicrograph) but in other cases (not shown) clusters of neurons appear to be completely independent of these cells. Calibration bars for A, B, 100 gm
spectively (Salter et al., in preparation). In addition, clusters of neuronal somata were particularly obvious when cultures were stained with neuron specific enolase (Fig. 3A). The majority of cells within each retinal cluster stained positive for N S E as we would have predicted from the general m o r p h o l o g y and the electrophysiological properties of the cells. Considerable morphological differentiation of the neurons occurred over several weeks in culture including extensive process growth and apparent appositions between m a n y of these neurons. However, no attempt was made to identify the different neuronal subtypes present within the clusters. M a n y synaptic connections were made by these neurons in culture as judged by electrophysiological recordings (Salter et al., in preparation). Other cells in or near the clusters, including a population of large and very flat cells, stained positive for G F A P suggesting that they were astrocytes. The G F A P positive cells were usually associated with clusters (Fig. 3B). However, some clusters appeared to contain no GFAP-positive cells suggesting that astrocytes were not essential for cluster formation (not shown). The neurons in each cluster were located upon a population of flat cells similar to those described by A k a g a w a and Barnstable (1986) in cultured rat monolayer cultures.
background cell layer up until the first week in culture. At this time the neurons gradually became more differentiated with monopolar, bipolar and multipolar processes, m a n y of which displayed varicosities. Some had thin non-beaded processes reminiscent of axons. During the second to third week in culture the neurons or the background cells themselves migrated to f o r m clusters. In some cases, cells migrated distances in excess of a millimeter and could be observed day by day until a cluster was formed. Migration ceased after 3 to 4 weeks in culture. Each cluster of neurons was underlaid by a thin layer of flat cells m a n y of which appeared to contain pigment granules and thus were tentatively identified as R P E cells. The cat retinal clusters were much more irregular in shape than guinea pig or mouse clusters. Experiments performed using cat retinae proved to be impractical for subsequent experimental manipulations because of the difficulty and the excessive expense o f attaining timed-pregnant cats. A prolonged migration of cells into clusters was not observed in cultures of retina from mouse; instead the clusters seemed to develop rapidly during the first several days in culture.
Quantification of clustering Clusters are formed by cellular migration Cultures of cat retina initially had small spherical neurons which were randomly distributed over the confluent
The presence or absence of clustering in the cultures was sufficiently striking that in most cases it was possible to classity each culture dish as either clustered or unclustered; the results of this classification are presented in
649 Table 2. Quantification of the distribution of neurons under various culture conditions. The (*) indicates the presence of non-neuronal retinal tisue. DIS gives the number of the dissection and (n) the number of plates. The crowding index is dependent upon the density of the cultures but the clustering index should be independent of the density of cells (from Pielou, 1977) DIS
n
Density (mean 4- S.D.)
Crowding (mean + S.D.)
A
Hippocampal Side 0.3[S~30bse~ed --
Clustering (mean 4- S.D.)
Poisson
0.2O
Hippocampus 1
3
2 3
15 16
1.68• 3.514-1.36 1.33•
2.824- 0.77 5.67• 1 . 9 7 2.334- 0 . 8 1
1.664-0.32 1.64• 1.76:50.42
.s O EL
0.1
Retina*
0.0
1
4
2 3
7 9
7.88• 7.06• 1.97 7.584- 1.84
25.144- 9.06 26.44• 7.59 28.994- 11.03
3 . 4 0 i 1.28 3.80• 3.944-0.54
35.204- 8.40 16.784- 3.93
3.75• 3.444-0.99
o
5
~o
15
2o
25
30+
Number of c e l l s / s q u a r e
Retinal* side of split plate 1
6
2
9
10.184-3.90 4.94•
B
Hippocampal side of split plate 1
6
2
9
2.07-k 0.58 2.024-0.65
Retinal Side 0 . 6 84
2.66• 0.70 2.894- 1.32
1.30• 1.37•
Observed
0.5 -
Retina* on hippocampal feeder plate 1 2 3
3 17 9
4.74• 2.83• 13.04•
5.38• 1.15 4.124- 1.78 13.80• 1.52
1.13i0.07 1.46• 1.07•
Hippocampus* 1
10
3.70• 1.06
11.29+ 2.65
-
Poisson
0.4
3.124-0.48
/3 0
_Q 0 EL
0.3 0.2 0.1 0.0 0
Table 1. A g r o u p o f experiments, whose results were not included in Table 1, were analyzed in a m o r e objective, statistical m a n n e r as described in the M e t h o d s ; these results are presented in Table 2. Figure 4 A illustrates that h i p p o c a m p a l n e u r o n s demonstrated a spatial distribution n o t greatly different f r o m that predicted by chance. In contrast, retinal cultures clearly deviated dramatically f r o m a r a n d o m distribution (Fig. 4B). Calculation o f the clustering index (Table 2) also confirms that h i p p o c a m p a l cultures show only a slight tendency to f o r m groupings, whereas retinal cultures are strongly clustered. It can be seen that the m e a n density for the h i p p o c a m p a l cultures was m u c h lower than for the retinal cultures. H o w e v e r , even with higher plating densities, h i p p o c a m p a l cultures did n o t f o r m clusters.
Cluster formation is not due to a diffusible factor In order to examine the possible role o f m e d i a - b o u r n e substances in the f o r m a t i o n o f clusters we grew cultures in conditioned medium. C o n d i t i o n e d media f r o m clustered retina did n o t induce the f o r m a t i o n o f clustering in h i p p o c a m p a l cultures, neither did conditioned m e d i a f r o m h i p p o c a m p a l cultures suppress the f o r m a t i o n o f retinal clusters (Table 1). A n additional series o f experiments was p e r f o r m e d
5
I0
15
20
25
30+
Number of cells/square Fig. 4. A Data from a culture of hippocampus (grown on a split-
plate beside clustered retina) have been analyzed as described in the Methods. The number of cells in each grid square has been plotted against the normalized probability of finding a cell in any particular square. The bar graph gives the actual data and the line the distribution predicted for the actual mean if it were to follow a Poisson Distribution. The data did not differ much from the predicted random distribution. B Data from the retinal side of the plate were analyzed and plotted as above. The Poisson Distribution for the experimental mean was used to generate the curve and the data appear as a bar graph. In contrast to the hippocampal side, the retinal side of the culture has neurons widely differing in their distribution from that predicted by chance
using the "split-plate" technique where h i p p o c a m p a l and retinal cultures shared the same culture dish and media. The presence o f shared m e d i u m had no effect on retinal clusters n o r on the lack o f clustering in h i p p o c a m p a l cultures (Table 2). Therefore, simple sharing o f the media apparently was not responsible for the differences between these two types o f cultures.
Putative role of R P E in cluster formation F r o m our previous experience with cultures o f other C N S regions we anticipated that the retinal cultures f r o m
650
Fig. 5
651 the mouse would be more viable if non-neuronal tissue was included to bolster the background layers of cells. As described in the Methods, a non-specified portion of additional tissue from behind the retina, consisting of the RPE cell layer as well as other tissue, was included in the culture. Cultures prepared in this manner demonstrated a high degree of clustering as previously described (see Tables 1, 2). When the neural retina was dissected as cleanly as possible, and cultured without the inclusion of this additional tissue, cluster formation was absent or dramatically reduced. When analyzed using our statistical approach, mouse retinal cultures did demonstrate some degree of clustering although it was very much less than for cultures containing the extra tissue. In spite of this statistical indication that some degree of clustering was occurring, it was qualitatively clear that the clustering was strikingly less in the absence of this additional tissue (Fig. 5A). Furthermore, when the split-plate technique was used, cluster formation occurred only on the side containing the non-neuronal component (not shown). One possible explanation for our results was that the RPE layer contained a population of cells whose presence was required for the induction of clustering. Therefore, we used foetal guinea pigs which possess an easily identifiable cell layer containing the RPE which we could readily isolate from both the surrounding tissue and the neural retina itself. When guinea pig retinae were cultured with or without the RPE layer, the results were unequivocal. In the absence of the RPE layer we saw virtually no clustering (Fig. 5B) while the inclusion of RPE induced the most striking clustering we have observed in any cultures (Fig. 2B). The guinea pig clusters contained so many neurons and the background so few that it was unnecessary to subject them to statistical analysis. In some cases, the neurons within clusters gradually stacked up to form multiple layers of neurons (see Fig. 2B). This process began about one week from initial plating and usually commenced simultaneously at the two ends of the cluster. After two to three weeks the entire cluster was composed of multiple layers of neurons even though similar cultures grown without RPE con-
tinued as monolayers with randomly distributed somata. The high density of neurons in these clustered cultures obscured the background cells underlying each cluster.
