Subsets of retinal progenitors display temporally regulated and distinct biases in the fates of their progeny

Transcription

1 Development 124, (1997) Printed in Great Britain The Company of Biologists Limited 1997 DEV Subsets of retinal progenitors display temporally regulated and distinct biases in the fates of their progeny Macrene R. Alexiades and Constance L. Cepko* Department of Genetics, and Howard Hughes Medical Institute, Harvard Medical School, 200 Longwood Avenue, Boston, MA 02115, USA *Author for correspondence ( SUMMARY Cell fate determination in the developing vertebrate retina is characterized by the sequential generation of seven classes of cells by multipotent progenitor cells. Despite this order of genesis, more than one cell type is generated at any time; for example, in the rat, several cell types are born during the prenatal period, while others are born postnatally. In order to examine whether there are classes of progenitor cells with distinct developmental properties contributing to this developmental progression, we examined antigen expression in progenitor cells during rat retinal development. Two markers of amacrine and horizontal cells, the VC1.1 epitope and syntaxin, were found to be expressed on a subset of progenitors in a temporally regulated manner that closely paralleled the birthdays of these cell types. In order to investigate which cell types were produced by the progenitors expressing these markers, fluorescent latex microspheres covalently coupled to VC1.1 antibodies were used to indelibly label VC1.1 + progenitor cells and their progeny. Early in retinal development, VC1.1 + progenitors generated a high percentage of amacrine and horizontal cells, but no cone photoreceptors. During this same period, a comparable number of cone photoreceptors were generated by VC1.1 progenitors. In the late embryonic and early postnatal period, VC1.1 + progenitors continued to generate predominantly amacrine cells, but also gave rise to an increasing number of rod photoreceptors. These findings demonstrate that expression of these two markers by progenitors is highly correlated with a bias towards the production of amacrine and horizontal cells. The fact that subsets of progenitors with temporally regulated and distinct biases are intermingled within the retinal neuroepithelium provides a basis for understanding how different cell types are generated both simultaneously and in a particular order by multipotent progenitors during retinal development. Key words: retina, progenitor, cell fate, VC1.1, amacrine cell, rat INTRODUCTION The retina has served as a model system for the vertebrate central nervous system (CNS). It develops from a pseudostratified neuroepithelium originating as an evagination from the ventral diencephalon. Across diverse species, [ 3 H]thymidine birthdating studies have shown that retinal cell types are generated in a characteristic order (reviewed in Altshuler et al., 1991). These studies demonstrated that the day that a retinal cell becomes postmitotic is highly correlated with its eventual fate. For example, ganglion cells are the first cell type to be produced, but are quickly followed by cone photoreceptors, horizontal cells and amacrine cells. Despite this order, multiple cell types are generated at any one time. For example, in mammals, rod photoreceptor birthdays overlap with those of the early cells as well as with those of the latest born cell types, bipolar cells and Müller glia (Sidman, 1961; Carter-Dawson and LaVail, 1979; Young, 1985a). Lineage analysis has shown that early retinal progenitor cells are totipotent, as clones with all retinal cell types were observed when marking was initiated early in development (Wetts and Fraser, 1988; Fekete et al., 1994). Later in development, progenitors are at least multipotent, in that marking of progenitors revealed clones containing multiple cell types (Turner and Cepko, 1987; Holt et al., 1988; Wetts and Fraser, 1988; Wetts et al., 1989; Fields-Berry et al., 1992; Fekete et al., 1994). Clones initiated late in development have a reduced spectrum of cell types relative to those generated early. For example, postnatal clones in the rodent retina containing two cells can have a rod and an amacrine, a rod and a bipolar, or a rod and a Müller glial cell (Turner and Cepko, 1987; Turner et al., 1990; Fields-Berry et al., 1992). Two models, which are not mutually exclusive, can be postulated to account for the order in cell type generation as well as the observation of complex, small clones (Cepko et al., 1996). One is that totipotent progenitors or their postmitotic daughters are instructed to produce or become different cell types at different times by environmental cues that vary during development. Alternatively, progenitors may undergo changes in their intrinsic properties that affect their ability to generate different cell types over the course of development. Several lines of evidence suggest that CNS progenitors change in their properties during development. In an in vitro study of the developing rat retina, embryonic retinal progen-

2 1120 M. R. Alexiades and C. L. Cepko itor cells differed from neonatal progenitor cells in proliferation and in kinetics of differentiation into rods (Watanabe and Raff, 1990). Similarly, cultures from embryonic and neonatal rat retina differed in their sensitivity to mitogens (Lillien and Cepko, 1992). Progenitor changes also have been assessed in studies of gene expression. In situ hybridization revealed a change in the RNA level for MASH-1 during retinal development (Guillemot and Joyner, 1993; Jasoni et al., 1994) and for otx-1 during cortical development (Frantz et al., 1994). In at least one CNS region, changes in progenitors contribute to the sequential production of cell types. Heterochronic transplantation in the mammalian cerebral cortex, which also displays an order of genesis for different cell types, demonstrated that cortical progenitors change during the course of the cell cycle in their plasticity (McConnell and Kaznowski, 1991) and that late cortical progenitors are limited in their ability to make early cell