Formation of the Retinal Ganglion Cell and Optic Fiber Layers Michiko Watanabe,' * Urs R u t i s h a ~ s e rand , ~ ~ Jerry ~ Silver3 Department of 'Pediatrics, 'Genetics, and 3Neurosciences,Case Western Reserve University School of Medicine, Cleveland, Ohio 441 06
SUMMARY The early development of retinal ganglion cell and the optic fiber layers has been studied by examining the morphology of differentiating retinal ganglion cells using immunoelectron microscopy and a monoclonal antibody against neuron-specific beta-tubulin. This antibody identified retinal ganglion cells during the stages of their most active differentiation and axonogenesis prior to maturation of other retinal neurons. The changing morphology of retinal ganglion cells during these early
stages is consistent with a differentiation sequence in which axonogenesis and translocation of the cell body to the vitreal surface occur while the cell is still attached to the vitreal margin through its vitreal endfeet. Thus, the mechanism of retinal ganglion cell axon generation and soma migration to the vitreal surface appears to involve maintenance of this attachment which may act as both a focus for axon differentiation and an anchor for directed nuclear translocation to the vitreal margin.
INTRODUCTION
reconstructions (Hinds and Hinds, 1974). It is established that an RGC precursor, with a neuroepithelial cell morphology indistinguishable from neighboring cells, undergoes mitosis at the ventricular surface (Ramon y Cajal, 1960). However, questions still exist regarding the sequence of morphological changes that occur as a differentiating neuron moves toward the vitreal margin, how it is guided there, and how its rounded cell body stabilizes and extends an axon specifically and with stereotypic directionality and position along this margin. At least two sequences of RGC differentiation have been proposed. In one model (Ramon y Cajal, 1960), RGC begin differentiation at the ventricular surface where they have undergone their last mitosis and enter into a freely migrating bipolar stage where the vitreally projecting process acts as a growth cone. The process, apparently guided to the vitreal surface by the radial cells, takes a right angle turn at the vitreal margin, as it is deflected by obstacles (endfeet, axons, the limiting membrane), and extends as an axon. In this scheme, the attachment of RGC endfeet at the vitreal margin does not serve to orient the RGC soma
The neural retina begins as a simple pseudo-stratified neuroepithelium, and thereafter differentiates into an orderly stratified structure with alternating cell body and neurite domains. Ganglion cells are the first neurons to reach their final position in the neural retina and together with their processes comprise the first detectable layers, a row of cell bodies with associated axons coursing along the vitreal surface and overlying the somata. The morphological sequence of retinal ganglion cell (RGC) development in a variety of species has been described based on the interpretation of Golgi-impregnated material (Morest, 1970; Nishimura, 1980; Prada, Puelles, and Genis-Galvez, 1981 ), horse radish peroxidase (HRP) back-filling (Maslim, Webster, and Stone, 1986), electron microscope (EM) sections (Rager, 1980), and EM serial Received June 28, 1990; accepted August 22, 1990 Journal of Neurobiology, Vo1. 22, No. I , pp. 85-96 (1991) 0 1991 John Wiley & Sons, Inc. CCC 0022-3034/91/0 10085-I2$04.00 * To whom correspondence should be addressed.
85
86
Watanabe el ul.
or axon. In contrast, Morest ( 1970) proposed that RGC differentiation progressed after the cell had regained its neuroepithelial morphology and, following commitment to a neuronal fate, generates the axon at the correct vitreal level directly from the vitreal endfoot. Endfeet all along the route would serve as a preferred pathway for growth out of the retina (Silver and Robb, 1979; Rager, 1980; Silver and Rutishauser, 1984). In this sequence, the endfoot plays an essential role by serving as an anchor for nuclear translocation and guide during axonogenesis. Clarification of the sequence of RGC morphology is a prerequisite for evaluating mechanisms proposed to govern orderly retinal histogenesis. Furthermore, an understanding of normal development can serve as a basis for uncovering mechanisms that produce retinal abnormalities. In this study, the details of RGC differentiation were examined by immunoelectron microscopy using a monoclonal antibody (TuJl ) directed to an isoform of beta-tubdin. Because the isoform is detectable in neurons as early as terminal mitosis (Moody, Quigg, and Frankfurter, 19891, the antibody allowed detection of RGC and their processes during the earliest stages of their differentiation. These methods have yielded evidence in support of the hypothesis that vitreal endfeet play a role in the establishment of RGC and axon position at the inner margin of the retina.
ImmunoPeroxidase Staining of Intact Retina Tissue The primary antibodies used were ( 1 ) purified IgC of monoclonal antibodies SE directed to embryonic chicken brain neural cell adhesion molecule NCAM (Frelinger and Rutishauser, I 986; Watanabe, Frelinger, and Rutishauser, 1986) and ( 2 ) TuJ1, a monoclonal antibody directed to the neuron-specific isotype of betatubulin (Moody et al., 1989). The monoclonal antibody TuJ 1 was generously provided by Dr. Frankfurter (University of Virginia, Charlottesville) . Tissues were incubated overnight in primary antibodies, washed three times for 1 h each in HBSS, incubated overnight in secondary antibodies conjugated to peroxidase, washed extensively in HBSS, and reacted with the peroxidase-substrate diaminobenzidine (D.4B) in the presence of metal ions ( A d a m , 1981 ).
Electron Microscopy Fixed and immunostained tissues were subsequently prepared for electron microscopy by a procedure developed by McDonald ( 1984). The tissues were embedded
METHODS Tissue Preparation Neural retina of white Leghorn chicken embryos (Gullus gulfus) were staged (Hamburger and Hamilton, 195 1 ) and dissected in Dulbecco’s phosphate-buffered saline (PBS) or HBSS (Hanks Balanced Salt Solutions, GIBCO). The eyes of stage 17-19 embryonic day three (E3) embryos were dissected and the lens and the surrounding mesenchyme removed. For embryos stage 23 (E3-4) and older, pie-shaped wedges of neural retina tissues dorsal to the optic fissure (Fig. 1 ) were dissected free from the pigmented epithelium and other tissues to allow thorough penetration of fixative. antibodies, and diaminobenzidine substrate. Most of our observations were made between E3 and E7, stages during which RGC are most active in migrating to the vitreal surface (Kahn, 1973, 1974) and, therefore, when we were more likely to capture an RGC in its early phases of differentiation. The tissues were immersed overnight (9- 15 h ) in a fixative developed by McClean and Nakane ( 1974), rinsed in phosphate buffer, and immunostained.
W Figure 1 Eyes were dissected for whole-tissue immunostaining. For embryos stage 23 (E3-4) and older, pieshaped pieces (dotted area) of neural retinas were dissected from an area dorsal to the optic disc and nerve ( O N ) . The shape allowed orientation of tissue during sectioning for light or electron microscopy. Sections perpendicular to the plane of the retina were taken along the central-peripheral axis in order to increase the chances of getting an RCC axon and its cell body in the same section. The smaller size of the tissue facilitated penetration of fixative, antibodies, and the peroxidase substrate DAB. L = lens; ON = optic nerve.
Formation ofRetina1 Layers in low-viscosity plastic (Spurrs: Electron Microscopy Supplies) and sectioned for both light (2+m sections) and electron microscopy. The tannic acid treatment in McDonald’s ( 1984) procedure was eliminated in some preparations to prevent intense counterstaining of basement membranes.
lmmunofluorescence Staining of Frozen Sections Tissues were fixed and cryoprotected for frozen-sectioning as described by McClean and Nakane ( 1974). Sections ( 10 p ) were collected on gelatin-coated glass slides and stained by the indirect immunostaining procedure using the biotinylated second antibody (Vector) followed by streptavidin conjugated to Texas Red (Amersham). Photographs were taken with KODAK TMAX film (ASA 400) through an epifluorescence microscope (Nikon, Optiphot).
