ABSTRACT
Previous in vitro studies have convincingly demonstrated the involvement of diffusible factors in the regulation of photoreceptor development. We now provide evidence that ciliary neurotrophic factor (CNTF) represents one of these regulatory molecules. In low density monolayer cultures prepared from embryonic day 8 chick retina, photoreceptor development was studied using the monoclonal antiopsin antibody rho-4D2 as a differentiation marker. The number of cells aquiring opsin immunoreactivity, determined after 3 days in vitro, was increased up to 4-fold in the presence of CNTF to maximally 10.5% of all cells. Basic fibroblast growth factor or taurine both of which have been reported to stimulate opsin expression in rat retinal cultures and other neurotrophic factors tested (nerve growth factor, brain derived neurotrophic factor) had no effect. The EC50 of the CNTF effect (2.6 pM) was virtually identical to that measured for other CNTF receptor mediated cellular responses. Conditioned medium produced by cultured retinal cells (most likely glial cells) exhibited opsin stimulating activity identical to that of CNTF. Stimulation of opsin expression was specific for morphologically less mature photoreceptors and obviously restricted to rods, since changes in the number of identifiable cone photoreceptors expressing opsin immunoreactivity (10% of all cones) were not detectable. Measurement of the kinetics of the CNTF reponse revealed that the factor acted on immature opsin-negative progenitors and that CNTF effects were unlikely to reflect enhanced cell survival. Proliferation of photoreceptors was also unaffected, as demonstrated by [3H]thymidine autoradiography. With prolonged culture periods a gradual decrease in the number of opsin-positive cells was observed both in controls and in the continuous presence of CNTF. This decrease could be partly prevented by the addition of 1 mM taurine. Our results suggest that CNTF acted as an inductive signal for uncommitted progenitor cells or during early stages of rod photoreceptor differentiation, whereas other extrinsic stimulatory activities seemed to be required for further maturation.
INTRODUCTION
The vertebrate retina, like other parts of the brain, is composed of a variety of neuronal cells types which are generated from a supposedly homogeneous population of neuroepithelial cells. Due to its relatively simple laminar structure and its experimental accessibility, the morphological and functional properties of retinal neurons are well characterized compared to other regions of the central nervous system (Wässle and Boycott, 1991). Therefore, the retina is a preferred model system for investigating the regulatory mechanisms underlying the generation of the multiplicity of phenotypes from neuronal precursor cells.
Birthdating studies in different species have demonstrated that the major retinal cell classes (ganglion, amacrine, bipolar and horizontal cells, rod and cone photoreceptors and Müller glia) are born in a fixed temporal sequence during development, although with considerable overlap between the different cell types (for review see Altshuler et al., 1991; Reh, 1991). Analysis of the progeny of single proliferating precursor cells has shown that these progenitor cells remain multipotent even late in development and can give rise to many different retinal cell types (Turner and Cepko, 1987; Holt et al., 1988; Wetts and Fraser, 1988). Based on these findings and on those from in vitro studies (Adler and Hatlee, 1989; Reh and Kljavin, 1989; Sparrow et al., 1990), the concept emerged that cell fate in the retina is largely lineage-independent and that it is the changing microenvironment which determines both the developmental potency of the progenitor cells and the cellular phenotype expressed after terminal mitosis (Altshuler et al., 1991; Harris, 1991; Reh, 1991).
Results from recent in vitro studies using opsin immunoreactivity as a marker for photoreceptor development in rodents have provided convincing evidence for the importance of environmental signals during the determination and differentiation of retinal phenotypes. In aggregate cultures but not in dissociated cultures interactions via cell contacts with appropriate neighbours promoted rod development (Reh, 1992; see also Harris and Messersmith, 1992). By co-culturing retinal cells of different age Watanabe and Raff (1990, 1992) demonstrated that expression of the rod-specific protein rhodopsin was stimulated by a developmentally regulated diffusible factor. Very similar results were obtained by Altshuler and Cepko (1992) using a gel culture system in which cell-cell contacts were minimized. In these cultures expression of several markers of rod differentiation was dependent on cell density and the production of the rod promoting activity was temporally correlated with rod development in vivo.
To understand the mechanisms underlying these regulatory cellular interactions and to study their in vivo relevance requires the identification of the active molecules. Basic fibroblast growth factor (bFGF) has been reported to support rod differentiation in vitro (Hicks and Courtois, 1992) and rod regeneration in vivo (Faktorovich et al., 1992). Recently, it has been shown that retina conditioned media and retinal extracts contain at least two separable activities with stimulatory and inhibitory effects on rod development and one of the active (stimulatory) molecules has been demonstrated to be taurine (Altshuler et al., 1993). While these results indicate that more than one extrinsic signal might be involved in the control of photoreceptor development, cone differentiation in monolayer cultures from chick retina was reported to occur in the absence of environmental signals as a developmental default pathway (Adler and Hatlee, 1989).
In other neuronal systems several protein factors have been described which influence neuronal determination and/or differentiation (for review see Yamamori, 1992). Ciliary neurotrophic factor (CNTF) which originally had been identified as a neurotrophic factor for ciliary ganglion neurons and purified from chick eye tissue (Barbin et al., 1984) represents one of these proteins. Besides its action as a trophic molecule on a variety of neurons, CNTF also acts as a neuronal differentiation factor (Yamamori, 1992). It stimulates the expression of phenotypic markers like low-affinity neurotrophin receptor, choline acetyltransferase or tyrosine hydroxylase in selected CNS neurons in vivo and in vitro (reviewed by Burnham et al., 1994) and, interestingly, it inhibits proliferation of sympathetic precursor cells and induces a change from adrenergic to cholinergic phenotype, indicating a potential function during the process of phenotypic determination (Ernsberger et al., 1989). In the rat, CNTF and its receptor are expressed during development in many regions of the CNS including the retina (Stöckli et al., 1991; Ip et al., 1993; Kirsch and Hofmann, 1994). In monolayer cultures from chicken retina CNTF has been reported to stimulate the survival of ganglion cells (Lehwalder et al., 1989) and the differentiation of cholinergic amacrine cells and evidence has been provided that retinal cells produce a CNTF-like molecule both in vivo and in vitro (Hofmann, 1988a,b).
