|Home | About | Journals | Submit | Contact Us | Français|
In the 1970s, several human retinoblastoma cell lines were developed from cultures of primary tumors. As the human retinoblastoma cell lines were established in culture, growth properties and changes in cell adhesion were described. Those changes correlated with the ability of the human retinoblastoma cell lines to invade the optic nerve and metastasize in orthotopic xenograft studies. However, the mechanisms that underlie these changes were not determined. We used the recently developed knockout mouse models of retinoblastoma to begin to characterize the molecular, cellular, and genetic changes associated with retinoblastoma tumor progression and optic nerve invasion. Here we report the isolation and characterization of the first mouse retinoblastoma cell lines with targeted deletions of the Rb family. Our detailed analysis of these cells as they were propagated in culture from the primary tumor shows that changes in cadherin-mediated cell adhesion are associated with retinoblastoma invasion of the optic nerve prior to metastasis. In addition, the same changes in cadherin-mediated cell adhesion correlate with the invasive properties of the human retinoblastoma cell lines isolated decades ago, providing a molecular mechanism for these earlier observations. Most importantly, our studies are in agreement with genetic studies on human retinoblastomas, suggesting that changes in this pathway are involved in tumor progression.
Retinoblastoma is a rare childhood cancer of the eye that can begin in utero and is diagnosed during the first few years of life. Genetic studies of families with inherited retinoblastoma led to the cloning of the first tumor suppressor gene, RB1 (7). It is now well established that the initiating event in retinoblastomas is RB1 gene inactivation, which leads to deregulated proliferation of retinal cells in the developing eye. More recent research has demonstrated that the p53 pathway suppresses retinoblastoma tumor progression, and inactivation of the p53 pathway is an important genetic event in this cancer (16, 31). Specifically, the MDMX or MDM2 gene is amplified in about 75% of human retinoblastomas (16), and this amplification suppresses p53-mediated cell death. The definition of amplification in this study was a ratio of MDMX to its corresponding centromere of >2, as described in other cancer genetic studies (13). These studies illustrate how molecular, cellular, and genetic studies on primary human retinoblastoma tumors are essential for elucidating the genetic lesions that contribute to tumor progression. In addition, studies of human retinoblastoma cell lines were essential for analyzing chemotherapeutic drug sensitivity to develop more effective therapies for this debilitating childhood cancer (16, 17).
To date, two human retinoblastoma cell lines, Weri1 and Y79, are widely used in research. The Weri1 cell line was derived from a 1-year-old girl with no family history of retinoblastoma (21). The primary tumor was maintained in culture at a high cell density, and within a few weeks, two distinct cell populations developed, adherent cells and nonadherent cells. This heterogeneity in cell adherence has been reported in virtually every description of primary human retinoblastoma cultures (10, 21, 26). The immortal-suspension cells from this patient were eventually called Weri1 cells, and the adherent cells were not maintained. The Weri1 cells have a stable diploid chromosome number of 46 (21).
The Y79 cell line was derived from a 2-year-old girl with a family history of retinoblastoma (26). As with the primary retinoblastoma culture that gave rise to the Weri1 cell line, the Y79 primary culture was made up of adherent and nonadherent cells. After several weeks in culture, the nonadherent cells were isolated and grown separately. During the first several months in culture, they organized into rosettes that resembled the classic histologic feature of primary retinoblastoma tumors (26). However, this property was lost as the primary culture expanded during subsequent months. It has been well established that retinoblastoma rosettes are formed by extensive cell-cell junctions (14); thus, the Y79 cells may have lost some of their cell adhesion properties during the establishment of the cell line.
It is important to note that in the original isolation and characterization of Y79 and Weri1, the cells were not clonally derived. This suggests that the primary cultures were most likely heterogeneous, and over subsequent decades, more homogeneous cell lines or populations have emerged to give rise to what we currently refer to as the Y79 and Weri1 cell lines.
A more recent effort to identify and characterize retinoblastoma cell lines improved the efficiency of establishing cell lines by using human fibroblast feeder layers during the primary culture period (10) and provided a more careful analysis of the cells during the early phases of growth in culture. These data support the idea that the Y79 cells have undergone significant changes and/or selection in culture that distinguish them from the primary human retinoblastoma. Unfortunately, the cell lines established by Griegel et al. are not available from the ATCC, and due to their slow growth, these cells are much more difficult to work with in the laboratory (10).
