Retinoblastomas initiate in the developing retina in utero and are diagnosed during the first few years of life. We have recently generated a series of knockout mouse models of retinoblastoma that recapitulate the timing, location, and progression of human retinoblastoma. One of the most important benefits of these preclinical models is that we can study the earliest stages of tumor initiation and expansion. This is not possible in human retinoblastoma because tumors initiate in utero and are not diagnosed until they are at an advanced stage. We found that mouse retinoblastoma cells exhibit a surprising degree of differentiation, which has not been previously reported for any neural tumor. Early-stage mouse retinoblastoma cells express proteins found normally in retinal plexiform layers. They also extend neurites and form synapses. All of these features, which were characterized by immunostaining, Golgi-Cox staining, scanning electron microscopy, and transmission electron microscopy, suggest that mouse retinoblastoma cells resemble amacrine/horizontal cells from the retina. As late-stage retinoblastoma cells expand and invade the surrounding tissue, they lose their differentiated morphology and become indistinguishable from human retinoblastomas. Taken together, our data suggest that neuronal differentiation is a hallmark of early-stage retinoblastoma and is lost as cells become more aggressive and invasive. We also show that rosette formation is not a hallmark of retinoblastoma differentiation, as previously believed. Instead, rosette formation reflects extensive cell-cell contacts between retinoblastoma cells in both early-stage (differentiated) and late-stage (dedifferentiated) tumors. [Cancer Res 2007;67(6):2701–11]

Mouse models of human cancer allow us to study early tumor formation and determine whether tumor cell differentiation is a hallmark of carcinogenesis. However, these models must recapitulate the histopathologic and genetic features of the human disease. We and others have generated knockout mouse models of retinoblastoma by inactivating the Rb and p107 genes in the developing mouse retina of Chx10-Cre;RbLox/−;p107−/− mice (14). MacPherson et al. showed that inactivation of Rb and p130 also led to retinoblastoma in mice (2). Although these mouse models are not the focus of this report, future analyses will compare data on the Rb;p107-deficient and Rb;p130-deficient retinoblastomas.

Chx10-Cre;RbLox/;p107−/− tumors exhibit only limited penetrance (50–60%) and minimal invasion of the surrounding tissue (5). When we simultaneously inactivated Rb, p107, and p53 in the developing Chx10-Cre;RbLox/−;p53Lox/−;p107−/− retina, aggressive, invasive bilateral retinoblastoma formed in 100% of the mice by a few months of age. We also recapitulated the loss of heterozygosity at the Rb locus in Chx10-Cre;RbLox/+;p53Lox/−;p107−/− mice (5, 6).

In human retinoblastoma, the p53 gene is not mutated (7, 8). These data suggest that the p53 pathway is inactivated in human retinoblastoma by a genetic lesion not involving the p53 locus. We recently found that amplification of the MDMX and MDM2 genes suppresses the p53 pathway in 75% of human retinoblastomas (9). Currently, no inducible, tissue-specific MDMX transgenic mouse line is available to recapitulate MDMX amplification in mouse retinoblastoma. However, our studies have shown that MDMX acts exclusively through the p53 protein; therefore, the Chx10-Cre;RbLox/+;p53Lox/−;p107−/− mice are the best mouse model of human retinoblastoma in that they recapitulate the genetic lesions in the tumor suppressor pathways (i.e., Rb and p53 pathways) and the histopathologic features of human retinoblastoma. These data, combined with the extensive characterization of the role of the retinoblastoma family in retinal development (1, 2, 1012), suggest that Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice faithfully recapitulate human retinoblastoma, and that Chx10-Cre;RbLox/−;p107−/− mice have a more mild form of retinoblastoma characteristic of the early stages of tumorigenesis.

In this study, we used Chx10-Cre;RbLox/−;p53Lox/−;p107−/− and Chx10-Cre;RbLox/−;p107−/− mice to study retinoblastoma progression, with particular emphasis on tumor cell differentiation during early-stage retinoblastoma and the loss of differentiation as the disease progresses and the cells invade the surrounding tissue. One feature of the Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mouse is that Cre-mediated recombination of the RbLox and p53Lox alleles occurs in a mosaic pattern of apical-basal stripes across the entire retina during development (11, 13). Not only does this mimic the developmental environment of human retinoblastoma, which initiates in utero, but also it leads to several neoplastic foci in each retina that facilitate the characterization of early-stage retinoblastoma tumorigenesis. Unlike other retinoblastoma models that rely on Nestin-Cre (2) or Pax6-Cre (1), the mosaic pattern of Rb and p53 inactivation using Chx10-Cre minimizes the complication of non-cell autonomous effects. Specifically, the Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice have multiple tumor foci interspersed throughout nontransformed tissue, which permits a direct comparison of cell autonomous and non-cell autonomous effects.

We analyzed the differentiation of mouse retinoblastomas by using immunostaining, real-time reverse transcription-PCR (RT-PCR), dissociated cell scoring, Golgi-Cox staining, scanning electron microscopy (SEM), and transmission electron microscopy (TEM). Although some investigators have used the term “tumor differentiation” to describe the progressive change in general appearance of the tumor as it assumes more neoplastic histologic features, we used a variety of morphologic assessments to define cellular differentiation (and possible dedifferentiation) during early and late stages of tumorigenesis. This rigorous assessment indicated that tumor progression reflects a changing ratio of highly differentiated and less-differentiated cell types, with the most advanced tumors composed primarily of undifferentiated cells.

Our data indicate that early-stage mouse retinoblastomas exhibit an extraordinary degree of differentiation, including neurite extension and synapse formation. This is a hallmark of early-stage tumorigenesis. As human or mouse retinoblastoma cells progress to late-stage tumorigenesis and invade the surrounding tissue, they lose their ability to differentiate. This is the first report of synaptogenesis in a central nervous system tumor and provides insight into the cell of origin, stem cell mechanism of growth, and the molecular/cellular changes that accompany tumor progression in the developing retina. These studies set the stage for analysis of other retinoblastoma models that have different gene deletions.

Animals. The Institutional Animal Care and Use Committee at St. Jude Children's Research Hospital approved all procedures for animal use. Generation of Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice and Chx10-Cre;RbLox/−;p107−/− mice has been described elsewhere (4).

Immunostaining. Retinae were freshly dissected and fixed overnight in 4% paraformaldehyde/1× PBS. Tissue was embedded in 4% agarose, and 50-μm sections were blocked and immunostained as described previously (14, 15). Primary antibodies and dilutions are provided in Table 2. Biotin-conjugated secondary antibodies (Vector Laboratories, Burlingame, CA) were diluted to 1:500, with the exception of mGluR6, which was diluted to 1:1,000. The Vectastain ABC reagent (Vector Laboratories) was used according to the manufacturer's instructions. Antigen detection was carried out using Cy3-tyramide (Perkin-Elmer, Wellesley, MA) according to the manufacturer's instructions. Nuclei were counterstained with SYTOX Green (Invitrogen, Carlsbad, CA) diluted to 1:15,000. All images were acquired on a Leica TCSNT confocal microscope.

Electron microscopy. Animals were anesthetized with avertin until deep tendon reflexes were lost. Transcardial perfusion was done first with carboxygenated Ames medium supplemented with 40 mmol/L glucose to clear the vasculature and then with Sorenson's phosphate buffer (pH 7.2) with 2% electron microscopy–grade paraformaldehyde and 1% electron microscopy–grade glutaraldehyde. Eyes were then harvested. A slit was made in the cornea to aid in diffusion of fixative, and the tissue was placed in Sorenson's phosphate buffer with 3% glutaraldehyde overnight. For TEM analysis, tissue was washed with 0.2 mol/L cacodylate buffer in 5% sucrose, post-fixed in 1% OsO4, embedded, sectioned, stained with uranyl acetate and Reynold's lead citrate, and mounted for imaging. For SEM analysis, samples were rinsed in 0.1 mol/L sodium cacodylate buffer with 5% sucrose and post-fixed in 1% OsO4 in the same buffer for 1 h. After rinsing, the samples were stained en bloc with 4% uranyl acetate in water for 2 h and dehydrated in a graded series of ethanol solutions to a final rinse in absolute ethanol. The samples were exposed to critical point drying in a Samdri 520 system (Tousimis, Rockville, MD) and viewed using an FEI Phillips XL30 Environmental Scanning Electron Microscope.