In the present study, we have investigated the effects of CNTF on photoreceptor development in low density monolayer cultures from embryonic chick retina using opsinimmunoreactivity as a differentiation marker. The monclonal antibody (rho-4D2) used has been described to specifically recognize rods in bovine, frog and newt retina (Molday, 1989; Bugra et al., 1992), but, as shown here, staines rods and a selected population of cones in chick retina both in vivo and in vitro. In culture, CNTF is demonstrated to stimulate opsin expression by rods most likely acting by inducing differentiation in immature progenitor cells.
MATERIAL AND METHODS
Cell cultures
Fertilized eggs were incubated in a humidified egg incubator at 37°C. Embryos were staged according to Hamburger and Hamilton (Fawcett and O’Leary, 1985) and used for culture preparation after 8 days of incubation at stages 33-35. Retinae were carefully dissected free of pigment epithelium and then incubated for 10 minutes in Ca2+/Mg2+-free Hanks’ balanced salt solution (HBSS). The tissue was treated with 0.25% trypsin (Sigma) in HBSS for 20-25 minutes at 37°C. The enzyme was inactivated by washing with culture medium containing 5% fetal calf serum (FCS). Retinae were washed once with culture medium and then dissociated by gentle trituration through a flamenarrowed glass pipette. Cells were seeded on poly-L-lysine (0.1 mg/ml) coated coverslips (14 mm diameter placed in a 24-multiwell plate) at a density of 100,000 cells/cm2 in 600 μl. Culture medium consisted of Dulbecco’s modified Eagle medium (DMEM) supplemented with 2 mM glutamine, 100 U/ml penicillin, 100 U/ml streptomycin and 1% heat-inactivated FCS. Factors and conditioned media were added before seeding. Cultures were grown for 1-6 days at 37°C in 5% CO2/95% air. To determine the effects of CNTF and other factors cultures were routinely grown for 3 days, since this period was found sufficient to measure maximum stimulation by CNTF. Recombinant rat CNTF and BDNF were kindly supplied by Regeneron Pharmaceuticals (Tarrytown, NY). Basic FGF was purchased from Boehringer (Mannheim) and NGF was purified from mouse submaxillary glands (Suda et al., 1978). Biological activity of the factors was tested by survival assays using cultures of appropriate peripheral neurons.
For the preparation of retina-conditioned medium (CM), retinal cells were seeded at a density of 200,000 cells/cm2 on poly-L-lysine coated wells in DMEM with 10% FCS. Every third day 50% of the culture medium was replaced by fresh medium. After about 14 days in vitro (DIV) flat non-neuronal cells had grown to confluency, while the number of neurons had considerably decreased. At that time the culture medium was completely replaced with DMEM containing N2 supplements (Bottenstein and Sato, 1979). Three days later the conditioned medium was removed, centrifuged for 5 minutes at 1000 g and stored frozen at −20°C before use.
Immunocytochemistry
For immunolabeling coverslips were fixed either in 4% paraformaldehyde alone or with additional 0.1% of glutaraldehyde, rinsed and preincubated for 30 minutes in 0.5% Triton X-100 in 0.1 M phosphate buffer (PB), pH 7.35 with 10% normal goat serum. They were then successively incubated with primary antibodies diluted in PB with 1% normal goat serum overnight at 4°C: rho-4D2, 1:50; anti-calbindin, 1:5000; anti-GABA, 1:8000. Antibody was visualized using biotinylated goat anti-mouse or goat anti-rabbit antibodies and FITC-conjugated avidin. The rho-4D2 antibody raised against bovine rhodopsin was a generous gift of Dr R. S. Molday (for characterization of the antibody see Molday, 1989). The monoclonal anti-calbindin antibody was kindly supplied by Dr M. R. Celio. The anti-GABA antibody was raised in rabbit against bovine serum albumin-conjugated GABA. Cultures were mounted in potassium iodide/glycerol and viewed with a Zeiss Axiovert equipped with epifluorescence optics.
Immunostaining of cryostat sections (8-12 μm) or retinal whole mounts immersion fixed in 4% paraformaldehyde for 1-2 hours was performed as described above for the cultures except that pretreatment was in 1% Triton X-100 for 1 hour and the primary antibody solution contained 0.1% Triton X-100.
For combined immunocytochemistry and autoradiographic visualization of thymidine incorporation, cultures were grown in the presence of [3H]thymidine (Amersham; 0.5 μCi/ml; 5 Ci/mmol) for 3 days and then processed for immunocytochemistry as described above. Coverslips were glued onto glass slides, dipped in photoemulsion (NTB 2, Kodak) and exposed in light-tight boxes for 3-6 days at 4°C. The slides were developed (D-19, Kodak) fixed, washed and mounted for microscopic observation.
For immunoelectron microscopy, pieces of the central part of adult retinae were fixed as above, preincubated with 0.1% Triton for 30 minutes, washed in bidest for 5 minutes, incubated for 30 minutes in 0.3% H2O2 in methanol and sequentially incubated in blocking solution (10% normal goat serum in PB), primary antibody (rho-4D2, 1:100), biotinylated secondary antibody (1:100) and avidin peroxidase (1:200; Sigma). Antibody binding was visualized using diaminobenzidine (0.05%) as substrate. The tissue was then postfixed in 1% osmium, dehydrated and embedded in Epon. Ultrathin sections, stained with uranyl acetate and lead citrate, were examined in a Zeiss 109 electron microscope.
Cell counting
To determine the number of cells surviving in culture and the percentage of cells expressing rho-4D2-immunoreactivity, the total number of cells and the number of immunopositive photoreceptors was counted along two diameters of each coverslip in some experiments. A minimum of 2800 cells per coverslip was evaluated in this way. In most experiments only the number of immunopositive cells was determined and cultures stimulated under various conditions were compared to control cultures which were set as 100%. To determine the number of [3H]thymidine incorporating cells differentiating into opsin-positive cells in vitro, at least 200 immunostained cells per coverslip were evaluated with respect to autoradiographic labeling. All determinations were performed in triplicate and each experiment was repeated at least twice. Differences between measurements, when mentioned in results were statistically significant with P<0.01 (Student’s t-test), if not otherwise indicated.