There is one report of a mouse cell line from the T-antigen transgenic model (15), but no mouse retinoblastoma cell lines from knockout mouse models (3, 18, 31) have been reported. Using a conditional-inactivation approach, we developed the first knockout mouse model of retinoblastoma in Chx10-Cre; RbLox/Lox; p107−/− mice and showed that the p53 pathway suppressed retinoblastoma tumorigenesis in Chx10-Cre; RbLox/−; p107−/−; p53Lox/− mice (16, 31). These new mouse models of retinoblastoma provided us with a unique opportunity to develop mouse retinoblastoma cell lines to complement the existing human retinoblastoma cell lines. More importantly, they provide us with an opportunity to study the changes in cell adhesion that occur in cultured retinoblastoma cell lines and determine if these changes affect their ability to invade the optic nerve and metastasize in vivo. Here we describe the isolation and characterization of two nonadherent retinoblastoma cell lines, SJmRBL-3 and SJmRBL-8, which have molecular, cellular, and morphological similarities to the mouse primary tumors and human retinoblastoma cell lines. We also present the characterization of two adherent cell lines, SJmRBL-10 and SJmRBL-12, which resemble the tumor-associated cells identified previously in human primary tumor cultures. We also show evidence for changes in cadherin-mediated cell adhesion associated with retinoblastoma growth in culture, and we demonstrate that these changes may play a role in optic nerve invasion in vivo. These new cell lines will serve as a valuable tool to study retinoblastoma metastasis and provide direct evidence for changes in cadherin-mediated cell adhesion associated with retinoblastoma invasion. Our data also complement the previous data on cadherin expression in human retinoblastoma cell lines and tumors (28, 30) and extend these studies to include Cdh11, mouse retinoblastomas, and an in vivo demonstration of the role of cadherins in retinoblastoma optic nerve invasion.
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/+; p107−/−; p53Lox/− mice has been described elsewhere (31).
An 8-month-old Chx10-Cre; RbLox/+; p107−/−; p53Lox/Lox mouse with advanced bilateral retinoblastoma was euthanized. One eye was removed, and 1.2 × 106 tumor cells were dispersed and maintained in RPMI medium supplemented with 10% fetal calf serum and penicillin-streptomycin-l-glutamine at a density of 1 × 105/ml at 37°C with 5% CO2. One week later, the culture was expanded, and when those cultures reached confluence 2 weeks after the tumor was harvested, the adherent and nonadherent pools were separated. Each pool was then expanded over several weeks, and individual clones were isolated by plating at low density (10,000 cells/ml). For the nonadherent cells, the tissue culture dishes were treated with poly-l-lysine to allow the cells to adhere to the plastic. As individual colonies grew from single cells on the plates, they were transferred to 48-well dishes and expanded individually to ultimately give rise to two nonadherent clones (SJmRBL-3 and SJmRBL-8) and two adherent clones (SJmRBL-10 and SJmRBL-12).
To immunostain the retinoblastoma cells, we treated 8-chamber glass slides with poly-l-lysine to facilitate adhesion of the cells. The cells on the slides were fixed in 4% paraformaldehyde and processed for immunostaining as described previously (5). The dilutions of each antibody were described previously, and more detailed information is available at www.stjude.org/dyer.
Real-time reverse transcriptase (RT)-PCR experiments were performed using the ABI 7900 HT sequence detection system (Applied Biosystems, Foster City, CA). Primers and probes were designed using Primer Express software (Applied Biosystems). TaqMan probes were synthesized with 5′ 6-carboxyfluorescein and 3′ black hole quencher. RNA was prepared using Trizol, and cDNA was synthesized using the Superscript system (Invitrogen, Carlsbad, CA). Samples were analyzed in duplicate and normalized to Gapdh, Gpi1, and Mmt2 expression levels.
The 3T3 and SJmRBL-8 cells were infected with the retroviruses murine stem cell virus (MSCV)-RasD12-IRES-EGFP and MSCV-IRES-EGFP (generous gifts of Martine Roussel). The day before infection, 3T3 cells were plated at 2 × 105 cells per 10-cm dish and retinoblastoma cells at 2 × 106 per 10-cm dish. Three days after infection, green fluorescent protein-positive (GFP+) cells were purified by fluorescence-activated cell sorting (FACS), and 1.0 × 105 cells were resuspended in 1 ml 0.3% low-melting-point agarose in RPMI 1640 medium containing 5% fetal bovine serum and plated on agar plates. One day after plating the cells, we added 2 ml RPMI 1640 containing 10% fetal bovine serum to the agar plates. Fourteen days after plating the cells, we visualized the colonies with MTT [3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide] and scored them.
Spectral karyotyping of mouse chromosomes was performed using a SkyPaint kit (Applied Spectra Imaging, Vista, CA) per the procedure recommended by the manufacturer. Probes were detected using the SkyPaint probes, as described by the manufacturer. Images were acquired with a fluorescence microscope (Nikon) equipped with an interferometer (Spectra Cube; Applied Spectral Imaging) and a custom-designed filter cube (Chroma Technology Corporation, Rockingham, VT). SKY analysis was done using SKY View version 2.1 software (Applied Spectral Imaging).
Array comparative genome hybridization (CGH) analysis was performed using a 244A mouse genome CGH microarray kit (Agilent Technologies, Santa Clara, CA). This high-resolution 60-mer-oligonucleotide-based microarray contains approximately 235,000 probes representing coding and noncoding mouse sequences, with a 6.4-kb average probe spatial resolution. Probe content was obtained from the UCSC mm7 mouse genome (NCBI build 35; August 2005). Labeling and hybridization were performed according to the protocol (v4.0) provided by Agilent Technologies. Briefly, 2 μg test DNA and 2 μg reference DNA (female C57BL/6J; 0000664; Jackson Labs) were digested with AluI and RsaI (Promega, Madison, WI) for 2 h at 37°C. The digested DNA was labeled by random priming using an Agilent Plus genomic DNA labeling kit. Reference DNA and test DNA were labeled with Cy3-dUTP and Cy5-dUTP, respectively. Labeled products were purified by Microcon YM-30 filters (Millipore, Billerica, MA). Reference DNA and test DNA were combined and hybridized with 50 μg mouse CotI DNA at 65°C with rotation for 40 h. Washing was performed according to the Agilent protocol. Arrays were analyzed using a Genepix 4000B scanner (Molecular Devices, Sunnyvale, CA) and Agilent feature extraction software (v9.1). Results were generated using Agilent CGH Analytics software (v.3.4).
