CE-CAM model for evaluating CD133lo Cancer Stem Cells in Retinoblastoma

Background Cancer Stem Cells (CSCs) reported in various tumors, play a crucial role in tumorigenesis and metastasis. Following the efforts to reduce, replace and rene the use of mammalian models, we aimed to establish a short-term xenograft for Retinoblastoma (Rb) to evaluate the tumorigenic and metastatic potential of CD133 lo CSCs in Rb Y79 cells, using the well-established chick embryo (CE) model. Methods Total and CD133 sorted Rb Y79 cells, labelled with eGFP/CM-Dil tracking dye, were transplanted onto the chorioallantoic membrane (CAM) of day-7 chick embryos and incubated for 7 days. The tumor formation on CAM and metastasis to the embryos were evaluated by confocal microscopy, in-vivo imaging, and histopathology.

were cultured at 5% CO 2 and 37 0 C until they reached over 80% con uence following which they were subcultured for the experiments.

Magnetic Activated Cell Sorting (MACS) and Flow cytometry
The Y79 cultured cells were sorted using CD133 micro bead kit according to manufacturer's protocol (Miltenyi Biotec Inc., auburn, CA) as described previously (18). Brie y, Cultured Rb Y79 cells were washed with MACS buffer (2mM EDTA and 0.5% FBS containing PBS; pH-7.2), and re-suspended with 300µl of MACS buffer. 100µl of FCR blocking reagent and 100µl of CD133 micro beads were added and incubated for 30 minutes at 4°C for CD133 magnetic labelling. The cell mixture was passed through the LS mini MACS columns followed by the repeated washes with MACS buffer, CD133 lo cells were eluted rst and collected in tube with media. The columns were removed from MACS magnet and CD133 hi cell fractions were eluted out using MACS buffer. Both the cell populations were concentrated and re-suspended in 1 ml of serum free media. The cell count and cell viability of post MACS were assessed by haemocytometer. The purity of the sorted cells was analysed by ow cytometry (BD LSRFortessa™).

Chick Embryo CAM assay
Embryonated white leghorn Gramapriya eggs (G. gallus) were procured from ICAR-Directorate of Poultry Research, Hyderabad, India. Ethics approval was obtained by the Institutional Ethics Committee at the University of Hyderabad and all the experiments were conducted in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The eggs were sterilized with 70% alcohol and incubated at 38 0 C and 68% relative humidity in an egg incubator (Sun Engineering, India). Growth, viability and vascularization of eggs were monitored by candling every alternate day post E3 stage. On E7-8 stage, a small circular window was made into the air sac of the egg using sterile micro-scissors. The CAM was identi ed by lifting the inner shell membrane and a small capillary was gently abraded. Two million cells of CM-Dil Y79 total (n=38), CD133 lo (n=41) and CD133 hi (n=53) and eGFP Y79 total (10), CD133 lo (n=10) and CD133 hi (n=10) cells in media containing Matrigel (BD Biosciences TM ) were transplanted on the CAM. The window was sealed back with sterile tape and was placed in the incubator undisturbed for 7 days until dissection.

CAM Tumor Volume measurement
The tumor volume measurements were performed for the eGFP Y79 cells transplanted tumor nodules. Brie y, the window was reopened to identify tumor nodules on day 14, the eGFP Y79 transplanted cells tumor volumes were measured using a Vernier caliper and volume was estimated by the following formula(23): Tumor Volume = 4/3×π×r 3 (r =1/2Ödiameter 1 ×diameter 2 ) CAM whole mount and Confocal Imaging The CAM tissues with the tumor nodules were gently rinsed with cold Phosphate buffered saline (PBS), placed on a glass slide and mounted with 50% Glycerol for Confocal analysis. The CAM was scanned on Laser Scanning Confocal microscope (Carl Zeiss NLO-710) at 550/570 excitation and emission spectrum for CM-Dil Y79 transplanted cells and 488/509 excitation and emission spectrum for eGFP Y79 transplanted cells. The 3D images were taken at different magni cations and the CAM was also scanned at different range of depths by using Z-stack method. Localized depths having strong uorescent signals were selected for image analysis by Image J software and parameters were adjusted using Zen 2010 software.
In-vivo imaging of embryos The whole embryos were analysed using IVIS Spectrum in-vivo optical imaging system (PerkinElmer) to track the presence of CM-Dil Y79 cells and eGFP Y79 cells. The embryos transplanted with CM-DilY79 cells and eGFP Y79 cells were exposed to excitation/emission spectrum of 550/570 and 488/509 for 30 seconds respectively. Embryos injected with PBS solution were used as controls. Image display analyses were performed using Living Image software (version 4.3.1, Xenogen, Alameda, USA). Data was obtained from the uorescent images by selecting the region of interest (ROI) and the number of photons emitted was measured as average radiance (photons/sec/cm 2 /sr).

