Role of B lymphoma Mo-MLV insertion region 1 in the oncogenic behavior of retinoblastomas

Retinoblastoma cases

The study protocol of human retinoblastoma archive specimens was approved by the Ethics Committee for Human Research, Beijing Tongren Hospital, Capital Medical University, China and adhered to the tenets of Declaration of Helsinki. Paraffin-embedded retinoblastoma eyeballs from 34 patients were archived in Beijing Tongren Hospital from 2007 to 2008. These patients had no associated medical and family history. All specimens were serially sectioned along the pupillary–optic nerve head axis and the sections were used for immunohistochemistry.

Immunohistochemistry of BMI-1 on archived sections

Immunoreactive BMI-1 was detected using the labeled horseradish peroxidase method. On paraffin sections, antigen was retrieved by trypsin and endogenous peroxidase inactivated by 0.3% hydrogen peroxide in methanol. After blocking with 5% normal goat serum and 0.1% Triton X-100, the section was incubated with or without 0.5 to 1 μg/ml antihuman BMI-1 monoclonal antibody (Clone F6, Millipore, Billerica, MA). Control sections were incubated with buffer without primary antibody. Following PBS rinses, sections were treated with goat anit-mouse immunoglobulin horseradish peroxidase conjugate (Jackson ImmunoRes Lab, West Grove, PA). Signals were revealed by 3,3'-diaminobenzidine (DAB) reduction and examined under light microscopy (DMRB, Leica, Wetzlar, Germany) equipped with Spot RT color system (Diagnostic Instruments Inc., Sterling Heights, MI). Stained sections were examined by two masked pathologists (RRJ, LTL). In 10 random images taken at 50× magnification, the percentage of BMI-1 positive cells in tumor or retinal layers was graded as (0) no staining, (1) <5%, (2) 5%–25%, (3) 25%–50%, or (4) >50%.

Expression constructs of BMI-1

BMI-1 expression construct was prepared by cloning a 990 bp EcoR1/Xho1 fragment encompassing full-length 981 bp open reading frame of wild-type human BMI-1 to EcoR1/Xho1 site of a mammalian expression vector pCMV-HA (Clontech, Mountain View, CA) to create pHA-BMI-1. Alternatively, for specific knockdown, synthesized 64 bp oligonucleotide containing human BMI-1 small interfering RNA (siRNA) sequence (5′-ATG AAG AGA AGA AGG GAT T-3′, position 269–287 bp from the start codon) was cloned into the HindIII/BglII site in the pSuper vector (Oligoengine, Seattle, WA) to generate pSuper-BMI-1. All constructs were verified by direct sequencing.

Cell transfection

Y79 cells (American Tissue Cell Collection, Manassas, VA) were maintained in RPMI-1640 (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS, Invitrogen), 100 U/ml penicillin G, and 100 μg/ml streptomycin sulfate at 37 °C under humidified conditions in 5% CO₂ balanced with air. The BMI-1 construct was transfected to cells at 5×10⁵ cells/ml by Lipofectamine 2000 (Invitrogen) at a ratio of 3 µl reagent per µg DNA in Opti-MEM® Reduced Serum Medium, GlutaMAX™ (Invitrogen). One day after transfection, the cells were maintained in 80 µg/ml Geneticin-418 (Invitrogen) for 10 days. Drug-resistant cells were pooled for protein and RNA analyses.

Cell growth, viability, and apoptosis assays

Transfected cells at a density of 5×10⁵ cells/µl were cultured in a six-well plate. Every 24 h, 200 µl of cell suspension was collected for for tetrazolium dye (MTT) cell viability/proliferation assay. Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay using ApopTaqIn Situ Apoptosis Detection Kit (Oncor, Gaithersburg, MD) was performed on paraformaldehyde-fixed cytospinned cells. The TUNEL-positive and 4',6-diamidino-2-phenylindole (DAPI)-labeled cells were counted in 10 random images captured under fluorescence microscopy with a 20x objective. The apoptosis rate was determined as the percentage of TUNEL-positive cells. All experiments were carried out in triplicate.

Multicellular sphere assay

Single transfected cells at 50 cells/ml were passed through 40 µm nylon mesh and incubated in a culture dish (100 mm diameter) in serum-free RPMI-1640 medium supplemented with 10 ng/ml basic fibroblast growth factor (Invitrogen). After 7 days, the culture was examined for sphere formation and the number and sizes of spheres were recorded. Cell proliferation was assayed by BrdU incorporation for 2 h, followed by immunofluorescence using anti-BrdU antibody (Abcam, Cambridge, UK), followed by Alexa 488-conjugated immuneglopbulin G (IgG) secondary antibody (Jackson ImmunoRes Lab, West Grove, PA).

Propidium iodide cell cycle analysis

Transfected cells were harvested and fixed with 70% ice-cold ethanol with vigorous shaking to prevent cell clustering. The cells were incubated with 50 μg/ml propidium iodide (Invitrogen) in the presence of 0.1 mg/ml RNase A (Sigma) and 0.05% Triton X-100. After washes, the cells were resuspended in PBS for flow analysis using a FACS Calibur Flow Cytometer (BD).