Expression analysis of RNA samples was conducted using the MEEBO (Mouse Exonic Evidence Based Oligonucleotide) microarray obtained from the Washington University School of Medicine Genome Sequencing Center (Seattle, WA). This microarray contains 38,467 70-mer oligonucleotide probes derived mostly from constitutively expressed exons representing approximately 25,000 mouse genes and was designed to enable the study of mouse transcription patterns. Reference RNA and test RNA samples were amplified using an Ambion Message Amp II RNA amplification kit (Applied Biosystems, Foster City, CA) according to the manufacturer's procedure (version 0309a). Reference RNA (5 μg) and test-amplified RNA (5 μg) were labeled with aminoallyl-labeled nucleotides via first-strand cDNA synthesis followed by a coupling of the aminoallyl groups to either Cy3 for the reference sample or Cy5 for the test sample. After labeling, the reference and test samples were combined for hybridization to the microarray. Details of the labeling protocol can be found at http://www.hartwellcenter.org/bio_services/fungen/docs/M004.10%20SJCRH%20Revised%20amino-allyl%20labeling%20of%20RNA%20for%20web.doc.
MEEBO microarrays were pretreated for 1 h at 42°C in 25% formamide, 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), 0.1% sodium dodecyl sulfate, 0.1 mM dithiothreitol, and 0.1 mg/ml sheared salmon sperm DNA to reduce background before hybridization with cyanine-labeled samples at 42°C for 18 h. The microarrays were then washed via shaking in 2× SSC and 0.1% sodium dodecyl sulfate at 42°C for 5 min, then in 1× SSC at room temperature for 5 min, and then in 0.1× SSC at room temperature for 2 min. Microarrays were spun dry in a slide spinner microcentrifuge (CLP-PGC Scientifics, San Diego, CA). The arrays were analyzed using a Genepix 4000B scanner (Molecular Devices, Sunnyvale, CA) with Genepix Pro version 6.0 software.
The orthotopic retinoblastoma xenograft model has been described previously (17). Briefly, 1,000 retinoblastoma cells in 1 μl RPMI medium were injected into the vitreous of the eyes of newborn Sprague-Dawley rat pups. Three weeks the later, the cells had proliferated, reorganized the retinal vasculature, and filled the eyes with tumor. Eyes were harvested at that time and processed for histopathologic analysis.
For electron microscopy (EM), animals were anesthetized with avertin until a loss of deep tendon reflexes was detected. Transcardial perfusion was performed with carboxygenated Ames medium supplemented with 40 mM glucose to clear the vasculature. Animals were then perfused with Sorenson's phosphate buffer (pH 7.2) with 2% EM-grade paraformaldehyde and 1% EM-grade glutaraldehyde. Eyes were then harvested. A slit was made in the cornea to aid in diffusion, and the tissue was placed in 3% glutaraldehyde in Sorenson's phosphate buffer overnight. Tissue was washed with 0.2 M cacodylate buffer in 5% sucrose, postfixed in 1% OsO4, embedded, sectioned, and viewed by transmission EM (TEM).
Cell cultures were split into 300,000 cells/ml. Then 100 μl of the cell suspension was added in triplicate to a 96-well, clear-bottomed white plate (Costar 3610; Corning, Lowell, MA). For each concentration point, 1 μl of 100× racemic nutlin-3 (N6287; Sigma, St. Louis, MO) in dimethyl sulfoxide was added 2 h later. Cells were incubated at 37°C for 48 h. The plate was set out for 20 min to reach room temperature, and then 100 μl Cell Titer Glo (G7571; Promega, Madison, WI) was added. The plate was immediately mixed for 2 min and then was placed in a dark drawer for 15 min. The plate was then read on an Envision 2103 multilabel plate reader (Perkin Elmer, Waltham, MA). Data were processed with GraphPad Prism software (v4.03; GraphPad Software, La Jolla, CA).
Dual-color fluorescence in situ hybridization (FISH) was performed as previously described (16) except that whole-eye sections were used instead of tissue microarrays. Sections showing sufficient hybridization efficiency (majority of nuclei with signals) were considered informative and were scored. Cutoffs for abnormalities were based on counts from nonneoplastic control specimens (normal brain from autopsy cases) for each probe. Images were captured using a high-resolution black-and-white Cohu charge-coupled device camera and a Cytovision basic workstation (Applied Imaging, Santa Clara, CA). A Z-stack motor allowed sequential DAPI (4′,6′-diamidino-2-phenylindole) (1 level), FITC (16 levels), and rhodamine (16 levels) filter settings to be captured, and the resulting images were reconstructed with blue, green, and red pseudocolors.
