A Stochastic Model for Cancer Stem Cell Origin in Metastatic Colon Cancer Free

Colon cancer is one of the most common forms of cancer in the industrialized world (1). As with most cancers, colon cancer is a heterogeneous disease that starts locally but generally spreads through the lymphatic or portal venous system to different organs (2). The liver is the most frequent visceral site of metastatic dissemination (3). It is the initial site of metastasis in the majority of recurring colon cancers, with most patients showing liver involvement at the time of death. It is widely believed that the successful metastatic cancer cell must leave the primary tumor, enter the lymphatic or circulatory systems, survive within the circulation, overcome host defenses, extravasate, and then grow as a vascularized metastatic tumor (4). We know that primary tumors shed large numbers of tumor cells, but only a few of these develop into metastases (5). Two models have been proposed to explain the heterogeneous potential of tumor cells and the process of metastasis in general: the stochastic model, whereby a distinct population of tumor cells acquires the appropriate set of somatic mutations and develops metastatic capability, and the hierarchical model, in which primary tumors and metastatic cancer are initiated by rare cancer stem cells (6, 7). Recently, several studies have identified the presence of cancer-initiating cells in breast (8), brain tumors (9–11), and, most recently, colon cancer (12–14). With this new insight into the cellular mechanism leading to cancer, it has been suggested that the metastatic potential of tumor cells may be a reflection of the ability of the cancer stem cells to clonally initiate tumorigenicity at distant sites (15, 16). Although strong evidence has been reported for the presence of cancer-initiating cells in colon cancer, limited support has surfaced for a possible role of tissue-derived stem cells in the initiation of this cancer and its metastatic process. One of the technical difficulties in answering this question is that normal and abnormal colon epithelial cells thus far have proven resistant to growth in culture. To investigate the presence of stem-like tumor cells in metastatic colon cancers, we used an in vitro assay, which provides efficient propagation of tumor cells and a functional demonstration of self-renewal and multipotency at the clonal level.

Tumor cell acquisition and preparation. Following informed consent, tumor tissues were collected after surgical removal from the liver of metastatic colon cancer patients. Tumor tissues were minced into small fragments with scalpels and then digested using a two-step collagenase method derived from the isolation of liver cells (17). A 20-min trypsin incubation was added to complete dissociation of the tumor tissue. If enough cells were collected from the tissue, flow cytometric analysis was performed on each tissue fraction. In most cases, tumor cells were plated on a monolayer of previously irradiated (80 Gy) rodent epithelial feeder cells (18).

Flow cytometry. Cells were counted and aliquoted at 10⁶ per tube and incubated with monoclonal antibodies at the appropriate dilution. Dead cells were detected with propidium iodide (10 μg/mL). Postacquisition analysis of the fluorescence-activated cell sorting data and desktop publishing were accomplished using the third-party flow cytometry software FlowJo.⁵

Clonal expansion. Metastatic colon tumor cells from the different patients were stained with cell surface markers and sorted for the desired phenotype. The sorted populations were resuspended in 2 mL of medium and deposited at different concentrations in a six-well plate. To initiate a clonogenic expansion, cells were observed using an inverted microscope at 100× amplification and single cells were individually aspirated using a micropipette. Each single cell was then manually deposited in a unique well of a 96-well plate previously plated with feeder cells. Four weeks later, plates were screened for colonies.

Transplantation in mice. Six- to 8-wk-old Rag-2/γc^−/− mice were anesthetized and transplanted with tumor cells. In general, tumor cells isolated from primary tissue or culture expansion were mixed with medium/Matrigel at a ratio of 1:1 and injected. Tumorigenicity was determined by transplanting tumor cells (10⁶). Six to 8 wk after transplantation, mice were bled and the serum samples were analyzed for the presence of carcinoembryonic antigen (CEA) by ELISA (Biomeda).

Immunostaining and transmission electron microscopy. Primary antibodies Muc-1 (BD Biosciences), Muc-2 (Santa Cruz Biotechnology), villin (Santa Cruz Biotechnology), and chromogranin A (CgA; Santa Cruz Biotechnology) were incubated for an hour followed by chicken anti-mouse Alexa Fluor (488 or 594, Invitrogen). HLA-biotin (Ancell) was followed by streptavidin-FITC (PharMingen). Transmission electron microscopy was done as described previously (19).

See additional Materials and Methods in Supplementary Data.