Cell-cell contact with background cells induces cluster formation In order to investigate the role of background cells in the establishment of clusters we used feeder plates derived from the mouse spinal cord, hippocampus, cerebral cortex or the retina as a substrate for culturing dissociated retinal neurons. These feeder plates contained background cells but were devoid of neurons, either because of their age or because of treatment with kainic acid. Cultures in excess of one or two months often have complete loss of neurons while the background support cells remain. Bath application of kainic acid is able to destroy neurons (Rothman 1985) and ensures the elimination of most remaining neurons. Feeder plates ranged from 1 week to 2 months in age but the reported observations were similar regardless of the age of the feeder plates. Clusters devoid of neurons could easily be recognized by the swirling pattern or thickened halo of background cells which had previously surrounded each cluster. When dissociated mouse retinal neurons (with or without the non-neuronal retinal tissue) were plated on such feeder cultures from CNS regions other than the retina itself, clusters were not formed and instead neurons were distributed in an almost random fashion across the plate (Fig. 6 and Table 2 for hippocampal feeder plates as an example). In contrast, when dissociated retinal cells were plated on feeder cultures derived from the retina, clusters were again formed as the newly seeded neurons preferentially re-populated pre-existing clusters (not shown). Furthermore, when hippocampal neurons were plated on cultured retinal plates, the hippocampal neurons also preferentially re-populated the cluster regions rather than the regions between the clusters.
Inclusion of nonmeural retinal tissue induces clusterin 9 in hippocampal cultures %-
Fig. 5. A Phase-contrast photomicrograph of mouse retinal culture
grown with as little contamination as possible from non-neuronal retinal tissue. Neurons did not demonstrate an overt formation of clusters although occasionally a number of small clumps can be seen. Note that these groups of cells are not tightly packed and do not show any overt architectonic organization. In addition there are no associated whorls of background cells with the groupings of phase-bright neurons. In other cultures (not shown) there was some suggestion of a whorling pattern of background cells although the neurons did not appear to be associated with these cells. B Cultures of foetal guinea pig retina grown with little or no contamination of RPE cells. These cultures displayed no evidence of cluster formation and this photomicrograph shows the only field which even resembled a cluster. Compare this distribution of cells with that shown in Fig. 3 where the RPE cell layer has been included with the cultured neural retina, and has resulted in the formation of distinct clusters. Calibration bars for A, B, 100 gm
The previous observation suggested to us that neurons regardless of their origin either preferentially adhered to the background cells in the cluster or that those between the cluster presented an unfavorable substrate for neuronal attachment, growth or survival. If this were the case we reasoned that if we included neurons from regions of the CNS that did not form clusters with some of the non-specific tissue or the RPE layer at the time of dissociation, then during the sorting out of cells these neurons should also be induced form clusters. This possibility was bourne out by our observation that when the non-neuronal retinal tissue was included with dissociated cells from the hippocampus the resulting monolayer cultures demonstrated a high degree of cluster formation (Table 2 and Fig. 7).
652
Fig. 6. Mouse retinal cells prepared for culture with non-neuronal retinal tissue and plated onto hippocampal feeder plates. The distribution of the neurons in this culture was approximately random.
Note that despite a relatively dense plating of cells, no clusters are discernable and no whorls of background cells can be seen. Calibration bar, 100 gm
Discussion
represent an organized spatial pattern of cell placement one with respect to the other. In previous studies it has been found that if retinal cells are prevented from attaching to the substratum of the culture dish there is a strong tendency for them to form re-aggregates which sort themselves out and form the major laminations of the retina (Akagawa et al. 1987; Vollmer and Layer 1986A, B; Vollmer et al. 1984). This re-aggregation can be of a rudimentary form ranging from the initial adhesion of cells to each other, to a more complex rosette formation, to the beginnings of cell laminations (Fujisawa 1973; Sheffield and Moscona 1969, 1970). Re-aggregate cultures may even demonstrate complete reconstruction of the major cell layers of the retina, if RPE cells are included with the dissociated cells (Vollmer and Layer 1986A, B; Vollmer et al. 1984). We suggest here that cluster formation, which may also depend upon the presence of RPE, represents the maximal organizational capacities of the cells when they are restricted to the two dimensional tissue culture plate as opposed to the three dimensional growth domain of re-aggregates. Therefore, we feel that cluster formation can serve as a developmental model of the early stages of cell sorting and migration. This model presents several advantages over re-aggregates. For example, the cellular organiza-
We have shown that monolayer cultures of mouse, cat and guinea pig retina differ from those of other laminated CNS regions. Retinal neurons in such cultures demonstrate a rudimentary architectonics unlike the apparently stochastic placement of neuronal somata characteristic of the other kinds of CNS monolayer cultures studied. The architectonics shown by retinal cultures takes the form of cell clusters. Each cluster consists of a monolayer grouping of neurons in some cases surrounded by whorls of background cells but in others consists simply of neurons, placed on top of unidentified flat cells. One possibility is that the flat cells are of glial origin (Li and Sheffield 1984, 1986A). Even though we often observed GFAP-positive astrocytes intimately associated with some mouse retinal clusters it was also clear that many clusters did not contain such immunoreactive cells. Therefore, it appears unlikely that this subpopulation of glial cells plays a primary role in the formation of clusters. It is important to note that the clusters are not simply clumps of neurons which have adhered to each other during the dissociation, settled to the culture dish, and subsequently developed in place. Rather, the clusters
653
Fig. 7. Non-neuronal tissue was included with dissociated hippocampal cells at the time of dissociation. In this phase-contrast photomicrograph it is clear that unlike conventional hippocampal
cultures the neurons in this cuIture environment were distinctly clustered. Calibration bar, 100 gm
tion in the clusters may be simpler, and because they are restricted to two dimensions histological and immunohistochemical techniques can easily be applied. Furthermore, the neurons within the cluster are in a monolayer and are readily accessible for electrophysiological and pharmacological analysis. The initial cell adhesions and the formation of cellular junctions in re-aggregate cultures are thought to be important for the subsequent laminar reconstruction (Sheffield 1970). Specific cell ligands on the extracellular surface may also contribute to this process (Barnstable 1987A, B; Moscona 1974). Even though retinal cell clusters do not obviously resemble re-aggregates they also may have arisen as a consequence of some preferential cell adhesion following plating but just prior to attachment to the substratum. Thus, retinal neurons might recognize each other, adhere and bias their eventual location on the culture plate. Similarly, non-neuronal cells might also be subject to a preference for "like" cells, according to the hypothesis of "cell affinities" (Steinberg 1963) where "alike" cells and tissue have a marked preference to reorganize and form contacts with each other. Recent experiments suggest an expansion of this hypothesis to include the view that each type of cell
expresses on its extracellular surface unique cell-typespecific factors qualitatively different from those expressed on other kinds of cells. Various glycosamines and other types of adhesion molecules might serve this purpose (see Getch and Steinberg 1986; Mendez-Otero et al. 1988; Barnstable 1987A, B). Presumably neurons from regions of the CNS other than the retina would also be subject to the same principle of "cell affinities". However, cluster formation was never observed ifi CNS cultures from a wide variety of locations unless retinal tissue containing the RPE was included at the time of dissociation. Background cells from the retina but from no other region of the CNS examined could support the formation of clusters. This new finding may indicate that the cultures derived from these other CNS tissues lack the types of neuroepithelial cells and adhesion factors (or their expression) required for lamination in vivo. Our observations with cat retina have clearly demonstrated that the formation of clusters is more complex than just cell adhesion and must also consist of a distinct phase of cellular migration. We have not yet attempted to determine if this migratory phase occurs as the result of active migration of neurons or alternatively, may be due to the movements of backgrounds cells to which they are attached. We did not observe a similar phase in