types (Frantz and McConnell 1996). We have examined the expression of antigens by retinal progenitors to assess whether progenitors change over time, and whether subsets of progenitors that differ in antigen expression also differ in their production of different cell types. We found that monoclonal antibodies VC1.1 and HPC-1, which mark two types of interneurons in the mature rat retina, are expressed on a subset of progenitors in a temporally regulated manner during retinal development. VC1.1 recognizes an N-linked carbohydrate epitope which is present on two forms of N-CAM and two high molecular weight proteoglycans expressed in postmitotic, differentiated amacrine and horizontal cells in the rat retina (Arimatsu et al., 1987; Zaremba et al., 1990; Naegele and Barnstable, 1991). HPC-1 recognizes the p35a protein syntaxin, a docking protein for synaptic vesicles, which is anchored by its C terminus in the presynaptic plasma membrane terminals of some neurons (Bennett et al., 1992). This marker had been shown to label amacrine cells in the mature rat retina (Barnstable et al., 1985). The temporal expression of the VC1.1 epitope and syntaxin in progenitor cells was found to correlate with the period of genesis of horizontal and amacrine cells. In order to investigate whether progenitors that express these antigens are biased towards production of amacrine and horizontal cells, the fates of cells produced by progenitors expressing the VC1.1 epitope were determined. This was accomplished using a novel approach to label the postmitotic daughters of marker-expressing progenitors. We found that the majority of the postmitotic daughters of VC1.1 + progenitors adopted the amacrine or horizontal cell fate, while the progeny of embryonic VC1.1 progenitors became cone photoreceptors. However, as development proceeded VC1.1 + progenitors were found to also generate an increasing number of postmitotic daughters that adopted the rod fate, and mitotic progeny that did not express the VC1.1 epitope. Expression of these markers thus indicates a bias, but not a commitment, to producing only amacrine and horizontal cells. These findings provide a basis for understanding how distinct cell types are generated in an overlapping fashion in the developing retina. At least two types of progenitors coexist in the retinal neuroepithelium. These two types are distinguishable by the expression of VC1.1 and syntaxin and possess distinct biases in the developmental fate of their postmitotic progeny. There is a dynamic relationship within the progenitor pool with respect to these markers as progenitors can gain and lose expression and the corresponding cell fate bias as development proceeds. MATERIALS AND METHODS Animals Sprague-Dawley rats were purchased from Taconic Laboratories. Average gestational length was 23 days. Day of birth was considered postnatal day 0 (P0). Embryonic age determinations were based on plug date, crown-rump length (Alexiades and Cepko, 1996) and morphological criteria (Long and Burlingame, 1936). [ 3 H]thymidine labeling of progenitors In vivo labeling entailed single intraperitoneal (i.p.) injection of 5 µci [ 3 H]thymidine/g body weight (BW). In vitro, retinae were dissected free of surrounding tissues and placed into 5 µci [ 3 H]thymidine/ml 45% Dulbecco s Modified Essential Medium (DMEM), 45% Ham s nutrient F12 medium (F12), 10% fetal calf serum (FCS), penicillin (Pen) (100 units/ml), and streptomycin (Str) (100 units/ml) for 1 hour. Retinae were subsequently rinsed by three media changes, dissociated, plated and fixed as described previously (Altshuler and Cepko, 1992). Immunolabeling and autoradiography Slides were blocked 1 hour in 10% FCS, 5% donkey serum (DS), 5% goat serum (GS) and 0.3% triton-x detergent for intracellular antigens, in DMEM. Primary antibody incubation for 1 hour to overnight employed blocking solution containing: VC1.1 (1:1000, Sigma, Arimatsu et al., 1987), HPC-1 (Sigma, anti-syntaxin, 1:750, Barnstable et al., 1985), or rabbit anti-syntaxin for double labeling with anti-vc1.1 (Bennett et al., 1992), anti-crabp (C-1, 1:500, Milam et al., 1990), calbindin (1:1000, Chu et al., 1993), Smi32 (1:1000, Sigma), recoverin (anti-p36) (1:100, Dizhoor et al., 1991; Kutuzov et al., 1991) and 115A10 (undiluted, Onoda and Fujita, 1987). Anti-rhodopsin staining with Rho4D2 (1:250, Hicks and Molday, 1986) was performed as described previously (Altshuler and Cepko, 1992). This was followed by three PBS-rinses and 1 hour in blocking solution containing secondary antibodies: Texas Red (TR)- conjugated donkey anti-mouse and donkey anti-rabbit (affinity purified, Jackson Immunoresearch Laboratories, 1:100); for double stains, TR-conjugated goat anti-mouse µ chain-specific Fab fragments and fluorescein (FITC) goat anti-mouse γ chain-specific Fab fragments (affinity purified, Jackson, 1:100) were used. Autoradiography was performed as previously described (Lillien and Cepko, 1992). Gelvatol-mounted slides were examined using a Zeiss Axiophot. [ 3 H]thymidine birthdating Cells becoming postmitotic the day of [ 3 H]thymidine pulse administration were identified as described previously (Young, 1985b). In vivo birthdating entailed single i.p. injection into gravid females or postnatal pups of 5 µci [ 3 H]thymidine/g BW (Young, 1985b; Guillemot and Cepko, 1992). Following various intervals, retinae were dissociated, immunostained, processed for autoradiography, and scored for cell fate and grain counts. In vitro birthdating entailed placing explants in 5 µci [ 3 H]thymidine/ml DMEM/F12/FCS for 1 hour at 37 C, followed by media washes and culture. To titrate the optimal dose for in vitro labeling, a series of retinae were labeled for 1 hour at doses of 0.1 to 10.0 µ Ci [ 3 H]thymidine/ml followed by culture. The necessary exposure lengths differed between doses, from 1-2 days to 4-6 weeks. Birthdating rates did not differ with dose, while maximum grain counts were higher with increased dosage. Proliferation and marker expression were similar between labeled and unlabeled retinae. The 5 µci/ml in vitro dose used provided optimal exposure length and grain counts.

