Main

Colorectal carcinoma (CRC) is one of the three most common malignant diseases and the second leading cause of cancer-related deaths.1, 2, 3 More than 40% of CRC are located in the rectum. Although the local recurrence rate of rectal carcinoma is less than 10% after preoperative radiotherapy/chemoradiotherapy and surgery based on total mesorectal excision, the development of distant metastases leads to a high rate of treatment failure of 30–35%.4 The hypothesis that certain subpopulations, frequently termed tumor-initiating, tumor-propagating or cancer stem cells (TIC/TPC/CSC), are resistant to classical treatment strategies and responsible for recurrence and metastasis of the disease is under intense investigation in CRC.5, 6, 7, 8, 9, 10, 11, 12 Several cell surface markers have been described to enrich for TIC/TPC/CSC subpopulations from CRC, such as CD133, CD44 or CD166.5, 12, 13, 14, 15, 16, 17, 18 In 2007, two groups independently described CD133-positive CRC cells from primary material but not their negative counterparts to be capable of producing tumors in mouse models.19, 20 In one of these studies, CD133+ subpopulations could be maintained in vitro as spheres under serum-free conditions, but CD133 expression was downregulated when cultured adherently upon serum supplementation.20 Interestingly though, established CRC cell lines routinely grown as adherent monolayers under serum-containing conditions show a highly variable CD133 surface expression.21 The argument that cell lines no longer reflect the original in vivo tumor cell geno- and/or phenotype is legitimate but challenges all previous and ongoing functional studies based on such cancer cell lines notwithstanding that they are tumorigenic in mouse models and have indeed led to numerous insights in cancer cell biology. In addition, primary and secondary anti-tumor treatment testings are not routinely performed with primary cells from patients’ material but most frequently relies on established cell lines. Therefore, it is of utmost interest to further verify the expression profiles of putative CSC biomarkers not only in primary tumor cells but also in established cell lines and to evaluate the tumorigenic potential and therapy response of CD133-positive vs CD133-negative cell lines and subpopulations.

We previously found CD133+ and CD133− subpopulations of the near-diploid and microsatellite-instable CRC cell line HCT-116 to neither differ in clonogenic cell survival or radioresponse in vitro nor in tumorigenic capacity.21 In contrast, HT29 were described by others to show a CD133-related CSC behavior based on small cohort animal experiments22 and characteristics consistent with those of CSCs are also stated for the CD133+ subfraction of SW620 cells, although no difference in tumor formation rate but in tumor size was observed.23 HT29 and SW620 are aneuploid or hyperdiploid, respectively, with a microsatellite-stable, chromosome instable (MSS/CIN) phenotype, which is seen in the majority (about 85%) of spontaneous CRC.24 We therefore hypothesized that the difference to our HCT-116 data may relate to the different carcinogenesis pathways. The current study was thus performed to verify the presence of a CD133-associated CSC phenotype in MSS/CIN CRC cell lines. The particular aim was a subsequent systematic comparison of the in vitro and in vivo behavior of CD133+ vs CD133−/low CRC cell line subpopulations. The results are striking, allude to various highly relevant methodological inconsistencies related to CD133 detection and isolation of subpopulations with differential CD133 expression, and indicate a high plasticity in CD133 positivity.

MATERIALS AND METHODS

Cell Lines and Culture Routines

Six colon carcinoma cell lines purchased from the ATCC (American Type Culture Collection, Manassas, USA) were studied (HCT-116—MSI; HT29, SW620, SW480, SW1417, LS1034—all MSS, see also Supplementary Table 1). Cell lines were free of mycoplasma as tested using a PCR Mycoplasma Kit (Applichem, Germany) and checked for correct genetic profile. Routine verification of cell line purity and clonality was detailed earlier for HCT-116 cells.21 Authenticity of cell lines was confirmed by microsatellite analyses at the Institute of Legal Medicine (Dresden University of Technology) using the commercial multiplex PCR kits Mentype NonaplexQS Twin (Biotype, Germany) and PowerPlex 16 (Promega Corporation, USA). Amplicons were detected by capillary electrophoresis in denaturing polymer POP4 in an ABI 310 sequencer (Perkin-Elmer, USA) according to the manufacturer’s instructions.

Cells were routinely grown in DMEM (Dulbecco's Modified Eagle Medium, PAN Biotech, Germany) containing antibiotics, phenol red, 25 mM HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), 5 mM glucose, 1 mM sodium pyruvate and 4 mM L-glutamine and supplemented with 10% FCS (Fetal Calf Serum, PAN Biotech). Culturing was performed at 37°C in a humidified 5–8% CO2 in air atmosphere. For experimental setup, cells were thawed from frozen stocks and passaged at least twice but no more than 10 times in 2-D culture before use. Exponential monolayer cultures were dissociated by enzymatic and mechanic means using a 0.05% trypsin/0.02% EDTA (Ethylenediaminetetraacetic acid) in PBS (Phosphate-buffered saline, PAN Biotech) solution to obtain single-cell suspensions for cell transfer, flow cytometric analyses and cell sorting. Cell numbers and volumes were monitored with a Casy1 cell analyzer system (Roche Innovatis).

Flow Cytometric Analyses and Fluorescence-Activated Cell Sorting (FACS)

CD133 cell surface expression was analyzed in single-cell suspensions from monolayer and spheroid cultures as well as from HT29, SW480 and SW620 xenografts. An advanced anti-CD133 staining protocol was applied as described earlier.21 Briefly, cells were incubated for 30 min in the dark at 4°C with a PE (phycoerythrin)-conjugated anti-CD133 antibody (CD133/1-PE or isotype, Miltenyi Biotec) diluted 1:100 in PBS containing 0.5% FCS and 2 mM EDTA followed by signal enhancement via a two-step FASER series (Fluorescence Amplification by Sequential Employment of Reagents) according to the manufacturer’s instructions (Miltenyi Biotec). Cell suspensions originated from xenograft tumors were also co-stained with an anti-human CD326-FITC (fluorescein isothiocyanate) antibody (dilution 1:25; Miltenyi Biotec) to discriminate human epithelial and mouse stromal cells. Adequate isotope antibodies at equivalent working concentrations served as controls. Propidium iodide was added at 2 μg/ml per sample directly before measurement to discriminate membrane-defect cells. CD133 profiles were recorded with a FACScan or a FACSCanto™ II flow cytometer (BD Biosciences, USA); signals of a minimum of 2 × 104 events or 1 × 104 viable cells of interest were collected per sample.