654 cultures of the mouse retina presumably because clusters were established much more rapidly in this species than in the cat. Strong influences of background cells upon retinal neuronal growth and migration have been noted by Li and Sheffield (1986A, B). Determination of the precise role of migration in cluster formation will require further study. One alternative explanation for the formation of retinal clusters would be a selective die off of "non-clustered" neurons or conversely the selective induction of neuronal survival by neurotrophic growth factors. For example, nerve growth factor (Adler 1986) and brainderived neurotrophic factor (Hofer and Barde 1988) both increase the survivability of peripheral and central neurons respectively and promote outgrowth of processes. It is possible that cells within the clusters or a population of background cells expresses or secretes a factor responsible for stimulating neuronal proliferation. However, our initial attempts at reproducing cell clustering by using conditioned media or by a constant sharing of same media in the split-plate cultures gave essentially negative results indicating that a freely diffusible factor which can act at long distances was not responsible for clustering. It may be necessary for retinal neurons to be exposed to concentration gradients of inductive factors released by support cells located within a few microns of each other. Therefore, it is possible that clustering failed to occur in the split-plate cultures because cells would simply not be close enough to each other to permit this kind of interaction. Further experimentation would be required to determine if this were the case. However, we believe that diffusible neurotrophic factors are unlikely to play a major part in cluster formation for several reasons. The clear migration of cells in feline retinal cultures argues strongly against a simple selective die off. Furthermore, retinal clusters in mouse and guinea pig appear to be associated with specific groupings of background cells and depended upon the unique presence of RPE cells. Our results are consistent with the hypothesis that retinal cell cluster formation is dependent upon direct cell-cell contact factors associated with non-neuronal cells specifically from the retina. Retinal neurons cultured on top of feeder plates of hippocampal background cells failed to form clusters. On the other hand, there was a preferential repopulation of clusters in feeder plates of background cells from retina which had previously formed clusters. In addition, we were able to demonstrate that cluster formation was strongly dependent upon the inclusion of non-neuronal retinal cells in the original dissociation. The experiments using guinea pig retina further indicated that the RPE cell layer was sufficient to induce the formation of clusters. Similarly, Vollmer and Layer (1986B) report that the formation of retinal laminations in re-aggregates depends strongly upon the presence of RPE cells. Furthermore, neither extracts nor conditioned medium from RPE cultures could induce laminations in the re-aggregates and direct physical contact between the retinal neurons and RPE cells was necessary. Another possibility for the induction of clustering formation may be the inclusion of mesenchymal cells in
the non-neuronal tissue. While the inductive properties of embryonic mesenchyme have been well described in the process of limb formation (Balinsky 1981), with regards to the retina, there is strong evidence for the role of the RPE in the induction of retinal cytoarchitectonics. Based on this literature and the present evidence that inclusion the RPE cell layer is sufficient to cause clustering, we feel that the RPE is the most likely candidate for the induction of cluster formation. The extracellular surfaces of neurons and other cells of the developing retina express a wide variety of molecules many of which are important for cell attachment and adhesion not only to the culture substrate but also of cells t o each other. Furthermore, the expression of such specific adhesion molecules is often topographically oriented within the retina in vivo suggesting that the architectonics of cell-cell interactions are determined, at least in part, by such molecules (for reviews see Getch and Steinberg 1986 and Barnstable 1987A, B). In addition, a number of important components of the extracellular matrix, such as laminin and fibronectin (Heaton and Swanson 1988; Pixley 1987) can serve to attach specific cell types and/or guide the outgrowth of neuronal processes. We are presently examining whether or not some of these cell-contact factors play an important role in the formation of retinal cell clusters.
Acknowledgement. This work has been supported by the Medical Research Council of Canada.
References Adler R (1986) Trophic interactions in retinal development and in retinal degeneration: in vivo and in vitro studies. In: Adler R, Farber D (eds) The retina: a model for cell biology studies, Part I. Academic Press, New York, pp 112-151 Akagawa K, Barnstable CJ (1986) Identification and characterization of cell types in monolayer cultures of rat retina using monoclonal antibodies. Brain Res 383 : 110-120 Akagawa K, Hicks D, Barnstable CJ (1987) Histiotypic organization and cell differentiation in rat retinal reaggregate cultures. Brain Res 437 : 298-308 Balinsky BI (1981) An introduction to embryology. Saunders Publishing Co, Philadelphia Barnstable CJ (1987A) Immunological studies of the diversity and development of the mammalian visual system. Immunol Rev 100:47-78 Barnstable CJ (1987B) A Molecular view of vertebrate retinal development. Mol Neurobiol 1: 9-46 Fujisawa H (1973) The process of reconstruction of histological architecture from dissociated retinal cells. Wilhelm Roux' Arch 171:312-330 Getch CC, Steinberg MS (1986) Differential adhesion in neuronal development. In: Adler R, Farber D (eds) The retina: a model for cell biology studies, Part I. Academic Press, New York, pp 67 109 Goffinet AM (1985) Events governing organization of postmigratory neurons: studies on brain development in normal and reeler mice. Brain Res Rev 319:261-296 Heaton MB, Swanson DJ (1988) The influence of laminin on the initial differentiation of cultured neural tube neurons. J Neurosci Res 19:212-218 Hofer MM, Barde YA (1988) Brain-derived neurotrophic factor prevents neuronal death in vivo. Nature 6153:261-262
655 Li H-P, Sheffield JB (1984) Isolation of flat cells, a subpopulation of the embryonic chick retina. Tissue Cell 16 : 843 Li H-P Sheffield JB (1986A) Retinal flat cells are a substrate that facilitates retinal neuron growth and fibre formation. Invest Ophthalmol Vis Sci 27:296-306 Li H-P Sheffield JB (1986B) Retinal flat cells participate in the formation of fibers by retinal neuroblasts in vitro. Invest Ophthalmol Vis Sci 27:307-315 Liu L, Cheng SH, Jiang LZ, Hansmann G, Layer PG (1988) The pigmented epithelium sustains growth and tissue differentiation of chicken retinal explants in vitro. Exp Eye Res 46:801-812 MacDonald JF, Miljkovic Z, Pennefather P (1987) Use-dependent blockade of excitatory amino acid currents by ketamine. J Neurophysiol 58 : 251-266 MacDonald JF, Wojtowicz JM (1982) The effects of L-glutamate and its analogues upon the membrane conductance of central murine neurons in culture. Can J Physiol Pharmacol 60 : 282-296 Mendez-Otero R, Schlosshauer B, Barnstable CJ ConstantinePaton (1988) A developmentally regulated antigen associated with neural cell and process migration. J Neurosci 8:564-579 Moscona AA (1974) Surface specification of embryonic cells: lectin receptors, cell recognition and specific cell ligands. In: Moscona AA (ed) The cell surface in development. John Wiley & Sons, New York, pp 67-99 Pixley SKR, Nieto-Sampedro M, Cotman CW (1987) Preferential
adhesion of brain astrocytes to laminin and central neurites to astrocytes. J Neurosci Res 18:402-406 Pielou EC (1977) Mathematical ecology. John Wiley & Sons, New York Rothman SM (1985) The neurotoxicity of excitatory amino acids is produced by passive chloride influx. J Neurosci 5:1483-1489 Sheffield JB (1970) Studies on aggregation of embryonic cells: initial cell adhesions and the formation of intercellular junctions. J Morphol 132 : 245-264 Sheffield JB, Moscona AA (1969) Early stages in the reaggregation of embryonic chick neural retinal cells. Exp Cell Res 57: 463-466 Sheffield JB, Moscona AA (1970) Electron microscopic analysis of aggregation of embryonic cells: the structure and differentiation of aggregates of neural retina cells. Dev Biol 23:36-61 Steinberg MS (1963) Reconstruction of tissues by dissociated cells. Science 141:401-408 Vollmer G, Layer PG (1986A) An in vitro model of proliferation and differentiation of the chick retina: coaggregates of retinal and pigment epithelial cells. J Neurosci 6:1885-1896 Vollmer G, Layer PG (1986B) Reaggregation of chick retinal and mixtures of retinal and pigment epithelial cells: the degree of laminar organization is dependent on age. Neurosci Lett 63 : 91-95 Vollmer G, Layer PG, Gierer A (1984) Reaggregation of embryonic chick retinal cells: pigment epithelial cells induce a high order of stratification. Neurosci Lett 48:191-196