3 Retinal progenitors with cell fate bias 1121 Bead preparation and labeling Yellow-green fluorescent microspheric latex beads (0.2 µm), purchased from Molecular Probes (cat. no. L-2151), were carboxylate-modified to minimize non-specific hydrophobic interactions with cell surfaces and biotin-labeled to maximize immunodetection possibilities. VC1.1 primary antibody (Sigma) was covalently coupled to beads according to the specifications provided by the supplier using VC1.1 antibody (1:10 dilution in MES buffer containing 1% bovine serum albumin (ph 6.0)). In vitro labeling involved placing retinae in 0.1% (100 µg/ml) VC1.1-coupled beads in DMEM/F12/FCS for 1 hour at 37 C, then medium-rinsing four times. Explants were then placed on 0.2 µm membrane filters (Nucleopore) floating on DMEM/F12/FCS/Pen/Str at 37 C for up to 2 weeks. In vivo labeling was performed by injecting 1.0 ml VC1.1-coated beads (1.0%) subretinally as described previously (Turner and Cepko, 1987). [ 3 H]thymidine and beads co-labeling In vivo experiments entailed concurrent injection of 5 µci [ 3 H]thymidine/g BW i.p. and 1.0 µl VC1.1-coupled beads (1.0%) subretinally. Explants were co-incubated 1 hour with 5 µci/ml [ 3 H]thymidine and 0.1% beads in DMEM/F12/FCS, medium-rinsed, cultured and harvested up to 2 weeks. Postmitotic daughters produced by all progenitors ([ 3 H]thymidine-birthdated) and VC1.1 + progenitors ([ 3 H]thymidine-birthdated and bead + ) were scored for marker expression. Controls in vitro and in vivo compared unlabeled retinae with: (1) VC1.1-coupled, (2) uncoupled and (3) control IgM-coupled beadlabeled retinae. Anti-alpha-sarcomeric actin (IgM) antibodies were obtained for coupled-bead controls (Sigma A-2172). Proliferation, marker expression and birthdating were analyzed. Controls for intercellular bead transfer involved co-cultured explants consisting of explant fragments labeled with beads and those labeled with PKH26 Red Fluorescent General Cell Linker Kit (Sigma). Explants were dissected into <0.5 mm 3 pieces; half VC1.1-coupled were beadlabeled, and half were labeled with PKH26 according to the specifications provided by the supplier. PKH26-labeled fragments were mixed with VC1.1-coupled bead-labeled fragments; these annealed into coherent explants as described previously (Lillien and Cepko 1992). Explants were harvested postculture and analyzed as whole mounts, as cryosections and as dissociated cells for PKH26 + bead + cells. Cell fate analysis Expression kinetics of the cell-type-specific antigens were determined for birthdating analysis. Marker expression onset and plateau in a birthdated cohort allowed age determinations for cell fate analysis. The amount and kinetics of cell death were also measured. Birthdated cohorts generated on each day during development were analyzed in the pre-cell-death period (i.e. to P2), mid-cell-death period (i.e. P4- P8), and the post-cell-death period (i.e. to P12) (Table A, Appendix). Postmitotic amacrine cells were identified using VC1.1, antisyntaxin and anti-crabp (Gaur et al. 1990; Milam et al., 1990), which reached maximum expression in birthdated cohorts by P2. Horizontal cells were identified by anti-calbindin, which recognizes a M r Ca 2+ -binding protein expressed by horizontal cells (Chu et al., 1993). Calbindin expression plateaued by P2. Cone photoreceptors were identified using anti-recoverin, as determined previously (Dizhoor et al., 1991; Stepanik et al., 1993). Recoverin also marks rod photoreceptors, which are born later (Young, 1985a). Among cells born on E14, recoverin expression plateaued by P3; among those born on E16 and E18, by P8. Rod photoreceptors were identified using antirhodopsin (Watanabe and Raff, 1990). Among cells born on E16, E18 and P0, rhodopsin expression plateaued by P8. Smi32, used to identify ganglion cells (Wu and Cepko, 1993), reached maximum expression among the in vivo E14-P0 birthdated population by P3. 115A10 expression, utilized to identify bipolar cells (Onoda and Fujita, 1987), plateaued by P9 among cells born on P0. RESULTS A subset of retinal progenitors express amacrine/horizontal cell markers in a temporally regulated manner during development The expression kinetics and cell-type specificity of two previously described markers of retinal cells, the VC1.1 epitope and syntaxin, were investigated during rat retinal development. Immunohistochemistry performed on early postnatal and mature retinae confirmed previous findings, which indicated expression of these markers in amacrine and horizontal cells (Barnstable et al., 1985; Arimatsu et al., 1987; Naegele and Barnstable, 1991). In the postnatal retina, the VC1.1 epitope and syntaxin were expressed by inner nuclear layer (INL) cells in two locations: primarily along the inner plexiform layer (IPL) border, in cells with extensive processes in the IPL, and along the outer plexiform layer (OPL) border, in cells with Fig. 1. Expression of the VC1.1 epitope and syntaxin during retinal development. The expression of the VC1.1 epitope and syntaxin in retinal sections was characterized using immunohistochemistry on different days during prenatal and postnatal retinal development. (A-C) E18 retinae immunostained with VC1.1 (B) and anti-syntaxin (C) antibodies. Expression of the VC1.1 epitope and syntaxin was observed in cells located in the VZ, as well as the inner nuclear layer (INL). (D-F) P4 and (G-I) P8 retinae immunostained with VC1.1 (E,H) and anti-syntaxin (F,I). At these ages, expression of VC1.1 and syntaxin was no longer detected in the VZ, but was observed in INL cells distributed along the inner plexiform layer (IPL) border, and along the outer plexiform layer (OPL) border (see Results). DAPI staining is shown in A, D, G and Texas-Red immunostaining is shown in B, C, E, F, H and I. VZ, ventricular zone; onl, outer nuclear layer; opl, outer plexiform layer; inl, inner nuclear layer; ipl, inner plexiform layer; gcl, ganglion cell layer. Sections are 10 µm thick.

