Volume 145, Issue 3, Pages 636-646.e5 (September 2013) Isolation and Phenotypic Characterization of Colorectal Cancer Stem Cells With Organ- Specific Metastatic Potential  Wenchao Gao, Lu Chen, Zhenyu Ma, Zunguo Du, Zhonghua Zhao, Zhiqian Hu, Qingquan Li  Gastroenterology  Volume 145, Issue 3, Pages 636-646.e5 (September 2013) DOI: 10.1053/j.gastro.2013.05.049 Copyright © 2013 AGA Institute Terms and Conditions

Figure 1 CD110 and CDCP1 are markers of colon CSCs mediating organ-specific metastasis. (A) Flow chart of in vivo selection of organ-specific metastatic subpopulations. (B) CRC108 and its metastatic variants were incubated in the presence of growth factors for 14 days. The percentage of oncosphere-forming cells was determined by sphere-forming assays. (C) After orthotopic implantation (n = 5 mice/group), organ-specific metastatic potentials of the indicated oncospheres were evaluated by bioluminescent imaging and gross examination (arrows indicate metastatic foci). (D) Comparison of gene expression profiles between oncospheres from CRC108 and its metastatic variants identified signatures that classify the in vivo selected metastatic oncospheres into a liver metastatic group (CRC102-LM, CRC105-LM) and a pulmonary metastatic group (CRC102-PM, CRC105-PM). (E) CD110 and CDCP1 messenger RNA levels in oncospheres from CRC108 and its metastatic variants were determined by quantitative PCR. (F) The indicated sphere cells were orthotopically implanted as described in C, after which CD110 and CDCP1 expressions in primary xenograft tumors were examined by immunohistochemistry. All experiments were performed in triplicate. **P < .05, ***P < .01 versus the corresponding controls. Bars = 100 μm; insets = 50 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Figure 2 Role of CD110 and CDCP1 in predicting CRLM and CRPM. (A) Representative images of double staining for CD110 and CDCP1 on tissues from liver or lung metastases and their matched primary CRC tumors. (B) Representative images of double staining for CD110 and CDCP1 in primary CRC tumors with synchronous liver and lung metastases. (C) Liver and lung metastasis-free survival rate in groups with the indicated phenotypes were analyzed by Kaplan–Meier survival analysis. Bars = 100 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Figure 3 The CD110+ and CDCP1+ subpopulations are cancer cells with stem cell properties. (A) Representation of the overlap between the CD110+ or CDCP1+ subpopulation and the CD133+ or CD44+ subpopulation of LOVO cells. (B) The CD133+CD110+, CD133+CDCP1+, and CD133+CD110-CDCP1 subpopulations were subcutaneously injected (n = 3 mice/group), after which primary tumor development at the site of injection was observed, and tumor size was assessed every 5 days after implantation. (C) H&E staining of the parental tumor and primary xenografts developed by various subpopulations. (D) Staining for markers associated with different cell lineages in the parental tumor and primary xenografts developed by the indicated subpopulations. (E) Representative images of double staining for CD110 and CDCP1 in the parental tumor and primary xenografts developed by the indicated subpopulations. (F) The indicated subpopulations isolated from CRC105 were incubated with 100 μmol/L 5-FU for 72 hours, followed by 24 hours of recovery after withdrawal of treatment. Cells were then stained with bromodeoxyuridine as a marker for DNA synthesis and with 7-aminoactinomycin D as a marker for apoptosis. The CRC102 cells were exposed to 100 μmol/L 5-FU for 14 days, after which the percentage of the indicated subpopulations was assessed by flow cytometry. All experiments were performed in triplicate. ∗∗P < .05, ∗∗∗P < .01 vs the corresponding controls. Bars = 50 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Figure 4 CD110+ and CDCP1+ cells possess organ-specific metastatic potential. (A) The number of the CD110+, CDCP1+, and CD133+ cells in a panel of CRC cell lines was analyzed by flow cytometry. (B) Orthotopic implantation of the CD133+CD110+ and CD133+CDCP1+ cells, as determined by bioluminescent imaging, from CRC102 or CRC105 led to metastases in the liver and lung, respectively (n = 5 mice/group). Arrows indicate metastatic foci. (C) H&E staining of liver and lung metastases shows moderately differentiated adenocarcinomas that are histologically similar to the matched primary tumor. (D) Schematic representation of the retroviral vectors cis-CD110-HSV1-tk/GFP and CDCP1-tdRFP-cmvFLuc. (E) CRC102 cells bearing the cis-CD110-HSV1-tk/GFP and CDCP1-tdRFP-cmvFLuc were orthotopically implanted (n = 5 mice/group). Liver and lung metastases were analyzed by fluorescence microscopy and two-color fluorescence-activated cell sorting. All experiments were performed in triplicate. ∗∗∗P < .01 versus the corresponding controls. Bars: (C) 100 μm, insets = 50 μm; (E) 50 μm, insets = 10 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Figure 5 TPO acts as a chemoattractant of the CD110+ CSCs and increases its self-renewal via binding to CD110. (A) Representative bioluminescence images of liver sections (upper panel) and bioluminescence quantification (lower panel) of NOG mice after injection of the CD110+ cells with or without CD110 knockdown (n = 5 mice/group). (B) CD110+ cells derived from CRC102 and CRC105 were orthotopically transplanted into NOG mice (n = 5 mice/group). At 2 weeks after transplantation, mice were injected intraperitoneally with anti-TPO antibody or control immunoglobulin G (10 mg/kg) twice weekly for 3 weeks, after which the number of invaded tumor cells in each group on days 21, 35 and 49 was evaluated. (C) Scheme of the in vitro invasion assay. (D) Semiquantitative PCR analysis for invasion of the indicated tumor cells. Liver specimens from wild-type and TPO−/− mice were used. (E) CD110+ cells from CRC102, CRC105, and CRC108 were incubated in the absence of growth factors with 50 ng/mL TPO and/or AMM2 for 7 days. The percentage of oncosphere-forming cells and the total number of cells were determined. (F) The CD110+ cells isolated from CRC105 were transfected with or without lenti-TPO and subcutaneously implanted (n = 3 mice/group). Tumor size was assessed 40 days after implantation. All experiments were performed in triplicate. **P < .05, ***P < .01 versus the corresponding controls. Bars = 100 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Figure 6 CDCP1 facilitates the adhesion of CDCP1+ CSCs to lung endothelial cells. (A) Representative bioluminescence images of lung sections (upper panel) and bioluminescence quantification (lower panel) of NOG mice after injection of CDCP1+ cells with or without CDCD1 knockdown (n = 10 mice/group). (B) Kaplan–Meier survival analysis of mice injected with CDCP1+ cells with or without CDCP1 knockdown. (C) CDCP1+ cells isolated from CRC108, LOVO, or SW480 with or without CDCP1 knockdown were seeded on top of a monolayer of endothelial cells from lung (HMVEC-L and LEISVO), umbilical vein (HUVEC), or control lung fibroblast cells (MRC-5) and liver sinusoidal endothelial cells (LSECs), and the attached cells were quantified 3 hours later. All experiments were performed in triplicate. **P < .05 versus the corresponding controls. Bars = 100 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Supplementary Figure 1 The CD110+, CDCP1+, and CD133+CD110−CDCP1− subpopulations exhibited no difference in tumor growth rate. The CD110+, CDCP1+, and CD133+CD110−CDCP1− cells isolated from CRC102 and CRC105 were orthotopically injected in NOG mice (n = 5 mice/group). Tumor growth was monitored by bioluminescence quantification every 10 days after implantation. Experiments were performed in triplicate. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Supplementary Figure 2 Validation of the responsiveness of the cis-CD110-HSV1-tk/GFP and CDCP1-tdRFP-cmvFLuc vector. After CRC102 cells were transduced with the cis-CD110-HSV1-tk/GFP and CDCP1-tdRFP-cmvFLuc, the sorted CD133+CD110+ and CD133+CDCP1+ cells were analyzed by fluorescence microscopy. The experiments were performed 3 times independently, and representative data are shown. Bar = 50 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Supplementary Figure 3 CD110 exerts no impact on primary xenograft tumor growth or tumor cell intravasation and extravasation. (A) Oncospheres derived from CRC102-PM cells