Experimental BrainResearch 9 Springer-Verlag1991
Mammalian neurons in dissociated cultures form clusters in the presence of retinal pigment epithelium J.F. MacDonald, L. Brandes, M. Deverill, I. Mody, M.W. Salter, and E. Theriault Playfair Neuroscience Unit, Departments of Physiologyand Surgery, University of Toronto, and Toronto Western Hospital, 399 Bathurst Street, Toronto, Ont M5T 2S8, Canada Received June 21, 1990 / Accepted August 31, 1990
Summary. The objective of this study was to investigate
Key words: Retina - Cytoarchitecture
the cellular processes involved in the formation of the cytoarchitectonics of the retina. Neurons derived from the retina, spinal cord, cerebral cortex and hippocampus were grown in dissociated monolayer tissue culture using standard techniques. The cultures of retina were unique in that the neurons actively formed into cell clusters. On the other hand, cultures of neurons from the other regions of the CNS grew without forming any obvious histotypical pattern. Cell clusters consisted of an apparent monolayer of neurons above a population of flat cells and clusters were observed in retinal cultures derived from all species studied (mouse, cat and guinea pig). Each cluster was surrounded by whorls of fibroblasts; astrocytes (GFAP-positive cells) were often closely associated with clusters. Formation of clusters appeared to depend strongly upon the presence of cells derived from the retinal pigment epithelium (RPE) because the ability of retinal cells to form clusters was markedly impaired when the RPE was omitted from the cultures. Interestingly, monolayer cultures of neurons from other regions of the CNS could be induced to form clusters, but only when cells of the RPE layer were included at the time of plating. In cultures grown without the RPE layer, clusters did not form when media taken from cultures expressing clusters was used, indicating that the formation of clusters was not caused by a mediabourne factor. On the other hand, clusters did form when neurons without RPE were grown on feeder plates in which clusters had previously been expressed and the neurons subsequently killed by prolonged culturing or by treatment with kainic acid. Hence, physical contact between neurons and cells derived from the RPE appears critical for the formation of clusters. Our results suggest that the cellular processes underlying the formation of clusters may reflect those in the development of the retina in vivo. Thus, cluster formation may be a useful model for investigating the initial stages in the development of retinal cytoarchitecture.
Retinal pigment epithelium - Mouse
Offprint requests to." E. Theriault, Playfair Neuroscience Unit,
Room 11-416, Toronto Western Hospital (address see above)
Development Cats Guinea pig
Introduction The organization of neuronal somata into distinct nuclei or laminae is a characteristic feature of many regions of the CNS. Such organization is considered functionally important because neurons in particular nuclei or laminae often have unique physiological properties. The complexities of the intact CNS make it difficult to determine the factors responsible for establishing neuronal cytoarchitecture, although the interactions of developing neurons with neuroepithelial tissue are thought to play a major role (Goffinet 1985). Simplified in vitro tissue culture systems have been used to study the mechanisms responsible for the development of cytoarchitechtonic organization. In this regard, the retina has been particularly useful because unlike cultures derived from other regions of the CNS, re-aggregate cultures of retinal cells demonstrate a remarkable ability to spontaneously reconstruct major cellular laminations (Akagawa and Barnstable 1986; Liu et al. 1988; Vollmer and Layer 1986A, B; Vollmer et al. 1984). A key element in this reconstruction process may be the participation of retinal pigment epithelial (RPE) cells (Vollmer and Layer 1986A, B; Vollmer et al. 1984). Monolayer cultures of CNS tissue provide a model system which is even more simplified than that of reaggregate cultures. However, monolayer cultures have thus far provided limited information of the development of cytoarchitectural organization because primary cultures of neurons from brain and spinal cord do not demonstrate overt signs of cytoarchitectonics. Here we present evidence that monolayer cultures of the mammalian retina differ significantly in their architectonics from dissociated cultures of other regions of the CNS in that retinal neurons actively sort themselves into discrete clusters of neurons. Formation of the clusters appears to be dependent upon physical contact with cells derived from the RPE cell layer.
644
Material and methods
Tissue culture CNS tissue was taken from foetal mice (Swiss white), cats and guinea pigs. All foetal animals were killed by cervical transection. Prior to surgical removal of foetuses pregnant mice were also killed by cervical dislocation. Pregnant cats were anaesthetized using halothane (Somnothane, Hoechst) and euthanized with an overdose of sodium pentobarbital. Guinea pigs were overdosed with ether anaesthesia. Retinae were dissected from foetuses of mice of embryonic day 18 (El8) and near-term guinea pigs (about E40). The eyes were removed and each retina was dissected away from the sclera and other tissue. Retinae were dissociated by trituration with a Pasteur pipette. Tissues were cooled during the dissection and dissociation of cells by placing containers on crushed ice. In a separate series of experiments cat retinae were cultured from approximately E40 foetuses (7.0 to 9.0 cm in length). Eleven pregnant cats were used and dissections were performed with a total of 52 foetuses. Cat retinae were treated with trypsin prior to dissociation. Cells were dissociated and plated in minimum essential media (MEM) supplemented with 10 % foetal bovine and 10 % horse serum and standard culture procedures followed (MacDonald and Wojtowicz 1982). Cultures from mouse and guinea pig were maintained for approximately 3 to 5 weeks while cat retinae were cultured for as long as 4 months. In order to enrich cultures of mouse retina with background cells approximately one millimeter length of tissue behind the retina was separately dissected and dissociated by treatment with trypsin followed by trituration. This tissue included RPE, sclera, mesenchyme and muscular choroid. With the mouse foetal retina we were unable to dissect out the RPE layer itself because it was not pigmented and did not easily separate into a distinct layer during the dissection. The cells dissociated from the non-neuronal tissue were combined with cells from 10 full retinae and plated on collagencoated 35 mm plastic culture dishes. In other dishes, retinae were also cultured without this additional tissue. The concentrations of dissociated cells was kept low ( < 106 cells per ml) to reduce aggregation and clumping of cells during and following dissociation. This also helped to minimize the clumping of developing neurons in culture in later weeks and favored the formation of a monolayer of neurons lying on top of a confluent multilayer of background cells. All cultures were treated with fluorodeoxyuridine (Hoffmann-La Roche) after one week in culture in order to terminate active mitosis of background cells. Cultures were grown for two to three weeks prior to analysis. A total of 68 dissections of timed pregnant mice were performed (approximately 500 foetal retinae) giving the 200 cultured plates used in these experiments. Some experiments employed guinea pig retinae (24 culture plates) because it was possible to separate the RPE from other ocular tissue. Guinea pig retina was cultured with or without retinal pigment epithelium. Cultures were also prepared from hippocampus, spinal cord and cerebral cortex of foetal mice using standard techniques previously described (MacDonald et al. 1987; MacDonald and Wojtowicz 1982). In the vast majority of experiments hippocampal cultures (250 plates) were used for comparison with retinal cultures. To investigate whether a media-bourne factor might be involved in cluster formation,"conditioned" media was withdrawn from the cultured mouse retina plates which had formed cell clusters. This conditioned media was used to support the growth of hippocampal neurons. Conditioned media from hippocampal cultures was fed to retinal cultures and vice versa. For example, media conditioned for 3 days in hippocampal cultures was diluted 2:1 (conditioned to fresh media) and fed for two weeks to retinal cultures from the time of the initial plating. Similarly, conditioned media from retina was fed to hippocampal cultures. In addition, a simple "split-plate" method was developed to permit growth media to be shared chronically between cultures originating from two different CNS regions or retina with or with-
out non-neuronal retina. This method involved scoring a line across the middle of the bottom of a plastic culture dish which effectively isolated the halves of the dish. A drop of dissociated retinal cells was then placed on one side of the dish and hippocampal cells on the other. Surface tension prevented the drops from mixing. Once the cells had attached, the dish was filled with media and cultured as previously described. Although both sides of the dish became confluent with background cells it was clear that cultured cells did not appreciably migrate across the score mark.