In vitro selection and expansion of tumor cells derived from metastatic colon cancer. Human colorectal tumor specimens, typically moderately to well-differentiated gland-producing adenocarcinomas, were collected from the metastatic liver site of 13 different Caucasian patients who underwent surgical resection (Table 1). None of the patients in this study had an identified genetic predisposition to colon cancer, such as familial adenomatous polyposis or hereditary nonpolyposis colorectal cancer. Tumor cells were cultured under conditions that favored epithelial cell growth (18). We postulated that in vitro cultures of tumor cells and their expansion would select for a population of tumor cells best adapted for culture and self-renewal. The rationale for this approach is that cancer stem cells have the capacity for self-renewal and this method should allow us to expand these primitive cells in vitro. Within 2 to 4 weeks of culture, a minor fraction of tumor cells in each of the 13 tumors studied showed growth into what seemed to be clonally derived colonies of epithelial cells. The generation of cell colonies from disaggregated tumors provides an index of tumor cell clonogenicity. In our hands, the frequency of tumor cell colony formation was estimated at <0.1% of the total cells, with anywhere from a few to tens of colonies per million cells isolated from the tumors. Long-term (more than three passages) patient-specific cultures of tumor cells were established for all 13 cultures (Fig. 1A). When observed in vitro, the majority of tumor cells exhibited a similar undifferentiated phenotype, with colonies composed of hundreds of epithelial cells. In each of the cultures, some of the tumor cell colonies changed gradually over time, from an undifferentiated to a more mature phenotype, ranging from the formation of an organized “ridge” of epithelial cells to large glandular formations resembling aberrant crypts (Fig. 1A). This pleomorphic maturation observed in the cultures of human colon tumor cells provides support for cellular organization in metastatic colon cancer and may have some relevance to the model of cancer stem cells. It seems that a hierarchy of tumor cells is emerging among the cultured cells from undifferentiated primitive cells to more mature cells.

Tumor cells derived from metastatic colon cancer display distinct colon cell lineage commitment in vitro. Several markers are associated with more differentiated phenotypes and mature cells of the intestine. These include the epithelial mucins, secretory and nonsecretory proteins, within colorectal tissues that have been traced to different lineages (20). Muc-2 is secreted by differentiated intestinal cells and is predominant within colorectal mucus-producing goblet cells, whereas Muc-1 (CD227), a nonsecretory and cell membrane–associated glycoprotein, is expressed by certain absorptive enterocytes along the columnar cell apexes in the glandular tissue (20, 21). Villin, a component of the intestinal brush border cytoskeleton, is also a marker of enterocytes but can be expressed weakly in immature colon cells (22). Lastly, CgA expression is restricted to the rare enteroendocrine cells (23). We first determine whether tumor cells expanded in vitro expressed these distinct lineage-specific markers. Muc-1⁺, Muc-2⁺ (Fig. 1B), and villin⁺ cells (data not shown) were present in all metastatic colon tumor cultures. The mucin-secreting cells were not well differentiated, with few vacuoles visible. No positive CgA cells could be identified in the cultured tumor cells. Although endocrine differentiation is detected in colorectal carcinomas (23) and in our xenograft model (Fig. 2C), only few cell lines are capable of such lineage differentiation in vitro (22). Overall, our result indicates the presence of two possible lineage-restricted cell types, enterocytes and goblet cells, after in vitro expansion of tumor cells from metastatic colon cancer.

Tumor cells derived from metastatic colon cancer share similar flow cytometric profiles after in vitro expansion. We analyzed the flow cytometric profile of 11 of the 13 patient-specific tumor cells (Fig. 1C). After in vitro expansion, single-cell suspensions of the various cultured tumor cells were prepared and analyzed by flow cytometry using a panel of 24 different cell surface markers. Nineteen of the markers showed consistent results among all 11 cultures (7 positive, 9 negative, and 3 heterogeneous markers), whereas 5 markers differed from patient to patient. Analyses of sequential passages per culture revealed a stable phenotype of the cell surface marker repertoire among tumor cell populations in culture (Fig. 1C). Based on the recent findings described in the literature, we have included other markers in our study. The cell surface markers CD133, CD44, CD49f, EpCam, and CD166 were recently reported as present on human colon cancer stem cells (12–14). CD44 and CD49f were present in all of our tumor cell cultures, whereas CD133 antigen (CD133.1 and CD133.2) was not consistently expressed among the cultured samples and was absent in at least one culture. Interestingly, another more recent study (13) also reported that CD133 was not consistently expressed in all colon cancer samples. A limited analysis of several tumor cultures for CD166 and EpCam showed that both markers are positive on our tumor cell cultures (Fig. 3A). Based on the above results, cell surface markers identified on colon cancer stem cells are expressed on the tumor cells propagated in vitro.

