Cancer stem cells from colorectal cancer-derived cell lines

6
Cancer stem cells from colorectal cancer-derived cell lines Trevor M. Yeung a , Shaan C. Gandhi a , Jennifer L. Wilding a , Ruth Muschel b , and Walter F. Bodmer a,1 a Cancer and Immunogenetics Laboratory, Weatherall Institute of Molecular Medicine, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DS, United Kingdom; and b Radiation Oncology and Biology, University of Oxford, Oxford OX3 7LJ, United Kingdom Contributed by Walter Fred Bodmer, January 7, 2010 (sent for review November 22, 2009) Cancer stem cells (CSCs) are the subpopulation of cells within a tumor that can self-renew, differentiate into multiple lineages, and drive tumor growth. Here we describe a two-pronged approach for the identication and characterization of CSCs from colorectal cancer cell lines, using a Matrigel-based differentiation assay, and cell surface markers CD44 and CD24. About 20 to 30% of cells from the SW1222 cell line form megacolonies in Matrigel that have complex 3D structures resembling colonic crypts. The megacoloniescapacity to self-renew in vitro is direct evidence that they contain the CSCs. Furthermore, just 200 cells from SW1222 megacolonies initiate tumors in NOD/SCID mice. We also showed that CD44 + CD24 + cells enriched for colorectal CSCs in the HT29 and SW1222 cell lines, which can self-renew and reform all four CD44/CD24 subpopula- tions, are the most clonogenic in vitro and can initiate tumors in vivo. A single SW1222 CD44 + CD24 + CSC, when grown in Matrigel, can form large megacolonies that differentiate into enterocyte, enteroendocrine, and goblet cell lineages. The HCT116 line does not differentiate or express CDX1, nor does it contain subpopula- tions of cells with greater tumor-forming capacity, suggesting that HCT116 contains mainly CSCs. However, forced expression of CDX1 in HCT116 leads to reduced clonogenicity and production of differ- entiating crypt-containing colonies, which can explain the selection for reduced CDX1 expression in many colorectal cancers. In sum- mary, colorectal cancer cell lines contain subpopulations of CSCs, characterized by their cell surface markers and colony morphology, which can self-renew and differentiate into multiple lineages. clonogenicity | matrigel | CD44 | goblet cells T he idea that cancers are driven by cancer stem cells (CSCs) that may differentiate into a variety of cell types corre- sponding to the tissue of origin of the cancer, while maintaining the ability to self-renew, is now widely accepted for essentially all cancers (1). Most CSC assays have so far depended on the use of a variety of different cell surface markers, including CD133, CD44, and CD24, to enrich for CSCs from primary tumor tissue using FACS technology, and then testing propagation in immu- nodecient mice (24). Such assays are cumbersome and expensive, and do not readily enable either the characterization of the CSCs or their functional study. A reliable in vitro assay for CSCs derived from cell lines would have many advantages; for example, enabling relatively high throughput testing of drugs and antibody effects, as well as gene knockdown and re-expression studies on specic CSC functions, rather than on all of the cells in a cancer. An in vitro assay would also enable the unequivocal identication of the specic genes that control the stem-cell state, which has so far proved relatively elusive. Cell lines represent a valuable resource that can be reused repeatedly over time, unlike primary human tissue. They can readily be characterized with respect to mutational content (5) and mRNA expression. They are not contaminated with stromal tissue, which can affect the interpretation of data obtained from primary human tissue (6). There is strong evidence that cell lines represent the tumors from which they were originally isolated (7) and can develop structures similar to those found in the original tissue (8). The same percentages of mutations in genes seen in primary colonic cancers are also seen across our panel of cell lines. Furthermore, data from our laboratory demonstrate that duplicate cell lines obtained from different laboratories continue to have the same mutations and their microarray-expression proles cluster together. Our aim in this article is to show how the CSCs in colorectal cancer-derived cell lines can be clearly identied. We concen- trate on three colorectal cell lines with contrasting abilities to differentiate, which correlate inversely with the proportion of CSCs found within them. We demonstrate that colorectal cell lines contain CSC populations that can be enriched by the use of an in vitro Matrigel-based differentiation assay together with selection for expression of the CD44 and CD24 cell surface markers. We show that megacoloniesfrom one of the lines, SW1222, containing complex internal crypt-like structures, must be derived from the CSC population, and that cells from these colonies are able to self-renew and initiate tumors in immuno- decient NOD/SCID mice. Furthermore, the CD44 + CD24 + cells are the most clonogenic in vitro, form the greatest numbers of megacolonies, and can initiate tumors in NOD/SCID mice. Results Single Cells from the SW1222 Cell Line Form Colonies with Different Morphologies when Grown in Matrigel. Earlier work from our laboratory showed that the colorectal carcinoma cell line SW1222 could form lumen like glandular structures when grown in three dimensions, either in collagen I gels or in Matrigel (9, 10). The two main types of colonies formed by single SW1222 cells, obtained by ltration and plated by limiting dilution in Matrigel, are shown in Fig. 1A. Approximately 20% of the cells give rise to large, complex structured colonies with clear internal crypt-like formations and these we call megacolonies.Another 15% of cells form smaller unstructured colonies less than 400 micrometers in diameter. Megacolonies increase signicantly in size between 7 and 14 days, and express increasing amounts of two key markers of differentiation, CDX1 and CK20 (11), at day 14 as compared to day 7 (Fig. 1 B and C). The small colonies, however, have limited growth potential. When live megacolony colonies were examined using multiphoton microscopy, they contained clear central lumens, with adjacent secretory vesicles and outlines of goblet cells (Fig. S1, Movie S1 and Movie S2). Staining using uorescein diacetate/propidium iodide to detect viable and dead cells within the megacolonies showed that most of the cells within these structures were still viable, even after 4 weeks in culture (Fig. S2). These data suggest that the megacolonies arise from CSCs in the cell line. To test this hypothesis, megacolonies and small colonies were subcloned in Matrigel. Author contributions: T.M.Y., S.C.G., J.L.W., R.M., and W.F.B. designed research; T.M.Y. and S.C.G. performed research; T.M.Y., S.C.G., J.L.W., and W.F.B. analyzed data; and T.M. Y. and W.F.B. wrote the paper. The authors declare no conict of interest. 1 To whom correspondence should be addressed. E-mail: [email protected]. This article contains supporting information online at www.pnas.org/cgi/content/full/ 0915135107/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.0915135107 PNAS Early Edition | 1 of 6 MEDICAL SCIENCES

