Main

Colon cancer is the third leading cause of cancer-related deaths in the United States.1 A subpopulation of cancer stem cells (CSCs), residing within the tumor bulk, are believed to be resistant to radiation and chemotherapeutic treatments and are likely responsible for recurrence of the disease in cancer patients.2

A number of non-invasive diagnostic assays/procedures are being developed in order to either diagnose patients with metastatic disease and/or to screen patients for relapse after treatment. A novel diagnostic tool, developed in recent years, consists of measuring the rare circulating tumor cells (CTCs) in the blood of patients for detecting early stage metastasis or recurrence of cancer in treated patients. This assay is based on the notion that metastatic cancer cells, going through epithelial–mesenchymal transition (EMT), invade surrounding stroma and lymphatic/blood vessels by intravasation. The tumor cell emboli move through the vasculature to either lymph nodes or other organs and initiate tumor growth in secondary sites by extravasation.3, 4 Therefore, detection of CTCs in circulation is diagnostic for on-going metastasis and/or relapse.5

However, developing reproducible and accurate methods for detecting and isolating CTCs has remained a challenge. Epithelial cell membrane markers used for identification and separation of the rare population of CTCs may provide erroneous results, as CTCs undergoing EMT during metastasis are downregulated for the expression of epithelial cell markers.6 The rare CTCs represent a minute percentage of total blood cells in circulation, ranging from only 1 to 100 CTCs/ml blood, among >109 red and white blood cells.7, 8 Thus, a sensitive bioassay is required for accurately detecting and isolating all CTCs/circulating CSCs (CCSCs), while excluding contaminating white blood cells.

We now know that CSCs are critically required for initiating and maintaining the growth of primary and metastatic tumors.9, 10, 11 Here, we present, for the first time, a method for identifying CCSCs in the blood of colon cancer patients, using a combination of epithelial cell markers and CSC markers, such as DCLK1,12, 13, 14 LGR515, 16 and CD44.14, 17 Epithelial cell surface marker CK1918, 19 and epithelial cell adhesion membrane protein, EpCAM,20 were used to confirm epithelial origin of the isolated CCSCs; combining surface markers has been shown to increase accuracy of detecting CTCs.18, 19

Progastrin (PG) peptide is expressed by a majority of human colon cancers (reviewed in Singh et al21). Cell surface Annexin A2 (CS-ANXA2) represents a non-conventional receptor for PG22 and is required for mediating mitogenic/co-carcinogenic effects of PG on target cells.23 Both PG and CS-ANXA2 regulate tumorigenic and metastatic potential of colon cancer cells.14, 21, 22, 23 CS-ANXA2 is increasingly expressed by aggressive epithelial tumor cells and is required for metastasis of many cancer cells.14, 21, 22, 23, 24, 25, 26 Overexpression of PG by embryonic epithelial cells transforms the cells, associated with a significant increase in the population of CSCs, co-expressing stem cell markers DCLK1/CD44 with CS-ANXA2.14 Co-expression of DCLK1 with CD44/CS-ANAX2 may dictate tumorigenic/metastatic potential of CSCs.14 Therefore, PG and CS-ANXA2 were also used as novel markers of CCSCs, in combination with CSC and epithelial cell markers, in order to develop a more robust and accurate assay for identifying CCSCs.

Using the methods and markers described here, we report the development of relatively simple assay(s) for detecting CCSCs in the blood of colon cancer patients. CCSCs isolated from the blood of patients, positive for colonic adenocarcinomas (AdCAs), formed tumorospheres in vitro, confirming that the isolated CCSCs represent tumor-initiating cells, which can potentially grow into metastatic tumors in vivo. Tumorospheres generated from CCSCs can be used in the future to test the efficacy of specific inhibitors, including inhibitors of stem cell markers, for eliminating CCSCs, and hence metastatic tumor growths, in order to improve clinical outcomes for the patients, as an approach towards personalized medicine.

MATERIALS AND METHODS

Reagents Used

Antibodies used include: anti-CD44 (Cell Signaling Technology, Danvers, MA, USA); anti-DCLK1, anti-CD45, anti-EpCAM, anti-CK19 and anti-GPCR GPR49 (Lgr5) (Abcam, Cambridge, MA, USA); anti-AnxA2 (BD Biosciences, Carlsbad, CA, USA), and anti-β-actin (total) (Sigma, St Louis, MO, USA). Anti-PG antibody was generated in our laboratory as described27). Alexa Fluor-594 and Alexa Fluor-488 coupled secondary IgG were from Invitrogen (Carlsbad, CA, USA). Three kits from StemCell Technologies were used: (1) RosetteSep Human CD45 Depletion Cocktail (no. 1522), (2) RosetteSep Human Circulating Epithelial Tumor Cell Enrichment Cocktail (no. 15127), and (3) EasySep Human Whole Blood CD45 Depletion Kit (no. 18289), (Vancouver, Canada).

