Spheroid Culture Models Imitating the Tumor Microenvironment of Renal and Melanoma Cancer

Background: Tumor development studies need higher degree of adaptation to the cancer cells specic mechanisms of plasticity in connection with their microenvironment. It appears that standard two-dimensional (2D) cultures and gas composition are not relevant to the real cancer environment. Existing three-dimensional (3D) models are often requiring dicult sophisticated conditions. Two distinct cancer models were chosen: melanoma (B16F10) and kidney cancer (RenCa) for their different biological reactions in terms of cancer progression. We proposed 3D method which brings simplicity, reproducibility and remarkable mimicry of the in vivo tumor reactions. We characterize and compare the 3D models with standard 2D culture in normoxic and hypoxic condition, depending on presence of hypoxia related genes/proteins and aggressiveness mechanism (EMT- Epithelial Mesenchymal Transition and CSC- Cancer Stem Cells). We validate proposed 3D method by comparing it with in vivo obtained tumors.

universal protocol allowing sphere formation. We characterize the obtained 3D structures compared to standard monolayer culture with a special emphasis on EMT and CSC related markers. The goal of the study was to assess whether cells undergo similar molecular changes in spheroids as in the tumor mass in vivo. Such well-characterized models may contribute to obtaining more reliable data in easy way, which will bring research tools to provide data closer to the possible translation into clinical practice.
Material And Methods
1.2. Three-dimensional cell cultures by spheroid formation 500 cells resuspended in 20 µL medium supplemented with 0.25% methylcellulose (vol/vol in medium), (R&D Systems, USA) were seeded as a single hanging drop, on the cover of culture plate in standard culture conditions (5% CO 2 , 21% pO 2 ). After 72h drops were individually transferred to a 96-well plate previously covered with 1% agarose dissolved in double distilled water (w/vol, VWR, Belgium), and 50 µL of fresh medium (supplemented with 0,25% methylcellulose, vol/vol) were added concomitantly. Cells were allowed to expand for another four days period.

In vivo tumor implantation and development
The experiments were carried out with female mice of the C57Bl6 strain -syngeneic model of melanoma and BALB /c -syngeneic model of kidney tumor. Cancer cell suspensions: RenCa − 100,000 cells; B16F10-200,000 cells in 100 µL of PBS mixed with Matrigel™ (50%, vol/vol) (Corning, USA) and maintained at 4 C, were subcutaneously administered to the animal's thigh. The renal cancer model experiment was terminated after 22 days, while the melanoma cancer model experiment was completed 18 days after the injection of the tumor cells. Experimental groups consisted of 3-4 mice kept in one cage, and experiments was performed using two separate sets of animals. All experiments were conducted after obtaining approval for procedures from Second Warsaw Local Ethics Committee for Animal Experimentation (no. WAW2/76/2017) and following Directive 2010/63/EU regulations.

Western blot
The two-dimensional cell cultures were washed twice with ice-cold PBS and collected by scraping.

Flow cytometry detection of intracellular enzymes activity and cell surface antigens
Cell suspensions from 2D cultured cells were obtained using Accutase solution (Biolegend, USA). Individually cultured spheroids were treated with 0,1 mg/ml collagenase type II (Gibco by Life Technologies) and incubated at 37° C with simultaneous stirring until the spheres disintegrated into single cells.

Aldehyde dehydrogenase-1 activity assessment
The ALDH1 enzyme activity was measured intracellularly with the help of Alde uor uorescent reagent. It is a BODIPY-aminoacetaldehyde uorescent ALDH substrate able to diffuse into intact and viable cells. Converted into BODIPY-aminoacetate by ALDH in cells, the negatively charged product is retained inside the cells increasing uorescence for ow cytometry analysis (λex max = 493 nm, λem max = 588 nm). Protocol was followed according to the manufacturer's recommendations (STEMCELL Technologies).
Brie y, 0.5 x 10 6 cells were suspended in 500 µL of Alde uor assay buffer. Assay samples received 2,5 µl of the Alde uor reagent, mixed and immediately 250 µl of the suspension was taken out and put in the control tube with the DEAB (N, N'-dimethylaminobenzaldehyde) inhibitor 2,5 µL. Cells were further incubated for 45 minutes at 37°C in the dark. After washing cells, viability was determined by staining with 7-Amino-Actinomycin D (BD Biosciences) a membrane permeant dye which is excluded from viable cells, it binds to double stranded DNA of dead cells (λex max = 546 nm, λem max = 647 nm). All samples were analyzed using a Calibur ow cytometer (from Becton Dickinson, Sunnyvale Ca, USA). Cells treated with inhibitor were used to set the gate for ALDHneg population and positive cells were analyzed.