Determination of Antibody Penetration Fluorescently labelled goat anti-mouse IgG penetrated neural retina tissues from E3 to E 1 1 (stage 1 9 /20-stage 36/37) embryos under conditions used for immunostaining with neuron-specific TuJ 1 and anti-NCAM (neural cell adhesion molecule) monoclonal antibody (mAB) 5E both of which are mouse IgGs. The staining of the E3-E8 neural retinas was of equal intensity throughout the width of the tissues, indicating even penetration of the labelled antibody. whereas the staining of El 1 neural retina was more intense at the vitreal and ventricular edges than in the center of the tissue, indicating resistence to antibody penetration. Goat anti-mouse IgG antibodies conjugated to fluorescein were incubated with E3-El I central neural retina pieces overnight as for the incubation of first antibody during the normal staining procedure. These retina pieces varied in thickness and in the maturation of vitreal and ventricular junctions, which may have affected penetration of the antibodics. Despite these variations with age, all retina pieces had fluorescent antibody-penetration to the center of the tissue. The intensity of fluorescence staining for E l 1 retinas was not as even across the tissue as for the younger tissues. Therefore, preferential staining of the vitreal edge is not caused by limited first antibody penetration for retinas up to E8. Further evidence exists for even penetration of immunological reagents into neural retinas. Other monoclonal antibodies of the IgG sublype drected to NCAM evenly stain the width of E4-E6 central neural retina pieces which are thicker and have more mature junctions. Bipolar, neuroepithelial cells that stain with TuJ 1 by a whole-mount technique do so evenly throughout the span of the retina from vitreal to ventricular side. Penetration of antibodies into neural retina explants
87
have been demonstrated for Fabs and antibodies (Silver and Rutishauser, 1984; Halfter and Chen, 1987). Furthermore, neural retinas stained with T d l , after the width of the tissue was exposed by frozen-sectioning, had identical staining patterns to that of whole tissues.
RESULTS The immunological method used in this study for staining RGC has advantages over previously used techniques. One advantage is that immature RGC can be identified before they have migrated to the vitreal surface. Another advantage is that the majority of postmitotic RGC rather than a subset are stained with this technique. In addition, with the ability to see the immunostain and ultrastructure at the same time, even parts of RGC can be detected. Immunoelectron microscopy with the TuJl antibody also allows identification of fine or immature RGC processes that are not well stained by reduced silver methods (Rager, Lausmann, and Gallyas, 1979). At the ultrastructural level, it is possible to determine in some cases, that a particular process projects from a particular cell body rather than just traveling adjacent to it. The relative extent of RGC maturation can also be inferred by comparing the intensity of TuJ 1-staining, which increases with differentiation.
Identification of Chicken Embryo Retinal Ganglion Cells by lmmunostaining with TuJl Cells were clearly visible in whole mounts of E4-E7 retina after immunoperoxidase staining for neuron-specific beta-tubulin ( TuJ 1 ). They were, in addition, identified as RGC by the vitreal position of their soma and axons. Flattened wedges of neural retina taken from the posterior pole of the eye (Fig. I ) immunostained with TuJl displayed a gradient of staining [Fig. 2( a ) ] , which was intense at the central region and gradually faded towards the periphery, reflecting the well-known gradient of maturation of the developing retina (Halfter et al., 1985: Kahn, 1973, 1974; Rager, 1980: Ramon y Cajal, 1960). The centrally located, intensely stained cell bodies could be focused at the same level [Fig. 2(b)], whereas the peripheral, more lightly stained cell bodies had to be focused at various levels within the neural retina [Fig. 2 ( c ) J . These results are due to the movement of RGC toward the vitreal side, the increase in their inten-
88
Wutunabe et al.
Figure 2 TuJ 1-stained cell bodies and axons of RGC. In stage 25 (E4.5) neural retina, it was possible to observe at the light microscope level a gradual increase in number of the cells stained with T d 1 and intensity of stain per cell from the peripheral to central retina ( a ) . Single axons from RGC bodies interacted with other axons soon after they protruded from the cell body and traveled with other axons (b,c). When immunoperoxidase-stainedneural retina pieces embedded in resin were sectioned perpendicular to the plane of the retina (d,e), it was revealed that most of the intensely stained cell bodies and axons were located along the vitreal margin. In these sections, it was possible to identify RGC with a length of its axon protruding from the cell body (d). More often axons were sectioned in cross section and appeared as black dots (e). (RGC cell bodies are 10-1 5 hm in diameter.)
89
Formation of Retinal Layers
-of I
-in1
* Ph
Figure 3 TuJl specifically stained RGC up to EX (st. 34) neural retinas. Intensely immunostained cells could be detected only at the vitreal surfaces of retinas from E6 (a) and E8 ( b ) animals. This is the location of RGC that are the first to differentiate and arrive at their final position in the retina. By E 1 1 (c), TuJ 1 also stained photoreceptor cells (ph) and cell bodies adjacent to the inner plexiform layer (ipl), which may belong to amacrine or bipolar cells. The arrowhead points to the vitreal surface. ofl = optic fiber layer; in1 = inner nuclear layer; ph = photoreceptors. Scale bar = 100 hm.
sity of TuJ 1-immunostaining with maturation, and the more peripheral and less mature cells differentiating at different levels. The TuJ l-immunostained cells were detected as early as stage 14, which is when the first RGC have been detected by silver impregnation (Halfter et al., 1985). The antigen recognized by TuJ 1 accumulated in both cell bodies and axons during differentiation. When the immunostained tissue wedges were sectioned perpendicular to the plane of the neural retina, it was revealed that many of the intensely stained cells and their axons were located at the vitreal surface [Fig. 2 (d,e)] . TuJ 1 specifically stained RGC in the vitreal margin up to and including E8 (Fig. 3 ) . By E 1 1 [ Fig. 3 (c)] the antigen was no longer specific for RGC and appeared in photoreceptor cells as well as cell bodies located in the inner nuclear layer adjacent to the inner plexiform layer. Therefore, during E2-E6, those stages examined in detail in this study, Tul 1 could be used to identify RGC.
bound to the cell bodies and axons of RGC that had migrated to the vitreal surface of the retina. When TuJ 1-stained cell bodies were counted in E4 and E6 retina (Table 1 ), 96%-97% were located in the vitreal third. The rest (3%-4%) were located in the ventricular two-thirds of the retina. Even at the light microscope level it was possible to detect TuJl-stained cell bodies within the neuroepithelium that were still attached to vitreal
Neuroepithelial Morphology TuJl bound to cell bodies as well as vitreal and ventricular processes in primitive RGC. It also
lmmunostained tissues were embedded in plastic and 2-pm serial sections were collected. Stained cells in every fifth section were counted to reduce the probability of counting the same cell twice (RGC somata are 10-1 5 p m in diameter).
Table 1 TuJ1-Stained Cell Bodies in E4 and E6 Retina
E4 (st. 25/26)
E6 (st. 28/29)
No. of Stained Cells
%
Region
533 15 8 1030 26 1
96 2.4 1.4 97 2.5 1
Vitreal third Middle third Ventricular third Vitreal third Middle third Ventricular third
90
Wutanuhe et al.
Figure 4 TuJ 1-stained primitive-appearing RGC in the process of nuclear translocation. In addition to staining cells that had the characteristic morphology of differentiating RGC, that is, with rounded cell bodies and axons at the vitreal surface, TuJ 1 also stained cells with the morphology of primitive neurons. A 2-pm plastic section of an immunoperoxidase-stained E4 neural retina revealed that a TuJ 1-positive RGC can have a soma in the middle of the retina and an endfoot bound to the vitreal limiting membrane (a). This particular endfoot appears to contain many vesicles when observed under phase and Nomarski optics. A computer-enhanced mirror-image (image analysis) of the same cell emphasizes the continuity of the process between the cell body and the endfoot ( b ) . Arrowheads point to the vitreal surface. RGC with primitive morphologies were also detected in 10-pm frozen sections of embryonic retina that were stained for fluorescence microscopy using the sensitive biotin/ streptavidin/ Texas Red procedure. One of these RGC extends a process almost halfway across the neural
Formalion of Retinal Layers
and/or ventricular processes (Fig. 4j. Because the antigen accumulated in neuroepithelial-shaped cells, it was possible in certain sections to follow the cytoplasmic processes extending from cell bodies to endfeet at the tissue limits. Presumptive transitional RGC with cell bodies more than three somatic diameters (30-50 pm) away from the vitreal surface were present in neural retinas of embryos as old as E8 and E 12 in both the central and peripheral regions. These often had long trailing ventricular processes. Faint staining was also apparent in cells with cell bodies located in the ventricular side of the retina adjacent to the pigmented epithelium. Vitreal Endfeet and Axonogenesis
In order to determine whether axons were emerging from these primitive-appearing RGC, it was necessary for us to examine the tissue at the electron microscopic level. Embryonic eyes and retinal tissues immunostained before embedding had intact vitreal basement membranes, plasma membranes of cell bodies, endfeet, and axons, nuclear membranes, and mitochondria1 membranes. Specialized junctions could be recognized including the apical junctions of the vitreal side and zonula adherens in the ventricular region. Even in retina pieces that were not immunostained, cells could be identified at the electron microscopic level which had the characteristics of retinal ganglion cells, that is, cells with rounded bodies adjacent to the vitreal surface. Some of these had thick vitreal attachments to the basement membrane (Fig. 5 j . TuJl stained the vitreal endfeet, of which some had adjoining processes of the caliber and immunostaining intensity of axons (Fig. 6 ) . Several TuJ 1-negative endfeet had TuJ 1-positive neurites protruding from them. Clear vacuoles were often observed in endfeet with adjoining TuJl positive neurites.