Growth of Primary and Metastatic Tumors from HCT-116 Cells in Athymic (SCID/nude) Mice

Sub-confluent HCT-116 cells in culture were processed for inoculating athymic nude mice as described previously.14 To induce the growth of sub-dermal xenografts (primary tumors), 5 × 106 HCT-116 cells/100 μl phosphate-buffered saline (PBS) were inoculated on the right and left flanks of female mice (Harland Sprague Dolly). For developing metastatic tumors, athymic female nude mice (5–6-weeks old) were subjected to intrasplenic inoculations with 2 × 106 HCT-116 cells/50–100 μl PBS as described previously.14 Mice receiving intrasplenic inoculations were subjected to splenectomy after 24 h of inoculation to avoid splenic/peritoneal growths. Mice in group I were inoculated with heat-inactivated colon cancer cells and served as the control group, with no tumor growths. Mice in group II were inoculated subcutaneously to develop primary tumor xenografts. Mice in group III were inoculated intrasplenically in order to develop metastatic tumors in the liver and lung, as described previously.14 Approximately 4 weeks postinoculation, when sub-dermal xenografts were clearly visible in the mice from group II (∼750 mm3 in size), mice in all three groups were anesthetized for blood collection as per our approved IACUC protocols and euthanatized by CO2 asphyxiation. The size of sub-dermal tumors (group II) and the total number of metastatic lesions in the liver of each mouse in group III was similar to the numbers previously reported.14 Approximately 1-ml blood was collected from one to two mice in a lithium heparin tube, representing one sample/group. Three samples of blood were collected per group from five mice/group and processed for isolating CTCs as described below.

Procurement of Blood Samples from Patients

‘Normal’ blood samples were obtained from consented UTMB patients scheduled for screening colonoscopy, aged ≥50 years, irrespective of gender and ethnicity, as per our approved IRB Protocol. Blood was also obtained from consented patients scheduled for surgical removal of their AdCAs and lymph nodes at UTMB. In all, 20–30 ml of blood was collected at the time of intravenous line placement, as per our approved IRB Protocol. Consent was obtained from alert patients, who had not received any medication as yet. Clinicopathological information of patients enrolled in this study is provided in Supplementary Figure S1. Blood samples were collected from seven patients with CRCs and seven patients with ‘normal’ colons, with no adenocarcinomas or adenomas but only hyperplastic polyps. Blood samples, in each case, were divided into three samples and processed separately as triplicate samples. Samples in triplicate from all the patients were subjected to CCSC analysis. Samples from two to four patients were also used for comparing the different isolation kits, recovery analysis and spheroid growths as indicated in the legend of the figures.

Generation of GFP-Expressing HCT-116 Colon Cancer Cells (HCT-116-GFP)

Preparation of lentiviral plasmids expressing eGFP

Plasmids encoding eGFP in lentivirus packaging plasmid was generated by amplifying the gene by PCR. The plasmid pEGFP-C1 (Clontech; Mountain View, CA, USA) was used as a template. Primers used contained the sequences for flanking restriction endonucleases (BamHI and ApaI), and the product was ligated into the lentivirus packaging plasmid, pLenti6 (Invitrogen), to create pLenti6-GFP constructs, which was confirmed by DNA sequencing in the recombinant core facility at UTMB.

Generation of stable clones of HCT-116-GFP cells

HCT-116 cells were obtained in late 1980s from Dr Michael Brattain’s laboratory and have been maintained in our laboratory since then. The HCT-116 cells were recently authenticated by BioSynthesis DNA Identity Center in 2012. The authenticated HCT-116 cells were grown in DMEM medium supplemented with 2.0 mM L-Glutamine (Invitrogen), penicillin/streptomycin (Invitrogen) and 10% heat-inactivated fetal bovine serum (HyClone; Logan, UT, USA). Sub-confluent HCT-116 cells were transfected with pLenti6-GFP plasmid. Stable clones of HCT-116 cells, expressing GFP, were generated by drug selection, as previously described.14 The drug-resistant GFP-expressing HCT-116 colonies were retrieved and expanded. The clones were further FACSorted to isolate HCT-116 cells expressing high levels of GFP, which were then maintained in drug selection medium in culture.