Measurement of transmitter uptake
The uptake of [3H]glutamate, [3H]GABA and [3H]glycine was measured by incubating the cultures with the respective radioactive amino acid (0.1 μM) dissolved in Hepes buffered Ringer’s solution for 10 minutes at 37°C (Möckel et al., 1994). After incubation cultures were rapidly washed several times with incubation medium and dissolved in 0.5% sodium dodecyl sulphate. Cell associated radioactivity was determined by liquid scintillation counting.
RESULTS
Rho-4D2-immunoreactvity is expressed by rods and a subpopulation of cones in chicken retina
As a marker for photoreceptor differentiation, we used the monoclonal antibody rho-4D2 which binds to the N-terminal part of the bovine rhodopsin molecule and has been demonstrated to specifically recognize rod photoreceptors in different species including pig and frog (Molday, 1989). It has also been used previously in rodents to study rod development in vitro (Reh, 1992; Hicks and Courtois, 1992). We have examined the expression of rho-4D2-immunoreactivity in the developing chicken retina before using the antibody for in vitro studies.
During in ovo development opsin immunoreactivity became detectable at embryonic day 14 (E14) and increased during further development (Fig. 1). At earlier stages (E14/E15) the entire cell membrane of many immature photoreceptors was uniformly stained including the basal process ending in the outer plexiform layer (Fig. 1A,B). Part of the immunoreactive cells at this stage possessed an intensily stained short apical process representing a rudimentary outer segment. With ongoing maturation immunoreactivity became increasingly restricted to this cellular compartment and in the adult, with the exception of a few cells remaining uniformly labeled, it was confined to mature outer segments which were of varying length and shape (Fig. 1C,D). Immunocytochemistry on retinal whole-mount preparations (not shown) and electron microscopic examination of stained sections confirmed that in chicken, in addition to rods, a subpopulation of cones was recognized by the rho-4D2 antibody which was in contrast to observations in other species. An opsin-positive rod flanked by two negative cones is shown in Fig. 1E. The immunoreactive cone in Fig. 1F is identifiable by a shorter outer segment as compared to the neighbouring rod and by the characteristic oil droplet located in the outer part of the inner segment. The rho-4D2-positive cones most likely represent green sensitive cones, since the protein moiety of the photopigment of this cone population shares more than 70% homology to the mammalian rhodopsin (Okano et al., 1992). In particular, the N-terminal part of the two proteins which is recognized by rho-4D2 is nearly identical.
In situ development and specificity of opsin immunoreactivity in the chicken retina. (A-D) Immunofluorescence staining was carried out on cryostat sections from E14 (A), E15 (B), E18 (C) and adult chicken retinae using rho-4D2 as primary antibody. At early stages opsin immuoreactivity is observed over the entire cell membrane (A). With ongoing development opsin expression becomes increasingly restricted to the developing outer segment (B-D). (E,F) Electron micrographs of immunostained adult chicken retina. An immunopositive rod with two neighbouring negative cones is visible in E. A subpopulation of cones with characteristic oil droplets was also found to be labeled (center cell in F). Scale bars in A-D, 10 μm; E, F, 1 μm.
In situ development and specificity of opsin immunoreactivity in the chicken retina. (A-D) Immunofluorescence staining was carried out on cryostat sections from E14 (A), E15 (B), E18 (C) and adult chicken retinae using rho-4D2 as primary antibody. At early stages opsin immuoreactivity is observed over the entire cell membrane (A). With ongoing development opsin expression becomes increasingly restricted to the developing outer segment (B-D). (E,F) Electron micrographs of immunostained adult chicken retina. An immunopositive rod with two neighbouring negative cones is visible in E. A subpopulation of cones with characteristic oil droplets was also found to be labeled (center cell in F). Scale bars in A-D, 10 μm; E, F, 1 μm.
Opsin expression in culture was enhanced specifically by CNTF
Under the conditions used (low serum concentration, 1000 cells/mm2) most of the cultured cells dissocated from E8 chick retinae grew isolated, and up to 1 week in vitro the number of flat non-neuronal cells remained low (<4%). Although the cells tended to form small clusters after some days, no organized aggregates formed and direct cell contacts were minimal. Cell survival determined after 3 DIV was 75% ± 11 of the original seeding density in control cultures (n=14; 5 independent experiments) and 75% ± 13 in CNTF treated cultures (n=16). After 6 DIV survival was 65% ± 9 (n= 8; 3 independent experiments) and 63% ± 8 (n=9), respectively. When cultures were immunostained with rho-4D2 after 3 DIV, positive cells were already present as shown in Fig. 2A. This was several days earlier than expected from the in vivo time course of opsin expression (see also Fig. 5). In cultures grown in the presence of CNTF (5 ng/ml) the number of opsin-positive cells was significantly increased (Fig. 2B). As shown in Fig. 3A CNTF increased the number of immunoreactive cells by about 300%. In different experiments the percentage of positive cells in control cultures varied between 1.6 and 4.4% and CNTF treatment resulted in a 2.3-to 4-fold increase. Other neurotrophic factors which have been shown to be expressed in the retina (bFGF, NGF and BDNF) had no effect (Fig. 3A). Taurine has been identified as one of the active components in rat retina conditioned medium which effectively stimulate opsin expression in low density cultures from rat retina (Altshuler et al., 1993). In chick retinal cultures, taurine in concentrations up to 10 mM was without any effect on the number of opsin-positive cells present after 3 DIV (Fig. 3B). Stimulation by CNTF was dose dependent and half maximal activity was measured at a concentration of 0.06 ng/ml corresponding to 2.6 pM (Fig. 3C). This value was virtually identical to that determined by the ciliary ganglion cell survival assay (Barbin et al., 1984; data not shown) indicating that both effects were mediated by the same receptor-coupled mechanism.
Immunofluorescence staining with rho-4D2 in control and CNTF treated retinal cultures. (A) Control cultures immunolabeled after 3 DIV. (B) Cultures grown for 3 days in the presence of CNTF (5 ng/ml). Scale bars, 10 μm.
