Despite the well recognized expression of the cell surface markers cluster of differentiation 44 (homing cell adhesion molecule) and CD133 (Prominin 1) on human colorectal cancer stem cells (CCSCs), these molecules do not appear to be effective targets for stem cell-directed therapies. Because the surface marker CD66c (also known as carcinoembryonic antigen-related cell adhesion molecule 6) has demonstrated promise as a therapeutic target in pancreatic malignancy, the authors evaluated its potential as a target for stem cell-directed treatment of colorectal cancer.
First, the authors characterized CD66c expression by flow cytometry and immunohistochemistry in colon cancer samples and in normal colon tissues. Then, the coexpression of CD66c and CD133 was evaluated on putative CCSCs. CD66c expression also was measured in stem cell-enriched colon spheres. Finally, the effects of small-interfering RNA-mediated CD66c silencing on the in vitro and in vivo growth of Caco2 colon cancer cells were evaluated.
CD66c expression was significantly higher in colon cancers than in contiguous normal colon tissues and paralleled cancer stage. CD66c was absent in CD133-positive cells that were isolated from normal colon, whereas its expression was brightest (CD66cbright) in CD133-positive cells from colon cancer samples. In vitro experiments demonstrated that colon spheres were considerably enriched in a CD66cbright population in a fashion comparable to the enrichment observed in fresh liver metastases. In vitro proliferation and clonogenic potential were hampered when CD66c was silenced in Caco2 cells. Finally, in vivo xenograft experiments demonstrated that CD66c silencing almost completely abrogated the tumorigenic potential of Caco2 cells.
The isolation and characterization of tumorigenic colon cancer cells play major roles in the development of new diagnostic and therapeutic procedures.1 A small subset of cancer stem cells has been identified in several cancers, including colon malignancies.2, 3 In colon cancer, a putative stem cell subset was identified based on cluster of differentiation 133 (CD133; Prominin 1)2-5 and CD44 (homing cell adhesion molecule)3, 4 expression, and these molecules reportedly play a functional role, albeit limited, in colon cancer cells.6-9 Although these markers have been proposed as therapeutic targets, they do not appear to be suitable for tailored stem cell-directed treatment in colon cancer.10, 11 Consequently, the search is on for other markers of colon cancer stem cells.
The carcinoembryonic antigen-related cell adhesion molecule (CEACAM) family member 6 (CD66c/CEACAM6) is currently being investigated for its potential function in tumorigenesis.12-14 CD66c expression, measured by immunohistochemistry, was reportedly increased in colorectal malignancies, and high CD66c expression was identified as an independent prognostic factor in patients with resectable colon cancers.15-17 In pancreatic cancer, CD66c is a promising therapeutic target of new monoclonal antibody-based and small-interfering RNA (siRNA)-based approaches.18, 19 In the current study, we characterized CD66c expression in colorectal cancer stem cells and investigated whether CD66c was involved in colon cancer growth thus determining its potential as a target for cancer stem cell-directed therapy.
MATERIALS AND METHODS
Forty-six colorectal cancer biopsy samples were obtained from the Department of Clinical and Experimental Medicine (Second University of Naples). Informed consent was obtained from all patients who were included in the study. Paired with each tumor sample, a biopsy of normal colon mucosa from the same patient was analyzed as a control except in the case of samples of hepatic metastases. All biopsies were treated and analyzed immediately after surgical resection. The samples were processed according to previously described procedures.3
Cells and Cell Culture
Caco2 cell line culture
The Caco2 human colorectal cancer cell line was purchased from the European Collection of Cell Cultures, which authenticated it. Caco2 cells were maintained at the CEINGE-Advanced Biotechnology (Naples, Italy) cell culture facility according to the original datasheet.
Colorectal cancer cell culture as spheres
Human colorectal cancer cells obtained from enzymatic digestion of fresh biopsies were cultured in a specific serum-free medium and an ultralow adhesion flask according to previously described protocols.3 Medium was replaced twice a week. From 6 primary tumors that were processed, we grew spheroid cultures for analysis.
