**1. Introduction**

Retinoblastoma is the most common intraocular tumor of childhood affecting 1 in 15,000 to 1 in 18,000 live births [1–3]. Retinoblastoma is highly curable if diagnosed in the early stages. For decades,

67

surgical removal of the affected eye (or both eyes in cases of bilateral retinoblastoma) has been the first choice. Later on, the introduction of chemotherapy provided the basis for eye preservation. Over the last decade, retinoblastoma treatment has radically changed from using systemic chemotherapy infusion, with low bioavailability in the ocular tissues but high in plasma, and thus severe systemic adverse events, to highly-selective novel techniques of drug delivery, including direct ophthalmic artery chemotherapy and intravitreous injection [3–5]. Though a striking increase in ocular survival has been attained, eyes with tumors that grow from the retina to the vitreous humor, namely vitreous seeds, are more difficult to cure and may relapse. Therefore, vitreous seeds remain a challenge in the managemen<sup>t</sup> of intraocular retinoblastoma and removal of the affected eye may be the only treatment option [5,6]. Recently, Munier et al. classified vitreous seeds into "dust", "spheres", and "clouds" based on their heterogeneous appearance at fundoscopy [6]. Dusts are composed of loose tumor cells in the vitreous, clouds are dense tumor fragments formed by translocation of the primary tumor content to the vitreous, and spheres are translucent solid tumors formed by further clonal growth of the dust or the cloud, or by sprouting of the primary retinal tumor [7]. Each class of seeds required a different cumulative dose and number of intravitreal injections of melphalan to achieve complete response to treatment [7–10]. Later on, each class of seeds is also correlated with histopathological features [11]. In general, clouds need the greatest number of melphalan injections (and cumulative doses) followed by spheres, and finally dusts [8–10]. Dusts might be more sensitive to treatment because they are more accessible to drugs, as they are composed of clusters of loose cells, while spheres or clouds grow as tight clusters with different layers of viable cells and may hamper homogeneous distribution of the drug after intravitreal injections [8,9]. Thus, assessment of the capacity of chemotherapy to penetrate into the tumor seeds, and thereby to become available to exert its cytotoxic effect would certainly add to the knowledge on and improvement of chemotherapy use and patient management.

Among the antineoplastic agents used for retinoblastoma, topotecan has been widely used based on its effect in preclinical models and clinical studies [12]. To quantify topotecan penetration into living tumorspheres, confocal microscopy could be used as topotecan in a fluorescent drug [13,14]. This powerful technique allows for the visualization of structures, including living cells and even thick living specimens, after noninvasive serial optical sectioning [15].

Over the last years, the development of three-dimensional (3D) in vitro tumor models allowed the establishment of structural complex cell-cell interactions in systems that resemble in vivo tumors [16,17]. Still, further developments using 3D tumor culture technology are needed for translational studies in retinoblastoma. Thus, we explored a scaffold-free culture method allowing self-aggregation of tumor cells into a 3D structure to closely resemble vitreous seeds. In serum-free, growth factor-supplemented culture conditions, retinoblastoma cells grow in suspension and form tumorspheres that may recapitulate vitreous seeds of the spheres class observed in vivo, providing a suitable in vitro tumor model to study drug penetration [18,19]. In contrast, commercial retinoblastoma cell lines grow as loose aggregates, resembling vitreous seeds classified as dust.

Thus, the aim of our study was to establish and characterize tumorpsheres derived from patients with intraocular retinoblastoma and use them to assess if differences in tridimensional conformations and sphere size affect topotecan penetration using confocal microscopy. The developed tridimensional cell model resembles several characteristics of vitreous seeds in pediatric patients with intraocular tumors. This model was useful to assess live-cell imaging of chemotherapy penetration for drug distribution studies and cytotoxicity assessment.