Intracellular CD133 positivity was determined in cell suspensions (LS1034) pre-stained for membraneous CD133 as described. These cells were fixed and permeabilized using freshly prepared 4% paraformaldehyde in PBS, followed by two washing steps with PBS containing 1% FCS and a 20 min exposure to sterile permeabilization buffer (0.05% saponin in PBS/1% FCS buffered with 0.12% HEPES) at room temperature (RT). Membrane-permeabilized cells were then exposed to the mouse monoclonal W6B3C1 antibody (Miltenyi Biotec) at a dilution of 1:100 in permeabilization buffer or equivalent concentration of an isotype control (Immunotools). An Alexa-488-conjugated goat-anti-mouse secondary antibody (1:100; Invitrogen) was applied for visualizing intracellular CD133 in the CD133−/low-sorted LS1034 fraction. Cells were washed twice and kept on ice between the incubation steps and were finally resuspended at a concentration of 1 × 106 cells in 0.5 ml PBS/1% FCS for flow cytometric analysis.

For FACSorting using a FACSAria (BD Biosciences), single-cell suspensions from exponentially grown monolayer cultures were stained accordingly and resuspended at a concentration of (0.5–1) × 107 cells/ml in PBS plus 2 mM EDTA. Sorted subpopulations were routinely reanalyzed by flow cytometry for purity; cell numbers and quality were reassessed via the Casy1 cell analyzer before any in vitro culturing or in vivo implantation.

Western Blot Analysis

Whole cell protein was extracted using RIPA buffer (1 × PBS, 1% Nonidet P40, 0.5% Sodiumdeoxycholat, 0.1% SDS) after addition of 0.1 M PMSF and a 1:100 protease inhibitor cocktail (P2714, Sigma-Aldrich, Germany). Protein content was determined via the BCA Protein Assay Kit (Pierce, Germany) as described by the manufacturer and aliquots were stored at −80°C. Two antibodies were used to visualize CD133 protein. An extracellular domain (AC133) is recognized by the anti-human W6B3C1 antibody (Miltenyi Biotec), which was applied in a TBS-buffered system, whereas the Ab19898 anti-human CD133 antibody (Abcam, UK) in PBS-buffered solutions binds to the intracellular C-terminus of CD133. Proteins were separated by 10% SDS-PAGE and transferred onto PVDF membranes (Millipore, Germany), which were blocked overnight in TBS and PBS, respectively, containing 0.1% Tween20 and 5% milk powder. After washing three times with the respective buffers supplemented with 0.1% Tween20, membranes were exposed to either the mouse monoclonal W6B3C1 antibody (1:100 diluted in blocking buffer) or the rabbit polyclonal anti-human CD133 antibody Ab19898 (1:100 diluted in PBS containing 0.1% Tween20) for 1 h at RT. Membranes were then washed, incubated with secondary anti-mouse or anti-rabbit immunglobulin/HRP (dilution 1:5000, 1 h, RT; Dako, Germany), subjected to western blotting luminol reagent (Santa Cruz, Germany) for 1 min and film exposure for 10 min. β-Actin was probed as control using the monoclonal antibody Ab6276-100 (Abcam, UK) in a TBS-buffered system (1:2000 in TBS containing 3% milk powder) and the secondary system as described. Semi-quantitative analysis of CD133 protein relative to β-actin was performed with ImageJ Software (http://rsbweb.nih.gov/ij/); the CD133 expression of originally distributed cell populations was set as 100%.

2-D assays of FACSorted Subpopulations

Subpopulations of selected cell lines, sorted according to their CD133 expression, were studied in 2-D culture with respect to colony formation capacity, growth behavior and CD133 profile. For colony formation assays (CFA), 400 FACSorted cells were seeded into each well of a six-well plate in standard DMEM and incubated under routine culture conditions for a period of 17 days. Cells were then stained with 0.5% crystal violet or 0.05% Coomassie Brilliant Blue and plating efficiencies were determined by manually counting colonies containing more than 50 cells under an AxioVert200 microscope (Carl Zeiss MicroImaging GmbH, Germany). A minimum of three wells per condition and subpopulation, respectively, was analyzed. To evaluate the impact of tissue normoxic conditions on colony formation of HT29 cells, adherent cells were relocated 4 h after plating to an Invivo300 Incubator (IUL, Germany) to be kept in a humidified 5% CO2 atmosphere at 4% oxygen.

A total of 1 × 104 CD133+, CD133−/low and original (run-through sorter) HT29 and SW620 cells were seeded in 2 ml supplemented DMEM per well into six-well culture plates to monitor cell growth and alterations in CD133 profile of sorted subpopulations. At defined time points, cells were enzymatically dissociated and cell numbers of three individual samples per condition were determined as described in routine culturing. Single-cell suspensions were analyzed for their CD133 cell surface expression via flow cytometry (see above). To obtain sufficient cell numbers for the analysis of CD133 expression at day 4 in 2-D culture, 6.5 × 104 sorted cells of each subpopulation in 13 ml DMEM were inoculated in 10 cm culture dishes and were handled according to the six-well plate approach.

