Identication of an appropriate cell line for the renal cell carcinoma model

Background Although renal cell carcinoma (RCC) is known to be susceptible to ferroptosis, we found primary RCC cells showed resistance to ferroptosis and aimed to investigate a feasible candidate for an appropriate cell line for the RCC model. Methods Glutathione peroxidase 4 (GPX4) immunostaining was adopted in the RCC tissue microarrays. Normal human proximal tubule cells (HK-2) and RCC cell lines were used for the MTT assay, Western blotting, sphere-forming assay, and orthotopic injection of athymic Balb/c-nude mice. Results GPX4 immunostaining showed low intensity compared to the normal kidney, which coincided with the ferroptosis-susceptibility of RCC. Primary RCC cell lines (Caki-2, SNU-333, SNU-349, and SNU-1272) showed resistance to 5-uorouracil and a GPX4 inhibitor compared to the HK-2 cells and to metastatic RCC cells (Caki-1). The Caki-2 cells showed increased GPX4 and xCT, and the SNU-333 cells showed increased ferritin heavy chain (FTH1) compared to the other RCC cells. The Caki-2 cells showed increased aSMA, bronectin, vimentin, and SNAIL, and the SNU-333 cells showed increased aSMA, E-cadherin, and EpCAM. The Caki-2 cells showed increased Sox-2 and CD105, and the SNU-333 cells showed increased c-Myc and Lgr5. The Caki-1 and SNU-333 cells formed spheres in vitro and orthotopic RCC masses in vivo. The injected SNU-333 tumor only showed high intensities of CD10 and PAX8, consistent with the diagnostic criteria for RCC. Conclusions The primary RCC cell lines used in this study were more resistant to ferroptosis and 5-uorouracil, and the SNU-333 cells showed tumor-initiating capacities in vitro and in vivo. These results suggest that SNU-333 might be suitable as a orthotopic RCC model for future


Background
Renal cell carcinoma (RCC), originates from the proximal tubule [1], shows an increasing incidence of ~ 21.7 per 100,000 people over the last decade in the Republic of Korea [2]. RCC is not sensitive to conventional chemotherapy and is at least partly resistant to apoptosis [3], but known among the cancer cell lines to be susceptible to lipid repair enzyme GPX4-regulated ferroptosis [4]. Ferroptosis is a speci c form of programmed cell death, which is initiated by an increase in the labile iron pool and lipid reactive oxygen species (ROS) production [5][6][7]. The chelation of iron by deferoxamine rescues the experimental induction of ferroptosis [8], but inhibition of the cysteine uptake (system Xc − (xCT)) by erastin or sulfasalazine or the inactivation of GPX4 by (1S,3R)-2-(2-chloroacetyl)-2,3,4,9-tetrahydro-1- [4-(methoxycarbonyl)phenyl]-1H-pyrido [3,4-b]indole-3-carboxylic acid, methyl ester (RSL3) induces ferroptosis [5,7]. In this context, changes in iron pro les, such as serum iron, ferritin, and transferrin receptor (TfRC), have been suggested as successful markers for chemotherapy for metastatic RCC [9], and a variety of iron metabolism genes were signi cantly associated with the survival of RCC patients [10].
The resistance of RCC might be associated with the characteristics of cancer stem cells (CSCs). CSCs, which present characteristics reminiscent of normal stem cells (i.e., Oct4, Sox2, c-Myc, and Lgr5), show common features that include maintenance of the stem cell pool (self-renewal), tumorigenesis, and metastasis in vivo, and treatment resistance and recurrence [11,12]. According to the CSC hypothesis, conventional chemotherapies usually cannot eliminate the CSC pool [13], which may cause tumor recurrence. Several RCC markers were found to be speci cally expressed in CSCs, such as CD105, ALDH1, and Oct4, although CSC markers are not unique across tumor types [12,14]. In addition, tumor cells in the non-CSC subpopulation can spontaneously undergo epithelial-mesenchymal transition (EMT) and acquire a CSC-like phenotype and surface marker expression [15][16][17]. High expression of CD44 in RCC also correlated with recurrence and poor prognosis for 5-year overall survival [18].
In our previous report [19], yeast extract treatment decreased TfRC and increased ferritin in metastatic RCC cells, but decreased GPX4 in primary RCC cells. Altered iron metabolism may differentially contribute to growth inhibition or ferroptosis induction in various RCC cell lines since iron itself contributes to mutagenicity and malignant transformation, and the transformed malignant cells require high amounts of iron for proliferation [6,20]. Herein, we aimed to investigate a feasible candidate for an appropriate cell line for RCC model since most RCCs remain resistant cancers against various cell death-related therapeutic strategies.
Epidermal growth factor (EGF) and basic broblast growth factor (bFGF) were purchased from Gibco (ThermoFisher Scienti c, Seoul, South Korea), and B27 supplement was purchased from Invitrogen (Thermo Fisher Scienti c). The ALDH activity assay kit (colorimetric) was purchased from Abcam (Cambridge, MA, USA).

