Isolation and characterization of renal cancer stem cells from patient-derived xenografts

As rapidly developing patient-derived xenografts (PDX) could represent potential sources of cancer stem cells (CSC), we selected and characterized non-cultured PDX cell suspensions from four different renal carcinomas (RCC). Only the cell suspensions from the serial xenografts (PDX-1 and PDX-2) of an undifferentiated RCC (RCC-41) adapted to the selective CSC medium. The cell suspension derived from the original tumor specimen (RCC-41-P-0) did not adapt to the selective medium and strongly expressed CSC-like markers (CD133 and CD105) together with the non-CSC tumor marker E-cadherin. In comparison, PDX-1 and PDX-2 cells exhibited evolution in their phenotype since PDX-1 cells were CD133high/CD105-/Ecadlow and PDX-2 cells were CD133low/CD105-/Ecad-. Both PDX subsets expressed additional stem cell markers (CD146/CD29/OCT4/NANOG/Nestin) but still contained non-CSC tumor cells. Therefore, using different cell sorting strategies, we characterized 3 different putative CSC subsets (RCC-41-PDX-1/CD132+, RCC-41-PDX-2/CD133-/EpCAMlow and RCC-41-PDX-2/CD133+/EpCAMbright). In addition, transcriptomic analysis showed that RCC-41-PDX-2/CD133− over-expressed the pluripotency gene ERBB4, while RCC-41-PDX-2/CD133+ over-expressed several tumor suppressor genes. These three CSC subsets displayed ALDH activity, formed serial spheroids and developed serial tumors in SCID mice, although RCC-41-PDX-1/CD132+ and RCC-41-PDX-2/CD133+ displayed less efficiently the above CSC properties. RCC-41-PDX-1/CD132+ tumors showed vessels of human origin with CSC displaying peri-vascular distribution. By contrast, RCC-41-PDX-2 originated tumors exhibiting only vessels of mouse origin without CSC peri-vascular distribution. Altogether, our results indicate that PDX murine microenvironment promotes a continuous redesign of CSC phenotype, unmasking CSC subsets potentially present in a single RCC or generating ex novo different CSC-like subsets.


INTRODUCTION
Current knowledge indicates that the initiation, growth, metastasis, chemo-resistance and recurrence of cancers are driven by a subset of cells endowed with stem-like properties called cancer stem cells (CSCs) or tumor-initiating cells. These cells are defined by their capability to self-renew and to recapitulate tumor formation when injected in mice [1][2][3]. Therefore, the identification of CSCs and a better understanding of their complex characteristics will provide very important diagnostic, therapeutic and prognostic targets for clinical application [1,2].
Renal clear cell carcinoma (RCC) is a very aggressive cancer resistant to conventional chemo-and radiotherapy with an early metastatic evolution [4], and it seems likely that renal CSCs may have a relevant role in tumor establishment, progression, and recurrence [5]. At present, only few reports have investigated the presence of CSCs in renal carcinoma. In this context, recent histochemistry observations illustrate that the human prominin-1 (CD133) antigen, in particular of the AC133 epitope frequently employed to isolate CSCs from different tumors [6], does not seem a reliable CSC marker in RCC [7,8]. Some groups attempted to identify renal CSCs using functional assays such as the ability to form spheres in serum-free medium [9] or the presence of the so-called Side Population [10]. Following an alternative sorting strategy, our group has recently identified, from different human RCC specimens a small subset of CSCs cells expressing the mesenchymal marker CD105 [11][12][13]. However, further immunohistochemistry studies on 102 RCC biopsies have shown that CD105 staining could only be detected in the cytoplasm of isolated tumor cells in the specimens derived from patients with high tumor grade and at the highest tumor stage [14]. Finally, it has been demonstrated that tumor-initiating cell frequencies were remarkably rare in well-differentiated tumors [15][16][17] whereas cancers in which differentiation programs are impaired would be comprised of only cells with stem and progenitor cell phenotypes [18,19]. Overall, the abovementioned data suggest that the current understanding of the renal stem cell system is not complete and that the kidney cancer could harbor different CSC pools displaying different phenotypes and functions according likely to tumor grade, stage and differentiation [14,19]. Thus, it is of major interest to develop new approaches in order to identify new putative renal CSCs subsets.
In these studies, we developed an alternative strategy for identifying renal CSCs by adapting to in vitro culture, primary cell suspensions from serial Patient-Derived Xenografts (PDX). Of note, PDX were obtained by serially grafting tumor samples characterized according to their different degrees of differentiation, tumor stage, and aggressiveness in SCID mice [19]. Cell suspensions from PDX of four different RCC patients, characterized by the shortest latency for tumor formation in SCID mice, were chosen from the Gustave Roussy Institute cell collection. The above-mentioned cell suspensions had been immediately frozen without in vitro culture (P-0) or after few in vitro passages (P-1; P-3), representing therefore invaluable material for this type of study [19]. Only the PDX cell suspension from one (RCC-41) out of four patients was able to adapt to the selective medium growth conditions. RCC-41 is an undifferentiated RCC, and from its serial xenografts (RCC-41-PDX-1 and RCC-41-PDX-2) we isolated, sorted, and cloned three novel renal CSCs subsets that diverge from each other in phenotypic and functional properties, fulfilling however most of the criteria used to identify CSCs. These data indicate that even using PDX model, which has been reported as a necessary step for the successful isolation of renal CSCs from Wilm's xenograft [20], it is very difficult to purify CSCs from RCC. Nevertheless, our data strengthen the idea that RCC carcinomas harbor different CSC pools displaying different phenotype and functions. In addition, the serial PDX derived from a single tumor may help to unmask different CSC subsets potentially expressed by a single RCC during its in vivo progression, or to generate ex novo different CSC-like subsets.