3-D Assays of FACSorted Subpopulations

3-D spheroid formation capacity of HT29 subpopulations after FACS was determined in liquid overlay using a standardized setup and culture conditions,25 as well as a limiting dilution approach. In the standardized procedure 1500 CD133+, CD133−/low and mixed HT29 cells in 200 μl medium were seeded per well into 1.5% agarose-coated 96-well plates. Spheroid formation, spheroid integrity and volume growth kinetics were monitored by semi-automated measurement of diameters and volumes from phase-contrast images as described earlier.25 A minimum of 18 spheroids per seeded population were analyzed. To study the CD133 profile of HT29 subpopulations throughout 3-D growth, 12–48 spheroids per condition were collected, enzymatically dissociated using a 0.1% trypsin/0.04% EDTA solution and cell numbers were evaluated with the Casy1 system. Resulting single-cell suspensions were processed for flow cytometric analysis as described above. The CD133 distribution at day 4 in culture (initiation period) was analyzed in duplicate whereas triplicates could be measured at later time points during 3-D culturing.

In limiting dilution experiments, cell concentrations of FACSorted HT29 subpopulations were reduced in consecutive 1:10 dilution steps to provide cell suspensions with an average concentration of 5 and 25 cells/ml standard medium. 200 μl of these suspensions were then pipetted into each well of an agarose-coated 96-well plate; 48–60 wells were prepared for each suspension and three independent experiments were performed. Wells were microscopically imaged every 2–4 days and the development of sphere colonies with a size of ≥100 μm was monitored over a period of 4 weeks. Spheroid formation capacity (%) was calculated as the number of wells with single-sphere colonies divided by the number of wells seeded.

Xenograft Formation Assay

HT29, HCT-116 and SW620 cells were sorted according to their CD133 expression profile, subpopulations were reanalyzed for purity and viable cells were counted. Cell suspensions between 10 and 2.5 × 103 were prepared in 100 μl of a 50% Matrigel (BD Biosciences) in PBS solution and immediately injected subcutaneously into the hind limb of 8- to 10-week-old female NMRI (nu/nu) mice (Experimental Center, Medical Faculty, University of Technology Dresden). The injection of low cell numbers was performed as block experiments, each with a total of four animals and injections per cell number (100 and 10) and subpopulation, respectively. A total of four blocks with cells derived from independent preparations and sort procedures were averaged. SW480 cells were implanted without separation using 10 000 or 500 cells per injection site to determine tumor formation rates. Palpable tumors were measured twice a week using a manual caliper to determine longest and perpendicular tumor axes and calculate tumor volumes by the formula of a rotational ellipsoid: v=π/6 × a × b2. Animals were killed when tumors reached a mean diameter of 1.2–1.5 cm, upon completion of the observation period of 120 days after injection or when animals appeared to suffer. Animal facilities and all experiments were approved in accordance to institutional guidelines and German animal welfare regulations. Animals were fed with commercial laboratory animal diet and water ad libitum.

For the analysis of CD133 distribution, selected xenograft tumors were dissected at defined diameters, non-tumor tissue was removed and tumor material was then weighted and minced with scalpels. Tumor pieces were incubated in enzyme dissociation solution containing Collagenase NB 4G (Serva, Germany) with or without Hyaluronidase, DNAse (both Sigma, Germany) and trypsin/EDTA (PAN Biotech) overnight. The obtained cell suspensions were filtered through a 70-μm mesh and cell numbers were determined. Aliquots were stained for flow cytometric analysis as described.

Evaluation of Drug and Radiation Response

Drug and radiation response of cell subpopulations were studied in classical 2-D CFA. In principle, HT29 cells were seeded and cultured as described in 2-D assays and HCT-116 cells were processed as depicted earlier.21 For drug treatment, however, cells (400 HT29 or 500 HCT-116) were inoculated in only 1 ml culture medium per well and allowed to adhere over a period of 4 h under standard conditions before 1 ml of conditioned DMEM containing double concentrated chemotherapeutics (Sigma-Aldrich, Germany) was added resulting in final concentrations of 0.1–25 μM for 5-Fluorouracil and 1–50 μM for Oxaliplatin. After an incubation time of 72 h, drugs were removed by 100% medium exchange and cultures were kept in the incubator for the time indicated and processed as detailed previously. Cell survival curves were documented and IC50 values were determined for each experiment using a Hill-Fit with variable slope in the software GraphPadPrism (La Jolla, CA, USA). Experiments were performed at least in triplicate and the IC50 values from the individual Hill fits for each sorted subpopulation were averaged.

To evaluate radioresponse, cell subpopulations were seeded in CFA, allowed to adhere for 4 h and then irradiated at RT with single doses of 0.5–12 Gy (1.3 Gy/min, 200 kV X-rays; 0.5 mm Cu filter; YxlonY.TU 320; Yxlon International, Germany). Colony formation capacity was analyzed as described and surviving fractions at 2 Gy (SF2 Gy) were calculated and averaged from dose–response curves of three individual experiments. A paired, two sided t-test was used to assess statistical significance.

RESULTS

In a previous study, we documented that CD133+ and CD133− HCT-116 cells do not differ in any of the parameters of interest, including in vitro radioresponse and subcutaneous tumor formation, when injecting ≥2500 cells per mouse.21 Because all of the animals developed a tumor, verification of our observation was required using a limiting dilution approach with 500, 100 and 10 cells per injection. As summarized in Table 1a, HCT-116 cells are highly potent and tumor take rate is still 100% at 100 cells injected per animal. A reduction in tumor formation was seen at 10 cells per animal. However, there was no significant difference in the tumor formation capacity of CD133+ vs CD133− HCT-116. To also upgrade our previous data on therapy response, sensitivity of the respective HCT-116 subpopulations to 5-FU and Oxaliplatin, which are frequently implemented in clinical treatment regimes for advanced CRC, was analyzed. Dose–response curves (n=3 for each drug) from 2-D CFAs revealed that CD133+ vs CD133− HCT-116 subpopulations do not differ in drug sensitivity (Supplementary Figure 1). The mean IC50 values (±s.d.) calculated from the individual dose–response curves are listed in Table 2a. These data confirm that HCT-116 cells do not show a therapy-relevant CD133-related phenotype.

