1 INTRODUCTION

Retinoblastoma cancer develops in the retina tissue and is the most common ocular cancer in younger children.1 It is diagnosed in approximately 300 cases in the United States, and 5000–10,000 cases worldwide per year.2 Disease mortality varies significantly according to the countries' development status; 40%–70% in Asia and Africa compared with 3%–5% in Western developed countries.1

Currently, treatments for retinoblastoma include surgery, focal therapy, radiotherapy, and chemotherapy. Multiple factors contribute to the selection of treatment strategies including disease stage, tumor laterality, RB1 mutation status, and vision status.1, 2 Systematic chemotherapy for managing intraocular retinoblastoma commonly consists of vincristine, carboplatin, and etoposide (VCE) combination.1, 3 However, this combination-based modality exhibits limited applications owing to its late and side effects.4-6 Moreover, in patients with advanced bilateral retinoblastoma, blindness, and enucleation events occur at a rate of approximately 50%.7 Therefore, as the major patient populations with retinoblastoma are young, therapies that preserve vision and improve quality of life are urgently needed.

Recent advances in the treatment of retinoblastoma have emerged including improvement of ocular local drug delivery, development of animal models in preclinical efficacy testing, and discovery of novel therapeutic targets.8 Several druggable molecular targets have been investigated to understand the genetic etiology underlying retinoblastoma resulting in the development of several small molecular inhibitors such as targeting Mdm2-p53 (Nutlin-3), spleen tyrosine kinase (SYK), and histone deacetylase (HDAC).8-11

Loss-of-function (LOF) mutations in RB1 are common in many cancers including small cell lung cancer (SCLC), triple-negative breast cancer (TNBC), and retinoblastoma. In both heritable and nonheritable retinoblastomas, homozygous RB1 gene loss leads to chromosome instability and initiates tumor growth.1 Until now, limited knowledge has made the therapeutic development of LOF mutation-driven tumors difficult. Nevertheless, synthetic lethality between RB1-loss and inhibition of the Aurora kinase family (AURKA or AURKB) has been found in lung cancer and TNBC,12, 13 thus providing a potential approach for the treatment of RB1-deficient malignancies. Tumors with RB1-loss exhibited hyperdependency on Aurora kinase activity for cell survival, and rendered sensitivity to AURKA or AURKB inhibitors.

In this study, we focused on RB1-LOF in advanced retinoblastoma and validated the synthetic lethality between RB1-LOF and Aurora kinase inhibition in this subtype. We showed that Aurora kinase inhibition could lead to cell mitotic abnormalities and apoptosis and exhibit in vivo efficacy in RB1-deficient tumors. Therefore, this finding may indicate a promising druggable target to address the unmet clinical needs of retinoblastoma disease. We present the following article in accordance with the ARRIVE reporting checklist.

2 METHODS 2.1 Cell lines, cell culture, and compounds

Cell lines Y79 and WERI-Rb1 were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ, Germany). RB355, RB3823, and RB522 cell lines were kindly provided by Professor Brenda Gallie (Department of Ophthalmology and Vision Sciences, The Hospital for Sick Children, Toronto, Canada). All cell lines were maintained in RPMI-1640 medium (Invitrogen) supplemented with 10% fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin in a humidified atmosphere of 5% CO2 at 37°C. All cells were cultured within 10 passages after recovery and were routinely tested for mycoplasma. LY3295668 and cisplatin were purchased from Selleck Chemicals (Shanghai, China).

2.2 Cell viability assay

Cell viability was determined using the CellTiter-Glo (CTG) luminescent cell viability assay kit (G5750, Promega) according to the manufacturer's instructions. Briefly, the tested cells were seeded onto 96-well plates at 8000 cells/well overnight. The compound LY3295668 (10 μM in DMSO) in nine three-fold serial dilutions was administered to assay plates in triplicate. The plates were incubated in a humidified atmosphere of 5% CO2 at 37°C for 72 h. CellTiter-Glo reagent was added to induce cell lysis at room temperature for 30 min. Luminescence was measured using a plate reader (Envision, Perkin Elmer). Relative viability was determined by normalizing the luminescence units to the untreated controls. The dose-inhibition curve and IC50 values were calculated using GraphPad Prism software (version 7, San Diego, CA).