91
Stained cell bodies within the marginal zone sometimes possessed two processes protruding towards the vitreal surface, with one process more intensely TuJ 1-positive than the other. Those RGC with cell bodies at the vitreal surface possessed axon-like processes that were intensely stained with TuJ 1. DISCUSSION
The morphology of chicken retinal ganglion cells was observed during early phases of differentiation in order to deduce the cellular mechanics that generate the characteristic positioning of RGC soma and axons at the vitreal surface. We used antibody TuJl to identify retinal neurons at a stage when RGC, exclusive of other neurons, were undergoing terminal mitosis and differentiation. Retinal ganglion cells were identified during the early phases of neurogenesis using this antibody which binds to a neuron-specific isoform of beta-tubulin. This protein is found in the soma and most processes of neurons. The antibody T d l is particularly useful for the study of neurogenesis because it recognizes the protein when it is expressed during or shortly after terminal mitosis (Moody et al., 1989). With this antibody it was possible to identify and mark the retinal ganglion cells at a primitive state of differentiation. In retinal tissues immunostained with TuJl, cells with somata at the ventricular side of the retina and displaying a primitive neuroepithelial morphology were stained throughout their soma and both vitreal and ventricular processes. These were identified as primitive RGC presumed to be at a stage before or during the process of nuclear migration. TuJ 1-positive cells of this primitive morphology were detected in E3-El2 embryos in both central and peripheral neural retina. Previous identifications of primitive RGC have been based
retina (c) which may be part of a vitreal process or a trailing ventricular process. Another cell ( d ) spans the neural retina, with its cell body in the middle of the neural retina, its ventricular process ending in an endfoot, and its vitreal process projecting towards the vitreal surface. In this photograph, the ventricular and vitreal processes appear to be stained, and more intensely so than the cell soma (small arrow), which is out of the plane of focus. In the peripheral E6 ( st. 29) neural retina, cell bodies and axons located at the vitreal margin (arrow) were intensely stained (c). Similar morphologies of RGC in the process of translocation with vitreal and/or ventricular processes could be detected in E6 peripheral retina ( e) TuJ 1 -immunofluorescently stained cryostat sections. The periphery of the tissue is to the right and the central region is to the left of each photograph. The arrows point to the vitreal surface. Scale bar = 50 Fm.
92
Watanabe ef a/.
Figure 5 Cells with their nuclei at the vitreal surface were attached to this surface by basal processes or endfeet. These cells had different cytoplasmic consistencies from neighboring cells. The cytoplasm of these cells appeared less dense than neighboring cells (a-d) but some were more dense (e). The arrowheads point to the vitreal surface. Scale bars = 1 Km.
on other criteria. For example, Prada et al. ( 1981 ) identified early RGC as radial cells without ventricular endfeet. Hinds and Hinds ( 1974) used EM serial sections to identify RGC based on migration of centrioles and cilia from the ventricular to vitreal side of nuclei within elongated neuroepithelial cells. According to Hinds and Hinds ( 1974) these did not have vitreal endfeet. Maslim et al. ( 1986) used HRP backfill to locate RGC and observed RGC each with an axon and both vitreal and ventricular processes. They also observed radial cells presumably backfilled with HRP through their axons with processes connecting them both to vitreal and ventricular margins of the retina. However, these labelled cells were close to the site of contact of HRP crystals and could, therefore, have been stained by direct contact with crystals rather than by axonal transport. A cytoplasmic protein RA4 has been identified in developing chicken retina that is selectively expressed in the optic fiber
layer in later stages of development ( McLoon and Barnes, 1989). At earlier stages antibodies to this antigen stain cells similar to those described here that have a primitive, bipolar appearance. However, no analysis at the electron microscope level has been carried out to confirm the presence or absence of endfeet or axons on these RA4-positive cells. The neuron-specific immunostaining procedure used in this study provided independent criteria for identifying RGC, provided evidence that differentiating RGC can have morphologies even more primitive than previously described, and allowed ultrastructural analysis of axons and endfeet. We do not know if every TuJl-positive cell including the neuroepithelial-shaped cells are necessarily undergoing axonogenesis. This determination may only be obtained by serial-section reconstructions at the electron microscope level of many TuJ 1-immunostained cells. Axons have been proposed to arise from neuro-
Formation ofRetinal Layers
Figure 6 The cytoplasm of certain endfeet were distinctly positive with TuJl immunostaining. These RGC endfeet were attached to the basal lamina and contained mitochondria. The vitreal surface (arrowhead) is at the top of each micrograph. Some samples (a,b) were simultaneously stained with anti-NCAM mAB 5E,which stains plasma membranes thus delineating the outline of the cells. Processes projected from TuJ 1-positive endfeet (b,c). The projection in ( b ) is similar in diameter and staining intensity to an axon in cross section (white *). T d 1-positive processes in ( c ) include transverse sections through axons and a fan-shaped region that may be a portion of a growth cone. The plasma membrane of one axon is continuous with an endfoot filled with vesicles (star). The arrowhead points to the vitreal surface. An RGC with a cell body already translocated to the vitreal margin possesses two processes (d). The right process, which is intensely TuJ 1-stained, ends in a club-like structure that resembles the base of a growth cone. The other process on the left is shorter and a portion of it is unstained with TuJ1. The end of this process is indicated by an asterisk (*). This cytoplasmic process was followed through serial sections (not shown) and extended no further than shown in this figure. The arrowhead indicates the vitreal surface. Scale bar = 1 pm.
93
94
Watanabe et al.
epithelial cells by different mechanisms (Fig. 7): ( 1 ) by lateral growth of specialized processes from the side of vitreal endfeet (Morest, 1970; Nishimura, 1980; Prada et al., 1981) or soma (Maslim et al., 1986): ( 2 ) as transformations of the vitreal endfeet which detach from the vitreal surface; or ( 3 ) as transformations of vitreally directed processes that do not reestablish stable contact with the basal lamina but are deflected by it (Ramon y Cajal, 1960; Hinds and Hinds, 1974; McLoon and Barnes, 1989). Our results support the idea that in the embryonic chicken retina, axons grow out of endfeet. One line of evidence is that elongated processes with the proper caliber for axons and positive for the neuron-specific antigen TuJ 1 were detected protruding perpendicularly from intact vitreal endfeet. We also recognized a number of
end foot
4
1
2
.,.J..&. *y.. . ..,a.
3
.-..,.....
4
.-::.:
.I. I . ’ : . .
A:.:;.