In vitro self-renewing capacity and clonogenic potential of malignant colon epithelia derived from metastatic colon cancer. The clonogenic capacity of metastatic colon tumor cell cultures was assayed in vitro by limiting dilution analysis (LDA). LDA is a quantitative measurement to study the enrichment of stem cells with stem cell–enriched fractions, giving a higher response as measured by the frequency of colony formation (24–27). This classic stem cell assay has been described by others for the presence of hematopoietic stem cells in long-term bone marrow cultures (28) or the presence of neural stem cells in the formation of neurosphere cultures for the central nervous system (29). We applied this stem cell assay to our candidate cancer stem cell population. Tumor cells from seven different cultures were sorted and plated by limiting dilution. Three to 4 weeks after plating, the colony frequency of tumor cells was calculated by linear regression (27). The frequency of colony-forming tumor cells ranged from 1 of 7 (Tu-27) to 1 of 92 (Tu-18; Fig. 1D). These data revealed a dramatic enrichment of colony-forming cells after in vitro expansion compared with the low number of colonies generated by the initial primary tumor cell cultures. This result suggests that we have enriched for a primitive population of clonogenic tumor cells capable of physiologic self-renewal in vitro, as defined by the ability to go through numerous cell division cycles while maintaining the undifferentiated state.

Clonogenic tumor cells propagated in vitro are capable of self-renewal and multilineage differentiation. Next, we determined if, like the parental tumor cell cultures, clonally derived tumor cell cultures are capable of multilineage differentiation. We expanded 23 clonally derived tumor cell cultures from five different patients (Fig. 4A). These cultures, each derived from a single cell, behaved similarly and showed the same pleomorphic colonies as the parental cultures (Fig. 4B). Clonally derived cultures were repeatedly passaged with no sign of replicative senescence, as shown by their high efficiency of forming colonies in vitro (Fig. 1D). Distinct tumor cells stained positively for Muc-2 or Muc-1, indicating the presence of both goblets and enterocytes (Fig. 4C), similar to the parental cultures (Fig. 1B). Villin was weakly positive in numerous undifferentiated cells (data not shown), but cells at the apical membrane on the edge of the colonies were often strongly villin⁺, indicating the presence of differentiated enterocytes with organized microvilli and a mature brush border. Staining for CgA, which marks enteroendocrine cells, did not reveal evidence of cells from this lineage, as in the parental cultures, although their presence was detected in xenograft tumors (Fig. 2C). This result indicates that heterogeneous tumor cells, with the presence of at least two lineages, enterocyte and goblet cells, can be derived from a single cell in culture. This outcome also suggests that simultaneous activation of differentiation lineage pathways may occur in the pathogenesis of metastatic colorectal cancer as previously reported for primary colon adenocarcinomas (22, 30, 31). Flow cytometry of cultured cells derived from clonogenic tumor cells resulted in profiles identical to those previously seen in the parental cultures (data not shown). Clonogenic cultures were repeatedly passaged, showing the physiologic capacity for self-renewal as well as lineage maturation potential, properties that are expected from stem cells.

Clonogenic, multipotent, self-renewing colon cancer cells are tumorigenic. Because the most critical attribute of any cancer is the capacity to generate and perpetuate tumors, we examined whether the isolated stem cells derived from metastatic colon tumors also have the capacity to initiate tumors. To that extent, we assessed the ability of these cells to give rise to tumors in vivo. Tumor cells isolated from fresh tissues, newly established parental cultures, and clonally derived cultures from metastatic colon cancer samples were transplanted into immunodeficient Rag-2/γc^−/− mice (Fig. 2). The dorsal fat pad region under the skin was found to be the most reliable place to grow colorectal tumors, as previously reported by others (32). Our goal was to show the tumorigenicity of the isolated tumor cell population before and after expansion on feeder cells. Mice transplanted with human tumor cells were tested for xenograft tumors. CEA was screened in the serum of mice by ELISA. CEA is a tumor marker secreted by cells and typically associated with colon cancer (33). We showed that the human tumor cells, expanded in vitro, expressed high levels of CEA in the culture medium (Fig. 2A,, blue columns). Using the same approach, sera of mice transplanted with different tumor cell cultures were screened for CEA expression. Mice showed positive expression of human CEA, ranging from a few ng/mL to 117 ng/mL (Fig. 2A,, red columns). All of these mice were sacrificed and found to carry human tumors. Tumors derived from the fresh tissues, parental cultures, as well as the clonally derived cultures were morphologically identical to the patient's isolated tumors. Histopathologic analysis of the xenografts showed a striking adenocarcinoma-like tissue pattern (Fig. 2B and C). Xenograft tumors were moderately to well-differentiated adenocarcinomas with expansive growth, without evidence of invasion. Both parental and clonally derived tumors were serially transplanted and shown to be tumorigenic on secondary transplantation. The secondary tumors expanded faster than the initial xenograft tumors, suggesting that the malignancy might have been exacerbated in both the parental and the clonally derived tumor cells as a result of selection. When posttransplanted tumor cells were cultured back on stroma, flow cytometric analyses identified the same expression of the 24 cell surface markers as detected on the cells before transplantation (data not shown).