Opsin expression in chick retinal cultures is specifically enhanced by CNTF. Cultures were grown under the conditions indicated and the number of opsin-positive cells was determined after 3 DIV. The number of immunoreactive cells in control cultures (Con) was set as 100%. (A) Effect of various neurotrophic proteins which have been described to be expressed in the retina; CNTF (10 ng/ml), basic fibroblast growth factor (bFGF, 10 ng/ml), nerve growth factor (NGF, 50 ng/ml), brain derived neurotrophic factor (BDNF, 20 ng/ml). (B) Taurine at concentrations up to 10 mM did not promote opsin expression. (C)Concentration dependence of the CNTF effect (EC50: 62 pg/ml). Values represent means ± s.d. of 3-4 determinations of one representative experiment.
Opsin expression in chick retinal cultures is specifically enhanced by CNTF. Cultures were grown under the conditions indicated and the number of opsin-positive cells was determined after 3 DIV. The number of immunoreactive cells in control cultures (Con) was set as 100%. (A) Effect of various neurotrophic proteins which have been described to be expressed in the retina; CNTF (10 ng/ml), basic fibroblast growth factor (bFGF, 10 ng/ml), nerve growth factor (NGF, 50 ng/ml), brain derived neurotrophic factor (BDNF, 20 ng/ml). (B) Taurine at concentrations up to 10 mM did not promote opsin expression. (C)Concentration dependence of the CNTF effect (EC50: 62 pg/ml). Values represent means ± s.d. of 3-4 determinations of one representative experiment.
The differentiation of rod photoreceptors in rat retinal cultures has been shown to critically depend on cell density, indicating the presence of stimulatory cellular interactions by diffusible molecules (Altshuler and Cepko, 1992). In our cultures, increasing the cell density by a factor of three did not result in an enhanced appearance of opsin-positive cells in control cultures (Fig. 4A). Thus, there was no evidence for the production of a CNTF-like activity in the largely glia-free neuronal cultures. However, when conditioned medium (CM) produced by 14-to 17-day old confluent glial cultures (containing a minor portion of neurons) was added to cultures of normal density at a concentration of 10%, stimulation of opsin expression was identical to that observed in the presence of saturating concentrations of CNTF (Fig. 4B). The effect of the CM was dose dependent and its activity was half maximal at 100-fold dilution. Again, this specific activity was virtually identical to that for the survival of ciliary ganglion neurons (Hofmann, 1988b) supporting the assumption that cultured retinal cells – most likely glial cells – produce a CNTF-like molecule.
Effect of cell density and retina conditioned medium (CM) on opsin expression. (A) Number of opsin positive cells in cultures grown for 3 days at the indicated seeding density. Increasing the the cell density did not significantly influence the percentage of cultured cells expressing opsin. (B) Dose dependence of the effect of CM on opsin expression. Medium conditioned by cultures grown for 2 weeks at higher density (see Methods) was added before seeding at the indicated concentrations. In the same experiment CNTF (5 ng/ml) enhanced the number of opsin-positive cells to 375% of the controls. This was identical to the maximal CM effect (at a concentration of 10%). Values in A represent means ± s.d. of triplicate determinations; means of duplicate determinations are given in B.
Effect of cell density and retina conditioned medium (CM) on opsin expression. (A) Number of opsin positive cells in cultures grown for 3 days at the indicated seeding density. Increasing the the cell density did not significantly influence the percentage of cultured cells expressing opsin. (B) Dose dependence of the effect of CM on opsin expression. Medium conditioned by cultures grown for 2 weeks at higher density (see Methods) was added before seeding at the indicated concentrations. In the same experiment CNTF (5 ng/ml) enhanced the number of opsin-positive cells to 375% of the controls. This was identical to the maximal CM effect (at a concentration of 10%). Values in A represent means ± s.d. of triplicate determinations; means of duplicate determinations are given in B.
To investigate whether CNTF has a general differentiation or survival promoting effect on cultured retinal neurons, we studied the expression of other phenotypical markers. As shown in Table 1, CNTF did not enhance the number of GABA immunoreactive amacrine cells (see Huba and Hofmann, 1990), peanut lectin-binding cone photoreceptors (Blanks and Johnson, 1984) or calbindin-positive neurons (probably sub-populations of amacrine cells, bipolar cells or cones; see Ellis et al., 1991). Similarly, the high affinity uptake of various radiolabeled neurotransmitters (GABA, glutamate, glycine) which can be used to characterize the development of different cell populations in retinal cultures (Huba and Hofmann, 1990; Möckel et al., 1994) was not increased in CNTF-treated as compared to control cultures after 3 DIV.
CNTF promoted early stages of photoreceptor development
Most of the cultured opsin-positive cells were oval or more elongated in shape with a single neuritic process growing out from one of the cell poles. Both in CNTF-treated and untreated cultures three types of immunoreactive cells were distinguishable with respect to their staining pattern (Fig. 5A-E). One type exhibited a uniformly labeled cell membrane, occasionally with a cytoplasmic condensation of immunoreactivity located opposite the neuritic process (type 1; Fig. 5A,C). In the second type an intensely stained short process had formed emanating from the apical cell pole (type 2; Fig. 5A,D). This process obviously represented a rudimentary outer segment similar to those that have already been described as being produced by cones differentiating in dissociated cultures, similar to ours (Adler, 1986). Labeled photoreceptors of the third type were characterized by immunonegative pericaria and neuritic processes with the immunoreactive material being entirely confined to the outer segment-like process (type 3; Fig. 5A,E-G). Since these different labeling patterns resembled those occurring during in vivo development, as illustrated in Fig. 1, we interpreted them as increasingly mature stages of differentiating photoreceptors. To examine whether CNTF promoted photoreceptor differentiation by stimulating the transition between the developmental stages, we separately determined the number of cells exhibiting the different staining patterns. As shown in Fig. 5H, each of the different types made up roughly one third of the stained cells in control cultures after 3 DIV and the proportions did not change significantly after prolonged culture periods (not shown). Addition of CNTF markedly stimulated the appearance of labeled type I and II photoreceptors, but had absolutely no effect on the number of type III cells. Thus, CNTF apparently had no general effect on survival or differentiation of developing photoreceptors, but seemed to promote the generation of immature photoreceptors from opsin-negative precursor cells.