RNA isolation, cDNA synthesis, and real-time RT-PCR. Tumors were removed from the surrounding sclera, lens, cornea, and iris and immediately flash frozen and stored at −80°C. RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer's instructions and reverse transcribed using the SuperScript II cDNA synthesis kit (Invitrogen). Real-time RT-PCR primers and probes were designed using Primer Express software (Applied Biosystems, Foster City, CA); the probe reporter was FAM, and the quencher was BHQ. The real-time RT-PCR reactions were done in 20-μL reactions containing 200 nmol/L primers, 200 nmol/L probe, and Taqman Universal PCR Master Mix (Applied Biosystems) on an ABI 7900 HT Sequence Detection System (Applied Biosystems). The following PCR variables were used: incubation at 50°C for 2 min followed by 95 °C for 10 min and 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Real-time RT-PCR data for Gapd (glyceraldehyde-3-phosphate dehydrogenase; Mm.317779) and Gpi1 (glucose phosphate isomerase; Mm.589) were used as internal references.

Golgi-Cox staining. Rapid Golgi-Cox staining was done on whole eyes by using the FD Rapid GolgiStain kit (FD NeuroTechnologies, Inc., Ellicott City, MD), as described by the manufacturer.

Immunostaining of dissociated cells. Dissociated cells were fixed in 4% paraformaldehyde in PBS, washed, and treated with 1% hydrogen peroxide in PBS before incubation in blocking solution (PBS containing 0.5% Triton X-100 and 2% normal donkey serum or 2% normal goat serum). Biotin-conjugated secondary antibodies (donkey anti-mouse IgG and goat anti-rabbit IgG) were diluted to 1:500 in blocking solution. After secondary antibody binding, the dissociated cells were incubated with an avidin-biotin-peroxidase complex (Vectastain ABC) and then detected with Cy3-tyramide (Perkin-Elmer) according to the manufacturer's instructions. For nuclear staining, dissociated cells were incubated with either SYTOX green or 4′,6-diamidino-2-phenylindole (Invitrogen; 1:20,000). Labeled cells were visualized using a Zeiss Axioplan 2 microscope, and images were captured with an Axiocam digital camera (Zeiss, Thornwood, NY).

Bromodeoxyuridine and [3H]thymidine labeling. To label S-phase cells, we incubated retinal tumors and their surrounding normal retinae in explant culture medium containing 10 μmol/L bromodeoxyuridine (BrdUrd; Boehringer Mannheim, Mannheim, Germany) or [3H]thymidine (5 μCi/mL; 89 Ci/mmol; Amersham Biosciences, Piscataway, NJ) for the indicated times at 37°C. Autoradiography and BrdUrd detection with anti-BrdUrd antibody (Amersham Biosciences) were carried out as described previously (14, 15). Detailed protocols are available online.6

Cell viability and apoptosis. Dissociated cells or retinal sections (14-μm thick) obtained on a Leica cryostat (Leica, Bannockburn, IL) were stained using the terminal deoxynucleotidyl transferase–mediated nick-end labeling (TUNEL) apoptosis system (Promega, Madison, WI) according to the manufacturer's instructions, with the exception that we used Cy3-tyramide for detection rather than the colorimetric substrate. The TUNEL assay was complemented by staining for activated caspase-3. Stained retinal sections were imaged using a Leica TLSNT confocal microscope, and the percentage of labeled nuclei was determined from micrographs. Analysis of cell viability and Annexin staining was done by fluorescence-activated cell sorting using the Viacount kit (Guava Technologies, Hayward, CA) according to the manufacturer's instructions.

Expansion of retinoblastoma into the vitreous, subretinal space, and anterior chamber. To study how retinoblastomas expand and invade the surrounding tissues, we perfused 21 mice ages postnatal day 4 (P4) to P257 (Table 1) from three genotypes (Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox, Chx10-Cre;RbLox/−;p107+/−;p53Lox/Lox, and Chx10-Cre;RbLox/−;p107−/−) with glutaraldehyde/formaldehyde in Sorenson's phosphate buffer and analyzed the excised globes using light microscopy, SEM, and TEM. Chx10-Cre;RbLox/−;p107−/− mice have a mild form of retinoblastoma, whereas Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox mice have a much more aggressive, invasive form. The Chx10-Cre;RbLox/−;p107+/−;p53Lox/Lox retinoblastoma is also aggressive and invasive, but the time of onset is delayed compared with Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox retinoblastoma (5, 6). In most cases, both eyes were examined, except when one was needed for comparative genetic or confocal studies. Three eyes were bisected through the optical nerve for direct comparison (i.e., one half was used for SEM, and the other half was used for TEM).

Table 1.

Genotypes of mice used for light and electron microscopy

Abbreviation*Animal IDGenotypeAgeAnalysis
TKO V7-42 Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox P13 SEM, LM 
TKO V7-44 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P13 SEM, LM 
TKO V6-36 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P33 TEM, LM 
TKO V3-21 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P64 TEM, LM 
TKO VIII10-71 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/− P70 SEM, LM 
TKO IV32-21 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/− P79 TEM, LM 
TKO VIII8-59 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P90 SEM, LM 
TKO V1-4 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/− P94 TEM 
TKO BS-441 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P107 TEM, LM 
TKO VI13-88 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/− P157 SEM, LM 
5 allele VIII2-11 Chx10-Cre;RbLox/Lox;p107+/−;p53Lox/Lox P133 SEM, LM, TEM 
5 allele VII11-76 Chx10-Cre;RbLox/−;p107+/−;p53Lox/Lox P157 SEM, LM, TEM 
5 allele VII8-44 Chx10-Cre;RbLox/Lox;p107+/−;p53Lox/Lox P196 SEM, LM, TEM 
5 allele IV7-38 Chx10-Cre;RbLox/−;p107+/−;p53Lox/− P233 TEM, LM 
DKO BS-6A Chx10-Cre;RbLox/Lox;p107−/− P4 TEM 
DKO BS-6 Chx10-Cre;RbLox/Lox;p107−/− P4 TEM 
DKO BS-9 Chx10-Cre;RbLox/−;p107−/− P12 TEM 
DKO BS-344 Chx10-Cre;RbLox/Lox;p107−/− P43 LM, TEM 
DKO BS-439 Chx10-Cre;RbLox/Lox;p107−/− P73 LM, TEM 
DKO BS-160 Chx10-Cre;RbLox/−;p107−/− P233 LM, TEM 
DKO BS-179 Chx10-Cre;RbLox/Lox;p107−/− P257 LM, TEM 
Abbreviation*Animal IDGenotypeAgeAnalysis
TKO V7-42 Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox P13 SEM, LM 
TKO V7-44 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P13 SEM, LM 
TKO V6-36 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P33 TEM, LM 
TKO V3-21 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P64 TEM, LM 
TKO VIII10-71 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/− P70 SEM, LM 
TKO IV32-21 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/− P79 TEM, LM 
TKO VIII8-59 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P90 SEM, LM 
TKO V1-4 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/− P94 TEM 
TKO BS-441 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/Lox P107 TEM, LM 
TKO VI13-88 Chx10-Cre;RbLox/Lox;p107−/−;p53Lox/− P157 SEM, LM 
5 allele VIII2-11 Chx10-Cre;RbLox/Lox;p107+/−;p53Lox/Lox P133 SEM, LM, TEM 
5 allele VII11-76 Chx10-Cre;RbLox/−;p107+/−;p53Lox/Lox P157 SEM, LM, TEM 
5 allele VII8-44 Chx10-Cre;RbLox/Lox;p107+/−;p53Lox/Lox P196 SEM, LM, TEM 
5 allele IV7-38 Chx10-Cre;RbLox/−;p107+/−;p53Lox/− P233 TEM, LM 
DKO BS-6A Chx10-Cre;RbLox/Lox;p107−/− P4 TEM 
DKO BS-6 Chx10-Cre;RbLox/Lox;p107−/− P4 TEM 
DKO BS-9 Chx10-Cre;RbLox/−;p107−/− P12 TEM 
DKO BS-344 Chx10-Cre;RbLox/Lox;p107−/− P43 LM, TEM 
DKO BS-439 Chx10-Cre;RbLox/Lox;p107−/− P73 LM, TEM 
DKO BS-160 Chx10-Cre;RbLox/−;p107−/− P233 LM, TEM 
DKO BS-179 Chx10-Cre;RbLox/Lox;p107−/− P257 LM, TEM 

Abbreviation: LM, light microscopy.