2.3 Cell transfection and CRISPR knockout RB3823 cells were transfected with the indicated plasmids using Lipofectamine 2000 transfection reagent (Thermo Fisher) according to the manufacturer's instructions. The RB1 knockout cell line was generated by genome editing using the CRISPR/Cas9 system. The pair of sgRNA targeting RB1 was designed as previously reported14 and is listed below (target sequences are underlined): 5′-CACCGAAGTGAACGACATCTCATCT-3′, 5′- AAACAGATGAGATGTCGTTCACTTC-3′.

Knockout clones were validated after genomic DNA extraction and Sanger sequencing, and confirmed clones were maintained in the culture medium. The RB1-KO strain was also transfected with plasmid pCMV-HA-hRB1-wt (kindly provided by Steven Dowdy, Addgene plasmid # 58905) using Lipofectamine 2000 reagent to generate the RB1-KO/rescue strain.

2.4 FACS-based Annexin V-FITC apoptosis and cell cycle analysis

RB3823 cells were plated in RPMI-1640 medium at the following cell densities: RB1-wt (200,000/ml), RB1-KO (100,000/ml), and RB1-KO/rescue (200,000/ml). For Annexin V staining assay, cells were treated with LY3295668 at 1 μM or vehicle for 24 h. Cells were stained with Annexin V-FITC and propidium iodide (PI) according to the manufacturer's instructions (BD Biosciences). Determination using a FACSCalibur analyzer (BD Biosciences) was then performed and the percentage of Annexin V-positive cells was calculated using FlowJo software. For cell cycle analysis, plated cells were treated with LY3295668 at 1 μM or vehicle for 48 h, washed with PBS, and fixed in ice-cold ethanol at a final concentration of 70% overnight at −20°C. Cells were stained with PI at room temperature in the dark for 20 min and analyzed with the FACSCalibur analyzer (BD Biosciences) and FlowJo software. The percentage of polyploidy was determined by gating the cells with >4 N content.

2.5 Cell colony formation

The cells including parental RB3823 cells, genetically modified RB1-KO cells, and RB1-KO/rescue transfectant cells were treated with LY3295668 at 1 μM or DMSO for 24 h. Cells were then pre-mixed with 0.3% agarose in the growth medium and seeded onto a bottom layer of 0.6% agarose in the growth medium in six-well plates. After 2-weeks of culture, the colonies were stained with 0.05% crystal violet solution and counted under a microscope.

2.6 Western blotting

Post-treatment cells were washed with ice-cold PBS and then lysed in cell lysis buffer containing protease inhibitor cocktail (Roche) and phosphatase inhibitors (Thermo Fisher) on ice. After lysing on ice for 30 min, the cells were centrifuged at 22,500 × g for 20 min to remove the cell debris and denatured by boiling. Cell lysates were quantified using a BCA protein assay kit (Thermo Fisher). The lysate proteins were separated by SDS-PAGE and subjected to immunoblot analysis using the following specific antibodies: RB1 (Cell Signaling Technology, 9309), AURKA (Cell Signaling Technology, 14475), AURKA pT288 (Cell Signaling Technology, 3079), cleaved caspase-3 (Cell Signaling Technology, 9661), cleaved PARP (Cell Signaling Technology, 5625), cyclin B1 (Cell Signaling Technology, 4138), phospho-histone H3 (ser10) (Cell Signaling Technology, 53348), and actin (Sigma-Aldrich, A5441).

