Time course of opsin expression in developing rod photoreceptors

The pattern of retinal neurogenesis is similar in all vertebrates. In general, differentiation of the retina into its distinct layers proceeds from basal (inner) to apical (outer) layers: the ganglion cell layer is formed first, and last to mature is the outer nuclear layer of photoreceptors (Sidman, 1960). In mice, for example, the onset of production of all cell types occurs prenatally, with cone photoreceptors first appearing about embryonic day 10 (E10), while production of rod photoreceptors begins later (≈E12) and extends into the first two postnatal weeks (Young, 1985; Carter-Dawson and LaVail, 1979). Similarly, in many species of teleost fish, cones appear first, and the larval forms initially have no rods (Blaxter and Staines, 1970). Teleost fish are also unusual in that as they grow, they continuously add new retina to the enlarging eye (Müller, 1952). Neurogenesis of all cell types thus continues beyond embryonic stages, with new retina added at the junction between the neural retina and iris epithelium, in a region called the germinal zone (Müller, 1952; Johns, 1977). In addition, there is a second site of neurogenesis in retinas of postembryonic fish: rod photoreceptors are continuously inserted into the otherwise post-mitotic retina as the eye grows (Lyall, 1957). These new rods are produced by scattered dividing cells called rod precursors, which are present in the outer nuclear layer (Sandy and Blaxter, 1980; Johns and Fernald, 1981). Rod precursors are most abundant in the young retina adjacent to the germinal zone, where the initial population of rod photoreceptors is added to the new retina. We call this area the ‘growth zone.’ Studies using thymidine autoradiographic birthdating with electron microscopy (Raymond and Rivlin, 1987) have shown that by at least nine days after the terminal mitosis, rods have begun to develop an axon terminal with synaptic specializations and an outer segment – concrete morphological evidence of differentiation.

In this paper, we address the question of whether dividing rod precursors show evidence of commitment to the rod cell fate. The only conclusive test of a cell’s degree of commitment is to transplant it into a different environment and see whether its fate remains the same or is altered. Since rod precursors are scattered among other retinal neurons, and we have not been able to isolate them, we have not yet done this experiment. Furthermore, we have been unable to find or generate an antibody to mark rod precursor cells specifically; for example, monoclonal antibody R24, which recognizes GD3 ganglioside of mammalian neuroepithelial cells (Goldman et al. 1984), does not label rod precursors. They are also not labelled by antibodies directed against mammalian intermediate filament proteins such as vimentin and neurofilaments, or by antibodies directed against intermediate filament proteins isolated from goldfish optic nerve (Jones et al. 1986). However, because rod precursors are mitotically active, they can be labelled with some non-specific markers for dividing cells, such as bromodeoxyuridine (BrdU) and human antiserum against proliferating cell nuclear antigen (PCNA; Stell et al. 1988), although a third mitotic cell marker (anti-sera against chick topoisomerase II; Heck and Earnshaw, 1986) fails to label them.

As a first step in examining commitment of rod precursors, we chose to look for expression of opsin, which we know is present in the mature rod photoreceptor. We take the expression of rhodopsin by a cell to indicate commitment to the rod pathway of development. Previous work investigating cell lineage in retinal development has addressed the same question: are retinal precursor cells committed to a particular cell fate? Turner and Cepko (1987), for example, traced retrovirally marked lineages (Sanes, 1989) in rat retina, demonstrating that a single progenitor cell in the retina is capable of giving rise to many types of retinal neurons and glia. Similar conclusions have been reached in developing amphibian retinas by the use of intracellular dye injections (Wetts and Fraser, 1988; Holt et al. 1988). Although these earlier studies suggested that mitotic retinal neuroepithelial cells are pluripotent, in all cases the lineage analysis was performed at a time in development when multiple cell types were being produced. In contrast, in the adult teleost fish, rod photoreceptors are the only neuronal cells normally produced in the otherwise mature regions of the retina. Therefore, if committed precursors exist anywhere in the retina, rod precursor cells are good candidates, since they normally produce only rods (Raymond and Rivlin, 1987). We had reason to suspect that if rod precursors were committed to the rod pathway of development, they might express opsin: both Hicks and Barnstable (1987),in vivo, and Araki et al. (1987) in vitro, showed in rat retina that cell bodies of immature rods express immunoreactivity for rhodopsin antibodies. Opsin is localized at detectable levels in the plasma membrane before morphological maturation of the outer segment.