Recovery of Colon Cancer Cells from the Human Blood, Spiked with HCT-116-GFP Cells

HCT-116-GFP clones, growing in culture, were collected into centrifuge tubes by scraping. An aliquot was used for confirming viability of cells to be >99% by trypan blue exclusion method using Cellometer Auto T4 (Nexcelom Bioscience, Lawrence, MA, USA). Blood from patients who were either negative for colonic tumors, other than small hyperplastic polyps (termed normal samples) or who were positive for AdCAs, were spiked with 0-500 HCT-116-GFP cells/ml blood, within 2–4 h of collecting the samples. The spiked blood was gently mixed manually and subjected to negative selection for RBCs using the HetaSep kit (as described in the legend of Supplementary Figure S2). Blood free of RBCs was subjected to further negative selection for CD45+ WBCs using the EasySep kit, as per the manufacturer’s recommendations. Human blood samples thus processed for negative selection of RBCs and WBCs were cytospun at 200 g for 5 min onto Superfrost/Plus B microscope slides (Fisher Scientific, Pittsburgh, PA, USA) using a CytoSpin III cytocentrifuge (Cheshire, England), followed by processing the cells for IF staining for the indicated markers; DAPI (4′,6-diamidino-2-phenylindole; Sigma) was used as a counterstain for identifying nucleated (epithelial/WBC) cells. HCT-116-GFP+ cells were further confirmed by measuring green fluorescent GFP-expressing cells, which allowed us to calculate the percentage of recovery of the spiked cells from the patient blood samples.

Isolation and Characterization of CTCs from the Blood of Mice and Humans

Mice

Blood samples collected from mice were centrifuged at 200 g for 5 min, giving rise to three layers: (1) pelleted red blood cells, (2) buffy coat containing epithelial cells, and (3) supernatant containing WBCs. Buffy coat and supernatant were analyzed together for total number of CTCs/ml, by subjecting the samples to negative selection using the EasySep kit, in order to eliminate WBCs from the samples. The elute was cytospun onto microscope slides using cyto-centrifuge, and slides were processed for immunostaining with mouse antibodies against CD45 and stem cell markers; nucleated cells were stained with DAPI.

Determination of the percentage of cells expressing CD45 in the three layers of mouse blood by FACS analysis

Cells present in the three layers of blood after centrifugation at 200 g were subjected to FACS analysis to identify the layer in which the majority of WBCs, positive for CD45, were located. Cells within the indicated layers were carefully re-suspended at 1 × 106 cells/ml of PBS with 2% BSA at room temperature for 1 h each. The cell suspensions were then incubated with anti-CD45-antibody (1:100) tagged to fluorophore Alexa-488 (DyLighd 488 NHS-Ester Die Light 488 NHS-Ester) for 2 h at 4 °C on a gently rocking platform. Cells were then pelleted at 200 g and washed 3 × with PBS containing 2% BSA to remove unbound primary antibody, followed by FACS analysis using FACSariaII (Carlsbad, CA, USA) in the core facility at UTMB Health, as previously described.14

Isolation and characterization of CTCs from human blood

Blood (∼30 ml) collected from patients was subjected to negative selection for RBCs and WBCs using three separate kits from Stem Cell Technologies, as per the manufacturer’s recommendations. The combined use of HetaSep and EasySep Kit gave the best results and was used to analyze the blood from all patients. Blood plasma samples negatively selected for RBCs and WBCs were processed for IF staining with the indicated antibodies and co-stained with DAPI. Images of IF staining were acquired using an epifluorescent microscope, and cells were analyzed using the Metamorph, V6.0 software (Molecular Devices), as described previously.14, 28

Growth of CTCs as Primary Spheroids In Vitro

In a separate set of experiments, CTCs isolated from the blood of patients positive for AdCAs were subjected to negative selection for RBCs/WBCs and plated to grow primary spheroids in low-attachment plates using the serum-free spheroid assay buffer as described previously.14, 28 Blood samples collected from patients free of colonic growths were similarly processed. The spheroids were imaged daily at 4 × and 40 × magnification using a white light microscope (Nikon Instruments, Melville, NY, USA). At day 25, spheroids were processed for western blotting (WB).28 Blots were cut into horizontal strips containing either the target or the loading control protein (β-actin) and processed for detection of the antigen–antibody complexes by chemiluminescence.14, 28 Membrane strips containing target/loading control proteins were simultaneously exposed to autoradiographic films. The loading control, β-actin, was measured in corresponding samples containing equivalent protein. Relative band density on scanned autoradiograms was analyzed using the Image J program (rsbweb.nih.gov/ij/download) and was expressed as a ratio of the target protein to β-actin in the corresponding sample.

