Morphologic and neurochemical target ... - Wiley Online Library

3 downloads 0 Views 2MB Size Report
Dec 16, 1996 - College of Optometry (D.M.S.), University of Houston, Houston, Texas, ...... to affect neurite extension by hippocampal neurons (Matt- son, 1988 ...
THE JOURNAL OF COMPARATIVE NEUROLOGY 376:476488 (1996)

Morphologic and Neurochemical Target Selectivity of Regenerating Adult Photoreceptors In Vitro DAVID M. SHERRY, ROBERT S. ST. JULES, AND ELLEN TOWNES-ANDERSON College of Optometry (D.M.S.), University of Houston, Houston, Texas, 77204-6052; Department of Neurosciences (R.S.St.J., E.T.-A.), University of Medicine and Dentistry-New Jersey Medical School, Newark, New Jersey 07103-2714

ABSTRACT Regenerating adult central nervous system (CNS) neurons must re-establish synaptic circuits in an environment very different from that present during development. However, the complexity of CNS circuitry has made it extremely difficult to assess the selectivity and mechanisms of synaptic regeneration at the cellular level in vivo. The synaptic preferences of adult photoreceptors were examined by using a defined cell culture system known to support regenerative process growth, presynaptic varicosity formation, and establishment of functional synapses. Immunolabeling for synaptic vesicle protein 2 and ultrastructural analysis demonstrated that cell-cell contacts made by photoreceptors were synaptic in nature. Target selectivity was determined by quantitative analysis of contacts onto normal and novel target cell types in cultures in which opportunities to contact all retinal cell types were present. Target cells were identified by morphology and immunolabeling for the amino acid neurotransmitters glutamate, aspartate, gamma-aminobutyric acid (GABA),and glycine. Regenerating photoreceptors showed a strong preference for novel multipolar cell targets (amacrine and ganglion cells) over normal photoreceptor, horizontal, and bipolar cell targets. Additionally, photoreceptors were selective for targets containing the transmitter GABA. These results indicate first, that the normal synaptic partners for photoreceptors are not intrinsically the optimal targets for regenerative synapse formation, and second, that GABA may modulate synaptic targeting by adult photoreceptors. o 1996 Wiley-Liss, Inc. Indexing terms: synaptogenesis, retina, GABA, amphibian, regeneration

Disruption of the complex synaptic circuitry of the central nervous system (CNS) by trauma or disease often produces devastating and permanent functional deficits. Reports that adult CNS neurons can regenerate axons and synapses under some circumstances (see review by Aguayo et al., 1990) has led to intense interest in CNS transplant strategies to overcome these deficits. In the retina, transplant therapies have been used to replace photoreceptors that are lost in degenerative disease (Sheedlo et al., 1991). However, photoreceptor transplants to date have provided little, if any, recovery of discriminative visual function. Lack of recovery seems due, at least in part, to poor synaptic integration with the host retina; transplanted photoreceptors form a limited number of synapses, even though they are apposed to the dendrites of their normal synaptic targets, the horizontal and bipolar cells (Gouras et al., 1991; Silverman et al., 1992; Mosinger Ogilvie et al., 1994). Most transplant strategies are based on the assumption that transplanted neurons will reconstitute synaptic circuits with the adult host neurons similar to those that arise

o 1996 WILEY-LISS, INC.

during development. However, synapse formation by adult neurons occurs in an environment that differs in its cellular and noncellular elements from the developmental milieu. In the adult CNS, transient cell populations-such as the cortical subplate cells, specific trophic factors, and matrix proteins associated with developmental synaptogenesisare absent (for review see Goodman and Shatz, 1993). Furthermore, constellations of membrane receptors and other cell-specific molecules in the mature CNS differ from those present in the developing CNS. Such differences could result in a different set of stimulatory and inhibitory signals regulating synapse formation during regeneration than during development. Little is known about the degree of specificity at the cellular level during synaptic regeneration (see Aguayo et al., 1990). In addition, the environmental Accepted August 20, 1996. Address reprint requests to David M. Sherry, College of Optometry, University of Houston, 4901 Calhoun, Houston, TX 77204-6052. E-mail: dsherrywmail-gw.opt.uh.edu

TARGETING OF PHOTORECEPTOR SYNAPSE REGENERATION and intrinsic cell factors regulating the specificity of synaptic regeneration are poorly understood and difficult to investigate in the complex CNS synaptic circuitry in vivo. We have examined the cellular selectivity of synaptic regeneration by adult photoreceptors by using an in vitro system of isolated retinal cells in which normal and novel retinal neuron targets were available to regenerating photoreceptors. This approach allowed investigation of target preferences in a system in which substrate and medium were defined, highlighting the intrinsic regenerative capabilities and preferences of adult neurons. Moreover, results could be subjected to the stringency of statistical analysis. Photoreceptors preferred novel multipolar cell targets (amacrine and ganglion cells) over the horizontal, bipolar, and photoreceptor targets they normally contact in vivo. Photoreceptors also preferred targets containing the neurotransmitter gamma-aminobutyric acid (GABA)over cells containing other amino acid transmitters. These results provide some explanation for the limited regenerative synaptogenesis by rod cells observed in retinas after transplantation (Gouras et al., 1991; Silverman et al., 1992; Mosinger Ogilvie et al., 1994 ) and in retinal disease (Jansen and Sanyal, 1992; Sanyal et al., 1992; Li et al., 1995). Some of these data have been presented in abstract form (Sherry and Townes-Anderson, 1994).

477

Enzyme Ringer (in mM): 85 NaCl; 1.5 KCl; 25 NaHCO,; 0.5 NaHzP04;0.9 CaC1,; 1.0 Na pyruvate; 16 glucose; 0.43 mg/ml cysteine; 14 units/ml papain. High-glucose Ringer: 10 ml normal salamander Ringer plus 0.1 ml30% glucose solution. Salamander Medium: 90 ml normal salamander Ringer; 7 ml medium 199; 1 ml 1 0 0 ~modified Eagle's medium (MEM) vitamin mix; 1 ml 200 mM glutamine; 0.2 ml 50x MEM essential amino acids; 0.1 ml 1 0 0 MEM ~ nonessential amino acids (all from Gibco BRL, Grand Island, NY);2 kg/ml insulin; 1 pg/ml transferrin; 1 kg/ml DL-thyroxine; 5 mM taurine; 10 pg/ml gentamycin; 1pg/ml bovine serum albumin (all from Sigma, St. Louis, MO).