*

TKO represents mice with conditional inactivation of Rb and p53 in the developing retina and p107 null. Five-allele mice are the same, except they have one wild-type copy of p107. DKO mice are lacking Rb in the developing retina, and the background is p107-null.

Age refers to postnatal (P) days.

All three genotypes showed significant disruptions in retinal morphology by P12 to P13 (Table 1; data not shown; ref. 10). Tumor size varied, depending on the animal's age and genotype. One of the most informative SEM samples (P196: Chx10-Cre;RbLox/−;p107+/−;p53Lox/Lox) contained an early-stage tumor with a pronounced vitreal protrusion (Fig. 1A). In normal eyes, contiguous Müller glial cells form the inner limiting membrane (ILM), which appears as a smooth, unbroken surface on the SEM image. In diseased eyes, retinoblastoma cells ruptured the ILM, particularly near the apex of the protrusion, where tumor cells could invade the vitreous (Fig. 1B). At these breach points, the retinoblastoma cell bodies and associated vasculature were exposed to view (Fig. 1B, enlargement 1). This early-stage tumor also contained structures that resembled neuronal processes (Fig. 1B, enlargement 2) similar to those found in the plexiform layers of a normal retina (Fig. 1C). These neurite-like structures were also present in more advanced tumors but were not found in the cells that had invaded the surrounding tissue (e.g., the anterior chamber; Table 1; Supplementary Figs. S1 and S2).

Figure 1.

Scanning electron micrographs of an early-stage tumor. A, SEM image of an early-stage tumor extending into the vitreous of a Chx10-Cre;RbLox/−;p53Lox/−p107−/− mouse. The anterior chamber of this eye also contained tumor (arrow) that invaded from a tumor in the other half of the orbit. From a different angle (middle), the main tumor extension (arrow) and secondary tumor peak (open arrowhead) were clearly visible. Right, higher-magnification view of the boxed region in (A). The ILM was ruptured in several places, including the peak of the tumor mass extension into the vitreous (arrows). In addition, there were visible stress marks in the ILM surrounding this tumor peak (open arrowheads). B, a magnified view of the boxed region in (A, right) shows a tumor-associated blood vessel and the disruption of the ILM as this retinoblastoma emerges into the vitreous. Adjacent to the tumor-associated blood vessel (enlargement 1) are tumor cell bodies (arrows). The tumor also contained regions of plexus (enlargement 2) similar to that seen in the inner plexiform layer (IPL) of the normal retina. C, SEM of a normal retina processed side-by-side with the tumor-bearing eyes. Amacrine cell bodies (arrow) and inner plexiform layer (open arrowhead) resemble the tumor cell bodies and tumor-associated plexus, respectively, in mouse retinoblastomas. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer.

Figure 1.

Scanning electron micrographs of an early-stage tumor. A, SEM image of an early-stage tumor extending into the vitreous of a Chx10-Cre;RbLox/−;p53Lox/−p107−/− mouse. The anterior chamber of this eye also contained tumor (arrow) that invaded from a tumor in the other half of the orbit. From a different angle (middle), the main tumor extension (arrow) and secondary tumor peak (open arrowhead) were clearly visible. Right, higher-magnification view of the boxed region in (A). The ILM was ruptured in several places, including the peak of the tumor mass extension into the vitreous (arrows). In addition, there were visible stress marks in the ILM surrounding this tumor peak (open arrowheads). B, a magnified view of the boxed region in (A, right) shows a tumor-associated blood vessel and the disruption of the ILM as this retinoblastoma emerges into the vitreous. Adjacent to the tumor-associated blood vessel (enlargement 1) are tumor cell bodies (arrows). The tumor also contained regions of plexus (enlargement 2) similar to that seen in the inner plexiform layer (IPL) of the normal retina. C, SEM of a normal retina processed side-by-side with the tumor-bearing eyes. Amacrine cell bodies (arrow) and inner plexiform layer (open arrowhead) resemble the tumor cell bodies and tumor-associated plexus, respectively, in mouse retinoblastomas. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer.

Close modal

Mouse retinoblastomas express markers of differentiated neurons. To determine whether proteins associated with retinal synapses are expressed in mouse retinoblastoma, we immunostained sections of early-stage and late-stage tumors from Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox mice by using antibodies that recognized proteins associated with the seven major classes of retinal cell types (Table 2). We also included antibodies that recognize synapse-associated proteins expressed in different neuronal cell types (Table 2). Consistent with a previous characterization of mouse retinoblastomas (16), the following proteins typically found in horizontal/amacrine cells were expressed in the mouse retinoblastoma: Calbindin (Fig. 2A), Syntaxin-1 (Fig. 2B), synapsin-1 (Fig. 2C), Gad65, Snap25, vGlut-1, E-cadherin, and N-cadherin (Supplementary Fig. S3). The heterogeneous expression of these proteins suggests that mouse retinoblastomas contain restricted regions where tumor cells differentiation along the horizontal/amacrine cell lineage. The portions of the retinoblastomas immunonegative for amacrine/horizontal markers contained more densely packed nuclei and lacked expression of markers for rods, cones, bipolar cells, ganglion cells, Müller glia, and astrocytes. Real-time RT-PCR was used to confirm the immunostaining results and to analyze retinal progenitor cell markers Eya2, Fgf15, and Sfrp1 (Supplementary Fig. S4). The immunostaining data suggest that mouse retinoblastomas are heterogeneous: some cells differentiate and resemble amacrine/horizontal cells, and others are less differentiated.

Table 2.