Table 1 Xenograft formation of (a) HCT-116, (b) HT29 and (c) SW620 colorectal cancer cell subpopulations sorted according to their CD133 expression and injected subcutaneously with matrigel into the hind limb of NMRI (nu/nu) mice
Full size table
Table 2 IC50 values (±s.d.) for (a) HCT-116 and (b) HT29 cell subpopulations sorted according to their CD133 expression profile, seeded in colony-forming assays and treated after attachment with different concentrations of either 5-FU or Oxaliplatin
Full size table

The MSS/CIN cell line HT29 has been described to contain highly variable CD133+ fractions in different laboratories ranging from a few to >95%.18, 22, 26, 27, 28, 29, 30 Because the fluorescence-staining procedure critically impacts the identification and quantification of CD133+ and CD133− cell fractions, we applied the same advanced staining protocol as for HCT-116 cells. The influence of signal amplification on flow cytometric discrimination of CD133−/low vs CD133+ in HT29 cells is documented in Figure 1a. With two signal amplification steps, only a small HT29 subfraction remained in the isotype gate (CD133−/low), whereas 90.9±6.4% of exponentially growing HT29 cells in serum-containing DMEM clearly expressed CD133 on the cell surface. The resulting sort layout and a representative reanalysis after separation of the subpopulations are shown in Figure 1b. CD133+ and CD133−/low separated subpopulations did not overlap in fluorescence signal intensity. Subpopulations were studied in vitro and in vivo compared with originally distributed HT29 cells, which had undergone the same staining and sorting procedure (original run-through sorter).

Figure 1
figure 1

The experimental setup to separate CD133−/low and CD133+ HT29 cell subpopulations includes signal amplification, which critically affects the quality and quantity of CD133 detection by flow cytometry. (a) Impact of Fluorescence Amplification by Sequential Employment of Reagents (FASER) on quantification and discrimination of HT29 subpopulations by flow cytometry after staining with an anti-CD133/1-PE antibody. Dot blot diagrams show HT29 cells stained without (left), with one (central) and with two rounds of fluorescence signal amplification (right). Only a small subset of ∼6% of the cells remains in the CD133−/low gate after two rounds of signal amplification and relative to the respective isotype control (grey), which underwent the same procedure. Propidium iodide (PI)-positive, membrane-defect cells and doublets were excluded. (b) The sort layout for fluorescence-activated cell sorting (FACS) was established according to the CD133 expression level of HT29 cells after staining of single-cell suspensions with 2 × FASER amplification. Sorted subpopulations were routinely reanalyzed as shown in the right panel for one representative experiment. The subfractions could be clearly identified; the purity according to the reanalysis gates was always ≥97% for both fractions.

Full size image

CD133−/low vs CD133+ HT29 cells were first analyzed in 2-D culture using colony formation and growth assays. The staining and sorting procedure in general reduced the plating efficiency (data not shown). However, independent of this, we observed a significant (P<0.005) difference in colony formation capacity of the two subpopulations (Figure 2a). The plating efficiency of CD133+ HT29 cells was always higher than in their CD133−/low counterparts (29.9±4.9% vs 17.0±4.8%); the originally distributed control was intermediate (26.6±5.7%; n=6 individual experiments). Despite the difference in colony formation, there was no growth advantage of the CD133+ HT29 cells; doubling times of the sorted HT29 cells were 23.3±2.2 h (CD133−/low), 23.7±1.7 h (CD133+) and 24±2.5 h (originally distributed population; n=3; Figure 2b).

Figure 2
figure 2

CD133+ and CD133−/low HT29 cells differ in colony formation capacity and initial spheroid size but exhibit comparable 2-D and 3-D growth kinetics. 2-D and 3-D growth characteristics were monitored for FACSorted CD133+ vs CD133−/low HT29 cells in comparison to original/mixed HT29 cells that underwent the same experimental procedure (original run-through sorter). (a) Plating efficiencies (+s.d.) of HT29 subpopulations averaged from n=6 independent experiments. CD133+ HT29 cells show a higher colony-forming capacity than their CD133−/low counterparts. Representative wells with colonies derived from each subpopulation (CD133+, CD133−/low, original) are shown in the right panel. (b) Cell numbers for the CD133−/low seeded fraction remain lower throughout culturing as compared with the CD133+ and the original populations, but growth kinetics and doubling times are comparable. A representative monolayer growth experiment from n=3 is shown; three samples were analyzed for each time point per experimental series. (c) Spheroids derived from 1500 CD133+ HT29 cells are significantly larger at day 4 in liquid overlay culture than those originated from the same number of CD133−/low HT29 cells. Spheroids were monitored by phase-contrast imaging and spheroid diameters were analyzed and averaged for ≥16 spheroids per experimental series. The graph shows mean spheroid sizes (+s.d.) from n=4 independent experiments. (d) Spheroid volume growth recorded after seeding of CD133+, CD133−/low and original HT29 subpopulations according to (c) does not show a difference in 3-D growth kinetics. A representative experiment from n=4 independent series is shown.

