**3. Discussion**

In this study, we established patient-derived retinoblastoma tridimensional culture resembling the architecture of spheres, a class of vitreous seeds, and for the first time used live tumorsphere imaging to gain insight into the penetration process of a chemotherapeutic agen<sup>t</sup> in such tumor models. After a thorough characterization of the cell lines, we took advantage of the fluorescence of topotecan in conditions compatible with live-cell monitoring to visualize and quantify its penetration into the core of tumorspheres derived from two patients with intraocular retinoblastoma and in cell clusters of Y79 resembling spheres and vitreous dust seeds, respectively [7,8,11]. The quantification of topotecan at different times based on the fluorescence intensity showed that drug penetration into the core of the cell aggregates was immediate in dust, and faster in small than in large spheres. This result is consistent with pathological observations stating that larger seeds consist of a greater number of cells arranged in multilayers around a central core. Therefore, spheres are not mere aggregates of cells but 3D structures that may hamper drug accessibility.

The property of topotecan fluorescence has been used to quantify the drug in several biological matrices for pharmacokinetic studies [12,14]. In addition, different researchers have exploited this property to study the release rates of topotecan from liposomes and nanoparticles as part of the development of new formulations, or to assess temporal variations in the disposition of topotecan in

living cells or even whole animals using non-invasive fluorescence microscopy [21–24]. As a novel application, we previously reported the use of fluorescent visualization of topotecan to evaluate the safety of the injection technique and adequate drug delivery in retinoblastoma patients using a portable Wood's lamp [13].

Ours is the first study on the imaging of topotecan distribution in living retinoblastoma cells spontaneously clustered in a 3D structure, as occurs in vivo. Previous studies on topotecan penetration in tumorspheres were performed in other tumor cell lines forced to aggregate into a sphere by gravity using the hanging drop technique [24,25]. In brief, the hanging drop technique consists of plating droplets of cells in culture media on a lid, inverting the lid over a dish, and place it in an incubator to allow for the 3D cell structure to form. Other researchers have worked with bioengineered spheres consisting of commercial Y79 cells and gelatin microparticles as a scaffold to obtain the 3D structure [26]. Specifically in that study, Y79 cells showed genomic alterations, including changes in the expression of the collagen family when using microparticles as the matrix on which they were grown. Thus, it may be expected that an altered drug penetration into the Y79-microparticle induced by the culture conditions may not resemble the in vivo behavior of tumor seeds. In contrast, our patient-derived tumor cells spontaneously grow as tumorspheres without the need for external forces, or polymer support resembling floating vitreous seeds, as occurs in patients.

After upfront enucleation of two patients affected with retinoblastoma, we established two patient-derived cell models. Then, we first confirmed the origin of the cell lines by different genomic analysis and the identification of the expression of markers of retinal and neuronal tumor origin to sustain the basis of all further developments. Importantly, cell-growth conditions avoiding the use of serum for culturing the cells from the biopsy have been shown to dramatically improve tumor cell selection and to avoid differentiation of the cell culture to a different phenotype than the parental tumors [18,19]. Moreover, it is important to mention that cell-growth conditions and the cell line passage number may result in different pharmacological sensitivity of the cell cultures as previously reported for retinoblastoma and other tumors [27,28]. As shown in the present study, commercial retinoblastoma cells spontaneously grow in loose aggregates of cells that allow almost instantaneous drug penetration, likely resembling the dust vitreous seeds. Therefore, caution should be exercised if using commercial cell lines but also different cell growth conditions to establish primary cell cultures of retinoblastoma to use them as a basis of pharmacological screening and translation into the clinics.

The structural properties of the patient-derived tumorspheres were similar to vitreous seeds classified as "spheres", based on morphological and histopathological observations [7,8,11]. Patient sphere sizes were reported to be in the range of 15 to 300 μm or even larger. In line with this observation, our tumorspheres obtained from in vitro culture are also within that range (Figure 3). In addition, small patient spheres displayed a homogeneous positive staining for Ki-67 as a marker of cell proliferation widely used in tumor tissue biopsies. We also showed that large tumorspheres are composed of multilayers of cells and that all cells located in the core or the surroundings were proliferative with almost identical expression levels of Ki-67. Thus, our model was able to recapitulate the morphology and phenotype of dusts and spheres of human vitreous seeds.