Transcript of Cancer stem cells from colorectal cancer-derived cell lines

Page 1: Cancer stem cells from colorectal cancer-derived cell lines

Cancer stem cells from colorectal cancer-derivedcell linesTrevor M. Yeunga, Shaan C. Gandhia, Jennifer L. Wildinga, Ruth Muschelb, and Walter F. Bodmera,1

aCancer and Immunogenetics Laboratory, Weatherall Institute of Molecular Medicine, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DS, UnitedKingdom; and bRadiation Oncology and Biology, University of Oxford, Oxford OX3 7LJ, United Kingdom

Contributed by Walter Fred Bodmer, January 7, 2010 (sent for review November 22, 2009)

Cancer stem cells (CSCs) are the subpopulation of cells within atumor that can self-renew, differentiate into multiple lineages, anddrive tumor growth. Here we describe a two-pronged approach forthe identification and characterization of CSCs from colorectalcancer cell lines, using a Matrigel-based differentiation assay, andcell surface markers CD44 and CD24. About 20 to 30% of cells fromthe SW1222 cell line form megacolonies in Matrigel that havecomplex 3D structures resembling colonic crypts. Themegacolonies’capacity to self-renew in vitro is direct evidence that they containthe CSCs. Furthermore, just 200 cells from SW1222 megacoloniesinitiate tumors inNOD/SCIDmice.Wealso showed thatCD44+CD24+

cells enriched for colorectal CSCs in the HT29 and SW1222 cell lines,which can self-renew and reform all four CD44/CD24 subpopula-tions, are the most clonogenic in vitro and can initiate tumors invivo. A single SW1222 CD44+CD24+ CSC, when grown in Matrigel,can form large megacolonies that differentiate into enterocyte,enteroendocrine, and goblet cell lineages. The HCT116 line doesnot differentiate or express CDX1, nor does it contain subpopula-tions of cells with greater tumor-forming capacity, suggesting thatHCT116 contains mainly CSCs. However, forced expression of CDX1in HCT116 leads to reduced clonogenicity and production of differ-entiating crypt-containing colonies, which can explain the selectionfor reduced CDX1 expression in many colorectal cancers. In sum-mary, colorectal cancer cell lines contain subpopulations of CSCs,characterized by their cell surface markers and colonymorphology,which can self-renew and differentiate into multiple lineages.

clonogenicity | matrigel | CD44 | goblet cells

The idea that cancers are driven by cancer stem cells (CSCs)that may differentiate into a variety of cell types corre-

sponding to the tissue of origin of the cancer, while maintainingthe ability to self-renew, is now widely accepted for essentially allcancers (1). Most CSC assays have so far depended on the use ofa variety of different cell surface markers, including CD133,CD44, and CD24, to enrich for CSCs from primary tumor tissueusing FACS technology, and then testing propagation in immu-nodeficient mice (2–4). Such assays are cumbersome andexpensive, and do not readily enable either the characterizationof the CSCs or their functional study. A reliable in vitro assay forCSCs derived from cell lines would have many advantages; forexample, enabling relatively high throughput testing of drugs andantibody effects, as well as gene knockdown and re-expressionstudies on specific CSC functions, rather than on all of the cellsin a cancer. An in vitro assay would also enable the unequivocalidentification of the specific genes that control the stem-cellstate, which has so far proved relatively elusive.Cell lines represent a valuable resource that can be reused

