Aldehyde Dehydrogenase Expressing Colon Stem Cells Contribute to Tumorigenesis in the Transition from Colitis to Cancer

Immunohistochemistry

All tissues were retrieved under pathologic supervision with IRB approval at University of Michigan and the University of Florida. Normal colon tissues were retrieved from one of two sources: Either control tissue at least 10 cm away from a malignancy, or from patients presenting for resection of benign disorders such as slow-transit constipation. Routine immunohistochemistry/immunofluorescence for ALDH1 (1:100 BD Biosciences), CD44 (1:50 Clone G44-26, Bectin-Dickinson), IL8R (1:100 Abcam) and MUC2 (1:100 clone CCP58, Zymed) from paraffin embedded sections were completed. Either MOM (Vector labs for ALDH) or ARK (DAKOCytomation, for MUC2 or CD44) antigen retrieval methods were used in xenograft sections. Other antibodies include CD30 (1:10, Catalog #Z0751, DAKO), CDX2 (1:50 Abcam), CD45 (1:100, Catalog #M0701, DAKO), CAM 5.2 (1:1, Catalog #A0452, BD Biosciences), AE1/AE3 (1:100, Catalog #M4502, Millipore), CK19 (1:100 Abcam), CK7 (1:100, Catalog #M7018, DAKO), CD2 (1:100, Catalog #A0452, DAKO), CD3 (1:100, Catalog #M7254, DAKO), c-kit/CD117 (1:100, Catalog #A4502, DAKO) and S100 (1:250, Catalog Z0311, DAKO). The stromal isolates were stained with SMA (1:600 R&D Systems). For sphere studies, the spheres were concentrated into a 1% low melting point agarose plug prior to embedding in paraffin.

Patient subjects and xenograft assays

The University of Michigan and the University of Florida Institutional Animal Care and Use Committees approved the following protocols (UM 9510 and UF F090, respectively). For colon cancer tissue, cancer and matched normal tissues were retrieved from the operating room and maintained sterilely. For colitic tissues, both distal and proximal regions of the colon were implanted. Only those patients with ulcerative colitis were used. Patients with Crohn’s disease were not evaluated in this study. All tissue retrieval was completed under pathologic supervision. Tissues were minced and placed subcutaneously into the flanks of NOD-SCID mice. For cellular xenografts, the selected numbers of cells were pelleted post flow cytometry. The pellets were resuspended in Matrigel (BD Biosciences) such that the total volume for injection was not greater than 100 μl. The opposite flank received the control injection. For the antibody blocking experiments, commercially available pellets (Innovation Research of America, Sarasota, Florida) were embedded with either anti-IL8 or anti-IL6 (R&D Systems, 5 micrograms/day) or placebo (control) and implanted one day prior to placement of the xenografts. To account for variability, at least 8 mice/condition were initiated. Mice were monitored daily for tumor growth and measured at least weekly to establish tumor volume.

Flow cytometric sorting

The tissues were digested with collagenase. Antibodies were added at a concentration of 1:100 and maintained for 30 minutes on ice. The cell pellets were washed with HBSS/2%FBS, and in circumstances where a secondary antibody is required, the cells are resuspended in the appropriate secondary antibody. For identification of CD44 and ESA, the anti-CD44 allophycocyanin (APC, Pharmingen, Franklin Lakes, NJ) and anti-epithelial specific antigen (ESA)-FITC (Biomeda, Foster City, CA) were employed at a dilution of 1:40. Nonviable cells are eliminated by using the viability dye, DAPI, just prior to submission for flow cytometry. Murine cells were eliminated by staining with H2K^d (Southern biotech). Flow cytometry was performed on the FACS Aria (BD Immunocytometry Systems, Franklin Lakes, NJ). Side scatter and forward scatter profiles was utilized to eliminate cell doublets. Cells were routinely subjected to double sorting, and reanalyzed for purity, which is typically greater than 97%. The Aldefluor kit (Stem Cell Technologies, Vancouver, Canada) was used to identify the ALDH positive population. Controls included the viability stain, DAPI, the Aldefluor stained cells treated with inhibitor of the enzyme (DEAB), and staining with the secondary antibody alone.