To isolate mouse retinoblastoma cell lines, we cultured a primary retinoblastoma from an 8-month-old female Chx10-Cre; RbLox/+; p107−/−; p53Lox/Lox mouse (Fig. (Fig.1A).1A). The tumor was dissected from the surrounding sclera, and the lens was removed. There were 6.2 × 106 cells in this tumor with 87% viability, as measured with the Guava Viacount assay. Approximately 5 × 106 cells were solubilized in Trizol for genomic DNA and RNA analysis. The remaining 1.2 × 106 cells were dispersed and maintained in RPMI medium at a density of 1 × 105 cells/ml. Two weeks later, the maximum doubling time of the cells was found to be 25 ± 2 h (Fig. (Fig.1B).1B). As with other retinoblastoma cell lines, the mouse retinoblastoma cells were sensitive to cell density in culture. There was increased early- and late-stage apoptosis in cultures at densities greater than 9 × 105 cells/ml and those less than 1 × 104 cells/ml (Fig. (Fig.1B1B).
Next, we characterized the expression of genes and proteins normally expressed in retinal progenitor cells and differentiated retinal neurons. The mouse retinoblastomas expressed several markers of retinal progenitor cells and those of horizontal or amacrine cells (1, 14, 27, 31). For example, the primary tumor expressed syntaxin (Fig. (Fig.1C),1C), which is a marker for progenitor, amacrine, and horizontal cells (2, 8), and Snap25 (Fig. (Fig.1D),1D), which is a synaptosome-associated protein expressed by amacrine cells in the developing retina (8). Furthermore, the tumor expressed Pax6, a marker of retinal progenitor cells and amacrine cells (Fig. (Fig.1E)1E) (20). The primary mouse retinoblastoma cells established in culture also expressed these markers (Fig. 1F to H), indicating that at least some of the differentiation characteristics of the primary tumor cells were preserved in the cell lines. Syntaxin was expressed in 73% ± 4%, Snap25 in 46% ± 6%, and Pax6 in 92% ± 7% of cells in the primary culture.
One feature of our mouse primary retinoblastoma cultures was the presence of both adherent and nonadherent cells (Fig. (Fig.2A).2A). As mentioned above, a similar heterogeneity of cellular morphology was observed in the human primary retinoblastomas from which the nonadherent Weri1 and Y79 cell lines were derived (21, 26). To further characterize the adherent and nonadherent cell populations of mouse retinoblastoma cells, we separated the two fractions (Fig. 2B and C) and generated individual clones of each (Fig. 2D to F). Two representative nonadherent clones (SJmRBL-3 and SJmRBL-8) and two adherent clones (SJmRBL-10 and SJmRBL-12) were used for subsequent characterization.
Cell number was scored over a period to 150 h for each of the four cell lines to obtain growth curves (Fig. 2G to J). Our analysis of early and late apoptosis suggested that nonadherent clones are more sensitive to cell density than are adherent clones, as indicated by the higher percentage of apoptosis in the nonadherent clones (Fig. 2G to H) at higher cell densities. We also analyzed protein expression patterns by immunofluorescence in each of the clones. For example, we detected a difference in the expression patterns of the progenitor protein Pax6 among the clones. The nonadherent clones SJmRBL-3 and SJmRBL-8 expressed Pax6, but the adherent SJmRBL-10 and SJmRBL-12 clones did not (Fig. 2L and N). Furthermore, like Weri1 cells, the SJmRBL-3 and SJmRBL-8 cells expressed Snap25, but the adherent clones did not (Fig. 2M and O). CD44, a marker for fibroblasts (6), was expressed in adherent SJmRBL-10 and SJmRBL-12 clones but not in the nonadherent SJmRBL-3 and SJmRBL-8 clones (Fig. 2K and P). Syntaxin-1 was detected in the nonadherent clones and the human retinoblastoma cell lines (Fig. (Fig.2Q).2Q). Together, these data suggest that the markers expressed by the nonadherent clones are most similar to those expressed by the human retinoblastoma cell lines Weri1 and Y79.
The retinoblastoma lines were derived from Chx10-Cre; RbLox/+; p107−/−; p53Lox/Lox mice. Due to the mosaic expression of the Cre transgene from the Chx10 promoter, it is possible that the retinoblastoma cell lines are wild type, heterozygous, or null at the p53 locus. PCR analysis indicated that all of the cell lines were p53 null (see Fig. S1 in the supplemental material). Similarly, the RbLox allele was also recombined in all four cell lines (see Fig. S1 in the supplemental material). Next, we tested the functional p53 response of the cell lines by treating them with a small-molecule inhibitor of MDM2, nutlin-3, which can lead to p53-mediated cell death. We previously showed that SJmRBL-8 cells are insensitive to nutlin-3 (16). The other mouse retinoblastoma cell lines (SJmRBL-3, SJmRBL10, and SJmRBL12) were also insensitive to nutlin-3, but a human retinoblastoma cell line with an intact p53 gene (Weri1) was sensitive to this MDM2 inhibitor (see Fig. S1 in the supplemental material).
To begin to determine if the adherent clones (SJmRBL-10 and SJmRBL-12) were tumor-associated fibroblasts or some other retinoblastoma cell population, we performed Affymetrix gene expression array analysis on three independent samples for SJmRBL-10, SJmRBL-8, and 3T3 cells (see Fig. S2 in the supplemental material). The principal component analysis and hierarchical clustering clearly showed that SJmRBL-10 is more similar to 3T3 cells than to SJmRBL-8 cells (see Fig. S2A and B in the supplemental material). The correlation table for all three samples (45,101 probe sets each) revealed that SJmRBL-10 is a 94% match to 3T3 cells while SJmRBL-8 is only an 84% match (see Fig. S2C in the supplemental material). A detailed comparison of the genes that are statistically different between SJmRBL-8, SJmRBL-10, and 3T3 cells is available upon request.