In vitro selection of RCC cell suspensions derived from primary RCC xenografts in SCID mice
To test the hypothesis that patient-derived xenografts [18] could represent a source of CSCs in renal cell carcinoma, we employed never cultured or firstpassage cell suspensions derived from four primary RCC xenografts. These PDXs (RCC-28-PDX-1 and -PDX-2, RCC-17-PDX-1 and -PDX-2, RCC-41-PDX-1 and -PDX-2, and RCC-47-PDX-1 and -PDX-2) characterized by different tumor stage, differentiation, histopathology and in vivo aggressiveness [19] (Table 1), were cultured in vitro with a selective medium (DMEM-LG) designed to preserve CSC stemness properties [12]. Only two cell suspensions out of eight (RCC-41-PDX-1 and -PDX-2) adapted to the selective medium and could be serially sub-cultured (Table 1). Cryopreserved cell suspensions were seeded at vessels of human origin with CSC displaying peri-vascular distribution. By contrast, RCC-41-PDX-2 originated tumors exhibiting only vessels of mouse origin without CSC peri-vascular distribution.
Altogether, our results indicate that PDX murine microenvironment promotes a continuous redesign of CSC phenotype, unmasking CSC subsets potentially present in a single RCC or generating ex novo different CSC-like subsets. 5 × 10 5 cells per 25 cm 2 flask. RCC-41-PDX-1 and -PDX-2 cells adhered to the plastic surface with an efficiency of about 40%. After two weeks, RCC-41-PDX-1 and -PDX-2 cells started to proliferate forming isolated colonies exhibiting epithelioid morphology. Upon subculture, about 80% of RCC-41-PDX-1 and -PDX-2 cells adhered to the plastic surface, started to proliferate, and could be serially sub-cultured. Interestingly, the P-0 cell suspension derived from the original tumor (RCC-41-P-0) adapted to DMEM-LG medium but subsequently could not be serially passaged.

Phenotype of RCC-41-P-0 versus RCC-41-PDX-1 and RCC-41-PDX-2
Flow cytometry analysis of primary RCC-41-P-0 cells shows that the majority of these cells strongly express two CSC stem-like markers: CD133 and CD105, while nearly 50% express E-cadherin ( Figure 1A). The expression of E-cadherin in RCC is a good prognostic marker that indicates a tendency towards differentiation [21]. CSCs do not express differentiation markers [1][2][3], therefore the expression of E-cadherin suggests the persistence of a non-CSC cell fraction [12,13] in the RCC-41-P-0 cell suspension.