MTT assay
The effect of drugs on cell viability was evaluated by MTT reduction to its formazan product. The cells were seeded in triplicate wells of 96-well plates (2 x 10 3 cells/well), and treated with 5-uorouracil, RSL3, sulfasalazine, and deferoxamine at various concentrations. The cells were incubated for three days and then 10 µl of the MTT reagent (5 mg/ml in PBS) was added to each well, dissolved in DMSO for 15 min, and the MTT reduction was measured 3 h later spectrophotometrically at 595 using the absorbance at 620 nm as background by a VERSAmax microplate reader (Molecular Devices Korea LLC). The absorbance values obtained from the wells of the untreated cells represented 100 % cell viability and were used for comparison to the treated cells.

Sphere-forming assay
The sphere-forming assays were conducted as previously reported [21,22] with a slight modi cation. In brief, single-cell suspensions were plated in 96-well plates covered with poly-2-hydroxyethyl methacylate to prevent cell attachment, at a density of 5 x 10 2 cells in serum-free DMEM/F12 medium supplemented with 1% B27 supplement, 20 ng/ml EGF, and 20 ng/ml bFGF. After 15 days in culture, the number and the size of the spheres per well were acquired.

ALDH1 activity assay
The alde uor assay was performed according to the manufacturer's instructions (Abcam) [23]. Brie y, we labeled one TEST and one CONTROL tube for each sample. We lysed the cells and add activated ALDEFLUOR reagent to the lysed cell suspensions and 5 μl of DEAB to the CONTROL tube and then transferred to the TEST tube, mixed and immediately transferred 0.5 ml of the mixture to the CONTROL tube. The test samples were incubated at 37 °C, centrifuged, and resuspended in assay buffer. Finally, ALDH activity was measured spectrophotometrically at 450 nm using a VERSAmax microplate reader (Molecular Devices Korea, LLC). The absorbance values obtained from the HK-2 cells represent 100% activity and were used for comparison to the RCC cells.

Western blotting
To obtain the intracellular proteins, cultured cells were harvested in M-PER mammalian protein extraction reagent (Thermo Fisher Scienti c) including 1% protease inhibitor cocktail set III (EMD Millipore), 0.5% phosphatase inhibitor cocktail 2, and 0.5% phosphatase inhibitor cocktail 3 (both from Sigma-Aldrich).
The protein concentrations were assessed using BCA protein assay (ThermoFisher Scienti c) according to the manufacturer's instructions.

Electrophoresis of the protein in cell lysates was performed with the TGX Stain-Free FastCast™
Acrylamide Starter Kit (Bio-Rad Laboratories, Inc., Seoul, South Korea) using a Tris/glycine buffer system (Bio-Rad Laboratories) and transferred onto PVDF membranes as previously described [19]. The membranes were blocked with 5% skim milk for 1 h and then incubated with primary antibodies overnight at 4 °C. After washing, peroxidase anti-mouse or anti-rabbit IgG antibodies (Vector Laboratories, Inc., Seoul, South Korea) were applied for 1 h at room temperature. Next, Western Lightning Chemiluminescence Reagent (PerkinElmer, Inc., Daejeon, South Korea) was used to detect the proteins. Anti-GAPDH antibody was used as a loading control on the stripped membranes. The bands were quanti ed using AzureSpot analysis software (version 14.2; Azure™ c300; Azure Biosystems, Inc. Dublin, CA, USA).