Importantly, the 3D multilayer cell structures may hinder drug penetration, and thus, diffusion studies of active drugs in each in vitro model, and eventually the evaluation of drug sensitivity in 3D models should be performed [29,30]. We demonstrated, by means of confocal microscopy of topotecan-treated spheres, that the drug penetrated upon the core of the tridimensional structures in all cases. Also, we observed different rates of drug penetration, depending on the tridimensional structure, resembling dusts or spheres, and the size of the in vitro spheres. Topotecan penetration was almost immediate in clusters of Y79 cells as they gathered in a loose packing 3D structure. Thus, although we observed a lower sensitivity to topotecan (higher topotecan IC50) of Y79 compared with patient-derived tumor cells, the drug exerts its cytotoxic activity immediately after exposure of the loose aggregates of Y79 cells. On the contrary, as a result of the multicellular architecture of the 3D structures that allow cell-cell interactions, patient-derived spheres presented longer times of drug penetration into the core and also differences between sphere sizes. Nonetheless, topotecan penetrated through all the evaluated 3D models and exerted a cytotoxic effect. These results obtained in vitro show a strong cytotoxic activity of topotecan in Y79, and primary cell cultures may differ from the clinical observation in retinoblastoma patients. Topotecan is usually administered with carboplatin or melphalan based on the favorable ocular disposition, the synergistic cytotoxicity, and because of the lack of documented clinical efficacy if administered as a single-agent [31].

Finally, and based on previous reports on the classification of vitreous seeds and the present results, we may speculate that the longer it takes for topotecan to reach the inner layers of the tumorspheres, the higher the number of doses of chemotherapy required for tumor control, as topotecan penetration may be used as a surrogate for cytotoxicity. This assumption is based on the cytotoxic effect of topotecan revealed by ethidium bromide after its penetration into the spheres. Nonetheless, the relation between the number of doses and the actual dose of topotecan with the response in dusts, spheres, and clouds should be assessed in clinics.

There are some limitations of the present 3D cell culture model that could be acknowledged. Heterogeneity of patient samples may limit the generalization of drug response and penetration rates through spheres that could be addressed if a larger number of samples are obtained to estimate interindividual variability in drug response. However, the lack of fresh primary tumors, specifically from patients enucleated upfront, and the difficulties in establishing primary tumor cell cultures limit the availability of a wide range of 3D models for retinoblastoma. In addition, technical challenges of the 3D cell culture technique relates to the ability of generating the same morphology and size of the aggregates and to the mobility of free-floating spheroids in suspension, posing a challenge to temporal imaging of the same spheroid. Also, it should be taken into account that the present study was performed with topotecan and that other drugs used for retinoblastoma treatment, including melphalan, may show different penetration rates into the 3D structures depending on the physicochemical properties among other factors.

Altogether, we developed an in vitro tumor model that resembles retinoblastoma vitreous seeds based on a close phenotype-genotype correlation. Under the cell culture conditions, these cells grow as tumorspheres with the ability to interact in a 3D structure. Tumorspheres provide a valuable model to study in vitro drug penetration as a surrogate for drug exposure in vitreous seeds that may be useful to optimize drug therapy, and ultimately, to improve the efficacy of retinoblastoma treatment.

### **4. Materials and Methods**

### *4.1. Ethics Statement*

Patient retinoblastoma samples were obtained after protocol and informed consent approval by Hospital de Pediatria JP Garrahan Institutional Review Board (protocol number 904, date of approval: 26 February 2016). Written informed consent was obtained from parents or guardians before sample collection.

### *4.2. Retinoblastoma Cell Line*

The commercial cell line Y79 was obtained from the American Type Culture Collection (HTB-18, Manassas, VA, USA). Cells were cultured at 37 ◦C with 5% CO2 in RPMI-1640 medium (Paisley, Scotland, UK) with 20% fetal bovine serum (FBS, Internegocios, Cordoba, Argentina), as previously reported elsewhere [19,27].

### *4.3. Establishment of Patient-Derived Cell Cultures*

An ophthalmologist collected fresh primary tumor samples from patients with intraocular retinoblastoma who underwent upfront enucleation as primary treatment. Tumor samples were mechanically disaggregated and after centrifugation, and cells were cultured in serum-free neural stem-cell medium as previously described for other pediatric stem cells culture conditions [19,27]. Briefly, cells were grown in serum-free neurobasal medium (Thermo Fisher Scientific, Grand Island, NY, USA) containing DMEM/F12 (Thermo Fisher Scientific), supplemented with B-27 (Thermo Fisher Scientific), heparin (Sigma-Aldrich), EGF and FGF (epidermal and fibroblast growth factors, respectively, Thermo Fisher Scientific), and PDGF (Platelet derived growth factor, PeproTech, Rocky Hill, NJ, USA). Cultures were maintained at 37 ◦C in an incubator with a humidified atmosphere of 5% CO2 and 95% air in T25 flasks at a density of approximately 10<sup>6</sup> cells/mL before starting the experiment.