Western blotting

The cells were lysed at 2.5×10⁶ cells/µl in radioimmunoprecipitation buffer containing 50 mM Tris-HCl (Sigma, St Louis, MI), 150 mM sodium chloride, 1% Nonidet P40, 0.25% sodium deoxycholate, protease inhibitor cocktail (Roche, Basel, Switzerland), and 1 mM phenylmethyl sulfonylfluoride. The clear supernatant was denatured in 2% sodium dodecyl sulfate (BioRad, Hercules, CA) and 50 mM DL-dithiothreitol. Proteins (equivalent to 2×10⁵ cells) were resolved by 10% or 13% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and analyzed by enhanced chemiluminescence using antibodies against HA (Millipore), BMI-1 (Millipore), p14ARF (Santa Cruz Biotechnol., Santa Cruz, CA), p16INK4a (Santa Cruz), cyclin D1 (Santa Cruz), cyclin D2 (Santa Cruz), proliferating cell nuclear antigen (PCNA, Sigma), caspase-3 (Millipore), and β-actin (Sigma).

Expression of ribonucleic acid

Total RNA was extracted with RNeasy kit (Qiagen, Valencia, CA), quantified and reverse transcribed with random primer (Roche) and SuperScript™ III reverse transcriptase (Invitrogen). Semiquantitative PCR was performed with specific primers for CHX10, Rx, NESTIN, neurofilament-M and glyceraldehyde phosphate dehydrogenase (Table 1). PCR products were resolved by 2% agarose gel electrophoresis and analysed by Quantity One Image Analysis. Gene expression was normalized with glyceraldehyde phosphate dehydrogenase. Three independent experiments were performed.

Statistical analyses

The analysis was performed using a commercially available statistical software package (SPSS for Windows, version 17.0, SPSS, Chicago, IL). The data were presented as mean and standard deviation, as well as in median and ranges. For the comparison of study groups, the Mann–Whitney U test was applied for unpaired samples. If there was one variable possibly associated with the dependent parameter, a Spearman correlation analysis was performed. A value of p<0.05 was considered as statistical significance.

Demography and retinoblastoma pathology

A total of 34 retinoblastoma specimens were recruited in this study. They were collected at Beijing Tongren Hospital from April 2007 to April 2008 (Table 2). The patients had no associated medical or family history. Twenty-one were males and 15 were females, and the average age was 2.27±1.13 years. There were 31 unilateral retinoblastomas (15 left and 16 right eyes) and 5 bilateral retinoblastomas. Leukocoria was commonly presented (31 eyes), while others exhibited pink eye (2 eyes), esotropia (1 eye), reduced vision with esotropia (1 eye), and leukocoria with esotropia (1 eye).

BMI-1 expression and retinoblastoma differentiation

Twenty-four tumors without Flexner-Wintersteiner rosettes and showing small-sized cells with a large nucleus-to-cytoplasm ratio were classified as undifferentiated type (column A in Figure 1). Ten tumors were characterized as differentiated with typical Flexner-Wintersteiner rosettes with cells surrounding the central lumen (column A in Figure 2). Homer-Wright rosettes bearing eosinophilic neurofibrillary cores were occasionally observed. Intratumor necrosis was infrequently observed.

Nuclear BMI-1 staining was present in the majority of retinoblastoma specimens (33/34 tumors, 97.1%). The staining intensity varied with the differentiation status. Distinct BMI-1 expression was observed in >50% tumor cells in 13 undifferentiated retinoblastoma cases (13/24, 54.2%; representative pictures in column B of Figure 1, magnified images in column C), but not in any differentiated cases (0/10, 0%; Table 3). Ten undifferentiated cases (10/24, 41.7%) had milder BMI-1 staining (<50% tumor cells). Only one case was BMI-1 negative. On the other hand, BMI-1 was not clearly observed in the differentiated tumors. All cases exhibited BMI-1 immunoreactivity in <25% tumor cells (representative pictures in column B of Figure 2, magnified images in column C). Hence, the expression of BMI-1 was shown to be significantly associated with undifferentiated retinoblastomas (p=0.05, Mann–Whitney U test).

We studied BMI-1 expression in 27 retinal tissues adjacent to retinoblastoma (Table 3). Notably, it was not widely detected in all retinas. Most retinal tissue next to undifferentiated retinoblastomas had positive BMI-1 staining in <25% retinal cells (14/17, 82.3%; representative pictures in upper panel of Figure 3). A similar observation was made for differentiated cases (9/10, 90%; representative pictures in lower panel of Figure 3). The remaining was BMI-1 negative. Hence, there was no significant correlation between BMI-1 expression in retina in terms of the differentiation status of retinoblastomas (p=0.86, Mann–Whitney U test). Furthermore, altered retinal layers, including an indistinct inner nuclear layer, was infrequently observed in both differentiated and undifferentiated tumors (marked by red brackets in Figure 3). There was no association between BMI-1 expression in the outer and inner nuclear layer with retinoblastoma differentiation status (p=0.49 and p=0.24, respectively; Table 3).