Antibodies used for analysis of mouse retinoblastomas

AntigenSpeciesCell type/structureDilution*Vendor
VC1.1 Mouse Amacrine 1:100/1:4,000 Sigma (St. Louis, MO) 
Syntaxin Mouse Amacrine/horizontal 1:100/1:4,000 Sigma (St. Louis, MO) 
Calbindin Mouse Horizontal/amacrine subset 1:100/1:2,000 Sigma (St. Louis, MO) 
PKC-α Mouse Bipolar 1:100/1:4,000 Upstate (Billerica, MA) 
Bassoon Mouse Photoreceptor synapses/IPL 1:500/1:1,000 Stressgen (Victoria, BC Canada) 
Kinesin II Mouse Photoreceptor synapses 1:300/1:1,000 Covance (Princeton, NJ) 
PSD-95 Mouse Photoreceptor synapses/IPL 1:200/1:1,000 ABR (Golden, CO) 
Calretinin Mouse Amacrine subset 1:100/1:2,000 Chemicon (Temecula, CA) 
GAP43 Mouse Ganglion cells 1:100/ND Sigma (St. Louis, MO) 
MAP1 Mouse Ganglion cells 1:100/ND Chemicon (Temecula, CA) 
GFAP Mouse Astrocytes/activated Müller glia 1:100/1:1,000 Sigma (St. Louis, MO) 
Rhodopsin Mouse Rod photoreceptors 1:500/1:2,000 Dr. R.S. Molday 
Glutamine synthetase Rabbit Müller glia 1:100/1:1,000 BD Bioscience (San Jose, CA) 
Pax6 Mouse Progenitor cells/amacrine cells 1:20/1:25 Hybridoma bank (Iowa City, IA) 
Chx10 Sheep Progenitor cells/bipolar cells 1:500/1:1,000 Exalpha 
CD44 Rat Müller glia 1:100/ND Hybridoma bank (Iowa City, IA) 
Snap25 Rabbit IPL/OPL 1:1,000/1:10,000 Stressgen (Victoria, BC Canada) 
Gα0 Rabbit Bipolar cells 1:100/1:2,000 Santa Cruz Biotechnology (Santa Cruz, CA) 
GAD65 Rabbit Amacrine/IPL 1:100/1:2,000 Chemicon (Temecula, CA) 
GluR1 Rabbit Amacrine/IPL 1:100/1:2,000 Sigma (St. Louis, MO) 
Cone arrestin Rabbit Cone photoreceptors 1:5000/1:10,000 Zymed (San Francisco, CA) 
Synapsin-1 Rabbit IPL/horizontal cell processes 1:500/1:1,000 Chemicon (Temecula, CA) 
vGlut-1 Mouse IPL/OPL 1:500/1:1,000 Chemicon (Temecula, CA) 
E-cadherin Rabbit Transmembrane adhesion molecule 1:500/1:1,000 Santa Cruz Biotechnology (Santa Cruz, CA) 
N-cadherin Rabbit Transmembrane adhesion molecule 1:500/1:1,000 Santa Cruz Biotechnology (Santa Cruz, CA) 
Recoverin Rabbit Photoreceptors/bipolar subset 1:5,000/1:10,000 Chemicon (Temecula, CA) 
AntigenSpeciesCell type/structureDilution*Vendor
VC1.1 Mouse Amacrine 1:100/1:4,000 Sigma (St. Louis, MO) 
Syntaxin Mouse Amacrine/horizontal 1:100/1:4,000 Sigma (St. Louis, MO) 
Calbindin Mouse Horizontal/amacrine subset 1:100/1:2,000 Sigma (St. Louis, MO) 
PKC-α Mouse Bipolar 1:100/1:4,000 Upstate (Billerica, MA) 
Bassoon Mouse Photoreceptor synapses/IPL 1:500/1:1,000 Stressgen (Victoria, BC Canada) 
Kinesin II Mouse Photoreceptor synapses 1:300/1:1,000 Covance (Princeton, NJ) 
PSD-95 Mouse Photoreceptor synapses/IPL 1:200/1:1,000 ABR (Golden, CO) 
Calretinin Mouse Amacrine subset 1:100/1:2,000 Chemicon (Temecula, CA) 
GAP43 Mouse Ganglion cells 1:100/ND Sigma (St. Louis, MO) 
MAP1 Mouse Ganglion cells 1:100/ND Chemicon (Temecula, CA) 
GFAP Mouse Astrocytes/activated Müller glia 1:100/1:1,000 Sigma (St. Louis, MO) 
Rhodopsin Mouse Rod photoreceptors 1:500/1:2,000 Dr. R.S. Molday 
Glutamine synthetase Rabbit Müller glia 1:100/1:1,000 BD Bioscience (San Jose, CA) 
Pax6 Mouse Progenitor cells/amacrine cells 1:20/1:25 Hybridoma bank (Iowa City, IA) 
Chx10 Sheep Progenitor cells/bipolar cells 1:500/1:1,000 Exalpha 
CD44 Rat Müller glia 1:100/ND Hybridoma bank (Iowa City, IA) 
Snap25 Rabbit IPL/OPL 1:1,000/1:10,000 Stressgen (Victoria, BC Canada) 
Gα0 Rabbit Bipolar cells 1:100/1:2,000 Santa Cruz Biotechnology (Santa Cruz, CA) 
GAD65 Rabbit Amacrine/IPL 1:100/1:2,000 Chemicon (Temecula, CA) 
GluR1 Rabbit Amacrine/IPL 1:100/1:2,000 Sigma (St. Louis, MO) 
Cone arrestin Rabbit Cone photoreceptors 1:5000/1:10,000 Zymed (San Francisco, CA) 
Synapsin-1 Rabbit IPL/horizontal cell processes 1:500/1:1,000 Chemicon (Temecula, CA) 
vGlut-1 Mouse IPL/OPL 1:500/1:1,000 Chemicon (Temecula, CA) 
E-cadherin Rabbit Transmembrane adhesion molecule 1:500/1:1,000 Santa Cruz Biotechnology (Santa Cruz, CA) 
N-cadherin Rabbit Transmembrane adhesion molecule 1:500/1:1,000 Santa Cruz Biotechnology (Santa Cruz, CA) 
Recoverin Rabbit Photoreceptors/bipolar subset 1:5,000/1:10,000 Chemicon (Temecula, CA) 

Abbreviation: ND, not determined.

*

The first number is for vibratome tissue sections, and the second number is for dissociated cells.

Figure 2.

Proliferating mouse retinoblastoma cells resemble horizontal/amacrine cells. A, calbindin immunofluorescence (red) in retinoblastoma from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. Arrows indicated immunopositive cells (nuclei counterstained with SYTOX green) in the region of tumor cell differentiation (marked by the dashed line). Right, higher-magnification view of the boxed region. Calbindin is expressed in the tumor cell bodies and the surrounding plexus associated with the cells. In the normal retina, calbindin is expressed in horizontal cells and a subset of amacrine cells. B, syntaxin-1 immunofluorescence in retinoblastoma from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. Arrows indicate an immunopositive plexus in the region of tumor cell differentiation (marked by the dashed line). Right, higher magnification view of the boxed region. Syntaxin-1 is expressed in the plexus with extensive tumor cell differentiation. In the normal retina, syntaxin-1 is expressed in horizontal cells and amacrine cells. C, synapsin-1 immunofluorescence in retinoblastoma from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. Arrows indicated immunopositive plexus in the region of tumor cell differentiation (marked by the dashed line). Right, higher-magnification view. Synapsin-1 is expressed in the plexus associated with the tumor cells undergoing differentiation. In the normal retina, synapsin-1 is expressed in the plexiform layers. D, four independent mouse retinoblastomas (7-68L, 7-68R, 4-43L, and 4-43R) were pulse-labeled for 1 h with [3H]thymidine, dissociated, and immunostained using antibodies against proteins found in the different types of neurons and glia in the normal retina. Cells (n = 250) were scored in duplicate for each sample and each antibody. The proportion of [3H]thymidine-labeled cells varied from 7% to 28%. The vast majority of proliferating retinoblastoma cells expressed amacrine/horizontal cell– or retinal progenitor cell–specific markers. An example of a syntaxin-1+ cell that incorporated [3H]thymidine is shown with the total proportion of syntaxin-1+ cells (histogram). The proportion of syntaxin-1+ cells that also incorporated [3H]thymidine (white) for each sample. Rhodopsin+ rod photoreceptors never incorporated [3H]thymidine. Bar, 25 μm (AC) and 10 μm (D). DAPI, 4′,6-diamidino-2-phenylindole; [3H]-thy, [3H]thymidine.

Figure 2.