4 1122 M. R. Alexiades and C. L. Cepko Fig. 2. Percentages of VC1.1 + and syntaxin + cells and progenitors during rat retinal development from E12 through P11. The expression of the VC1.1 epitope and syntaxin on all retinal cells and on mitotic cells (i.e. progenitors) was quantified for each day during retinal development, as described in Materials and Methods and presented in Appendix, Tables A, B. (A) Percentage of VC1.1 + and syntaxin + cells during development. Retinae of prenatal and postnatal rats were dissociated and analyzed for marker expression by immunocytochemical analysis. Each point represents the mean percentage of all retinal cells expressing each marker plotted with the corresponding standards of error (s.e.m.) for 3 experiments, each of which consisted of at least one litter of embryonic or postnatal pups. (B) Percentage of VC1.1 + and syntaxin + retinal progenitors during development. Retinae were administered a [ 3 H]thymidine pulse, dissociated and analyzed immunocytochemically for marker expression. The percentage of VC1.1 + and syntaxin + S-phase cells (progenitors) was measured as 100 [ 3 H]thymidine + marker + / [ 3 H]thymidine + cells. Each point represents the mean percentage of progenitors expressing VC1.1 or syntaxin plotted with the corresponding s.e.m. of a range of 3 experiments, each corresponding to at least one litter of embryonic or postnatal pups. (C) Comparison between postmitotic daughters generated by VC1.1 + progenitors and VC1.1 + postmitotic cells born. The number of postmitotic daughters generated by VC1.1 + progenitors were quantified and compared to the number of VC1.1 + postmitotic cells born (see Appendix Tables A and B). This analysis was conducted for 1 to 2 day intervals throughout development. In order to determine the number of postmitotic daughters generated by VC1.1 + progenitors, the rates at which VC1.1 + and VC1.1 progenitors gave rise to postmitotic cells were measured by birthdating analysis (see Material and Methods) and were found to be approximately equal for all ages measured (Table 1). The number of postmitotic cells generated by VC1.1 + progenitors on each day during development (black bars) was calculated as the product of the proportion of progenitors expressing the VC1.1 eptitope and the number of postmitotic cells generated on each day by all progenitors (Alexiades and Cepko, 1996; Appendix, Table B). In order to determine the number of cells born on each day expressing the VC1.1 epitope following differentiation (white bars), the birthdated cohort was immunocytochemically analyzed in the postnatal period. Scoring was conducted on P2, prior to the onset of cell death, except for those born on P0, for which post-death measurements were made on P9 (Appendix Table A). The percentage of VC1.1 + cells born ranged from 9% (born on P0) to 64% (born on E15) (Appendix Table A). The numbers of birthdated cells generated on E14 through E21 that survived to the post-death period demonstrated a mean percentage of cell death for birthdated VC1.1 + cells of 19.8±2.5% (Appendix Table A). The numbers of VC1.1 + cells born on each day (white bars) were calculated as the product of the number of postmitotic cells generated on each day and the proportion of birthdated cells that expressed VC1.1 + in the postnatal period (Appendix, Table B). large somata and horizontal processes (Fig. 1), consistent with expression by amacrine and horizontal cells, respectively. Multiple labeling with a panel of cell-type-specific antibodies demonstrated that cells expressing the VC1.1 epitope or syntaxin usually do not express antigens specific to the other retinal cell types. The VC1.1 epitope and syntaxin were also expressed during embryonic development in cells located in the ventricular zone (VZ), the area where proliferating cells reside (Fig. 1B,C). The VZ cells expressing the VC1.1 epitope and syntaxin possessed the characteristic bipolar morphology of proliferating neuroblasts undergoing interkinetic nuclear migration with nuclei traversing the radial dimension of the VZ (Sauer, 1937). VZ Number of cells (X 10 5 ) % of all cells expressing marker % of progenitors expressing marker C A %VC1.1 + %Syntaxin + E13 E15 E17 E19 E21 P0 P2 P4 P6 P8 P10 Age B %VC1.1 + %Syntaxin + E13 E15 E17 E19 E21 P0 P2 P4 P6 P8 Age E14 E15 E16 E17 E18 E19 E20 E21 P0-P2 P2-P4 P4-P7 Age cells expressing these antigens were found at the ventricular surface and within the VZ with processes extending to the ventricular surface. By the early postnatal period, cells expressing the VC1.1 epitope and syntaxin were no longer present in the VZ; rather, they were present only in the INL (Fig. 1D-I). Expression of the VC1.1 epitope and syntaxin on retinal cells were quantified during retinal development. Freshly dissected retinae were dissociated, immediately immunostained and scored for marker expression. The percentages of cells expressing the VC1.1 epitope or syntaxin were moderate as retinal development began, increased to high levels in the midembryonic period and decreased postnatally (Fig. 2A). The number of VC1.1 + or syntaxin + cells exceeded the number of postmitotic cells during this period (Alexiades and Cepko, 1996; Appendix Tables A,B), suggesting that the VC1.1 epitope and syntaxin were expressed by a proportion of mitotic cells. The expression of the VC1.1 epitope and syntaxin by mitotic cells (hereafter referred to as progenitor cells) was directly assessed. S-phase cells were identified by labeling with a [ 3 H]thymidine pulse and were immediately assayed for marker

5 Retinal progenitors with cell fate bias 1123 expression. A subset of [ 3 H]thymidine + cells expressing the markers was observed (Figs 2B, 3). In the early embryonic period (E12), 35-40% of progenitors expressed the VC1.1 epitope or syntaxin. The percentage increased as development proceeded, peaking at 70-80% on E14-15, then decreased in the late embryonic period, reaching low levels by the early postnatal period. Double-staining with the antibodies at E17 and E21 revealed that 100% of VC1.1 + cells were stained with anti-syntaxin, and 90% of syntaxin + cells also stained with VC1.1. These data indicated that the VC1.1 epitope and syntaxin were co-expressed by a subset of retinal progenitors in a temporally regulated manner during development. Quantitative analysis of VC1.1 + progenitors and production of amacrine and horizontal cells To begin to examine whether progenitors that express the VC1.1 epitope and syntaxin produce amacrine or horizontal cells, we first compared the numbers of cells in two categories for each day of development: (1) postmitotic cells generated on a given day by VC1.1 + progenitors and (2) postmitotic cells born on the same day that express VC1.1 after differentiation is complete. There was a close correlation between the numbers in the two categories throughout the embryonic period, as shown in Fig. 2C, suggesting that VC1.1 progenitors may produce primarily amacrine and horizontal cells. In contrast, from P0-P2, the number of postmitotic cells generated by VC1.1 + progenitors exceeded the number