Volume 141, Issue 1, Pages 269-278 (July 2011) Differentiated Human Colorectal Cancer Cells Protect Tumor-Initiating Cells From Irinotecan  Benjamin L. Emmink, Winan J. Van Houdt, Robert G. Vries, Frederik J.H. Hoogwater, Klaas M. Govaert, Andre Verheem, Maarten W. Nijkamp, Ernst J.A. Steller, Connie R. Jimenez, Hans Clevers, Inne H.M. Borel Rinkes, Onno Kranenburg  Gastroenterology  Volume 141, Issue 1, Pages 269-278 (July 2011) DOI: 10.1053/j.gastro.2011.03.052 Copyright © 2011 AGA Institute Terms and Conditions

Figure 1 Differentiation of colonosphere-forming cells from human colorectal tumors. (A) Light microscopic images of a representative colonosphere and a population of colonosphere-derived adherent tumor cells. Bar, 20 μm. The lower panel shows periodic acid–Schiff–stained cultures. (B) Western blot analysis of the expression of cytokeratin 20 (CK20) and actin in isogenic colonosphere and adherent tumor cell cultures. (C) Adherent tumor cells were grown on glass coverslips and were analyzed by immunofluorescence to detect the differentiation markers CK20 (in green) and phospho-ezrin (red). Bar, 20 μm. Gastroenterology 2011 141, 269-278DOI: (10.1053/j.gastro.2011.03.052) Copyright © 2011 AGA Institute Terms and Conditions

Figure 2 ALDH1 marks tumor-initiating cells in colonospheres. (A) Lysates of paired colonosphere and differentiated cultures derived from primary colorectal tumors and liver metastases were analyzed for the expression of ALDH1 by Western blotting. (B) Cultures enriched for tumor-initiating cells were analyzed for ALDH activity by making use of the fluorescent ALDH substrate Aldefluor. Clone forming potential of cells with high and low ALDH activity was determined after 3 weeks. Representative examples of clones formed in ALDH-high and ALDH-low cultures are shown. (C) ALDHhigh and ALDHlow cell populations within the spheroids were separated by making use of the fluorescent ALDH substrate Aldefluor. ALDHhigh and ALDHlow cells (1000) were injected into the flanks of nude mice (n = 8). Tumors generated by ALDHhigh cells were harvested and processed to generate single-cell suspensions. Again ALDHhigh and ALDHlow cell populations were separated by FACS and these were injected into the flanks of nude mice (1000 cells). Tumor growth was analyzed by caliper measurements. The mean tumor volume of all tumors is shown. (D) Single-cell populations were generated from the indicated colonosphere lines and from adherent tumor cell cultures and these were suspended in Matrigel (BD Biosciences, Franklin Lakes, NJ) at 1000 cells/mL. Equal numbers of Matrigel-embedded cells (n = 100 in 100 μL) were allowed to set in 48-well plates and clone formation was analyzed after 3 weeks of culture. (E) Single-cell populations were generated from L145 colonospheres and from adherent tumor cell cultures and these were suspended in Matrigel. Mice were injected with 10,000 (n = 3), 1000 (n = 3), or 200 (n = 3) tumor cells as indicated and tumor growth was followed up over time by caliper measurements. *Statistical significance (unpaired, 2-tailed t test: P < .05). Gastroenterology 2011 141, 269-278DOI: (10.1053/j.gastro.2011.03.052) Copyright © 2011 AGA Institute Terms and Conditions

Figure 3 Colonosphere resistance to irinotecan. (A) Colonospheres (open bars) and isogenic differentiated tumor cells (closed bars) were seeded in 48-well plates in triplicate and were exposed to the indicated concentrations of irinotecan for 5 days. Mitochondrial activity then was tested by CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay (MTS). Absorbance values are expressed as percentages of vehicle-treated control wells. (B) L145 colonospheres and differentiated tumor cells were treated either with vehicle (open bars) or with 125 μg/mL irinotecan (closed bars) for 72 hours. The cells then were pulse-labeled with bromodeoxyuridine for 30 minutes and the percentage of bromodeoxyuridine-positive cells was determined by FACS analysis. (C) L145 colonospheres and differentiated tumor cells were treated with irinotecan for the indicated periods of time. Cell lysates were prepared and were analyzed for the presence of p53 and activated (cleaved) caspase-3 (a-caspase-3). *Statistical significance (unpaired, 2-tailed t test: P < .05). Gastroenterology 2011 141, 269-278DOI: (10.1053/j.gastro.2011.03.052) Copyright © 2011 AGA Institute Terms and Conditions

Figure 4 Irinotecan resistance is mediated by a subpopulation of tumor cells expressing ABCB1. (A) L145 and L167 colonospheres were cultured in the presence of irinotecan (50 μg/mL), the ABCB1 inhibitor PSC833 (2 μmol/L), the ABCG2 inhibitor Ko143 (200 nmol/L), or the ABCC1 inhibitor MK571 (30 μmol/L), either alone or in combination as indicated. Mitochondrial activity then was assessed by CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay (MTS) for 3 consecutive days. Absorbance values are expressed as percentages of vehicle-treated control wells. *Statistical significance (unpaired, 2-tailed t test: P < .05). (B) L145 colonospheres were cultured in the presence of irinotecan either in the absence or presence of the ABCB1 inhibitor PSC833. Cell lysates were prepared at the indicated periods of time and these were analyzed for the presence of activated (cleaved) caspase-3 (a-caspase-3). (C) L145 colonospheres were treated with vehicle, PSC833 alone (2 μmol/L), irinotecan alone (75 μg/mL), or with the combination. The resulting cell populations were injected into the flanks of nude mice to assess tumor-initiating potential. Tumor growth was measured by caliper measurements. (D) Cell lysates of colonosphere cultures and differentiated tumor cells of the indicated tumors were lysed and analyzed for the presence of ABCB1 by Western blotting. L145 colonospheres were treated with vehicle or with irinotecan for 72 hours and lysates were prepared for analysis of ABCB1 expression by Western blotting. (E) FACS analysis of ABCB1 and Aldefluor in 3 different colonosphere lines and in 2 spheroid-initiated tumors. (F) FACS analysis of ABCB1 and Aldefluor in irinotecan-treated L145 colonosphere cultures (75 μg/mL; 24 h). Gastroenterology 2011 141, 269-278DOI: (10.1053/j.gastro.2011.03.052) Copyright © 2011 AGA Institute Terms and Conditions