### **2. Results**

### *2.1. Patient-Derived Tumorspheres Resemble the Original Tumor*

Early-passage cell cultures derived from intraocular tumors of two patients who underwent upfront enucleation without receiving previous treatment were obtained and named after the codes

HPG-RBT-12L and HPG-RBT-26. They were considered established after three passages, each one performed during the log-phase growth.

Analysis of DNA showed that in both primary cells the *RB1* gene mutations were germline and single base substitutions. *RB1* mutation in HPG-RBT-12L cells, as well as in the parental tumor was identified as a point mutation in exon 15 (NM\_000321.2(RB1):c.1421G>T) associated with altered splicing, while for the HPG-RBT-26 cells and tumor in exon 23 (NM\_000321.2(RB1):c.2359C>T(p.Arg787\*)), it was associated with a premature stop codon. Moreover, the short tandem profile (STR) for the DNA from the cell lines was identical to that obtained for the tumor DNA (Tumor samples: HPG-RBT-12T and HPG-RBT-26T), confirming the origin of the cell cultures (Table 1). There was no significant overlap between both primary cell lines and no cell line corresponded to the DNA profile of the present retinoblastoma cell lines in the STR database of the American Type Culture Collection.


**Table 1.** Short Tandem Repeat analysis of the cell lines and matched tumors.

As shown in Figure 1, both primary cell cultures were positive for cone-rod homebox transcription factor (CRX) by RT-qPCR and presented similar expression levels with respect to the commercial cell line Y79.

Both small and large patient-derived tumorspheres, as well as Y79 cells, were positive stained for arrestin 3 and synaptophysin, confirming the retinal and neuroectodermic tumor cell origin (Figure 2A,B). Notwithstanding the size of the 3D model, all of them were histologically composed of viable and proliferative tumor cells, as shown in Figure 2C,D, respectively.

Taking into account the morphology, HPG-RBT-12L and HPG-RBT-26 cell cultures grew as freely floating tridimensional structures in serum-free medium, matching the classification of seeding for "spheres". These 3D structures formed spontaneously in the culture media resulting in different sizes. A representative image of small and large tumorspheres from both patients is depicted in Figure 3 showing the morphological features. Small tumorspheres from patient 1 and 2 (Figure 3A,B) showed a median (range) diameter of 91 μm (53–109) and 53 μm (41–79), respectively. In the case of large tumorspheres (Figure 3C,D), median diameters (range) were 353 μm (341–381) and 356 μm (334–385) for patients 1 and 2, respectively.

**Figure 1.** Expression of cone-rod homebox transcription factor in patient-derived cells.Expression of the cone-rod homebox transcription factor (CRX) was determined by RT-qPCR and transcript levels were quantified relative to the housekeeping gene and then normalized by the level of mRNA CRX detected in Y79 cells. Data is shown as mean (SEM).

**Figure 2.** *Cont.*

**Figure 2.** Characterization of the tridimensional retinoblastoma cell structures. Representative small and large 3D tumor cell cultures resembling both spheres (HPG-RBT-12L) and dusts (Y79) show: (**A**) Cone photoreceptor-specific staining (ARR3). (**B**) synaptophysin stain confirming neuronal characteristics of the cells. (**C**) strong nuclear basophilic staining demonstrating they are composed of viable tumor cells; and (**D**) Ki-67 expression showing positive brown-staining all through the 3D structure. Scale bar, 100 μm. Images taken at 20× magnification. Abbreviations: ARR3, arrestin3; H&E, hematoxylin and eosin.

**Figure 3.** Representative confocal microscope images of in vitro tumorspheres derived from intraocular retinoblastoma tumors. Small (**A**,**B**) and large (**C**,**D**) spherical aggregates of primary tumor cells (HPG-RBT-12L and HPG-RBT-26 cells) obtained by culturing samples from intraocular tumors of patients. Images taken at 20× magnification. Scale bar, 100 μm.