Orthotopic RCC model establishment
All animal experiments were conducted in accordance with and approved by the Jeju National University Institutional Animal Care and Use Committee (2019-0015).
Athymic Balb/c nude mice (7-week-old males) were purchased from OrientBio (Seongnam, Republic of Korea). To induce an orthotopic RCC model, the mice were divided into ve groups (n = 5 mice/group) according to the different RCC cell lines (Caki-1, Caki-2, SNU-333, SNU-349, and SNU-1272), and anesthetized using an intraperitoneal injection of sodium pentobarbital (Entobar TM , Hanlim Pharm. Co. Ltd., Seoul, South Korea). Under aseptic conditions, a small longitudinal incision was made in the left lower back, and 2 x 10 5 cells in 20 µl of Matrigel (BD Biosciences, Becton Dickinson, Seoul, South Korea) were injected with a 27-gauge needle into the renal capsule. Thirty days following the injection, the mice were euthanized, the tumor masses explanted and xed for 24 h in 4% paraformaldehyde, and the xed tissues were dehydrated and embedded in para n wax. Sections (4 µm-thick) were cut with a microtome and prepared for hematoxylin/eosin (H/E) staining and immunohistochemistry.

Immunohistochemistry
Commercially available patient kidney tissue arrays (CL2) from SuperBioChips Laboratories (Seoul, Republic of Korea) were used for immunohistochemistry (IHC). GPX4 IHC was performed according to the supplier's protocol. Each array contains 50 tumor sections and nine matched control sections. As a primary antibody, we used a polyclonal rabbit anti-human GPX4 antibody (1:2000, Abcam). GPX4 was visualized by staining with a biotinylated anti-rabbit secondary antibody (1:200, Vector Laboratories), developed with diaminobenzidine, and counterstained with hematoxylin. Blind-scoring for the GPX4 immunostaining was performed by two independent observers who rst assessed the overall staining intensity on a single array. Next, we used a 4-step scoring scale to score the individual sections: 0 for absent or very low, 1 for mild staining, 2 for moderate staining, and 3 for strong staining.
For the primary antibodies, CD 10 and paired box G8 (PAX8), used for diagnosing RCC [24], we used the Benchmark XT staining system (Roche Diagnostics) on the orthotopic RCC masses, developed with DAB and counterstained with hematoxylin.

Statistical analysis
All data were compiled from a minimum of three replicate experiments. The data are expressed as mean values ± SD. Statistically signi cant differences (p < 0.05) were obtained using the Student t-test or oneway analysis of variance (ANOVA) with a Bonferroni post-hoc test (MS excel 2010).

Results
RCCs are known to be susceptible to ferroptosis, but primary RCCs show resistance to ferroptosis To con rm the susceptibility of RCC to ferroptosis, GPX4 immunostaining of human RCC tissues was done. The GPX4 immunostaining scores of the human RCC tissues were 2.89 ± 0.11 for normal, 1.98 ± 0.15 for Stage I, 2.14 ± 0.15 for Stage II, 2.14 ± 0.14 for Stage III, and 1.75 ± 0.25 for Stage IV. The staining intensity was signi cantly lower in RCC tissue compared to normal human kidney (p < 0.001), in which metastatic stage IV was the lowest, without signi cance (Fig. 1A).
To con rm the known characteristics of RCC, the effects of an anti-cancer drug (5-FU), inducers of ferroptosis (RSL3 and sulfasalazine), and an inhibitor of ferroptosis (deferoxamine) on cell viability were assessed using the MTT assay (Fig. 1B).