Quantification of spatial distributions of neurons Specific groups of cultures were subjected to tests of the randomness of the spatial arrangement of neuronal somata. Photographs were taken of the cultures under a phase contrast microscope. A standard square grid (85 gm x 85 gm, calibrated with respect to the magnification) was then placed over each photograph and the number of neurons in each square recorded. Initially, the randomness of the spatial arrangement of neurons was determined by fitting the observed data to a Poisson distribution possessing the same mean. Deviations from the predicted Poisson distribution would indicate some degree of spatial orientation of the cell bodies with each other. To further quantify the spatial arrangement cell density was measured; crowding and clustering indices were then calculated according to previously described methods (Pielou 1977). The crowding index was taken as the mean number per neuron of other neurons in a given square. The clustering index, or "patchiness" index of Lloyd (cf Pielou 1977), served to quantify the degree of spatial patterning of the neuronal cell bodies and was calculated as the ratio of crowding index to mean density. The clustering index is a measure which is a property of the spatial pattern itself independent of density. Theoretically, this measurement should also be independent of random losses of the population due to die off, which is particularly appropriate for the cultured neurons examined in the present study. Note that a value of 1 for the clustering index indicates that no clustering is present, whereas values greater than 1 indicate increasing degrees of clustering.
Immunocytochemistry In some experiments murine retinal neurons were cultured on plastic coverslips that were collagen-coated, sterilized and placed in plastic tissue culture dishes. Cells were grown for 10 to 14 days as described above, then removed from the incubator, washed in serum-free M E M and fixed at room temperature for 30 to 40 rain in a fixative solution consisting of either 2% or 4% paraformaldehyde in 0.1 M phosphate-buffered saline. Cells were then washed for 30 min in buffer, incubated in a 10% blocking serum and incubated for a further 60 min at room temperature in primary antiserum. Commercially available antisera to neuron-specific enolase (NSE; Dakopatts, Inc.) and glial fibrillary acidic protein (GFAP; Dakopatts, Inc.) were used in a 1:50 or 1:100 dilution.
Fig. 1. A Dissociated monolayer culture of mouse hippocampus (phase-contrast microscopy). Note that the cells are dispersed over the surface of the culture plate and show no evidence of cluster formation. Individual neurons, however, do show clear evidence of morphological differentiation after 10 days in culture, as evidenced by extensive process growth and by synaptic contacts (as determined electrophysiologically). There is no apparent organization either of the neurons nor of the surrounding support cells. Compare this with a similar field of mouse retinal neurons B. The figure shows three separate clusters of small phase-bright retinal neurons. Each cluster is separated and surrounded by whorls of phase-dark background cells. Occasionally individual neurons can be found outside the clusters. Calibration bars for A, B, 100 pm
645
Fig. 1
646 Table I. Comparison of the frequency of clustered and non-clustered dishes under various culturing conditions. All results from mouse cultures except for those indicated otherwise. The (*) indicates the presence of non-neuronal retinal tissue in the dissociation Culture conditions
No. of dissections
No. of plates (clustered/unclustered)
Retina* Hippocampus Hippocampus with conditioned retinal medium* Retina* with conditioned hippocampat medium Retina* split-plate with hippocampus Hippocampus split-plate with retina* Retina* on feeder plates from hippocampus Retina* on feeder plates from cerebral cortex Retina* on feeder plates from spinal cord Guinea pig retina with RPE Guinea pig retina no RPE
20 25 2
141/0 0/200 0/4
1
2/0
5
13/0
Cultures were then stained for immunoreactivity using the avidinbiotin system (Vectastain Kits, Vector Labs) with diaminobenzidine as the chromagen. Following a final wash in buffer, the coverslips were inverted onto glass slides and photographed in brightfield. Controls included omission of the primary antiserum and replacement with blocking serum.
Results Cultures, regardless of the tissue of origin, became confluent with background cells consisting of fibroblasts, glia and other, unidentified, cells after about one week in culture. Provided the initial plating density was kept low, neurons formed a monolayer upon this diverse non-neuronal background. Unlike the background cells, the monolayer of neurons did not become confluent. Instead, neurons grown from hippocampus, spinal cord and cerebral cortex were distributed over the surface layer of cells with no obvious relationship of one neuronal cell body to another (i.e. Fig. 1A). Some subtle architectonic relationships m a y have developed in such cultures but these were not immediately obvious to us as a pattern.
5
0/13
12
0/21
1
0/2
2
0/3
3 3
12/0 0/12
pigs (Fig. 2B) each cluster was oval in shape and separated from other clusters by regions of background cells either containing no neurons, or sparsely populated by small numbers of individual neurons. The diameter of clusters varied from about 100 to 1000 ~tm. In mouse cultures the border of each cluster was associated with a circular whorl of flat background cells (Fig. 2A). H o w ever, such whorls were not a distinctive feature of clusters of the cat (not shown) or guinea pig retina (Fig. 2B). The neurons within each cluster formed a monolayer but particularly in the guinea pig, with longer periods in culture, they tended to pack more tightly and occasionally form multiple neuronal layers (Fig. 2B).
Identification of neurons and 9lia in cell clusters Cells considered to be neurons on the basis of the appearance of the s o m a under phase-contrast microscopy were confirmed as neurons using electrophysiological and immunocytochemical techniques. These cells showed voltage-dependent sodium and calcium currents and were excited or inhibited with appropriate changes in conductance by excitatory amino acids or G A B A , re-
Retinal cultures >
When retinae from foetal mice, cats or guinea pigs were cultured using standard techniques a highly distinctive pattern of neuronal architectonics was observed. Following 1 to 2 weeks in culture, the majority of neurons were found in loosely-packed monolayer groups which we have called "clusters" (Fig. 1B). Clusters were never observed in hippocampal, spinal cord or cerebral cortical cultures from the mouse (i.e. Fig. 1A; see also Table 1). Retinal cultures could be identified with 100% accuracy from other CNS cultures under phase contrast microscopy by this distinct pattern of clustering, In retinal cultures grown from foetal mice (Fig. 2A) and guinea
Fig. 2. A Dissociated monolayer of mouse retina prepared with non-neuronal retinal tissues and photographed with phase-contrast microscopy. One cluster is shown clearly surrounded by regions of background cells largely devoid of other neurons. Note the small circular or spherical phase-bright neurons within the cluster, some of which are out of focus due to the multilayer composition of the cluster. B Phase-contrast photomicrograph of dissociated monolayers of guinea pig retina which were cultured with the identified retinal pigment epithelial cell layer. A typical cluster of retinal neurons is presented, with neuronal cell bodies appearing phasebright. The cluster appears to be surrounded by a clearly defined phase-bright membrane. The neurons within the cluster appear tightly packed and seem to form multiple layers. Calibration bars for A, B, 100 lam
647
Fig. 2
648
Fig. 3. A This micrograph has been taken using brightfield photomicroscopy. The mouse retinal culture has been fixed with aldehydes and stained for NSE. Note that the spherical or circularshaped neurons are positively stained and they are located primarily within the clusters (small white arrows indicate several neuronal somata within a cluster). The flat background cells underlying the clusters do not show specific staining for NSE in these preparations
(compare with Fig. 5B). A single neuron with bipolar processes is indicated in the top left corner (arrowhead). B In contrast to the neurons, immunoreactivity for GFAP is found in a population of large thin background cells. These cells are often found associated with clusters (as shown in this photomicrograph) but in other cases (not shown) clusters of neurons appear to be completely independent of these cells. Calibration bars for A, B, 100 gm
spectively (Salter et al., in preparation). In addition, clusters of neuronal somata were particularly obvious when cultures were stained with neuron specific enolase (Fig. 3A). The majority of cells within each retinal cluster stained positive for N S E as we would have predicted from the general m o r p h o l o g y and the electrophysiological properties of the cells. Considerable morphological differentiation of the neurons occurred over several weeks in culture including extensive process growth and apparent appositions between m a n y of these neurons. However, no attempt was made to identify the different neuronal subtypes present within the clusters. M a n y synaptic connections were made by these neurons in culture as judged by electrophysiological recordings (Salter et al., in preparation). Other cells in or near the clusters, including a population of large and very flat cells, stained positive for G F A P suggesting that they were astrocytes. The G F A P positive cells were usually associated with clusters (Fig. 3B). However, some clusters appeared to contain no GFAP-positive cells suggesting that astrocytes were not essential for cluster formation (not shown). The neurons in each cluster were located upon a population of flat cells similar to those described by A k a g a w a and Barnstable (1986) in cultured rat monolayer cultures.