$6” 1
2
3
4
vent v- i_t ..I:-’;
vent
Figure 7 Proposed sequences for retinal ganglion cell differentiation. Sequence A: A cell undergoes mitosis at the ventricular surface ( 1), reextends to span the neuroepithelium ( 2 ) , sends an axon from the vitreal endfoot while retracting the ventricular endfoot and translocating the nucleus ( 3 ) , and attains the final vitreal position with an axon growing along the margin and cell body stabilized at the vitreal region ( 4 ) . Sequence B: A cell undergoes mitosis at the ventricular surface ( 1 ) , sends a process towards the vitreal surface (2), attains a freely migrating bipolar stage during which the cell body, led by a vitreal process, migrates to the vitreal surface where the process makes a right angle turn to project along the margin, transforms into an axon ( 3 ) , and reaches the vitreal surface (4).
retinal ganglion-like cells with rounded bodies at the vitreal surface still connected to the vitreal basal lamina by thick stalks of cytoplasm. We suggest that these may be RGC that have translocated nuclei but have not eliminated their endfeet. Thus, findings from this study support the proposal that RGC send out axons from their vitreal endfeet parallel to the basal lamina (Morest, 1970). The vitreal endfoot can be retained during the early part of axonogenesis and is conjectured to be a stabilizing structure for nuclear translocation. However, axonogenesis, release of the ventricular endfeet, and cell body translocation must occur rapidly. This is suggested by the small number of TuJ 1-positive cells with radial morphology and with soma within the middle and ventricular third of the neural retina even during the most active time of RGC differentiation. The small percentages of the “transitional cells” are consistent with the frequency of similar cells detected by Hinds and Hinds (1974) using EM reconstructions of serial sections as well as the observations by Morest ( 1970), who also concluded that translocation must be a rapid process. Eventually, the endfeet of RGC are eliminated. Observations from this study would be consistent with either their retraction into the RGC soma or elimination by autolytic disintegration. We have observed endfeet devoid of T d l staining that are filled with vesicles from which a TuJ 1-positive neurite protruded. One interpretation for these observations is that the endfoot degenerates soon after axonogenesis and in the process loses the protein recognized by anti-TuJ 1, whereas the axon retains it. Such redistributions in the subcellular localization of proteins within differentiating neurons have been noted for such axon-associated proteins as GAP-43 ( G o s h , Schreyer, Skene, and Banker, 1990) and RA4 (McLoon and Barnes, 1989). The numerous vesicles within the endfeet may be lysosomes containing autolytic enzymes. Lysosomes have also been detected by Hinds and Hinds ( 1974). In addition, they observed pinching off of cytoplasm by neuroepithelial cells during elimination of ventricular processes. which could be another mechanism used for vitreal endfeet elimination. We observed TuJ 1-stained cell bodies possessing two processes protruding towards the vitreal surface with one process more intensely stained than the other. Maslim et al. (1986) also noticed cells with two processes, one being an endfoot and another an elongated axon. Ramon y Cajal ( 1960) proposed that bifurcation yielding
Formation of Retinal Layers
two processes may arise when the axon hits the basal lamina and is deflected in two directions. Ours is an additional interpretation in which one process would be the developing axon and the other process a retracting endfoot. Although we are still limited to inference of dynamic processes from observations of static preparations, the use of a neuron-specific antibody has allowed us to describe RGC development starting from a very early phase of differentiation. The results provide convincing evidence that RGC reach their vitreal position by reading cues they receive while spanning the retina with an endfoot at each side. This then opens the possibility that in addition to cues that could be read by a freely migrating neuron, using radial cell surfaces as guides (Rakic, 1972, 1981; Hatten, 1990), the primitive RGC could also use cues available through its two endfeet. These contacts may allow the cell to distinguish the vitreal from the ventricular surface and, thereby, tether the cell body to the vitreal endfoot that remains attached during axon generation and soma translocation. The type of cues used by the primitive RGC for its eventual vitreal location of the soma and axons are unknown. They could include surface molecules on neighboring radial cells as well as components adjacent to its own plasma membranes at the marginal surfaces. Because of its radial neuroepithelial shape, an RGC may be able to detect differences between vitreal and ventricular margins and preferentially detach from the ventricular side while retaining attachments on the vitreal surface. The authors thank Denice Major and Catherine Doller for their excellent technical assistance, Dr. James R. Unnerstall for the image-analysis, Dr. Jeremy Tuttle for alerting us to the properties of antibody TuJ1, and Dr. Anthony Frankfurter for providing us with antibodies to TuJ 1. This study was supported by NIH grant No. EY06 107 (U.R., principal investigator), NIH grant No. EY05952 (J.S., principal investigator), and grant No. B1934 from the Eppley Foundation for Research (M.W., principal investigator).
REFERENCES ADAMS,J. C. ( 1981). Heavy metal intensification of DAB-based HRP reaction product. J . Histochem. Cytochem. 29(6):775. FRELINGER, A. L. and RUTISHAUSER, U. (1986). Topography of N-CAM structural and functional deter-
95
minants. 11. Placement of monoclonal antibody epitopes. J. Cell Biol. 1031729-1737. GOSLIN,K., SCHREYER, D. J., SKENE,J. H. P., and BANKER, G. ( 1990). Changes in the distribution of GAP-43 during the development of neuronal polarity. J. Neurosci. 10(2):588-602. HALFTER,W. and CHEN,S. F. (1987). Immunohistochemical localization of laminin, neural cell adhesion molecule, collagen type IV and T-61 antigen in the embryonic retina of the Japanese quail by in vivo injection of antibodies. Ceil Tissue Rex 249:487-496. HALFTER,W., DEN, S., and SCHWARTZ, U. (1985). The formation of the axonal pattern in the embryonic avian retina. J. Comp. Neurol. 232:466-480. HAMBURGER, V. and HAMILTON, H. L. ( 195 1). A series of normal stages in the development of the chick embryo. J. Morphol. 88:49-92. HATTEN,M. E. (1990). Riding the glial monorail: a common mechanism for glial-guided neuronal migration in different regions of the developing mammalian brain. Trends Neurosci. 13( 5):179-184. HINDS,J. W. and HINDS,P. L. (1974). Early ganglion cell differentiation in the mouse retina: An electron microscopic analysis utilizing serial sections. Dev. Biol. 37:38 1-4 16. KAHN,A. J. (1973). Ganglion cell formation in the chick neural retina. Bruin Rex 63:285-290. KAHN,A. J. (1974). An autoradiographic analysis of the time of appearance of neurons in the developing chick neural retina. Dev. B i d . 38:30-40. MASLIM:J . , W E B S ~ E R M, . , and STONE,J . ( 1986). Stages in the structural differentiation of retinal ganglion cells. J. Comp. Neurol. 254:382-402. MCCLEAN.I. W. and NAKANE,P. K. (1974). Periodate-lysine-paraformaldehyde fixative: A new fixative for immunoelectron microscopy. J. Histochem. Cytochern. 22( 12):1077-1083. MCDONALD.K. ( 1984). Osmium femcyanide fixation improves microfilament preservation and membrane visualization in a variety of animal cell types. J. Ultrastruct. Res. 86: 107-1 18. MCLOON,S. C. and BARNES,R. B. (1989). Early differentiation of retinal ganglion cells: an axonal protein expressed by premigratory and migratory retinal ganglion cells. J. Neurosci. 9 ( 4): 1424-1432. MOODY,S. A., QUGG, M. S., and FRANKFURTER, A. ( 1989). Development of the peripheral trigeminal system in the chick revealed by an isotype-specific anti-beta-tubulin monoclonal antibody. J. Comp. Neurol. 279567-580. MOREST,D. K. ( 1970). The pattern of neurogenesis in the retina of the rat. Z.Anat. Entwick1.-Gesch. 131:45-67. NISHIMURA, Y. ( 1980). Determination of the developmental pattern of retinal ganglion cells in chick embryos by Golgi impregnation and other methods. Anat. Ernbryol. 158:329-347. PRADA,C., PUELLES,L., and GENIS-GALVEZ, J . M.
96
Watanabe et al.
( 1981 ). A Golgi study on the early sequence of differ-
entiation of ganglion cells in the chick embryo retina. Anat. Embryol. 161:305-3 17. RACER, G. H. ( 1980). Development of the retinoteclal projection in the chicken. Adv. Anal. Embryol. Cell Biol. 63: 1-83. RACER, G., LAUSMANN, S., and GALLYAS. F. (1979). An improved silver stain for developing nervous tissue. Stain Techol. 54(4):193-200. RAKIC, P. ( 1972). Mode of cell migration to the superficial layers of fetal monkey neocortex. J . Comp Neurol. 1 4 5 6 1-84. RAKIC, P. ( 198 1 ) . Neuronal-glial interaction during brain development. Trends Neurosci. 4: 184- 187.
RAMONY CAJAL,S. (1960). Studies on Vertebrate Neurogenesis, Springfield: Ill: Charles C. Thomas, pp. 72. SILVER,J. and ROBB, R. M. (1979). Studies on the development of the eye cup and optic nerve in normal mice and in mutants with congenital optic nerve aplasia. Dev. Biol. 68: 175- 190. SILVER,J. and R u r I S H A u s E R , U. (1984). Guidance of optic axons in vivo by a preformed adhesive pathway on neuroepithelial endfeet. Dev. Biol. 106:485-499. WATANABE, M., FRELINGER, A. L., and RUTISHAUSER, U. (1986). Topography of N-CAM structural and functional determinants. 1. Classification of monoclonal antibody epitopes. J. Cell Biol. 103: 1721-1727.