2.7 Mouse xenograft studies

All animal studies were performed in strict accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) of Nanjing Medical University and were approved by the ethics committee of the Affiliated Changzhou No. 2 People's Hospital. Female Balb/c nude mice (6–8 weeks old) provided by Shanghai Lingchang Biotech were housed in groups (3–5 per cage) in a pathogen-free facility with controlled temperature and a standard 12 h/12 h light–dark cycle, and food and water were provided ad libitum. For tumor models, cells resuspended in a 1:1 mixture (v/v) of PBS and Matrigel (BD Biosciences) were inoculated subcutaneously into the right flanks of mice (5 × 106 cells/site). After tumor volumes grew to approximately 100–150 mm3, the mice with similar tumor volumes were selected and randomly assigned with seven mice into different groups and administered the vehicle, cisplatin, or LY3295668 as indicated in the main text. Tumor volume was determined using calipers every 3–4 days and tumor volumes were calculated according to the formula: V = 0.5 × length × (width)2 where length represents the largest tumor diameter and width represents the perpendicular tumor width. The mice were euthanized when the tumor volume reached 2000 mm3. Blood was collected from mice on post-treatment day 15 and 30 using the submandibular bleeding method, and white blood cell count (WBC) and hematocrits (Hct) were examined on ProCyte Dx Hematology Analyzer (IDEXX, USA).

2.8 Statistical analysis

All cell-based assays were independently repeated thrice. Animal data were analyzed using GraphPad Prism software (version 7, San Diego, CA) and results were presented as mean ± SEM. The difference was considered significant when p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****).

3 RESULTS 3.1 RB1-deficient retinoblastoma cells are sensitive to Aurora-A kinase inhibition

Predominant RB1 gene mutations in human retinoblastoma tumors have been reported. In our collection of five retinoblastoma cell lines, biallelic loss of the RB1 gene was found in three cell lines, Y79, RB355, and WERI-Rb1, and led to complete loss of RB1 protein expression (Figure 1A). The other two cell lines, RB3823 and RB522, harbored normal RB1 gene levels (Figure 1A) but with amplified MYCN.15 As previous findings suggested a dependency of RB1 mutant cells on the AURKA gene,12, 13 we initially examined the dose–response effect of a specific AURKA inhibitor (AURKAi), LY3295668, across the five cell lines. As shown in Figure 1B, cell lines with RB1-deficiency exhibited higher sensitivities to AURKAi (IC50 varied from 0.06 to 0.10 μM), while weak responses to AURKAi in RB1-wild type (wt) cell lines (IC50 > 1 μM).

AURKA inhibitor sensitized RB1-deficient retinoblastoma cells. (A) Differential RB1 protein levels were detected in a panel of retinoblastoma cell lines using western blot assay. (B) Different dose–response curves from different retinoblastoma cell lines upon treatment of a range of LY3295668 concentrations were analyzed with CTG assay. (C) Annexin V staining assay was performed upon treatment with 0 or 0.1 μM LY3295668 for 24 h in different cell lines. Data were obtained from two independent experiments. (D) Effects of LY3295668 treatment on Y79 and RB522 cell lines with different RB1 protein expression levels were analyzed in western blot assay

In the above-mentioned five cell lines, as indicated by Annexin V positive staining, the basal apoptotic rates were nearly the same (Figure 1C). Remarkably, LY3295668 treatment contributed to a 2- to 3-fold increase in Annexin V-positive percentages in RB1-deficient cell lines (Y79, RB355, and WERI-Rb1) compared to RB1-wt cells (RB3823 and RB522) (Figure 1C). Notably, in the RB1-deficient cell line Y79, AURKA phosphorylation (p-AURKA) at the T288 site was inhibited upon treatment with 0.1 μM LY3295668, with a concomitant increase in cleaved caspase-3 (Figure 1D). However, there were undetectable changes in caspase-3 cleavage in RB1-wt cell line RB522 despite p-AURKA being reduced to the same level. These findings suggest that RB1-deficient retinoblastoma cells exhibit differential sensitivity to AURKA inhibitors, accompanied by cell proliferation inhibition and apoptosis.

3.2 Dependence of RB1 on the response to Aurora-A kinase inhibitor treatment

To avoid genetic heterogeneities among the tumor cell lines mentioned above and evaluate the cell fate under different RB1 protein dosages, a genetic knock-out was introduced into the RB3823 cell line using the CRISPR/Cas9 system. As shown in Figure 2A, RB1 protein expression was low in the sgRNA-RB1 KO strain but unaltered in the scrambled sgRNA-edited strain. Moreover, the RB1 gene was re-introduced into the sgRNA-RB1 KO strain to generate an RB1-KO/rescue strain with similar RB1 protein levels as the scramble strain (Figure 2A). Therefore, these three strains originated from the same parental cell line and harbored an identical genetic background, except for RB1 status.