In vitro CSC functional properties of RCC-41-PDX-1 and -PDX-2 subsets
CSCs are more resistant to conventional chemotherapy than differentiated cancer cells. Indeed, CSCs employ several mechanisms to protect themselves against cytotoxic agents. For example, they have high levels of activity of the detoxifying enzymes ALDH, which enable them to resist the effects of chemotherapy [29]. Therefore, we measured the ALDH activity in RCC-41 by flow cytometry, using an ALDEFLUOR kit. As  figure 3A, sorted RCC-41-PDX-1/CD132 + and RCC-41-PDX-2/CD133 + displayed intermediate levels of ALDH activity (40 and 60% positive cells respectively), while unsorted RCC-41-PDX-2 and sorted RCC-41-PDX-2/CD133displayed high ALDH activity (> 80% positive cells) ( Figure 3A). We also investigated in the PDX-1 and PDX-2 CSC clones the presence of the most primitive cells (Side Population, SP) able to extrude the Vybrant violet dye. The experiments were performed in the presence or not of the transporter inhibitor Verapamil: no SP population could be detected in PDX-1 or PDX-2 CSC clones and PDX-1 and PDX-2 clones did not show significant phenotypic differences (data not shown). A remarkable feature of CSCs is their capability of self-renewal. Indeed this mechanism is required for preservation of the CSC pool [1][2][3]. In order to test their in vitro self-renewal potential, sorted RCC-41-PDX-1/CD132 + and unsorted RCC-41-PDX-2 cells were cultured at concentrations ranging from 10 2 to 1 cell per well in DMEM-LG serum-free medium and assessed for spheroid formation ( Figure 3B). Within 21 days, RCC-41-PDX-1/CD132 + grew forming small spheroids with an efficiency of about 40% at the limiting dilution of 10 2 cells per well, while no spheres developed at the limiting dilution of 1 cell/well. Unsorted RCC-41-PDX-2 produced spheroids in about 100% of the wells at 10 2 cells per well, while at the limiting dilution of 1 cell per well, RCC-41-PDX-2, sorted RCC-41-PDX-2/CD133and  RCC-41-PDX-2/CD133 + formed spheroids with respective efficiencies of 6, 3, and 1% ( Figures 3B and 3C). RCC-41-PDX-1/CD132 + and -PDX-2 primary spheroids were subsequently dissociated using trypsin and were able to grow serially as spheroids at their respective limiting dilutions.

Tumorigenic potential of human renal RCC-41-PDX-1/CD132 + in SCID mice
The results depicted in Table 2 shows that subcutaneous injection of 10 3 RCC-41-PDX-1/CD132 + -1.0 cells into SCID mice induced the formation of tumors within 3 weeks in all studied mice (n = 8 mice), while subcutaneous injection of 10 2 RCC-41-PDX-1/ CD132 + -1.0 induced the formation of tumors in 2 out of 8 mice. RCC-41-PDX-1/CD132 + -1.0 xenografts were enzymatically dissociated and the derived cell suspension (RCC-41-PDX-1/CD132 + -1.1) was phenotypically characterized and subsequently re-injected in SCID mice at 10 3 and 10 2 cells per mouse. RCC-41-PDX-1/CD132 + -1.1 induced the formation of tumors within 3 weeks in 5 out of 6 mice at both cell concentrations, indicating the acquisition of an increased tumor forming efficiency at the lower cell concentration. RCC-41-PDX-1/CD132 + -1.1 xenografts subjected to the same enzymatic dissociation procedure produced the derived cell suspension termed RCC-41-PDX-1/CD132 + -1.2. These cells exhibited a highly decreased capacity to form tumors in the SCID mice: after 3 weeks only 1 out of 6 mice developed a tumor and only at the concentration of 10 3 cells. These data show that in SCID mice, RCC-41-PDX-1/CD132 + generates serially transplantable tumors with a variable efficiency. The important differences in the tumor-forming ability observed in the serial RCC-41-PDX-1/CD132 + xenografts led us to investigate whether these variations could be associated with a modified expression of CSC markers. For this purpose, we analyzed the phenotypes of all injected cell suspensions. Flow cytometry analysis indicates that the percentage of CD133 + cells present in the cell suspension injected in the SCID mice directly correlated to the tumor forming ability of the cells ( Figure 5A) suggesting therefore that, contrary to what previously reported [7,8], CD133 may help to identify subsets of CSCs in RCC.

Progressive epithelial differentiation and presence of human vessels in RCC-41-PDX-1 serial tumors
Histologic analysis of tumors serially generated in SCID mice (RCC-41-PDX-1/CD132 + -1.0, -1.1 and -1.2) shows a homogenous histopathology consisting of highly mitotic undifferentiated carcinomas, and rare clear cells ( Figure 5B). Co-staining with anti-hCD31 and anti-hCD133 antibodies shows that several vessels of human origin are detectable in the serial xenografts ( Figure 5B) and that CD133 + cells display a peri-vascular distribution ( Figure 5B) that is maintained in the serial xenografts even if their frequency progressively decreases. In serial xenografts starting from PDX-1.1, we observed the appearance of cells expressing the human epithelial marker EpCAM ( Figure 5B), which strongly increases in PDX-1.2, suggesting a natural tendency towards epithelial differentiation, that is associated with a rarefied frequency of CD133 + cells ( Figure 5A and 5B). Thus, RCC-41-PDX-1/CD132 + cells are able to generate serially transplantable tumors in SCID mice, which recapitulate the histopathology of the original RCC and show a tendency for a progressive engagement in an epithelial differentiation pathway. This was confirmed by the increased expression of E-cadherin and the decreased expression of vimentin analyzed by RT-qPCR ( Figure 5C). Interestingly, flow cytometry analysis performed on www.impactjournals.com/oncotarget