Immunocytochemistry

Cultures were fixed in 2.5% glutaraldehyde + 2.5% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.4) at 4°C. Cultures were rinsed in phosphate-buffered saline (PBS, pH 7.41, exchanged to deionized water, and pretreated with 0.5%NaBH4 in deionized water for 1minute at room temperature to reduce autofluorescence. Pretreated cultures were rinsed in water, exchanged to PBS, and nonspecific labeling was blocked with 2% normal goat serum in PBS for 45 minutes at room temperature. Blocking agent was removed and primary antiserum was applied overnight at 4°C. Each culture was immunoreacted for one of four putative amino acid transmitters using rabbit MATERIALS AND METHODS polyclonal sera against glutaraldehyde conjugates of glutaAnimals mate (GLU), aspartate (ASP), glycine (GLY), or GABA. All experiments were conducted on neotenic tiger sala- Antisera against GLU, GABA, and ASP were purchased manders (Ambystomatigrinum; 16-23 cm length). Animals commercially (Chemicon International, Temecula, CA) and were maintained in aerated tanks at 5°C on a 12-hours-of- were diluted 1:2000, 1:1500, and 1:1000, respectively. The light: 12-hours-of-dark cycle for several weeks before use. GLY antiserum and a second GLU antiserum were the gift of Dr. Robert Marc (University of Utah, Salt Lake City, UT; Retinal dissociation and culture Marc et al., 1990; Kalloniatis and Fletcher, 1994) and were Retinal dissociation, culture methods, and solutions for used at a dilution of 1:2000. Cultures immunoreacted for adult salamander retinal neurons were as described previ- GABA or GLY were double-labeled simultaneously for ously (MacLeish et al., 1983; Townes-Anderson et al., 1985; synaptic vesicle protein 2 (SV2) by using a mouse monocloMacLeish and Townes-Anderson, 1988; Mandell et al., nal antibody diluted 1 : l O (Buckley and Kelly, 1985; the gift 1993). Light-adapted salamanders were decapitated and of Dr. Kathleen Buckley, Harvard Medical School, Boston, pithed within 1hour of the midpoint of the light portion of MA) to help visualize synaptic varicosities. Cultures were the cycle. The head was rinsed in 70% ethanol, then sterile rinsed in PBS, blocked, and secondary antiserum was Ringer's solution (Ringer), and the eyes were removed. The applied for 45 minutes at room temperature. Binding of cornea, lens, and vitreous humor were removed, and the rabbit polyclonal antisera was visualized with goat antiretina, free of the retinal pigmented epithelium, was dis- rabbit-fluorescein; binding of mouse monoclonal antibodies sected from the eyecup. Both retinas from one animal were was visualized using goat anti-mouse-rhodamine (secondincubated together in 7 ml of enzyme Ringer containing 14 ary antisera were diluted 1:30; Boeringer-Mannheim, Indiaunits/ml papain (Worthington, Freehold, NJ) and 0.43 napolis, IN). Labeled cultures were rinsed in PBS, rinsed in mgiml DL-cysteine HC1 for 5 minutes a t 37"C, then for 30 water, then covered with a fade-retardant mounting meminutes at room temperature with gentle agitation. En- dium. All antisera were diluted in 2% normal goat serum + zyme Ringer was bubbled with 5%Coz/95%02 for 15 0.1% Triton X-100 in PBS. The specificities of the anti-GLU, anti-ASP, anti-GABA, minutes before use. The retinas were rinsed in Ringer, exchanged to high-glucose Ringer, and triturated. Isolated and anti-GLY sera have been established by several methcells were plated onto glass coverslips coated with Sal-1 ods. First, preadsorption of the primary antiserum with monoclonal antibody, which recognizes a plasma mem- appropriate or inappropriate conjugates (Marc et al., 1990; brane epitope present on all salamander retinal cell types, Sherry and Ulshafer, 1992a,b; Kalloniatis and Fletcher, allowing cells to adhere (MacLeish et al., 1983; the gift of 1994). Second, a postembedding specificity test with resinDr. Peter MacLeish, Morehouse School of Medicine, At- embedded amino acid-glutaraldehyde-bovineserum albulanta, GA). Cells were maintained in 2 ml of serum-free min conjugates of GLU, ASP, GABA, and taurine (Sherry salamander medium at 10°C. Culture medium was replaced and Ulshafer, 1992a,b). Third, immunoblot tests of crossreactivity with amino acid conjugates (Marc et al., 1990; weekly. Culture solutions were as follows. Salamander Ringer (in Kalloniatis and Fletcher, 1994). All amino acid antisera mM): 108 NaC1; 2.5 KC1; 1.0 NaHCO,; 0.5 NaH2P04;0.5 show excellent specificity for the appropriate conjugate. MgS04; 0.9 CaCl,; 1.0 Na pyruvate; 24 glucose; 2 HEPES; The specificity of the mouse monoclonal antibody directed against SV2 has been described previously (Buckley and pH adjusted to 7.7 with NaOH.

D.M. SHERRY ET AL.

478 Kelly, 1985). The specificity of the immunocytochemical method was tested in the following two ways: substitution of normal rabbit serum for primary antisera raised in rabbit, and omission of primary antibody or antiserum. Both treatments eliminated immunolabeling, indicating that labeling was specific.

Electron microscopy Preparation of cultured cells for electron microscopy was as described previously (Mandell et al., 1993). Briefly, cultures were fixed at 2 weeks postisolation in 2.5% parafor2.5% glutaraldehyde + 0.05% picric acid in maldehyde 0.1M sodium cacodylate (pH 7.4) overnight at 4°C. Cultures were postfixed in osmium and potassium ferrocyanide, stained en bloc with uranyl acetate, dehydrated,and embedded in epoxy resin. Photoreceptors making varicosity contacts were identified by visual inspection, excised, and remounted on resin stubs. Serial sections were collected onto thin bar grids, stained with lead citrate and uranyl acetate, and examined in a JEOL 1OOCX-I1microscope.

+

Statistical analysis Statistical analyses were performed using Statview 4.02 software (Abacus Concepts, Berkeley, CAI. Analytical procedures are described in the Results section.

RESULTS The synaptology and amino acid neurochemistry of the adult amphibian retina are well established (c.f., Dowling, 1968; Werblin and Dowling, 1969; Lasansky, 1973, 1978; Wong-Riley, 1974; Wu, 1986, 1992; Yang and Yazulla, 1988a,b, 1994a,b;Yazulla and Yang, 1988). Photoreceptors, the focus of this study, use the neurotransmitter GLU and synapse onto GABAergic horizontal cells, bipolar cells that are glutamatergic although some small subpopulations also contain GABA, GLY, or ASP and other photoreceptors. Photoreceptors do not synapse with amacrine or ganglion cells. Amacrine cells are predominantly GABAergic or glycinergic although some subpopulations also may contain GLU or ASP. Ganglion cells are probably glutamatergic although some subpopulations also may contain ASP.

Identification of retinal cell types and regenerative growth patterns The unique morphologic traits that characterize the major cell types in the intact retina are retained by isolated salamander retinal neurons (Fig. 11, allowing unequivocal identification of four cell types-photoreceptor, horizontal, bipolar, and multipolar cells. Multipolar cells include the amacrine and ganglion cell types. The accuracy of cell type identification based on light microscopic morphology has been confirmed by extensive studies of isolated retinal cell physiology (Bader et al., 1978, 1982; Attwell et al., 1987; MacLeish and Townes-Anderson, 1988; Tessier-Lavigne et al., 1988; Eliasof and Werblin, 1993), neurochemistry (Sherry and Townes-Anderson, 1995) and ultrastructure (Townes-Anderson et al., 1985; MacLeish and TownesAnderson, 1988; Townes-Anderson and Vogt, 1989). Further, isolated salamander retinal neurons maintained in defined, serum-free medium exhibit cell-type-specific growth (Fig. 1; MacLeish and Townes-Anderson, 1988; Mandell et al., 19931, which allows continued cell type identification for weeks in culture. Similar cell-type-specific growth has

been described in other retinal systems as well (Dowling et al., 1985; Ishida and Cheng, 1991a,b). Morphologically identifiable regenerating photoreceptor, horizontal, bipolar, and multipolar cells from salamander retina also have been shown to re-establish functional synapses in culture (MacLeish and Townes-Anderson, 1988). Finally, the distribution of amino acid transmitters in isolated retinal cells during regenerative growth has been compared with their distribution in the retina (Sherry and Townes-Anderson, 1995).Amino acid transmitter localization in isolated retinal neurons was similar to the cell-type-specific distribution of the intact retina and was consistent for at least 2 weeks in culture. There were only two exceptions. First, the intensity of GABA-immunoreactivity (IR) in isolated horizontal cells tended to decrease over time in culture (Fig. 2B). Second, ASP-IR was elevated in photoreceptors (see Figs. 2E, 4A,B). Photoreceptors. Isolated photoreceptors were easily identified (even if the outer segment was absent) by the presence of the ellipsoid, a collection of mitochondria unique to photoreceptors (Fig. lA,B). Many photoreceptors lose the outer segment during isolation or the first few days in culture, although a few cells retain the outer segment for several weeks. Once lost, outer segments are not regenerated, but cell-type-specific expression of visual pigments in the plasma membrane continues (Mandell et al., 1993). Regenerating photoreceptors showed stereotypical patterns of growth (Fig. 1A,B; MacLeish and Townes-Anderson, 1988; Mandell et al., 1993). Shortly after isolation, photoreceptors extend filopodial processes; thicker, neuritic processes are present by approximately 24 hours postisolation. Neuritic processes can arise from any point on the circumference of the soma. Formation of synaptic varicosities