To rule out the possibility that SJmRBL-10 and SJmRBL-12 cells were immortalized Müller glial cells, we purified Müller glia from a transgenic mouse line that expresses a recombinant fluorescent protein (Kusabira) in Müller glia in the retina (see Fig. S2D in the supplemental material). The RNA prepared from purified Müller glia was then used for real-time RT-PCR analysis and compared to that from SJmRBL-10, SJmRBL-8, and 3T3 cells. Two Müller genes (encoding Cralbp and glutamine synthetase) were enriched in the purified normal Müller glia but were not expressed in SJmRBL-10 cells (see Fig. S2E and F in the supplemental material). In contrast, cyclin E was expressed in each of the cell lines but not the purified Müller glia, which are quiescent. Taken together, these data suggest that the adherent retinoblastoma cell lines are more similar to an immortalized mouse fibroblast cell line (3T3) than to a normal retinal cell type or a mouse retinoblastoma.
The lack of contact inhibition and anchorage-independent growth is an indicator of the transformed malignant phenotype of cancer cells. The ability of cells to grow in soft agar is one method to evaluate these characteristics in retinoblastoma cells (9). The four cell lines were plated in soft agar (0.3%) and grown for 14 days. NIH 3T3 cells infected with a GFP-expressing virus (MSCV-GFP) were plated alongside them as a negative control (Fig. 3A to C), and NIH 3T3 cells infected with a Ras-expressing retrovirus (MSCV Ras-GFP) were used as a positive control (Fig. 3D to F). Seven days after being plated in soft agar, SJmRBL-8 cells formed colonies with or without Ras expression (Fig. 3G to J). After 14 days in culture, the colonies were stained with MTT and scored in triplicate (Fig. (Fig.3K).3K). These results suggest that SJmRBL-8 and SJmRBL-3 (data not shown) cells express a transformed malignant phenotype, as indicated by their loss of contact inhibition and anchorage-independent growth in soft agar. SJmRBL-10 and SJmRBL-12 cell lines showed no evidence of colony formation (data not shown).
To generate an orthotopic xenograft model, we transplanted 1,000 SJmRBL-8 cells into the vitreous of newborn rats (Fig. (Fig.4A).4A). The methods for generating this model are described elsewhere (17). After 2 weeks, dissection of the engrafted cells confirmed that they had proliferated, filled the vitreous chamber, and reorganized the retinal vasculature. Blood vessels within the tumor were visible upon microscopic analysis, and the viable tumor was surrounded by a larger, necrotic tumor (Fig. 4B to D). Hematoxylin-eosin staining also showed a viable tumor mass and blood vessels within that mass (Fig. 4E to G). SJmRBL-3 cells also proliferated and reorganized the retinal vasculature, similar to what was seen with the SJmRBL-8 cells (data not shown).
TEM analysis of SJmRBL-3 and SJmRBL-8 orthotopic xenografts revealed that these tumors are highly invasive and can pass through the retina to establish tumors in the subretinal space (Fig. 4H to J). Importantly, the invading tumor cells lacked any of the cell adhesion junctions characteristic of rosettes in primary human retinoblastomas (14), which is consistent with their ability to grow in soft agar and invade the retinal architecture. SJmRBL-10 and SJmRBL-12 cells did not efficiently form primary tumor lesions in this orthotopic xenograft model (data not shown). We complemented these studies by generating orthotopic xenografts in adult immunocompromised SCID mice (B6.CB17-Prkdcscid/SzJ). Five adult animals received an intravitreal injection of 1 × 104 SJmRBL-3 or SJmRBL-8 cells (Table (Table1).1). All of the injected eyes showed rapid tumor growth by 4 weeks, and the animals were euthanized. Histopathologic analysis showed that the tumors had features of very aggressive, invasive retinoblastoma, including invasion of the anterior chamber, choroid, optic nerves, and subretinal space (see Fig. S3 in the supplemental material). These data are consistent with the data from Y79 cells showing extensive optic nerve invasion (Table (Table1;1; see Fig. S3 in the supplemental material) (4). As reported previously, Weri1 cells did not show optic nerve invasion (Table (Table1;1; see Fig. S3 in the supplemental material) (4). Together, these data suggest that the mouse retinoblastoma cell lines SJmRBL-3 and SJmRBL-8 are aggressive and invasive tumor cells similar to Y79 cells and represent the most severe form of retinoblastoma. TEM analysis of SJmRBL-3, SJmRBL-8, Weri1, and Y79 cells in culture confirmed that the invasive cell lines do not form extensive cell-cell junctions; however, the noninvasive Weri1 cell line formed extensive cell-cell junctions (see Fig. S4 and S5 in the supplemental material).