Full size image

Because 3-D cultures better reflect in vivo micromilieu conditions than monolayer cultures, we also studied the growth behavior of HT29 subpopulations in a spheroid formation and growth assay. When 1500 CD133−/low, CD133+ and originally distributed HT29 cells were seeded per well into agarose-coated 96-well plates, spheroids were formed in all wells in liquid overlay but the size of the spheroids after a defined initiation period differed significantly (Figure 2c). At day 4 in culture, spheroids derived from CD133+ HT29 were significantly larger than those formed from the CD133−/low fraction (298.1±38.3 mm vs 227.9±21.3 μm in diameter; P<0.005; n=5). The size of spheroids after inoculation of originally distributed control HT29 was intermediate (274.5±42.7 μm), hence reflecting the contribution of the different fractions of ∼90% CD133+ and 10% CD133−/low cells. Despite this difference in initial spheroid diameter, spheroid volume growth kinetics were similar (Figure 2d, n=5), indicating that the discrepancy in spheroid size at day 4 does not result from a different proliferative activity but rather from the difference in clonogenic cell survival. We therefore evaluated the spheroid formation capacity of CD133+ vs CD133−/low HT29 cells in a limiting dilution approach by seeding mean cell numbers between 1 and 1500 and monitoring the formation of growing spheroids over a period of several weeks. There was no difference in spheroid formation rate of the different HT29 subpopulations if cell numbers of 5 and higher were seeded per well. With an average number of 5 cells/well, spheroid formation was achieved in 79.3±1.5%, 77.3±7.4% and 77.6±9.2% of the wells inoculated with CD133+, CD133−/low and original HT29 cells, respectively. In contrast, spheroid formation rate clearly differed if cell suspensions were seeded to average 1 cell/well. Here, CD133+ cells showed a significantly higher capacity to form spheroids than the CD133−/low HT29 fraction with 37.5±6.0 vs 25.5±7.8% of the wells containing growing spheroids (P<0.05). The original run-through sorter 1 cell/well control led to the formation of spheroids in 34.1±12.3% of the wells.

The CSC concept assumes that cells with the particular phenotype are more resistant to anticancer treatment and are responsible for recurrence of the disease.5, 31 Further CFAs were thus performed to evaluate the susceptibility of HT29 subpopulations against 5-FU, Oxaliplatin and single-dose irradiation. According to HCT-116 cells, CD133+ and CD133−/low HT29 cell populations did neither significantly differ in drug response nor in radiosensitivity (Table 2b and Supplementary Figure 2). Because 2-D colony and spheroid formation capacity after single-cell seeding of CD133+ HT29 cells was higher than for their CD133−/low counterparts, and on the basis of literature data, we hypothesized a different tumor formation capacity of the subpopulations. Subcutaneous injection of 2500 and 500 cells (CD133+, CD133−/low and original run-through sorter) into NMRI (nu/nu) mice using a matrigel/PBS mixture resulted in 100% tumor formation. We then performed block experiments with four independent staining, sorting and dilution series to inject 100 and 10 cells per mouse (four mice per block for each population). In the 100 cell cohorts, only one injection failed to produce a tumor. By contrast, the injection of only 10 cells resulted in a clearly reduced tumor take. However, no difference in tumor formation rate was observed for the different HT29 subpopulations sorted according to their flow cytometric CD133 expression profile (Table 1b).

Significant phenotypic differences of CD133+ and CD133−/low HT29 cell subpopulations in vitro were observed but were not as pronounced as expected and did not correlate with in vivo tumor formation. We therefore studied colony formation under tissue normoxia (4% oxygen) to more closely reflect the oxygen conditions upon subcutaneous injection of the cells and found HT29 clonogenic cell survival after sorting in general to be enhanced in tissue normoxic vs culture normoxic, tissue hyperoxic conditions (4 vs 21% O2). The benefit was significantly higher for the CD133−/low HT29 fraction, partly but not entirely compensating for the difference in clonogenic survival of CD133+ vs CD133−/low HT29 cells plating (Supplementary Figure 3).

We further raised the question whether the particular CD133 phenotypes are conserved throughout extended 2-D and 3-D culturing as well as in vivo. The distribution of CD133+ and CD133−/low fractions was thus investigated in more detail by flow cytometry in cells cultured for various times after FACSorting. There was no significant change in the CD133+ sorted fraction throughout 2-D culturing; this cell population remained positive for CD133. However, already at day nine in 2-D cultures, 81.1±3.5% of the cells derived from the CD133−/low HT29 population had to be classified as CD133+ (n=3). Ten days later (day 19 after FACS), the population had readjusted to the original distribution with only 4.8±2.2% HT29 cells appearing in the CD133−/low, isotype gate (Figure 3a). To prove that the unexpected redistribution of the CD133−/low-sorted cells in culture was not due to contamination by minor fractions of CD133+ HT29 cells, we also analyzed single-cell suspensions from two individual sorts already at day 4 after sorting and found 48.4% and 55.7%, respectively, of the cells derived from the CD133−/low subpopulation to express CD133 on the cell surface. The massive increase in the CD133+ fraction over a period of only 4 days, ie with a maximum of four possible culture doublings in 2-D culture (doubling time ∼22–24 h as documented earlier), proved that CD133−/low cells re-express CD133 on their surface. Similar changes were recorded in 3-D culture (Figure 3b). In two independent experiments, we found only 47.8% and 61%, respectively, of the cells to have maintained a CD133−/low phenotype in 4-day-old spheroids derived from the FACSorted CD133−/low HT29 population. The proportion of CD133−/low cells further decreased to reach 13.2±5.6% and 2.5±1.5% at days 9 and 19 after seeding (n=3) leading to the assumption that CD133 is rapidly re-expressed in both 2-D and 3-D culture. Six xenograft tumors with a diameter of 1.0–1.2 cm originated from 1 × 104 CD133−/low HT29 cells were extracted (day 30–40 after s.c. injection), dissociated and analyzed by flow cytometry to substantiate a potential redistribution in vivo. Indeed, the majority of the human cells in suspensions derived from 6/6 xenografts were positive for CD133. Three examples are shown in Figure 4b.