Fig. 1. Identification of isolated retinal cell types immediately after dissociation (A,C,E,H) and after 2 weeks in culture (B,D,F,G,I). A,B. Photoreceptors. A Freshly dissociated photoreceptors can be identified unequivocally by the presence of the ellipsoid (El, even when the outer segment is absent (arrow, rod photoreceptor with an intact outer segment). Other cells are examples of neurons in the undetermined category. B: Photoreceptors, identified by the ellipsoid (El, showing stereotyped regenerative process growth at 2 weeks after isolation. Photoreceptor PI has several neurites with synapticvaricosities (arrowheads) in addition to slender filopodial processes. A varicosity contact (arrow) from photoreceptor PI onto the soma of a nearby cell of undetermined type (asterisk) is visible. Photo receptor Pz has more limited process growth. C,D. Horizontal cells. C: Horizontal cell shortly after isolation characterized by a large soma and several thick processes emerging from the cell body. The processes have a somewhat flattened morphology. D: An isolated horizontal cell after 2 weeks in culture retaining its large somatic size and broad, flattened processes. E-G. Bipolar cells. E: A freshly isolated bipolar cell showing the characteristic stout primary dendrite (arrow) with several secondary dendrites, relatively small cell body, axon, and axon terminal (arrowheads). F: An isolated bipolar cell after 2 weeks in culture. The primary dendrite (arrow)has flattened into a large, complex lamellipodium growing away from a nearby photoreceptor (PI. A slender axon ending in a small axon terminal (arrowheads) is also present. Note that the photoreceptor cell has not established any varicosity contacts onto the bipolar cell despite its proximity. G Isolated bipolar cells lacking the axon and axon terminal can often be identified on the basis of the characteristic stout primary dendrite (arrow) even after 2 weeks in culture. H,I. Multipolar cells. H: Example of a small multipolar cell, identified by its complex process arbor, shortly after isolation. Note the differences in shape and diameter between the primary process (arrow) of the multipolar cell and the stout primary dendrite of the bipolar cell shown in panel E. I: Example of a multipolar cell maintained in culture for 2 weeks. The cell is characterized by several long, relatively slender processes. Scale bar = 50 km.

TARGETING OF PHOTORECEPTOR SYNAPSE REGENERATION

Figure 1

479

D.M. SHERRY ET AL.

480

along neuritic processes begins after 1-2 days in culture. Synaptic varicosities represent regenerated presynaptic terminals, containing synaptic vesicles that can undergo depolarization-induced recycling (Mandell et al., 1993) and the photoreceptor transmitter GLU (Fig. 2G). Horizontal, bipolar, and multipolar cells. Horizontal cells were characterized by relatively short, thick processes radiating from a large soma with a large ovoid nucleus (Fig. lC,D). Horizontal cells tended to flatten over time in culture but retained the characteristic large soma size and radial organization of short, broad processes. Bipolar cells (Fig. 1E,F) were distinguished by a relatively small cell body, a short stout primary dendrite, and a slender axon ending in a discrete terminal. Bipolar cell primary dendrites frequently formed elaborate lamellipodia with time in culture. Bipolar cell axon terminals also tended to form lamellipodia of variable size and complexity. Bipolar cells that lost the axon during isolation often could be identified by the characteristic stout primary dendrite (Fig. 1G). Multipolar cells, which included the amacrine, interplexiform, and ganglion cells, were characterized by one or more relatively slender processes extending from the soma and forming process arbors (Fig. 1HJ). Regenerating processes from multipolar cells often displayed growth cones, and the process arbors often became larger and more elaborate as process regeneration proceeded. Undetermined cells. Cells that suffered extensive process loss during isolation and showed little process outgrowth in culture, could not be identified unequivocally at the light microscopic level and were classified as undetermined. Undetermined cells clearly were not photoreceptor or horizontal cells because they lacked the ellipsoid characteristic of photoreceptors and the large flattened soma characteristic of horizontal cells. At the electron microscopic level, some undetermined cells have ultrastructural characteristics of bipolar cells, such as synaptic ribbons, whereas other undetermined cells show multipolar cell ultrastructure (Sherry and Townes-Anderson, unpublished data). Transmitter localization studies indicated that undetermined cells with transmitter content typical of bipolar or multipolar cells were both present (Sherry and TownesAnderson, 1995, and below). The undetermined cell category, therefore, was considered to be a mixture of bipolar and multipolar cells.

TABLE 1. Formation of Varicosity Contactsby Regenerating Photoreceptorsat 1and 2 Weeks Postisolation 1Week postisolation'

2 Weeks postisolation2

100% 0% 1.1 i 0.1 1-3 1.4 i 0.8 1-4

82.370 17.7% 1.1 5 0.1 1-2 1.3 ? 0.7 1-4

Contacts onto target soma (%) Contacts onto target process (56) Mean no. target cellsiphotoreceptor Range of target cellsiphotoreceptor Mean no. varicositiesitarget cell Range of varicositiesitarget cell

'One-week postisolation data derived from 3 cultures from 3 animals; 34 photoreceptors making contacts; 38 target cells. 2Two-week postisolation data derived from 17 cultures from 14 animals; 124 photoreceptors making contacts; 135 target cells. Mean values are expressed as mean standard deviation.

*

TABLE 2. Varicosity Formation by Regenerating PhotoreceptorsThat Do and Do Not Contact a Target Cell at 1and 2 Weeks Postisolation 1Week postisolation' Mean varicositiesiphotoreceptor making contact Mean varicosities1 photoreceptor not making contact

5.5 5 5.1 (n

=

34 cells)

1.7 i 2.3 (n = 116 cells)

2 Weeks postisolation2 5.5 5 2.9 (n = 110 cells) 1.7 i 2.4

(0 =

251 cells)

'One-week postisolation data derived from 3 cultures from 3 animals. Two-week postisolation data derived from 17 cultures from 14 animals Values expressed as mean 5 standard deviation.

Photoreceptors making varicosity contacts regenerated significantly more varicosities than photoreceptors that did not make varicosity contacts (Table 2). The number of varicositiesicell regenerated by photoreceptors making contacts was similar at 1and 2 weeks postisolation, as was the number of varicositiesicell regenerated by photoreceptors lacking varicosity contacts. These data are consistent with the data presented in Table 1, suggesting that varicosity number is stable by 1week postisolation, before formation of functional synapses. Two-way analysis of variance (ANOVA) analysis followed by Scheffe's post hoc test showed that the number of varicosities regeneratedicell correlated with the establishment of varicosity contacts and was independent of time in culture (P < 0.0001). Therefore, establishment of varicosity contacts may have enhanced varicosity formation. Alternatively, photoreceptors that regenerated large numbers of varicosities may have been more likely to form contacts.

Regeneration of synaptic varicosities and synaptic varicosity contacts The stability of intercellular contact is critical to regeneration of functional synapses. Functional synapses between salamander neurons have been reported to appear beginning 10 days after cell plating (MacLeish and TownesAnderson, 1988). Therefore, the number of contacts made by the presynaptic varicosities of photoreceptors was examined at 1 and 2 weeks postisolation, before and after the appearance of functional synapses in culture, respectively. Varicosity contact parameters were virtually identical at 1 and 2 weeks postisolation (Table I), suggesting that varicosity contacts established by 1week in vitro were maintained during the period of synapse formation in culture. Target cells typically received a single varicosity contact, but several examples of target cells receiving multiple contacts were observed. The increase in contacts onto processes at 2 weeks postisolation may reflect the increase in process outgrowth by the target cells. Photoreceptors usually contacted a single target cell, but cases of a single photoreceptor contacting multiple targets also were observed.

Fig. 2. Regenerative photoreceptor growth and synaptic varicosity formation in vitro. A-C. A Phase contrast micrograph of a rod photoreceptor (P) retaining the outer segment (arrow) making a varicosity contact (arrowhead) onto a multipolar cell (asterisk), a novel target type. A horizontal cell (H) is partially visible at the top of the micrograph. B: The multipolar target cell shows GABA-immunoreactivity, but the photoreceptor and horizontal cell show only background labeling. C: SV2-immunoreactivity indicates the presence of synaptic vesicles at the contact site (arrowhead). D,E. D: Phase contrast micrograph of a regenerating photoreceptor lacking an outer segment (Pz)making two varicosity contacts (arrowheads) onto a flat, lamellar process from a horizontal target cell (asterisk), a normal target cell type. Another photoreceptor (PI) does not establish varicosity contacts onto the horizontal cell. E: Both photoreceptors and their synaptic varicosities exhibit the intense ASP-immunoreactivitytypical of salamander photoreceptors regenerating in culture (Sherry and Townes-Anderson, 1995). The horizontal cell shows only background ASP-immunoreactivity. F,G. F: Phase contrast micrograph of a photoreceptor lacking an outer segment (P) making a varicosity contact (arrowhead) onto the soma of an inappropriate multipolar target cell (asterisk). G The photoreceptor, its synaptic varicosities, and the multipolar target cell all show GLU immunoreactivity. Scale bar = 50 (rm.