Ferroptosis-resistant Caki-2 and SNU-333 cells show increased levels of CSC or tumor markers
As the primary RCC cells showed resistance to ferroptosis and an anti-cancer drug compared to the HK-2 and metastatic RCC cells, ferroptosis-related markers were rst assessed in the RCC cell lines (Fig. 2A). Surface markers were also assessed in the RCC cell lines (Fig. 2C). Compared to the HK-2 cells, the Caki-2 cells showed signi cantly increased levels of CD24 (p < 0.05/ 0.0123) and CD44 (p < 0.01/ 0.0036). The others were not changed considerably or decreased.
ALDH activity was assessed using the HK-2 cells as a control. The RCCs did not show any signi cant increase but were considerably decreased, except for the Caki-2 cells. The Caki-2 (p < 0.001) and SNU-333 (p < 0.01; 0.0017) cells were signi cantly increased compared to the Caki-1 cells.

Tumorigenic capacity was further con rmed in Caki-1 and SNU-333 cells in vivo
To establish an orthotopic RCC model similar to human RCC, RCC cells were orthotopically injected into Balb/c nude mice (Fig. 4A). Among the RCC cell lines, the Caki-1 and SNU-333 cells successfully formed tumor masses.
We evaluated the morphological features by H/E staining, and the expression of markers with diagnostic relevance, such as CD10 and PAX8 (Fig. 4B). The Caki-1-bearing tumors consisted of spindle tumor cells with marked nuclear pleomorphism and desmoplasia and showed focal CD10 and PAX8 immunopositivity. The SNU-333-bearing tumors consisted of epithelioid tumor cells with clear or granular cytoplasm and showed strong CD10 and PAX8 immunopositivity. Based on the H/E and CD10 and PAX8 immunostaining results, the SNU-333 cell line showed the typical staining patterns of human RCC, whereas it was unclear if the Caki-1 cells originated from RCC. In addition, the GPX4 immunostaining was considerably weaker in both orthotopic tumor masses compared to the normal murine kidney.