Figure 3
figure 3

CD133 distribution after seeding of CD133−/low FACSorted HT29 cells is readjusted during short-term 2-D and 3-D culturing. The distribution of CD133 fractions was monitored in 2-D (a) and 3-D (b) culture at days 4, 9 and 19 after FACSorting of the respective CD133+ and CD133−/low subpopulations. Original HT29 cells (o) as well as CD133+ (+) and CD133−/low (−) subpopulations directly after sort are documented in the left set of columns. Most notably, cultures derived from the CD133−/low population contained a high percentage of CD133+ cells already 4 days after FACSorting and reached >90% at day 19 after sort both in 2-D and 3-D culture. Representative flow cytometric dot plot diagrams of the seeded CD133−/low subpopulation after 4, 9, and 19 days in culture are shown in the lower panels. Single-cell suspensions were prepared and stained according to Figure 1; Propidium iodide (PI)-positive, membrane-defect cells and doublets were excluded.

Full size image
Figure 4
figure 4

Xenograft formation is not different or delayed in any of the CD133-related HT29 subpopulations and CD133 is re-expressed in xenograft tumors originated from CD133−/low FACSorted HT29 cells. (a) Proportion of palpable tumors as a function of time after injection of 100 or 10 CD133+, CD133−/low and original mixed HT29 fractions. (b) Dot plot diagrams documenting the gating strategy including FCS-A vs SSC-A, SSC-A vs SSC-W and CD326 vs propidium iodide (PI) (upper panel from left to right) to identify single membrane intact, human cells in xenograft preparations for CD133 surface expression analysis. Three representative CD133 histogram overlays of dissociated xenograft tumors originated from CD133−/low HT29 cells are shown (lower panel) to visualize the in vivo re-expression of CD133 relative to the respective isotype controls.

Full size image

A similarly rapid and massive change in CD133 expression profile throughout in vitro growth and also in in vivo xenografts was not seen in the HCT-116 CD133− fraction.21, 32 We therefore reassessed the findings in HT29 cell cultures. Flow cytometric analyses directly after separation of CD133+ and CD133−/low HT29 cells always showed two clearly distinct populations with negligible signal overlap (Figure 1b). However, we frequently observed a striking right shift of the fluorescence signal of the entire CD133−/low population relative to the respective sorting gate, which may not exclusively be attributed to methodological constraints. Western blot analyses revealed a clear enrichment of the respective CD133−/low and CD133+ HT29 populations upon sorting. However, an extremely faint CD133 signal was reproducibly seen in whole cell protein extracts from the CD133−/low-sorted HT29 subpopulation using two different antibodies directed either against the extracellular CD133 epitope AC133 or an intracellular C-terminal region of the CD133 molecule (Figure 5a). At this time of analysis it was not possible to verify if the signal came from a minor contaminating CD133+ cell fraction or if the entire CD133−/low HT29 population showed marginal CD133 expression. The contamination hypothesis, however, is unlikely in light of the rapid CD133 re-expression in the CD133−/low HT29 cell population throughout culturing, which could also be verified in protein extracts from 2-D cultures at days 4 and 14 after FACSorting (Figure 5b and c).

Figure 5
figure 5

Western blot analyses reveal rapid re-expression of CD133 in CD133−/low-sorted HT29 but not HCT-116 fractions during short-term culture. (a) Histogram overlay (upper panel) of flow cytometric CD133 fluorescence signals from FACSorted HT29 subpopulations and originally distributed cell line and representative western blots (lower panel) of whole cell protein extracts from the subpopulations using antibodies (1) W6B3C1 (directed against an extracellular epitope) and (2) Ab19898 (directed against a C-terminal, intracellular epitope) for CD133 detection; a weak CD133 signal in the CD133−/low HT29 subpopulation is seen. (b) Representative western blots using Ab19898 to determine CD133 protein in CD133+, CD133−/low and originally distributed HT29 populations that were seeded and cultured for 4 and 14 days. The respective HCT-116 subpopulations after 14 days in 2-D culture were analyzed in parallel. (c) Densitometric analysis of n=3 independent western blots and protein extracts, respectively, for each HT29 subpopulation using antibody (2) Ab19898 is shown. Relative mean values (+s.d.) are given with CD133 protein expression normalized to β-actin and originally distributed controls set to 100%.

Full size image

Based on this observation and the awareness of the discrepant binding (epitopes) of the different anti-CD133 antibodies, we reassessed the CD133 protein expression of several cell lines that were classified to contain CD133+ and CD133−/low fractions in flow cytometry. Exponentially grown SW1417 and LS1034 cells contain subfractions of 69.7±9.1% and 38.5±10.9%, respectively, with clear CD133 cell surface expression. Western blot analyses of whole cell protein extracts from the CD133−/low-sorted fractions of these two cell lines, however, identified CD133 protein bands according to HT29 cells (Sort layout, representative western blots and quantification of relative signal intensities are given in Supplementary Figure 4A–D. We used one of these cell models, LS1034, to verify by flow cytometry if the CD133−/low population contains intracellular CD133 protein. For this purpose, we established a co-staining for two CD133-specific antibodies (AC133 and W6B3C1) in viable and fixed, permeabilized cells. As expected, the CD133+ (AC133+) population was also positive upon staining with the W6B3C1 (data not shown). The CD133−/low population, however, was found to be negative for W6B3C1 only if cell membranes were intact, whereas cell permeabilization resulted in a clear positive CD133 protein signal in the majority of these cells (Supplementary Figure 4E). This reveals that CD133 cell surface presentation and intracellular protein can differ critically.