Flow Cytometry Analysis and Cell Sorting
Sample staining procedure and antibodies used in the study
After the appropriate preparation of samples, 50 μL of cell suspensions were incubated with appropriate amounts of each monoclonal antibody for 30 minutes at 4°C. The cells were then washed once with 1 mL 2% fetal bovine serum/phosphate-buffered saline, resuspended in 500 μL 2% fetal bovine serum/phosphate-buffered saline, stained with 0.5 μL of Sytox Blue for 5 minutes at room temperature, and analyzed by flow cytometry. We used the Becton Dickinson FACSAria cell sorter (Becton Dickinson, Franklin Lakes, NJ) for the analysis. We used the following monoclonal antibodies in this study: CD133-allophycocyanin (APC) (clone AC133; Miltenyi Biotec, Auburn, Calif); CD66c-phycoerythrin (PE); CD326–peridinin-chlorophyll protein complex (PerCP)–cyanine 5.5 (Cy5.5); human and mouse CD45-APC-cyanine 7 (Cy7); and human leukocyte antigen (HLA)-ABC-fluorescein isothiocyanate (FITC) (BD Biosciences and BD Pharmingen, San Jose, Calif). For vital dye, we used Sytox Blue (Invitrogen Life Technologies, Carlsbad, Calif). The background level for each fluorochrome used was assessed with the “fluorescence minus 1” technique.20
Analysis of fresh human normal colon, colorectal cancer, and colon spheres
Before the evaluation of CD66c and CD133 expression, the appropriate cell population was selected using the following gating strategy: 1) Cells were first gated on physical parameters (forward scatter [FSC] and side scatter [SSC]) to exclude most of the debris and dead cells; 2) doublets and aggregates were eliminated using FSC-area versus FSC-height pattern; and 3) viable epithelial cells were then selected as CD45-negative/CD326-positive/Sytox Blue-negative.
Caco2 cell line analysis
To evaluate CD66c expression on Caco2 cells: 1) the cells were first gated on physical parameters (FSC and SSC) to exclude debris; then 2) doublets were excluded from the selected population in a FSC-area versus FSC-height dot plot; and finally 3) viable cells were identified as Sytox blue-negative.
Analysis of Caco2-derived xenografts
For the cytometric evaluation of the percentage of Caco2 cells inside mouse xenografts, 1) cells were first gated on physical parameters to exclude debris; 2) in an FSC-area versus an FSC-height dot plot, doublets were excluded; and 3) viable cells were identified as Sytox blue-negative. Caco2 cells were identified as CD326-positive/human HLA-ABC–positive/mouse CD45-negative cells.
Mouse xenograft models
We used female athymic nu/nu mice aged 4 to 6 weeks (Harlan Laboratory, Udine, Italy) for in vivo xenograft experiments with the Caco2 cell line. The mice were housed in a specific pathogen-free environment, and food and water were provided ad libitum. A quantity of 1 × 106 siRNA-treated cells was injected subcutaneously into the flanks of the mice. CD66c protein expression and tumor cell viability were assessed immediately before the injection of siRNA-treated cells into mice. According to a protocol established by our ethics committee, the mice were killed 12 to 20 weeks after the injection of Caco2 cells, when the tumors reached a size of approximately 1 cm. Because it was not possible to wait for the “clinical death” of the animals, we were unable to produce survival curves. A portion of each tumor recovered was fixed in formalin for histochemical analysis (if the size of the sample permitted), and another portion was subjected to enzymatic digestion for flow cytometric analysis.
CD66c stealth RNA interference (RNAi) siRNA and aspecific, high-guanine/cytosine (GC)-content, stealth RNAi siRNA negative control were purchased from Invitrogen Life Technologies. The sequence of CD66c-specific siRNA used (141626H03) was GGAAGGAGGUUCUUCUACUCGCCCA. Transfection was performed according to the manufacturer's instructions. Selected siRNAs were transfected at a concentration of 50 nM in T-25 cell culture plates (Falcon; Becton Dickinson) by using Lipofectamine 2000 reagent (Invitrogen Life Technologies) when Caco2 cells reached 40% to 50% confluence.
Proliferation, Apoptosis, and Colony-Formation Assays
Caco2 siRNA-transfected cells were detached from culture flasks after 48 hours of incubation and seeded in 96-well plates (Falcon; Becton Dickinson). Cell proliferation was assessed, as indicated by the datasheet, using the water-soluble tetrazolium-1 (WST-1) for 1 week. The formazan formed from WST-1 cleavage was quantified by absorbance measurement at 450 nm with the Ultramark Plate Reader (BioRad, Hercules, Calif), and 640 nm absorbance was subtracted to diminish background level.
To evaluate the apoptosis and necrosis rate in siRNA-treated cells, we used the Vybrant® apoptosis assay kit with YO-PRO-1 (a carbocyanine nucleic acid stain) and propidium iodide (Invitrogen Life Technologies). According to the datasheet, cells were resuspended in phosphate-buffered saline at a concentration of 1 × 106 cells/mL in 1-mL volume and then 1 μL propidium iodide (1 mg/mL), and 1 μL of YO-PRO-1 were added to the suspension. The cells were incubated on ice for 30 minutes and then analyzed by flow cytometry.