Proliferating mouse retinoblastoma cells resemble horizontal/amacrine cells. A, calbindin immunofluorescence (red) in retinoblastoma from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. Arrows indicated immunopositive cells (nuclei counterstained with SYTOX green) in the region of tumor cell differentiation (marked by the dashed line). Right, higher-magnification view of the boxed region. Calbindin is expressed in the tumor cell bodies and the surrounding plexus associated with the cells. In the normal retina, calbindin is expressed in horizontal cells and a subset of amacrine cells. B, syntaxin-1 immunofluorescence in retinoblastoma from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. Arrows indicate an immunopositive plexus in the region of tumor cell differentiation (marked by the dashed line). Right, higher magnification view of the boxed region. Syntaxin-1 is expressed in the plexus with extensive tumor cell differentiation. In the normal retina, syntaxin-1 is expressed in horizontal cells and amacrine cells. C, synapsin-1 immunofluorescence in retinoblastoma from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. Arrows indicated immunopositive plexus in the region of tumor cell differentiation (marked by the dashed line). Right, higher-magnification view. Synapsin-1 is expressed in the plexus associated with the tumor cells undergoing differentiation. In the normal retina, synapsin-1 is expressed in the plexiform layers. D, four independent mouse retinoblastomas (7-68L, 7-68R, 4-43L, and 4-43R) were pulse-labeled for 1 h with [3H]thymidine, dissociated, and immunostained using antibodies against proteins found in the different types of neurons and glia in the normal retina. Cells (n = 250) were scored in duplicate for each sample and each antibody. The proportion of [3H]thymidine-labeled cells varied from 7% to 28%. The vast majority of proliferating retinoblastoma cells expressed amacrine/horizontal cell– or retinal progenitor cell–specific markers. An example of a syntaxin-1+ cell that incorporated [3H]thymidine is shown with the total proportion of syntaxin-1+ cells (histogram). The proportion of syntaxin-1+ cells that also incorporated [3H]thymidine (white) for each sample. Rhodopsin+ rod photoreceptors never incorporated [3H]thymidine. Bar, 25 μm (AC) and 10 μm (D). DAPI, 4′,6-diamidino-2-phenylindole; [3H]-thy, [3H]thymidine.

Close modal

Based on the number of cells immunopositive for each of the amacrine/horizontal cell markers (1.7–3.1 × 106 per tumor), it is unlikely that normal retinal neurons were embedded in the growing tumors. Far fewer amacrine/horizontal cells are present in the normal mouse retina (17).

To determine whether the differentiated tumor cells proliferate or are quiescent, we injected [3H]thymidine into tumor-bearing Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. One hour later, tumors were isolated, dissociated, and immunostained for the markers of amacrine/horizontal cell differentiation (Fig. 2D; Supplementary Fig. S5). A substantial proportion of cells expressing amacrine/horizontal cell differentiation markers incorporated [3H]thymidine (Fig. 2D; Supplementary Fig. S5). For example, 20% to 34% of syntaxin-1+ cells incorporated [3H]thymidine (Fig. 2D). A small proportion of cells expressing photoreceptor markers were adjacent to the retinal pigmented epithelium (data not shown); these cells did not incorporate [3H]thymidine (Fig. 2D; Supplementary Fig. S5). Similar data were obtained for bipolar cells and Müller glia (Supplementary Fig. S5). The GFAP+ cells that were [3H]thymidine+ were probably astrocytes rather than Müller glia undergoing reactive gliosis because very few (if any) glutamine synthetase+ Müller glia incorporated [3H]thymidine. All of the genotypes tested provided similar results.

Murine retinoblastoma cells extend processes. Results from the SEM analysis, immunostaining, real-time RT-PCR, and dissociated cell scoring indicated that mouse retinoblastoma cells differentiate along the amacrine/horizontal cell lineages. However, the term “differentiation,” when applied to these cells in normal retina, means that they not only express certain markers but also extend processes and form synapses. To determine whether retinoblastoma cells extend processes characteristic of horizontal/amacrine cells, we Golgi-Cox stained six tumors from Chx10-Cre;RbLox/−;p107−/− mice and six from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. Golgi-Cox staining allows for the visualization of individual neuronal processes and neurites. All of the labeled cells extended processes (Fig. 3A). Approximately 18% (45 of 250) extended one to three long (>5 cell body lengths) main processes from the cell body, with further neurite branching characteristic of horizontal cells or wide-field amacrine cells (Fig. 3A and B; ref. 18). Forty-six percent (115 of 250) of the Golgi-Cox–labeled cells extended a main process with extensive neurite outgrowth, which is characteristic of amacrine cells (Fig. 3C). The remaining cells were less differentiated with short (<1 cell body length) unbranched neurites (data not shown). Just as in the immunostained tumor sections, the Golgi-Cox–stained cells were found near the tumor origin; fewer Golgi-Cox–labeled cells were present toward the lens and anterior chamber (Fig. 3D). This finding is consistent with the loss of the differentiated tumor phenotype as retinoblastoma cells become more invasive and aggressive.

Figure 3.

Golgi-Cox staining of mouse retinoblastoma. A, typical Golgi-Cox–stained mouse retinoblastoma cells from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. The image from one plane of focus is adjacent to a tracing from all planes of focus compressed into a two-dimensional image. B, tracing of Golgi-Cox–labeled tumor cells with large cell bodies, which is consistent with horizontal cell morphology. In one example, a few large, long main processes had neurites extending off of them. C, in contrast, some of the tumor cells extended smaller neurites from a single location on the cell body, which is consistent with amacrine cell differentiation. D, map of Golgi-Cox–labeled cells in three serial sections from a Chx10-Cre;RbLox/−;p53Lox/−;p107−/− tumor. The pigmented epithelium is outlined in black, and the location of the lens would be near the bottom of each trace. Each dot represents a single Golgi-Cox–labeled tumor cell. Bar, 10 μm (A–J) and 1 mm (K–M). RPE, retinal pigment epithelium.

Figure 3.

Golgi-Cox staining of mouse retinoblastoma. A, typical Golgi-Cox–stained mouse retinoblastoma cells from Chx10-Cre;RbLox/−;p53Lox/−;p107−/− mice. The image from one plane of focus is adjacent to a tracing from all planes of focus compressed into a two-dimensional image. B, tracing of Golgi-Cox–labeled tumor cells with large cell bodies, which is consistent with horizontal cell morphology. In one example, a few large, long main processes had neurites extending off of them. C, in contrast, some of the tumor cells extended smaller neurites from a single location on the cell body, which is consistent with amacrine cell differentiation. D, map of Golgi-Cox–labeled cells in three serial sections from a Chx10-Cre;RbLox/−;p53Lox/−;p107−/− tumor. The pigmented epithelium is outlined in black, and the location of the lens would be near the bottom of each trace. Each dot represents a single Golgi-Cox–labeled tumor cell. Bar, 10 μm (A–J) and 1 mm (K–M). RPE, retinal pigment epithelium.

Close modal

Synapse formation in mouse retinoblastoma. Retinoblastoma tumors in Chx10-Cre;RbLox/−;p107−/− mice were characterized by pronounced loss of photoreceptor cells, appearance of unique populations of undifferentiated tumor-like cells, and extensive formation of plexiform regions. Smaller tumors were more common, but in some instances, a large tumor filled much of the vitreal space (Fig. 4A). TEM images of a Chx10-Cre;RbLox/−;p107−/− tumor revealed two morphologically distinct cell types: stage I retinoblastoma cells which had pale, round nuclei and resembled differentiated neurons (Fig. 4B), were almost exclusively associated with a plexus. This is consistent with a differentiated tumor phenotype along amacrine/horizontal cell lineages. Stage II retinoblastoma cells contained more darkly stained, irregular nuclei and were packed tightly together with little or no plexus associated with individual cells (Fig. 4B). This morphology is characteristic of a less differentiated tumor cell phenotype.

Figure 4.