Histological analysis
The nodules with the surrounding CAM tissues were rinsed with PBS and xed in 10% buffered formalin for 24 hours. The embryos were lifted from the egg and cleaned with PBS to remove the yolk and extra embryonic membranes. The embryos were decapitated and the embryonic organs (brain, eye, femur-bone marrow, and liver) were dissected and placed in formalin xative for 24 hours. Following xation, they were processed for routine histological analysis. Immunohistochemistry was performed with KI67 antibody (Roche Life Sciences) to detect Rb Y79 proliferating tumor cells using automated benchmark ultra IHC diagnostic system (Roche Life Sciences). The slides were then analysed under a light microscope and the staining was independently assessed by an experienced ocular pathologist (GKV).

Statistical Analysis
The quantitative data were stated as Mean±SEM and GraphPad Prism (GraphPad Software, La Jolla, CA) was used for unpaired Student's t-test and ANOVA with Tukey's Post-hoc multiple comparison tests. The representative images were analysed using ImageJ software. The experiments were repeated at least thrice with biological replicates and p<0.05 was considered as statistically signi cant difference between the groups.

Results
Analysis of CSCs in Rb Y79 cell line Using CD133 surface marker, ow cytometry analysis showed that 15.5 ± 0.32% of Y79 cells were CD133 lo (Fig. 1). The percentage of CD133 lo CSCs within the Y79 cell line is concordant with our earlier study. The cells were sorted for both the populations using magnetic activated cell sorting (MACS) with a purity of ≥ 90%. Cell viability was found to be over 85% for all the cell groups.

In-vitro CM-Dil uorescent labelling and GFP transduction e ciency
For xenograft studies, two approaches of uorescent cell tracking were utilized and their staining intensities were analysed in-vitro prior transplantation. Y79 cells, labeled with uorescent cell tracker CM-Dil dye were assessed for staining intensity and were observed to retain the dye for 17 days in culture ( Fig. 2a-c). However, uniformity of staining within individual cells were noted to reduce as the number of days in-vitro increased, possibly owing to cell proliferation.
The enhanced green uorescent protein (eGFP) labeled Y79 cells were tracked and assessed for GFP expression up to 2 weeks in culture by uorescence microscopy. The intensity of GFP uorescence varied cell to cell and majority of cells expressed GFP in-vitro ( Fig. 2d-f).
Embryo viability and assessment of tumor formation Y79 cells sorted using CD133 marker was examined for their tumorigenicity and metastasis in-vivo.
Brie y, 2 million total Y79, CD133 lo and CD133 hi cells were transplanted separately onto the abraded CAM and the embryos were sacri ced after 7 days. The survival percentages of CM-Dil Y79 transplanted embryos were 69.32% (n = 27), 69.14% (n = 28) and 79.71% (n = 41), and eGFPY79 cells transplanted embryo viability were 50% (n = 5), 90% (n = 9) and 90% (n = 9) for total Y79, CD133 lo and CD133 hi cells respectively. Both the labeled cells of total Y79 and CD133 lo groups formed pinkish-white raised wet perivascular nodules with feeder vessels on the CAM (Fig. 3a, c, e and g), whereas CD133 hi formed smaller plaque-like growths on the CAM ( Fig. 3i and k), as observed upon gross dissection. The tumor volume of eGFP Y79 CD133 lo nodules were signi cantly higher (40.44 ± 7.74mm 3 ) when compared to total (p = 0.02, 5.3 ± 1.01mm 3 ) and CD133 hi (p = 0.005, 2.56 ± 0.66mm 3 ) nodules (Fig. 5a). Confocal imaging con rmed the presence of growing tumor nodules within the CAM (Fig. 3). The uorescence intensity from the region of interest (ROI) analysis of CM-Dil Y79 cells transplanted CAM tissues showed that CD133 lo group (AUF = 6.37 × 10 7 ±7.7 × 10 6 ) (Supplementary Material: Video I) had higher localization of cells when compared to total Y79 group (AUF = 3.33 × 10 7 ±0.2 × 10 6 ) and CD133 hi group (p < 0.0001, AUF = 1.08 × 10 7 ±1.6 × 10 6 ) (Supplementary Material: Video II) (Fig. 5b). Similarly, the eGFP In-vivo imaging analysis of spontaneous metastasis to the embryo The IVIS-spectral images of the chick embryos transplanted with tumor cells on the CAM revealed uorescence signals in the cephalic, abdominal areas and within the bones of hind limbs. The control group (injected with PBS) did not show any uorescence signals within the embryo (Fig. 4a and e). The embryos transplanted with CM-Dil Y79 CD133 lo cells showed intense uorescence and more spread out signal compared to total Y79 and CD133 hi group (p = 0.0277) (Fig. 4b-d and 5d). Similarly, eGFP Y79 CD133 lo cells showed intense uorescence and more spread out signal compared to total Y79 (p = 0.0168) and CD133 hi group (p = 0.0049) (Fig. 4f-h and 5e). Further analyses of the three areas of the embryos showed that abdominal region had higher uorescence followed by the cephalic region and the limbs. The embryos transplanted with CD133 lo CSCs had increased metastases to the abdominal, cephalic and limbs when compared to CD133 hi non-CSCs of both the labeled cell groups.