Responses of different RB1 protein abundance in retinoblastoma cells upon AURKA inhibitor treatment. (A) sgRNAs with scramble or RB1 target were used to knockout RB1 gene in RB3823 cell line using CRISPR-Cas9 method. For the sgRNA-RB1 knockout strain, the RB1 gene was re-introduced using the plasmid-based lipo-transfection method. (B) after treatment of LY3295668, dose–response curves were analyzed for cell lines with different RB1 protein levels using CTG assay. (C) after 0 μM or 0.1 μM LY3295668 treatment for 24 h, cell apoptosis rates were detected in the indicated cell lines using Annexin V staining assay. (D) Apoptosis biomarkers changes from the indicated cell lines after LY3295668 treatment or not were investigated in western blot assay

Treatment with AURKAi LY3295668 exerted a significant inhibitory effect on the RB1-KO strain (IC50 = 0.03 μM) but RB1-scramble and RB1-KO/rescue strains showed compromised potencies with IC50 values of 0.30 μM and 0.24 μM, respectively (Figure 2B). In the Annexin V staining assay, apoptotic rates increased by nearly 60% in the RB1-KO strain after AURKAi induction but only 10% in RB1-scramble and RB1-KO/rescue strains (Figure 2C). Consistent results were observed when LY3295668 upregulated the markers for apoptosis and cleaved forms of caspase-3 and PARP-1 proteins in the RB1-KO strain. In contrast, cell lines with intact or ectopic RB1 proteins showed moderate cleavage of caspase-3 and PARP-1 after AURKA inhibition (Figure 2D). Altogether, these data indicate a dose dependence of RB1 on AURKA kinase blockade-induced cell growth inhibition and apoptosis.

3.3 Aurora-A kinase inhibition exacerbates mitotic abnormalities in the absence of RB1

Previously, Aurora kinase family members were found to coordinate normal centrosome maturation and priming mitotic spindle assembly (SAC). To understand why RB1-deficient retinoblastoma cells showed high sensitivity to AURKA inhibitors, we analyzed whether LY3295668 affected cell cycle arrest in the absence of the RB1 gene. Under normal culture conditions, all cultured cell lines remained in a basal state with minimal polyploid cells. After treatment with 0.1 μM LY3295668, the RB1 knockout strain produced a significantly higher percentage of polyploid cells than the parental RB3823 cell line (Figure 3A). In contrast, recovery of RB1 protein expression in the RB1-KO strain alleviated its polyploidy state to the levels in the parental strain. Therefore, RB1-deficient retinoblastoma cells showed delayed mitosis in response to AURKA inhibition compared to a minor effect exerted on RB1-sufficient cells.

Mitotic arrest and colony formation impairment effects upon Aurora-A kinase inhibition. (A) the polyploidy rates were analyzed from the indicated cell lines upon LY3295668 treatment in two independently repeated cell cycle assays. Cells with more than 4 N DNA content in cell cycle analysis were included as the polyploidy state. (B) Changes of cell mitotic-related biomarkers were detected within RB1-wt or RB1-KO cell lines after different concentrations of LY3295668 treatment via western blot assay. (C) Analysis of cell colony formation was performed in three different cell lines upon treatment with LY3295668 or not for 2-weeks. (D) Cell colonies in Figure 3C were visualized and counted in microscope from six random visual fields, and then quantitatively analyzed

We found that LY3295668 caused a dramatic increase in the phosphorylation of histone H3 and stabilization of the APC/C substrate cyclin B1 in RB1-KO cell lines (Figure 3B). As these two markers are indicators of cell mitotic arrest, the observed concentration-dependent increases in phospho-histone H3 and cyclin B1 further confirmed the exacerbation of mitotic arrest in RB1-deficient cells upon AURKA kinase inhibition.