Surface antigens detection by uorescent antibodies
The evaluation of the expression of CD105 and CD133 was determined using commercially available antibodies: PerCP-eFluor710 conjugate rat-anti-mo

Reverse Transcriptase-Polymerase Chain Reaction
RNA samples from in vitro and in vivo cultures were isolated using the commercially available RNeasy Mini Kit (Qiagen, Hilden, Germany), according to the manufacture protocol. The obtained cell lysates were either stored at -80 ° C or RNA isolation was started. The isolated RNA was then puri ed using the TURBOTM DNA-free Kit (Thermo Fisher Scienti c, USA) according to the manufacturer's protocol. The concentration and purity of RNA were determined by measuring the absorbance in the wavelength range: 230, 260 and 280 nm. RNA integrity was assessed using a uorometer based on the RIN (RNA integrity number) coe cient (acceptable values 7-9). After obtaining high purity mRNA, a reverse transcription process, using a commercially available High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scienti c, USA) was performed according to the manufacturer's instructions. The resulting singlestranded cDNA was used as real-time polymerase chain reaction templates.
6. Next Generation Sequencing (NGS) NGS was performed outsourced service by Lexogen GmbH (Vienna, Austria) with NextSeq 500 system (Illumina, USA). Brie y, RNA was extracted and puri ed using methods described in Reverse Transcriptase-Polymerase Chain Reaction section. RNA BR Assay Kit (Thermo Fisher Scienti c, USA) was used to measure total RNA concentration and Qubit RNA IQ Assay Kit (Thermo Fisher Scienti c, USA) to measure RNA integrity and quality. To obtain cDNA libraries 1µg of total RNA with IQ > 8,5 was used. Preparation of libraries was performed according to manufactures instructions as followed: The NEBNext® Poly(A) mRNA Magnetic Isolation Module -mRNA isolation and fragmentation (New England Biolabs) Ultra RNA Library Prep Kit for Illumina -cDNA Library synthesis, New England Biolabs); NEBNext Multiplex Oligos for Ilumina -adding Adaptors (New England Biolabs). All steps were performed along with the puri cation of the reaction products using NEBNext Sample puri cation Beads (New England Biolabs). Libraries were quali ed on Bioanalyzer (Agilent, USA), with High Sensitivity DNA Kit (Agilent, USA). Differentially expressed genes (DEGs) were classi ed according to -1,5 < logFC > 1,5 and p-value < 0.5. Gene symbols were translated into UniProt accession numbers using the UniProt Knowledgebase (UniProtKB). Protein networks were retrieved from the STRING (Search Tool for the Retrieval of Interacting Genes) database using the list of protein accession numbers as a query and then analyzed using the Cytoscape software.
Microscopy Imaging 3D cultures were performed according to the above-mentioned procedure. Imaging to assess live and dead cells in the spheres was performed after labelling by calcein acetoxymethyl ester (Calcein AM, Biolegend) which is hydrolyzed by cellular esterases to give uorescent calcein in the cytoplasm of live cells and propidium iodide (Exbio, Czech Republic) DNA intercalating agent which is highly uorescent after intercalation. Brie y, after adding the dyes spheroids and cells grown in two-dimensions were incubated for 15 minutes in the dark at room temperature. Images were acquired using a Zeiss AxioObserver.7, uorescence, and inverted microscope (5X magni cation) and analysis performed with the Zen 2.6 blue edition software (Zeiss, Germany).