Histological analysis of the xenograft formation and metastases
Histological examination of the tumor nodules showed areas of viable cells as well as necrotic tumor areas within the CAM along with surface ulceration (Fig. 6a-c). In contrast to the host cells within the CAM, the tumor cells were larger with high nuclear-cytoplasmic ratio and mitotic gures (Fig. 6b and c). The surrounding CAM also revealed focus areas of in ammatory cells and granulation tissue with giant cells (Fig. 6a). Metastatic tumor deposits were seen in the embryos transplanted with CD133 lo cells as pockets of in ltrating round cells next to the blood vessels within the embryonic liver tissues, brain and the eye (Fig. 6d-f). The metastatic Y79 cells within the liver, brain and eye showed immunoreactivity for Ki-67 indicating proliferation in the distant sites ( Fig. 6g-i).

Discussion
Retinoblastoma, a childhood ocular cancer, accounts for 3% of all childhood tumors and can be fatal if left untreated owing to the rapidly growing tumor cells within the developing mutated retina (24). Although several in-vitro and in-vivo animal model systems have been explored to study the underlying mechanisms of Rb tumorigenesis and metastasis, there is a challenge in investigating the tumor in a developmental microenvironment. In this study, we demonstrate the establishment of CE-CAM xenograft model for Rb Y79 cell line with formation of tumor nodules on CAM, that exhibit spontaneous metastasis to the embryo. Our study also provides preliminary evidence for CD133 lo , being the putative stem cells in Rb with higher potential to form tumor and metastases as documented by the in-vivo study using CM-Dil labeled and eGFP transduced Rb Y79 cells on chick embryo.
This novel work demonstrates that the CE model is a cost effective, time saving (procurement and maintenance), and a suitable model for visualization and examining Rb tumorigenesis and metastasis that could pave the way for exploring targeted therapy. This study provides in-vivo evidence and validates the results of our in-vitro studies that cancer stem cells properties are evident in CD133 lo population of RbY79 cells (18).
Animal models using transgenic and xenograft approaches have been well-established for Rb (13) (25). Most of them include the mammalian system in order to resemble the tumor microenvironment. However, a developmental model is di cult to recapitulate in this system due to many technical di culties (13). There is a dire need to simulate Rb tumorigenesis and metastasis within a developmental microenvironment in order to develop strategies for better therapeutic targeting. The CE-CAM, therefore, offers not only a temporal investigation but also provides a reliable tool for screening therapeutics in a short duration.  (15)(16)(25). Though the purpose of their study was towards cell line characterization and chemoresistance, the data does not provide any evidence for metastasis within the embryo. One of the possible reasons could be that the authors checked for invasion on E17/18, by which time the immune system of the embryo is observed to be fully functional and reactive (3). We concur with their observation of Y79 cells being capable of forming nodules on CAM and having the potential of spontaneous invasion within the CAM tissue. This observation is also similar to studies done on other solid tumors, such as glioblastoma, ovarian cancer, prostate cancer and uveal melanoma(7)(17).  (26). This could also serve as an important model to enhance or suppress Rb angiogenesis in tumor models due to its ease of visualization.
The spontaneous metastasis was observed in the cephalic, abdominal and limb regions of the embryo using the whole embryo in-vivo imaging and con rmed by histology in the sections of embryonic organs such as the brain, eye and liver. Rb is clinically known to metastasize to the brain by invading through optic nerve and to the bone and liver via haematogenous spread and, therefore, involvement of optic nerve, uvea and sclera are considered as one of the important prognostic factors to predict metastasis(10) (27). Our ndings also concur with the study done by Palmer and co-workers who investigated spontaneous metastasis of Human Epidermoid Carcinoma cell line (HEp3) to the chick liver and lung (28). Though the imaging of CAM and embryo was possible after dissection, there was an observed signal in the beak area which did not reveal any presence of tumor cells upon histological analysis. This could be due to auto ourescence due to the density of the bone and keratin present on the beak.