Distinct cellular staining patterns of opsin-positive cells correlate with responsiveness to CNTF. (A) Fluorescence micrograph of an immunolabeled control culture demonstrating the three different patterns of opsin expression observed in vitro. In B the same visual field is shown under phase contrast optics. Large arrowhead indicates an immunonegative cone photoreceptor with an oil droplet, small arrowhead shows a typical opsin-positive process of a photoreceptor cell (type 3). The different staining patterns visible in A are shown at higher magnification in C-F. (C) Type 1 cell with immunoreactivity located on the entire cell membrane and (unevenly distributed) in the cytoplasm. (C)Type 2 cell which additionally possesses an intensily stained apical process. (E) Type 3 cell with immunoreactivity entirely confined to its presumed rudimentary outer segment. (F,G) Another example of a type 3 cell photographed using fluorescence (F) and phase contrast (G) optics. Here the immunostained ‘outer segment’ lying in the plane of focus is clearly recognizable as an appendage of the cell labeled with an arrowhead in G. Scale bars in A-G, 10 μm. (H) Effect of CNTF on the different cell types defined in C-G. Cultures were grown for 3 days in the absence (open bars) or presence (hatched bars) of CNTF (10 ng/ml) and the percentage of immunoreactive cells exhibiting the different staining patterns was determined separately. Values are given as means ± s.d. (n = 3).
Distinct cellular staining patterns of opsin-positive cells correlate with responsiveness to CNTF. (A) Fluorescence micrograph of an immunolabeled control culture demonstrating the three different patterns of opsin expression observed in vitro. In B the same visual field is shown under phase contrast optics. Large arrowhead indicates an immunonegative cone photoreceptor with an oil droplet, small arrowhead shows a typical opsin-positive process of a photoreceptor cell (type 3). The different staining patterns visible in A are shown at higher magnification in C-F. (C) Type 1 cell with immunoreactivity located on the entire cell membrane and (unevenly distributed) in the cytoplasm. (C)Type 2 cell which additionally possesses an intensily stained apical process. (E) Type 3 cell with immunoreactivity entirely confined to its presumed rudimentary outer segment. (F,G) Another example of a type 3 cell photographed using fluorescence (F) and phase contrast (G) optics. Here the immunostained ‘outer segment’ lying in the plane of focus is clearly recognizable as an appendage of the cell labeled with an arrowhead in G. Scale bars in A-G, 10 μm. (H) Effect of CNTF on the different cell types defined in C-G. Cultures were grown for 3 days in the absence (open bars) or presence (hatched bars) of CNTF (10 ng/ml) and the percentage of immunoreactive cells exhibiting the different staining patterns was determined separately. Values are given as means ± s.d. (n = 3).
Since rho-4D2 recognized both rods and cones in situ, it remained unclear which of the two cell types responded to CNTF. To address this question we examined the effect of the factor on the appearance of identifiable cone photoreceptors in monolayer cultures. Adler and coworkers have extensively studied the morphological differentiation of chick retinal cones which takes place in dissociated cultures in the absence of added factors (Adler et al., 1984; Adler, 1986). On the light microscopic level, differentiated cones are identifiable by their elongated and polarized morphology and, in particular, by the presence of an intracellular ‘oil droplet’ which is characteristic for avian cones, but is not produced by rods (see Fig. 1F). Applying these criteria, it was possible to count the number of differentiated cone-like cells after 6 DIV. Less differentiated cones (without oil droplets) may have escaped identification. As shown in Table 2, cone-like cells made up 10% of all cells and this fraction was not influenced by CNTF. Of this cell population only 10% were immunopositive, demonstrating that only a minor fraction of cone photoreceptors is recognized by rho-4D2. CNTF did not increase the proportion of opsinpositive cones as would be expected, if a substantial part of the CNTF responsive cells had differentiated into cones. In control cultures more than half (56.0% ± 3.5) of the opsin-positive cells showed cone-like morphology. With the increase of the total number of opsin-positive cells caused by CNTF the proportion with cone-like morphology decreased (28.7% ± 3.8) reflecting an unchanged number of opsin-immunoreactive cones in the presence of CNTF (as can be demonstrated by calculation). Thus, the results summarized in Table 2 suggested that CNTF did not stimulate cone development, but exclusively affected rod progenitors. This notion was supported by the observation that the number of cells binding peanut lectin which supposedly is a specific marker of cone photoreceptor cells in the retina was unaltered in the presence of CNTF (Table 1).
CNTF has no detectable effect on morphologically identified cone photoreceptors differentiating in vitro

Results on the time course of the CNTF effect on opsin expression demonstrated in Fig. 6 supported the conclusion that CNTF was effective at early stages of photoreceptor development. Immunoreactivity was first detectable after 2 DIV independent of the presence of CNTF and increased between day 2 and 3 in vitro (Fig. 6A). No further increase was observed in control cultures. Comparison with Fig. 1 reveals that opsin expression started earlier in vitro than in vivo by about 4 days. Such premature differentiation has been ascribed to the process of dissociation which has been shown to attenuate cell proliferation and to promote phenotypic development in monolayer cultures from immature nervous tissue (Reh and Kljavin, 1989; Ahmed and Fellows, 1987). CNTF apparently did not accelerate differentiation, since the time course of opsin expression during the first 3 DIV in treated cultures was very similar to that in controls. After 3 days a maximum of 7-10.3% of all cells was opsin-positive in CNTF-treated cultures. Fig. 6B demonstrates that the continuous presence of CNTF was not necessary to achieve maximum stimulatory effects during the first 3 DIV. When the factor was added at the time of seeding and removed by changing the culture medium 24 hours later, i.e., before the appearence of the first immunoreactive positive photoreceptors, stimulation of immunoreactivity observed after 3 DIV was identical to that in cultures stimulated permanently for 3 days. With CNTF added for 24 hours at later stages (after 24 and 48 hours, respectively) there was a slight decrease in the effectiveness of stimulation. These results showed that it was opsin-negative cells that responded to CNTF and that a transient action of CNTF on these precursors was sufficient to cause an increase in the number of cells expressing opsin. The observation that addition of CNTF after 48 hours in vitro resulted in a significant stimulation of opsin expression only 24 hours later contrasted with results presented in Fig. 6A. There, CNTF was unable to induce a detectable opsin expression within 24 hours, when added before seeding. This suggests that cultured opsinnegative cells underwent developmental changes, altering their reponses to CNTF. Between 3 and 6 DIV, even in the continued presence of CNTF, the number of immunoreactive cells slowly decreased by about 50% (Fig. 7). This means either that opsin-expressing photoreceptors died or that the presence of CNTF was not sufficient to maintain their state of differentiation. The same was true, when CM instead of CNTF was used to stimulate opsin expression (not shown). In control cultures the same relative decrease of opsin-positive cells occurred between 3 and 6 DIV supporting the conclusion that differentiated photoreceptors required additional factors for survival or maintenance of differentiation. When taurine, which had no effect on opsin expression during the first 3 DIV (Fig. 3B), was added after 3 DIV to CNTF-stimulated cultures either together with CNTF or alone, the loss of immunoreactive cells was attenuated (Fig. 7). Taurine prevented only about 30% of the decrease in immunoreactivity. The effect, however, was significant and reproducible. Apparently, photoreceptor precursors became responsive to taurine after they had differentiated into opsin-positive cells.