of VC1.1 + cells born by at least 40% (Appendix Table B). These findings suggested that, while VC1.1 + progenitors may be the source of the postmitotic VC1.1 + daughters, they were not committed to producing only daughters that achieved the amacrine or horizontal cell fate. A method to follow the fate of the postmitotic daughters produced by VC1.1 + progenitors In order to directly identify the fate of the postmitotic progeny generated by VC1.1 + progenitors, a labeling method was devised for tracking these cells (Fig. 4A). The method took advantage of the fact that the VC1.1 epitope is on the cell surface and accessible to binding of VC1.1 antibody in live tissue. To label VC1.1 + cells with an easily traced, stable marker, fluorescent beads were covalently coupled to VC1.1 antibodies. To test the method for labeling specificity and kinetics, several controls were performed. Fluorescent beads were covalently coupled either to VC1.1 or to an IgM antibody specific for actin, an intracellular antigen, which controlled for IgM-mediated labeling, or were used uncoupled. Explants were bead-labeled for up to 1 hour, rinsed and analyzed. After labeling for 30 minutes, the VC1.1-coupled beads were primarily attached to the cell surface; after labeling for 1 hour, the majority were intracellular (Fig. 4B). Following 1 hour of labeling, the explants were placed in culture and harvested at intervals. By 4 hours in culture, all beads were intracellular. Based on previous determinations of cell cycle and S-phase lengths (Alexiades and Cepko, 1996), the 4 hour maximum internalization time preceded M phase and postmitotic cell production. The VC1.1-coupled beads heavily labeled those cells that they labeled (Fig. 4B). Labeling occurred within 30 minutes and no increase in percent labeling was observed after 60 minutes (Table 1A). The efficiency of labeling VC1.1 + cells (i.e. percentages of VC1.1 + cells bead-labeled after 30 and 60 minute intervals) were calculated to be 45.5% and 43.2%, Table 1. Bead-labeling and [ 3 H]thymidine birthdating of retinae A. Labeling rate of E19 retinae with VC1.1-coupled beads 1 Labeling % of total cells % of VC1.1+ cells time (min) labeled with beads labeled with beads ± ± E19 explanted retinae were labeled in vitro with VC1.1-coupled beads, as described in Materials and Methods. Retinae were harvested after 30 and 60 minutes of labeling, dissociated and the percentage of bead+ cells scored. Each value represents the mean and s.e.m. of 3 experiments, each consisting of at least 1 litter of embryos. The percentage of VC1.1+ cells labeled with beads was calculated as the % bead+ cells/% VC1.1+ cells (see Fig. 4a and Table A). B. Rates of producing postmitotic progeny in developing retinae 2 % birthdated progeny In vitro In vivo Total Bead Bead+ Total Bead Bead+ E ± ± ± PO ± ± ± ± ± ±5.2 2 [ 3 H]thymidine birthdating rates for retinae of ages spanning development were measured for all progenitors, bead and bead+ progenitors both in vitro and in vivo whenever possible. [ 3 H]thymidine birthdating was performed as described in Materials and Methods. The rate for all progenitors was measured as 100 #birthdated cells/total # cells at the time of harvest. For bead progenitors this was 100 #birthdated bead cells/# bead cells, and for bead+ progenitors, 100 #birthdated bead+ cells/# bead+ cells. 3 E14 retinae were [ 3 H]thymidine birthdated in vitro and in vivo and scored at P3. For the in vitro studies, each value represents the mean and s.e.m. of 3 experiments, each experiment corresponding to retinae obtained from 1 litter. For the in vivo studies, the value represents the data obtained from retinae of 1 litter. 4 PO retinae were birthdated in vitro and in vivo in conjunction with VC1.1+-coupled bead-labeling. Each value represents the mean and s.e.m. of 3 experiments, each experiment corresponding to retinae obtained from 1 litter. respectively. Following developmental periods of up to 2 weeks, the clonal descendants of bead + progenitors could be identified as bead + (Figs 4A, 5). To examine the specificity of bead-labeling (i.e. whether all bead-labeled cells were VC1.1 + ), bead-labeled cells were assayed for VC1.1 expression. Ideally, the expression of the VC1.1 epitope by bead-labeled cells would be assessed by binding VC1.1 antibodies, followed by Texas Red-coupled secondary antibodies, immediately following bead-labeling. However, this could not be done until 12 hours post-beadlabeling for the following reasons. First, the assay needed to be performed following internalization of the bead-antibody complexes and destruction of the primary antibody bound to the beads, in order to preclude binding of the Texas Redcoupled secondary antibodies. Second, the beads were intensely fluorescent, emitting brightly at all wavelengths, including those detected using filters for Texas Red and other fluors. They also filled the cytoplasm initially, such that the entire cell was brightly fluorescent. By 12 hours post-labeling, the beads no longer filled the cytoplasm, but were in large intracellular aggregates, whose distribution did not obscure immunofluorescent detection on the cell surface, the typical location of VC1.1. The specificity of bead-labeling determined

6 1124 M. R. Alexiades and C. L. Cepko A B Fig. 3. A subset of retinal progenitors express the VC1.1 epitope and syntaxin during development. In order to determine directly whether retinal progenitor cells expressed the VC1.1 epitope or syntaxin, retinae of ages spanning development were administered a [ 3 H]thymidine pulse to mark cells in S phase. Immediately following the pulse, the labeled progenitors were assayed for expression of the VC1.1 epitope and syntaxin by immunocytochemistry. The DAPIstained nuclei of representative cells are shown in A and D. The silver grains present over several of the nuclei following [ 3 H]thymidine autoradiography (B,E) mark the S phase cells. A subset of the [ 3 H]thymidine + progenitors are shown to express the VC1.1 epitope (C) and syntaxin (F). Quantification of these data is presented in the Appendix, Table A and Fig. 2B. for E18 retinae in vitro (i.e. the percentage of bead + cells that were VC hours post-labeling) was observed to be between 77.4±4.4 and 87.9±2.6%. This was in keeping with the observation that a proportion of VC1.1 + cells lose VC1.1 expression during this time period. The predicted percentage of cells generated by E18 VC1.1 + progenitors expected to maintain VC1.1 expression was calculated to be 78.4% by 24 hours in vivo (Appendix Table B). Beads that were uncoupled and those coupled to anti-actin antibody produced a very low level of non-specific labeling, consisting of single bead particles randomly distributed over the retina. Controls for possible bead transfer between cells were also performed. Retinal fragments labeled with VC1.1- coupled beads were mixed with fragments labeled with the red dye, PKH26. These mixed explants were cultured and examined for double-labeled cells in whole mounts, on cryosections and among dissociated cells. In each case, the rate of transfer of beads