### *4.4. Cell Authentication and Retinal Lineage Markers*

Genomic DNA was isolated from cell lines, blood, and tumor samples using PureLink Genomic DNA mini kit (Invitrogen, Carlsband, CA, USA) following the manufacturer's instructions. After checking the quality and quantity of the DNA by Nanodrop 2000 (Thermo Scientific, Waltham, MA, USA), short tandem repeat (STR) profiling was performed in tumor samples and cell lines for authenticating the origin of the cell lines by means of the analysis of 15 autosomal STR loci and amelogenin. RB1 mutations were assessed by direct sequencing of the 27 exons and the promoter region of the RB1 gene, and mutations were described with reference to GenBank accession # L11910 [32]. Multiplex Ligation-dependent Probe Amplification assay (MLPA, MRC, Amsterdam, The Netherlands) was performed according to the manufacturer´s protocol to screen for deletions or duplications in the *RB1* gene.

The expression of the photoreceptor lineage marker cone-rod homeobox transcription factor (CRX) was analyzed to confirm the retinal tumor origin. Real-time quantitative PCR (RT-qPCR) was used to evaluate the expression of CRX mRNA in the patient-derived cell cultures using TaqMan technology in a 7500 Sequence Detection System (Applied Biosystems, Foster city, CA, USA) [33]. Briefly, total RNAs was isolated from tumor cells using PureLink RNA Mini Kit (Thermo Fisher Scientific) following the manufacturer´s instructions. After RNA quantitation using NanoDrop spectrophotometer, RNA was reverse-transcribed into cDNA using random primers and the SuperScript III kit (Invitrogen). The sequences of primers and probes used for RT-qPCR to analyze CRX mRNA expression were as follows: Forward: 5-AGGTGGCTCTGAAGATCAATCTG-3, Reverse: 5-TTAGCCCTCCGGTTCTTGAA-3, and probe 5-FAM-CTGAGTCCAGGGTTC-3-MGB. Relative expression of CRX mRNA was determined in cell cultures using the 2−ΔΔ*C*<sup>t</sup> method, where Ct is the threshold cycle of the target product and that of the housekeeping gene. Then, primary cell culture data was normalized against CRX mRNA expression levels obtained in Y79.

Additionally, immunohistochemistry was performed in paraffin embedded cells for synaptophysin (NCL-L-SYNAP-299, Leica BioSystems, Newcastle, UK) and arrestin 3 (ARR3, 11100-2-AP, Proteintech group, Chicago, IL, USA) expression to confirm the neuronal and cone-specific origin of the cells, and hematoxylin-eosin staining for morphological assessment [19]. We also evaluated the immunohistochemical staining for Ki-67 as a marker of proliferating but not quiescent (G0 phase) cells using Ki-67 antibody (Ki-67 anti-human clone, Dako, Denmark), and the percentage of Ki-67 positive cells was compared between tumorspheres of different sizes and origins. The percentage of Ki-67 positive cells was calculated using ImageJ software (NIH, Bethesda, MD, USA) [34].

### *4.5. Live-Cell Confocal Microscopy*

To quantify the penetration of topotecan into retinoblastoma patient-derived spheres, two sizes of tumorspheres were considered based on visual inspection with the microscope. "Small" (< 110 μm) and "large" (110–400 μm) tumorspheres were classified according to the diameter calculated from an optical slice corresponding to approximately the middle of each sphere.