SUMMARY The early development of retinal ganglion cell and the optic fiber layers has been studied by examining the morphology of differentiating retinal ganglion cells using immunoelectron microscopy and a monoclonal antibody against neuron-specific beta-tubulin. This antibody identified retinal ganglion cells during the stages of their most active differentiation and axonogenesis prior to maturation of other retinal neurons. The changing morphology of retinal ganglion cells during these early
stages is consistent with a differentiation sequence in which axonogenesis and translocation of the cell body to the vitreal surface occur while the cell is still attached to the vitreal margin through its vitreal endfeet. Thus, the mechanism of retinal ganglion cell axon generation and soma migration to the vitreal surface appears to involve maintenance of this attachment which may act as both a focus for axon differentiation and an anchor for directed nuclear translocation to the vitreal margin.
INTRODUCTION
reconstructions (Hinds and Hinds, 1974). It is established that an RGC precursor, with a neuroepithelial cell morphology indistinguishable from neighboring cells, undergoes mitosis at the ventricular surface (Ramon y Cajal, 1960). However, questions still exist regarding the sequence of morphological changes that occur as a differentiating neuron moves toward the vitreal margin, how it is guided there, and how its rounded cell body stabilizes and extends an axon specifically and with stereotypic directionality and position along this margin. At least two sequences of RGC differentiation have been proposed. In one model (Ramon y Cajal, 1960), RGC begin differentiation at the ventricular surface where they have undergone their last mitosis and enter into a freely migrating bipolar stage where the vitreally projecting process acts as a growth cone. The process, apparently guided to the vitreal surface by the radial cells, takes a right angle turn at the vitreal margin, as it is deflected by obstacles (endfeet, axons, the limiting membrane), and extends as an axon. In this scheme, the attachment of RGC endfeet at the vitreal margin does not serve to orient the RGC soma
The neural retina begins as a simple pseudo-stratified neuroepithelium, and thereafter differentiates into an orderly stratified structure with alternating cell body and neurite domains. Ganglion cells are the first neurons to reach their final position in the neural retina and together with their processes comprise the first detectable layers, a row of cell bodies with associated axons coursing along the vitreal surface and overlying the somata. The morphological sequence of retinal ganglion cell (RGC) development in a variety of species has been described based on the interpretation of Golgi-impregnated material (Morest, 1970; Nishimura, 1980; Prada, Puelles, and Genis-Galvez, 1981 ), horse radish peroxidase (HRP) back-filling (Maslim, Webster, and Stone, 1986), electron microscope (EM) sections (Rager, 1980), and EM serial Received June 28, 1990; accepted August 22, 1990 Journal of Neurobiology, Vo1. 22, No. I , pp. 85-96 (1991) 0 1991 John Wiley & Sons, Inc. CCC 0022-3034/91/0 10085-I2$04.00 * To whom correspondence should be addressed.
85
86
Watanabe el ul.
or axon. In contrast, Morest ( 1970) proposed that RGC differentiation progressed after the cell had regained its neuroepithelial morphology and, following commitment to a neuronal fate, generates the axon at the correct vitreal level directly from the vitreal endfoot. Endfeet all along the route would serve as a preferred pathway for growth out of the retina (Silver and Robb, 1979; Rager, 1980; Silver and Rutishauser, 1984). In this sequence, the endfoot plays an essential role by serving as an anchor for nuclear translocation and guide during axonogenesis. Clarification of the sequence of RGC morphology is a prerequisite for evaluating mechanisms proposed to govern orderly retinal histogenesis. Furthermore, an understanding of normal development can serve as a basis for uncovering mechanisms that produce retinal abnormalities. In this study, the details of RGC differentiation were examined by immunoelectron microscopy using a monoclonal antibody (TuJl ) directed to an isoform of beta-tubdin. Because the isoform is detectable in neurons as early as terminal mitosis (Moody, Quigg, and Frankfurter, 19891, the antibody allowed detection of RGC and their processes during the earliest stages of their differentiation. These methods have yielded evidence in support of the hypothesis that vitreal endfeet play a role in the establishment of RGC and axon position at the inner margin of the retina.
ImmunoPeroxidase Staining of Intact Retina Tissue The primary antibodies used were ( 1 ) purified IgC of monoclonal antibodies SE directed to embryonic chicken brain neural cell adhesion molecule NCAM (Frelinger and Rutishauser, I 986; Watanabe, Frelinger, and Rutishauser, 1986) and ( 2 ) TuJ1, a monoclonal antibody directed to the neuron-specific isotype of betatubulin (Moody et al., 1989). The monoclonal antibody TuJ 1 was generously provided by Dr. Frankfurter (University of Virginia, Charlottesville) . Tissues were incubated overnight in primary antibodies, washed three times for 1 h each in HBSS, incubated overnight in secondary antibodies conjugated to peroxidase, washed extensively in HBSS, and reacted with the peroxidase-substrate diaminobenzidine (D.4B) in the presence of metal ions ( A d a m , 1981 ).
Electron Microscopy Fixed and immunostained tissues were subsequently prepared for electron microscopy by a procedure developed by McDonald ( 1984). The tissues were embedded
METHODS Tissue Preparation Neural retina of white Leghorn chicken embryos (Gullus gulfus) were staged (Hamburger and Hamilton, 195 1 ) and dissected in Dulbecco’s phosphate-buffered saline (PBS) or HBSS (Hanks Balanced Salt Solutions, GIBCO). The eyes of stage 17-19 embryonic day three (E3) embryos were dissected and the lens and the surrounding mesenchyme removed. For embryos stage 23 (E3-4) and older, pie-shaped wedges of neural retina tissues dorsal to the optic fissure (Fig. 1 ) were dissected free from the pigmented epithelium and other tissues to allow thorough penetration of fixative. antibodies, and diaminobenzidine substrate. Most of our observations were made between E3 and E7, stages during which RGC are most active in migrating to the vitreal surface (Kahn, 1973, 1974) and, therefore, when we were more likely to capture an RGC in its early phases of differentiation. The tissues were immersed overnight (9- 15 h ) in a fixative developed by McClean and Nakane ( 1974), rinsed in phosphate buffer, and immunostained.
W Figure 1 Eyes were dissected for whole-tissue immunostaining. For embryos stage 23 (E3-4) and older, pieshaped pieces (dotted area) of neural retinas were dissected from an area dorsal to the optic disc and nerve ( O N ) . The shape allowed orientation of tissue during sectioning for light or electron microscopy. Sections perpendicular to the plane of the retina were taken along the central-peripheral axis in order to increase the chances of getting an RCC axon and its cell body in the same section. The smaller size of the tissue facilitated penetration of fixative, antibodies, and the peroxidase substrate DAB. L = lens; ON = optic nerve.
Formation ofRetina1 Layers in low-viscosity plastic (Spurrs: Electron Microscopy Supplies) and sectioned for both light (2+m sections) and electron microscopy. The tannic acid treatment in McDonald’s ( 1984) procedure was eliminated in some preparations to prevent intense counterstaining of basement membranes.
lmmunofluorescence Staining of Frozen Sections Tissues were fixed and cryoprotected for frozen-sectioning as described by McClean and Nakane ( 1974). Sections ( 10 p ) were collected on gelatin-coated glass slides and stained by the indirect immunostaining procedure using the biotinylated second antibody (Vector) followed by streptavidin conjugated to Texas Red (Amersham). Photographs were taken with KODAK TMAX film (ASA 400) through an epifluorescence microscope (Nikon, Optiphot).