Statistical Analysis of Data

Quantitative analysis of data is presented as mean±s.e.m. of values obtained from the indicated number of samples in each experiment. To test for significant differences between values obtained from normal vs CRC samples, nonparametric Student’s T-test and/or Mann–Whitney test was employed using the GraphPad Prism software (La Jolla, CA, USA); P-values <0.05 were considered statistically significant.

RESULTS

Detection of CCSCs in the Blood of Athymic Nude Mice Bearing Metastatic Colon Cancers

Athymic nude mice (5 mice/group) were inoculated with HCT-116 cells as described in the Methods section. Blood collected from all the three groups was centrifuged and FACSsorted as diagrammatically presented in Figure 1a. Population of CD45+/−FACSorted cells in supernatant+buffy coat and in RBC pellet are shown as a forward scatter plot in Figure 1b; the average percentages of CD45+ cells in the fractions is presented in Figure 1a. Majority of CD45+ (>98%) and CD45− (>99%) cells were present in the supernatant+buffy coat and RBC pellet layers, respectively. A small percentage of cells in the supernatant+buffy coat fraction were CD45− (1.1%), which likely represents CTCs, as reported by others.29, 30 CD45− cells from supernatant+buffy coat layers were cytospun on slides and processed for IF staining for CSC markers (DCLK1/CD44/Lgr5) and ANXA2 (Figure 1c). Approximately 1.5–3% of CD45− cells in the buffy coat+supernatant layers of plasma from the group III mice expressed DCLK1, CD44, Lgr5 and ANXA2 (Figure 1c). In contrast, <0.5–1% of CD45− cells in the plasma of mice in groups I and II were positive for the indicated markers (Figure 1c). A slightly higher percentage of CD45− cells (∼0.7–1%) in groups I/II expressed CD44 and ANXA2, compared with stem cell markers DCLK1/Lgr5 (Figure 1c). The remaining CD45− cells (>97%) likely represent CTCs, which are not CCSCs. Some of the ANXA2+/CD44+ cells may also represent contaminating CD45+ cells in these fractions, as negative selection for WBCs is not 100% efficient. CD45+ cells are known to express CD44 and ANXA2, as previously reported.31, 32, 33 We have previously reported co-expression of stem cell marker DCLK1 with CD44/ANXA2 by human colon cancer cells as a marker of transformation/metastatic potential.14 CD45− cells in the plasma of group III mice were found to similarly co-express DCLK1/ANXA2 and DCLK1/CD44 (Figure 1c, right panels), confirming that CSCs in circulation maintain the malignant phenotype of CSCs reported in primary tumors.14

Figure 1
figure 1

Detection of metastatic circulating tumor cells in the blood of athymic nude mice. (a) Diagnostic scheme for isolating nucleated epithelial cells from the blood of mice. Three samples of blood/mouse group were collected and analyzed by FACSorting for percentage of cells positive or negative for CD45 in the supernatant/buffy coat and red blood cell layers. CD45− cells isolated from the supernatant/buffy coat layer are stained by IF staining for the indicated markers. (b) Representative data from FACSorting of the cells into CD45+/− populations, showing forward scatter plots for the two fractions of mouse plasma from GroupIII mice. (c) Bar graphs depicting percentage of cells positive for the specific markers/ml blood, collected from all three groups. Images of CD45− cells stained by IF for DCLK1/ANXA2/CD44/LGR5 markers are shown above the bar graphs, and co-expression is depicted as merged images on the right. Mean±s.e.m. of data obtained from ∼10 slides/3 samples/group. *P<0.05 vs normal values.

Full size image

Recovery of Colon Cancer Cells from Human Plasma

The efficiency of EasySep human whole blood CD45-depletion kit for recovering colon cancer epithelial cells from human plasma samples was analyzed by spiking the samples from normal and colon cancer patients with HCT-116-GFP cells. Majority of the HCT-116-GFP cells (>97%) were recovered from the spiked plasma samples using the kit (Supplementary Figure S2). The number of nucleated cells (epithelial+CD45+) recovered from spiked plasma was higher than the number of colon cancer cells used for spiking. The latter increase was due to contamination with CD45+ WBCs (Supplementary Figures S2i and ii); however, none of the GFP+ cells co-stained for CD45 (Supplementary Figures S2i and ii).

