Transition from Colitis to Cancer: High Wnt Activity Sustains the Tumor-Initiating Potential of Colon Cancer Stem Cell Precursors

Chronic inflammation is associated with many cancers (1). For example, patients with ulcerative colitis, a chronic inflammatory disease of the large bowel, are at substantially increased risk of developing a form of colorectal cancer (CRC) known as colitis-associated cancer (CAC; refs. 2, 3). Moreover, few tools are available for detecting the colitis-to-cancer transition that allows early diagnoses at more treatable stages. Although CAC is thought to involve mutational events, epigenetic modifications, and influences from the microenvironment, the pathogenesis of CAC is unclear.

A gene frequently found to be mutated early in the adenoma-to-carcinoma sequence in sporadic CRC is Adenomatous Polyposis Coli (APC), a tumor suppressor, which is a negative regulator of Wnt (wingless)/β-catenin signaling (4, 5). APC is also found to be mutated at the germline level in familial adenomatous polyposis (FAP), an autosomal dominant colon disorder, which promotes the development of CRC (6). APC is an important component of the destruction complex in the Wnt/β-catenin signaling pathway, which includes glycogen synthase kinase 3β (GSK3β), axin 2, and casein kinase 1α (CK1α) and β-catenin.

In the absence of Wnt ligand, β-catenin is located primarily at the cell membrane, along with a small, dynamic cytoplasmic pool, which is targeted for proteosomal degradation when bound to the destruction complex. In the presence of Wnt ligands, the destruction complex undergoes dissociation, resulting in the stabilization and accumulation of free β-catenin in the cytoplasm, and subsequent translocation into the nucleus. Following nuclear translocation, β-catenin displaces transcriptional corepressors, allowing direct binding to the transcription factor, T-cell factor (TCF)/lymphocyte enhancer factor (LEF) and subsequent transcription of Wnt-target genes including axin2, LGR5, c-Myc, and cyclin D1 (4). Indeed, activating mutations in any one of the components of the Wnt/β-catenin signaling cascade results in constitutive activation of this pathway as observed in cases of sporadic CRC and FAP (4). Furthermore, high Wnt/β-catenin signaling in sporadic CRC cells has been shown to mark the cancer stem cell (CSC) compartment (7). Other roles for Wnt/β-catenin signaling in the intestinal tract include maintenance of adult crypt structure and proliferation of intestinal epithelial progenitor cells (5, 8).

The contribution of Wnt/β-catenin signaling to the colitis-to-cancer transition is poorly understood. Initial mutational studies suggested that activation of Wnt/β-catenin signaling in the colitis-to-cancer transition is much less frequent than in sporadic CRC, and occurs "later" in the pathogenic cascade (9). As discussed below, our findings suggest that Wnt/β-catenin signaling may occur earlier and be more prevalent than previously thought. We and others reported that sporadic CRC is initiated by a rare population of crypt cells called colon cancer stem cells (CCSC; refs. 10–13). We showed that high aldehyde dehydrogenase (ALDH) expression marked the CCSCs and enriched the tumor-initiating cell (TIC) population (10). We recently identified a similar ALDH^high population in the normal-appearing, nondysplastic colonic epithelium of ulcerative colitis (UC) patients (14). These results suggested for the first time that CAC might have a CSC origin. Because they have the capacity to initiate the colitis-dysplasia-cancer transition, we refer to these cells as precursor-CCSCs (pCCSC). pCCSCs can be propagated in vivo as tumor xenografts and in vitro as nonadherent spheres. The success rate of generating spheres from ALDH^high cells derived from nondysplastic colitic colon is low, 5% to 13% (14) and similar to the incidence of CAC in the UC population, approximately 2% to 19% (2).

Our findings to date indicate that pCCSCs are a valuable experimental model with which to interrogate the colitis-to-cancer transition, especially when considering early, preneoplastic events (14). In particular, use of this model could pave a path for the development of methods that could aid early disease diagnosis and targeted drug therapy, which together may prevent the progression from colitis to cancer. Although pCCSCs might be involved in the colitis-to-cancer transition (14), mechanisms underlying this transformation are unknown. Because initiation of sporadic CRC has been associated with activating mutations in the Wnt/β-catenin signaling pathway (15, 16), and CCSCs exhibit high Wnt/β-catenin signaling (7), we reexamined the role of Wnt/β-catenin signaling in the colitis-to-cancer transition. We proposed the existence of a Wnt/β-catenin–dependent CSC hierarchy that is operative "early" in the pathogenesis of CAC. We show that high Wnt/β-catenin signaling enriches tumor initiation activity from pCCSCs to such a degree that recapitulation of the adenocarcinoma phenotype is possible from a single cell.