background cell layer up until the first week in culture. At this time the neurons gradually became more differentiated with monopolar, bipolar and multipolar processes, m a n y of which displayed varicosities. Some had thin non-beaded processes reminiscent of axons. During the second to third week in culture the neurons or the background cells themselves migrated to f o r m clusters. In some cases, cells migrated distances in excess of a millimeter and could be observed day by day until a cluster was formed. Migration ceased after 3 to 4 weeks in culture. Each cluster of neurons was underlaid by a thin layer of flat cells m a n y of which appeared to contain pigment granules and thus were tentatively identified as R P E cells. The cat retinal clusters were much more irregular in shape than guinea pig or mouse clusters. Experiments performed using cat retinae proved to be impractical for subsequent experimental manipulations because of the difficulty and the excessive expense o f attaining timed-pregnant cats. A prolonged migration of cells into clusters was not observed in cultures of retina from mouse; instead the clusters seemed to develop rapidly during the first several days in culture.
Quantification of clustering Clusters are formed by cellular migration Cultures of cat retina initially had small spherical neurons which were randomly distributed over the confluent
The presence or absence of clustering in the cultures was sufficiently striking that in most cases it was possible to classity each culture dish as either clustered or unclustered; the results of this classification are presented in
649 Table 2. Quantification of the distribution of neurons under various culture conditions. The (*) indicates the presence of non-neuronal retinal tisue. DIS gives the number of the dissection and (n) the number of plates. The crowding index is dependent upon the density of the cultures but the clustering index should be independent of the density of cells (from Pielou, 1977) DIS
n
Density (mean 4- S.D.)
Crowding (mean + S.D.)
A
Hippocampal Side 0.3[S~30bse~ed --
Clustering (mean 4- S.D.)
Poisson
0.2O
Hippocampus 1
3
2 3
15 16
1.68• 3.514-1.36 1.33•
2.824- 0.77 5.67• 1 . 9 7 2.334- 0 . 8 1
1.664-0.32 1.64• 1.76:50.42
.s O EL
0.1
Retina*
0.0
1
4
2 3
7 9
7.88• 7.06• 1.97 7.584- 1.84
25.144- 9.06 26.44• 7.59 28.994- 11.03
3 . 4 0 i 1.28 3.80• 3.944-0.54
35.204- 8.40 16.784- 3.93
3.75• 3.444-0.99
o
5
~o
15
2o
25
30+
Number of c e l l s / s q u a r e
Retinal* side of split plate 1
6
2
9
10.184-3.90 4.94•
B
Hippocampal side of split plate 1
6
2
9
2.07-k 0.58 2.024-0.65
Retinal Side 0 . 6 84
2.66• 0.70 2.894- 1.32
1.30• 1.37•
Observed
0.5 -
Retina* on hippocampal feeder plate 1 2 3
3 17 9
4.74• 2.83• 13.04•
5.38• 1.15 4.124- 1.78 13.80• 1.52
1.13i0.07 1.46• 1.07•
Hippocampus* 1
10
3.70• 1.06
11.29+ 2.65
-
Poisson
0.4
3.124-0.48
/3 0
_Q 0 EL
0.3 0.2 0.1 0.0 0
Table 1. A g r o u p o f experiments, whose results were not included in Table 1, were analyzed in a m o r e objective, statistical m a n n e r as described in the M e t h o d s ; these results are presented in Table 2. Figure 4 A illustrates that h i p p o c a m p a l n e u r o n s demonstrated a spatial distribution n o t greatly different f r o m that predicted by chance. In contrast, retinal cultures clearly deviated dramatically f r o m a r a n d o m distribution (Fig. 4B). Calculation o f the clustering index (Table 2) also confirms that h i p p o c a m p a l cultures show only a slight tendency to f o r m groupings, whereas retinal cultures are strongly clustered. It can be seen that the m e a n density for the h i p p o c a m p a l cultures was m u c h lower than for the retinal cultures. H o w e v e r , even with higher plating densities, h i p p o c a m p a l cultures did n o t f o r m clusters.
Cluster formation is not due to a diffusible factor In order to examine the possible role o f m e d i a - b o u r n e substances in the f o r m a t i o n o f clusters we grew cultures in conditioned medium. C o n d i t i o n e d media f r o m clustered retina did n o t induce the f o r m a t i o n o f clustering in h i p p o c a m p a l cultures, neither did conditioned m e d i a f r o m h i p p o c a m p a l cultures suppress the f o r m a t i o n o f retinal clusters (Table 1). A n additional series o f experiments was p e r f o r m e d
5
I0
15
20
25
30+
Number of cells/square Fig. 4. A Data from a culture of hippocampus (grown on a split-
plate beside clustered retina) have been analyzed as described in the Methods. The number of cells in each grid square has been plotted against the normalized probability of finding a cell in any particular square. The bar graph gives the actual data and the line the distribution predicted for the actual mean if it were to follow a Poisson Distribution. The data did not differ much from the predicted random distribution. B Data from the retinal side of the plate were analyzed and plotted as above. The Poisson Distribution for the experimental mean was used to generate the curve and the data appear as a bar graph. In contrast to the hippocampal side, the retinal side of the culture has neurons widely differing in their distribution from that predicted by chance
using the "split-plate" technique where h i p p o c a m p a l and retinal cultures shared the same culture dish and media. The presence o f shared m e d i u m had no effect on retinal clusters n o r on the lack o f clustering in h i p p o c a m p a l cultures (Table 2). Therefore, simple sharing o f the media apparently was not responsible for the differences between these two types o f cultures.
Putative role of R P E in cluster formation F r o m our previous experience with cultures o f other C N S regions we anticipated that the retinal cultures f r o m
650
Fig. 5
651 the mouse would be more viable if non-neuronal tissue was included to bolster the background layers of cells. As described in the Methods, a non-specified portion of additional tissue from behind the retina, consisting of the RPE cell layer as well as other tissue, was included in the culture. Cultures prepared in this manner demonstrated a high degree of clustering as previously described (see Tables 1, 2). When the neural retina was dissected as cleanly as possible, and cultured without the inclusion of this additional tissue, cluster formation was absent or dramatically reduced. When analyzed using our statistical approach, mouse retinal cultures did demonstrate some degree of clustering although it was very much less than for cultures containing the extra tissue. In spite of this statistical indication that some degree of clustering was occurring, it was qualitatively clear that the clustering was strikingly less in the absence of this additional tissue (Fig. 5A). Furthermore, when the split-plate technique was used, cluster formation occurred only on the side containing the non-neuronal component (not shown). One possible explanation for our results was that the RPE layer contained a population of cells whose presence was required for the induction of clustering. Therefore, we used foetal guinea pigs which possess an easily identifiable cell layer containing the RPE which we could readily isolate from both the surrounding tissue and the neural retina itself. When guinea pig retinae were cultured with or without the RPE layer, the results were unequivocal. In the absence of the RPE layer we saw virtually no clustering (Fig. 5B) while the inclusion of RPE induced the most striking clustering we have observed in any cultures (Fig. 2B). The guinea pig clusters contained so many neurons and the background so few that it was unnecessary to subject them to statistical analysis. In some cases, the neurons within clusters gradually stacked up to form multiple layers of neurons (see Fig. 2B). This process began about one week from initial plating and usually commenced simultaneously at the two ends of the cluster. After two to three weeks the entire cluster was composed of multiple layers of neurons even though similar cultures grown without RPE con-
tinued as monolayers with randomly distributed somata. The high density of neurons in these clustered cultures obscured the background cells underlying each cluster.