Percentage of Contamination of Processed Plasma Samples with CD45+ cells in the Final Elute

On an average, 82±10 CD45+ cells/ml were present in the sample from cancer patients, while 70±10 CD45+ cells/ml were present in the sample from normal patients (Figure 2a), suggesting that a minute population of CD45+ WBCs remains as a contaminant in the final elute. As CD44 is also expressed by CD45+ WBCs,31 on an average, 50 and 85 CD44+ cells/ml were present in the blood of normal and AdCA patients, respectively (Figure 2b). Interestingly, ∼98% of CD44+ cells in the blood from normal patients co-stained with CD45, suggesting that almost all the CD44+ cells present in the blood of normal patients were WBCs (Figure 2c). However, only ∼70% of CD44+ cells co-stained with CD45 in the blood of cancer patients (Figure 2c), suggesting that ∼30% of CD44+ cells circulating in the blood of colon cancer patients represents CCSCs. Numerically, an average of 9 cells/ml were CD44+/CD45− cells in AdCA patients, while 0 cells/ml were CD44+/CD45− in normal patients (Figure 2d).

Figure 2
figure 2

The presence of CD44+/CD45− cells in the blood of patients with colon-adenocarcinomas. Bar graphs illustrating: (a) The number of recovered CD45+ cells/ml in the blood of patients, who were either free of colonic growths (normal) or positive for adenocarcinomas (AdCA); IF images of remaining CD45+ cells collected in the final elution are shown above the bar graphs. (b) The number of CD44+ cells/ml in the blood of normal vs AdCA patients. (c) The number of cells co-localizing for CD45/CD44 per ml of blood. Representative IF staining of circulating cells stained with the indicated markers are shown above bar graphs, after negative selection of the samples. (d) The number of CD44+ cells/ml, which do not co-localize with CD45 in the blood of normal vs AdCA patients; representative IF images are shown above bar graphs. Each bar graph=mean±s.e.m. of data from 5 to 7 patients analyzed in triplicate as described in the Methods section. *P<0.05 vs normal values.

Full size image

CCSCs in the Blood of Normal and Cancer Patients

The number of cells positive for DCLK1, CD44, Lgr5, ANXA2, PG, EpCAM and CK19 were significantly higher in the blood of AdCA patients compared with that in the blood of normal patients (Figure 3a). Besides CD44, a minute subpopulation of leukocytes also expressed DCLK1/ANXA2 (Figure 3b), suggesting for the first time that CD45+ cells can potentially express DCLK1. A subpopulation of CD45+ peripheral blood monocytes co-express ANXA232, 33 and CD44.31 However, majority of DCLK1/ANXA2-positive cells in the final elute of plasma samples from AdCA patients were negative for CD45, suggesting that these markers are largely expressed by epithelial CCSCs in CRC patients. Comparative tables, enumerating average number of CTCs/ml sample from normal vs cancer patients, with or without contaminating CD45, are presented in Supplementary Figures S3A and B.

Figure 3
figure 3

A sub-population of leukocytes is also positive for DCLK1, ANXA2 and CD44. (a) Blood from normal and AdCA patients were processed with the EasySep kit, catalog no. 18289. Cells collected after final elution were cytospun onto glass slides, followed by IF staining for the indicated markers. Based on the staining results, the number of cells/ml of blood positive for the indicated markers was determined, as shown in the bar graphs. (b) Representative IF images showing co-expression of DCLK1/CD45, CD44/CD45 and ANXA2/CD45. Images were taken at × 10 magnification, and a single cell image was magnified and is presented in an inset. (n=7 normal samples and n=7 AdCA samples).

Full size image

Representative IF images of CTCs in the final elute of plasma samples from AdCA patients, stained for the indicated markers, are shown in Figure 4a. Number of cells/ml sample that were positive for the indicated markers in the absence of CD45 staining are presented in Figure 4b. After excluding CD45+ cells, normal samples were almost completely devoid of CTCs. However, plasma samples from AdCA patients were positive for a significant number of CTCs, which stained for the indicated epithelial and stem cell markers (Figure 4b). Majority of CTCs were positive for CD44 (∼10 cells/ml) while only a minority were positive for EpCAM (1.3 cells/ml), providing further evidence that majority of CTCs, including CCSCs, in cancer patients are downregulated for the expression of the adhesion protein, EpCAM. The majority of CD44+/CD45− cells were positive for epithelial cell marker CK19 (∼7 per 10 CD44+ cells/ml) suggesting that CK19 is a better marker for epithelial tumor cells, but CK19 may also be underexpressed by a sub-population of CTCs.29 On an average, ∼5, 10, 4, 5, 4 and 7 CTCs/ml blood from AdCA patients expressed DCLK1, CD44, Lgr5, ANXA2, PG and CK19, respectively (Figure 4b).

