Gastric Cancer Originating from Bone Marrow-Derived Cells

See allHide authors and affiliations

Science  26 Nov 2004:
Vol. 306, Issue 5701, pp. 1568-1571
DOI: 10.1126/science.1099513


Epithelial cancers are believed to originate from transformation of tissue stem cells. However, bone marrow–derived cells (BMDCs), which are frequently recruited to sites of tissue injury and inflammation, might also represent a potential source of malignancy. We show that although acute injury, acute inflammation, or transient parietal cell loss within the stomach do not lead to BMDC recruitment, chronic infection of C57BL/6 mice with Helicobacter, a known carcinogen, induces repopulation of the stomach with BMDCs. Subsequently, these cells progress through metaplasia and dysplasia to intraepithelial cancer. These findings suggest that epithelial cancers can originate from marrow-derived sources and thus have broad implications for the multistep model of cancer progression.

The link between infection, chronic inflammation, and cancer has long been recognized (1), a prime example being infection with Helicobacter pylori and gastric cancer (2). Chronic gastric inflammation, which develops as a consequence of H. pylori, leads over time to repetitive injury and repair resulting in hyperproliferation, an increased rate of mitotic error, and progression to adenocarcinoma. The same inflammatory environment that favors the development of cancer has also been linked to homing and engraftment in peripheral tissue by BMDCs. Recent studies have suggested that BMDCs may possess an unexpected degree of plasticity and often home to sites of chronic injury or inflammation (3, 4). In particular, this may occur where tissue injury induces excessive apoptosis that overwhelms or compromises the supply of endogenous tissue stem cells (5). Although the mechanism and extent of subsequent BMDC differentiation is not established (6), it is clear that engrafting cells rely on external environmental cues for the orderly inactivation of growth programs and progression of appropriate differentiation (3, 4, 7). There is little information on the long-term consequences of recruiting pluripotent cells to areas of chronic inflammation where signals for cell growth and differentiation may be altered.

To investigate the possible role of BMDCs in the metaplasia/dysplasia/carcinoma progression associated with chronic Helicobacter infection, we employed the established H. felis/C57BL/6 mouse model of gastric cancer (8). After undergoing lethal irradiation, mice were transplanted with bone marrow from C57BL/6JGtrosa26 (ROSA26) transgenic mice expressing a nonmammalian beta-galactosidase enzyme or from control wild-type littermates (9). Engraftment of ROSA26 marrow-derived cells was tracked with X-galactosidase (X-gal) staining. X-gal staining (blue) was not detected in wild-type mice (Fig. 1A) or wild-type infected mice (Fig. 1C). Uninfected ROSA26 transplanted mice did not demonstrate BMDC engraftment into gastric glands (Fig. 1B). Although acute (3 week) H. felis infection was associated with intense bone marrow–derived inflammation, it did not produce major architectural destruction and was not sufficient stimulus for stomach repopulation with BMDCs (table S1). In this model, gastric mucosal apoptosis increases at 6 to 8 weeks after inoculation (2) and, consistent with this, beta-galactosidase–positive (blue-staining) glands appeared after this peak of apoptosis. These cells were initially detectable at 20 weeks of infection, but their numbers increased dramatically with the length of time of infection, such that 90% of the gastric mucosa at the squamocolumnar junction was replaced with cells derived from donor marrow at 52 weeks after infection (Fig. 1D and table S1).

Fig. 1.

(A and C) X-gal staining (blue) of C57BL/6 mouse transplanted with wild-type marrow and (A) mock infected or (C) infected with H. felis for 30 weeks. (B and D) C57BL/6 mouse transplanted with ROSA26 marrow and (B) mock infected or (D) infected with H. felis for 30 weeks. (E) Wild-type mouse with chronic H. felis infection shows TFF2 (red) staining and is X-gal negative (blue). (F) In the infected ROSA26-transplanted mouse, BMDCs are positive for both beta-galactosidase (blue) and TFF2 (red). (G) Dysplastic glands in the infected ROSA26 mouse express abundant beta-galactosidase activity. (H) Mitotic activity in BMD epithelial cells demonstrated by coexpression of cytoplasmic beta-galactosidase activity (short arrows; blue) and chromosomal BrdU incorporation (long arrows; brown). 10-μm frozen sections. Magnification: [(A) to (G)], 600X; (H), 1,000X.