Statistical Analysis
The results are shown as a mean +/-SEM. All experiment was performed at least in 3 biological replicates. All statistical analyses were performed using GraphPad Prism 9.0 software (RRID:SCR_002798). Depending on the gaussian distribution we performed One-Way Analysis of variance (ANOVA) with Tukey post-hoc test or Kruskal-Wallis with post-hoc Dunn's test.

Results
Optimization, characterization, and comparison of 3D model with 2D culture models.

Establishment of 3D cultures
Irregular cell aggregates were observed especially for RCC model after using standard culture methods, e.g.: hanging drop in regular culture medium ( Supplementary Fig. 1A). Consistent spheroids from renal and melanoma murine cancer cells in their long-term culture in normoxic conditions were obtained by combining the hanging drop method with culture in semi-solid matrix (Fig. 1B). 500 cells in a 20 µL drop were seeded for both cell lines to compare the proliferative capacity of the cells during the sphere's formation. Spheroids were cultured for seven days in total: three days in hanging drops and then four days in agarose coated bottom plate in standard culture medium supplemented with methylcellulose (Fig. 1A). After three days of culture in hanging drops irregular cell aggregates were observed for both cell lines: RenCa and B16F10 (Supplementary Fig. 1B). Round and regular spheroids were obtained after seven days of culture were completed ( Fig. 1. B.). As measured by diameter, the RCC cells at the end of culture formed smaller spheroids (~ 400µm) as compared to spheres (~ 600 µm) (Fig. 1C) from B16F10 despite the initial number of seeded cells was the same. This suggests a slower growth of RenCa cells in 3D. Using ow cytometry, we determined the cell viability in 2D, and 3D cultures conducted in normoxic and hypoxic conditions. Higher proportion of viable cells for both cell lines was recovered from twodimensional cultures, than from spheroids in both oxygen tension conditions (Fig. 1D). Microscopic observation of spheres stained by calcein and propidium iodide showed a necrotic core of 3D structures in normoxia ( Fig. 1.E). The cells from spheroids cultured in hypoxia showed a very low viability ( Fig. 1.D; Supplementary Fig. 1C) and were smaller than spheres cultured in normoxia. It suggests that hypoxia which is known to develop in the center of the sphere may be in uencing the viability of cells recovered from 3D structures where cells have an impaired access to oxygen (29). Consequently, spheroid culture in additional/external hypoxic conditions may cause an intense cellular stress drastically reducing the cell survival. As very few cells were recovered from hypoxia cultured spheres, we decided to conduct further research on cells cultured under normoxic condition which seems more appropriately reconstitute natural development of a tumor.
3D culture condition induces melanoma cancer cell cycle arrest at the G0/G1 phase Spheroid culture affected the cell cycle distribution in both cell lines with no direct effect of hypoxia only ( Fig. 2A-D). In the case of B16F10 cells, 3D culture caused accumulation of cells in G0/1 phase with concurrent reduction of S phase mostly ( Fig. 2A, C). On the contrary, spheroid recovered RenCa cells were enriched in proliferating, G2/M cells with non-signi cant drop of cells in both G0/1 and S phase (Fig. 2B,  D). As activation of p53 tumor suppressor can lead to the cell cycle arrest (33) we checked the expression of p53 and its inhibitor mdm2. For both cancer models we observed a similar downregulation of p53 expression by both hypoxia and 3D formation, as compared with 2D normoxic culture (Fig. 2E). Expression of mdm2 in melanoma cancer for 2D hypoxia and 3D was down regulated as compared with 2D normoxia, with a signi cantly stronger effect due to the spheroid type of culture (Fig. 2E). For RCC cultures no statistically signi cant changes of mdm2 expression were observed (Fig. 2E). 3D melanoma cancer model induces hypoxia.
Presence of necrotic core in 3D structures of RCC and melanoma models after staining with calcein and propidium iodide suggested the induction of hypoxia in the middle of the spheres (29). We checked several hypoxia related genes/proteins such as: HIF-1α, vegf-a, and VHL ( Fig. 3. A, B). In 3D melanoma model we observed upregulation of hif-1α expression and a tendency for higher protein level as compared to both 2D normoxia and hypoxia (Fig. 3. A, B). Vhl expression was not affected by any of tested culture conditions. Vegf expression increased both in 3D and 2D culture in hypoxia, although in the later condition, no HIF-1α upregulation occurred (Fig. 3. A, B). In RCC model we observed an opposite tendency; 3D culture downregulated hif1α, vhl and vegf-a genes, however hypoxia alone tended to increase HIF-1α protein and vhl and vegf gene expression (Fig. 3. A, B). Global gene expression analysis with NGS was performed for RCC model to identify hypoxia gene signatures in 3D as compared with 2D cultures (Fig. 3. C). String protein networks indicated upregulation of Endothelial PAS Domain Protein 1 (Epas1) for 3D culture as compared with both 2D normoxia and hypoxia culture conditions. Also, upregulation of matrix metalloproteinase-10 (Mmp10), Matrix metallopeptidase 13 (Mmp13) and Lysyl oxidase (Lox) was observed as compared to 2DN culture ( Fig. 3. C).