TARGETING OF PHOTORECEPTOR SYNAPSE REGENERATION

Figure 2

481

482

D.M. SHERRY ET AL.

Fig. 3. Ultrastructural features of a synaptic varicosity contact. The synaptic varicosity is at the top of the micrograph; the soma of the target cell is at the bottom of the micrograph. The varicosity is filled with synaptic vesicles. A synaptic ribbon with a halo of synaptic vesicles is present (arrow). Presynaptic and postsynaptic densities similar to the symmetrical contacts of developing photoreceptor terminals in the intact retina are also present (arrowheads). A coated pit also is visible (heavy arrow), suggestingactive vesicular recycling. Inset: The varicos-

ity contact (arrowhead) was made by a photoreceptor (P) onto an undetermined target cell (asterisk).Identification of the target cell type at the electron microscopic level is tentative, but the ultrastructural characteristics of the target cell were consistent with multipolar cell identification-an invaginated nucleus, typical of amacrine cells, and an absence of the small synaptic ribbons typical of bipolar cells. Scale bar = 0.5 pm.

Synaptic varicosities regenerated by photoreceptors are sites of synapse formation when apposed to a neighboring neuron (Figs. 2,3). In cultures immunoreacted for SV2, all synaptic varicosity contacts were immunoreactive (67 contacts in 8 cultures from 8 animals). All varicosity contacts in cultures immunoreacted for GLU also were immunoreactive (20 contacts in 3 cultures from 3 animals). These data indicate that both synaptic vesicles and the photoreceptor transmitter were invariably present in varicosity contacts, consistent with a synaptic function. Electron microscopic examination of four varicosity contacts demonstrated the presence of ultrastructural features characteristic of photoreceptor synapses in all contacts (Fig. 3). Numerous small clear vesicles, dense-core vesicles, and short, free-floating synaptic ribbons were present; in addition, presynaptic and postsynaptic densities similar to the symmetrical contacts of developing photoreceptor synapses (Blanks et al., 1974) occurred along the membranes at the site of contact. This ultrastructural organization is characteristic of developing photoreceptor synapses (Blanks et al., 1974; Chen and Witkovsky, 1978). The attachment of ribbons to the plasma membrane, indicative of a more mature synapse, has been observed in later stages of regeneration at approximately

30-40 days postisolation (MacLeish and Townes-Anderson, 1988).

Photoreceptor target selectivity Photoreceptors at 2 weeks postisolation made varicosity contacts onto all types of retinal neurons, indicating that photoreceptors were capable of regenerating synapses with any type of retinal neuron (Table 3). However, the most frequently identified targets were multipolar cells, neurons that would not be contacted by photoreceptors in vivo. Very few targets were of the expected cell types-horizontal, bipolar, or photoreceptor cells. The rarity of varicosity contact formation onto bipolar and horizontal cells was not due to a loss of competence for intercellular contact formation by second-order neurons, because several examples of bipolar and horizontal cells making intercellular contacts with neurons other than photoreceptors were observed. Most target cells, however, fell into the undetermined category. This target cell type distribution was very consistent across cultures. At 1 week postisolation, all observed synaptic varicosity contacts formed onto multipolar or undetermined cells, although several instances of photoreceptors with potential horizontal or bipolar cell targets

TARGETING OF PHOTORECEPTOR SYNAPSE REGENERATION

483

TABLE 3. Comparison of Target Cell Types at 1and 2 Weeks Postisolation

Target cells at 1 week’ Target cells at 2 weeks2

PR

HZ

BP

MULTI

UND

Total targets

0 7 (5.2%)

0 2 (1.5%)

0 1(0.7%)

2 (5.3%) 43 (31.9%)

36 (94.7%) 82 (60.71)