In contrast to the CD133−/low fractions in HT29, SW1417 and LS1034 cultures, CD133 protein in CD133−-sorted SW620 cells was at or below detection level in western blot analyses (Figure 6b). SW620 cells were therefore chosen as an additional model to evaluate if CD133+ and CD133−-sorted subpopulations from CRC cell lines differ in in vitro and in vivo behavior and to substantiate our finding that CD133 can be re-expressed on the surface of CD133−/low cell line populations. In exponential culture, 67.4±2.5% of the SW620 cells present CD133 on their surface; the subpopulations are clearly distinguishable (Figure 6a upper panel). Reanalysis of sorted subpopulations revealed high purities of ≥97%. No significant difference was found in colony formation capacity of CD133− vs CD133+ SW620 cells. Similar to the results with HT29, CD133− SW620 cells rapidly re-expressed CD133 protein on their surface in 2-D culture as documented in Figure 6d. The proportion of CD133-expressing cells increased from 9.3±4.3% at day 4 to 55.6±7.5% and 63.8±0.9% at days 9 and 18, respectively. In addition, in vivo tumor formation capacity of CD133+ and CD133− SW620 populations did not differ (Table 1c) and also no subpopulation-dependent delay in tumor formation was observed (Figure 7a). Roughly all (95.4±4.0% in n=10 xenografts) of the CD326-positive human cells from dissociated SW620 xenografts originated from the CD133− SW620 population re-expressed CD133 in vivo (examples are documented in Figure 7b). Hence, CD133 re-expression in the SW620 in vivo model was even more pronounced than in HT29 xenografts.

Figure 6
figure 6

Reanalysis of FACSorted SW620 CD133 subpopulations by western blotting reveal that CD133 protein is under the detection level in CD133− SW620 cells. Separated SW620 CD133+/CD133− subpopulations do not differ in colony formation capacity and rapid re-expression/upregulation of CD133 protein in the CD133− fraction was seen in 2-D culture. (a) Sort layout and representative reanalysis of SW620 cells, which exhibit clearly distinguishable subpopulations according to their CD133 expression. A purity >98% was reached after FACS. (b) Western blot analyses using whole cell protein extracts from separated subpopulations and (1) W6B3C1 or (2) Ab19898 anti-CD133 Ab for detection. Representative western blots and densitometric analyses (+s.d.) of (2) from n=3 independent experiments are shown. Analyses were performed according to Figure 5c. (c) Colony formation of CD133+, CD133− and original SW620 cells. (d) CD133+/CD133− ratios in SW620 cells directly after sort and in descendant 2-D cultures. Representative dot plot diagrams visualizing CD133 re-expression on days 4, 9 and 18 are documented (lower panel); analytical strategy and quantification according to Figure 3.

Full size image
Figure 7
figure 7

Xenograft formation is not different or delayed in any of the CD133-related SW620 subpopulations. CD133 is re-expressed in xenograft tumors originating from CD133−/low FACSorted SW620 cells and is also induced in xenografts of the CD133− cell line SW480. (a) Proportion of palpable tumors as a function of time after injection of 10 CD133+, CD133−/low and original mixed SW620 fractions. (b) Representative histogram overlays of three dissociated xenograft tumors originated from CD133− SW620 cells showing CD133 cell surface presentation relative to the respective isotype controls; analytical strategy as described in Figure 4b. (c) Histogram overlays according to (b) of three dissociated xenograft tumors originated from SW480 cells classified in vitro as CD133− cell line.

Full size image

Finally, we analyzed the CD133 expression pattern of dissociated xenograft tumors developed after the injection of SW480 colorectal cancer cells, which were negative for CD133 cell surface expression in vitro in 2-D culture and also showed a lower tumor formation rate in NMRI (nu/nu) mice than the SW620 cell line, which were derived from a lymph node metastasis of the same patient. Figure 7c documents the CD133 profile of three different xenografts showing a clear positive shift in the CD133 fluorescence and an increase in the fraction with a signal intensity above isotype. Taken together the re-expression or upregulation of CD133 in vivo appears to be a common phenomenon in CD133−/low MSS CRC cell lines and subpopulations.

DISCUSSION

Numerous studies indicate that the CD133-expressing fraction is a relevant prognostic factor for disease-free and overall survival33, 34, 35, 36, 37, 38, 39, 40, 41 and also relates to response to treatment.33, 39, 41, 42, 43, 44, 45 However, there has been an ongoing discussion on the usefulness of CD133 on the one hand and established CRC cell lines on the other hand as tools in cancer stem cell research and treatment. Indeed, the functional and causal relation between CD133 protein expression and putative cancer stemness and plasticity is still a matter of debate.6, 10, 46, 47, 48, 49, 50, 51 This prompted us to further investigate the relevance of CD133 expression in CRC cell lines, which are integral to many drug development test platforms.

The decision to examine HT29 cells reflecting the more common MSS pathway to colorectal carcinogenesis as opposed to the MSI cell line HCT-116 studied earlier21 was primarily driven by literature data indicating a difference in tumorigenic potential of HT29 CD133+ vs CD133− subfractions.22 In addition, a decrease in CD133 expression in HT29 populations with a higher grade of differentiation was reported,18 and CD133 expression was found to be enhanced in this cell line upon irradiation.43 Furthermore, CD133 knockdown was shown to result in greater susceptibility to staurosporine-induced apoptosis and reduced cell motility.29 Taken together, these findings suggested that CD133 has a role in response to treatment and/or CSC behavior of HT29 cells. A literature review, however, also indicated considerable differences in CD133 expression levels in HT29 cultures in various laboratories. This may be attributed to (i) different growth and environmental conditions, (ii) selection of particular HT29 clones, or (iii) different staining protocols, detection technologies and analytical tools.18, 22, 26, 27, 28, 29, 30 Because CD133 expression was found to be elevated upon infection with Mycoplasma hyorhinis,28 our cell lines were routinely tested negative for mycoplasma and verified for genetic origin to exclude artificial modifications.