repeatedly over time, unlike primary human tissue. They canreadily be characterized with respect to mutational content (5)and mRNA expression. They are not contaminated with stromaltissue, which can affect the interpretation of data obtained fromprimary human tissue (6). There is strong evidence that cell linesrepresent the tumors from which they were originally isolated (7)and can develop structures similar to those found in the originaltissue (8). The same percentages of mutations in genes seen in

primary colonic cancers are also seen across our panel of celllines. Furthermore, data from our laboratory demonstrate thatduplicate cell lines obtained from different laboratories continueto have the same mutations and their microarray-expressionprofiles cluster together.Our aim in this article is to show how the CSCs in colorectal

cancer-derived cell lines can be clearly identified. We concen-trate on three colorectal cell lines with contrasting abilities todifferentiate, which correlate inversely with the proportion ofCSCs found within them. We demonstrate that colorectal celllines contain CSC populations that can be enriched by the use ofan in vitro Matrigel-based differentiation assay together withselection for expression of the CD44 and CD24 cell surfacemarkers. We show that “megacolonies” from one of the lines,SW1222, containing complex internal crypt-like structures, mustbe derived from the CSC population, and that cells from thesecolonies are able to self-renew and initiate tumors in immuno-deficient NOD/SCID mice. Furthermore, the CD44+CD24+

cells are the most clonogenic in vitro, form the greatest numbersof megacolonies, and can initiate tumors in NOD/SCID mice.

ResultsSingle Cells from the SW1222 Cell Line Form Colonies with DifferentMorphologies when Grown in Matrigel. Earlier work from ourlaboratory showed that the colorectal carcinoma cell lineSW1222 could form lumen like glandular structures when grownin three dimensions, either in collagen I gels or in Matrigel (9,10). The two main types of colonies formed by single SW1222cells, obtained by filtration and plated by limiting dilution inMatrigel, are shown in Fig. 1A. Approximately 20% of the cellsgive rise to large, complex structured colonies with clear internalcrypt-like formations and these we call “megacolonies.” Another≈15% of cells form smaller unstructured colonies less than 400micrometers in diameter. Megacolonies increase significantly insize between 7 and 14 days, and express increasing amounts oftwo key markers of differentiation, CDX1 and CK20 (11), at day14 as compared to day 7 (Fig. 1 B and C). The small colonies,however, have limited growth potential. When live megacolonycolonies were examined using multiphoton microscopy, theycontained clear central lumens, with adjacent secretory vesiclesand outlines of goblet cells (Fig. S1, Movie S1 and Movie S2).Staining using fluorescein diacetate/propidium iodide to detect

viable and dead cells within the megacolonies showed that most ofthe cells within these structures were still viable, even after 4 weeksin culture (Fig. S2). These data suggest that themegacolonies arisefrom CSCs in the cell line. To test this hypothesis, megacoloniesand small colonies were subcloned in Matrigel.

Author contributions: T.M.Y., S.C.G., J.L.W., R.M., and W.F.B. designed research; T.M.Y.and S.C.G. performed research; T.M.Y., S.C.G., J.L.W., and W.F.B. analyzed data; and T.M.Y. and W.F.B. wrote the paper.

The authors declare no conflict of interest.1Towhomcorrespondence shouldbeaddressed. E-mail:[email protected].

This article contains supporting information online at www.pnas.org/cgi/content/full/0915135107/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.0915135107 PNAS Early Edition | 1 of 6

MED

ICALSC

IENCE

S

Page 2: Cancer stem cells from colorectal cancer-derived cell lines

SW1222 Megacolonies Give Rise to Megacolonies and Small Colonies,Although Small Colonies only Give Rise to More Small Colonies.SW1222 single cells obtained by filtration were plated in limitingdilutions in 96-well Matrigel-containing plates. Megacolonies andsmall colonies growing in single wells were pooled separately,disaggregated, and filtered into a single-cell suspension and thenreplated intoMatrigel. Fig. 2A shows that cells frommegacolonieswere able to reform megacolonies (27.8% on average), as well assmall colonies (59.4%), although cells from small colonies couldonly reform themselves (36.7%) and could not formmegacolonies.Cells from both types of colonies were next injected into NOD/SCID mice. As few as 200 cells from megacolonies were able to

initiate tumors; however, even 1,000 cells from small colonies wereunable to do so (Fig. 2B). These results strongly support the viewthat the megacolonies are formed from CSCs in the SW1222 cellline. Xenografts derived from megacolonies are able to recapit-ulate a well-differentiated adenocarcinoma (Fig. S3).