Tissues from colitis patients and colon cancer patients were retrieved under pathologic supervision with Institutional Review Board approvals at the University of Florida and the University of Michigan. Normal colon tissues were obtained from a local organ procurement organization (Life Quest).

Inbred nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice (5–6 weeks old) were used. Mice were maintained under pathogen-free conditions. Experiments were approved by the University of Florida Institutional Animal Care Committee.

ALDH^high sphere isolates were obtained from UC and CRC patients. The former are pCCSC; the latter are CCSC. The sphere isolates used in the study are CT-1, CT-2, and CA. CT-1 and CT-2 are 2 colitis sphere isolates (pCCSC) obtained from 2 different colitic patients, CT-1 suffered from colitis for 8 years and CT-2 for 3 years. CA is a cancer sphere isolate (CCSC) obtained from a sporadic CRC patient. Isolated cells were cultured in serum-free media as previously described by Carpentino and colleagues (14). Serum-free media is referred to as defined media.

Cells with high or low eGFP intensities were deposited at 1, 2, 3, 4, 6, 8, 10, 12, 16, 18, 20, and 24 cells per well of 96-well, ultralow adhesion plates (Corning) containing defined media. For each cell density, 8 wells were plated. Clonal frequency and statistical significance were evaluated with the Extreme Limiting Dilution Analysis "limdil" function (7). This assay was carried out with cells obtained from Wnt^high-pCCSCs and Wnt^high-CCSCs and from ESA⁺/H2K^d− cells derived from dissociated tumor xenografts.

For primary tumor xenografts, 10, 100, and 1,000 Wnt^high-pCCSCs and Wnt^high-CCSCs (sphere isolates) with the 2% lowest and the 2% highest eGFP expression levels, corresponding to Wnt^high and Wnt^low cells, were deposited by fluorescence-activated cell sorting (FACS) into a 96-well plate containing defined medium admixed with Matrigel at a 1:1 ratio such that the total volume was 100 μL and injected as previously described (10). For secondary tumors, the ALDH^high-Wnt^high and ALDH^high-Wnt^low primary tumor xenografts were dissociated and the 10, 100, and 1,000 ESA⁺/H2K^d− cells with the 10% highest and 10% lowest eGFP expression levels (Supplementary Fig. S2A)—corresponding to Wnt^high and Wnt^low—were injected as described above (Supplementary Fig. S2B). Single cell injections were carried out as described for the secondary tumors for both Wnt^high and Wnt^low cells enriched from primary tumor xenografts. Likewise, single cell ALDH^high secondary tumors were generated. Tertiary tumors were generated with 10, 100, and 1,000 ESA⁺/H2K^d− cells with the 10% highest and lowest eGFP intensities enriched from Wnt^high secondary tumors (Supplementary Fig. S2B). The mice were sacrificed either after 12 weeks posttumor injection or if the tumor size reached 1 cm × 1 cm. The tumors were measured twice a week using digital calipers.

One hundred Wnt^high-pCCSCs and Wnt^high-CCSCs (sphere isolates) with the 2% lowest and highest eGFP intensities (corresponding to Wnt^high and Wnt^low cells), admixed with Matrigel at 1:1 ratio in a total volume of 100 μL, were injected subcutaneously into the hind flanks of NOD/SCID mice. At day 4 postinjection, the mice were injected with 2.5-mg/kg indomethacin (Calbiochem) or DMSO (control) i.p. every 12 hours for 6 weeks. For details of β-catenin knockdown refer to the supplementary methods, available online. For β-catenin knockdown tumors, 100 cells each of scrambled (Sc) [control], #1 and #2 shRNA transduced Wnt^high-pCCSCs, and Wnt^high-CCSCs were prepared for injection as mentioned above. Tumor size was measured twice a week for 6 weeks using calipers.

Data are presented as means ± SE as indicated in the figure legends. Statistical significance was defined as P ≤ 0.05 determined by an unpaired Student's t test or a 1-way analysis of variance. To compare tumor growth rates, a mixed linear model was used with tumor volume as the response variable and with time and group as explanatory variables. We included subject (mouse) as a random effect, and we assumed a compound symmetric covariance structure.