Caco2 clonogenic assay
A clonogenic assay was performed according to the protocol published by Franken et al.21 Briefly, Caco2 siRNA-treated cells were seeded in triplicate at a concentration of 200 cells per well in a 6-well plate. Two weeks later, the colonies were stained with 6% glutaraldehyde and 0.5% crystal violet for 30 minutes and counted under a phase-contrast microscope (Leica DMIL; Leica Microsystems, Wetzlar, Germany).
Immunohistochemical and Histologic Analysis
CD66c expression was evaluated in a different set of 22 colorectal carcinomas, in 12 adenomas, and in adjacent normal tissues. The slides were mounted and examined under a Leica DC500 compound microscope (Nussloch, Germany), as described by Andolfo et al.22 Two operators analyzed the samples and evaluated the expression. An intensity score was used to represent the average intensity of the positive cells with scores of −/+ (none or very faint intensity, at the limit of detection, in >80% of cells), + (weak in >50% of cells), ++ (intermediate in >50% of cells), and +++ (strong in >80% of cells). For the general histopathology of Caco2-derived xenografts, paraffin-embedded sections were stained with hematoxylin solution for 8 minutes, then counterstained with eosin solution, mounted on glass slides, and examined under a Leica DC500 compound microscope.
Differences between groups were evaluated with paired t tests, analyses of variance, Mann-Whitney tests, or Wilcoxon tests, as appropriate, using GraphPad Prism (GraphPad Software, Inc., San Diego, Calif). Comparisons with controls were taken into account only if they were considered statistically significant (P < .05)
CD66c Expression in Normal Colon and in Colorectal Cancers
In this study, we examined 41 primary colorectal cancer samples and 5 hepatic metastatic samples from 44 patients (we were able to obtain both the primary tumor and the hepatic metastasis in 2 patients at the time of surgical resection of the primary tumor and 1 year later, respectively). Table 1 lists the anatomic locations of the 46 colorectal cancer specimens and the patients' TNM classification.
Table 1. Clinicopathologic Characteristics of Patients Affected by Colorectal Carcinoma in the Current Study
No. of Patients
Colon, right side
Colon, left side
Lymph node classification
The expression of CD66c was significantly higher in colorectal cancer specimens than in normal colon samples in terms of both the percentage of positive cells (P < .0001) and the mean intensity of expression (P < .0001) (Fig. 1Aa, Ab). In parallel, we used immunohistochemistry to evaluate the expression and localization of CD66c in a different set of 22 colorectal cancers, in 12 adenoma specimens, and in adjacent normal tissues. CD66c expression was higher in colorectal cancers than in normal tissues, in which it was hardly detectable (Fig. 1B,C). Moreover, immunohistochemical analysis revealed that CD66c expression followed a gradient according to the malignity of the lesion, increasing from normal tissues to adenoma and from adenoma to carcinoma (Fig. 1B,C).
Using multidimensional flow cytometry, we next compared CD66c expression in the 41 primary tumors according to the patients' TNM classification (T1-T2 vs T3-T4 [Fig. 1D] and M0 vs M1 [Fig. 1E]). In both analyses, CD66c expression was significantly greater in advanced lesions (P = .0073 and P = .0290, respectively).
CD66c Expression on CD133-Positive Putative Stem Cells in Normal Colon and in Colorectal Cancers
We also used multidimensional flow cytometry to evaluate the expression of CD66c on the CD133-positive putative stem cell population in the 41 primary tumors and in the pair-matched normal colon tissues. Figure 2A shows that CD133 and CD66c expression levels were mutually exclusive in normal colon (Fig. 2Aa), whereas there was a striking correlation between CD133 and CD66c expression in colorectal cancer cells (Fig. 2Ab). Flow cytometric analysis revealed that CD66c was expressed by most tumor cells, but its intensity ranged from low to bright; and cells with the brightest CD66c expression (CD66cbright cells) fully corresponded to CD133-positive cells. Figure 2B shows that CD133-positive cells from normal tissues were almost completely CD66c-negative whereas almost all CD133-positive cells from tumors were CD66c-positive.
Immunohistochemistry performed on 22 colorectal cancers and adjacent normal tissues confirmed the absence of CD66c at the bottom of the normal colon crypts in which colon stem cells are located (Fig. 2Ca, inset).23, 24 Labeling was weak in the upper half of the crypt in normal tissues (Fig. 2Ca), whereas CD66c was strongly expressed in all the dysmorphic crypts in tumor tissues (Fig. 2Cb). It is noteworthy that, unlike flow cytometry, immunohistochemistry was unable to distinguish tumor cell populations according to their level of CD66c expression.