Cell-cell contact with background cells induces cluster formation In order to investigate the role of background cells in the establishment of clusters we used feeder plates derived from the mouse spinal cord, hippocampus, cerebral cortex or the retina as a substrate for culturing dissociated retinal neurons. These feeder plates contained background cells but were devoid of neurons, either because of their age or because of treatment with kainic acid. Cultures in excess of one or two months often have complete loss of neurons while the background support cells remain. Bath application of kainic acid is able to destroy neurons (Rothman 1985) and ensures the elimination of most remaining neurons. Feeder plates ranged from 1 week to 2 months in age but the reported observations were similar regardless of the age of the feeder plates. Clusters devoid of neurons could easily be recognized by the swirling pattern or thickened halo of background cells which had previously surrounded each cluster. When dissociated mouse retinal neurons (with or without the non-neuronal retinal tissue) were plated on such feeder cultures from CNS regions other than the retina itself, clusters were not formed and instead neurons were distributed in an almost random fashion across the plate (Fig. 6 and Table 2 for hippocampal feeder plates as an example). In contrast, when dissociated retinal cells were plated on feeder cultures derived from the retina, clusters were again formed as the newly seeded neurons preferentially re-populated pre-existing clusters (not shown). Furthermore, when hippocampal neurons were plated on cultured retinal plates, the hippocampal neurons also preferentially re-populated the cluster regions rather than the regions between the clusters.
Inclusion of nonmeural retinal tissue induces clusterin 9 in hippocampal cultures %-
Fig. 5. A Phase-contrast photomicrograph of mouse retinal culture
grown with as little contamination as possible from non-neuronal retinal tissue. Neurons did not demonstrate an overt formation of clusters although occasionally a number of small clumps can be seen. Note that these groups of cells are not tightly packed and do not show any overt architectonic organization. In addition there are no associated whorls of background cells with the groupings of phase-bright neurons. In other cultures (not shown) there was some suggestion of a whorling pattern of background cells although the neurons did not appear to be associated with these cells. B Cultures of foetal guinea pig retina grown with little or no contamination of RPE cells. These cultures displayed no evidence of cluster formation and this photomicrograph shows the only field which even resembled a cluster. Compare this distribution of cells with that shown in Fig. 3 where the RPE cell layer has been included with the cultured neural retina, and has resulted in the formation of distinct clusters. Calibration bars for A, B, 100 gm
The previous observation suggested to us that neurons regardless of their origin either preferentially adhered to the background cells in the cluster or that those between the cluster presented an unfavorable substrate for neuronal attachment, growth or survival. If this were the case we reasoned that if we included neurons from regions of the CNS that did not form clusters with some of the non-specific tissue or the RPE layer at the time of dissociation, then during the sorting out of cells these neurons should also be induced form clusters. This possibility was bourne out by our observation that when the non-neuronal retinal tissue was included with dissociated cells from the hippocampus the resulting monolayer cultures demonstrated a high degree of cluster formation (Table 2 and Fig. 7).
652
Fig. 6. Mouse retinal cells prepared for culture with non-neuronal retinal tissue and plated onto hippocampal feeder plates. The distribution of the neurons in this culture was approximately random.
Note that despite a relatively dense plating of cells, no clusters are discernable and no whorls of background cells can be seen. Calibration bar, 100 gm
Discussion
represent an organized spatial pattern of cell placement one with respect to the other. In previous studies it has been found that if retinal cells are prevented from attaching to the substratum of the culture dish there is a strong tendency for them to form re-aggregates which sort themselves out and form the major laminations of the retina (Akagawa et al. 1987; Vollmer and Layer 1986A, B; Vollmer et al. 1984). This re-aggregation can be of a rudimentary form ranging from the initial adhesion of cells to each other, to a more complex rosette formation, to the beginnings of cell laminations (Fujisawa 1973; Sheffield and Moscona 1969, 1970). Re-aggregate cultures may even demonstrate complete reconstruction of the major cell layers of the retina, if RPE cells are included with the dissociated cells (Vollmer and Layer 1986A, B; Vollmer et al. 1984). We suggest here that cluster formation, which may also depend upon the presence of RPE, represents the maximal organizational capacities of the cells when they are restricted to the two dimensional tissue culture plate as opposed to the three dimensional growth domain of re-aggregates. Therefore, we feel that cluster formation can serve as a developmental model of the early stages of cell sorting and migration. This model presents several advantages over re-aggregates. For example, the cellular organiza-
We have shown that monolayer cultures of mouse, cat and guinea pig retina differ from those of other laminated CNS regions. Retinal neurons in such cultures demonstrate a rudimentary architectonics unlike the apparently stochastic placement of neuronal somata characteristic of the other kinds of CNS monolayer cultures studied. The architectonics shown by retinal cultures takes the form of cell clusters. Each cluster consists of a monolayer grouping of neurons in some cases surrounded by whorls of background cells but in others consists simply of neurons, placed on top of unidentified flat cells. One possibility is that the flat cells are of glial origin (Li and Sheffield 1984, 1986A). Even though we often observed GFAP-positive astrocytes intimately associated with some mouse retinal clusters it was also clear that many clusters did not contain such immunoreactive cells. Therefore, it appears unlikely that this subpopulation of glial cells plays a primary role in the formation of clusters. It is important to note that the clusters are not simply clumps of neurons which have adhered to each other during the dissociation, settled to the culture dish, and subsequently developed in place. Rather, the clusters
653
Fig. 7. Non-neuronal tissue was included with dissociated hippocampal cells at the time of dissociation. In this phase-contrast photomicrograph it is clear that unlike conventional hippocampal
cultures the neurons in this cuIture environment were distinctly clustered. Calibration bar, 100 gm
tion in the clusters may be simpler, and because they are restricted to two dimensions histological and immunohistochemical techniques can easily be applied. Furthermore, the neurons within the cluster are in a monolayer and are readily accessible for electrophysiological and pharmacological analysis. The initial cell adhesions and the formation of cellular junctions in re-aggregate cultures are thought to be important for the subsequent laminar reconstruction (Sheffield 1970). Specific cell ligands on the extracellular surface may also contribute to this process (Barnstable 1987A, B; Moscona 1974). Even though retinal cell clusters do not obviously resemble re-aggregates they also may have arisen as a consequence of some preferential cell adhesion following plating but just prior to attachment to the substratum. Thus, retinal neurons might recognize each other, adhere and bias their eventual location on the culture plate. Similarly, non-neuronal cells might also be subject to a preference for "like" cells, according to the hypothesis of "cell affinities" (Steinberg 1963) where "alike" cells and tissue have a marked preference to reorganize and form contacts with each other. Recent experiments suggest an expansion of this hypothesis to include the view that each type of cell
expresses on its extracellular surface unique cell-typespecific factors qualitatively different from those expressed on other kinds of cells. Various glycosamines and other types of adhesion molecules might serve this purpose (see Getch and Steinberg 1986; Mendez-Otero et al. 1988; Barnstable 1987A, B). Presumably neurons from regions of the CNS other than the retina would also be subject to the same principle of "cell affinities". However, cluster formation was never observed ifi CNS cultures from a wide variety of locations unless retinal tissue containing the RPE was included at the time of dissociation. Background cells from the retina but from no other region of the CNS examined could support the formation of clusters. This new finding may indicate that the cultures derived from these other CNS tissues lack the types of neuroepithelial cells and adhesion factors (or their expression) required for lamination in vivo. Our observations with cat retina have clearly demonstrated that the formation of clusters is more complex than just cell adhesion and must also consist of a distinct phase of cellular migration. We have not yet attempted to determine if this migratory phase occurs as the result of active migration of neurons or alternatively, may be due to the movements of backgrounds cells to which they are attached. We did not observe a similar phase in