was negligible: <0.1% of PKH26-labeled cells were bead + at the end of the culture period and, in the few cases of PKH26-bead-labeled cells, only one bead was present. The criterion for specific bead-labeling in all experiments was therefore 2 bead particles in a cell. The effects of antibody-coupled beads and [ 3 H]thymidine on retinal development were investigated in vivo and in vitro. Fig. 4. A method to map the fate of postmitotic daughters generated by VC1.1 + progenitors. In order to assess directly the fate of the postmitotic progeny generated by VC1.1 + progenitors, a labeling method was devised that allowed the tracking of the fate of their postmitotic daughters. (A) Schematic of labeling of postmitotic daughters of VC1.1 + progenitors with beads and [ 3 H]thymidine. VC1.1 + progenitors, which express the VC1.1 epitope on their surface, were labeled with fluorescent microspheric beads to which VC1.1 antibodies were covalently coupled. The antibody-coated beads were bound to the antigen expressed on the cell surface of VC1.1 + cells, and the bead-antibody complexes were subsequently internalized by the cell, resulting in bead-labeling that was independent of maintenance or loss of VC1.1 epitope expression by the cell or its daughters. In order to identify the postmitotic daughters generated by this subset of progenitors, a [ 3 H]thymidine pulse was administered simultaneous with bead-labeling to retinal explants or in vivo. The postmitotic daughters generated on the day of labeling remained heavily labeled with [ 3 H]thymidine. Those generated by VC1.1 + progenitors on the day of labeling were identified as heavily labelled with [ 3 H]thymidine and labeled with beads. Following a developmental period of up to 2 weeks, fate analysis was conducted on the postmitotic daughters generated by VC1.1+ progenitors of these retinae in the postnatal period. (B) Labeling rates of retinae with VC1.1-coupled beads. Retinae were labeled as explants or in vivo with VC1.1-coupled beads as described in Materials and Methods. A time course for bead labeling and internalization was performed. Fig. 4B1 shows a bead-labeled cell from an E19 retinal explant after 30 minutes of labeling and dissociation. The majority of the beads are on the outside of the cell. Fig. 4B2 shows a beadlabeled cell following 60 minutes of labeling an E19 retinal explant. The majority of the beads appear to have been internalized. The corresponding labeling and internalization rates are presented in Table 1 and Results. Labeled and unlabeled explants displayed similar layering and cellular morphology as in vivo retinae, as reported previously (Sparrow et al., 1990). Labeled and unlabeled retinae yielded similar cell numbers, percentages of cell types and birthdating profiles (Table 2; Alexiades and Cepko, 1996). Birthdating rates in vitro and in vivo were also comparable (Table 1B). Birthdating percentages for experiments using E14 retinae, which were harvested postnatally, differed between in vivo and in vitro conditions likely due to an eventual decline in proliferation in

7 Retinal progenitors with cell fate bias 1125 Fig. 5. Fate of postmitotic progeny generated by VC1.1 + progenitors during retinal development. The fate of the postmitotic daughters of VC1.1 + progenitors were determined following co-labeling with VC1.1-coupled beads and [ 3 H]thymidine, as shown in Fig. 4. Labeled retinae were allowed to develop to maturity and birthdated cells were analyzed for cell fate by immunocytochemistry. The determination of birthdate was based upon [ 3 H]thymidine grain counts with histogram analysis conducted for each experiment to determine the grain number necessary to assess this. The fate of each birthdated cell was determined by marker expression. Representative photomicrographs of experiments demonstrating the fates of the immediate postmitotic daughters of VC1.1 + progenitors are shown. (A-C) Labeling of an E18 retina which was cultured until the equivalent of P6. (A) A bead-labeled cell that contained 8 bead particles; the criterion for bead labeling was 2 bead particles; (B) the same cell was heavily labeled with [ 3 H]thymidine (i.e. birthdated); the cell contained 36 silver grains, the criterion for birthdating in this experiment was 18 grains; (C) the bead + birthdated cell expressed the VC1.1 epitope. (D-I) Labeling of an E14 retina harvested on P3. (D) A bead-labeled that contained 3 bead particles, of which 2 are in the plane of focus; (E) the bead + cell was heavily labeled with [ 3 H]thymidine; (F) the bead + birthdated cell expressed calbindin. (G-I) Labeled at E14 and harvested at P3. (G) A bead-labeled cell that contained 2 bead particles of which 1 is in the plane of focus; (H) the bead-labeled cell as well as a cell in the lower right corner of the field are shown to be heavily labeled with [ 3 H]thymidine; (I) the bead + birthdated cell failed to stain with antirecoverin antibodies, whereas the birthdated cell that was not bead labeled in the lower right field was found to be recoverin +. (J-L) Labeled in vivo on P0 and harvested on P9. (J) A bead-labeled cell which contained 5 bead particles; (K) the same cell was heavily labeled with [ 3 H]thymidine; (L) the bead + birthdated cell was rhodopsin +. The fate analysis results of the experiments initiated on E14, E16, E18 and P0, and harvested in the postnatal period are shown in Table 2 A, B, C and D, respectively. DAPI staining in A, D, G and J and Texas-Red staining in C, F, I and L. such long-term cultures (Alexiades and Cepko, 1996). The only other difference noted was the absence of Smi32 + cells in explants (Table 2). Smi32 + cells are ganglion cells, which have been shown to undergo apoptosis within 24 hours following optic nerve resection without affecting the development of other retinal cells (Barron et al., 1986; Beazley et al., 1987; Yew et al., 1989; Villesgas-Perez et al., 1993; Wu and Cepko, 1993; Berkelaar et al., 1994 and current observation). VC1.1 + progenitors usually, but not always, make amacrine and horizontal cells The fates of the immediate postmitotic progeny of VC1.1 + progenitors were determined by labeling E14, E16, E18 and P0 retinae with VC1.1-coupled beads and a [ 3 H]thymidine pulse concurrently. Retinae were then cultured as explants or allowed to develop in vivo. Cell fate analysis of birthdated cells was conducted postnatally by immunocytochemistry and autoradiography. Markers were assessed for use in this assay by monitoring their expression kinetics relative to the terminal S phase and ages of analysis were chosen accordingly (Materials and Methods). Cell fate assignments based on marker expression were interpreted as follows (see Materials and Methods): VC1.1 + as amacrine and horizontal cells, calbindin + as horizontal cells, recoverin + as cone and rod photoreceptors, rhodopsin + as rod photoreceptors, Smi32 + as ganglion cells and 115A10 + as bipolar cells. The VC1.1 epitope and syntaxin were judged to be reliable markers of postmitotic cells that had achieved the amacrine and horizontal cell fates, rather than of progenitors or newly