Determination of Antibody Penetration Fluorescently labelled goat anti-mouse IgG penetrated neural retina tissues from E3 to E 1 1 (stage 1 9 /20-stage 36/37) embryos under conditions used for immunostaining with neuron-specific TuJ 1 and anti-NCAM (neural cell adhesion molecule) monoclonal antibody (mAB) 5E both of which are mouse IgGs. The staining of the E3-E8 neural retinas was of equal intensity throughout the width of the tissues, indicating even penetration of the labelled antibody. whereas the staining of El 1 neural retina was more intense at the vitreal and ventricular edges than in the center of the tissue, indicating resistence to antibody penetration. Goat anti-mouse IgG antibodies conjugated to fluorescein were incubated with E3-El I central neural retina pieces overnight as for the incubation of first antibody during the normal staining procedure. These retina pieces varied in thickness and in the maturation of vitreal and ventricular junctions, which may have affected penetration of the antibodics. Despite these variations with age, all retina pieces had fluorescent antibody-penetration to the center of the tissue. The intensity of fluorescence staining for E l 1 retinas was not as even across the tissue as for the younger tissues. Therefore, preferential staining of the vitreal edge is not caused by limited first antibody penetration for retinas up to E8. Further evidence exists for even penetration of immunological reagents into neural retinas. Other monoclonal antibodies of the IgG sublype drected to NCAM evenly stain the width of E4-E6 central neural retina pieces which are thicker and have more mature junctions. Bipolar, neuroepithelial cells that stain with TuJ 1 by a whole-mount technique do so evenly throughout the span of the retina from vitreal to ventricular side. Penetration of antibodies into neural retina explants
87
have been demonstrated for Fabs and antibodies (Silver and Rutishauser, 1984; Halfter and Chen, 1987). Furthermore, neural retinas stained with T d l , after the width of the tissue was exposed by frozen-sectioning, had identical staining patterns to that of whole tissues.
RESULTS The immunological method used in this study for staining RGC has advantages over previously used techniques. One advantage is that immature RGC can be identified before they have migrated to the vitreal surface. Another advantage is that the majority of postmitotic RGC rather than a subset are stained with this technique. In addition, with the ability to see the immunostain and ultrastructure at the same time, even parts of RGC can be detected. Immunoelectron microscopy with the TuJl antibody also allows identification of fine or immature RGC processes that are not well stained by reduced silver methods (Rager, Lausmann, and Gallyas, 1979). At the ultrastructural level, it is possible to determine in some cases, that a particular process projects from a particular cell body rather than just traveling adjacent to it. The relative extent of RGC maturation can also be inferred by comparing the intensity of TuJ 1-staining, which increases with differentiation.
Identification of Chicken Embryo Retinal Ganglion Cells by lmmunostaining with TuJl Cells were clearly visible in whole mounts of E4-E7 retina after immunoperoxidase staining for neuron-specific beta-tubulin ( TuJ 1 ). They were, in addition, identified as RGC by the vitreal position of their soma and axons. Flattened wedges of neural retina taken from the posterior pole of the eye (Fig. I ) immunostained with TuJl displayed a gradient of staining [Fig. 2( a ) ] , which was intense at the central region and gradually faded towards the periphery, reflecting the well-known gradient of maturation of the developing retina (Halfter et al., 1985: Kahn, 1973, 1974; Rager, 1980: Ramon y Cajal, 1960). The centrally located, intensely stained cell bodies could be focused at the same level [Fig. 2(b)], whereas the peripheral, more lightly stained cell bodies had to be focused at various levels within the neural retina [Fig. 2 ( c ) J . These results are due to the movement of RGC toward the vitreal side, the increase in their inten-
88
Wutunabe et al.
Figure 2 TuJ 1-stained cell bodies and axons of RGC. In stage 25 (E4.5) neural retina, it was possible to observe at the light microscope level a gradual increase in number of the cells stained with T d 1 and intensity of stain per cell from the peripheral to central retina ( a ) . Single axons from RGC bodies interacted with other axons soon after they protruded from the cell body and traveled with other axons (b,c). When immunoperoxidase-stainedneural retina pieces embedded in resin were sectioned perpendicular to the plane of the retina (d,e), it was revealed that most of the intensely stained cell bodies and axons were located along the vitreal margin. In these sections, it was possible to identify RGC with a length of its axon protruding from the cell body (d). More often axons were sectioned in cross section and appeared as black dots (e). (RGC cell bodies are 10-1 5 hm in diameter.)
89
Formation of Retinal Layers
-of I
-in1
* Ph
Figure 3 TuJl specifically stained RGC up to EX (st. 34) neural retinas. Intensely immunostained cells could be detected only at the vitreal surfaces of retinas from E6 (a) and E8 ( b ) animals. This is the location of RGC that are the first to differentiate and arrive at their final position in the retina. By E 1 1 (c), TuJ 1 also stained photoreceptor cells (ph) and cell bodies adjacent to the inner plexiform layer (ipl), which may belong to amacrine or bipolar cells. The arrowhead points to the vitreal surface. ofl = optic fiber layer; in1 = inner nuclear layer; ph = photoreceptors. Scale bar = 100 hm.
sity of TuJ 1-immunostaining with maturation, and the more peripheral and less mature cells differentiating at different levels. The TuJ l-immunostained cells were detected as early as stage 14, which is when the first RGC have been detected by silver impregnation (Halfter et al., 1985). The antigen recognized by TuJ 1 accumulated in both cell bodies and axons during differentiation. When the immunostained tissue wedges were sectioned perpendicular to the plane of the neural retina, it was revealed that many of the intensely stained cells and their axons were located at the vitreal surface [Fig. 2 (d,e)] . TuJ 1 specifically stained RGC in the vitreal margin up to and including E8 (Fig. 3 ) . By E 1 1 [ Fig. 3 (c)] the antigen was no longer specific for RGC and appeared in photoreceptor cells as well as cell bodies located in the inner nuclear layer adjacent to the inner plexiform layer. Therefore, during E2-E6, those stages examined in detail in this study, Tul 1 could be used to identify RGC.
bound to the cell bodies and axons of RGC that had migrated to the vitreal surface of the retina. When TuJ 1-stained cell bodies were counted in E4 and E6 retina (Table 1 ), 96%-97% were located in the vitreal third. The rest (3%-4%) were located in the ventricular two-thirds of the retina. Even at the light microscope level it was possible to detect TuJl-stained cell bodies within the neuroepithelium that were still attached to vitreal
Neuroepithelial Morphology TuJl bound to cell bodies as well as vitreal and ventricular processes in primitive RGC. It also
lmmunostained tissues were embedded in plastic and 2-pm serial sections were collected. Stained cells in every fifth section were counted to reduce the probability of counting the same cell twice (RGC somata are 10-1 5 p m in diameter).
Table 1 TuJ1-Stained Cell Bodies in E4 and E6 Retina
E4 (st. 25/26)
E6 (st. 28/29)
No. of Stained Cells
%
Region
533 15 8 1030 26 1
96 2.4 1.4 97 2.5 1
Vitreal third Middle third Ventricular third Vitreal third Middle third Ventricular third
90
Wutanuhe et al.
Figure 4 TuJ 1-stained primitive-appearing RGC in the process of nuclear translocation. In addition to staining cells that had the characteristic morphology of differentiating RGC, that is, with rounded cell bodies and axons at the vitreal surface, TuJ 1 also stained cells with the morphology of primitive neurons. A 2-pm plastic section of an immunoperoxidase-stained E4 neural retina revealed that a TuJ 1-positive RGC can have a soma in the middle of the retina and an endfoot bound to the vitreal limiting membrane (a). This particular endfoot appears to contain many vesicles when observed under phase and Nomarski optics. A computer-enhanced mirror-image (image analysis) of the same cell emphasizes the continuity of the process between the cell body and the endfoot ( b ) . Arrowheads point to the vitreal surface. RGC with primitive morphologies were also detected in 10-pm frozen sections of embryonic retina that were stained for fluorescence microscopy using the sensitive biotin/ streptavidin/ Texas Red procedure. One of these RGC extends a process almost halfway across the neural
Formalion of Retinal Layers
and/or ventricular processes (Fig. 4j. Because the antigen accumulated in neuroepithelial-shaped cells, it was possible in certain sections to follow the cytoplasmic processes extending from cell bodies to endfeet at the tissue limits. Presumptive transitional RGC with cell bodies more than three somatic diameters (30-50 pm) away from the vitreal surface were present in neural retinas of embryos as old as E8 and E 12 in both the central and peripheral regions. These often had long trailing ventricular processes. Faint staining was also apparent in cells with cell bodies located in the ventricular side of the retina adjacent to the pigmented epithelium. Vitreal Endfeet and Axonogenesis
In order to determine whether axons were emerging from these primitive-appearing RGC, it was necessary for us to examine the tissue at the electron microscopic level. Embryonic eyes and retinal tissues immunostained before embedding had intact vitreal basement membranes, plasma membranes of cell bodies, endfeet, and axons, nuclear membranes, and mitochondria1 membranes. Specialized junctions could be recognized including the apical junctions of the vitreal side and zonula adherens in the ventricular region. Even in retina pieces that were not immunostained, cells could be identified at the electron microscopic level which had the characteristics of retinal ganglion cells, that is, cells with rounded bodies adjacent to the vitreal surface. Some of these had thick vitreal attachments to the basement membrane (Fig. 5 j . TuJl stained the vitreal endfeet, of which some had adjoining processes of the caliber and immunostaining intensity of axons (Fig. 6 ) . Several TuJ 1-negative endfeet had TuJ 1-positive neurites protruding from them. Clear vacuoles were often observed in endfeet with adjoining TuJl positive neurites.