It has been well established in a variety of tumor models that as tumors progress, pathways involved in cell adhesion are often mutated, and as a result, cancer cells begin to invade the surrounding tissues. Our data suggest that in the process of growing the primary tumor in culture, the SJmRBL-3 and SJmRBL-8 cell lines underwent molecular and/or genetic changes in key pathways that regulate cell adhesion (Fig. (Fig.33 and and4;4; see Fig. S3, S4, and S5 in the supplemental material). To test this hypothesis, we analyzed the changes in gene expression and aberrations in chromosomes (Table (Table2)2) that occurred in these two cell lines compared with that seen in the primary tumor and the original pooled samples. Gene expression microarrays were used to analyze the changes in gene expression, and bacterial artificial chromosome CGH and chromosomal SKY analyses were used to monitor the genetic lesions in the tumor cell lines compared to the primary tumor (Table (Table22 and data not shown). Although the data showed important and interesting changes in various cellular pathways, the most frequent changes were in expression of genes involved in cell-cell adhesion and cell-matrix adhesion. Of particular interest were changes in the expression of the cadherins (N-cadherin and cadherin-11). These proteins were of particular interest because they have been implicated in retinoblastoma tumor progression and vitreal seeding (19, 22).
To further explore the changes in cadherin expression, we performed immunoblotting for N-cadherin and cadherin-11 in all four mouse retinoblastoma cell lines and compared those data to the patterns of cadherin expression in human retinoblastoma cell lines (Fig. (Fig.5A).5A). Interestingly, the SJmRBL-3 and SJmRBL-8 cell lines resembled the invasive Y79 cell line in that they had lost N-cadherin and cadherin-11 expression. Our analysis of the expression of cadherin-11 and N-cadherin mRNA by real-time RT-PCR showed a similar pattern (Fig. 5B and C).
To test the functional significance of cadherin-mediated cell-cell adhesion, we performed trypsin digestion in the presence and absence of calcium. Calcium stabilizes cadherin-cadherin interactions, which makes them resistant to trypsin digestion. Thus, cells that rely on cadherin for their cell adhesion remain aggregated during trypsin digestion in the presence of calcium but dissociate during trypsin digestion in its absence. Consistent with their cadherin-11 and N-cadherin expression profile, the Weri1 cells showed a marked dependence on cadherin-mediated cell adhesion (Fig. 5D to F and H), and in TEM images, this was dependent on the cell-cell junctions at the edges of the cells (see Fig. S6 in the supplemental material). However, Y79, SJmRBL-3, and SJmRBL-8 cells were just as likely to be dissociated by trypsin treatment whether calcium was present or absent (Fig. 5G and I to K).
To begin to test the role of cadherin-mediated cell adhesion in human retinoblastoma invasion, we selected 16 cases with extensive optic nerve, retinal, choroidal, or scleral invasion (Table (Table3).3). We performed FISH on whole-eye sections from these tumors for 10 genes believed to be involved in cell adhesion or signaling pathways typically associated with changes in invasive properties. The gene copy number in the invading cells was directly compared to the gene copy number in the main tumor to correlate such genetic lesions with any selective advantage for the cells with regard to invasion (see Fig. S7 in the supplemental material). The most striking observation (in 4 of 14 cases) was selective loss of Cdh11 in retinoblastoma cells invading the optic nerve and choroid (Table (Table3).3). There were also one case of selective Ncad gene loss and one case of selective Ecad loss (Table (Table33).
To begin to test the functional role of cadherin-mediated cell adhesion in retinoblastoma tumor invasion in vivo, we ectopically expressed cadherin-11 and N-cadherin in Y79, SJmRBL-3, and SJmRBL-8 cells. If cadherin-mediated cell adhesion is sufficient to form the types of junctions seen in Weri1 cells that express abundant cadherin-11 and N-cadherin, then we predict that we will be able to convert these aggressive and invasive retinoblastoma cell lines (Y79, SJmRBL-3, and SJmRBL-8) into more benign cells similar to Weri1. As a first step, we cloned the full-length human Cdh11 and Ncad cDNAs individually into a mammalian expression vector (pcDNA 3.1) and validated full-length protein expression using a heterologous cell culture system (COS and 293T cells) (Fig. 6A and B). The N-cadherin protein is fused to the Myc epitope tag as described previously (25) (Fig. (Fig.6A),6A), and the Cdh11 protein is fused to the FLAG epitope tag (Fig. (Fig.6B6B).
To test the sufficiency of these two proteins for retinoblastoma cell adhesion, we ectopically expressed them together with a nuclear GFP (nGFP) reporter gene in Y79, SJmRBL-8, and Weri1 cells. Cells expressing the GFP reporter gene were purified by FACS and cultured for an additional 48 h. As expected, ectopic expression of Cdh11 and Ncad had no effect on cluster size in Weri1 cells (Fig. (Fig.6C).6C). However, ectopic expression of these two proteins dramatically increased cluster size in both Y79 and SJmRBL-8 cells (Fig. 6D and E). More importantly, the cellular morphology of the Y79 and SJmRBL-8 cells was strikingly altered to resemble that of Weri1 cells, which endogenously express Cdh11 and Ncad (Fig. 6F to H and data not shown).
TEM analysis revealed that the Y79 cells expressing Cdh11 and Ncad were virtually indistinguishable from Weri1 cells with respect to their cell-cell junctions and their extensive interdigitations along the edge of the cells (compare Fig. 7A to C with Fig. 7D and E). More importantly, the increase in the ability of Y79 and SJmRBL-8 cells to form large clusters of cells following transfection of Cdh11 and Ncad could be selectively reversed by incubating with trypsin in the absence of Ca2+ (Fig. (Fig.7F),7F), which is a hallmark of cadherin-mediated cell adhesion.