Discussion
We found that among primary RCC cell lines, the SNU-333 cell line showed resistance to ferroptosis and was suitable for an orthotopic RCC model with tumorigenic capacities.
First, we con rmed that metastatic RCC was resistant to 5-FU, but susceptible to ferroptosis based on immunohistochemistry and MTT assay, as previously reported [3-5, 7, 8]. Interestingly, primary RCCs, especially the Caki-2 and SNU-333 cells, showed resistance to ferroptosis and 5-FU compared to metastatic RCC and/or the normal proximal tubule cells. The resistance to ferroptosis was related to GPX4, rather than xCT, and with deferoxamine (Fig. 1).
The Caki-2 cells showed signi cantly increased levels of GPX4 and xCT compared to the other cell lines, which may be the basic mechanism for the resistance to ferroptosis. However, the SNU-333 cells did not show considerable changes with GPX4 or xCT (Fig. 2A). The GPX4-mediated resistance to ferroptosis in the Caki-2 cells might be reinforced by recent studies [25,26] reporting that a GPX4-dependent cancer cell state underlies the clear cell morphology and confers sensitivity to ferroptosis. The clear cell morphology of RCCs is caused by highly active lipid and glycogen synthesis and deposition and marked by clear cytoplasm in histological analyses, which is hypersensitive to GPX4 knockdown [25]. The clear cell histotype shows vulnerability to ferroptosis via GPX4 dependency [26]. In addition, the Caki-2 cells showed signi cantly increased levels of TfRC and FTH1 and the SNU-333 cells showed increased levels of FTH1, which could be another mechanism for the resistance. Iron is usually dysregulated in cancers. Cancer cells often show higher levels of TfRC, down-regulation of FPN, and lower levels of ferritins, which lead to an increased intracellular labile iron pool [6,20]. Although it is mostly utilized for tumor growth by cytosolic and mitochondrial iron enzymes, excessive amounts can promote increased oxidative stress via ROS accumulation and result in ferroptosis. The Caki-2 and SNU-333 cells showed paradoxically increased levels of ferritins (FTH1), which is comparable to a decreased intracellular labile iron pool and may lead to the resistance to ferroptosis.
Other resistance proteins, including CSC markers and tumor markers, were analyzed ( Fig. 2B and 2C) because the primary RCCs also showed resistance to 5-FU compared to the metastatic RCC and/or HK-2 cells. Although the CSC markers are not unique across tumor types [12], normal stem cell markers (Oct4, Sox2, c-Myc, and Lgr5) and CSC markers (CD105) might share the characteristics of stem cells in RCCs [11,12]. The Caki-2 cells showed increased levels of Sox2 and CD105, whereas the SNU-333 cells showed increased levels of c-Myc and Lgr5 compared to the other RCC cell lines. In addition, the Caki-2 cells only showed increased levels of CD24 and CD44. In the cell viability assay, the Caki-2 and SNU-333 cells showed stemness, which is associated with resistance and/or recurrence.
As non-CSCs can acquire a CSC-like phenotype during EMT [15,16] and EMT-induced cells form spheres [17], a sphere-forming assay was conducted in the RCCs. As the Caki-1 cell line was reported to form more spheres compared to Caki-2 cells [22], the Caki-1 cells were considered a positive control. The SNU-333 cells only formed big renospheres, which were responsible for the signi cantly higher levels of EGFR ( Fig. 3A and 3B). Mizumoto et al. [27] suggested that acquired resistance to sunitinib in RCC cells may be related to the activation of EGFR, and thus, increased EGFR in the SNU-333 cells might have a role in resistance, as well as sphere formation.
Compared to the Caki-1 cells, the Caki-2 and SNU-333 cells showed signi cantly increased ALDH1 activity (Fig. 3B), consistent with previous reports that CSCs could be identi ed with ALDH1 activity [12,14]. Since the sphere-forming assay is anchorage-independent in culture [12], the EMT properties were examined (Fig. 3C). The RCCs showed signi cantly increased levels of αSMA, the Caki-2 cells showed increased bronectin, vimentin, and SNAIL, and the SNU-333 cells showed increased E-cadherin and EpCAM. Taken together, the Caki-2 cells might show resistance via stemness and increased EMT-like properties. The SNU-333 cells also showed resistance via stemness but did not show EMT-like properties.
Finally, to evaluate the tumorigenic potential of RCCs in vivo, cancer cells are transplanted into immunocompromised mice. The Caki-1 and SNU-333 cells successfully formed orthotopic tumor masses, but the Caki-2 cells did not form tumors (Fig. 4A). The orthotopic masses were rarely immunostained with GPX4 while the normal tubules were stained, which was similar to the results of the human RCC tissue microarray. The histologic and immunohistochemical results revealed that the SNU-333 cells had characteristics consistent with the diagnostic criteria of human RCC (Fig. 4B). Metastatic RCC cells have been used to induce an orthotopic RCC model [24,28,29]. However, the established orthotopic RCC model might not be a proper model for ferroptosis-related research because metastatic RCC cells are known to be susceptible to ferroptosis. Using the proper cell line is crucial to furthering our understanding of RCC.
Cancer is very complex and shows heterogeneity, and thus, choosing the right cell line for RCC research based on the molecular background has been suggested [30]. In addition to the well-known RCC cell lines (Caki-1 and Caki-2), we also used Korean RCC cell lines (SNU-333, SNU-349, and SNU-1272) primarily focused on the primary RCCs because we believe ferroptosis-resistant orthotopic RCC models are needed for further studies.

Conclusions
We clearly showed that the primary RCC cell lines were more resistant to ferroptosis than the metastatic RCC cells, in which the SNU-333 cell line showed resistance via iron metabolism and stemness, and further tumor-initiating capacities in vitro and in vivo. These results suggest that the SNU-333 cell line was the most suitable for a orthotopic RCC model for further research including the era of ferroptosis, drug resistance, and CSCs.

Consent for publication
Not applicable

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.