38 135

No. of cultures sampled

~

~~~

’One-week postisolation data derived from 3 cultures from 3 animals. 2Two-weekpostisolation data derived from 17 cultures from 14 animals. PR, photoreceptors; HZ, horizontal cells; BP, bipolar cells; MULTI, multipolar cells; UND, undetermined cells.

were seen. The high proportion of undetermined target cells at 1 week postisolation probably reflected lower levels of process outgrowth at the earlier time point. The similarity of target cell types at 1 and 2 weeks postisolation is consistent with the establishment of stable varicosity contacts by 1 week postisolation. Varicosity contacts onto Miiller glial cells were never observed. The predominance of multipolar and undetermined cell targets suggested that photoreceptors might selectively establish contacts with novel synaptic partners while avoiding normal synaptic partners. Further, regenerating photoreceptors often contacted one neighboring neuron without establishing contacts with other nearby neurons (Fig. 4A,B), also suggesting target selectivity. To determine if varicosity contact formation was truly cell type selective and not simply related to the number of available cells for each cell type, quantitative analysis of actual targets versus available targets for regenerating photoreceptors was performed at 2 weeks postisolation (Fig. 4C). The vast majority of varicosities form within 25 pm of the photoreceptor soma (Mandell et al., 1993). Therefore, any neuron within a 25-pm radius of a photoreceptor cell was defined as a potential target; neurons outside the 25-pm radius were excluded from analysis. Because a single neuron could be a potential target for more than one photoreceptor, the total number of times a neuron was available to be a photoreceptor target was defined as the number of target opportunities (for example, a neuron within 25 pm of two photoreceptors would represent two target opportunities). Neurons that received varicosity contacts were defined as actual targets. Target cell type was identified by morphology in cultures that were double-labeled for SV2, to highlight varicosities, and GABA (four cultures) or GLY (three cultures), to characterize target cell transmitter content. Analysis was performed for a total of 482 photoreceptors in 7 cultures from 7 animals. Fifty-two photoreceptors (10.8%) made varicosity contacts. No actual target cells in this sample received contacts from more than one photoreceptor. Nonrandom, target-selective contact formation would result in a disproportionate number of contacts onto specific cell types (Table 4). Chi-square analysis of observed versus expected targets showed extremely nonrandom formation of varicosity contacts onto the different target cell types (P < 0.0001). As a quantitative measure of the selectivity of synaptic varicosity contact formation, the “selectivity ratio,” defined as (% of total actual targets):(% of total target opportunities), was calculated for each target cell type. Random varicosity contact formation onto a target cell type would result in a selectivity ratio of 1.0 (i.e.,%total actual targets = % total target opportunities for that cell type). Selectivity ratios > 1.0 and < 1.0 would indicate above-random and below-random contact formation frequency, respectively. Selectivity ratios for each cell type were calculated in each culture and analyzed by one-way ANOVA analysis with Fisher’s protected least significant difference post hoc test. Varicosity contact formation onto

photoreceptor, horizontal, and bipolar cell targets showed similar, below-random selectivity ratios. These selectivity ratios differed significantly from the above-random selectivity ratios for multipolar ( P < 0.0001) and undetermined ( P 2 0.01) cell types. Selectivity ratios for multipolar and undetermined cell types also differed from one another (P= 0.003). Thus, regenerating photoreceptors distinguished between target cell types, exhibiting a strong selectivity for novel target cell types and actually avoiding normal target cell types. Culture densities ranged from 28.0-127.3 cells/ mm2 (mean = 71 cells/mm2), but target preferences were consistent across cultures, indicating that culture density did not determine target selectivity.

Neurochemical selectivity The factors determining cell-type-selective synapse formation are poorly understood, but neurotransmitters are excellent candidates for modulating regenerative synaptic selectivity. Transmitters in the adult CNS show cell-typespecific distribution, affect regenerative process growth (Lipton and Kater, 19891, and mediate synaptic plasticity (Collingridge and Singer, 1990). Transmitters also modulate neuronal cytoarchitecture, axonal growth, and synapse formation by developing CNS neurons (Mattson, 1988; Mattson and Kater, 1989; Redburn, 1992; Bodnarenko and Chalupa, 1993; Goodman and Shatz, 1993; Leslie, 1993; Zheng et al., 1994; Bodnarenko et al., 1995). To determine the potential relationship between synapse formation and transmitters, the amino acid transmitter content of the actual target cells (cells contacted by a photoreceptor varicosity) in each culture was determined immunocytochemically by labeling for one of four putative amino acid transmitters (GLU, GABA, GLY, or ASP). As described above, GABA-immunoreacted and GLY-immunoreacted cultures were double-labeled for SV2 to help visualize photoreceptor varicosities. In cultures labeled for GLU, 65.0% of the actual target cells showed GLU-IR; 52.9% of the actual target cells in cultures labeled for GABA showed GABA-IR.A total of 33.3% of actual target cells in cultures labeled for GLY showed GLY-IR; 12.5% of actual target cells in cultures immunoreacted for ASP showed ASP-IR. No colocalization studies were performed, but the large percentage of actual target cells containing GLU suggests that some postsynaptic cells may have contained more than one amino acid transmitter. Glutamate is found in the vast majority of neurons in the intact salamander retina (Yang and Yazulla, 1994a,b; Sherry and Townes-Anderson, 1995) and colocalizes extensively with other amino acids (Yang and Yazulla, 1994a,b). The transmitter selectivity of varicosity contact formation was determined with a “transmitter selectivity ratio,” similar to the selectivity ratio used to assess cell type selectivity (Table 5 ) . The transmitter selectivity ratio was defined as (% of total actual targets containing a given transmitter):(% of total target opportunities onto cells containing that transmitter). Similar to the selectivity ratio

D.M. SHERRY ET AL.

484

C

Fig. 4. Target selectivity of varicosity contact formation. A: A photoreceptor (PI) forming a varicosity contact (arrowhead) onto the soma of a multipolar neuron (asterisk), a novel target, while avoiding a nearby photoreceptor (Pz), a normal target. Both photoreceptors and the multipolar target cell show ASP-IR (ASP-IR is typical of regenerating photoreceptors; Sherry and Townes-Anderson, 1995). B: An ASPimmunoreactive photoreceptor (P) makes several varicosity contacts (arrowheads) onto a nearby ASP-immunoreactive undetermined cell (asterisk), but does not establish varicosity contacts with a nearby horizontal cell (H), an appropriate target cell type, or a neighboring Muller glial cell (MI. C: Illustration of selectivity analysis. Regenerating photoreceptors (PI and Pz) form most synaptic varicosities within 25 pm of the soma (represented by dotted circles).Any neuron within 25 pm of a regenerating photoreceptor was defined as a potential target and represented a target opportunity for that photoreceptor; cells further than 25 pm from a regenerating photoreceptor were not con-

sidered a target opportunity. A neuron within 25 pm of more than one photoreceptor (asterisk) was scored as a single potential target cell but represented multiple target opportunities (two target opportunities in this case). Cells receiving a varicosity contact were designated as actual targets. In the example illustrated, the photoreceptor on the left (PI) has two target opportunities, a bipolar cell (B) and a multipolar cell (asterisk),which receives a varicosity contact (arrowhead). Because the multipolar cell receives a varicosity contact, it is also an actual target. The photoreceptor on the right (Pz)also has two potential targets, the multipolar cell and a horizontal cell (H), but does not establish varicosity contacts with either cell. Therefore, it has no actual targets. The undetermined cell (U) at the bottom lies further than 25 pm from both photoreceptors; therefore, it is not a potential target and represents no target opportunities for either photoreceptor. Scale bar = 50 pm in A and B; C not to scale.

TARGETING OF PHOTORECEPTOR SYNAPSE REGENERATION

485

TABLE 4. Target Selectivity of Regenerating Photoreceptors'

No. potential target cells No. actual target cells No. target opportunities % of total actual target cells % of total target opportunities Selectivity ratio2(mean t SEMI

PR

HZ

BP

MULTI

UND

Total

214 5 250 9.1% 35.0% 0.3 i 0.2

43 1 45 1.8% 6.3% 0.2 2 0.2

13 0 15 0% 2.18 0.0 i- 0.0

158 25 167 45.5% 23.4% 2.1 ? 0.3

217 24 237 43.6% 33.2% 1.1 0.2

645 55 714 100% 100% 1.0

*

'Data derived from 482 photoreceptorswith potential targets in 7 2-week postisolation cultures derived from 7 animals. 2Selectivityratio = (% of total actual targets): (9%of total target opportunities); see text for details. PR, photoreceptors;HZ, horizontal cells; BP, bipolar cells; MULTI, multipolar cells; UND, undetermined cells.