Kinetics of responses to CNTF. (A) Time course of opsin expression in control (open bars) and in CNTF-treated (hatched bars) cultures. CNTF (5 ng/ml) was added before seeding. Cultures were grown for the indicated periods and the number of opsin-positive cells was counted. (B) Responses to CNTF present at different intervals during a 3 day culture period; Con, no CNTF added. (A) CNTF treatment for 3 days; (B) CNTF was present from 0-24 hours, cultures were then thoroughly washed twice and then grown for two more days in the absence of the factor; (C) CNTF added after 24 hours in vitro removed after 48 hours; (D) CNTF added after 48 hours. All cultures were fixed and immunostained after 72 hours. Values are given as means ± s.d. (n = 3).
Kinetics of responses to CNTF. (A) Time course of opsin expression in control (open bars) and in CNTF-treated (hatched bars) cultures. CNTF (5 ng/ml) was added before seeding. Cultures were grown for the indicated periods and the number of opsin-positive cells was counted. (B) Responses to CNTF present at different intervals during a 3 day culture period; Con, no CNTF added. (A) CNTF treatment for 3 days; (B) CNTF was present from 0-24 hours, cultures were then thoroughly washed twice and then grown for two more days in the absence of the factor; (C) CNTF added after 24 hours in vitro removed after 48 hours; (D) CNTF added after 48 hours. All cultures were fixed and immunostained after 72 hours. Values are given as means ± s.d. (n = 3).
Taurine partly prevents the reduction of opsin expression observed in control and CNTF-treated cultures during prolonged culture periods. Cultures were grown either for 3 days (3 DIV) or 6 days (6 DIV) under different stimulation conditions as indicated. Con, no additives; CNTF, continuous presence of CNTF (5 ng/ml); Tau, 1 mM taurine present for 6 DIV; CNTF→Tau: sequential stimulation with CNTF (0-3 DIV) and taurine (3-6 DIV); CNTF+Tau, continuous presence of CNTF (0-6 DIV) with taurine added after 3 DIV. Both in control and CNTF-treated cultures the number of opsin-positive cells declined between 3 and 6 DIV. This decline was partly prevented by taurine, independent of the presence of CNTF. Values represent means ± s.d. (n = 3). **, significantly different from CNTF-treated cultures after 6 DIV (P<0.05).
Taurine partly prevents the reduction of opsin expression observed in control and CNTF-treated cultures during prolonged culture periods. Cultures were grown either for 3 days (3 DIV) or 6 days (6 DIV) under different stimulation conditions as indicated. Con, no additives; CNTF, continuous presence of CNTF (5 ng/ml); Tau, 1 mM taurine present for 6 DIV; CNTF→Tau: sequential stimulation with CNTF (0-3 DIV) and taurine (3-6 DIV); CNTF+Tau, continuous presence of CNTF (0-6 DIV) with taurine added after 3 DIV. Both in control and CNTF-treated cultures the number of opsin-positive cells declined between 3 and 6 DIV. This decline was partly prevented by taurine, independent of the presence of CNTF. Values represent means ± s.d. (n = 3). **, significantly different from CNTF-treated cultures after 6 DIV (P<0.05).
One way by which CNTF could have increased the number of precursor cells acquiring opsin-immunoreactivity, was by stimulating the proliferation of progenitor cells or by inducing differentiation in proliferating precursors. Therefore, we carried out double labeling experiments ([3H]thymidine autoradiography and rho-4D2 immunocytochemistry) to examine whether the fraction of [3H]thymidine-incorporating progenitors that started to express opsin-immunoreactivity during the culture period was increased by CNTF. As demonstrated in Fig. 8, the number of opsin-positive cells showing autoradiographic labeling was very small (<2%) both in treated and untreated cultures, whereas about one third of all cultured cells had incorporated [3H]thymidine (Table 3). This means that virtually all of the cells that expressed opsin immunoreactivity in response to CNTF had undergone their last mitotic cycle in vivo. The total number of [3H]thymidine-incorporating neurons was slightly but significantly increased in the presence of CNTF (Table 3) indicating that there were other CNTF responsive cell populations. So far, however, we have no information on the developmental fate of these cells.
Culture double labeled for opsin immunoreactivity and [3H]thymidine autoradiography. CNTF stimulated cultures grown for 3 days in the presence of [3H]thymidine (0.5 μCi/ml) were immunostained and then processed for autoradiography. (A) Fluorescence micrograph demonstrating opsin immunoreactivity. (B) Phase contrast micrograph showing [3H]thymidine incorporating cells. Large arrowhead indicating an immunopositive, autoradiographically negative cell; small arrowhead indicating an opsin-negative, [3H]thymidine incorporating cell. Scale bars, 10 μm.