With chronic Helicobacter infection, a second proliferative zone forms deeper within the gastric mucosa, giving rise to metaplasia (2), designated SPEM (spasmolytic expressing metaplasia) because of positive staining for trefoil factor 2 (TFF2), also known as spasmolytic polypeptide (10, 11). In chronically infected wild-type mice, TFF2 (red staining) is prominent in deep antral and fundic glands (Fig. 1E). In mice transplanted with ROSA26 marrow and infected with H. felis, TFF2 expression is seen within blue beta-galactosidase–positive BMDCs (Fig. 1F). Histological alterations were similar in infected wild-type and ROSA26-transplanted mice, with both showing equivalent metaplasia and dysplasia. Of the few parietal cells or chief cells that persisted in the infected stomach, none were beta-galactosidase positive, which indicates that under these experimental conditions of H. felis infection, marrow cells do not differentiate toward the parietal or chief cell phenotype.

Epithelial dysplasia increased in severity over time, and by one year after inoculation resulted in carcinoma or high-grade gastrointestinal intraepithelial neoplasia (GIN) (12). In the mouse model of Helicobacter-mediated gastric cancer, dysplasia is considered a direct precursor of gastric adenocarcinoma and is found both at the squamocolumnar junction and at the antral-pyloric junction (13, 14). In the H. felis model, the majority of dysplastic glands stained blue with X-gal (Fig. 1G), and many BMDC within the epithelium, were bromodeoxyuridine (BrdU) positive (Fig. 1H), which demonstrates active proliferation. To further confirm the presence of beta-galactosidase, we used immunohistochemistry (IHC) for bacterial beta-galactosidase (9). Gastric tissue from wild-type mice did not stain for beta-galactosidase (Fig. 2, A and C), whereas all observed intraepithelial neoplasia in the 52-week infected mice were beta-galactosidase positive (Fig. 2, B and D; brown intracellular staining), which proves that these cells arose from donor marrow and strongly suggests an inherent vulnerability of this population to malignant progression. Bone marrow–derived GIN displayed features consistent with this histological diagnosis (12), including elongation and branching, crowding and distortion of gland structures, presence of hyperchromatic nuclei, pronounced cellular and nuclear atypia, and loss of polarity. Double-label immunofluorescence staining revealed that the beta-galactosidase–positive cells (red) within deep gastric glands were also pan-cytokeratin positive (green; merged seen as yellow) (Fig. 2F and fig. S1), with CD45 expression specifically restricted to infiltrating leukocytes (fig. S1). These studies confirmed that the marrow-derived cells had differentiated to a gastric epithelial phenotype, ruling out the unlikely possibility that the observed staining pattern was due to lymphocytes intercalating into the gland structure. Although these analyses directly showed beta-galactosidase enzyme activity (X-gal staining) and protein abundance (IHC), we further evaluated BMDCs within the epithelium for the lacZ/Neo fusion gene specific to donor cells. Laser-capture microdissection was used to capture and isolate entire X-gal–positive glands from chronically H. felis–infected ROSA26 and wild-type transplanted mice (9) (fig. S2, A and B). Polymerase chain reaction (PCR) with specific lacZ/Neo fusion gene primers followed by sequence analysis verified the cells to be of donor origin (9) (fig. S2C).

Fig. 2.