Sphere culture

Flow cytometry was utilized to sort cells from xenografts either from colorectal oncogenic or colitic sources. Isolated cells were placed into a low attachment six well plate at a density of 2000–10,000 cells/well. These cells were suspended in a serum-free medium containing DMEM/F12 (Gibco), 10 nM progesterone (Sigma), glutamine, Gibco), 10 μg/ml insulin (Sigma), 13 μg/ml transferrin (Sigma), 15 nM sodium selenite (Sigma), 4 mg/ml BSA (Sigma), 50 μM putrescine (Sigma), and 15mM HEPES (Gibco), supplemented with 10 ng/ml FGF and 20 ng/mlEGF (Sigma).

Isolation and maintenance of primary stromal isolates

Tissues from colitic, colon cancer, and matched tissues were minced and digested with Collagenase type 1A (Sigma) for 30 minutes at 37°C. The resulting suspension was plated in serum containing media (DMEM 10% FBS, Gibco). For these experiments, successful isolates were employed in passages 4–9. Prior to use, observation of phenotype and staining with α smooth muscle actin was used to confirm these cells as fibroblasts For these experiments, successful isolates were employed in passages 4–9. For comparison, normal colon fibroblasts (CRL 1541) and colon cancer fibroblasts (CRL7213) were purchased from the American type culture collection (ATCC, Manassas, Virginia)

Cytokine array analysis

Conditioned media was harvested from stromal isolates. Conditioned media was harvested after plating of successful isolates at a density of 10⁵ cells in a well of a 24-well tissue culture treated plate. After the cells had adhered, the media was removed and replaced with fresh media in the absence of serum. After 24 hours 37° C, 5% CO₂, the overlying media was harvested for use in the cytokine/chemokine array. The Chemicon cytokine array (Human Cytokine Array III, Millipore) was conducted per manufacturer’s instructions.

Statistical Analysis

The incidence of ALDH expressing epithelium was enumerated in normal vs. colitic epithelium. In total, at least 5000 epithelia per disease type were counted. The data were analyzed as a two sample t-test with p < 0.05 taken as significant. For comparison of tumor latencies, growth rates, and size, ANOVA using Duncan’s multiple comparison procedure at the 0.05 significance level was determined.

ALDH+ epithelial cells from select colitis patients progress to adenocarcinomas in NOD-scid mice xenografts