### *2.2. Topotecan Penetration and Live-Cell Imaging*

Confocal microscopy images were collected from 3D tumor models exposed to topotecan. Supplementary Video 1 shows sequentially acquired and recorded images from a representative experiment in a large HPG-RBT-12L tumorsphere before and 5 min after topotecan administration. The penetration of the green color corresponds to topotecan fluorescence from the bathing solution into the outer layers of the tumorsphere and thereafter into the core as a function of time. Different rates of topotecan penetration were observed in tridimensional structures formed by Y79 cells (fast penetration) and in patient-derived tumorspheres (slow penetration) (Figure 4). A uniform green signal corresponding to topotecan fluorescence was observed even in large clusters, as topotecan penetrated into the Y79 aggregates in less than 0.5 min due to the lack of strong cell-cell interactions, as shown in Figure 4A. Y79 cells grow as loose clusters in suspension, resembling dust vitreous seeds. This result was in contrast to the delayed penetration in patient-derived spheres. As shown in Figure 4B, after 0.5 min of topotecan exposure, the spheres are black due to the lack of topotecan penetration into the 3D structure. In large spheres, topotecan penetration into the tumorsphere was evidenced by the progressive acquisition of green fluorescence from the outer to the core of the sphere over 5 min of drug exposure (Supplementary Video 1).

**Figure 4.** Topotecan penetration and activity in retinoblastoma 3D structures. ( **A**) Topotecan fluorescence as green color after 0.5 min of incubation in Y79 cells (dust model) and (**B**) in small and large HPG-RBT-12L cells (sphere model). Note the absence of staining within small and large spheres after incubation with topotecan for 0.5 min. ( **C**) Cytotoxic activity of topotecan (1 μg/mL) shown as accumulation of ethidium bromide in a tumorsphere treated for 10 min; dead nucleated cells are stained red. ( **D**) Control tumorsphere (untreated control) stained green with acridine orange. Scale bar, 100 μm.

The time for maximum topotecan accumulation into tumorspheres (*t*max) was defined as the time to achieve maximum fluorescence in the core of the sphere. A decrease in tumorsphere size correlated with a shorter time for topotecan accumulation from the bathing solution into the core of the tumorsphere for both cell models (*p* < 0.05; Table 2). However, no significant differences in *t*max were found between cell models for a similar size range (*p* > 0.05).

**Table 2.** Time to achieve maximum fluorescence in the core of small and large retinoblastoma spheres from models HPG-RBT-12L and HPG-RBT-26.


Abbreviations: *t*max, time to achieve maximum fluorescence in the core of the sphere. Data is showed as mean (SEM) of three independent experiments for each cell model.

### *2.3. Topotecan Cytotoxicity*

In order to characterize the sensitivity of the cell lines to the chemotherapeutic agent, we determined the IC50 of topotecan for the three cell cultures. Mean (range) topotecan IC50 in Y79, HPG-RBT-12L, and HPG-RBT-26 cells was 28.6 nM (26.1–30.5), 12.3 nM (6.8–13.7), and 6.3 nM (5.4–8.6), respectively. Topotecan IC50 in Y79 was similar to the value previously reported and significantly higher than the values attained in HPG-RBT-12L and HPG-RBT-26 (*p* < 0.05). However, no significant difference could be detected in topotecan IC50 between primary cell cultures (*p* > 0.05, Figure S1).

Because topotecan needs at least 10 min to exert its cytotoxic activity in commercial cell cultures [20], we assessed the cytotoxic effect of topotecan penetration into the spheres by ethidium bromide staining after at least 10 min of topotecan exposure. As we used a concentration of topotecan that was clinically relevant and higher than the IC50 determined in the cell cultures, it was expected that topotecan would exert a cytotoxic effect in the tridimensional cell clusters. Complete cell damage was observed by ethidium bromide red staining as shown in Figure 4C, compared to a control (topotecan-free) tumorsphere stained as green with acridine orange (Figure 4D), supporting the hypothesis that once inside the cells that composed the spheres, topotecan exerted cytotoxic activity.