The 92–110 kDa glycoprotein CD133 (prominin-1) consists of an extracellular N-terminal domain with two large glycosylated extracellular loops, five transmembrane domains and a cytoplasmic C-terminus.52, 53, 54, 55 The anti-CD133/1 (clone AC133) has most frequently been applied to isolate cells with different CD133 surface presentation and also served as the tool in our study.21 The antibody was long thought to bind to an extracellular glycosylated epitope54, 56 and false-positivity/negativity has been supposed to be due to the presence of alternatively spliced isoforms, truncated proteins or others.52 However, the hypothesis that AC133 recognizes a glycosylated epitope was questioned,27 and Kemper et al50 recently showed that it binds to a non-glycosylated epitope on the second extracellular loop at exons 16–24. Because of these controversial issues and because protein processing or masking of epitope by protein folding may be of relevance for antibody binding capacity,50, 57 it is advisable, but not yet routine, to confirm flow cytometric data with independent protein detection methods and/or other anti-CD133 antibodies. We performed western blot analyses of whole cell protein extracts from subpopulations of interest using the AC133 as well as an alternative antibody clone, which binds to an intracellular C-terminal epitope to determine CD133 protein independent of its cellular location. In contrast to HCT-116, we frequently found a weak CD133 protein band in the CD133-negative-sorted HT29 cell subpopulation and also in the respective CD133− SW1417 and LS1034 cell fractions. This, together with the slight shift in the flow cytometric signal of the CD133−/low-sorted HT29 population to higher mean fluorescence values, and the finding that CD133 cell surface presentation and intracellular protein can critically differ—as shown herein via flow cytometry using the LS1034 cell model—implies that CD133−/low cells may indeed express CD133 protein. The plasticity of CD133−/low HT29 cells with respect to CD133 protein expression is further confirmed by the rapid re-expression of CD133 on the surface of this subpopulation during continued tissue culture and may explain the transient phenotype of CD133+ vs CD133−/low HT29 fractions and lack of difference in tumorigenic potential.

SW620 cells that exhibit clearly distinguishable CD133− and CD133+ subpopulations in vitro as verified in western blot analyses were included in our study because a CD133+ SW620 subfraction was also reported earlier to result in faster tumor formation and larger xenografts after subcutaneous or orthotopic injection into NOD/SCID or nude mice.23, 58 Our data, however, reveal, that CD133− and CD133+ SW620 subpopulations neither differ in clonogenic survival in vitro nor in tumor formation capacity in NMRI (nu/nu) mice. And, the CD133− population of this cell line also appeared to rapidly re-express CD133 protein for presentation on the cell surface, a modulation particularly induced after in vivo engraftment. Interestingly, even the CD133-negative cell line SW480 showed an increase in CD133 signal in xenograft tumors. From these data and because xenografts derived from CD133− HCT-116 cells also contained variable CD133+ fractions in spite of lack of rapid re-expression in vitro, we conclude that CD133 presentation in CRC cell lines is highly plastic independent of the underlying carcinogenesis pathway and common genetic alterations (see Table 3 and Supplementary Table 1). We did not find the CD133 in vitro profile of any of the cell lines under standard conditions to be associated with tumor-propagating potential but all cell lines and cell line populations studied were to some extent capable to (re-)express CD133 protein in vivo.

Table 3 Overview of data on cell line characteristics and behavior related to CD133 (AC133) cell surface expression
Full size table

The reason(s) for the significant and reproducible difference for CD133+ vs CD133−/low HT29 subpopulations with respect to colony and spheroid formation still remains elusive. In brain tumors, the distribution of CD133-positive vs -negative subpopulations was found to be modulated by the oxygen level59, 60, 61, 62 and hypoxic niches are discussed to support stemness phenotype in both normal tissues and solid cancers.63 In the present study, oxygen concentration is unlikely to cause rapid CD133 protein re-expression in HT29 2-D cultures. It may be speculated though, that such effects may further contribute to the reappearance of CD133+ populations in vivo and the lack of a non-tumorigenic CD133− phenotype because tissue normoxic conditions already partly neutralized the in vitro survival advantage of the CD133+ HT29 population. The impact of therapeutically relevant hypoxia and intermittent oxygen supply due to the chaotic, immature vascular network in solid tumors in the scenario of CD133 expression and plasticity is of particular interest in this context and subject to further systematic study.

Recently, a cell-density-related upregulation of CD133 expression in CaCo2 cells in vitro was described.64 We observed similar changes in SW620 cells. When CD133−-sorted SW620 cells were passaged according to a standard protocol and kept in an exponential state over a culture period of 18 days, the proportion of CD133+ cells increased to 14–20%, whereas without sub-culturing 63.8±0.9% of the cells became positive on their cell surface (Supplementary Figure 5). Such effect is likely to also contribute to the upregulation of CD133 in some xenografts but may be less relevant in others, eg CD133− HCT-116 populations did not become positive with increasing density over a period of 2 weeks in culture. Interestingly, 95% of the FACSorted CD133+ SW620 cells maintained their positivity independent of passaging, which implies that CD133 expression is density-dependent only in a subfraction of the cell line.

Analogous studies have to verify, if other biomarkers discussed with CRC stemness phenotype such as CD44 (refs. 65, 66) are also affected by such environmental constraints. The cell lines used herein clearly differ in their CD44 cell surface profile under identical conditions in exponential 2-D culture (Supplementary Figure 6) and some of them could be models of interest in this context.

Taken together, our data show that despite a potential phenotype in vitro, reduced or lack of CD133 presentation neither relate to loss of tumorigenic potential in the CRC cell lines studied herein nor does it correlate with a different response to treatment, most probably because the majority of the MSS cells is capable of re-expressing CD133 on the cell surface. These cell lines may in principle possess a cancer stem-like phenotype and the plasticity required to survive in a (patho)physiological in vivo environment. Because of their particularly high tumor take rate, they are poor models for scrutinizing the cancer stem cell hypothesis. However, they are still good tools for treatment testing in sophisticated culture systems or in vivo models reflecting the heterogeneous tumor environment.