To determine the functional importance of Wnt/β-catenin signaling in the colitis-to-cancer transition, we did immunohistochemistry to detect the presence of nuclear/cytoplasmic β-catenin and ALDH in colonic tissues from healthy controls, UC patients, and sporadic CRC patients. Patient characteristics for these tissues are in Supplementary Table 1. Active Wnt/β-catenin signaling was detected by nuclear/cytoplasmic β-catenin staining (Supplementary Fig. S1A). There were significantly more Wnt/β-catenin pathway active cells in colitis (2.5-fold increase) and CRC (4.5-fold increase) than in normal colon (Supplementary Fig. S1B). Similar to nuclear/cytoplasmic β-catenin, the percentage of ALDH⁺ cells in colitis samples was in between normal colon and colon cancer samples (Fig. 1A and B; ref. 14). This suggests that both Wnt/β-catenin signaling activity and ALDH expression in normal, colitic and CRC colon tissue samples may parallel the normal-to-colitis-to-CRC transition. The coimmunostaining results showed an expansion of the Wnt-active/ALDH⁺ TIC population from normal-to-colitis and colitis-to-CRC (Fig. 1A). Indeed, about 52% of the ALDH⁺ cells in colitis were Wnt-active, indicating that Wnt-active cells represent a major subpopulation of pCCSCs (Fig. 1B). Our immunostaining data thus suggest an "early" role for activation of Wnt/β-catenin signaling in the colitis-to-cancer transition.

To ascertain the functional importance of Wnt/β-catenin signaling activity in pCCSCs during the colitis-to-cancer transition, we used a lentiviral dual fusion Wnt reporter, TTLG (consisting of 6x TCF/LEF binding sites and a minimal thymidine kinase promoter regulating firefly luciferase and eGFP reporter genes; Fig. 2A). TTLG was transduced into pCCSCs (CT-1 and CT-2) as well as into CCSCs (CA) derived, respectively, from UC and sporadic CRC patients. CCSCs were operationally defined as ALDH^high cells derived from CRC tissues that have shown the ability to undergo serial passaging through immunocompromised mice with limited cell numbers while retaining the ability to recapitulate the primary tumor (10). CCSCs served as a control for pCCSCs throughout the study, as CCSCs with high Wnt/β-catenin signaling have been shown to display cancer SC properties (7). pCCSCs were operationally defined as ALDH^high cells derived from colitic colon that (i) were isolated from nondysplastic UC colon, (ii) could be serially passaged through immunocompromised mice, and (iii) developed, first an anaplastic phenotype, and on serial passage transformed into a poorly differentiated adenocarcinoma (14). TTLG specificity was verified by comparing the expression of Wnt target genes (Fig. 2B, Supplementary Fig. S3A), in TTLG-eGFP^high and TTLG-eGFP^low populations of pCCSCs and CCSCs. These correspond, respectively, to Wnt^high and Wnt^low cell populations. TTLG specificity was confirmed by nuclear/cytoplasmic active β-catenin (ABC) staining, which indicated active Wnt/β-catenin signaling (Fig. 2C, Supplementary Fig. S3B). Wnt^high-pCCSCs were distinguishable from CCSCs based on differences in Wnt target gene expression profiles (Fig. 3B). Collectively, our findings not only confirm that TTLG is a valid Wnt/β-catenin signaling reporter, but also supports the findings of immunochemical analysis demonstrating that active Wnt/β-catenin signaling demarcates a major subpopulation of pCCSCs.

To determine the functional significance of Wnt/β-catenin signaling in pCCSCs, we subjected ALDH^high-Wnt^high and ALDH^high-Wnt^low sphere cells (pCCSCs and CCSCs; Supplementary Fig. S2A) to in vitro clonogenic assays [limiting dilution assays (LDA)] and in vivo tumorigenic assays under limiting dilution conditions (Supplementary Fig. S2B). Tumors so obtained are designated as primary xenograft tumors. However, there was no significant difference in the overall frequency of clonogenicity (Fig. 3A–C) or rate of tumor formation (Supplementary Table 2) for the ALDH^high-Wnt^high versus ALDH^high-Wnt^low primary xenograft tumors. These results may be attributed to a starting cell population of ALDH^high-pCCSCs and CCSCs that are already enriched for progenitor and SCs (14). As our study involves enrichment of the TIC fraction, the purity of the cell populations may not be absolute. This was confirmed by eGFP expression studies: FACS analysis revealed heterogeneity of eGFP expression in ALDH^high-Wnt^low and ALDH^high-Wnt^high primary xenograft tumors (Supplementary Fig. S4A) where a portion of cells in the ALDH^high-Wnt^low primary xenograft tumor had increased eGFP expression corresponding to high Wnt-activity. However, the percentage of Wnt-active cells was significantly greater in ALDH^high-Wnt^high than in ALDH^high-Wnt^low primary xenograft tumors (Supplementary Fig. S4B-C). In addition, the resulting ALDH^high-Wnt^low primary xenograft tumors phenocopied the histological appearance of ALDH^high-Wnt^high primary xenograft tumors (Fig. 3D) showing a poorly differentiated adenocarcinoma phenotype. As tumor initiation is a property of both progenitor cells and SCs, we serially passaged these primary xenograft tumors based on 2 extreme levels of Wnt activity to further distinguish these 2 populations. Similar to CSCs, CCSCs retain the ability to both initiate tumors on serial passaging and to undergo self-renewal.