The xenograft model is widely used for testing tumorigenicity of putative cancerous cells. Therefore, we employed xenografting of ESA^high/ALDH+ primary tumor and colitic cells into NOD/SCID mice to determine tumor initiation and differentiation potential. In figure 1, we show staining for ALDH1 in normal (A), colitic, and malignant colon tissues and found increased ALDH1 immunoreactivity at the base of the crypt in a subset of colitic samples (B, and supplemental figure 1), and widespread expression throughout colitis-associated cancer tumor tissue (C). Quantification of epithelial cells in grossly normal and colitic tissues supported our observations that ALDH is overexpressed in colitic epithelium (D). In our studies, as few as 25 DAPI^low ESA^highALDH^high cancer cells reproducibly generated serial xenografts with typical adenocarcinoma phenotypes (Figure 2A). To determine whether colitic tissue might bear the same tumorigenic potential as frank colon cancer, random tissue was isolated from patients presenting for resection of their colorectum due to chronic ulcerative colitis, with the majority of these patients presenting for surgery due to medically refractory disease (19/22; Supplemental Table 1). Classically, ulcerative colitis may be more severe distally in the colon, and less severe or relatively normal in the proximal colon. Therefore, portions of the proximal colon were retained as a matching control tissue. Proximal biopsies often revealed quiescent or nearly normal histology (data not shown). For every specimen retrieved, the pathologic evaluation based on histology of the flanking sections to the distal implanted tissues showed only ulcerative colitis without evidence of either dysplasia or adenocarcinoma (Supplemental Table 1). The histopathology of the flanking sections revealed inflammation ranging from quiescent disease to severe chronic colitis (Supplemental Table 1). Colitis samples were implanted into the subcutaneous flank of NOD-SCID murine hosts. and were observed for up to 5 months. In contrast to frank colorectal cancer, in which we now routinely engraft >75% (2) of the cancer tissue successfully, xenografts evolved from only the three of twenty-two colitic specimens (13.6%). No tumors were found in the contralateral flank containing the proximal normal colonic tissue implants from the same patients (not shown). Similar to the frank colon cancer xenografts, all three of the primary xenografts resulting from the colitic tissues could form secondary xenografts (Figure 2B) and were serially passaged in mice. Secondary xenografts were formed from both bulk primary xenograft cells and flow cytometric sorted for DAPI^lowESA^highALDH^high cells. ESA was included to ensure isolation of predominantly epithelial cells rather than any potential ALDH^high hematopoietic cells in the colon/xenograft. As few as 50 DAPI^lowESA^highALDH^high cells were capable of tumorigenesis via serial passage of the xenograft. Further we were able to passage these tumors sequentially at least ten passages with continued enrichment for ALDH^high cells.

The colitic cells within the early passage anaplastic xenografts did not stain for a series of tissue specific markers including CDX2, c-kit, CD3, CD2, β-catenin, CD30, vimentin, CK20, cam52, and CD45 (data not shown). While CDX2 was employed to evaluate differentiated intestinal tissues, both CD3 and CD2 (pan T-cell and T-cell subset) markers were absent. Similarly, the CD45, or common leukocyte antigen, was absent. CD30, a marker of anaplastic lymphomas, was not seen. Staining for neither CK20, a marker of differentiated colorectal carcinomas, nor CAM 5.2, for low molecular weight cytokeratins, were visualized. Further evaluation included S100, evident in some melanomas and therefore included in the differential for anaplastic lesions, was absent. To evaluate the possibility that these tumors were variants of gastrointestinal stromal tumors, they were stained with the CD117 antibody (c-kit antibody), and they were negative for this marker. Instead, they displayed a classic anaplastic morphology of expanded, homogeneous cells (Figure 2B, middle). Continued xenograft passage of ESA^highALDH^high selected cells resulted in progression of tumor morphology to a phenotype which was diagnostic of adenocarcinoma, with hallmark production of mucus and glandular formation (Figure 2B, right). In those colitic tissues with successful engraftment, the latency of tumor growth from bulk cell populations was similar to cancer counterparts with onset of tumor growth averaging six weeks. Enrichment for ESA^highALDH^high selected cells resulted in palpable tumors in an average of two weeks. Once palpable, xenografts grew at similar rates regardless of source. Matched grossly normal tissues taken from the same patients did not engraft (N=22). This finding led us to categorize the original xenograft initiating cells isolated from colitis patients as precursor-Colon Cancer Stem Cells or pCCSC in order to distinguish them from the CCSC isolated from frank colon cancer.