postmitotic cells, due to the fact these antigens were stably expressed almost entirely by postmitotic cells identified as amacrine and horizontal cells in the INL and OPL by the early postnatal period (Fig. 1E, F; Discussion). The expression level of the markers varied, indicated by a + for high level and a +/ for low level, as described in Table 2. The postmitotic daughters of VC1.1 + progenitors were identified as bead-labeled and heavily [ 3 H]thymidine-labeled. The rates of postmitotic cell production were compared between VC1.1 + progenitors and all progenitors, and found to be equivalent for all ages tested, as shown for E14 and P0 in Table 1. The amount of cell death in the postnatal period was measured for each cell type. The mean cell death for VC1.1 + cells born from E14 through E21 was 19.8±2.5%, which occurred between P2 and P12 (Appendix Table A). The fate analysis of the postmitotic daughters of VC1.1 + progenitors compared to those of the progenitor population as a whole from E14 through P0 is shown in Table 2. Throughout development, VC1.1 + progenitors generated a significantly higher proportion of VC1.1 + postmitotic daughters than did the progenitor population as a whole. From E14-E18, 77-93% of postmitotic daughters arising from VC1.1 + progenitors were VC1.1 + postnatally, as compared to 42-52% of those arising from all progenitors. At P0, VC1.1 + birthdated daughters constituted 42-74% of those generated by VC1.1 + progenitors, as

8 1126 M. R. Alexiades and C. L. Cepko Table 2. Fate of postmitotic progeny generated by all progenitors and VC1.1 + progenitors during retinal development 1 All progenitors VC1.1 + progenitors in vivo in vitro in vivo in vitro Marker + + and +/ + + and +/ + + and +/ + + and +/ (A) Percentage of [ 3 H]thymidine birthdated daughters generated from all progenitors and VC1 + progenitors on E14 that are marker + on P3 2 VC ± ± ± ±5.1 calbindin ± ± ± ±2.9 recoverin ± ± ± ±0.0 smi (B) Percentage of [ 3 H]thymidine birthdated daughters generated from all progenitors and VC1.1 + progenitors on E16 that are marker + by P8 5 VC ± ± ± ±3.1 calbindin ± ± ± ±1.2 recoverin ± ± ± ±3.4 rhodopsin 21.4± ± ± ±2.3 smi (C) Percentage of [ 3 H]thymidine birthdated daughters generated from all progenitors and VC1 + progenitors on E18 that are marker + by P10 6 VC ± ± ± ±2.0 calbindin ± ± ± ±0.0 recoverin ± ± ± ±2.4 rhodopsin 36.6± ± ± ±3.5 smi (D) Percentage of [ 3 H]thymidine birthdated daughters generated from all progenitors and VC1 + progenitors on PO that are marker + on P9 7 VC1 9.0± ± ± ± ± ± ± ±4.1 rho 64.7± ± ± ± ± ± ± ± A10 4.6± ± ± ±1.6 smi32 4.5± ± ± ±4.5 1 The postmitotic cells generated during development by all progenitors and VC1.1 + progenitors were identified by [ 3 H]thymidine birthdating and co-labeling with VC1.1 coupled beads. The fate of each birthdated cell was determined by cell-type specific marker expression: cells that were unambiguously positive for each marker were designated as +, and those that expressed the marker at a low level were designated as +/. 2 For the in vivo controls, each value represents the data obtained from retinae of 1 litter. For the in vitro studies, each value represents the mean and s.e.m. of data obtained from three sets of experiments; each set consisted of explanted retinae obtained from 1-3 litters which were harvested on the age indicated. The minimum number of [ 3 H]thymidine birthdated cells scored per experiment for each set of progenitors were as follows: all progenitors, in vivo 200, in vitro 100; VC1.1 + progenitors Represents the results of three repetitions with the following results: 0/29, 0/24, and 0/25 (i.e. at total of 78 birthdated cells that were labeled with beads were scored and none were recoverin +. The possibility is not discounted that had a greater number been scored, recoverin + cells that were bead labeled and birthdated may have been found. 4 No Smi32 + cells were observed in the explanted retinae after culture in vitro. See results and discussion for details. 5 For the in vivo controls, each value represents the data obtained from retinae of 1 litter. For the in vitro studies, each value represents the mean and s.e.m. of data obtained from 3-5 sets of experiments; each set consisted of explanted retinae obtained from 1-3 litters which were harvested on P5-P8 for all markers, with the exception of rhodopsin which was harvested on P8. The minimum number of 3 H-thymidine birthdated cells scored per experiment for each set of progenitors were as follows: all progenitors, in vivo 200, in vitro 100; VC1.1 + progenitors For the in vivo controls, each value represents the data obtained from retinae of 1 litter. For the in vitro studies, each value represents the mean and s.e.m. of data obtained from 3-5 sets of experiments; each set consisted of explanted retinae obtained from 1-3 litters which were harvested at P10. The minimum number of [ 3 H]thymidine birthdated cells scored per experiment for each set of progenitors were as follows: all progenitors, in vivo 200, in vitro 100; VC1.1 + progenitors Each value represents the mean and s.e.m. of 3-4 sets of experiments; each set consisted of 6-12 retinae from 1 litter. The minimum number of [ 3 H]thymidine birthdated cells scored per experiment for each set of progenitors were as follows: all progenitors, in vivo 200, in vitro 100; VC1.1 + progenitors 35. compared to 9-19% of those generated by all progenitors. VC1.1 + progenitors also gave rise to a higher proportion of calbindin + postmitotic cells. On E14, 39-47% of those generated by VC1.1 + progenitors were calbindin +, as compared to 22-27% by all progenitors. At E16, 11-16% of birthdated cells generated by VC1.1 + progenitors were calbindin + as compared to 8-9% by all progenitors. From E18 onward, no calbindin + birthdated cells were observed. VC1.1 + progenitors gave rise to far fewer recoverin + postmitotic cells as compared to all progenitors. At E14, 0% of postmitotic daughters generated by VC1.1 + progenitors were recoverin +, as opposed to 26-27% of those generated by all progenitors. Recoverin + cells birthdated at E14 represent cone photoreceptors, whereas those birthdated later in development include rods (see Discussion). A small increase in the rate of recoverin + postmitotic cell production was observed for VC1.1 + progenitors as development proceeded. At E16, 5-9% of postmitotic daughters of VC1.1 + progenitors were recoverin +, as compared to 37-41% of those generated by all progenitors. At E18, these percentages increased to 10-13% and 44-50% recoverin +, respectively. The increase in recoverin + birthdated cells was correlated with a rise in the production of rhodopsin + postmitotic cells by VC1.1 + progenitors. At E16, 5-10% of birthdated cells generated by VC1.1 + progenitors were rhodopsin +, as compared to 21-31% of those made by all progenitors. At E18, these proportions were 4-17% and 37-44% rhodopsin +, respectively. The majority, if not all, of the recoverin + birthdated cells generated from VC1.1 + progenitors may be accounted for as rod photoreceptors, as measured by rhodopsin staining. Moreover, the absence of recoverin + cells generated from VC1.1 + progenitors at E14, when cone production is highest, also suggests that the recoverin + cells generated later in development are most likely to be rods. Alternatively, the VC1.1 +