CD66cBright Population Enrichment in In Vivo Metastases and in In Vitro Colon Sphere Cultures
Flow cytometry revealed significant enrichment of a CD66cbright cell population in the 5 liver metastases. Figure 3A illustrates CD66c and CD133 expression in a primary tumor (Fig. 3Aa) and in a hepatic metastasis (Fig. 3Ab) that occurred in the same patient 1 year after surgical resection of the primary tumor. A comparison of CD66c expression in 41 primary colorectal cancers and in 5 hepatic metastases showed that CD66c expression was significantly higher (P = .0102) in hepatic metastases (Fig. 3B).
In 6 independent in vitro experiments, we observed that CD66c expression was significantly brighter (P = .0313) in fresh tumor-derived cells that were grown in sphere culture conditions (in which cancer stem cell selection, proliferation, and maintenance are promoted)2, 3 than in cells that were obtained from fresh, pair-matched primary tumors (Fig. 3C). Moreover, CD66cbright cell selection was similar to that observed in hepatic metastases (Fig. 3Da). The brightfield image in Figure 3Db illustrates a typical colon sphere that was obtained from a primary colorectal cancer.
We expected that the CD66cbright tumor cell subset would be tumorigenic in mouse xenograft experiments, because these cells fully correspond to CD133-positive cells whose tumorigenic potential is well established.2, 3 We verified our hypothesis by injecting the CD66cbright cells sorted from 2 patients with colon cancer into 2 nonobese diabetic/severe combined immunodeficient mice. In both mice, the CD66cbright cells gave rise to macroscopic tumor formations, whereas CD66c-negative cells injected as a control into the same mice did not form tumors. Moreover, CD66cbright cell-derived tumors contained CD66c-positive cells and CD66c-negative cells that closely resembled the primary tumors (data not shown).
CD66c Silencing in the Caco2 Cell Line
Because the Caco2 cell line inherently expresses high levels of CD66c (Fig. 4A), we used it to evaluate the effects of CD66c-specific, siRNA-mediated silencing. To this aim, cells were transfected with either CD66c-specific siRNA or aspecific control siRNA to evaluate the specificity of the observed phenotypes. The silencing of CD66c was evaluated by flow cytometry 72 hours after siRNA exposure. At that time, CD66c expression was suppressed to a highly significant extent (P = .0013) in the cells that were transfected with CD66c-specific siRNA (Fig. 4B).
We also measured CD133 expression in CD66c-specific, siRNA-treated Caco2 cells and observed that it was unaffected by silencing, supporting the idea that the effects we observed depend on the specific inhibition of CD66c (data not shown). Proliferation was reduced significantly (P < .0001) in CD66c-silenced cells up to 144 hours after siRNA exposure (Fig. 5A). Apoptosis and necrosis were significantly higher (P = .0013 and P = .0214, respectively) in CD66c-specific, siRNA-treated cells than in control cells (Fig. 5B). In a colony-forming assay using Caco2 cells after siRNAs treatment, colony formation was inhibited significantly (P = .0009) in CD66c-specific, siRNA-treated cells versus control cells (Fig. 5C,D). These findings are indicative of reduced clonogenic potential in CD66c-specific siRNA-treated cells.
Next, we investigated the tumorigenic potential of Caco2 cells that were transfected with CD66c-specific siRNA or with aspecific siRNA. In 3 independent experiments, we injected aspecific, siRNA-treated cells into the flank of 2 mice and CD66c-specific siRNA into the other flank, for a total of 6 mice. The injection of aspecific, siRNA-treated cells gave rise to a palpable tumor in each mouse (except 1 that died of unknown causes the day after the injection), whereas the CD66c-specific, siRNA-treated cells either did not generate tumors or resulted in minute growths (Fig. 6Aa). Immunohistochemistry revealed that tumors arising from aspecific, siRNA-treated cells were full of proliferating glands (Fig. 6Ba). Conversely, the growths generated by CD66c-silenced cells appeared to be a mouse reaction to the injection, and the very few small tumor glands observed (Fig. 6Bc) probably derived from a residual percentage of unsilenced cells. Flow cytometry of the 5 tumors generated by aspecific, siRNA-treated cells revealed numerous human Caco2 cells (Fig. 6Bb), whereas the percentage of Caco2 cells in the growths generated by CD66c-silenced cells was negligible (Fig. 6Bd).