First, cell culture viability was evaluated by means of trypan blue exclusion assay. Then, ten microliters of each cell suspension of approximately 10<sup>6</sup> cells/mL were seeded on a slide covered by a clean 15 mm round coverslip, and an appropriate field with tumorspheres was located. Untreated tumorspheres were evaluated for natural fluorescence in a separate experiment. As shown in Supplemental Figure S2A, there was no acquisition of signal before topotecan exposure, while after

6 min of drug addition to the medium (Figure S2B), the sphere acquired a complete green fluorescence due to topotecan penetration. Then, 1 μL of a 10 μg/mL solution of topotecan (final concentration 1 μg/mL) was added. This concentration was selected based on the dose of topotecan commonly used for intravitreal injection (30 μg) and the volume of distribution in human vitreous humor (4 mL), resulting in an in vivo concentration similar to that used for in vitro evaluation [35,36]. Immediately after, and thereafter every 30 s, fluorescent images were obtained using an Olympus Fluoview FV1000 confocal laser scanning microscope (Olympus, Tokyio, Japan) with imaging software (Olympus Fluoview FV10-ASW v1.7c, Melville, NY, USA), and equipped with a UPlanSApo 20X/0.75 NA objective. Excitation was provided by a 458 nm line of a multiline argon laser at a 25% transmittance, and fluorescent images were collected with a 505–605 nm emission filter. Images were collected with a format of 1024 × 1024 pixels. Topotecan penetration into the tumorsphere was determined in terms of fluorescence intensity and images were processed using ImageJ software. At least three tumorspheres from each size were assessed for each patient-derived model. Autofluorescence contribution (natural fluorescence contribution of untreated tumorspheres) was evaluated in a different experiment. All optical images collected from each tumorsphere were sequentially acquired and recorded on video using ImageJ software.

Images were processed to quantify the penetration of topotecan into the sphere, as follows. First, anisotropic diffusion filtering was applied to smooth noise and defined region boundaries of each tumorsphere [37]. Then, the Otsu's thresholding segmentation method was used for segmenting the spheres from the background and also inside each tumorsphere to calculate the percentage of change of black and white pixels as a function of time as a surrogate of topotecan penetration [38,39]. The time to achieve maximum fluorescence in the core of the tumorsphere as a surrogate of complete penetration of topotecan was defined as the time needed for obtaining at least 90% of white pixels.

### *4.6. Topotecan Cytotoxicity and Cell Viability Assay*

Topotecan cytotoxicity was evaluated in each cell culture to determine the concentration of drug that causes a 50% decrease in cell proliferation or IC50, as previously reported [20,27]. Briefly, cells were counted, seeded in 96-well plates, cultured for 24 hours, and thereafter exposed to different concentrations of topotecan (0.001–1.000 nM). After incubation for 72 h, cell proliferation was determined by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) colorimetric assay. The IC50 was calculated using GraphPad Prism v.6 (GraphPad, San Diego, CA, USA).

At the end of the study, tumorspheres were stained using ethidium bromide (100 μg/mL, Sigma Aldrich) and acridine orange (100 μg/mL) double-staining for the visualization of nucleic acids of membrane-damaged cells (necrotic or cells in late apoptosis induced by topotecan, observed as stained red) and live cells (stained green), respectively, and observed under fluorescence microscope. As both reagents were added simultaneously, non-specific signals to acridine orange of ethidium bromide could be assessed in topotecan-treated and untreated 3D structures, respectively.

### *4.7. Statistical Analysis*

Comparison of the time needed for topotecan to fully penetrate into the core of the two sizes of tumorspheres derived from patients was performed by means of a *t* test with a significance *p*-value of 0.05.

### **Supplementary Materials:** Supplementary materials can be found at http://www.mdpi.com/1422-0067/20/5/ 1077/s1.

**Author Contributions:** Conceptualization, U.W., D.H.A., A.M.C., and P.S.; methodology, U.W., F.F., F.B., S.Z., R.A., M.S., C.S., A.M.C., and P.S.; cell line establishment and analysis, U.W., S.Z., and R.A.; confocal microscopy, U.W. and F.F.; formal analysis, U.W., F.F., F.B., D.H.A., A.M.C., and P.S.; patient treatment, M.S. and C.S.; clinical data analysis, M.S. and C.S. All authors significantly contributed to the design of the study, to the interpretation of the data, and to the writing and editing.

**Funding:** This work was supported by Agencia Nacional de PromociónCientífica-FONCYT (PICT 2016- 1505), Consejo Nacional de InvestigacionesCientificas y Tecnológicas (CONICET), FondationNelia et Amadeo Barletta, Fundación Natali DafneFlexer, and Fund for Ophthalmic Knowledge, NY, USA.

**Acknowledgments:** The authors want to thank Guillermo Chantada for his thorough discussions in preclinical modeling and clinical therapeutics, and Debora Chan and Juliana Gambini for their technical support with image processing. We also acknowledge the technical support with genomic studies of Irene Szijan and Marcela Menna.

**Conflicts of Interest:** The authors declare no conflict of interest.