Figure 4
figure 4

Circulating cancer stem cells (CCSCs) in the blood of patients. (a) Representative IF images of circulating cancer stem cells, positive for the indicated markers. Images were digitally enhanced, as shown by dotted arrows. (b) Bar graphs illustrating the number of cells/ml, positive for the indicated markers, excluding CD45+ cells. (n=7 normal samples and n=7 AdCA samples).

Full size image

CCSCs in Blood Samples from AdCA Patients Co-Express Markers for Stem Cells (DCLK1/Lgr5) with CD44/ANXA2, Representing a Malignant Phenotype

After negative selection of plasma samples for RBCs/WBCs, remaining cells in the final elute were co-stained with antibodies against specific markers. IF images from representative AdCA samples are presented in Figures 5a–c. Total number of cells that co-stained with the indicated markers/ml of sample from several patients is presented as bar graphs in Figure 5d. A comparative table enumerating average number of CTCs/ml co-expressing the specific markers from normal vs cancer patients is presented in Supplementary Figure S3C. Normal blood samples were largely negative for cells that co-stained with the indicated markers. However, <0.5 cells co-stained with DCLK1 and ANXA2/CD44 in normal plasma samples suggesting lingering contamination with non-epithelial cells, as described above. Plasma samples from AdCA patients, on the other hand, were positive for a significant number of cells (∼1–6/ml) that co-stained with the indicated markers. Co-staining of cells with stem cell markers and EpCAM continued to be the least robust marker. Interestingly, a significant number of stem cells (DCLK1+) and ANXA2+ cells, co-stained with colon cancer cell marker, PG, providing a unique marker for colon CSCs in circulation. To confirm the epithelial nature of CCSCs, cells were co-stained with CK19, a well-established epithelial marker for CTC recognition18, 34 (Figure 5c; Supplementary Figure S4).

Figure 5
figure 5

Co-expression of CSC markers (DCLK1/LGR5) with AnxA2/PG/CD44/EpCAM/CK19 by CCSCs in the blood of colon cancer patients. (a) Co-expression (IF staining) of DCLK1+/LGR5+ cells with EpCAM, CD44 or ANAX2 in AdCA samples; white arrows depict stained cells. (b) Co-expression of DCLK1/PG or ANXA2/PG in CCSCs. (c) CCSCs co-expressing CK19 with either ANXA2, DCLK1 or LGR5. Merged images are enhanced in the inset. (d) Bar graphs illustrating the number of CCSC cells co-expressing the indicated markers in per ml of blood. (n=7 normal samples and n=7 AdCA samples).

Full size image

CCSCs Likely Represent Tumor-Initiating Cells

Cells in the final elute were seeded in low-attachment plates to grow as spheroids, to test stemness of the cells (as an indirect indicator of tumor initiating potential). By day 14, epithelial cells enriched from the blood of AdCA patients started growing as distinct spheroids, whereas cells remaining in the elute of normal patients did not form spheroidal structures (Figure 6a). On day 25 of culture, all cells were collected from normal and AdCA plates, pelleted by centrifugation and processed for western blotting analysis. The spheroidal cells from AdCA patients were positive for Lgr5/DCLK1, whereas normal samples were negative (Figures 6bi and ii), confirming that normal blood lacks CCSCs, whereas AdCA samples are positive for CSCs, which likely have the potential for initiating metastatic tumors in vivo.35

Figure 6
figure 6

Formation of spheroids, in vitro, from circulating tumor cells, isolated from AdCA patients. (a) Representative images of spheroids formed from enriched circulating epithelial cells, isolated from the blood of normal and AdCA patients. Images were taken at × 4 and × 40 magnification at days 3, 14 and 25 after seeding the cells. (bi) Western blotting analysis demonstrating the increased expression of LGR5 and DCLK1 in spheroids from CTCs isolated from a representative AdCA patient, compared with that obtained from a representative normal patient. (bii) Bar graphs demonstrating the percentage of change in ratio of target proteins to β-actin. (n=3 normal patient samples; n=3 AdCA patient samples).