To establish the functional significance of cadherin-mediated cell adhesion for retinoblastoma invasion of the optic nerve during metastasis, we injected cells (1 × 106) into the vitreous of adult SCID mice (B6.CB17-Prkdcscid/SzJ) and monitored tumor growth with a digital retinal camera. We tested Y79 and SJmRBL-8 cells transfected with Cdh11/N-cad, Y79 and SJmRBL-8 cells transfected with a control vector, and Weri1 cells as a negative control (Fig. (Fig.7G7G and data not shown). One to two weeks later, when the tumor cells had filled the vitreous, we harvested the eyes and optic nerves to determine if ectopic expression of Cdh11/Ncad could prevent Y79 and SJmRBl-8 cells from invading the optic nerve. We were able to detect extensive optic nerve invasion (up to 3 mm) in the Y79 and SJmRBL-8 xenografts only with control vector (Fig. 7H to K). In the three eyes that did show some evidence of optic nerve invasion by cells expressing Cdh11 and Ncad, invasion was much less extensive than that by the Y79 cells (~200 to 300 μm beyond the optic nerve head). In a separate series of experiments, we tested whether both Cdh11 and Ncad were required by transfecting them individually and together in Y79 and SJmRBL-8 cells. We found that either cadherin was sufficient to cause the cell adhesion phenotype described above.
In this report we describe the isolation and characterization of the first mouse retinoblastoma cell lines derived from a primary tumor removed from a Chx10-Cre; RbLox/+; p107−/−; p53Lox/− mouse. We isolated two nonadherent cell lines with properties similar to those of the previously described human retinoblastoma cell lines Weri1 and Y79 and two adherent cell lines that resemble the adherent cells found in all primary human retinoblastoma cultures described previously. The nonadherent mouse retinoblastoma cell lines (SJmRBL-3 and SJmRBL-8) express markers found in the primary mouse tumor and have many of the cellular features of the Y79 cells. More importantly, we show that the process of selecting clones in culture from primary retinoblastoma tumors enriches the population of cells that have undergone distinct changes in their cell adhesion properties that correlate with tumor cell invasion. Specifically, these cells express low levels of N-cadherin and cadherin-11; thus, they show little cell adhesion, a property that is important for vitreal seeding and ocular invasion of retinoblastoma cells before metastasis. It is important to note that these changes are similar to those seen in human retinoblastomas (19, 22).
Tumorigenesis is a multistep process that involves sequential genetic lesions in pathways that control proliferation, cell death, cell adhesion, growth factor dependence, and telomere maintenance (11). The initiating genetic lesion in retinoblastoma is inactivation of the Rb pathway, which leads to deregulated proliferation of retinoblasts and activation of p53-mediated cell death through the p14ARF/MDM2/MDMX/p53 pathway. To overcome p53-mediated cell death, most human retinoblastomas sustain genetic amplification of their MDMX gene, which suppresses p53-mediated cell death (16). Inactivation of the Rb and p53 pathways is an early genetic event in retinoblastoma tumorigenesis, because these signals are essential for rapid tumor cell expansion.
In contrast to Rb and p53 pathway inactivation, genetic changes that promote local invasion and retinoblastoma metastasis may occur later during tumorigenesis and produce a tumor that is heterogeneous with respect to its ability to invade ocular structures and metastasize. Consistent with this model, very large tumors that show little or no sign of local ocular invasion develop in some children. Moreover, invading tumor cells in humans and mice lack extensive cell-cell contacts in electron micrographs (14); thus, invading tumor cells are distinct from the cells in the main tumor mass. This finding confirms that retinoblastoma tumors are heterogeneous with respect to cell adhesion (14). Previous studies have shown that the Weri1 and Y79 retinoblastoma cell lines exhibit differences in N-cadherin expression, that cell adhesion is sensitive to calcium in the medium, and that an N-cadherin antibody can block aggregation and collagen adhesion in culture (28, 30). Indeed, our data regarding the genetic changes in the retinoblastoma cells invading the optic nerve compared to the body of the tumor suggest that the lesions in the cadherin loci are heterogenous in the primary tumor but have a selective advantage for invading the optic nerve. We also demonstrate for the first time the functional significance of changes in cadherin-mediated cell adhesion for cell invasion in the optic nerve in vivo providing a robust model system to attempt to disrupt this process in human retinoblastomas.
It is also interesting to contrast our findings related to the role of cadherins in retinoblastoma progression to those obtained with other tumor types. Reminiscent of cadherin switching during development, E-cadherin downregulation is correlated with de novo expression of N-cadherin or cadherin-11 in many carcinomas (24, 29). Downregulation of E-cadherin and sustained N-cadherin expression is also observed in a Rip1Tag2 mouse model of pancreatic β-cell tumor progression from adenoma to invasive carcinoma (23). Hazan and colleagues have suggested that in invasive breast cancer, the association between N-cadherin-expressing tumor cells and stromal cells contributes to invasion and metastasis (12). These studies and many others suggest that the switch from E-cadherin to N-cadherin is one mechanism to promote tumor invasion and progression. In contrast, sustained N-cadherin and/or cadherin-11 expression is not permissive for optic nerve invasion and progression for retinoblastoma. This may reflect differences in the microenvironment of the eye and/or optic nerve relative to other tissues where cadherin expression has been studied during tumorigenesis.