We reasoned that cell cycle arrest would lead to the inhibition of long-term cellular colony formation. To test this, we first exposed the RB3823 cell line and its two genetically edited strains to 0.1 μM LY3295668 for 14 days and then examined with crystal violet staining (Figure 3C). Treatment of RB1-deficient cells with the AURKA inhibitor induced dramatic inhibitory effects on cellular colony formation, with this effect being less pronounced in the parental cell line and RB1-KO/rescue strains in the presence of RB1 expression (Figure 3C,D). Collectively, AURKA inhibition caused lasting mitotic arrest effects in RB1-KO cells and impeding long-term colony formation; this may explain the synthetic lethality between RB1-LOF and AURKA inhibition in retinoblastoma tumors.

3.4 In vivo efficacy of Aurora-A kinase inhibition in RB1-deficient retinoblastoma xenografts

Next, we investigated the efficacy of LY3295668 in mouse retinoblastoma xenografts. NOD-SCID mice were engrafted with RB1-deficient cell line Y79 and randomized into four groups before oral administration of 25 or 50 mg/kg LY3295668 twice daily, or intravenously injected with 2.5 mg/kg cisplatin once a week or vehicle (saline) for 30 days. In comparison with the vehicle and cisplatin groups, LY3295668 exhibited a significant dose-dependent tumor growth inhibitory efficacy (p < 0.01 for 25 mg/kg; p < 0.001 for 50 mg/kg) (Figure 4A). At 50 mg/kg BID of LY3295668, xenografted tumors reached a static status. These two doses of LY3295668 were well tolerated in the tested mice, with no obvious bodyweight loss and hematological toxicities, consistent with previous findings that AURKA-specific inhibitors produced fewer side effects in animal studies16, 17 (Figure 4B,E,F). Notably, the classic cytotoxic agent cisplatin did not exhibit tumor inhibitory effects in the Y79-xenograft model even at higher doses that led to maximal tolerance toxicity with 23.2% loss of body weight (Figures 4A,B).

Retinoblastoma xenografts exhibited a differential response to Aurora-A kinase inhibitor LY3295668. (A, B) in Y79 cell xenograft mice, the effects of two doses of LY3295668 (25 and 50 mg/kg BID) were compared with cisplatin (2.5 mg/kg, Q1W) after administration for 30 days (A). The related animal body weights were measured and compared within different groups (B). N = 7 in three treatment groups, respectively. (C, D) in RB522 cell xenograft mice, two doses of LY3295668 (25 and 50 mg/kg BID) and cisplatin (2.5 mg/kg, Q1W) were treated, and tumor volumes were compared until 30 days (C). The related animal body weights were measured and compared (D). N = 7 in three treatment groups, respectively. (E, F) in Y79 cell xenograft nude mice upon continuous dosing of vehicle or LY3295668 (25 and 50 mg/kg BID), blood was collected on day 15 and day 30 and analyzed white blood cell count (E) and hematocrits (F). N = 5 in each group. (G) Schematic chart exhibits the molecular machnism underlying synthetic lethality between RB1 deletion and AURKA inhibition in retinoblastoma. The left part shows normal cell state with RB1 proficiency; the middle one depicts RB1 deletion could cause SAC hyperactivation, but tolerated defective mitosis in tumor; and the right section exhibits the RB1-AURKAi synthetic lethality-induced mitotic arrest and cell apoptosis

Importantly, the same dosing conditions were applied in the mouse model xenografted with RB1-sufficient cell line RB522. However, all dosing groups showed limited therapeutic efficacy (Figure 4C) under maximal tolerant dosages that resembled the Y79 xenograft model (Figure 4D). Based on the above findings, we suppose that the AURKA-specific inhibitor may target specifically RB1-deficient retinoblastomas in vivo with less toxicity than clinically approved cytotoxic agents but exhibit lower efficacy in other tumor models with RB1-wt status.

4 DISCUSSION

Retinoblastoma, an aggressive childhood cancer of the retina, is characterized by the biallelic loss of the tumor suppressor RB1 gene in nearly all cases. Although the mortality is relatively low upon multimodal treatment in developed countries, approximately 50% of patients experience vision loss and late effects under therapy in developing regions. Unsolved clinical needs still require innovative translational research and drug development.