91
Stained cell bodies within the marginal zone sometimes possessed two processes protruding towards the vitreal surface, with one process more intensely TuJ 1-positive than the other. Those RGC with cell bodies at the vitreal surface possessed axon-like processes that were intensely stained with TuJ 1. DISCUSSION
The morphology of chicken retinal ganglion cells was observed during early phases of differentiation in order to deduce the cellular mechanics that generate the characteristic positioning of RGC soma and axons at the vitreal surface. We used antibody TuJl to identify retinal neurons at a stage when RGC, exclusive of other neurons, were undergoing terminal mitosis and differentiation. Retinal ganglion cells were identified during the early phases of neurogenesis using this antibody which binds to a neuron-specific isoform of beta-tubulin. This protein is found in the soma and most processes of neurons. The antibody T d l is particularly useful for the study of neurogenesis because it recognizes the protein when it is expressed during or shortly after terminal mitosis (Moody et al., 1989). With this antibody it was possible to identify and mark the retinal ganglion cells at a primitive state of differentiation. In retinal tissues immunostained with TuJl, cells with somata at the ventricular side of the retina and displaying a primitive neuroepithelial morphology were stained throughout their soma and both vitreal and ventricular processes. These were identified as primitive RGC presumed to be at a stage before or during the process of nuclear migration. TuJ 1-positive cells of this primitive morphology were detected in E3-El2 embryos in both central and peripheral neural retina. Previous identifications of primitive RGC have been based
retina (c) which may be part of a vitreal process or a trailing ventricular process. Another cell ( d ) spans the neural retina, with its cell body in the middle of the neural retina, its ventricular process ending in an endfoot, and its vitreal process projecting towards the vitreal surface. In this photograph, the ventricular and vitreal processes appear to be stained, and more intensely so than the cell soma (small arrow), which is out of the plane of focus. In the peripheral E6 ( st. 29) neural retina, cell bodies and axons located at the vitreal margin (arrow) were intensely stained (c). Similar morphologies of RGC in the process of translocation with vitreal and/or ventricular processes could be detected in E6 peripheral retina ( e) TuJ 1 -immunofluorescently stained cryostat sections. The periphery of the tissue is to the right and the central region is to the left of each photograph. The arrows point to the vitreal surface. Scale bar = 50 Fm.
92
Watanabe ef a/.
Figure 5 Cells with their nuclei at the vitreal surface were attached to this surface by basal processes or endfeet. These cells had different cytoplasmic consistencies from neighboring cells. The cytoplasm of these cells appeared less dense than neighboring cells (a-d) but some were more dense (e). The arrowheads point to the vitreal surface. Scale bars = 1 Km.
on other criteria. For example, Prada et al. ( 1981 ) identified early RGC as radial cells without ventricular endfeet. Hinds and Hinds ( 1974) used EM serial sections to identify RGC based on migration of centrioles and cilia from the ventricular to vitreal side of nuclei within elongated neuroepithelial cells. According to Hinds and Hinds ( 1974) these did not have vitreal endfeet. Maslim et al. ( 1986) used HRP backfill to locate RGC and observed RGC each with an axon and both vitreal and ventricular processes. They also observed radial cells presumably backfilled with HRP through their axons with processes connecting them both to vitreal and ventricular margins of the retina. However, these labelled cells were close to the site of contact of HRP crystals and could, therefore, have been stained by direct contact with crystals rather than by axonal transport. A cytoplasmic protein RA4 has been identified in developing chicken retina that is selectively expressed in the optic fiber
layer in later stages of development ( McLoon and Barnes, 1989). At earlier stages antibodies to this antigen stain cells similar to those described here that have a primitive, bipolar appearance. However, no analysis at the electron microscope level has been carried out to confirm the presence or absence of endfeet or axons on these RA4-positive cells. The neuron-specific immunostaining procedure used in this study provided independent criteria for identifying RGC, provided evidence that differentiating RGC can have morphologies even more primitive than previously described, and allowed ultrastructural analysis of axons and endfeet. We do not know if every TuJl-positive cell including the neuroepithelial-shaped cells are necessarily undergoing axonogenesis. This determination may only be obtained by serial-section reconstructions at the electron microscope level of many TuJ 1-immunostained cells. Axons have been proposed to arise from neuro-
Formation ofRetinal Layers
Figure 6 The cytoplasm of certain endfeet were distinctly positive with TuJl immunostaining. These RGC endfeet were attached to the basal lamina and contained mitochondria. The vitreal surface (arrowhead) is at the top of each micrograph. Some samples (a,b) were simultaneously stained with anti-NCAM mAB 5E,which stains plasma membranes thus delineating the outline of the cells. Processes projected from TuJ 1-positive endfeet (b,c). The projection in ( b ) is similar in diameter and staining intensity to an axon in cross section (white *). T d 1-positive processes in ( c ) include transverse sections through axons and a fan-shaped region that may be a portion of a growth cone. The plasma membrane of one axon is continuous with an endfoot filled with vesicles (star). The arrowhead points to the vitreal surface. An RGC with a cell body already translocated to the vitreal margin possesses two processes (d). The right process, which is intensely TuJ 1-stained, ends in a club-like structure that resembles the base of a growth cone. The other process on the left is shorter and a portion of it is unstained with TuJ1. The end of this process is indicated by an asterisk (*). This cytoplasmic process was followed through serial sections (not shown) and extended no further than shown in this figure. The arrowhead indicates the vitreal surface. Scale bar = 1 pm.
93
94
Watanabe et al.
epithelial cells by different mechanisms (Fig. 7): ( 1 ) by lateral growth of specialized processes from the side of vitreal endfeet (Morest, 1970; Nishimura, 1980; Prada et al., 1981) or soma (Maslim et al., 1986): ( 2 ) as transformations of the vitreal endfeet which detach from the vitreal surface; or ( 3 ) as transformations of vitreally directed processes that do not reestablish stable contact with the basal lamina but are deflected by it (Ramon y Cajal, 1960; Hinds and Hinds, 1974; McLoon and Barnes, 1989). Our results support the idea that in the embryonic chicken retina, axons grow out of endfeet. One line of evidence is that elongated processes with the proper caliber for axons and positive for the neuron-specific antigen TuJ 1 were detected protruding perpendicularly from intact vitreal endfeet. We also recognized a number of
end foot
4
1
2
.,.J..&. *y.. . ..,a.
3
.-..,.....
4
.-::.:
.I. I . ’ : . .
A:.:;.
$6” 1
2
3
4
vent v- i_t ..I:-’;
vent
Figure 7 Proposed sequences for retinal ganglion cell differentiation. Sequence A: A cell undergoes mitosis at the ventricular surface ( 1), reextends to span the neuroepithelium ( 2 ) , sends an axon from the vitreal endfoot while retracting the ventricular endfoot and translocating the nucleus ( 3 ) , and attains the final vitreal position with an axon growing along the margin and cell body stabilized at the vitreal region ( 4 ) . Sequence B: A cell undergoes mitosis at the ventricular surface ( 1 ) , sends a process towards the vitreal surface (2), attains a freely migrating bipolar stage during which the cell body, led by a vitreal process, migrates to the vitreal surface where the process makes a right angle turn to project along the margin, transforms into an axon ( 3 ) , and reaches the vitreal surface (4).