Full size image

DISCUSSION

The presence of CTCs in CRC patients has been reported to be associated with poor prognosis for survival/recurrence, irrespective of chemo/radiation therapy.36, 37, 38 Several CTC assays have been developed in the past, all of which use some combination of negative selection (removal of red and white blood cells) and positive selection (cells positive for epithelial and/or intracellular stem cell marker, such as ALDHA1).4, 8, 18, 39 CD133 has also been used for positive selection.39 However, stem cell markers, DCLK1/Lgr5, have not been used previously for detecting circulating CSCs. As DCLK1 has emerged as a specific marker of colon/pancreatic CSCs,13, 14, 28, 40, 41 we report for the first time detection of DCLK1+ CCSCs in the blood of CRC patients. In addition, we used the metastatic marker, AnexinA2. Thus the markers used in this study, for detecting CCSCs, are expected to be more valuable for diagnosis/prognosis of recurrence/relapse of the disease and for perhaps detecting even primary CRCs with metastatic potential.

Several laboratories have tested novel techniques to negatively select for leukocytes (WBCs) in the peripheral blood; however, due to the rarity of CTCs in the blood and the inability to entirely remove the abundant WBCs from blood specimens, this approach has remained challenging.42 Size-based purification,43 high-speed microscopic scanning of nucleated tumor cells44 and in vivo multi-photon imaging of CTCs45 also present technical challenges resulting in low CTC purity. We first optimized and standardized a relatively simple assay for measuring CCSCs. Our results suggest that negative selection for RBCs (using HetaSep solution) and WBCs (using the EasySep kit) followed by positive identification of CCSCs by immunostaining may represent a relatively simple and inexpensive method. Although the optimal assay used by us yielded the presence of <0.0015% CD45+ cells after negative selection, we were able to discount them by using a combination of leukocyte and CSC/epithelial cell markers.

ANXA232, 33 and CD4431, 46 have been previously reported to be expressed by peripheral blood monocytes; however, the expression of DCLK1 by CD45+ cells has not been reported and represents a new and novel finding of the current study. Majority of cells expressing CSCs/ANXA2 were CD45− (Figure 3b). Therefore, in order to distinguish rare CCSCs from white blood cells, negative selection by staining with anti-CD45 antibody needs to be added as a final step for most CTC assays, in order to eliminate false-positive results. We also analyzed isolated cells for EpCAM/CK19 staining to confirm the epithelial nature of isolated CCSCs. EpCAM is commonly used as a marker for enumerating CTCs and is generally absent in normal blood cells.4, 47 EpCAM has been previously reported to be expressed by many epithelial cancers, including that of the lung, gastric, breast and colon.48, 49 However, our results showed that EpCAM was not a robust marker for detecting CTCs in the blood from CRC patients (Figures 4 and 5). Several recently developed CTC-enrichment technologies, such as microfluidic-based CTC capture4 and independent density-gradient centrifugation,50 have utilized EpCAM as a marker of CTCs. However, the reliability of using EpCAM as a robust marker has been questioned, as this adhesion protein is downregulated during invasion and dissemination.51 The metastatic epithelial tumor cell likely looses adhesion due to downregulation of EpCAM and undergoes EMT.52 EpCAM antibodies were reported to be insufficient for detecting subtypes of breast cancer cells,53 as confirmed in our studies with CRC patients (Figures 3a and 4b). Thus antibodies against several cell surface antigens needs to be used for accurate detection of all epithelial CTCs. The latter point is especially important, as epithelial cells are not known to be present in the circulation of individuals who are free of cancerous growths. Normally, epithelial cells from the tissues of origin are eliminated after cell death with the help of macrophages and/or sloughed off into the lumen in the case of glandular cells lining the GI tract. Thus the presence of epithelial cells in the circulation, in itself, reflects the presence of cancerous tumors that have acquired invasive properties. However, in order to avoid false-negative results, one needs to carefully choose the epithelial cell marker for identifying these rare cells in the circulation, which are generally positive for EMT. We used CK19 as an additional marker to confirm the epithelial nature of CCSCs (Figure 5). Cytokeratin 19 is a well-established marker of epithelial cells, including primary tumor cells, and is known to be absent from mesenchymal and normal blood cells.54, 55, 56, 57 Presence of CK19+ cells in the blood has been shown to be associated with poor clinical outcome for patients with many types of epithelial cancers, including breast cancers.58 Importantly, almost all the CCSCs isolated by us expressed CK19 (Figure 5c), confirming that CCSCs, positive for one or more stem cell markers (such as DCLK1, LGR5, CD44), as isolated and enumerated by IF staining, were of epithelial origin. The current findings from <10 patients, however, needs to be validated by analyzing blood samples from a larger cohort of patients.