were transfected with pCMV-GFP or pCMV-CD110-GFP. After orthotopic implantation (n = 5 mice/group), organ-specific metastatic potentials of the indicated oncospheres were evaluated by bioluminescent imaging. The CD110+ cells isolated from CRC102 were transfected with or without shCD110 and orthotopically injected in NOG mice (n = 5 mice/group). (B) Tumor growth was monitored by bioluminescence quantification every 5 days after implantation. (C) Five milliliters of blood from tumor-bearing mice was collected and red blood cells were lysed. RNA from the remaining cells was extracted, after which the number of circulating tumor cells was assessed using quantitative PCR as a function of human GAPDH expression relative to murine β2-microglobulin. (D) At 36 hours after the CD110+ cells from CRC102 were transfected with or without shCD110, the Transwell migration/invasion assays were performed. (E) Representative immunofluorescence staining for CD31 (red) and GFP (green) in 5-μm-thick liver sections collected 48 hours after intravenous inoculation of NOG mice with 105 CRC105−CD110+ cells carrying GFP with/without shCD110 (n = 5 mice/group). (F) Quantification of in vitro transendothelial migration of parental CRC105 cells, the CRC105−CD110+ cells transfected with or without shCD110. All experiments were performed in triplicate. Bars = 50 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Supplementary Figure 4 Treatment with anti-TPO antibody exerts no impact on primary tumor growth. The CD110+ cells derived from CRC102 and CRC105 were orthotopically transplanted into NOG mice (n = 5 mice/group). Two weeks after transplant, mice were injected intraperitoneally with anti-TPO antibody or control immunoglobulin G (10 mg/kg) twice weekly for 3 weeks. Seven weeks after implantation, tumor growth was compared by bioluminescence quantification. Experiments were performed in triplicate. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Supplementary Figure 5 The CD110+ CSCs are induced to proliferate during liver colonization. Representative immunofluorescence staining for Ki-67 (red) and GFP (green) in the livers of mice that were intravenously inoculated with the CD110+ cells derived from CRC108 (n = 5 mice/group). Quantification of multiple liver sections at the indicated times showed significant differences in the rate of tumor cell proliferation. Experiments were performed in triplicate. ***P < .01 versus the corresponding controls. Bars = 50 μm. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Supplementary Figure 6 Treatment with TPO increases the CD110+ CSC population. CRC105 and CRC108 cells were treated with 50 ng/mL TPO for 48 hours, after which the percentage of the indicated subpopulations was assessed by flow cytometry. Experiments were performed in triplicate. **P < .05 versus the corresponding controls. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Supplementary Figure 7 CDCP1 expression has no effect on lung colonization. Oncospheres derived from CRC102-LM cells were transfected with pcDNA3.1-GFP or pcDNA3.1-CDCP1-GFP. After orthotopic implantation (n = 5 mice/group), organ-specific metastatic potentials of the indicated oncospheres were evaluated by bioluminescent imaging. Experiments were performed in triplicate. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions

Supplementary Figure 8 CDCP1 knockdown has no effect on either tumor growth or tumor cell intravasation. The CDCP1+ cells isolated from CRC108 were transfected with or without shCDCP1 and orthotopically injected in NOG mice (n = 5 mice/group). (A) Tumor growth was monitored by bioluminescence quantification every 10 days after implantation. (B) Five milliliters of blood from tumor-bearing mice was collected and red blood cells were removed using red blood cell lysing buffer. RNA from the remaining cells was extracted, after which the number of circulating tumor cells was assessed using quantitative PCR as a function of human GAPDH expression relative to murine β2-microglobulin. All experiments were performed in triplicate. Gastroenterology 2013 145, 636-646.e5DOI: (10.1053/j.gastro.2013.05.049) Copyright © 2013 AGA Institute Terms and Conditions