Synaptogenesis in Chx10-Cre;RbLox/−;p107−/− mouse retinoblastoma. A, light microscopic image of a large tumor from a Chx10-Cre;RbLox/−; p107−/− mouse. The remaining retina merged with an aggregate of densely packed tumor cell bodies surrounded by lighter-stained regions with a large, expanded plexus (upper panel). More commonly observed in this genotype were smaller focal tumors with limited vitreal invasion. Tumors from Chx10-Cre;RbLox/−; p107−/− mice were heterogeneous. They contained regions of densely packed cells and regions of more sparsely packed cells interspersed with plexus visible at the light microscopic level (lower panel). B, TEM images (upper and lower panels) revealed two distinct cellular morphologies. Stage I cells were more differentiated tumor cells with large round nuclei and prominent nucleoli. These cells were associated with plexus. Stage II cells were densely packed with little plexus and more irregularly shaped nuclei. C, the plexus regions in Chx10-Cre;RbLox/−; p107−/− tumors (shown at three increasing magnifications) were filled with synaptic densities among adjoining neurites that were associated with clusters of synaptic vesicles. Early-stage retinoblastoma cells extend neurites and seem to form synapses, which are characteristic of amacrine/horizontal cells.

Figure 4.

Synaptogenesis in Chx10-Cre;RbLox/−;p107−/− mouse retinoblastoma. A, light microscopic image of a large tumor from a Chx10-Cre;RbLox/−; p107−/− mouse. The remaining retina merged with an aggregate of densely packed tumor cell bodies surrounded by lighter-stained regions with a large, expanded plexus (upper panel). More commonly observed in this genotype were smaller focal tumors with limited vitreal invasion. Tumors from Chx10-Cre;RbLox/−; p107−/− mice were heterogeneous. They contained regions of densely packed cells and regions of more sparsely packed cells interspersed with plexus visible at the light microscopic level (lower panel). B, TEM images (upper and lower panels) revealed two distinct cellular morphologies. Stage I cells were more differentiated tumor cells with large round nuclei and prominent nucleoli. These cells were associated with plexus. Stage II cells were densely packed with little plexus and more irregularly shaped nuclei. C, the plexus regions in Chx10-Cre;RbLox/−; p107−/− tumors (shown at three increasing magnifications) were filled with synaptic densities among adjoining neurites that were associated with clusters of synaptic vesicles. Early-stage retinoblastoma cells extend neurites and seem to form synapses, which are characteristic of amacrine/horizontal cells.

Close modal

Rosettes in the Chx10-Cre;RbLox/−;p107−/− retinoblastomas were usually made up of stage I cells with a central plexus but were also found adjacent to clusters of stage II cells (Fig. 4B). TEM images of the Chx10-Cre;RbLox/−;p107−/− retinoblastoma also revealed mitotic figures and apoptotic cells, which are hallmarks of human retinoblastoma. Higher-magnification TEM images revealed that the plexus regions of Chx10-Cre;RbLox/−;p107−/− tumors not only contained processes that resembled those found in amacrine/horizontal cells but also all of the plexus regions contained synaptic densities and associated synaptic vesicles (Fig. 4C). These data are consistent with the immunofluorescence studies (Fig. 2; Supplementary Fig. S3) that suggested that these retinoblastoma cells express markers found in neurites and synapses. They also indicate that Chx10-Cre;RbLox/−;p107−/− retinoblastomas show an unexpected degree of differentiation.

Invasion and tumor cell differentiation in Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox retinoblastomas.Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox retinoblastoma is much more aggressive and invasive than Chx10-Cre;RbLox/−;p107−/− retinoblastoma (5, 6). This invasion includes the retina and subretinal space (Supplementary Fig. S6A), the anterior chamber, and the subretinal pigmented epithelium compartment (Supplementary Fig. S6B). As in human retinoblastoma, Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox mice showed rapid filling of the vitreal cavity with densely packed tumor cells and rosettes (Supplementary Fig. S6CE). A prominent feature of the Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox retinoblastomas was an overabundance of stage II cells. This finding is consistent with the hypothesis that inactivation of p53 promotes tumor progression to a less differentiated state or expansion of a progenitor cell during retinal development (9).

Regions of plexus with synaptic densities and synaptic vesicles were present near the outer surface of the neural retina, where the tumor initiated (Supplementary Fig. S7). The morphologic features of these neurites and synapses were indistinguishable from those in Chx10-Cre;RbLox/−;p107−/− tumors. When examined at high magnification (×10,000), areas of the plexus within the posterior chamber in all three genotypes consisted of neuron-like processes with synaptic structures reminiscent of horizontal/amacrine cells. This was the case in the extensive plexus areas characteristic of Chx10-Cre; RbLox/−;p107−/− tumors and in smaller areas of plexus characteristic of p53-deficient retinae. Most processes were small (<0.5 μm in diameter), but large ones (1–3 μm) were also noted occasionally. We observed a variety of synaptic arrangements (e.g., en passon, serial, and reciprocal contacts), all of which involved contacts among processes (Supplementary Fig. S7). No clear examples of axo-somatic contacts were observed. Rare instances of ribbon synapses were observed only in areas near remaining photoreceptor cell bodies. These data support the previous finding by Berns et al. that mouse retinoblastomas resemble amacrine/horizontal cells, at least during early tumorigenesis (16).

Tumor invasion and progression is accompanied by retinoblastoma cell dedifferentiation. We analyzed and compared the morphologic features of retinoblastoma cells in the anterior chamber of the Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox eye with those of early-stage Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox tumor cells and Chx10-Cre;RbLox/−;p107−/− tumor cells. Tumor cells that invaded the anterior chamber beneath the cornea were densely packed stage II cells surrounded by sparse regions of plexus (Fig. 5A and B). Interestingly, there were no synaptic densities or synaptic vesicles in the plexus or rosettes of the invading cells. In addition, metabolic activity was high, as indicated by their abundant mitochondria and mitotic figures (Fig. 5A and B; data not shown). Even in rosettes with a very large central plexus, there was no evidence of cell differentiation (Fig. 5B). In the center of the plexus were several large processes (1–3 μm), fewer small processes (<0.5 μm), and abundant mitochondria (Fig. 5B). This suggests that rosette formation is not necessarily a feature of differentiated retinoblastoma cells, as previously believed. We found rosettes with a central plexus made up of large, undifferentiated processes and rosettes with a central plexus containing neurites and synapses. Therefore, we propose that rosettes result from junctional complexes linking adjacent somata, which is a hallmark of a wide variety of primitive neuroectodermal tumors.

Figure 5.

Invasive, stage II retinoblastoma cells dedifferentiate. A, light micrographs of an anterior chamber tumor from a Chx10-Cre;RbLox/−; p107−/−;p53Lox/Lox mouse. The tumor was made up of densely packed cells and sparse regions of plexus. TEM analysis revealed that the plexus contained processes but no synaptic densities or synaptic vesicles. This finding is consistent with the less differentiated stage II cells. B, similarly, rosettes in the anterior chamber contained a central region of plexus made up of both large and small processes but lacking the features of the more differentiated tumors (i.e., synaptic densities and synaptic vesicles). These anterior chamber tumor cells also contained a large number of mitochondria and an increased proportion of mitotic figures. C, tumor cells in the posterior region of a late-stage retinoblastoma from a Chx10-Cre;RbLox/−; p107−/−;p53Lox/Lox mouse contained extensive processes, synaptic densities, and synaptic vesicles.

Figure 5.

Invasive, stage II retinoblastoma cells dedifferentiate. A, light micrographs of an anterior chamber tumor from a Chx10-Cre;RbLox/−; p107−/−;p53Lox/Lox mouse. The tumor was made up of densely packed cells and sparse regions of plexus. TEM analysis revealed that the plexus contained processes but no synaptic densities or synaptic vesicles. This finding is consistent with the less differentiated stage II cells. B, similarly, rosettes in the anterior chamber contained a central region of plexus made up of both large and small processes but lacking the features of the more differentiated tumors (i.e., synaptic densities and synaptic vesicles). These anterior chamber tumor cells also contained a large number of mitochondria and an increased proportion of mitotic figures. C, tumor cells in the posterior region of a late-stage retinoblastoma from a Chx10-Cre;RbLox/−; p107−/−;p53Lox/Lox mouse contained extensive processes, synaptic densities, and synaptic vesicles.