During the last 10 years, it has been demonstrated that CD66c is overexpressed in colon cancer, but its role in colon tumorigenesis and its involvement with cancer stem cells remain obscure.15-19 In the current report, we describe CD66c as a “stemness” marker in colorectal cancer. Hitherto, CD66c was reportedly identified as a differentiation marker in normal colonocytes, in which it plays a role in the maintenance of tissue architecture and in colonocyte differentiation. CD66c is normally up-regulated as colonocyte differentiation proceeds,12, 25 as witnessed by its restriction to the crypt apex, which we also describe in this report. This discrepancy may be explained by evoking the “asynchronous differentiation” model, which has been observed in several types of cancers. For instance, in the hematopoietic system, CD66c is expressed exclusively by maturing granulocytes, but in B-cell acute leukemia, it is expressed in abundance on the surface of early lymphoblasts.26 Alternatively, it is conceivable that CD66c is expressed at undetectable levels and/or that it is expressed in a very small subpopulation of normal colon cells. Cancerous transformation may amplify CD66c-expressing cells and may also increase its expression at the cell surface.
We considered CD66cbright to be a CCSC marker, first because it was perfectly coexpressed with CD133, which is a widely recognized marker of CCSCs. Second, because CD66cbright cells from fresh colorectal cancer tissues were tumorigenic in mice, whereas CD66c-negative cells were not. Third, CD66cbright cells were able to reform the CD66c-negative counterpart in vivo, which indicates that they are able to undergo asymmetric divisions that resulted in the formation of growths that resembled the original tumor. Finally, when CCSCs were maintained and enriched in colon sphere cultures, CD66cbright cells were spontaneously selected, thus reinforcing the concept that the CD66c expression level is associated with CCSCs.
Our data regarding the functional role of CD66c in colorectal tumorigenesis, based on the effects derived from its silencing, complement previous observations that CD66c overexpression may be involved in this tumorigenesis. Ilantzis et al. demonstrated that deregulated overexpression of CD66c in stable transfectant Caco2 cells blocked cell differentiation,25 thereby disrupting crypt architecture in vivo in nude mice and in vitro in monolayer or 3-dimensional culture.27 Moreover, CD66c expression in transgenic mice leads to an increase in colonocyte proliferation and strongly inhibits differentiation and anoikis.27 These effects can contribute to colorectal tumorigenesis. In fact, CD66c ectopic expression in rat myofibroblasts enabled them to form tumors in mice,28 whereas CD66c overexpression in Caco2 cells greatly reduced the latency time of tumor formation and resulted in larger tumors in nude mice.12
Despite its recognized role in normal colon architecture and in tumorigenesis, little is known about the function of CD66c. It was identified in specific membrane rafts in which the cross-linking of its external domain was able to recruit and then to activate other molecules, such as the α5β1 integrin, in which involvement in cell-cell and cell-matrix interactions is relevant in cell differentiation, proliferation, and survival. CD66c cross-linking also mediates the recruitment and activation of integrin-linked kinase, protein kinase B, and mitogen-activated protein kinase.27
Our current findings raise the possibility of exploiting CD66c silencing in the treatment of colorectal cancer. CD66c silencing has been used previously with encouraging results in the treatment of pancreatic adenocarcinoma.18 Moreover, a CD66c-directed antibody was successfully evaluated in preclinical models of pancreatic carcinomas.19 The latter study also revealed that nonhuman primates exposed to the CD66c-directed antibody were affected only by granulocyte-specific neutropenia, which regressed within a very short time.
Despite their role as CCSC markers, the use of CD133 and CD44 as therapeutic targets is questionable. In fact, CD133 is not involved in colon tumorigenesis, and its silencing did not impair tumor growth in the mouse.8 Moreover, in another study, CD133 was not expressed by all colorectal cancer samples,3 whereas CD66c was consistently expressed in all of our samples, suggesting that it is a more robust marker of CCSCs. With regard to CD44, it has been reported that its silencing impairs tumor growth.9 However, the therapeutic use of CD44 probably would be limited because of its toxic effects consequent to its ubiquitous expression. In conclusion, our results set the scene for further investigations of CD66c as a therapeutic target in human colorectal malignancy and particularly in colorectal cancer stem cell-directed treatment.
We thank Jean Ann Gilder (Scientific Communication srl) for writing assistance and Vittorio Lucignano for figure assembly.
This work was supported by grants from the Italian Association for Cancer Research (grant 10737 to Dr. Del Vecchio), from the Ministry of Health (Rome, Italy), from the Ministry of Education, University, and Research (grants PS 35-126 IND and PRIN 2007 to Dr. Salvatore), and from a CEINGE-Regione Campania contract (DGRC 1901/2009 to Dr. Salvatore).