Culture double labeled for opsin immunoreactivity and [3H]thymidine autoradiography. CNTF stimulated cultures grown for 3 days in the presence of [3H]thymidine (0.5 μCi/ml) were immunostained and then processed for autoradiography. (A) Fluorescence micrograph demonstrating opsin immunoreactivity. (B) Phase contrast micrograph showing [3H]thymidine incorporating cells. Large arrowhead indicating an immunopositive, autoradiographically negative cell; small arrowhead indicating an opsin-negative, [3H]thymidine incorporating cell. Scale bars, 10 μm.
DISCUSSION
In previous studies it has been convincingly demonstrated that diffusible factors produced by retinal cells play an important role in the regulation of photoreceptor development (Altshuler and Cepko, 1992; Watanabe and Raff, 1992; Altshuler et al., 1993). In the present study, we provided experimental evidence that CNTF represents one of these regulatory molecules. Our results show first, that CNTF, but not other factors tested, promote the appearance of opsin-positive cells in dissociated cultures from embryonic chick retina; second, that a CNTF-like activity with identical effects on photoreceptor development is produced by chick retinal cells, at least in vitro; third that CNTF acted on opsin-negative precursor cells without stimulating cell proliferation or survival.
There are several possible ways by which an exogenously added factor could cause an increase in the number of cultured cells expressing a particular phenotype. First, it could support the survival either of immature precursor cells or of the differentiated cells. Second, the effect could be on the proliferation of progenitor cells. Third, the factor could determine the developmental fate by acting on precursor cells which otherwise would not have entered the phase of phenotypic differentiation. Fourth, it could be required for the differentiation and maturation of immature cells which are already committed to a specific phenotype. In the case of the CNTF effects described here we believe that the results are most consistent with the third possibility, although, in principal, it is difficult to distinguish effects on cell determination from those on early steps of differentiation. Effects of CNTF on cell division in photoreceptor precursors can be excluded, since a negligible portion of opsin-positive cells developed in vitro from mitotically active precursors. Our experiments also provided evidence against a survival promoting activity of CNTF on photoreceptors or their precursors. CNTF enhanced the appearance of opsin-positive cells occurring at days 2 and 3 in vitro, but with prolonged time in culture the absolute number of cells dying in CNTF-treated cultures was even greater (Fig. 7). Moreover, the fact that the effectiveness was unchanged, when the factor was withdrawn after only 24 hours, and only slightly reduced when CNTF was added for only 24 hours after 2 DIV (Fig. 6B), confirmed that CNTF did not act by preventing the death of opsin-immunoreactive photoreceptors or their undifferentiated precursors. It rather demonstrated that a transient action of CNTF was sufficient to initiate opsin expression in undifferentiated precursor cells and that CNTF was not required during the actual differentiation process, reflected by the appearance of opsin immunoreactivity. Our results, however, cannot elucidate whether CNTF acted by determining the developmental fate of uncommitted precursor cells or by promoting the differentiation of immature cells already committed to become rod photoreceptors. In agreement with in vivo studies (Prada et al., 1991) we found that virtually all rod precursors were postmitotic at embryonic day 8 (Table 3). Since opsin expression and outer segment formation does not start until E14 in vivo (Fig. 1), there is a delay of at least 5-6 days between the birthdate and the visible differentiation of rods in the chick. In vitro this period was obviously reduced, but still there was a delay of at least 2 days between the final mitosis of precursor cells and the appearance of the first opsinpositive photoreceptors (Fig. 6A). It was during this phase that the transient presence of CNTF resulted in the induction of opsin expression. From in vivo cell lineage analyses it has been concluded that the developmental fate of retinal precursors is determined during or shortly after the last mitosis of precursor cells (Turner and Cepko, 1987; Wetts and Fraser, 1988; Holt et al., 1988). This would suggest that the postmitotic cells responding to CNTF in our cultures were already committed to become photoreceptors. Phenotypic determination, however, is not necessarily a single step process as demonstrated for neural crest derived neurons (Anderson, 1993) and the variability of the delay between final mitosis and differentiation possibly reflects the complexity of influences eventually resulting in the induction of differentiation. We are presently investigating, whether enhanced photoreceptor differentiation in the presence of CNTF is accompanied by a reduction in the number of other cell types. Together with a more detailed analysis of the kinetics of CNTF effects these studies should make it possible to discriminate between the hypotheses that CNTF determines (or alters) cell fate or instead acts on precursor cells already committed to becoming rods.
Beyond its classical (survival promoting) activity, CNTF has already been shown to stimulate the expression of phenotypical properties in a variety of immature neurons in vivo and in vitro (reviewed by Manthorpe et al., 1993). In addition, it can influence early developmental processes like proliferation of sympathetic and oligodendrocyte precursors or the choice of transmitter phenotype in sympathetic neurons (Saadat et al., 1989; Ernsberger et al., 1989). The effects during early stages of photoreceptor differentiation described here further support the idea of a widespread and multiple role of CNTF during neuronal development. The EC50 value (2.6 pM) of photoreceptor responses was close to that for other biological actions of rat CNTF e.g. on embryonic chick ciliary (data not shown) and motoneurons (Arakawa et al., 1990) suggesting that they were mediated by the same receptor coupled mechanism. Although it still needs to be demonstrated that the cultured photoreceptors express functional CNTFR receptors, there is substantial experimental evidence now for the expression of CNTF (or a closely related protein) and of the ligand binding α-subunit of the CNTF receptor (CNTFRα) in the retina during the phase of neuronal differentiation. First, developmentally regulated expression of both CNTF and CNTFRα in the early postnatal rat retina has been demonstrated by PCR (Ip et al., 1993; Kirsch et al., unpublished data). Second, conditioned media from chick retinal cultures and extracts from embryonic (E8-E15) chick retinae contain a protein with stimulatory activity for ciliary neurons (the classical target for CNTF), cholinergic amacrine cells and rod photoreceptors identical to that of rat CNTF (Hofmann, 1988b; present study). Third, the active molecule in retinal CM had an apparent relative molecular mass of 21×103 (Hofmann, 1988b) which is close to that of mammalian CNTF (23×103) and identical to that of a neurotrophic protein termed ‘growth promoting activity’ (GPA) which has been cloned from chick eye tissue (Leung et al., 1992). GPA shows biological activities identical to those of CNTF (Heller et al., 1993) and although it shares only 50% sequence homology with rat CNTF it probably represents its chick homologue. Fourth, the binding subunit of the chick GPA/CNTF receptor has been cloned recently and shown to mediate both CNTF and GPA effects (Heller et al., 1994). By in situ hybridization the same authors could demonstrate that the GPA/CNTF receptor is expressed in the outer nuclear layer of the retina between E8 and E12 but not at E16 (Heller and Rohrer, personal communication). This corresponds exactly with our in vitro observations.