Renal cancer spheroids show upregulation of a cancer stem like cell -CSC-population
Upregulation of cells arrested in G0/G1 phase may suggest the presence of CSC (34), and hypoxia was shown to induce selection of CSC in cancer foci (29). Therefore, the levels of several potential CSC markers such as: ALDH1, CD133 and CD105 were evaluated (Fig. 4). In the case of B16F10 cells with G0/G1 arrest, we could not observe CSCs induction; CD105 positive cells dropped in 3D culture while CD133 positive cells remained unchanged (Fig. 4D, E). Similarly, ALDH1 was not signi cantly altered in those cells (Fig. 4A, B). Surprisingly, in spheroid growing RCC cells, that were characterized by increased G2/M accumulation, a strong increase of ALDH1 protein level and activity were observed (Fig. 4A, B). Analysis of String protein networks also indicated the upregulation of Aldehyde dehydrogenase 2 (ALDH2) expression in 3D (Fig. 4C). Additionally, in RenCa cell line an increased number of CD133 positive cells in 2D cell cultures in hypoxia and in 3D cultures was shown, whereas no statistically signi cant changes for CD105 levels were observed ( Fig. 4. D, E). Remarkedly, an opposite tendency was displayed by B16F10 cells.

3D induces EMT in melanoma model
Another mechanism of cancer aggressiveness is EMT (35). We tested whether this process is induced by spheroid culture by assessing the main EMT markers: Vimentin, N-cadherin, and B-catenin (Fig. 5). In the B16 F10 melanoma model, the upregulation of N-cadherin, Vimentin and B-catenin was observed in spheroids. In the RenCa model B-catenin in spheroids was downregulated and N-Cadherin could not be detected in any tested culture conditions (Fig. 5). Changes for vimentin expression in RCC models were not signi cant, however tendencies were opposite than those observed in melanoma model (Fig. 5), con rming the distinct reactions uncovered by the 3D mode of culturing the two cancer cell types.
3D spheroid formation as an alternative model to in vivo murine tumors To assess for the signi cance of spheroid formation to mimic some tumor characteristics, we compared the levels of above tested markers in 3D models to in vivo tumors, comparatively to standard monolayers type of culture in normoxia (2DN).

3D cell culture models induce similar pattern of expression of cancer suppressor genes as in vivo growing tumor
Melanoma and RCC spheroids and their corresponding tumors, showed similar level of downregulation of p53 expression as compared to monolayer cultured cells (Fig. 6). While, in the melanoma model a similar downregulation pattern was observed for mdm2 expression, this was not the case in the RCC model as mdm2 expression was not signi cantly altered (although tended to increase) (Fig. 6).

3D cultures and in vivo tumors show a similar expression pattern of proteins / genes associated with hypoxia
Melanoma spheroids and tumors displayed a similar upregulation of hif1α and vegf-a genes expression but the HIF-1 protein increase in spheroids was not as strong as in tumors ( Fig. 7A-C). A statistically signi cant downregulation of vhl was observed only for its transcripts in tumor (Fig. 7A). On the other hand, there was no effect on the corresponding protein level both in spheroid and tumor (Fig. 7B, C). For RCC we observed that 3D and tumors show similar changes in expression when compared to monolayer, but tendencies were opposite to melanoma model: both in tumor and 3D a downregulation of hif1α, vhl and vegf-a was observed (Fig. 7A-C). On the level of proteins, VHL increased in tumors only; in spheroid culture this upregulation did not reach a statistical signi cance (Fig. 7B, C).