Loss of human endothelial micro-vessels (EM) in RCC-41-PDX-2 tumors
H&E staining of tumors generated in SCID mice by RCC-41-PDX-2, and RCC-41-PDX-2/CD133 -, RCC-41-PDX-2/CD133 + and their respective clones, shows that PDX-2 xenografts exhibit histopathology similar to that observed in PDX-1 tumors, consisting of highly mitotic undifferentiated carcinomas, and rare clear cells ( Figure 7A). Immunofluorescence analysis of tumor vessels shows off a remarkable property that distinguishes RCC-41-PDX-1/CD132 + from RCC-41-PDX-2 xenografts: in all PDX-2 xenografts, vessels are entirely of murine origin as shown by EM counts using hCD31 and mCD34 mAbs proved to be rigorously species-specific in a previous study [47]. All vessels present in PDX-2 xenografts were lined with mCD34 + endothelial cells, whereas vessels lined with hCD31 + endothelial cells were never detected ( Figure 7A). Moreover PDX-2/CSCs, identified in vivo through the expression of CD146, were detected nearby murine vessels but did not show the strict peri-vascular distribution detected in PDX-1 xenografts ( Figure 7B).

DISCUSSION
The identification of renal CSCs represents a therapeutic priority that, for the moment, remains somewhat elusive due to RCC genetic and histological heterogeneity [4,5]. Recently, a CSC population expressing CD105 was purified from RCC patients [12]. An immunohistochemical study, performed on paraffin-embedded tumor samples derived from 102 RCC patients, showed that CD105 expression in tumor cells was found in high-grade tumors and highest tumor stages [14]. This would suggest that RCC may harbor CSCs with different phenotypes [5]. Thus we developed an alternative strategy for identifying putative renal CSCs, different from the CD105 + cell sorting [12]. For this purpose, we exploited the Gustave Roussy Institute cell bank; using never-cultured cell suspensions derived from the four more aggressive serial RCC-PDX displaying different phenotypes, tumor stage and grade [19,48]. Only RCC-41-PDX-1 and -PDX-2 cell suspensions out of four different RCC-PDX adapted to the selective medium conceived for preserving in vitro the stem cell properties [12]. Interestingly, we also tried to culture the primary cell suspension derived from the original tumor (RCC-41-P-0); however, while the cells attached and grew slowly up to near confluence, they could not be further passaged.
In order to further enrich RCC-41-PDX-1 and RCC-41-PDX-2 subsets in CSC, we developed different cell sorting strategies. Sorted cells were cultured and amplified over two in vitro passages, analyzed for their phenotype, and subsequently cloned at limiting dilution. Using the PDX model, we selected three potential CSCs subsets that could be further distinguished on the basis of the expression of the CSC-like marker EpCAM [26]: CSC/PDX-1/CD132 + / EpCAM -, CSC/PDX-2/CD133 -/EpCAM low , and CSC/PDX-2/CD133 + /EpCAM bright . The sorted CSC/PDX-1 and CSC/PDX-2 subsets, as well as their respective clones, display properties typical of CSCs detected in other solid tumors. Namely they: 1) express stem cell markers, 2) exhibit clonal growth at limiting dilution, 3) grow and can be serially transferred as floating spheroids, 4) express mid-high levels of the detoxifying enzyme ALDH, a typical CSC marker [51], and 5) generate serially transplantable tumors characterized by similar histopathology. However it must be stated that serial PDX-1/CD132 + xenografts displayed microvessels of human origin and strict peri-vascular distribution of CSCs that favors the preservation of stemness [52], while PDX-2 xenografts exhibited microvessels of murine origin without perivascular distribution of CSCs.
Finally, in PDX-2 xenografts, murine microenvironment induces a remodeling of the CSC properties favoring the appearance of two novel subsets of less primitive CSC since they have lost the potential to generate human microvessels and are characterized by different phenotypes: PDX-2/CD133 -/EpCAM -/ERBB4 + is the predominant aggressive subset while the PDX-2/ CD133 + /EpCAM bright is a minor, less aggressive CSC subset characterized by the expression of several tumor suppressor genes.
In conclusion, our data support the idea that the serial PDX derived from a single tumor may help to unmask the different CSC subsets potentially expressed by a single RCC during its in vivo progression. Alternatively, it is also possible to postulate that the PDX CSC niche, which is of murine origin, could redesign the properties of human CSC favoring the generation of modified CSCs that become unable to generate human microvessels hampering somehow the development of targeted anti-angiogenic preclinical approaches.