BMI-1 expression and retinoblastoma invasion to posterior tissues

In our 34 studied cases, 9 retinoblastomas remained in situ and no invasion to adjacent retinal tissue was observed (9/34, 26.5%; pT1 grade in Pathologic Classification of TNM [Tumor, Node, Metastasis] Classification for Retinoblastoma, Table 3). For the remaining 25 cases showing tumor cell invasion, involvement of the optic nerve was only detected in 14 retinoblastomas (38.9%; pT2a grade), 3 were choroid only (8.3%), and 8 were invaded to both optic nerve and choroid (22.2%; pT2b grade). The percentage of BMI-1 positive cells was significantly greater in retinoblastomas with invasion than those without (p=0.0023, Mann–Whitney U test; Figure 1 and Figure 2; Table 3). Among these malignant cases, more BMI-1 expressing cells were found in tumors invading both the optic nerve and choroid (p=0.0005), whereas fewer were found in tumors with invasion to either location. For retinoblastomas showing optic nerve invasion (n=22), 13 were in the optic nerve head (pT2 grade) and 9 were extended to the posterior surface of the lamina cribrosa (pT3 grade) and percentage of BMI-1 positive cells was similar in both categories (Table 3).

Effect of BMI-1 on retinoblastoma cell growth and proliferation

Human retinoblastoma Y79 cells were transfected with pCMV-HA/BMI-1 for BMI-1 overexpression or by pSuper-BMI-1-si for specific BMI-1 suppression. Stable transfectants were obtained after G418 selection for 10 days (Figure 4A). Cells transfected with empty vectors (pCMV-HA or pSuper) and nontransfected cells served as controls (Figure 4B).

Increased BMI-1 transcript and protein levels in Y79 cells with pCMV-HA/BMI-1 transfection were confirmed by semiquantitative PCR and western blotting of the HA epitope (Figure 4C,F). Nontransfected cells exhibited negligible HA staining. BMI-1-HA transfected Y79 cells showed increased cell viability and a shorter doubling time (17.9 h) when compared to nontransfected or mock-transfected control cells (both were 22.3 h; Figure 4D). They formed more multicellular spheres in 7 days than pSuper-BMI-1-si and nontransfected cells. The sphere size was generally greater and there were more spheres with a diameter greater than 100 µm in BMI-1-HA-transfected Y79 cells (63%) than nontransfected cells (40%). The BMI-1-transfected cells exhibited higher expression of PCNA (Figure 4F) and BrdU incorporation. The apoptosis rate, determined by the percentage of TUNEL-positive cells, was significantly decreased with BMI-1-HA expression (5.2±1%) when compared to control (8.2±2.7%; p=0.023, paired Student t test; Figure 5A). This observation was substantiated in the reduced expression of pro- and active caspase-3 in western blotting (Figure 5B). Using propidium iodide cell cycle analysis, there was no significant difference of S-phase population in BMI-1-HA-expressing Y79 cells (13.1%) compared to mock-transfected control cells (11.3%; Figure 5C).

In contrast, siRNA-mediated inhibition of BMI-1 impaired Y79 cell growth (Figure 4E). The cells showed longer cell doubling time (29 h) when compared to nontransfected or mock-transfected control cells (both were about 23.1 h). Further, they exhibited less multicellular sphere formation (less than one-third that of nontransfected cells; Figure 6A). A set of spheres (43%) were smaller in size (<40 µm in diameter). BrdU incorporation was also reduced (Figure 6C). There were more TUNEL-positive cells, and the apoptosis rate (14.5±5%) was significantly higher than in mock-transfected control cells (7±2.3%; p=0.0017, paired Student t test; Figure 5A). This was accompanied by an increased expression of active caspase-3 (Figure 5B). Propidium iodide cell cycle analysis showed that BMI-1 repression increased apoptotic cells (21.8%) when compared with mock-transfection (7.2%). There was no effect on S-phase alteration (13.7% in si-BMI-1 cells versus 11.3% in control; Figure 5C).

Western blotting showed that BMI-1 induced cyclin D1 (~37 kDa) and D2 (~34 kDa) expression (Figure 5D) and reduced cyclin-dependent kinase (CDK) inhibitor p14ARF and p16INK4 expression (Figure 4F). In contrast, its suppression by specific RNA interference caused a reduction in cyclin D1 and cyclin D2 (Figure 5D), accompanied by induction of p14ARF and p16INK4 (Figure 4F). BMI-1 also affected the expression of retinal development markers. The expression of CHX10 and Rx was elevated with its ectopic expression in Y79 cells and reduced after the specific knockdown (Figure 5E). However, it did not affect the expression of Nestin and neurofilament M.