retinal ganglion-like cells with rounded bodies at the vitreal surface still connected to the vitreal basal lamina by thick stalks of cytoplasm. We suggest that these may be RGC that have translocated nuclei but have not eliminated their endfeet. Thus, findings from this study support the proposal that RGC send out axons from their vitreal endfeet parallel to the basal lamina (Morest, 1970). The vitreal endfoot can be retained during the early part of axonogenesis and is conjectured to be a stabilizing structure for nuclear translocation. However, axonogenesis, release of the ventricular endfeet, and cell body translocation must occur rapidly. This is suggested by the small number of TuJ 1-positive cells with radial morphology and with soma within the middle and ventricular third of the neural retina even during the most active time of RGC differentiation. The small percentages of the “transitional cells” are consistent with the frequency of similar cells detected by Hinds and Hinds (1974) using EM reconstructions of serial sections as well as the observations by Morest ( 1970), who also concluded that translocation must be a rapid process. Eventually, the endfeet of RGC are eliminated. Observations from this study would be consistent with either their retraction into the RGC soma or elimination by autolytic disintegration. We have observed endfeet devoid of T d l staining that are filled with vesicles from which a TuJ 1-positive neurite protruded. One interpretation for these observations is that the endfoot degenerates soon after axonogenesis and in the process loses the protein recognized by anti-TuJ 1, whereas the axon retains it. Such redistributions in the subcellular localization of proteins within differentiating neurons have been noted for such axon-associated proteins as GAP-43 ( G o s h , Schreyer, Skene, and Banker, 1990) and RA4 (McLoon and Barnes, 1989). The numerous vesicles within the endfeet may be lysosomes containing autolytic enzymes. Lysosomes have also been detected by Hinds and Hinds ( 1974). In addition, they observed pinching off of cytoplasm by neuroepithelial cells during elimination of ventricular processes. which could be another mechanism used for vitreal endfeet elimination. We observed TuJ 1-stained cell bodies possessing two processes protruding towards the vitreal surface with one process more intensely stained than the other. Maslim et al. (1986) also noticed cells with two processes, one being an endfoot and another an elongated axon. Ramon y Cajal ( 1960) proposed that bifurcation yielding
Formation of Retinal Layers
two processes may arise when the axon hits the basal lamina and is deflected in two directions. Ours is an additional interpretation in which one process would be the developing axon and the other process a retracting endfoot. Although we are still limited to inference of dynamic processes from observations of static preparations, the use of a neuron-specific antibody has allowed us to describe RGC development starting from a very early phase of differentiation. The results provide convincing evidence that RGC reach their vitreal position by reading cues they receive while spanning the retina with an endfoot at each side. This then opens the possibility that in addition to cues that could be read by a freely migrating neuron, using radial cell surfaces as guides (Rakic, 1972, 1981; Hatten, 1990), the primitive RGC could also use cues available through its two endfeet. These contacts may allow the cell to distinguish the vitreal from the ventricular surface and, thereby, tether the cell body to the vitreal endfoot that remains attached during axon generation and soma translocation. The type of cues used by the primitive RGC for its eventual vitreal location of the soma and axons are unknown. They could include surface molecules on neighboring radial cells as well as components adjacent to its own plasma membranes at the marginal surfaces. Because of its radial neuroepithelial shape, an RGC may be able to detect differences between vitreal and ventricular margins and preferentially detach from the ventricular side while retaining attachments on the vitreal surface. The authors thank Denice Major and Catherine Doller for their excellent technical assistance, Dr. James R. Unnerstall for the image-analysis, Dr. Jeremy Tuttle for alerting us to the properties of antibody TuJ1, and Dr. Anthony Frankfurter for providing us with antibodies to TuJ 1. This study was supported by NIH grant No. EY06 107 (U.R., principal investigator), NIH grant No. EY05952 (J.S., principal investigator), and grant No. B1934 from the Eppley Foundation for Research (M.W., principal investigator).
REFERENCES ADAMS,J. C. ( 1981). Heavy metal intensification of DAB-based HRP reaction product. J . Histochem. Cytochem. 29(6):775. FRELINGER, A. L. and RUTISHAUSER, U. (1986). Topography of N-CAM structural and functional deter-
95
minants. 11. Placement of monoclonal antibody epitopes. J. Cell Biol. 1031729-1737. GOSLIN,K., SCHREYER, D. J., SKENE,J. H. P., and BANKER, G. ( 1990). Changes in the distribution of GAP-43 during the development of neuronal polarity. J. Neurosci. 10(2):588-602. HALFTER,W. and CHEN,S. F. (1987). Immunohistochemical localization of laminin, neural cell adhesion molecule, collagen type IV and T-61 antigen in the embryonic retina of the Japanese quail by in vivo injection of antibodies. Ceil Tissue Rex 249:487-496. HALFTER,W., DEN, S., and SCHWARTZ, U. (1985). The formation of the axonal pattern in the embryonic avian retina. J. Comp. Neurol. 232:466-480. HAMBURGER, V. and HAMILTON, H. L. ( 195 1). A series of normal stages in the development of the chick embryo. J. Morphol. 88:49-92. HATTEN,M. E. (1990). Riding the glial monorail: a common mechanism for glial-guided neuronal migration in different regions of the developing mammalian brain. Trends Neurosci. 13( 5):179-184. HINDS,J. W. and HINDS,P. L. (1974). Early ganglion cell differentiation in the mouse retina: An electron microscopic analysis utilizing serial sections. Dev. Biol. 37:38 1-4 16. KAHN,A. J. (1973). Ganglion cell formation in the chick neural retina. Bruin Rex 63:285-290. KAHN,A. J. (1974). An autoradiographic analysis of the time of appearance of neurons in the developing chick neural retina. Dev. B i d . 38:30-40. MASLIM:J . , W E B S ~ E R M, . , and STONE,J . ( 1986). Stages in the structural differentiation of retinal ganglion cells. J. Comp. Neurol. 254:382-402. MCCLEAN.I. W. and NAKANE,P. K. (1974). Periodate-lysine-paraformaldehyde fixative: A new fixative for immunoelectron microscopy. J. Histochem. Cytochern. 22( 12):1077-1083. MCDONALD.K. ( 1984). Osmium femcyanide fixation improves microfilament preservation and membrane visualization in a variety of animal cell types. J. Ultrastruct. Res. 86: 107-1 18. MCLOON,S. C. and BARNES,R. B. (1989). Early differentiation of retinal ganglion cells: an axonal protein expressed by premigratory and migratory retinal ganglion cells. J. Neurosci. 9 ( 4): 1424-1432. MOODY,S. A., QUGG, M. S., and FRANKFURTER, A. ( 1989). Development of the peripheral trigeminal system in the chick revealed by an isotype-specific anti-beta-tubulin monoclonal antibody. J. Comp. Neurol. 279567-580. MOREST,D. K. ( 1970). The pattern of neurogenesis in the retina of the rat. Z.Anat. Entwick1.-Gesch. 131:45-67. NISHIMURA, Y. ( 1980). Determination of the developmental pattern of retinal ganglion cells in chick embryos by Golgi impregnation and other methods. Anat. Ernbryol. 158:329-347. PRADA,C., PUELLES,L., and GENIS-GALVEZ, J . M.
96
Watanabe et al.
( 1981 ). A Golgi study on the early sequence of differ-
entiation of ganglion cells in the chick embryo retina. Anat. Embryol. 161:305-3 17. RACER, G. H. ( 1980). Development of the retinoteclal projection in the chicken. Adv. Anal. Embryol. Cell Biol. 63: 1-83. RACER, G., LAUSMANN, S., and GALLYAS. F. (1979). An improved silver stain for developing nervous tissue. Stain Techol. 54(4):193-200. RAKIC, P. ( 1972). Mode of cell migration to the superficial layers of fetal monkey neocortex. J . Comp Neurol. 1 4 5 6 1-84. RAKIC, P. ( 198 1 ) . Neuronal-glial interaction during brain development. Trends Neurosci. 4: 184- 187.
RAMONY CAJAL,S. (1960). Studies on Vertebrate Neurogenesis, Springfield: Ill: Charles C. Thomas, pp. 72. SILVER,J. and ROBB, R. M. (1979). Studies on the development of the eye cup and optic nerve in normal mice and in mutants with congenital optic nerve aplasia. Dev. Biol. 68: 175- 190. SILVER,J. and R u r I S H A u s E R , U. (1984). Guidance of optic axons in vivo by a preformed adhesive pathway on neuroepithelial endfeet. Dev. Biol. 106:485-499. WATANABE, M., FRELINGER, A. L., and RUTISHAUSER, U. (1986). Topography of N-CAM structural and functional determinants. 1. Classification of monoclonal antibody epitopes. J. Cell Biol. 103: 1721-1727.