Wnt/β-catenin signaling has been implicated in the self-renewal of adult colon SCs and CCSCs (17, 18). To show this phenomenon and also to enrich for stem-like cells in pCCSCs, secondary and tertiary tumors were generated from the 10% highest and lowest fluorescent cells derived from primary ALDH^high-Wnt^high xenograft tumor (Supplementary Fig. S1A). Simultaneously, an in vitro LDA was conducted to determine clonogenicity (Supplementary Fig. S2B). Clonal frequency correlated with tumor-forming potential wherein ALDH^high-Wnt^high tumor xenograft cells formed tumors at a greater frequency than ALDH^high-Wnt^low tumor xenograft cells. ALDH^high-Wnt^low xenograft cells largely failed to grow with subsequent passages (9 successes in 69 attempts; Fig. 4A–F and Table 1). Also, tumors derived from ALDH^high-Wnt^high sphere and tumor xenograft cells in subsequent passages grew at a faster rate following injections of 10 cells (Supplementary Fig. S5). Similar results were obtained with secondary tumors derived from primary ALDH^high-Wnt^low xenograft tumors (Supplementary Fig. 6A–C). These data indicate that pCCSCs generate self-renewing Wnt^high cells. Thus, high Wnt-activity associates with sustained tumor initiation and self-renewal.

To test whether high Wnt-activity confers an additional level of enrichment to the already existing pCCSC marker ALDH^high, single cell injections of ALDH^high cells enriched from ALDH^high primary tumor xenografts and Wnt^high and Wnt^low cells enriched from ALDH^high-Wnt^high primary tumor xenografts were conducted. We had a 25% tumor formation success rate (5 of 20 injections resulted in a tumor) from single ALDH^high-Wnt^high tumor cell, whereas none of the ALDH^high or the ALDH^high-Wnt^low single tumor cells developed into a palpable mass (Fig. 5A and Table 1). The resulting ALDH^high-Wnt^high tumor displayed histological characteristics of a well-differentiated adenocarcinoma in contrast to the poorly differentiated adenocarcinoma phenotype of primary ALDH^high tumor xenograft and Wnt^high tumor xenograft at the primary stage (Figs. 3D and 5B). Furthermore, besides the differential β-catenin expression, the tumor revealed rare cells with expression of the goblet cell marker Muc2, thus confirming tumor heterogeneity (Fig. 5C). These results confirm the greater level of CCSC enrichment bestowed by high Wnt-activity on ALDH^high cells.

To ascertain if high levels of sustained Wnt activity are necessary for tumor initiation and growth in CAC, we used an RNA interference approach in which we inhibited β-catenin using two shRNAs, #1 and #2 (19). A scrambled (Sc) shRNA was used as a control. Following lentiviral transduction and drug selection, the efficacy of the knockdown was confirmed by FACS. eGFP levels as a measure active Wnt/β-catenin signaling was reduced for both shRNA, with shRNA #1 showing the greatest reduction, compared the Sc shRNA for Wnt^high-pCCSCs (Supplementary Fig. S7A). These results were confirmed by western blotting analysis to detect β-catenin protein (Fig. 6A). In vivo studies were done with shRNA #2 transduced into the Wnt^high-pCCSC cell population. Tumors derived from shRNA #2 transduced Wnt^high-pCCSCs and -CCSCs grew at a significantly slower rate compared with Sc transduced Wnt^high-pCCSC and -CCSC–derived tumors (Fig. 6B, Supplementary Fig. S7B). Furthermore, we used a nonsteroidal anti-inflammatory drug (NSAID), indomethacin, to attenuate Wnt/β-catenin signaling in Wnt^high-pCCSCs by suppressing β-catenin expression (20, 21). Wnt^high-pCCSC tumors in vehicle-treated mice grew significantly faster than in indomethacin treated animals (Fig. 6C). The decreased Wnt activity in tumors from mice treated with indomethacin was confirmed by immunostaining for β-catenin (Fig. 6D). This suggests a definitive role for high-level Wnt signaling in determining the rate at which colitis progresses toward CAC. Thus, therapeutic targeting of Wnt/β-catenin signaling may abrogate the progression of colitis to cancer.