Interestingly, GPA, in contrast to rat CNTF, was found to be secreted by cells transfected with GPA cDNA (Leung et al., 1992). This could explain the presence of CNTF-like activity in media conditioned by chick retinal cells (Fig. 4). We have not specifically studied the source of this activity. Its absence in neuron-enriched cultures (no effect of cell density) and the cellular composition of the cultures used to produce CM (predominantly flat non-neuronal cells) indicated that glial cells were the main source. This would agree with previous findings (Hofmann, 1988a) and with the localization of CNTF which is preferentially, if not exclusively, expressed in glial cells including Müller cells of the retina (Stöckli et al., 1991).
In summary, there is increasing evidence that CNTF (or GPA) is involved in the regulation of retinal development. With the chick retina as a model system it should be possible to study the in vivo relevance of the CNTF effects on neuronal development observed in cultures by applying the factor or inhibitory agents to the vitreous at defined developmental stages.
Several other molecules have been reported to promote rod photoreceptor development in dissociated cultures prepared from the rod dominated rat retina. Acidic FGF (Hicks and Courtois, 1988), bFGF (Hicks and Courtois, 1992), taurine (Altshuler et al., 1993) and retinoic acid (Kelley et al., 1994) have been described to stimulate the differentiation of rods as indicated by the appearance of opsin-immunoreactive cells. In our experiments on chick cultures bFGF and taurine were completely ineffective. Remarkable differences between the two species also exist with respect to CNTF responses. In cultures of newborn rat retinae where considerable differentiation of opsin positive photoreceptor occurred in the absence of exogenously added factors we found that CNTF very effectively inhibited the expression of opsin immunoreactivity (Kirsch et al., unpublished data). Very similar results have been obtained by Cepko and coworkers (personal communication) with a gel culture system of dissociated rat retina. The two species differ with respect to their photoreceptors in that rat retinae are rod dominated with few cones, whereas in chick retinae cones prevail. Therefore, it was tempting to presume that CNTF had opposite effects on the two types of photoreceptors, particularly since a subpopulation of chick cones was found to be recognized by the opsin antibody used in our study. However, the absence of a detectable CNTF effect on peanut lectin-binding (Table 1) and morphologically identifiable cones (Table 2) largely excludes the possibility that the factor acted on cone photoreceptors. In addition, the expression of various other phenotypic markers (Table 1) was unaltered in the presence of CNTF. Cultured chick ganglion cells – in contrast to photoreceptors – have been reported to show enhanced survival in the presence of CNTF (Lehwalder et al., 1989). These observations support the conclusion that CNTF actions on rod photoreceptors do not reflect a general induction of the differentiation of immature retinal precursors.
That the same extrinsic signals (taurine, CNTF, bFGF) result in different responses of the same cell type depending on the species is a unique observation and it will be of interest to investigate its functional significance. So far, several explanations are conceivable. Since the effects were obtained in mixed cultures the cellular composition of which differs for chick and rat retina, CNTF might have acted indirectly inducing the production of different agents which then were responsible for the photoreceptor responses. Although we cannot exclude this possibility, we believe that with the low cell density used and the short incubation times necessary to observe CNTF effects this is not very likely to be the mechanism of CNTF effects in chick cultures. The differences in photoreceptor responses could be based on differences in the functional properties of GPA/CNTF receptors in the two species. However, the identity of CNTF and GPA effects on other neurons both in mammals and in chick and the homology of the receptor proteins (Heller et al., 1994) argue against this hypothesis.
Another reasonable, though largely hypothetical interpretation is to assume that precursor cells in chick and rat retinal cultures show different responses because the effects of neurotrophic or other regulatory factors depend on the developmental history and state of the responsive cell. Although very little is known about the situation in vivo, the in vitro effects of the various molecules mentioned above indicate that the generation of photoreceptors from uncommitted neuroblasts requires the action of multiple regulatory signals. Observations in other neuronal systems which demonstrate that individual neuronal cell types respond to a variety of neurotrophic factors and express the corresponding receptors support this concept (for a detailed discussion see Korsching, 1993). Recently, it has been shown that by the sequential action of three different neurotrophic proteins (CNTF, bFGF, NGF) immortalized sympathoadrenal progenitors of the MAH cell line can be driven to differentiate into sympathetic neurons (Ip et al., 1994). According to our results CNTF preferentially acts during earlier stages of photoreceptor development, whereas taurine responsiveness seems to appear later. Similar differences between taurine and CNTF (Kirsch et al., unpublished data) or other stimulatory activities in retina conditioned media (Altshuler et al., 1993) have been found in rat retinal cultures. Thus, CNTF, taurine and other molecules, all of which have been demonstrated to influence rod development in vitro, might be components of a multi-factorial regulatory process governing the generation and differentiation of the diversity of retinal phenotypes. According to this concept, in vitro effects of an individual factor and possibly its role in vivo would depend on the developmental context in which it becomes active. To get further support for this hypothesis it will be necessary to analyze in detail the specific cellular processes influenced by CNTF and the other effective molecules and to study possible interactions of different factors and their effects on the differentiation of other retinal cell types. It will be particularly interesting to compare these aspects during development in the cone dominated chick and the rod dominated rat retina.
ACKNOWLEDGEMENTS
The authors wish to thank G. Kaiser and I. Beckmann for their commitment and their excellent technical assistance. This work was supported by a grant from the Deutsche Forschungsgemeinschaft, SFB 325.