Close modal

Eyes from Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox mice with anterior chamber invasion still contained differentiated tumor cells (Fig. 5C). This finding suggests that stage I retinoblastoma cells persist during later stages of tumorigenesis. Retinoblastoma may also originate from different cells during development (e.g., one cell of origin may be a cell committed to the amacrine/horizontal cell lineage, and another may be an immature retinal progenitor cell).

Human retinoblastomas resemble late-stage mouse retinoblastomas. One advantage of studying tumorigenesis in animal models is that we can characterize the earliest stages of the disease that precede clinical diagnosis in humans. Early-stage mouse retinoblastoma is characterized by extensive cell differentiation, including neurite extension and synaptogenesis. In addition, inactivation of p53 promotes tumor progression and invasion, which is accompanied by expansion of less differentiated cells. Based on these preclinical studies, we predict that human retinoblastomas resemble stage II, less differentiated tumor phenotype. To test this hypothesis, we did TEM on seven human retinoblastomas from primary enucleations done before the patient receive any treatment.

At the light microscopic level, these tumors contained densely packed tumor cells and small plexiform regions (Fig. 6A). One unique feature of the human retinoblastoma cells was the extensive cell-cell junctions between cell bodies (Fig. 6B). Junctions were observed in mouse retinoblastomas (Supplementary Fig. S8), but they were not as extensive along the cell body as those found in human tumors. Interestingly, there were many examples of human tumor cells invading across the ILM, and those cells were devoid of junctional complexes (Fig. 6C). This observation raises the possibility that junctional stability is concomitantly lost during tumor cell invasion. Like mouse retinoblastoma cells invading the anterior chamber, human retinoblastoma cells contained abundant mitochondria and mitotic figures (Fig. 6B; data not shown) and often were organized into rosettes, which were made up of a central plexus and a surrounding row of tumor cells (Fig. 6D). As in the mouse tumors, the human tumors contained a central plexus made up of large- and small-diameter processes, but we found no evidence for synaptic densities or synaptic vesicles in any of the seven human retinoblastomas analyzed by TEM.

Figure 6.

Human retinoblastomas resemble late-stage mouse tumors. TEM analysis was done on seven human tumors that were removed from the patient before any treatment. A, light microscopic analysis revealed that the main mass of tumor cells was infiltrated with large blood vessels. It also showed little organized structure and no recognizable retinal lamination. The major cell type had pleomorphic nuclei and limited cytoplasm, but all of the tumors contained small regions of plexus. B, in TEM images, the retinoblastoma cells were remarkably similar to those of the late-stage mouse tumors in the anterior chamber (i.e., they contained abundant mitochondria and were devoid of synaptic densities and synaptic vesicles). Both large and small processes extended from the cells, but no synaptogenesis was detected. One major difference was the extensive junctional complexes found between cells in the human retinoblastomas; mouse retinoblastoma cells were characterized by much smaller cell-cell junctions. C, tumor cells that were actively invading the retina or surrounding tissues lacked extensive junctional complexes, suggesting a possible role in invasion and metastasis. D, rosettes were organized around a central plexus, which consisted of both large and small processes but lacked synaptic densities and synaptic vesicles.

Figure 6.

Human retinoblastomas resemble late-stage mouse tumors. TEM analysis was done on seven human tumors that were removed from the patient before any treatment. A, light microscopic analysis revealed that the main mass of tumor cells was infiltrated with large blood vessels. It also showed little organized structure and no recognizable retinal lamination. The major cell type had pleomorphic nuclei and limited cytoplasm, but all of the tumors contained small regions of plexus. B, in TEM images, the retinoblastoma cells were remarkably similar to those of the late-stage mouse tumors in the anterior chamber (i.e., they contained abundant mitochondria and were devoid of synaptic densities and synaptic vesicles). Both large and small processes extended from the cells, but no synaptogenesis was detected. One major difference was the extensive junctional complexes found between cells in the human retinoblastomas; mouse retinoblastoma cells were characterized by much smaller cell-cell junctions. C, tumor cells that were actively invading the retina or surrounding tissues lacked extensive junctional complexes, suggesting a possible role in invasion and metastasis. D, rosettes were organized around a central plexus, which consisted of both large and small processes but lacked synaptic densities and synaptic vesicles.

Close modal

Retinoblastoma differentiation. Early-stage retinoblastoma cells undergo considerable neuronal differentiation. We use the term “differentiation” to indicate that retinoblastoma cells exhibited morphological and biochemical markers indicative of a commitment to a distinctive neuronal phenotype. SEM, TEM, immunostaining, Golgi-Cox staining, and molecular analyses indicated that tumor cells in our retinoblastoma models extend neurites and form synaptic connections characteristic of amacrine/horizontal cells of the neural retina. Importantly, as the tumor progressed and invaded the anterior chamber of the eye, the cells lost these features (i.e., they still extended neurites and organized into rosettes, but they did not express proteins associated with amacrine/horizontal cell synapses, nor did they contain synaptic vesicles or form synapses). Based on these data, we propose that the loss of retinoblastoma cells' ability to differentiate is an important step in tumor progression in mice.

Human retinoblastomas resemble late-stage mouse retinoblastomas. They form rosettes and extend processes but do not form synapses. We propose that early-stage human retinoblastoma is characterized by amacrine/horizontal cell differentiation, similar to that seen in mouse tumors, but this differentiated phenotype is lost before the tumor is diagnosed. We recently presented evidence for this by showing that when the p53 pathway is inactivated in retinoblastoma cells by ectopically expressing MDMX, cells rapidly transition to a less differentiated phenotype (9).

Retinoblastoma cell of origin. Two models that are not mutually exclusive have been proposed to explain the transition from a tumor with features of differentiated amacrine/horizontal cell neurons to one with features of less differentiated cells. First, there may be one cell of origin for this tumor, and that cell expands, giving rise to differentiated tumor cells. Subsequent genetic lesions then lead to a phenotypic transition of the tumor cells to a less differentiated cell type that may subsequently invade the anterior chamber of the eye and progress to metastatic retinoblastoma. Inactivation of p53 promotes the progression to this less differentiated tumor cell phenotype in humans and mice, but the progression is incomplete. Chx10-Cre;RbLox/−;p107−/−;p53Lox/Lox animals contain neoplastic cells in invading tumors of the anterior chamber and a mixture of differentiated and neoplastic cells in the retinal regions.

The second model that explains the transition of retinoblastoma cells from differentiated to dedifferentiated cells includes two distinct cells of origin for retinoblastoma: one that is substantially more differentiated than the other. According to this model, the tumor cells that are less differentiated eventually outgrow the differentiated ones, invade the anterior chamber, and form metastatic retinoblastoma.

Synaptogenesis in retinoblastoma. There is clear evidence for synaptic connections among retinoblastoma cells in early-stage tumors. The absence of synaptic ribbons, the relatively small size of the presynaptic terminal, and the types of postsynaptic connections suggest that these synapses arise through differentiation of a retinal interneuron phenotype, such as horizontal or amacrine cells, and not from photoreceptors or bipolar cells. The pattern of synaptic connectivity within the tumor lacks the laminar arrangement seen in normal retina but nonetheless retains one of the key elements of retinal circuitry (i.e., the general absence of axo-somatic synapses). Tumor cell bodies are excluded from plexiform regions that are densely populated with synaptic connections, among processes of varying diameters.

Based on these observations, we propose that it is unlikely that retinoblastoma synaptogenesis results from nonspecific stimulation of synapse production of remaining neurons, but rather, that it reflects a specific program of large-scale, cellular differentiation of retinal interneuron-like cells. In future studies, analysis of the morphologic features of retinoblastomas in the model developed by MacPherson et al. (2), which relies on the inactivation of Rb and p130, may show distinct features of differentiation. If so, that model may provide insight into the unique and overlapping roles of Rb, p107, and p130 in the control of normal retinal proliferation and differentiation and in retinoblastoma.