The results presented here show that mitotic rod precursors do not express opsin. Rather, the onset of expression occurs in young rods between 3 and 4 days after BrdU injection, implying that cells do not express opsin until a few days after their terminal mitotic division.

Goldfish (Carassius auratus) were obtained from Ozark Fisheries (Richland, MO). Fish were on average 3–4cm in body length, with a nasotemporal eye diameter of 3.4–4.5 mm. They were anesthetized in 0.2% tricaine methanesulfonate (Sigma), wrapped in wet tissue and put on the stage of a Wild stereomicroscope. A small incision was made with a microknife (Tiemann & Co.) nasally through the sclera at the limbus. A 5 μl Hamilton syringe with a 33 gauge fixed blunt tip needle was used to inject 0.9–1.7μl of ImM 5-bromo-2’-deoxyuridine, BrdU (Sigma) into each eye to give an estimated final intraocular concentration of 50μM. At various intervals after BrdU injections (4h to 14 days), fish were killed by decapitation, and enucleated. Eyes were fixed in 4 % paraformaldehyde and 5 % sucrose in 0.1 M phosphate buffer (pH 7.4) for 30 min. The lens was removed, and the eye was again placed in fixative for 30 min.

Eyes were rinsed, cryoprotected, frozen in a 2:1 mixture of 20% sucrose and OCT (Tissue Tek), and then cryosectioned at 3 μm according to a previously published procedure (Barthel and Raymond, 1990). Frontal sections were cut parallel to the limbus and immediately behind the germinal zone. This plane of section yields the highest percentage of dividing rod precursors and young rods since each section is entirely within the growth zone of the retina. Sections were first incubated for 30 min with 20% normal goat serum, NGS (Sigma) diluted in phosphate-buffered saline (PBS) to block nonspecific binding. The mouse monoclonal antibody rho 4D2 (a gift from D. Hicks and R. Molday) was plated at a dilution of 1:200 in 1% NGS with 0.1% sodium azide and 0.5 % Triton X-100, and incubated overnight at 4°C. The following day, after rinsing off the primary antibody, the secondary antibody (Jackson Immunoresearch, pre-absorbed against rat immunoglobin and conjugated to fluorescein isothiocyanate, FITC), was applied and incubated for 30min. Sections were then treated with 4N HC1 for 30 min to denature the DNA in order to expose incorporated BrdU to the antibody (Taylor et al. 1989). Sections were then rinsed in PBS with 0.5% Triton (15 min) followed by PBS alone (15 min). After another 30 min blocking step, the same protocol was repeated using rat monoclonal anti-BrdU (1:20, Accurate Chemical). The secondary antibody (Jackson Immunoresearch, preabsorbed against mouse immunoglobin) was conjugated to either tetrarhodamine isothiocyanate (TRITC) or Texas Red. Slides were coverslipped with 60% glycerol in 0.1M sodium carbonate buffer (pH9.0), with 0.4 mg ml^-1 p-phenylenediamine, and viewed with a Leitz Aristoplan epifluorescence microscope, using an FITC cube (Leitz I 3) and a TRITC cube (Leitz N2.1).