Epithelial marker expression of primary tumors, xenografts, and spheres

Findings by others have shown that neural stem cells can be characterized by their ability to form sphere shaped clusters, called neurospheres, in serum free medium (28, 29). Sphere formation has subsequently been used to isolate both normal and cancer stem/progenitor cells from a wide variety of tissues including the colon (4, 30, 31). We have established conditions that support the formation of colon spheres from both frank colon cancer and colitis, but not from normal colon. We used immunohistochemistry to delineate the phenotypes of the original tissues compared with both xenografted tissues and with the colon spheres (Figures 3 and 4). MUC2 is present in goblet cells during mucin formation, and is indicative of epithelial differentiation (32). We observed MUC2 staining in original tissues and in the xenografts from both colon cancer and serially passaged colitis sources with an adenocarcinoma histological profile (Figures 3 and 4A–B right panels). Expression of MUC2 was only visualized at the outer edges of the spheroid colonies generated from the frankly malignant source and not in the colitis spheres (Figure 3 and 4C, right panels), reflecting a more anaplastic nature. ALDH1 and CD44 immunoreactivities were also compared revealing dispersed cytological localization in both the primary tumors and resulting xenografts. CD44 expression was seen in primary tissues, xenografted tissues, and spheres from both origins (Figures 3 and 4A–C, left panels). However, though ALDH expression was seen in sections of all primary patient samples, the expression was much more restricted in successful xenografts and in the spheres from colitic origins (Figure 4B and C middle panels).

The Stromal Microenvironment Changes in Colitis and Cancer

Recent findings suggest that alterations in normal stroma may support tumorigenesis, implying that the inflammatory environment found in some colon cancers and colitic tissues includes paracrine interactions that contribute to oncogenesis (18, 33). Indeed, cytokines including IL8 have been associated with inflammatory and oncologic processes in the colon (34). Therefore, we isolated colonic stromal cells (Supplemental Figure 2) from the colitic milieu and compared the cytokine profile of conditioned medium from the colitic fibroblasts to normal colon and malignant colon fibroblasts using cytokine array analysis. The expression of IL6 and IL8 was low to undetectable from the normal stromal cell medium (Figure 5A top), but was substantially increased from the medium taken from both the colitic (Figure 5A middle) and frankly malignant (Figure 5A bottom) stromal environments. In most cases the colitic stroma produced an intermediary amount of IL6 and IL8 compared to the baseline expression of normal colon stroma and high-level expression from colon cancer stroma. IL8 has at least one high affinity receptor, CXCR1, which we show is prominently expressed in malignant colon and colitic xenografts of ESA^highALDH^high sorted cells (Figure 5B). We next evaluated whether colitis or cancer derived stroma enhance the ability of colon cancer spheres to form xenografts. We co-injected sphere-derived single cell suspensions alone or in the presence of normal colon/colon cancer/colitis associated fibroblasts and monitored the mice for tumor volumes and latencies. Those mice with both sphere-derived cells and cancer-associated fibroblasts developed tumors with a markedly decreased latency - mean time to palpable tumor <2 weeks with cancer/colitis fibroblasts versus ~6 weeks for spheres alone or with normal fibroblasts (Figure 5C) providing strong evidence that fibroblast-derived IL6 and/or IL8 plays a definitive role in tumorigenesis. Moreover, we show that colon cancer stem cells, when transplanted alone, result in longer tumor latency and smallest tumor volume in xenograft experiments (ANOVA, Duncan’s multiple comparison p < 0.05). Mixed model analysis for tumor growth rate in the colon cancer sphere with colitic fibroblasts revealed a growth rate which was significantly higher than that in the colon cancer spheres co-injected with either the colon cancer fibroblasts, normal colon fibroblasts, or spheres alone (p = 0.0006, < 0.0001, and < 0.0001, respectively). The tumor growth rate for the colon cancer spheres with colon cancer fibroblasts group was significantly higher that that in colon cancer spheres alone group (p = 0.0006). Injections consisting of fibroblasts alone failed to generate a xenograft, which was confirmed at necropsy (not shown). As shown in the insets (Figure 5C), for this experiment, the colitic fibroblasts secreted more IL6 and IL8 than the colon cancer fibroblast cell line.

To further determine a role for these cytokines in tumorigenesis, we implanted slow-release pellets containing either anti-IL8 or anti-IL6 antibodies adjacent to the xenografted cells. Both conditions resulted in significantly reduced tumor volume compared to the control group (Figure 5D). Moreover, palpable tumors increased in size after 8 weeks, the period at which the antibody supply in the pellets was exhausted. Together, these data indicate a direct role for these cytokines in tumor growth in vivo.