RCC spheroids and tumors show a similar expression pattern of markers associated with stemness
Previously observed ALDH1 induction in renal cancer spheroids was also seen tumors; ALDH1 protein level was increased similarly in 3D culture as compared to monolayer (Fig. 8). Also, melanoma tumors were characterized by a very high expression of this protein, although in this model, spheroid culture could not upregulate ALDH1 signi cantly (Fig. 8).

Melanoma and RCC spheroids and tumors show a similar expression pattern of markers associated with EMT
Previously observed ß-catenin induction in melanoma and downregulation in renal cancer spheroids was also observed tumors (Fig. 8). The same tendencies for vimentin expression in 3D and the tumor was also observed for both models, were we observed upregulation of this protein in melanoma and downregulation in RCC model (Fig. 8).

Discussion
To increase validity and success rate of cancer research, it is necessary to handle parameters that mimic the most important features of the disease. Monolayer cell cultures provide useful and homogenous model, however many studies showed that 3D culture models display features that are closer to in vivo tumor characteristics (29,30). 3D models using B16F10 cells do exist for murine melanoma, but there are no well-optimized models for murine kidney cancer using RenCa cells for studies (29,(36)(37)(38). The present work performed long long-term 3D cultures allowing comparative models for melanoma and renal cancer and characterized them in terms of response to hypoxia, CSC selection and EMT induction. Additionally, we demonstrated the validity of the spheroid type of culture to re ect the mechanisms observed in tumors in vivo. The designed method does not require any special equipment nor expensive laboratory consumables and pipetting was reduced to minimum. We exploited 96-well format that can be easily used for cost-and time-effective drug screening but provide su cient cellular material for molecular experiments (western blot, qPCR or ow cytometry can be performed using 96 spheroids for each technique) (Fig. 1A).
It was shown that size of the sphere in uences the cellular processes induced within. The dimensions of spheroids are usually correlated with seeding density (32), however in our study we point out also intrinsic features of used cells for the outcome of 3D culture. The study, although restricted to two murine cancer cell type to reduce the variability of spheroid formation, con rms the universality of the method. Indeed, the present modi cation of hanging drop method was e ciently inducing 3D formation by both cell lines, despite the poor RenCa ability for sphere formation (Supplementary Fig. 1A, B) (Fig. 1C). RCC spheroids were signi cantly smaller than melanoma's, although initial cell number was same for both cell lines. Such differences may result from different proliferation capacity of cells, resulting is a less apparent necrotic core (Fig. 1E). Ability to form spheroids corelates with cancer stem-like cells characteristics of cell lines (39,40). We showed that melanoma cell line is initially characterized by higher levels of ALDH1 (protein / ALDH1 activity) and CD133 positive cells than RenCa cells (Fig. 4) what relates with the higher capacity of B16F10 cells to form spheroids.
Cells recovered from spheroids were less viable probably due to a necrotic core formed in 3D structures as observed by microscopy (Fig. 1E). This suggests that hypoxia develops in the center of the sphere (41). Indeed, in the melanoma spheres, we observed upregulation of Hif1α (42), inducer of the main proangiogenic factor VEGF-A, (Fig. 3A). In RCC 3D model we observed downregulation hif1α, vhl and vegf-a expression (not statistically signi cant), which suggest that hypoxic processes were not effectively induced probably because spheres were smaller (Fig. 3A, B). However, the same pattern of gene expression was observed in the subcutaneous tumor obtained in vivo (Fig. 7). This shows that 3D culture induces cellular mechanisms that are cell-speci c, in addition to the physical parameters that universal and due to spheroid architecture. This strongly support the conclusion that, in many aspects, spheroids are su cient to model tumor physiology.