TABLE 5. Transmitter Selectivity of Regenerating Photoreceptors Transmitter GABA GLY GLU ASP

Total actual targets

No. and 8 of actual targets, transmitter-IR(+)

No. total target opportunities

Total target opportunities, transmitter-IR(+)

34 21 20 48

18 (52.9%) 9 (42.9%) 13 (65.0%) 6 (12.58)

476 238

92 (19.3%) 89 (37.4%)

-

Selectivity ratio'

85.8%2

49.9952

2.7 1.1 0.8 0.3

'Selectivity ratio = [% oftotal actual targets transmitter-IR(+)l:l%of total target opportunities transmitter-IR(+ )I. ZEstimated, see text. GABA, gamma-aminobutyricacid; GLY, glycine; GLU, glutamate; ASP, aspartate.

used for cell type selectivity analysis, a ratio of 1.0 would indicate random transmitter selectivity, and ratios > 1.0 and < 1.0 would indicate above-random and below-random transmitter selectivity, respectively. GABA-selectivity and GLY-selectivitywere determined in the cultures immunoreacted for GABAlSV2 or GLYISV2 used for cell type selectivity analysis, because the morphologic characteristics and the GABA or GLY content of all potential and actual targets in these cultures were known. The GABA selectivity ratio was 2.7 "52.9% of actual targets showed GABA-IR]:L19.3% of target opportunities showed GABA-IR]), indicating a strong preference for formation of contacts onto GABAcontaining targets. In this light, horizontal cells-which contain GABA in the intact salamander retina (Yang and Yazulla, 1988b)-might be expected to be a favored photoreceptor target. However, horizontal cells in long-term culture often showed attenuated GABA-IR (Fig. 2B; Sherry and Townes-Anderson, 1995), consistent with photoreceptor selectivity against horizontal cell targets (Table 4). The GLY selectivity ratio was 1.1, indicating that GLYcontaining targets were neither selected against nor favored on the basis of their GLY content. Selectivity for GLU-containing and ASP-containing targets also can be estimated if the proportion of potential targets showing GLU-IR or ASP-IR is assumed to be similar to the proportion of all cells in culture showing GLU-IR or ASP-IR. In 2-week postisolation cultures labeled for GLU, 65.0%of actual target cells showed GLU-IR, but 85.8%of all cells in culture showed GLU-IR (?7.2% standard deviation; 369 total cells sampled in 3 cultures from 3 animals), yielding an estimated GLU selectivity ratio of 0.8. This ratio indicated approximately random selectivity for target cells on the basis of their GLU content. Thus, although a high proportion of the cells receiving contacts contained GLU, a high proportion of the available targets contained GLU as well. This analysis suggests that the observed avoidance of photoreceptor, horizontal, and bipolar cells (which all show GLU-IR) and the apparent selectivity for GLU-immunoreactive multipolar and undetermined target cells (12 of 19 cells) was probably not attributable to GLU. In 2-week postisolation cultures labeled for ASP, 12.5% of target cells showed ASP-IR (Table 4), but 49.9% of all cells in culture showed ASP-IR (25.8%

standard deviation; 445 total cells sampled in 3 cultures from 3 animals), yielding an ASP selectivity ratio of 0.3. This selectivity ratio indicated that photoreceptors selected against cells containing ASP as targets. These data suggest that GABA is a positive modulator of synaptic targeting by regenerating photoreceptors. However, it is also clear from these data that the GABA content of a target cell cannot be the sole determinant of target selectivity because contacts formed onto targets containing other amino acids as well.

DISCUSSION The current study corroborates the finding that adult photoreceptors can form synapses onto normal horizontal and bipolar cell targets (MacLeish and Townes-Anderson, 1988; Gouras et al., 1991; Silverman et al., 1992; Mosinger Ogilvie et al., 1994). However, this study extends those observations, showing quantitatively that adult photoreceptors regenerating in a defined culture system display pronounced target selectivity, preferring novel multipolar cell targets (amacrine and ganglion cells) over their normal synaptic targets. It is presumed that both rod and cone photoreceptors regenerated synaptic contacts in the current study, because both cell types are present in the cultures and both cell types regenerate processes and synaptic specializations (MacLeish and Townes-Anderson, 1988; Mandell et al., 1993). However, the regenerative growth patterns of rods and cones are similar, making unequivocal identification of rods versus cones difficult without the use of a marker such as rhodopsin immunolabeling. Our focus on synaptic contacts between morphologically and neurochemically defined neurons using doublelabeling with antibodies against SV2 and neurotransmitters precluded the additional use of photoreceptor-specific markers. In the future, it will be of interest to determine whether rod and cone cell synaptic selectivities differ significantly.

Preference for novel targets The target preferences observed in this study provide an explanation for results reported after photoreceptor transplants and in degenerating retinas. Incorporation of normal rod cells into host rat retina yields only a fraction of the

D.M. SHERRY ET AL.

486

usual number of synapses in the outer plexiform layer even Yazulla, 1994a), but did contain GABA or GLY, which are though the transplanted cells seem viable for many months found primarily in amacrine cells (Yang and Yazulla, and are closely apposed to the dendrites of the horizontal 1988a,b). Less than 1% of salamander ganglion cells conand bipolar cells (Gouras et al., 1991; Silverman et al., tain GABA (Watt et al., 1994), and salamander ganglion 1992; Mosinger Ogilvie et al., 1994). The current study cells containing GLY have not been reported (Yang and suggests that this low level of synapse formation is due to Yazulla, 1988a; Watt and Florack, 1993). There are small poor capacity for synapse regeneration between photorecep- subpopulations of salamander bipolar cells that contain tors and second-order neurons, specifically. More dramatic GABA or GLY (Yang and Yazulla, 1988a,b); however, these avoidance of normal synaptic target cells has been observed cell populations are much smaller than the populations of in the adult human retina (Li et al., 1995).In patients with GABA-containing and GLY-containing amacrine cells (Yang retinitis pigmentosa, a degenerative retinal condition, pho- and Yazulla, 1994a,b). Fourth, the selectivity ratio for the toreceptors in the peripheral retina extend neurites filled undetermined cell class exceeded the bipolar cell selectivity with synaptic vesicles past the second-order neurons into ratio but was less than the multipolar cell ratio, consistent the inner retina, where the amacrine and ganglion cells with selective contact formation onto a subset of multipolar reside. Although the formation of new synapses has not neurons within the undetermined group. Fifth, some multibeen reported, the growth toward the inner retina is polar and undetermined target cells had an indented nucleus consistent with the preference for multipolar cell targets (personal observations), which is typical of amacrine cells shown by cultured photoreceptors. Finally, in degenerating (Dowling, 1968). Resolution of this issue will require develretinas of rdslwild type or chimeric rdlwild type mice, opment of specific markers capable of distinguishing the bisurviving photoreceptors form new synapses with horizon- polar, amacrine, and ganglion cell populations unequivocally. Although the cell type of undetermined cells is not tal and bipolar cells; however, these new synapses occur only between photoreceptors and the second-order neurons known, an assessment of the effect of the undetermined they already contact and do not include establishment of cells on the analysis of target selectivity can be made using synapses with new synaptic partners (Jansen and Sanyal, neurotransmitter content as an indicator of cell type. All salamander bipolar cells contain glutamate (Yang and 1992; Sanyal et al., 1992). The reduced synaptic regeneration between first-order Yazulla, 1994a). If the assumption is made that all undeterand second-order retinal neurons in vitro is not due to a mined cells containing glutamate are bipolar cells, then generalized loss of regenerative capacity. Photoreceptors 52.5% of undetermined cells would be bipolar cells (from a did establish contacts with other neurons and interactions sample of 80 undetermined cells in 4 cultures; Sherry and between horizontal and bipolar cells with nonphotorecep- Townes-Anderson, 1995). This is an overestimate of the tors were observed in the cultures (personal observations) proportion of bipolar cells within the undetermined cell and have been reported previously (MacLeish and Townes- category because many amacrine and ganglion cells also Anderson, 1988). It is possible, however, that the comple- contain glutamate (Yang and Yazulla, 1994a,b; Sherry and ment of cell surface molecules on regenerated neurites Townes-Anderson, 1995). The remaining undetermined prevents or is insufficient for certain synaptic interactions. cells must be multipolar cells because undetermined cells Whatever the mechanism, establishment of contacts onto do not have the characteristics of photoreceptor or horizonnovel multipolar cell targets by adult photoreceptors is tal cells. Using the data in Table 4 and redistributing the consistent with the ability of other adult CNS neurons to undetermined cells as bipolar or multipolar cells, bipolar establish synapses onto novel targets (Ready and Nicholls, cells would become 22.9% of actual targets and multipolar 1979; Schacher et al., 1985; Arechiga et al., 1986; Cantore cells would become 66.2% of actual targets. The selectivity and Scalia, 1987; Haydon and Zoran, 1989; Aguayo et al., ratio for bipolar cells would then be 1.2, very near random, 1990; Nicholls et al., 1990; Raisman and Field, 1990; Zoran and 1.8 for multipolar cells, still well above random. Thus, and Haydon, 1990; Molnar and Blakemore, 1991; Zwimpfer even with an overly generous estimate of the proportion of et al., 1992). Thus, the data from the current study suggest bipolar cells within the undetermined category, the preferthat first-order and second-order neurons from the adult ence for multipolar cell targets remains. retina are not optimal .