Beta-galactosidase IHC of stomachs from C57BL/6 mice transplanted with ROSA26 marrow. (A and C) Mock-infected mice do not demonstrate any BMDC engraftment, as evidenced by lack of beta-galactosidase staining. (B and D) H. felis–infected mice have substantial architectural distortion and beta-galactosidase–positive (brown) GIN. Fluorescence IHC for cytokeratin (green) and beta-galactosidase (red). (E) Glands within GIN from an infected mouse transplanted with wild-type marrow do not express beta-galactosidase. (F) Glands within GIN from an infected mouse transplanted with ROSA26 marrow demonstrate beta-galactosidase expression (red), colocalized with cytokeratin (green) to form yellow, confirming epithelial differentiation of integrated BMDC. Occasional mononuclear leukocytes are beta-galactosidase positive (red) and cytokeratin negative. Scale bars, 400 μm [(A) and (B)], 160 μm [(C) and (D)], 40 μm [(E) and (F)].

As a further additional test for bone marrow origin, we used a completely independent model of labeled bone marrow reconstitution. Female C57BL/6 mice were lethally irradiated, transplanted with bone marrow from male transgenic mice expressing chicken beta-actin-EGFP (enhanced green fluorescent protein), and infected with H. felis for 15 to 16 months. Dispersed gastric mucosal cells from these mice were sorted by flow cytometry (Fig. 3A) into GFP+/CD45+ (P4), GFP+/CD45 (P5), and GFP/CD45 (P6) populations, subsequently stained for pan-cytokeratin, and also analyzed by fluorescent in situ hybridization (FISH) for X and Y chromosomes (9). The CD45 populations were consistently positive for pan-cytokeratin (Fig. 3, D and F), which indicates their epithelial nature, whereas CD45+ cells were negative for this marker (Fig. 3B). GFP/CD45 cells contained two X chromosomes (Fig. 3G), which confirms them to be of host origin, whereas all GFP+ cells were consistently positive for both X and Y chromosomes by FISH (Fig. 3, C and E), which demonstrates that they are of donor bone marrow origin. Analysis of tissue sections (Fig. 4) from these mice demonstrated that tumor cells were GFP positive (brown stain) and Y chromosome positive (green signal) and expressed cytokeratin (red signal) (9). These studies, using two independent markers (GFP and Y chromosome), confirmed that in Helicobacter-infected mice, bone marrow–derived cells can give rise to gastric epithelial cancer.

Fig. 3.

(A) FACS sorting of gastric mucosal cells from long-term H. felis–infected female mice transplanted with male GFP+ bone marrow. Three populations were sorted and characterized further. (B) The population sorted as GFP+/CD45+ (P4) does not stain for cytokeratin (hematoxylin counterstain). (C) X (red) and Y (green) FISH confirms that the GFP+/CD45+ (P4) population is made up of donor-derived leukocytes. (D) The population sorted as GFP+/CD45 (P5) is made up of donor-derived engrafted gastric mucosal cells that stain for cytokeratin (brown) and (E) contain both X and Y chromosomes by FISH. (F) GFP/CD45 cells (P6) are cytokeratin positive and (G) contain two X chromosomes, confirming that they are host-derived gastric epithelial cells.

Fig. 4.

(A) Female wild-type mouse transplanted with female wild-type marrow does not stain for GFP by IHC. (B and C) Female mouse transplanted with male GFP marrow shows IHC for GFP (brown staining) in tumor cells. (D) FISH for Y chromosome (green) is negative in the female-to-female transplant (cytokeratin; red). (E and F) Tumor from male-to-female transplants show numerous Y chromosomes (green) within the nuclei (black) of cytokeratin-positive (red) cells.

We examined whether lesser degrees of injury would result in BMDC engraftment in the stomach. Engraftment of BMDCs in the gastrointestinal tract has been reported to occur infrequently (6, 15). In our system, lethal irradiation and bone marrow reconstitution without H. felis infection did not induce appreciable BMDC engraftment; only 2 glands containing BMDCs were found in 1780 glands examined at 30 weeks after transplantation (fig. S3, A, B, and C). Next, we induced acute gastric ulceration by serosal cryoinjury or submucosal acetic acid application (9) and examined mice 4 (fig. S3, D and E), 10, or 20 (fig. S3, F and G) days later. We found few beta-galactosidase–positive leukocytes and fibroblast-like cells (blue) at the ulcer base and edge but no engraftment of BMDCs as epithelial cells.