To demonstrate that the monoclonal antibody rho 4D2 is specific for rhodopsin in fish, Western blots were prepared using isolated goldfish retinal rod outer segments. Rhodopsin was purified according to method A of Frank and Rodbard (1975) with minor modifications. Whole retinas were dissected from five fish and initially frozen. Retinas were later thawed and shaken in a microfuge tube for 1min in 1.5 mM Tris (pH 7.4) with protease inhibitors (5 ng ml^-1 Leupeptin, Pepstatin, Aprotinin), then pulsed once in the microfuge to spin down residual tissue. The supernatant was collected and spun for 25 min in a Sorvall RC2-B at 14×10³ revs min^-1 using an SS-34 rotor, and the pellet was resuspended in Tris buffer. A BCA protein assay (Pierce) at OD 560 showed a yield of 450 μg of total protein in 1.15 ml total volume. An equal volume of 20 % sodium dodecyl sulfate (SDS) was added, and the mixture was heated for 1 h in a water bath at 60°C. The mixture was then gently agitated for 18 h and dialyzed against 100 volumes of 1.0 mM EDTA, 1% SDS and 1% betamercaptoethanol; 20μ1 NaOH was added to raise the pH. After incubation for 2 h at room temperature, the final dialysis solution was changed to 0.5% SDS for 48 h. SDS–polyacrylamide gel electrophoresis was performed with a 6 % stacking and a 10% running gel. The resulting gel was removed and transferred to nitrocellulose. The membrane was blocked for 2h in Tween-20 in Tris-buffered saline (TTBS) and 5% NGS. Another incubation followed for 30 min in TTBS with 1 % NGS. A 1:1000 dilution of rho 4D2 in TIBS and 1 % NGS was applied for 30 min. After rinsing with TTBS, biotinylated secondary antibody (anti-mouse) and ABC reagent were applied according to the Vectastain (Vector Laboratories, Burlingame, CA) protocol, followed by a 5 min reaction with 3, 3’-diaminobenzadine tetrahydrochloride.

Western blots of outer segment proteins isolated from goldfish retina and reacted with the monoclonal antibody rho 4D2 showed a distinct band at a relative molecular mass of 38×10³ (Fig. 1). The relative molecular mass of bovine and Xenopus rhodopsins have been reported to be between 32 and 38×10³ (Papermaster and Dreyer, 1974), and therefore we are confident that rho 4D2 recognized opsin in the goldfish retina.

Approximately 20 sections were cut from each retina. Each section was scanned carefully for rho 4D2 reactive cell profiles in the outer nuclear layer. The immunofluorescent labelling had a ring-like appearance, characteristic of plasma membrane antigens. Each rho 4D2+ cell was then checked under the TRITC cube to determine whether the nucleus was labelled with anti-BrdU. BrdU+ cells in the outer nuclear layer, whether double labelled or not, were counted. BrdU labelling of mitotic cells was also seen in the choroid and occasionally in the inner nuclear layer. The morphology of double labelled cells was checked with Nomarski optics when necessary, in order to distinguish rod cell bodies from their outer segments, which are also labelled by rho 4D2. Some sections had such a large number of labelled cells that counting each one became impossible.

Table 1 shows the cell counts for each time point. The number of BrdU+ cells was usually greater than the number of rho 4D2+ cells in a given retina (15 out of 19 total retinas). At early times (from 4h to 3 days), the two labels were never localized to the same cell; a combined total of 1015 rho 4D2+ and 1974 BrdU+ cells were counted at these early times. Fig. 2 shows an example of a 4 h retina with no double labelled cells. We first observed double labelled cells at 4 days after BrdU injection, when on average, 5.7 % of the rho 4D2+ cells were also anti-BrdU+. (We use the rho 4D2 population to calculate the percent double labelled because it was usually the smaller of the two, and therefore limiting, since the maximum possible number of double labelled cells would equal the total number of rho 4D2+ cells.) At intervals longer than 4 days, the percentage of double labelled cells increased steadily, peaking at 10 days. Fig. 3 shows an example of a 7-day retina with a double labelled cell.

At 10 days, one of the three retinas examined had no double labelled cells (Table 1). In the other two, about 16% of the rho 4D2 population was double labelled. We consider the lack of double label in this one retina to be an aberrant result, and we have not included it in the averages, which are plotted in Fig. 4. We suggest that, because of unavoidable differences in the brightness of rho 4D2 staining among retinas, it is possible that some rho 4D2+ cell bodies were lightly labelled in this retina and therefore may have been missed. Because this one retina was so different from the other two retinas at the same interval and the other six retinas at intervals of 4 days and longer, all eight of which contained double labelled cells, we feel justified in excluding this outlier from the data set.

After peaking at 10 days, the percentage of double labelled cells began to fall off (Fig. 4). This decrease is expected, since young rods only transiently express high levels of opsin in the soma plasma membrane (Hicks and Barnstable, 1987). Preliminary results (Knight and Raymond, 1990) show that, by one month, there are no double labelled cells in the retina; presumably, in cells that continue to divide, the BrdU is diluted to background levels and the only remaining BrdU+ cells are differentiated rods with rho 4D2+ outer segments.