>ynaptictargets for regenerating Neurochemical selectivity and potential adult photoreceptors and, in conjunction with the in vivo synaptic targeting mechanisms reports, that adult neurons do not necessarily re-establish synaptic circuits that were formed during development. The pronounced selectivity of photoreceptors for GABAcontaining targets is consistent with transmitter receptor distribution and the role of GABA in photoreceptor developIdentity of target cells in the multipolar ment. Neurotransmitters, including GABA, directly affect and undetermined cell groups growth by those neurons that express the appropriate Although identification of specific cell types within the receptors (Michler-Stuke and Wolff, 1987; Mattson, 1988; multipolar and undetermined cell categories was not pos- Lipton and Kater, 1989; Mattson and Kater, 1989; Leslie, sible, several pieces of evidence suggest that varicosity 1993). Adult salamander photoreceptors express GABAA contacts formed selectively onto amacrine cells rather than receptors (Wu, 1986,1992;Yang et al., 1992). Furthermore, bipolar or ganglion cells. First, contacts onto identified GABA has been implicated as a positive regulator of bipolar cells were extremely rare in all cultures. Second, synaptogenesis by developing photoreceptors (Madtes and GABA and GLU colocalize in many salamander amacrine Redburn, 1983; Messersmith and Redburn, 1990, 1993; cells (Yang and Yazulla, 1994a,b),consistent with the high Redburn, 1992). GABAergic modulation of targeting may frequency of GABA-containing and GLU-containing multi- explain the pronounced preference for multipolar cell tarpolar target cells. Third, many unidentified target cells did gets and avoidance of horizontal and bipolar cell targets; not show GLU-IR, inconsistent with the known GLU multipolar cells are the principal GABA-containing cells in content of salamander bipolar and ganglion cells (Yang and culture, the GABA content of horizontal cells declines after

s

TARGETING OF PHOTORECEPTOR SYNAPSE REGENERATION the first few days in culture, and few bipolar cells express GABA (consistent with the intact retina) (Sherry and Townes-Anderson, 1995, unpublished data). The mechanisms by which GABA may guide synaptic targeting has not been determined, but chemotactic mechanisms are likely; transmitter gradients can guide axonal outgrowth (Zheng et al., 19941, and focal application of GABA has been shown to affect neurite extension by hippocampal neurons (Mattson, 1988; Mattson and Kater, 1989). The near-random selectivity of regenerating photoreceptors for targets on the basis of GLU and GLY content would be predicted by the reported absence of excitatory amino acid (Eliasof and Werblin, 1993) and GLY receptors (Miller et al., 1981) on salamander photoreceptors, which would render photoreceptors “blind” to GLU and GLY. The apparent selectivity for the GLU-containing subset of multipolar cells might be explained by the high frequency of GLU and GABA colocalization in salamander amacrine cells (Yang and Yazulla, 1994a) or might be related to other nontransmitter factors. Because the excitatory amino acid receptors that mediate ASP effects in addition to those of GLU appear to be absent from salamander photoreceptors (Eliasof and Werblin, 19931, the observed selectivity against ASP-containing targets probably is not due to ASP itself. This interpretation is consistent with the suggestion that ASP subserves primarily metabolic rather than signaling functions in the retina (Massey, 1990; Kalloniatis et al.,1994). Although this study suggests that GABA may be an important modulator of photoreceptor synaptic targeting, GABA is clearly not the only factor involved because photoreceptors also regenerated contacts onto non-GABAcontaining targets. This observation is consistent with photoreceptor synapse formation during development, in which contacts are established with bipolar cells that do not express GABA. Furthermore, these results are consistent with a growing body of data indicating that developmental synapse formation and targeting involves several presynaptic and postsynaptic factors (c.f., Collingridge and Singer, 1990; Goodman and Shatz, 1993; Leslie, 19931, of which neurotransmitters are only one. Retinotopically distributed cell surface proteins, such as TOPATand TOPD” (Trisler, 19901, have been proposed to guide developmental and regenerative synaptogenesis in the visual system and are present on all retinal cell types. Therefore, mismatches in TOP proteins, or similar retinotopically distributed proteins, could potentially inhibit synapse formation between neurons from different retinal regions (e.g., dorsal and ventral retina). However, in our dispersed cultures, such effects would be manifested as a decrease in the total number of contacts formed-because the cells are distributed at random throughout the culture-rather than a change in cell type selectivity. Other retinal cell surface markers that could account for the observed target selectivity have not been described. Extracellular substrate also can influence process guidance (see Goodman and Shatz, 1993). However, in this culture system, the extracellular substrate (Sal-1) is uniform and, therefore, would not produce directed process growth. Thus, GABA is the only candidate identified to date that may modulate the target selectivity of regenerating adult photoreceptors.

ACKNOWLEDGMENTS This work was supported by a David Warfield Fellowship in Ophthalmology from the New York Trust and New York

487

Academy of Medicine (D.M.S.) and a grant from the National Eye Institute (E.T.-A.; EY06135). The electron microscopy facility is supported in part by an unrestricted grant from Research to Prevent Blindness, Inc.

LITERATURE CITED Aguayo, A.J., G.M. Bray, M. Fiasminsky, T. Zwimpfer, D. Carter, and M. Vidal-Sanz (1990) Synaptic connections made by axons regenerating in the central nervous system ofadult mammals. J. Exp. Biol. 153:199-224. Arechiga, H., M. Chiquet, D.P. Kuffler, and J.G. Nicholls (1986) Formation of specific connections in culture by identified leech neurones containing serotonin, acetylcholineand peptide transmitters. J. Exp. Biol. 126:1531. Attwell, D., P. Mobhs, M. Tessier-Lavigne, and M. Wilson (1987)Neurotransmitter-induced currents in retinal bipolar cells of the axolotl, Ambystoma mexicanum. J. Physiol. (Lond.) 387:125-161. Bader, C.R., P.R. MacLeish, and E.A. Schwartz (1978) Responses to light of solitary rod photoreceptors isolated from tiger salamander retina. Proc. Natl. Acad. Sci. USA 75:3507-3511. Bader, C.R., D. Bertrand, and E.A. Schwartz (1982) Voltage-activated and calcium-activated currents studied in solitary rod inner segments from the tiger salamander. J. Physiol. (Lond.) 331.253-284. Blanks, J.C., A.M. Adinolfi, and R.N. Lolley (1974) Synaptogenesis in the photoreceptor terminal of the mouse retina. J. Comp. Neurol. 156:81-94. Bodnarenko, S.R., and L.M. Chalupa (1993) Stratification of ON and OFF ganglion cell dendrites depends on glutamate-mediated afferent activity in the developing retina. Nature 364:144-146. Bodnarenko, S.R., G. Jeyarasasingam, and L.M. Chalupa (1995) Development and regulation of dendritic stratification in retinal ganglion cells by glutamate-mediated afferent activity. J. Neurosci. 15:7037-7045. Buckley, K., and R.B. Kelly (1985) Identification of a transmembrane glycoprotein specific for secretory vesicles of neural and endocrine cells. J. Cell Biol. 100:1284-1294. Cantore, W.A., and F. Scalia (1987) Ultrastructural evidence of the formation of synapses by retinal ganglion cell axons in two non-standard targets. J. Comp. Neurol. 261t137-147. Chen, F., and P. Witkovsky (1978) The formation of photoreceptor synapses in the retina of larvalxenopus. J. Neurocytol. 7721-740. Collingridge, G.L., and W. Singer (1990) Excitatory amino acid receptors and synaptic plasticity. Trends Pharmacol. Sci. 11:290-296. Dowling, J.E. (1968) Synaptic organization of the frog retina: An electron microscopic analysis comparing the retinas of frogs and primates. Proc. R. SOC.Lond. B. Biol. Sci. 170:205-228. Dowling, J.E., M.W. Pak,and E.M. Lasater (1985) White perch horizontal cells in culture: Methods, morphology, and process growth. Brain Res. 360:33 1-338. Eliasof, S., and F.S. Werblin (1993) Characterization of the glutamate transporter in retinal cones of the tiger salamander. J. Neurosci. 13:402-412. Goodman, C.S., and C.J. Shatz (1993) Developmental mechanisms that generate precise patterns of neuronal connectivity. Neuron Suppl. 10: 77-98. Gouras, P., J. Du, M. Gelanze, R. Lopez, R. Kwun, H. Kjeldbye,and W. Krehs (1991) Survival and synapse formation of transplanted rat rods. J. Neural Transplant. Plast. 2:91-100. Haydon, P.G., and M.J. Zoran (1989) Formation and modulation of chemical connections: Evoked acetylcholine release from growth cones and neur i t e s of specificidentified neurons. Neuron 2:1483-1490. Ishida, A.T., and M.-H. Cheng (1991a) Cold inhibits neurite outgrowth from single retinal ganglion cells isolated from adult goldfish. Exp. Eye Res. 52;175-191. Ishida, A.T., and M.-H. Cheng (1991b) Synchronous neurite branchings in single goldfish retinal ganglion cells. Vis. Neurosci. 6:537-549. Jansen, H.G., and S. Sanyal (1992) Synaptic plasticity in the rod terminals after partial photoreceptor cell loss in the heterozygous rds mutant mouse. J. Comp. Neurol., 316:117-125. Kalloniatis, M., and E.L. Fletcher (1994) Immunocytochemical localization of the amino acid neurotransmitters in the chicken retina. J. Comp. Neurol. 336:174-193. Kalloniatis, M., R.E. Marc, G. Tomisch, and G. Barnes (1994) Pathways of glutamate production in the mammalian retina. Invest. Ophthalmol. Vis. Sci. Suppl. 351908. Lasansky, A. (1973) Organization of the outer synaptic layer in the retina of the larval tiger salamander. Philos. Trans. R. SOC.Lond. B. 265t471-489.