Previous studies have shown that targeted ablation of parietal cells with transgenic or chemical approaches results in increased cellular proliferation, altered patterns of differentiation, and mucous cell metaplasia (11, 16). To determine whether parietal cell loss plays a role in recruitment and retention of BMDCs, parietal cells were depleted with DMP777 (9, 11), followed by parietal cell repopulation after drug withdrawal. Tissue was examined for BMDC engraftment by X-gal staining at 9 days or 30 weeks after drug withdrawal. At 30 weeks, scattered leukocytes and fibroblast-like cells were bone marrow derived (blue staining), but we did not demonstrate any gastric epithelial cells of bone marrow origin (fig. S3, H and I).

Using three methods, we next addressed whether fusion occurred between BMDCs and the gastric epithelium. First, we examined histological sections containing bone marrow–derived epithelial cells and found that cells contained only a single nuclei; no binucleate cells were seen. Second, we used fluorescence-activated cell sorting (FACS) of propidium iodide–stained cells to determine DNA content in wild-type tissues, early infection prior to engraftment, and BMDC carcinoma, and did not demonstrate any difference between these groups (fig. S4). Third, in female mice transplanted with male GFP transgenic marrow, we evaluated 10,000 GFP+/CD45 pancytokeratin+ FACS-sorted gastric cells, using FISH, and showed a single X and a single Y chromosome in all cells examined (9) (Fig. 3E). These studies strongly suggest that stable fusion did not occur.

In our initial reconstitution studies, whole bone marrow was used to minimize cell manipulation, which can alter growth potential and behavior of stem cells (4, 6). To identify the population of cells within the bone marrow responsible for gastric mucosal engraftment, we cultured hematopoietic stem cells (HSC) (lineage depleted, RhodullHodull) or adherent mesenchymal stem cell (MSC) populations in transwell culture plates in contact with the soluble components of control medium or culture medium from primary gastric epithelial cell cultures (9). Neither HSC nor MSC populations expressed epithelial markers (TFF2 or KRT1-19) at the time of isolation or after culture with control medium. MSC cultures, but not HSC cultures, showed a marked up-regulation of both KRT1-19 and TFF2 at 24 and 48 hours when exposed to the soluble components of gastric epithelial tissue, which demonstrates that MSC (but not HSC) acquired a gastric mucosal cell–gene expression pattern without cell-to-cell contact or fusion (fig. S5). Western blot analysis performed on whole gastric mucosa from H. felis–infected C57BL/6 mice (12 and 16 months after infection) showed a substantial up-regulation of SDF-1 and SCF-1 (two factors identified in mobilization and migration of marrow progenitor cells) compared with uninfected age-matched controls (fig. S6). These data suggest that Helicobacter infection can give rise to an environment conducive to marrow stem cell recruitment, with the MSC the most likely candidate.

The experiments described here show that, in response to chronic Helicobacter infection, bone marrow–derived cells can home to and repopulate the gastric mucosa and contribute over time to metaplasia, dysplasia, and cancer. Demonstration of malignant progression of a marrow-derived progenitor cell in the setting of chronic inflammation offers the basis for a new model of epithelial cancer. Many features of cancer cells become much clearer when viewed within the context of our model: their undifferentiated nature, ability for self-renewal, relative resistance to apoptosis, and propensity for metastases and early spread. These are properties that may be inherent to BMDCs rather than characteristics acquired over time. Striking similarities have been noted between cancer cells and stem cells (17, 18), with recent in vitro studies suggesting that adult MSCs may be targets for malignant transformation (19). On the basis of the data presented here, we conclude that H. felis–induced gastric cancer originates from a cell population within the adherent MSC population. Our data further support that stable fusion between the BMDCs and the gastric mucosa does not occur; however, our experiments were not designed to evaluate fusion with reductive division. The concept that epithelial cancers can arise from BMDCs greatly alters our overall understanding of cancer initiation and progression and has broad implications for the development of anticancer therapies.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Table S1

References and Notes

Stay Connected to Science

Navigate This Article