Our results indicate that dividing rod precursor cells do not express opsin. The appearance of double labelled cells between 3 and 4 days after BrdU incorporation suggests that only cells that have completed their terminal mitotic division express opsin. Dividing rod precursors incorporate BrdU during the S-phase of the mitotic cycle. By 24 h after injection, BrdU labelled cells have completed at most one cell cycle, since the approximate length of the cell cycle is 20 h (Johns, 1982). From 3 to 7 days, some BrdU-labelled precursor cells continue to divide while some undergo their terminal mitotic division and become postmitotic. Therefore, there are two populations of cells that could potentially be double labelled at 4 days, when we first begin to see double labelling. However, if a mitotic cell were double labelled, we would have seen double labelled cells earlier, because the dividing cells at 4 days are replicas of the precursors from the previous generation. Since we do not see labelling before 4 days, we can conclude that the double labelled population is made up only of postmitotic cells, not dividing rod precursors.

At 10 and 14 days, we assume that the BrdU label has been diluted to background levels in all or most cells that have continued dividing. Thus, only postmitotic rods in the outer nuclear layer are labelled with BrdU. Once the rod differentiates it becomes impossible to recognize double labelled cells because the opsin that was present in the plasma membrane of the young rod is now targeted to the outer segment (Hicks and Barnstable, 1987). Since the outer segment and the cell body are several micrometers distant in the mature retina (Raymond, 1985), the two can no longer be associated with an individual cell.

Watanabe and Raff (1990), working in rat retina, have similarly found that only postmitotic cells express opsin. The onset of opsin expression in rat retinal cells at 44–54 h after exposure to BrdU is slightly shorter than, but comparable to, our value of 4 days. Watanabe and Raff also suggest, as do Reh and Kljavin (1989), that the age of the other cells that are in contact with a dividing cell can influence its fate: they put forth the hypothesis that a rod-producing signal is turned on at a specific time in development. Adler and Hatlee (1989) propose an alternative model suggesting that all cell types in the retina are derived from a single precursor cell, each cell type then arising according to its position in the retina. These authors suggest that photoreceptor differentiation is the default pathway, taken by cells in the absence of other signals. Their hypothesis is based on the differentiation of isolated cultured cells from embryonic chick retina. However, in the chick retina a substantial majority (86 % central, 67 % peripheral) of photoreceptors are cones, not rods (Morris, 1970). In contrast, the rat retina contains almost exclusively rods (98.5 % : LaVail, 1976) Therefore, it is possible that the results presented by Adler and Hatlee are specific for the differentiation of cones,’ and that cones, not photoreceptors in general, differentiate in the absence of inductive cues (Raymond, 1990).

If dividing rod precursors are responsive to signals for rod differentiation, then it is possible that changing the local microenvironment could induce the rod precursors to produce cells other than rods. The rod precursor cells are normally surrounded by differentiated cones, cone horizontal cells and emerging rods, and cues in this environment may ensure that rod precursors only produce rods. Earlier work on retinal regeneration (Raymond et al. 1988) is consistent with this idea. In these experiments, injections of ouabain destroyed most of the neural retina, but the rod precursors survived (as evidenced by pre-labelling with tritiated thymidine), and in fact showed enhanced mitotic activity (Kurz-Isler and Wolburg, 1982; Negishi et al. 1987, 1988; Raymond et al. 1988). The presence in the regenerating retina of scattered neurogenic foci with a distribution similar to that of rod precursors, and the observation that the outer nuclear layer must be destroyed in order to provoke regeneration in the inner layers of the retina, both point to the conclusion that rod precursors are the source of regenerated cells (Raymond et al. 1988; Raymond and Barthel, 1989). In this vastly different milieu created by the destruction of the outer nuclear layer, the rod precursors appear to be capable of regenerating the entire retina. This suggests that the fate of rod precursors can be altered only when their immediate microenvironment is perturbed, and, further, that the rod precursors are not committed to the rod cell fate.

We would like to thank Linda Barthel and Mary Ellen Rounsifer for their invaluable technical assistance. This work was supported by NIH grants T32EY07022 (J.K.K.) and R01EY04318 (P.A.R.). P.A.R. has published previously as P.R. Johns.