488 Lasansky, A. (1978) Contacts between receptors and electrophysiologically identified neurones in the retina of the larval tiger salamander retina. J. Physiol. (Lond.)285:531-542. Leslie, F.M. (1993) Neurotransmitters as neurotrophic factors. In S.E. Loughlin and J.H. Fallon (eds): Neurotrophic Factors. San Diego: Academic Press, pp. 565-598. Li, Z.-Y., I.J. Kljavin, and A.H. Milam (1995) Rod photoreceptor neurite sprouting in retinitis pigmentosa. J. Neurosci. 15:5429-5438. Lipton, S.A., and S.B. Kater (1989) Neurotransmitter regulation of neuronal outgrowth, plasticity and survival. Trends Neurosci. 12265-270. MacLeish, P.R., and E. Townes-Anderson (1988) Growth and synapse formation among major classes of adult salamander retinal neurons in vitro. Neuron 1:751-760. MacLeish, P.R., C.J. Barnstable, and E. Townes-Anderson (1983) Use of a monoclonal antibody as a substrate for mature neurons in uitro. Proc. Natl. Acad. Sci. USA80:7014-7018. Madtes, P. Jr., and D.A. Redburn (1983) GABA as trophic factor during development. Life Sci. 33:979-984. Mandell, J.W., P. MacLeish, and E. Townes-Anderson (1993) Process outgrowth and synaptic varicosity formation by adult photoreceptors in vitro. J. Neurosci. 13:3533-3548. Marc, R.E., W.-L.S. Liu, M. Kalloniatis, S.F. Raiguel, and E. Van Haesendonck (1990) Patterns of glutamate immunoreactivity in the goldfish retina. J. Neurosci. lOt4006-4034. Massey, S.C. (1990) Cell types using glutamate as a neurotransmitter in the vertebrate retina. Prog. Retinal Res. 9:399425. Mattson, M.P. (1988) Neurotransmitters in the regulation of neuronal cytoarchitecture. Brain Res. Rev. 13:179-212. Mattson, M.P., and S.B. Kater (1989) Excitatory and inhibitory neurotransmitters in the generation and degeneration of hippocampal neuroarchitecture. Brain Res. 478r337-348. Messersmith, E.K., and D.A. Redburn (1990) Kainic acid lesioning alters development ofthe outer plexiform layer in neonatal rabbit retina. Int. J. Dev. Neurosci. 8t447-461. Messersmith, E.K., and D.A. Redburn (1993) The role of GABA during development of the outer retina in the rabbit. Neurochem. Res. 18:463470. Michler-Stuke, A,, and J.R. Wolff (1987) Facilitation and inhibition of neurite elongation by GABA in chick tectal neurons. In V. Chan-Palay and S.L. Palay (eds): Neurology and Neurobiology. Vol. 32. New York Liss, pp. 253-266. Miller, R.F., T.E. Frumkes, M. Slaughter, and R.F. Dacheux (1981) Physiological and pharmacological basis of GABA and glycine action on neurons of the mudpuppy retina. I. Receptors, horizontal cells, bipolar cells, and G-cells.J. Neurophysiol. 4.5743-763. Molnar, Z., and C. Blakemore (1991) Lack of regional specificity for connections formed between thalamus and cortex in coculture. Nature 351 :475477. Mosinger Ogilvie, J., M. Landgraf, J.M. Lett, T. Tenkova, X. Wang, M. Wu, and M.S. Silverman (1994) Synaptic connectivity of transplanted photoreceptors in RD mice expressing the transgenic marker gene LacZ. Invest. Ophthalmol. Vis. Sci. Suppl. 35t1523. Nicholls, J.G., Y. Liu, B.W. Payton, and D.P. Kuffler (1990) The specificityof synapse formation by identified leech neurones in culture. J. Exp. Biol. 153t141-154. Raisman, G., and P.M. Field (1990) Synapse formation in the adult brain after lesions and after transplantation of embryonic tissue. J. Exp. Biol. 153:277-287. Ready, D.F., and J. Nicholls (1979) Identified neurones isolated from leech CNS make selective connections in culture. Nature 281r67-69. Redburn, D.A. (1992) Development of GABAergic neurons in the mammalian retina. Prog. Brain Res. 90:133-147. Sanyal, S., R.K. Hawkins, H.G. Jansen, and G.H. Zeilmaker (1992) Compensatory synaptic growth in the rod terminals as a sequel to partial photoreceptor cell loss in the retina of chimaeric mice. Development 114t797-803. Schacher, S., S.G. Rayport, and R.T. Ambron (1985) Giant Aplysia neuron R2 reliably forms strong chemical connections in vitro. J. Neurosci. 52851-2856.

D.M. SHERRY ET AL. Sheedlo, H., V. Gaur, L. Li, A.D. Seaton, and J.E. Turner (1991) Transplantation to the diseased and damaged retina. Trends Neurosci. 14:347-350. Sherry, D.M., and R.J. Ulshafer (1992a) Neurotransmitter-specific identification and characterization of neurons in the all-cone retina of Anolis carolinensis. I. Gamma-aminobutyricacid. Vis. Neurosci. 8:515-529. Sherry, D.M., and R.J. Ulshafer (199213)Neurotransmitter-specificidentification and characterization of neurons in the all-cone retina of Anolis carolinensis. 11. Glutamate and aspartate. Vis. Neurosci. 9:313-323. Sherry, D.M., and E. Townes-Anderson (1994) Morphological and neurochemical target preferences of adult photoreceptors regenerating in vitro. SOC.Neurosci. Abstr. 20:1088. Sherry, D.M., and E. Townes-Anderson (1995) Neurotransmitter content of isolated adult retinal neurons during regenerative growth. Invest. Ophthalmol. Vis. Sci. Suppl. 36:383. Silverman, M.S., S.E. Hughes, T.L. Valentino, andY. Liu (1992) Photoreceptor transplantation: Anatomic, electrophysiologic, and behavioral evidence for the functional reconstruction of retinas lacking photoreceptors. Exp. Neurol. 11537-94. Tessier-Lavigne, M., D. Attwell, P. Mobbs, and M. Wilson (1988) Membrane currents in retinal bipolar cells of the axolotl. J. Gen. Physiol. 91:49-72. Townes-Anderson, E., and B.A. Vogt (1989) Distribution of muscarinic acetylcholine receptors on processes of isolated retinal cells. J. Comp. Neurol. 29Ot369-383. Townes-Anderson, E., P.R. MacLeish, and E. Raviola (1985) Rod cells dissociated from mature salamander retina: Ultrastructure and uptake of horseradish peroxidase. J. Cell Biol. 100:175-188. Trisler, D. (1990) Cell recognition and pattern formation in the developing nervous system. J. Exp. Biol. 153:ll-27. Watt, C.B., and V.J. Florack (1993) Double label analyses of the coexistence of somatostatin with GABA and glycine in amacrine cells of the larval tiger salamander retina. Brain Res. 61 7:131-137. Watt, C.B., P.A. Glazebrook, and V.J. Florack (1994) Localization of substance P and GABA in retinotectal ganglion cells of the larval tiger salamander retina. Visual Neurosci. 11:355-362. Werblin, F.S., and J.E. Dowling (1969) Organization of the retina of the mudpuppy, Necturus maculosus. 11. Intracellular recording. J. Neurophysiol. 32t339-355. Wong-Riley, M.T.T. (1974) Synaptic organization of the inner plexiform layer in the retina of the tiger salamander. J. Neurocytol. 3:l-33. Wu, S.M. (1986) Effects of gamma-aminobutyric acid on cones and bipolar cells of the tiger salamander retina. Brain Res. 365:70-77. Wu, S.M. (1992) Functional organization of GABAergic circuitry in ectotherm retinas. Prog. Brain Res. 90:93-106. Yang, C.-Y., and S.Yazulla (1988a) Light microscopiclocalization of putative glycinergic neurons in the larval tiger salamander retina by immunocytochemical and autoradiographical methods. J. Comp. Neurol. 272:343-357. Yang, C.-Y., and S. Yazulla (198813) Localization of putative GABAergic neurons in the larval tiger salamander retina by immunocytochemical and autoradiographic methods. J. Comp. Neurol. 277:96-108. Yang, C.-Y., and S. Yazulla (1994a) Glutamate and GABA immunoreadivities in retinal bipolar cells of the tiger salamander. Vis. Neurosci. 11:1193-1204. Yang, C.-Y.,and S. Yazulla (1994b) Glutamate immunoreactivity in the tiger salamander retina differentiates between GABA-IR and glycine-IR amacrine cells. Invest. Ophthalmol. Vis. Sci. Suppl. 351361. Yang, C.-Y., Z . 3 . Lin, and S. Yazulla (1992) Localization of GABAAreceptor subtypes in the tiger salamander retina. Vis. Neurosci. 8:57-64. Yazulla, S., and C.-Y. Yang (1988) Colocalization of GABA- and glycineimmunoreactivities in a subset of retinal neurons in tiger salamander. Neurosci. Lett. 95:3741. Zheng, J.Q., M. Felder, J.A. Connor, and M.-m. Po0 (1994) Turning of nerve growth cones induced by neurotransmitters. Nature 368t140-144. P.G. Haydon (1990) Target dependent induces in an identified motoneuron during synaptogenesis. Dev. Biol. 138:202-213. Zwimpfer, T.J., A.J. Aguayo, and G.M. Bray (1992) Synapse formation and preferential distribution in the granule cell layer by regenerating retinal ganglion cell axons guided to the cerebellum of adult hamsters. J. Neurosci. 12: 1144-1 159.