Cell surface-associated ANXA2 (CS-ANXA2) is known to have a critical role in metastasis of breast and lung cancers (discussed in Sarkar et al59); downregulation of ANXA2 with shRNA-ANXA2-nanoparticles significantly inhibited the formation of metastatic lesions from lung CSCs.26 Embryonic stem cells, induced to overexpress PG (a potent autocrine growth factor for colon cancer cells,21), became transformed and developed metastatic potential and were found to overexpress CS-ANXA2, unlike the non-transformed embryonic stem cells.14 Downregulation of ANXA2 reduced the metastatic potential of the transformed epithelial embryonic cells.14 As CS-ANXA2 is required for metastasis of cancer cells (discussed in Andey et al26and Sarkar et al59), we examined the possible expression of ANXA2 by CCSCs and report for the first time co-expression of ANXA2 by CCSCs in the blood of nude mice bearing metastatic tumors and in the blood of CRC patients (Figures 1c and 5).

PG peptides are expressed by the majority of CRCs,21 and a high expression of PG by adenocarcinomas from patients was reported to be prognostic for poor survival.60 PG expression by CD133+ colon CSCs was recently reported.61 In the current studies, we report for the first time co-expression of PG by a significant population of CCSCs (Figure 5). PG expression by CCSCs was found to heavily co-localize with CS-ANXA2 (Figures 5b and d); co-localization and endocytosis of CS-ANXA2, bound to PG, was required for measuring the biological effects of PG.59 Thus the expression of PG in CCSCs suggests the potential autocrine activity of PG in CCSCs, which may be required for seeding and initiating the growth of CCSCs as metastatic lesions in the peripheral organs. PG may thus not only support the mitotic growth of colon cancer cells, by upregulating CSC populations as previously reported,23 but may also support dissemination and metastatic growth of colon CTCs, providing a CRC-specific CTC marker.

To date, ‘CellSearch System’ is the only automated CTC detection bioassay, which has been approved by the FDA.8, 62 The system utilizes a combination of well-established epithelial cell markers, such as EpCAM and cytokeratins; however, reproducibility between laboratories has remained an issue.52, 53, 54, 55 Laboratories using CellSearch systems for detecting CTCs in colorectal patients have reported the presence of <3 CTCs/7.5 ml blood;8 these results are consistent with our findings of <1 CTC/ml blood of CRC patients, when EpCAM alone is used as a marker (Figures 3 and 4). However, detection of CTCs with CSC markers (CD44/DCLK1) along with CK19, with or without ANXA2/PG, resulted in isolating a significantly higher number of CCSCs (>4–5/ml blood) (Figures 4 and 5). The latter results re-emphasize the unreliability of using EpCAM alone as a marker for CTC detection, as it can potentially result in false-negative results. Use of EMT markers significantly increased the numbers of CTCs measured in cancer patients.56 However, EMT markers alone may also give erroneous results, as many cells other than CSCs also express EMT markers during dissemination. Therefore, use of one or more CSC markers along with epithelial/EMT/metastatic markers will likely provide the most accurate assessment of the risk for recurrence and metastatic disease. Addition of PG may provide a CRC-specific marker and help diagnose patients for the type of cancer as well. Criterion, such as specificity and sensitivity, are generally important for diagnostic assays.63 However, these criteria do not apply to all types of tests, especially the cell-based assays, as is the case with CTCs. The latter point is well discussed in a rebuttal letter to the editor in response to the article by Parikh et al.63 As discussed above, the presence of CTCs (irrespective of number/ml blood), in itself, is diagnostic for the presence of cancerous growths and/or metastatic disease. However, to avoid false positives, we have suggested the manual inspection of the cells to eliminate false-positive results due to CD45 staining. To avoid false-negative results, the use of more robust markers are described above, including CK19. In order to further increase the value of the assay, we have suggested the analysis of the cells for several stem cell markers and markers such as ANXA2/PG, based on the results of our investigations.

Therefore, based on the results of our study, we report a novel approach for detecting CSCs in circulation, using a combination of CSC and epithelial cell markers. As CSCs are believed to be responsible for tumor initiation and metastases, we can potentially screen patients for relapse or for the presence of metastatic disease, by measuring CCSCs in the blood of patients, and improve the automated bioassays by including one or more CSC markers. We recently reported that targeted disruption of DCLK1 along with other non-toxic dietary agents can potentially eliminate CSCs and the possibility of relapse.28 Thus, based on the results of CCSC assays described here, personalized treatment regimens can be developed for targeting and eliminating circulating CSCs, to significantly reduce the possibility of relapse and improve clinical outcomes.