Figure 5 ABCB1 marks a population of proliferating tubule-forming cells displaying enterocyte-like differentiation. (A) L145 colonospheres were fixed in formalin and processed for immunofluorescence using validated antibodies directed at ABCB1, CD44, and EpCAM. Irinotecan-treated cultures (75 μg/mL; 72 h) also were analyzed for ABCB1 localization (lower panel). (B) Large colonospheres with signs of intrasphere tubule formation were fixed and stained for ABCB1 and phospho-ezrin as in panel A. (C) Paraffin-embedded and (D) frozen tissue sections of the same region in human colorectal tumor tissue (L150) were stained for expression of ABCB1 and ALDH1. Representative images are shown. ABCB1 is found mainly on the apical membrane of tubule-forming cells. ALDH1 is expressed by non–tubule-forming cells in poorly differentiated tumor areas that lie between the tubule-forming cells. Bar, 50 μm. (E) Colonospheres were fixed in formalin and processed for immunofluorescence using validated antibodies directed at cytokeratin 20 (CK20) and phospho-ezrin (p-ezrin). A representative L145 sphere is shown. (F) Expression of the proliferation marker Ki67 in the CK20/ABCB1/p-ezrin–positive outer cell layer relative to inner-sphere cells was assessed by immunofluorescence as in panel A. Quantification was performed by counting the number of outer and inner cells displaying Ki67 expression. A representative example is shown in the left panel, quantification of Ki67 in the 2 cell compartments was based on analysis of 16 spheres from L145, L146, and L150 each. *Statistical significance (P = .0002). All bars, 20 μm. Gastroenterology 2011 141, 269-278DOI: (10.1053/j.gastro.2011.03.052) Copyright © 2011 AGA Institute Terms and Conditions

Figure 6 Separation of irinotecan resistance and tumor-initiating potential on the basis of ABCB1 expression. (A) L145 colonospheres and sphere-derived single-cell populations were treated either with vehicle or with irinotecan in the presence or absence of PSC-833 for 5 days at the indicated concentrations. Mitochondrial activity then was assessed by CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay (MTS) for 3 consecutive days. Absorbance values are expressed as percentages of vehicle-treated control wells. (B) L145 colonospheres were either left as spheres or were processed to generate single-cell populations. Colonospheres and single-cell populations then were treated with irinotecan. All cultures were washed and the spheres were processed to generate single-cell populations. All single-cell populations were resuspended in Matrigel at 1000 live cells/mL. Equal numbers of Matrigel-embedded cells (n = 100 in 100 μL) were allowed to set in 48-well plates and clone formation was analyzed after 2 weeks (left panel), 4 weeks (middle panel), or 8 weeks (right panel) of cell culture. (C) Single-cell cultures were derived from L145 colonospheres and these cells were stained for expression of ABCB1 by using 2 independent antibodies (MM4.17 and F4; see also Supplementary Figure 7). The ABCB1-positive and ABCB1-negative subpopulations were separated by FACS sorting and analyzed for clone-forming and tumor-initiating capacity as described earlier. A part of the cells was re-sorted to assess their proper separation. (D) The clones generated by the ABCB1-negative cells were recovered from the Matrigel and single-cell populations were prepared and analyzed for ABCB1 expression by FACS analysis. (E) Tumors were excised and processed to generate single-cell suspensions. ABCB1-positive and ABCB1-negative epithelial (EpCAM+) cells were separated by FACS sorting and analyzed for expression of ALDH1 and cytokeratin 20 (CK20) by Western blotting. (F) Intestinal organoid cultures from Lgr5-EGFP-ires-CreERT2 mice 27 (n = 3, each containing 25–50 organoids) were treated for 24 hours with the indicated concentrations of irinotecan. The number of green fluorescent protein (GFP)-positive organoids was assessed by fluorescence microscopy using 4',6-diamidino-2-phenylindole (DAPI) as counterstain for healthy cells. *Statistical significance (unpaired, 2-tailed t test: P < .05). Gastroenterology 2011 141, 269-278DOI: (10.1053/j.gastro.2011.03.052) Copyright © 2011 AGA Institute Terms and Conditions

Figure 7 PSC-833 enhances the antitumor efficacy of irinotecan. (A) Subcutaneous tumor xenografts were generated by injection of colonospheres (n = 8) or by transplantation of unsegregated biopsies from a freshly resected human colorectal liver metastasis (n = 8). Tumor growth was observed in all injected animals. Treatment (vehicle, irinotecan, PSC-833, or the combination) was started (day 0) after tumors reached a volume of 250–350 mm3. Tumor volume was analyzed by caliper measurements. Tumors were harvested and pooled samples of each group were analyzed (B) for active caspase-3 by Western blotting, or (C) for ALDH activity by FACS analysis, using Aldefluor. Gastroenterology 2011 141, 269-278DOI: (10.1053/j.gastro.2011.03.052) Copyright © 2011 AGA Institute Terms and Conditions