Xenograft tumors derived from a single expanded tumor cell were further analyzed for lineage markers. Markers for goblet cells, enterocytes, and enteroendocrine cells were present in the xenografts derived from a single expanded tumor cell. Muc-2 expression was present on a limited number of aberrant crypt cells, showing the presence of a few mature goblet cells at the edge of the apical surface (Fig. 2B–D). Villin was strongly expressed, indicating the presence of enterocytes. The demonstration of the mature morphology of the clonally derived tumor cells was determined by transmission electron microscopy of glutaraldehyde-fixed tissue. Figure 2D shows the mature morphology of goblet cells with mucus-containing vacuoles surrounded by columnar cells with apical microvilli and tight junctions. In conclusion, both well-differentiated goblet cells and columnar cells were observed in clonally derived xenograft tumors along with undifferentiated tumor cells. These results satisfy all of the criteria necessary to define these tumor cells as multipotent cancer stem cells in both in vitro and in vivo environments.

CD133⁺ tumor cells mediate tumor growth in xenotransplant. The cancer stem cell theory states that only a subset of prospective tumor cells has the ability to initiate the tumor. To determine if the in vitro culture and expansion system described here could generate similar results as previously reported for colon cancer stem cells (12, 14), we isolated tumor cells from three different patients showing CD133 heterogeneous expression (Tu-12 clone F7, Tu-21 clone C2, and Tu-27). Tumor cells isolated from xenograft tumors in mice (Tu-12 clone F7 and Tu-21 clone C2) and tumor cells expanded in culture (Tu-27) were sorted for CD133.1⁺ and CD133.1⁻ subpopulations. The CD133.1⁺ and CD133.1⁻ tumor cell subpopulations were transplanted to the left and right flank of immunodeficient mice. Six to 12 weeks after transplantation, only the site injected with CD133⁺ cells developed visible tumors, confirming previous reports that CD133⁺ colon cancer cells can grow as xenograft transplants (Fig. 3B and C).

Chromosomal instability is present in clonogenic multipotent tumor cells. To assess the presence of chromosomal instability in tumor cells, karyotype analysis was performed on Tu-12 and Tu-21 by examining the parental and two clonally derived tumor cell cultures (Tu-12 F7, Tu-12 G12, Tu-21 H11, and Tu-21 E12). All tumor cells analyzed expressed human karyotypes with clonal and nonclonal numerical and structural aberrations (Fig. 4D; Table 2; see Supplementary Data for more information). Many of the same aberrations observed in the parental cells were also seen in the clonally derived tumor cells, confirming their common origin. Surprisingly, the clonally derived tumor cells also contained aberrations unique to each culture.

The variability in the karyotypes of these cells suggests the presence of chromosomal instability with ongoing selection. As a control, normal human fetal liver cells were cultured under the same conditions. After seven passages in culture, no chromosomal abnormalities could be detected in the normal human fetal liver cultures. This result indicates that the in vitro culture conditions are not inducing the chromosomal instability observed in the tumor cell cultures.