Interestingly, many of the studies on the isolation and characterization of retinoblastoma cell lines have described changes in cell adhesion over the course of culturing the samples (10, 21, 26). For example, it has been reported that Y79 retinoblastoma cells initially grow in clumps, but then they start to lose cell adhesion and grow as single cells (26). Moreover, these cells exhibit invasive properties when transplanted into the mouse eye, but cells with more adhesive properties do not (4). This finding suggests that changes in cell adhesion that occur during the process of culturing retinoblastomas to produce cell lines are similar to those associated with tumor progression, invasion, and metastasis. It is not known if the adherent cells (SJmRBL-10 and -12) in the primary cultures play any role in retinoblastoma cell invasion in vivo. In addition, it is not known if they were derived from a common tumor precursor that gave rise to the nonadherent and adherent cells or were distinct. That is, it is possible that these cells started as tumor-associated fibroblasts and that in the culturing process, they underwent genetic loss spontaneously at the Rb locus and became transformed during the culture period. However, our data show that these cells are not required for optic nerve invasion in vivo nor are they required to observe the dependence of cadherin-mediated cell adhesion on this process. We feel that these cell lines may be used to help develop additional retinoblastoma cell lines by using them in coculture experiments with primary tumors.
To test this hypothesis directly, we produced retinoblastoma cell lines from a recently described mouse model of retinoblastoma (31). As the tumor cells grew in culture, they stopped expressing genes that are important for cell adhesion, and when transplanted into the mouse eye, the cells exhibited invasive properties. Specifically, SJmRBL-3 and SJmRBL-8 cells downregulated the expression of cadherin-11 and N-cadherin and had little, if any, cadherin-based cell-cell adhesion. This property was also observed in retinoblastoma cells during optic nerve invasion in vivo. In contrast, Weri1 cells form extensive cell-cell junctions, express cadherin-11 and N-cadherin, and are not invasive or aggressive in vivo. These studies indicate that changes in cadherin-mediated cell adhesion may be important for retinoblastoma cell invasion and that by maintaining retinoblastoma tumors in culture, we can enrich them for these highly invasive and aggressive cells for further study.
It is also important to consider the contribution of the aforementioned heterogeneity in the starting tumor cell population. That is, there may have been a population of cells in the primary cultures of the mouse tumor that was more similar to Weri1 cells, with cadherin-mediated cell-cell junctions, and another population of cells more similar to Y79 cells, which lack these junctions. In the course of growing our cultures, it would not be surprising if the less-adhesive cells lacking cadherin expression outgrew those with cadherin-mediated cell-cell junctions. Even though we carefully monitored growth of these cultures from the primary tumor to individual clones and characterized their molecular, cellular, and genetic profiles during this process, it is impossible to rule out such selection in our experiments. In either case, we have shown that very aggressive mouse retinoblastoma cells can be isolated and cloned in culture and used to recapitulate optic nerve invasion in vivo.
Together, our data suggest that either the mouse retinoblastoma cell lines downregulated their expression of cadherins over the course of selection in culture or there was a heterogenous cell population in the primary tumor, and cells lacking cadherins had a selective growth advantage under these culture conditions. This may begin to explain the ability of SJmRBL-3 and SJmRBL-8 cells to grow and form colonies in soft agar and to invade the optic nerve. It is important to note that while the expression of N-cadherin and cadherin-11 was lost in both SJmRBL-3 and SJmRBL-8 cells, the Cadherin-11 locus was deleted in only one of the cell lines; thus, at least two alternative mechanisms for downregulating the cell-adhesion pathway are present in these tumors. It is important to emphasize that FISH analysis of cadherin genes described here is likely to underrepresent the proportion of cadherin gene lesions in retinoblastoma. Microdeletions, point mutations, and other genetic or epigenetic lesions will not be detected by FISH, but they may nonetheless lower cadherin expression and promote optic nerve invasion in retinoblastomas. Our preliminary deep sequencing analysis demonstrates that it may be feasible to sequence broadly across the retinoblastoma genome and ultimately identify these additional lesions.
We thank David Finkelstein for assistance with bioinformatics, Michael Wang for assistance with the analyses of gene expression microarrays and BAC-CGH arrays, and Jill Lahti for assistance with Sky analysis and karyotyping. We also thank Angie McArthur for editing the manuscript. We thank Fara Sudlow and Jackie Craft for expertise with TEM tissue processing and analysis. We thank C. H. Siu for the gift of the N-cadherin cDNA expression vector.
This work was supported by grants (to M.A.D.) from the National Institutes of Health, Cancer Center Support from the National Cancer Institute, the American Cancer Society, Research to Prevent Blindness, the Pearle Vision Foundation, the International Retinal Research Foundation, the Pew Charitable Trust, and the American Lebanese Syrian Associated Charities (ALSAC). M.A.D. is an HHMI early career scientist.
Published ahead of print on 28 September 2009.
†Supplemental material for this article may be found at http://mcb.asm.org/.