Sphere formation assay
We evaluated the self-renewal capacity of RCC-41-PDX-1/CD132 + , RCC-41-PDX-2, RCC-41-PDX-2/ CD133 + , and RCC-41-PDX-2/CD133by performing a limiting dilution assay for spheroid formation. Cells were plated using a FACS DiVa cell sorter equipped with autoclone software (Beckton Dickinson, Le Pont-de-Claix, France) at a density of one and 100 cells per well in ultralow attachment 96-well plates (Corning Life Sciences, Acton, MA). Each well was supplemented with 200 μl of serum-free complete DMEM-LG. After 3 weeks, each well was examined under a light microscope and the total number of wells containing spheroid colonies was determined. Five replicates were used for each condition, and the experiment was repeated two times.

RNA extraction and quantitative real time reverse transcriptase polymerase chain reaction
Total RNA was isolated from cells using TRIZOL reagent (Invitrogen), according to manufacturer's instructions. RNA was quantified spectrophotometrically (Nanodrop ND-1000, Wilmington DE). Primers used for qRT-PCR are shown in Table 4; other primer sequences are available upon request.
To detect mRNA expression, first-strand cDNA was produced from 200 ng of total RNA using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Real-time PCR experiments were performed in 20 µl reaction mixture containing 5 ng of cDNA template, the sequence-specific oligonucleotide primers (purchased from MWG-Biotech AG, Ebersberg, Germany, www.mwg-biotech.com) and the Power SYBR Green PCR Master Mix (Applied Biosystems). Negative cDNA controls (no cDNA) were cycled in parallel with each run. qRT-PCR was performed using a 96-well StepOne™ Real Time System (Applied Biosystems). mRNA comparison between samples was calculated on relative expression data normalized using TATA binding protein (TBP), as endogenous control. Fold change expression with respect to controls was calculated for all samples.

Flow cytometry
For all assays described below, we acquired fluorescence data for 10,000 events on a FACSCalibur flow cytometer (BD Biosciences, Oxford, UK) and analyzed the data with the use of CellQuest software (BD Biosciences) or FlowJo (Treestar). The experiments were repeated at least three times. In some experiments, cells were analyzed on a FACScan (Becton Dickinson, Franklin Lakes, NJ, USA).

Tumorigenic potential of RCC-41-PDX-1/ CD132 + and -PDX-2 cells in severe combined immunodeficient (SCID) mice
RCC-41-PDX-1/CD132 + and -PDX-2 subsets were harvested by incubation with trypsin-EDTA, washed in PBS and resuspended in 100 μL of DMEM. Aliquots of 10 2 or 10 3 cells were added to 100 μL of Matrigel (BD Biosciences), chilled on ice, and injected subcutaneously with a 1-mL syringe fitted with a 26-gauge needle into the left and right sides of 6-week-old male SCID mice (Charles River, Jackson Laboratories, Bar Harbor, ME). Four to eight mice were injected with each of the different RCC-41 subsets. We performed three independent experiments. At different times after cell injections (3, 6, 9, and 10 weeks) the mice were killed by carbon dioxide asphyxiation and their tumors were excised, weighed, and measured. Studies were approved by the Italian Ministry of Health and by the Institutional Review Board of the University of Turin and were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources: 7th ed. Washington, DC, Institute of Laboratory Animal Resources, Commission on Life Sciences, National Research Council, 1996).
The endothelial micro-vessels (EM) were assessed by anti-hCD31 and anti-mCD34 staining and examination of twenty microscopic fields (0.5 mm 2 ) per tumor. The most intense vascular areas (hotspots) were selected subjectively from each tumor section. The micro-vessels with a clearly defined lumen or well-defined linear vessel shape were taken into account.
For image analysis, digital images were collected using a Nikon E-1000 fluorescence microscope (Nikon Instruments, Tokyo, Japan) equipped with appropriate filter sets and the Genikon imaging system software (Nikon Instruments).

Statistical analysis
For tumor studies in mice, the number of animals per group was based on our acquired expertise after 4 years of experiments with this model (we did not perform power calculations to determine the number of mice per group). The unit of analysis was the mouse, and the average of all tumors per mouse was taken into account. For the survival curves, the log rank test was used to compare the survival of the different groups. Otherwise, the two-sided Student's t test was used to compare groups. A P value less than .05 was considered statistically significant.