Colon cancer is a complex and heterogeneous disease. Recently, the hypothesis that cancer stem cells drive the tumorigenic process in colorectal cancer has received support. The cancer stem cell hypothesis suggests that tumors are generated and maintained by a small subset of cancer cells capable of self-renewal and differentiation into bulk tumor cells. Both CD133⁺ (12, 14) and EpCam^high/CD44⁺/CD166⁺ (13) have been proposed as colon cancer stem cell surface markers. Collectively, these studies show the presence of cancer-initiating cells within colon tumors and raise questions about the origin of colon cancer stem cells. It has been anticipated that normal stem cells of the intestinal tract represent the natural target of tumorigenic mutations. Consequently, these cells would be the best candidates for the origin of cancer stem cells. However, not much information has been published on the possibility of cancerous stem cells in colon cancer. Our work reports the identification and isolation of tumor cells from human metastatic colon cancer with the properties of intestinal stem cells as well as tumor-initiating cells. The presence of single cells within metastatic tumors, which show self-renewal, multipotency, and the capacity to differentiate into enterocytes and goblet cells, broadens the current model of colon cancer stem cells by providing additional support for the potential role of natural stem cells in the development of colon cancer. However, the possible role of colon progenitor cells involved in this cancer cannot be ruled out until normal human colon stem cells are identified and characterized. Additionally, our data indicate that all 11 metastatic colon cancer samples extensively analyzed showed a remarkably consistent pattern of antigen staining. This result increases the number of previously proposed cancer stem cell markers and provides a robust and reproducible surface marker profile for colon cancer stem cells. Moreover, the presence of such tumor cells in each of the metastatic tumor samples isolated from patient liver sheds light on a common origin of the metastatic process in colon cancer. Despite extensive heterogeneity in such areas as tissue morphology and somatic mutations (34), evidence for the cellular origin of metastatic colon cancer is consistent with a common origin. The knowledge that tumor cells with similar characteristics are involved in the metastatic process of colon cancer is of vital importance to our understanding as to how these tumors metastasize, including, and especially, information on the key cellular and molecular mechanisms responsible for the metastatic process. Finally, the biology of the neoplastic stem cells in patients suffering from metastatic colon cancer is clearly important to future successful therapies and could also serve as a paradigm relevant to other types of malignancies that might originate from transformed stem cells.

Furthermore, chromosome instability has long been recognized as an engine of neoplastic transformation (35, 36) in the development and progression of the majority of colon cancers (37). Chromosomal abnormalities occur at a relatively early stage of colorectal neoplasia (38) and have been proposed as a significant force for clonal dominance in a dividing cell population (39). Our study suggests that chromosomal instability is present in colon cancer stem cells. The presence of genomic instability in cancer stem cells has many biological and therapeutic implications. From a biological point of view, genetic instability supports the stochastic model of cancer in the cancer stem cell population. This result points to a heterogeneous cancer stem cell population with an evolutionary process of tumor progression determined by two main events within the cancer stem cells: the generation of genetic heterogeneity in the cancer stem cell pool and the selection of genetic variant cancer stem cells most suited to survival, proliferation, invasion, and metastasis (40). This selection should, through the classic stochastic model of cancer origin, generate heterogeneous cancer stem cell populations, with some cancer stem cells better adapted and selected to reach the metastatic stage, whereas others are not. In this model, not every cancer stem cell would be capable of metastasis. From a therapeutic point of view, genetic instability introduces rapid mutations and adaptation to the selection process and may explain why many experimental therapeutic approaches have poor clinical results (41). Under imatinib therapy, a treatment that causes remission in most patients with chronic myeloid leukemia (CML) due to specific action on BCR-ABL1⁺ cells (42), patients relapse due to the inability of the treatment to eradicate CML cancer stem cells (43) and to the instability of BCR-ABL1 gene in the leukemia stem cells (44, 45). Furthermore, compared with normal stem cells, cancer stem cells with genetic instability may have acquired the formidable advantage of tumor cell survival fitness by combining the twin driving forces of mutation and natural selection (40).

This updated model of cancer stem cells once again raises the question of whether genetic instability is the initiating event and driving force that may transform colon stem cells or progenitor cells into lethal cancer-initiating cells. Although the ability of cancer stem cells to differentiate despite having chromosomal aberrations supports the development of new therapeutic approaches designed to promote tumor cell differentiation, the apparent capacity of cancer cells to acquire resistance to current treatments begs for a rationalized therapy targeting cells with chromosomal instability (35).

T.C. Gamblin: speakers bureau/honoraria, Aloka, Bayer, and Covidien. The other authors disclosed no potential conflicts of interest.

Grant support: Jeannik M. Littlefield-AACR Grant in Metastatic Colon Cancer Research (C. Odoux, H. Fohrer, T. Hoppo, L. Guzik, D.A. Geller, and E. Lagasse), Commonwealth of Pennsylvania (C. Odoux, H. Fohrer, T. Hoppo, L. Guzik, and E. Lagasse), and NIH grant RO1DE14729 (S.M. Gollin and D.W. Lewis). Cytogenetic studies were carried out in the University of Pittsburgh Cancer Institute Cytogenetics Facility, supported in part by NIH grant P30CA047904 to Ronald B. Herberman.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Alessandra Calvelli for her technical contribution.