Further characterization of RenCa spheroid model revealed upregulation of Epas1 (Hif2α) transcript (Fig. 3C). Studies using in vivo and in vitro models indicated that HIF-1α acts as a tumor suppressor, whereas HIF-2α has oncogenic potential (12). Together with increased expression of Hif2α, increase of Mmp10, Mmp13 and Lox expression was observed (Fig. 3C). These proteins are associated with cancer progression and metastasis in various types of tumors (43)(44)(45). Taken together that RCC spheroids have population of aggressive, highly proliferative cells that are selected upon 3D formation. Indeed, RenCa spheroids were enriched in proliferating, G2/M cells (Fig. 2), cancer stem-like cells (upregulation of ALDH1, ALDH2 expression and CD133 + increased population) (Fig. 4) and metastasis linked proteins.
In contrast melanoma spheroids were characterized by a reduced percentage of proliferating cells (G2/M) with a simultaneous G0/G1 phase shift in the cell cycle, despite p53 downregulation together with its regulator mdm2 (Fig. 2). G0/G1 phase arrest also suggests the presence of quiescent cancer stem cells (48), although other markers of CSC phenotype, as ALDH1 and CD133, were not induced in B16F10 spheroids (Fig. 4A, B, D) or down regulated as CD105 positive cells (49) (Fig. 4E). CD105 downregulation in these cells causes reduction of proliferative capacity, and promotes spheroids compaction (21). This may explain cell cycle arrest in G0/G1 phase of cells in the melanoma spheres ( Fig. 2A-D), and shape differences of both cell models (Fig. 1B, E). Although we do not observe any changes in CSC markers expression in melanoma spheres, EMT markers were upregulated (Fig. 5). The use of the 3D type of culture uncovered the distinct molecular mechanisms set by the melanoma and the RCC cells to reach a higher aggressiveness. As EMT is induced by hypoxia in several types of cancer (14), this might have not been reached by RenCa spheres in contrast to B16F10. Although RenCa spheroids are smaller and less hypoxic, their structure allowed a positive CSC selection and proliferation. B16F10 form bigger and hypoxic spheres which possess EMT features. Our result correspond with what was shown previously, that melanoma intra-tumor heterogeneity does not rely on the CSC hypothesis, but rather on EMT features (50).
To validate the developed models for mimicking the most important features of cancer, we compared the levels of the above tested markers in 3D models with in vivo tumors. Melanoma and RCC spheroids showed a similar pattern of regulation of p53 oncogene and its regulator -mdm2 (Fig. 6), which may indicate similar proliferative capacities of tumors and spheres (51). Hypoxia markers found in the melanoma spheroids were expressed at similar levels but the HIF1 protein, for which increase in spheroid lower than in the tumor (Fig. 7). This indicates that melanoma 3D culture is indeed able to mimic the features of tumor hypoxia. RCC spheres also show similar changes as observed in the corresponding tumor, although these two cancer types were often oppositely modulating molecular pathways (Fig. 7). The 3D type of cultures clearly bring the conditions much closer to in vivo ones. Indeed, the induction of CSC in RCC model was as e cient as in the tumor con rming the similarities between 3D and the tumor, in contrast to the melanoma where the CSC induction was lower than in the tumor but in both cases the EMT processes were largely mimicked.

Conclusion
In our studies, we observed that 3D models can mimic some features of the tumor such as hypoxia and CSC induction, adequately to the characteristics of corresponding in vivo tumor mass. Spheroid models of renal and melanoma cancers induce different mechanisms, related to the progression of cancer. RCC model displayed induction of CSC characteristic whereas, melanoma spheres contribute to EMT-related mechanism. Changes in cellular processes in these two cell lines in spheroids compared to standard monolayer were not universal, however were reproduced in the in vivo subcutaneous model, proving the validity of spheroid cultures to model cancer physiology.

Competing interests:
The authors declare that they have no competing interests.    Statistical analysis was performed by One-Way ANOVA/Tukey test or Kruskal-Wallis test/ Dunn's test -*p<0,05, N≥3.  Comparison of proliferation capacity regulation genes. Relative expression of mdm2 and p53 for B16F10 and RenCa cells. All data normalized to 2DN expression. β-Actin served as a quantitative internal control.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.