The tumor suppressor gene, RB1, is commonly recognized as a key player in repressing E2F-mediated transcription, thus controlling the G1-S transition in the cell cycle. Meanwhile, RB1 loss has also been found to induce hyperactivation or priming of the spindle assembly checkpoint (SAC).18 Mutated tumor suppressor genes, including RB1 deficiency, have thus far remained actionable in current therapeutic discovery; the discovery of synthetic lethality between RB1 LOF and Aurora kinase inhibition12, 13 offers a new opportunity (Figure 4G).

The Aurora kinase family that consists of serine/threonine kinases Aurora-A, -B, and -C has been found to play key roles in mitotic progression including centrosome maturation and mitotic spindle assembly. They are frequently overexpressed in human tumors and have been linked to oncogenic transformation, partially through the induction of chromosomal instability in cancer cells. Currently, Aurora kinases have been investigated as potential anticancer targets and have entered early clinical development stages.19 However, early clinical development data exhibited bone marrow-suppressive toxicity and displayed modest efficacy owing to the limited maximal tolerated doses.20 Therefore, there is a demand for sensitive biomarkers that allow effective therapy in predefined patients.

Recently, two companion publications identified that multiple types of tumor cells lacking the RB1 gene are hyperdependent on Aurora kinase activity for survival.12, 13 Moreover, RB1 deficiency has been validated as a predictive biomarker for sensitivity to AURKA or ARUKB inhibitors. Based on these pilot findings, we attempted to apply this knowledge to retinoblastoma cases in the current study.

In retinoblastoma cells, we found that AUKRA kinase inhibitor caused tumor growth inhibition and apoptosis both in vitro and in vivo in an RB1 dose-dependent manner. Analysis of the mechanism of action suggested that AURKA kinase inhibitor exerted increased mitotic arrest exacerbation in RB1 deficient status compared to wild-type status. These redundant roles in delaying cell mitosis between RB1 deficiency and Aurora kinase inhibition in retinoblastoma may potentially provide evidence of their synthetic lethal relationship (Figure 4G). In addition, the overlap of timepoints between cell apoptosis and mitotic arrest (Figures 2D and 3) suggested that AURKAi-induced RB1-deficient retinoblastoma cell death was caused by the failure to escape the cell cycle arrest.

Aurora kinase inhibitors have severe myelosuppression side effects that limit their clinical advancements. Here, we chose ARUKA inhibitor because this subtype of inhibitors including MLN-8054 and MK5108 seem to have fewer myelosuppression toxicities; however, pan-Aurora or AURKB inhibitors led to frequent hematological side effects.16, 17, 21-23 The AURKA inhibitor LY3295668 exhibited 1000-fold higher specificity targeting AURKA versus AURKB, and manifested nonobserved hematological toxicities in the mouse model (Figure 4E,F). The kinome profiling assay also showed its wide spectrum specificity with AURKA as the most potent kinase target,13 thus providing molecular evidence for the synthetic lethality between AURKA and RB1 in retinoblastoma.

On the other hand, the AURKA inhibitor achieved more than 10-fold potency in RB1 deficient cells when compared with RB1 proficient cells (Figure 1B). Combining this evidence, we assumed that specific AURKA inhibitors would provide a therapeutic window with lower effective doses and less toxicity, and would outperform current cytotoxic agents as the standard care in retinoblastoma systematic therapy.

However, when considering the blood-retinal barrier, there remain unique challenges in the intraocular distribution of candidate drugs that may compromise their treatment intensities.9, 10 To address this need, a tumor xenograft site shifted from the subcutaneous flank to the orthotopic intraocular site will be attempted in our forthcoming research.8 The potential benefits of systematic treatments and local drug delivery will also be investigated in retinoblastoma.

Retinoblastoma, as a highly unmet clinical need, is difficult to treat due to the limited availability of systematic regimens with few side effects. To this end, a promising approach using an AURKA inhibitor has been developed to fight against RB1 deficient retinoblastoma.

CONFLICT OF INTEREST

The authors have no conflicts of interest to declare.