Heterogeneity of retinoblastoma rosettes. Rosettes have been described as a histologic landmark of retinoblastoma. Some investigators have defined late-stage, differentiated tumor as a region in which 80% or more of the volume is occupied by rosettes, as determined by light microscopic analysis of H&E-stained sections. Our electron microscopy analysis showed that rosettes are heterogeneous structures, with respect to the morphology of cell bodies that constitute the outer ring and the cellular constituents of the central plexus. Stage I and II somata can form rosettes either separately or as a mixture; the plexus may contain a high density of synaptic connections or be devoid of them. It should also be noted that rosette formation, which often involves mature rod photoreceptors, occurs in almost all disruptive conditions within the retina. Therefore, this feature may simply reflect the inherent repulsive/attractive forces between retinal somata and neurites that, under normal circumstances, produce characteristic retinal lamination.

Given their heterogeneity and nonspecificity, rosettes may have limited value in diagnosing the stage of retinoblastoma tumor progression. Alternatively, they may be correlated with disease progression because the presence of rosettes indicates strong cell-cell adhesion rather than tumor cell differentiation. That is, tumors with rosettes are less aggressive because they have extensive cell-cell junctions that prevent invasion. Presumably, this difference in tumor histology reflects differences in genetic lesions in pathways that control cell-cell adhesion and invasion. These processes in mouse and human retinoblastomas will be addressed in future studies.

Tumor invasion. In the mouse models of retinoblastoma, we examined how emerging tumors invade surrounding tissue. The point of tumor origin invariably involved the outer plexiform layer (OPL) and cell bodies located within the OPL or along its borders. Abnormalities in this region were observed, even when the surrounding lamina seemed unchanged. Larger tumors extended toward both the vitreal and subretinal borders, along with extensive evidence of cell death in all three nuclear layers. The expansive force of the tumor was obvious in SEM images of tumors with sharp protrusions into the vitreal chamber. Rupture of the ILM at the apex, which resulted from physical stress and/or enzymatic activity, provided access of tumor cells to the vitreous and a potential avenue for vitreal seeding.

Human retinoblastoma samples contain highly advanced tumors and provide limited opportunity to examine the expanding borders between tumor and normal tissue. Nevertheless, we noted several similarities with the mouse models. In one sample, the leading edge of lateral expansion of the tumor was confined to the OPL in the adjacent retina, a finding that suggests that this region in both humans and mice is involved in (or susceptible to) early stages of tumorigenesis. In another human sample, tumor cells seemed to be in transit at the point of the ILM rupture. This interpretation is based on contortion of the cell body as it filled the small opening in the ILM. The free ends of the outer limiting membrane were also extended toward the vitreous, suggesting that the movement of the tumor cell was from the retina to the vitreous, as might be envisioned during vitreal seeding.

In summary, our data support the hypothesis that early-stage human and mouse retinoblastomas exhibit features of amacrine/horizontal cell differentiation, but these features are rapidly lost as cells become more invasive and aggressive. Although p53 does not directly regulate this process, it clearly accelerates the transition from differentiated retinoblastoma to less differentiated, invasive retinoblastoma in both humans and mice. Importantly, our observation that rosette formation does not relate to retinoblastoma tumor cell differentiation challenges the long-held belief that retinoblastomas with rosettes are less aggressive due to increased differentiation. These tumors may be less aggressive because they have extensive, stable cell-cell junctions that must be disrupted for invasion, but this does not correlate with neurite extension and synaptogenesis characteristic of early-stage retinoblastoma. This is an important distinction because it will help direct future studies toward analysis of the changes in cell-cell adhesion as retinoblastoma cells invade the optical nerve, choroids, and anterior chamber.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

M.A. Dyer is a Pew Scholar.

Grant support: NIH grant EY04867 (M.A. Dyer), National Cancer Institute Cancer Center Support grant CA21765 (M.A. Dyer), American Cancer Society grant RSG-06-03-01 (M.A. Dyer), Pearle Vision Foundation (M.A. Dyer), American Lebanese Syrian Associated Charities (M.A. Dyer), National Eye Institute Core Grant EY13080 (D.A. Johnson), and Research to Prevent Blindness.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Angie McArthur for editing the article, Dr. Marina Kedrov for various aspects of data management and for capturing many of the light microscopic images, and Lou Boykins for scanning electron microscopy imaging.

1
Chen D, Livne-Bar I, Vanderluit JL, Slack RS, Agochiya M, Bremner R. Cell-specific effects of RB or RB/p107 loss on retinal development implicate an intrinsically death-resistant cell-of-origin in retinoblastoma.
Cancer Cell
2004
;
5
:
539
–51.
2
MacPherson D, Sage J, Kim T, Ho D, McLaughlin ME, Jacks T. Cell type-specific effects of Rb deletion in the murine retina.
Genes Dev
2004
;
18
:
1681
–94.
3
Dyer MA, Bremner R. The search for the retinoblastoma cell of origin.
Nat Rev Cancer
2005
;
5
:
91
–101.
4
Zhang J, Schweers B, Dyer MA. The first knockout mouse model of retinoblastoma.
Cell Cycle
2004
;
3
:
952
–9.
5
Dyer MA, Harbour JW. Cellular events in tumorigenesis. In: Singh AD, Damato B, Murphree AL, Perry JD, editors. Clinical ocular oncology. London: Elsevier; 2006.
6
Dyer MA, Rodriguez-Galindo C, Wilson MW. Use of preclinical models to improve treatment of retinoblastoma.
PLoS Med
2005
;
2
:
e332
.
7
Nork TM, Poulsen GL, Millecchia LL, Jantz RG, Nickells RW. p53 regulates apoptosis in human retinoblastoma.
Arch Ophthalmol
1997
;
115
:
213
–9.
8
Kato MV, Shimizu T, Ishizaki K, et al. Loss of heterozygosity on chromosome 17 and mutation of the p53 gene in retinoblastoma.
Cancer Lett
1996
;
106
:
75
–82.
9
Laurie NA, Donovan SL, Shih CS, et al. Inactivation of the p53 pathway in retinoblastoma.
Nature
2006
;
444
:
61
–6.
10
Donovan S, Schweers B, Martins R, Johnson D, Dyer MA. Compensation by tumor suppressor genes during retinal development in mice and humans.
BMC Biol
2006
;
4
:
14
.
11
Donovan S, Dyer MA. Developmental defects in Rb-deficient retinae.
Vision Res
2004
;
44
:
3323
–33.
12
Zhang J, Gray J, Wu L, et al. Rb regulates proliferation and rod photoreceptor development in the mouse retina.
Nat Genet
2004
;
36
:
351
–60.
13
Rowan S, Cepko CL. Genetic analysis of the homeodomain transcription factor Chx10 in the retina using a novel multifunctional BAC transgenic mouse reporter.
Dev Biol
2004
;
271
:
388
–402.
14
Dyer MA, Cepko CL. p57(Kip2) regulates progenitor cell proliferation and amacrine interneuron development in the mouse retina.
Development
2000
;
127
:
3593
–605.
15
Dyer MA, Cepko CL. p27Kip1 and p57Kip2 regulate proliferation in distinct retinal progenitor cell populations.
J Neurosci
2001
;
21
:
4259
–71.
16
Robanus-Maandag E, Dekker M, van der Valk M, et al. p107 is a suppressor of retinoblastoma development in pRb-deficient mice.
Genes Dev
1998
;
12
:
1599
–609.
17
Jeon CJ, Strettoi E, Masland RH. The major cell populations of the mouse retina.
J Neurosci
1998
;
18
:
8936
–46.
18
MacNeil MA, Masland RH. Extreme diversity among amacrine cells: implications for function.
Neuron
1998
;
20
:
971
–82.