A further novel aspect of our study is the use of this model to study the tumorigenic and metastatic potential of CD133 lo CSCs within theY79 cell line. This is an important validating point as CD133 hi subset is projected as CSC in tumors of the brain (29), while the negative cells seem to harbour CSC properties in Rb (18). It would be logical to propose that while many CSC markers could be common to most tumors, some of them could vary in each tumor speci cally due to the lineage of the tumor or speci c aspects of differentiation of cells in that target tissue. For instance, CD133 is a marker of photoreceptor differentiation (30), hence their precursor cells are expected to be negative for this marker, which we have documented both in primary as well as Y79 cell line (18) (31).
The CE-CAM model is an ideal short-term in-vivo system to study different subsets of tumors and identify tumor initiating cells. We also demonstrate the use of a cell tracker dye CM-Dil, with comparison to eGFP labeled cells, in order to track Rb cells in-vivo. CM-Dil dye has been successfully used earlier in Rb zebra sh orthotopic xenograft model by Chen et al. who tracked the tumor cell invasion locally and along the optic nerve (32). However, considering that CM-Dil dye intensity could reduce with proliferation of cells (33), we also assessed the same using stable eGFP labeled Y79 cells. GFP labeled tumor cells were successfully used by several investigators to track in-vivo metastases within the chick embryos in several tumors(17)(34) (35). Our analyses showed that both the cell labeling strategies can be utilized for in-vivo tracking. To the best of our knowledge, this is the rst study that explores the use of in-vivo uorescence imaging of whole embryo for identifying spontaneous metastasis in a CE-CAM Rb xenograft model.
Use of other alternate imaging technologies such as in-vivo PET/CT has been demonstrated using radiotracers by Warnock et al in a glioblastoma xenograft model for studying tumor metastasis (36). Their studies showed the PET tracer uptake over time by the tumor within the embryo and were able to measure the tumor volume with improved accuracy using CT imaging. The CE-CAM, owing to the ease of visualization through the window from the eggshell makes it extremely suitable and convenient for tracking tumor cells in-vivo.
The clinical relevance of this study is the use of an inexpensive, easy to establish in-vivo model using cell lines and patient derived xenografts for exploring existing and novel therapeutics. However, there are certain limitations of the CE model and further studies are required to understand the cellular adaptations to avian microenvironment. The incubation period of about a week may not be suitable for slow growing tumor cells and it may be di cult to observe visible metastases in such cases. We believe that validating these results with early passage Rb primary cells and other Rb cell lines-WERI-Rb1, etc., in the future, would add more value to this in-vivo model. Use of CM-Dil dye for tracking rapidly growing cells may be a challenge owing to its loss with each successive tumor generations, which could possibly under represent the tumor load of the tissue. Parallel con rmation with GFP labeled Y79 cells helped overcome this limitation by demonstrating similar results in tumor formation and metastasis. Quantitative studies using human speci c gene expression assays could compliment these assays in future (28). Though the imaging of CAM and embryo was possible after dissection, the serial imaging of the entire live egg including the shell would be ideal for which different kinds of uorescent dyes and standardization of the software settings may be required, which we hope to address in our future studies.

Conclusions
In this study, we have successfully established an in-vivo tumor xenograft model for Rb CSCs using chick embryo, which simulates the developmental nature of Rb tumorigenesis. We showed evidence for tumor formation within the CAM and spontaneous metastasis to the embryo. CD133 lo cells showed higher tumor forming ability and metastatic potential when compared to total, CD133 hi Rb Y79 cells. Therefore, this model system can be used as a valuable in-vivo platform to understand Rb tumorigenesis, metastasis and pave the way for CSC speci c targeted therapies.      Imaging analysis of tumor nodules formed by total Y79 cells, CD133lo and CD133hi cells on the CAM and within the embryo (metastatic foci). a) CD133lo cells formed larger nodules on CAM when compared to total and CD133hi cells (*p=0.02, **p=0.005). Corresponding confocal analysis of tumor nodules on CAM showed that CD133lo cells displayed highest uorescence intensity when compared to the nodules formed by total and CD133hi cells labeled with b) CM-Dil (***p<0.0001) c) eGFP (**p=0.0018, ***p=0.0003). CD133lo cells also showed increased spontaneous metastasis to the embryo when compared to CD133hi cells labeled with d) CM-Dil (*p=0.0277) e) eGFP (*p=0.0168, **p=0.0049).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Video4.mp4 Video3.mp4 Page 21/21 Video2.avi Video1.mp4