9 Retinal progenitors with cell fate bias 1127 progenitors may change in their properties and acquire the ability to make cones, and the aforementioned percentage similarity may be coincidental. Postnatally, the generation of rhodopsin + cells by VC1.1 + progenitors increased, but remained significantly lower than that of the entire progenitor population. In vivo at P0, 36-49% of postmitotic daughters generated by VC1.1 + progenitors were rhodopsin +, as compared to 65-76% generated by all progenitors. In addition, VC1.1 + progenitors at P0 gave rise to a small percentage of 115A10 + cells, which was lower than that generated by all progenitors. The in vitro results were comparable to those obtained in vivo. A high percentage of birthdated daughters that expressed the VC1.1 epitope at the end of the culture period maintained the beads. These values are as follows: E14, 68.5±9.1%, E16, 55.5±9.3%, E18, 65.6±7.1%, P0 (in vivo) 80.6±13.2%, and P0 (in vitro) 59.0±11.5%. These values are the mean and s.e.m. of 3-6 experiments corresponding to those presented in Table 2. DISCUSSION Progenitors expressing the VC1.1 epitope generate a distinct subset of cell types In this study, we have shown that distinct subsets of retinal progenitors differ in the cell types that they produce. The progenitor subsets are distinguishable by the expression of two markers, the VC1.1 epitope and syntaxin, which identify amacrine and horizontal cells in the mature retina. The vast majority of postmitotic daughters produced by VC1.1 + progenitors during the embryonic period were identified as amacrine and horizontal cells. At this high rate of genesis, given the proportions of VC1.1 + progenitors observed, virtually all amacrine and horizontal cells could be accounted for as products of VC1.1 + progenitors and need not be descended from VC1.1 progenitors. VC1.1 + and VC1.1 progenitors also differed in cone production. A significant percentage of the postmitotic cells produced at E14 by all progenitors were identified as cones, yet no cones were identified as the products of VC1.1 + progenitors. Given that VC1.1 progenitors are a minority at this time, this finding suggests that most daughters of early VC1.1 progenitors are cones. Notably, as development proceeded, VC1.1 + progenitors could produce photoreceptors, as an increasing percentage of their birthdated daughters were identified as rods. In the early postnatal period, roughly 60% of the postmitotic daughters of VC1.1 + progenitors were identified as amacrine cells; the remainder were identified largely as rods, and a small number as bipolar cells. These ratios differed from those of the progenitor population as a whole, which generated mostly rods and few amacrine cells. All of these data suggest that there is a profound bias towards the amacrine and horizontal cell fate for the postmitotic progeny of the VC1.1 + progenitor. Given this propensity towards amacrine and horizontal cell production, why do VC1.1 + progenitors also produce rods and a few bipolar cells? Rod-inducing signals and/or signals that repress the amacrine/horizontal fates in the late embryonic and neonatal environment may sometimes override the bias to become an amacrine cell. Instead, or in addition, inductive signals that are required for amacrine/horizontal commitment may be limiting in these later environments. The need for such signals is unclear, but at least one extrinsic influence is important in amacrine cell genesis: introduction of a constitutively active Notch allele blocks amacrine development (Bao and Cepko, 1997). The data presented here are consistent with lineage analysis conducted previously on postnatal rat retinae in our laboratory (Turner and Cepko, 1987). Amacrine cells resulting from infection of postnatal retinae and present in two cell clones are usually accompanied by a rod (24/25 clones), and infrequently by another amacrine cell (1/25 clones). Larger clones generated postnatally often contain only one amacrine cell. Also in keeping with lineage analysis is our finding that roughly 10% of progenitors at P1 express the VC1.1 epitope and syntaxin. Approximately 6% of clones initiated by viral infection on P0 and analyzed in the post-cell-death period contained at least one amacrine cell; the difference may be attributable in part to cell death (Results and Appendix Table A). The data from lineage studies and the data presented here indicate that amacrine cells do not derive from a progenitor committed to making only amacrine cells. A recent cell death study in the rat retina indicated that a large number of displaced amacrine cells (i.e. those located in the ganglion cell layer) died during the embryonic period (Galli-Resta and Ensini, 1996). Blaschke et al. (1996) also found that cell death in the cortical proliferative zones was higher than appreciated previously. Our measurement of amacrine and horizontal cell death in the classic cell death period, between P2 and P10, was on average approximately 20%, which would not confound the interpretations presented (Appendix Table A). Cell death in the embryonic period could affect our interpretations if, for example, any cone produced by a VC1.1 + progenitor died, or any amacrine or horizontal produced by a VC1.1 progenitor died. Were this the case, there would still be a profound bias in the production of cell types that survive by distinct progenitors, perhaps reflecting fitness due to information passed from the progenitor cell. Assessing amacrine and horizontal fate Of importance in the experimental approach used here is the validity of the markers used to identify cell fates. VC1.1 and syntaxin expression on postmitotic cells was interpreted as an indication of amacrine and horizontal cell fates, rather than as markers transiently expressed on postmitotic progeny from VC1.1 + /syntaxin + progenitors. It was necessary to establish that marker expression on postmitotic cells was stable. Quantitation of amacrine and horizontal cell genesis by [ 3 H]thymidine birthdating analysis and VC1.1 and syntaxin labeling was consistent with previous determinations of amacrine and horizontal cell numbers and birthdating in mouse, which employed morphological criteria for cell identification in sections (Young, 1985a). Comparison of birthdating analysis with quantitation of marker-expressing cells through the pre-cell-death period suggested that the markers were stably expressed from the day of the final S phase through maturity (Appendix Table B). Furthermore, cells expressing the VC1.1 epitope and syntaxin in the postnatal period were identified as amacrine and horizontal cells based upon morphology and location in sections (Fig. 1). Finally, multiple labeling using VC1.1 and syntaxin and other cell-type-specific markers has shown that birthdated cells expressing these markers rarely co-expressed other markers.