In the present study, we showed early activation of Wnt/β-catenin signaling in the colitis-to-cancer transition based on immunohistochemical analysis as well as on in vitro and in vivo functional assays. Our initial immunohistochemical studies showed that colitic epithelial cells may exhibit tumor-initiating activity based on their expression of ALDH along with an elevated coexpression of nuclear/cytoplasmic β-catenin (Fig. 1A and B). We confirmed an association of high levels of Wnt/β-catenin signaling with self-renewal, sustained tumor initiation, and tumor heterogeneity, the 3 properties of CSCs. Moreover, we showed the importance of high Wnt activity as an additional enrichment marker of pCCSCs within the ALDH^high population that promotes the progression from colitis to cancer.

Initial reports suggested that the Wnt/β-catenin signaling was activated late during disease pathogenesis (22–24). However, more studies that are recent reported early Wnt/β-catenin pathway activation (25, 26) similar to that observed in CRC (27). Moreover, the conclusions of the initial reports were mainly based on mutational analysis of Wnt signaling pathway components, whereas recent findings (including our own) suggesting early Wnt/β-catenin pathway activation are based on immunohistochemical data. To further examine this finding, we generated a Wnt/β-catenin reporter, which was validated by immunostaining to detect active β-catenin and real-time qPCR for representative Wnt/β-catenin pathway target genes. We observed a consistently higher expression level of all Wnt/β-catenin pathway target genes except for c-Myc in CCSCs, which was consistent with the results of Vermeulen and colleagues (7). We speculate that the decrease in expression of some Wnt/β-catenin pathway target genes in the Wnt^high population (Fig. 2B, Supplementary Fig. S3A) could be a result of CpG island methylation as reported by de Sousa and colleagues in CRC (28).

Apart from colitis, cross-talk between Wnt/β-catenin signaling and other pathways [such as TGF-β/bone morphogenic protein, Hedgehog, Notch, and mitogen-activated protein kinase (MAPK)] have been reported during development, adult homeostasis, SC maintenance, and in other diseases (29–33). For example, in murine models of colitis, Lee and colleagues, reported activation of Wnt/β-catenin signaling in development of dysplasia, which was mediated through the phosphoinositide 3-kinase/PTEN cascade (34). On the basis of these findings, we speculate that “early” activation of Wnt/β-catenin pathway in the colitis-to-cancer transition could be the result of pathway cross-talk rather than mutational inactivation of pathway components.

For the first time in UC, our study highlighted 2 distinct functional TIC populations in pCCSCs based on serial passaging. The two populations may be analogous to long-term TIC (LT-TIC) and tumor transit amplifying cells (T-TAC) as described by Dieter and colleagues (35, 36). LT-TICs are CSCs that are characterized by sustained tumorigenicity as evidenced by in vivo self-renewal and recapitulation of tumor heterogeneity, whereas T-TAC may be the non–self-renewing progenitor population (35, 36). Previously, we showed that the minimum number of colitis-derived ALDH^high cells (pCCSCs) that initiated tumor formation in mice was 50 (14). In the present study, we compared the tumorigenic potential of ALDH^high and ALDH^high-Wnt^high cells at the single cell level and showed that a single ALDH^high-Wnt^high tumor cell was sufficient to initiate tumor formation and satisfied all three criteria for CSCs; indeed, the tumor initiating function was enriched by greater than 10-fold. The successful single cell injections suggest that the ALDH^high-Wnt^high cells reside at the apex of the CSC hierarchy (Supplementary Fig. S8). In this study, samples consisted of 2 colitis sphere isolates (pCCSCs) obtained from 2 individual colitic patients. This small sample size is because of the low percentage of colitic colons harboring pCCSCs, and attests to the difficulty of propagating these cells both in vivo and in vitro, which reflects the infrequently occurring transitions of colitis-to-cancer. In contrast, using the exact same methodology, our ability to propagate frank sporadic colon cancer as a tumor was nearly 90%, with a frequency of in vitro sphere formation of 50% to 60%.