654 cultures of the mouse retina presumably because clusters were established much more rapidly in this species than in the cat. Strong influences of background cells upon retinal neuronal growth and migration have been noted by Li and Sheffield (1986A, B). Determination of the precise role of migration in cluster formation will require further study. One alternative explanation for the formation of retinal clusters would be a selective die off of "non-clustered" neurons or conversely the selective induction of neuronal survival by neurotrophic growth factors. For example, nerve growth factor (Adler 1986) and brainderived neurotrophic factor (Hofer and Barde 1988) both increase the survivability of peripheral and central neurons respectively and promote outgrowth of processes. It is possible that cells within the clusters or a population of background cells expresses or secretes a factor responsible for stimulating neuronal proliferation. However, our initial attempts at reproducing cell clustering by using conditioned media or by a constant sharing of same media in the split-plate cultures gave essentially negative results indicating that a freely diffusible factor which can act at long distances was not responsible for clustering. It may be necessary for retinal neurons to be exposed to concentration gradients of inductive factors released by support cells located within a few microns of each other. Therefore, it is possible that clustering failed to occur in the split-plate cultures because cells would simply not be close enough to each other to permit this kind of interaction. Further experimentation would be required to determine if this were the case. However, we believe that diffusible neurotrophic factors are unlikely to play a major part in cluster formation for several reasons. The clear migration of cells in feline retinal cultures argues strongly against a simple selective die off. Furthermore, retinal clusters in mouse and guinea pig appear to be associated with specific groupings of background cells and depended upon the unique presence of RPE cells. Our results are consistent with the hypothesis that retinal cell cluster formation is dependent upon direct cell-cell contact factors associated with non-neuronal cells specifically from the retina. Retinal neurons cultured on top of feeder plates of hippocampal background cells failed to form clusters. On the other hand, there was a preferential repopulation of clusters in feeder plates of background cells from retina which had previously formed clusters. In addition, we were able to demonstrate that cluster formation was strongly dependent upon the inclusion of non-neuronal retinal cells in the original dissociation. The experiments using guinea pig retina further indicated that the RPE cell layer was sufficient to induce the formation of clusters. Similarly, Vollmer and Layer (1986B) report that the formation of retinal laminations in re-aggregates depends strongly upon the presence of RPE cells. Furthermore, neither extracts nor conditioned medium from RPE cultures could induce laminations in the re-aggregates and direct physical contact between the retinal neurons and RPE cells was necessary. Another possibility for the induction of clustering formation may be the inclusion of mesenchymal cells in
the non-neuronal tissue. While the inductive properties of embryonic mesenchyme have been well described in the process of limb formation (Balinsky 1981), with regards to the retina, there is strong evidence for the role of the RPE in the induction of retinal cytoarchitectonics. Based on this literature and the present evidence that inclusion the RPE cell layer is sufficient to cause clustering, we feel that the RPE is the most likely candidate for the induction of cluster formation. The extracellular surfaces of neurons and other cells of the developing retina express a wide variety of molecules many of which are important for cell attachment and adhesion not only to the culture substrate but also of cells t o each other. Furthermore, the expression of such specific adhesion molecules is often topographically oriented within the retina in vivo suggesting that the architectonics of cell-cell interactions are determined, at least in part, by such molecules (for reviews see Getch and Steinberg 1986 and Barnstable 1987A, B). In addition, a number of important components of the extracellular matrix, such as laminin and fibronectin (Heaton and Swanson 1988; Pixley 1987) can serve to attach specific cell types and/or guide the outgrowth of neuronal processes. We are presently examining whether or not some of these cell-contact factors play an important role in the formation of retinal cell clusters.
Acknowledgement. This work has been supported by the Medical Research Council of Canada.
References Adler R (1986) Trophic interactions in retinal development and in retinal degeneration: in vivo and in vitro studies. In: Adler R, Farber D (eds) The retina: a model for cell biology studies, Part I. Academic Press, New York, pp 112-151 Akagawa K, Barnstable CJ (1986) Identification and characterization of cell types in monolayer cultures of rat retina using monoclonal antibodies. Brain Res 383 : 110-120 Akagawa K, Hicks D, Barnstable CJ (1987) Histiotypic organization and cell differentiation in rat retinal reaggregate cultures. Brain Res 437 : 298-308 Balinsky BI (1981) An introduction to embryology. Saunders Publishing Co, Philadelphia Barnstable CJ (1987A) Immunological studies of the diversity and development of the mammalian visual system. Immunol Rev 100:47-78 Barnstable CJ (1987B) A Molecular view of vertebrate retinal development. Mol Neurobiol 1: 9-46 Fujisawa H (1973) The process of reconstruction of histological architecture from dissociated retinal cells. Wilhelm Roux' Arch 171:312-330 Getch CC, Steinberg MS (1986) Differential adhesion in neuronal development. In: Adler R, Farber D (eds) The retina: a model for cell biology studies, Part I. Academic Press, New York, pp 67 109 Goffinet AM (1985) Events governing organization of postmigratory neurons: studies on brain development in normal and reeler mice. Brain Res Rev 319:261-296 Heaton MB, Swanson DJ (1988) The influence of laminin on the initial differentiation of cultured neural tube neurons. J Neurosci Res 19:212-218 Hofer MM, Barde YA (1988) Brain-derived neurotrophic factor prevents neuronal death in vivo. Nature 6153:261-262
655 Li H-P, Sheffield JB (1984) Isolation of flat cells, a subpopulation of the embryonic chick retina. Tissue Cell 16 : 843 Li H-P Sheffield JB (1986A) Retinal flat cells are a substrate that facilitates retinal neuron growth and fibre formation. Invest Ophthalmol Vis Sci 27:296-306 Li H-P Sheffield JB (1986B) Retinal flat cells participate in the formation of fibers by retinal neuroblasts in vitro. Invest Ophthalmol Vis Sci 27:307-315 Liu L, Cheng SH, Jiang LZ, Hansmann G, Layer PG (1988) The pigmented epithelium sustains growth and tissue differentiation of chicken retinal explants in vitro. Exp Eye Res 46:801-812 MacDonald JF, Miljkovic Z, Pennefather P (1987) Use-dependent blockade of excitatory amino acid currents by ketamine. J Neurophysiol 58 : 251-266 MacDonald JF, Wojtowicz JM (1982) The effects of L-glutamate and its analogues upon the membrane conductance of central murine neurons in culture. Can J Physiol Pharmacol 60 : 282-296 Mendez-Otero R, Schlosshauer B, Barnstable CJ ConstantinePaton (1988) A developmentally regulated antigen associated with neural cell and process migration. J Neurosci 8:564-579 Moscona AA (1974) Surface specification of embryonic cells: lectin receptors, cell recognition and specific cell ligands. In: Moscona AA (ed) The cell surface in development. John Wiley & Sons, New York, pp 67-99 Pixley SKR, Nieto-Sampedro M, Cotman CW (1987) Preferential
adhesion of brain astrocytes to laminin and central neurites to astrocytes. J Neurosci Res 18:402-406 Pielou EC (1977) Mathematical ecology. John Wiley & Sons, New York Rothman SM (1985) The neurotoxicity of excitatory amino acids is produced by passive chloride influx. J Neurosci 5:1483-1489 Sheffield JB (1970) Studies on aggregation of embryonic cells: initial cell adhesions and the formation of intercellular junctions. J Morphol 132 : 245-264 Sheffield JB, Moscona AA (1969) Early stages in the reaggregation of embryonic chick neural retinal cells. Exp Cell Res 57: 463-466 Sheffield JB, Moscona AA (1970) Electron microscopic analysis of aggregation of embryonic cells: the structure and differentiation of aggregates of neural retina cells. Dev Biol 23:36-61 Steinberg MS (1963) Reconstruction of tissues by dissociated cells. Science 141:401-408 Vollmer G, Layer PG (1986A) An in vitro model of proliferation and differentiation of the chick retina: coaggregates of retinal and pigment epithelial cells. J Neurosci 6:1885-1896 Vollmer G, Layer PG (1986B) Reaggregation of chick retinal and mixtures of retinal and pigment epithelial cells: the degree of laminar organization is dependent on age. Neurosci Lett 63 : 91-95 Vollmer G, Layer PG, Gierer A (1984) Reaggregation of embryonic chick retinal cells: pigment epithelial cells induce a high order of stratification. Neurosci Lett 48:191-196