Progenitor cells are incapable of self-renewal and only contribute to tumor formation in primary mice, that is, they fail to generate tumors on subsequent passages (35, 36). Failure of Wnt^low cells to consistently generate tumors following in vivo serial passaging suggested that this cell population is dominated by progenitor cells. The equivalent tumorigenic potential between Wnt^high- and Wnt^low-pCCSC and -CCSC populations during primary xenograft transplantation is consistent with the findings of David and colleagues for CRC (37). However, in contrast to our study, those authors reported no correlation between the level of Wnt/β-catenin signaling and tumorigenicity. This discrepancy could possibly be explained by differences in the starting populations and enrichment techniques. Moreover, unlike our findings, their conclusions were based on primary tumor xenograft studies. We attribute the equivalent clonogenic and tumorigenic potential of Wnt^high and Wnt^low cell populations to the presence of both CSC and progenitor cells in the ALDH^high starting population, which was confirmed by serial transplantations of the Wnt^high and Wnt^low primary tumor xenografts. The ability of the control Wnt^high-CCSCs to propagate with serial in vivo passages are in agreement with the results of Vermeulen and colleagues, for sporadic CRCs, wherein they reported that Wnt^high-CCSCs are capable of self-renewal (7).

We also tested whether high Wnt activity could be used as a therapeutic target to mitigate the colitis-to-cancer transition. To this end, we knocked down β-catenin levels using an RNAi strategy to reduce Wnt activity. Knockdown by shRNA #1 was so effective that the transduced cells did not survive more than a week in culture, likely owing to role of Wnt/β-catenin pathway in survival (38). We observed a reduction in tumorigenicity of Wnt^high-pCCSCs, following β-catenin knockdown (shRNA #2; Fig. 6B), which was further confirmed by pharmacological inhibition of β-catenin using indomethacin, an NSAID. Previously, indomethacin was used as pharmacotherapy for treating UC. Although this drug is no longer used to treat UC (39–41), our study indicates the potential benefits of Wnt inhibition using pharmacological agents in attenuating the colitis-to-cancer transition.

Currently, there exists no direct evidence of a genetic cause for the increased risk of CRC in patients with UC. However, several molecular consequences such as generation of reactive oxygen species, microsatellite instability, telomere shortening, and chromosomal instability have been attributed to inflammation-driven genomic stress that leads to CRC (42). Although these studies have shed light on our understanding of inflammation-associated carcinogenesis, these markers lack sensitivity or specificity to be used as reliable biomarkers with which to assess the risk of CRC in patients with UC (42–46). Here, we not only identify the pCCSCs in UC colons, but also, we show that high Wnt/β-catenin signaling activity could be one of the mechanisms that drives the colitis-to-cancer transition. While ALDH may be a more inclusive marker for pCCSCs and CCSCs, the use of ALDH^high-Wnt^high as a marker panel may provide a more specific method of screening for pCCSCs in patients with colitis and indicative of an increased risk of malignant transformation in UC patients. Those chronic UC patients bearing an epithelial phenotype exhibiting high Wnt activation might be best served by a prophylactic colectomy. Additional clinical studies are warranted to rigorously show such a correlation.

No potential conflicts of interest were disclosed.

Conception and design: A.K. Shenoy, R.C. Fisher, E.W. Scott, E.H. Huang

Development of methodology: A.K. Shenoy, E.A. Butterworth, E.H. Huang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.K. Shenoy, E.A. Butterworth, L.-J. Chang, E.H. Huang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.K. Shenoy, M. Chang, E.W. Scott, E.H. Huang

Writing, review, and/or revision of the manuscript: A.K. Shenoy, R.C. Fisher, L. Pi, L.-J. Chang, M. Chang, E.W. Scott, E.H. Huang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.-J. Chang, E.H. Huang

Study supervision: E.W. Scott

Technical support related to molecular biology: L. Pi

Histologic analysis of the study specimens: H.D. Appelman

The authors thank Neal Benson and the Flow Cytometry Core (courtesy of UFSCC) at the University of Florida, Marda Jorgenson (immunohistochemistry), and Douglas Smith (imaging) as part of the Cell and Tissue Analysis Core, at the University of Florida. The authors would also like to thank Dr. Mark Krebs for his help with confocal imaging and Yilun Sun for helping with analyses of tumor growth rates.

E.H. Huang was supported by the Broad Medical Research Fund IBD-0239 and NIH R01 CA142808. E.W. Scott was supported by NIH R01 CA142808 and HL70738.

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.