SW1222 FACS-Sorted CD44+CD24+ Cells Are the Most ClonogenicSubpopulation, Form the Greatest Number of Megacolonies, WhichSelf-Renew and Differentiate, and Are the Most Efficient in ProducingTumors in NOD/SCID Mice. Using the markers CD44 and CD24 toenrich for putative colorectal CSCs from the SW1222 cell line byFACS sorting, we showed that the CD44+CD24+ cells (Fig. 3A,upper 0.5–1%,) were the most clonogenic and gave rise to thehighest proportion of megacolonies (Fig. 3B). Immuno-fluorescence analysis of megacolonies derived from singleCD44+CD24+ sorted cells expressed markers of all three types ofdifferentiated colonic epithelial cells, namely columnar, goblet,and enteroendocrine cells (Fig. 3C). The chromogranin-A stainingwas particularly strong around the lumen-like structures, sug-gesting that the product was secreted. The overall pattern ofstaining is as expected on the assumption that thesemegacolonies,containing multiple lineages, originally arise from singleCD44+CD24+ CSCs present in the SW1222 cell line.The sorted CD44+CD24+ cells were also the most efficient at

producing tumors in NOD/SCID mice, as would be expected ifthis subpopulation enriched the most for CSCs from the SW1222cell line (Fig. 3D).Cells from megacolonies analyzed by FACS for CD44 and

CD24 produced all four combinations of CD44±CD24± expres-sion. Reanalysis of megacolonies derived from singleCD44+CD24+ cells again produced all four combinations ofCD44±CD24± expression, and isolation of CD44+CD24+ cellsderived from this second cell sort were capable of reformingmegacolonies for a further generation. The ability of mega-colonies to self-renew and differentiate is again consistent withtheir containing the CSC subpopulation (Fig. S4).FACS analysis of xenografts derived from an initial injection of

CD44+CD24+ cells showed that these tumors were able toexpress all four CD44±CD24± subpopulations (Fig. S5). Thehistology of the SW1222-derived xenograft tumors was similar tothat of a well-differentiated primary human lumen-formingtumor, whether it was from a SW1222 CD44+CD24+ sorted cellpopulation or from unsorted SW1222 cells (Fig. 3E). The tumorsfrom the SW1222 CD44+CD24+ cells contained differentiatedcells, just as did the megacolonies studied in vitro (Fig. 3F).The overall results both from the clonal analysis of SW1222

cells and from the analysis of the SW1222 CD44+CD24+ sortedcells provide, in our view, unequivocal evidence for the existenceand identification of CSCs in the SW1222 cell line.

CSC Analysis in the HCT116 and HT29 Cell Lines. Two additional CRClines were chosen for CSC analysis along the same lines as thatdescribed above for SW1222. HCT116 is known to be a highlyaggressive cell line with little or no capacity to differentiate,although HT29 has an intermediate capacity to differentiate intoenterocytes and mucin-expressing lineages (12–14). The colony-forming abilities ofHCT116 andHT29 inMatrigel are shown inFig.4A. In this case, colonymorphology was classified by whether or notthey formed lumens. The colonies that were able to differentiate toform lumens were comparable to those that developed into mega-colonies in the SW1222 cell line. The data are consistent with thecharacterization of HCT116 as highly aggressive and non-differentiating, giving rise to no lumen-forming colonies; however,HT29 has retained at least some capacity to differentiate.Immunofluorescence analysis of colonies derived from these

cell lines, seeded in Matrigel and using the antibodies CDX1 andAUA1, is consistent with the above observations and the knownlack of CDX1 expression in HCT116 (11) (Fig. 4B). Both cell

Fig. 1. Immunofluorescence of SW1222 cell line colonies. (A) The SW1222cell line can differentiate and form two distinct colony types: megacolonieswith complex 3D lumen-like structures and small colonies. (Scale bar, 400μm.) (B) Immunofluorescence of SW1222 megacolonies in Matrigel showslow levels of CDX1 and CK20 expression at day 7. (C) By day 14, mega-colonies express much higher levels of CDX1 and CK20. Magnification: 20×.

A

B

Fig. 2. SW1222 subcloning in Matrigel. (A) Megacolonies and small coloniesgrowing in single wells in a 96-well plate were pooled separately, dis-aggregated into a single-cell suspension, and then replated in Matrigel in96-well plates, at a concentration of 60 cells per well. Megacolonies are ableto reform themselves, as well as small colonies, but small colonies can onlyreform themselves. (B) Cells from megacolonies are able to initiate tumors inNOD/SCID mice, but cells from small colonies are not tumorigenic.

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.0915135107 Yeung et al.

Page 3: Cancer stem cells from colorectal cancer-derived cell lines

lines produced colonies that were negative for PR4D4 staining.A few cells in HT29 colonies expressed CDX1, consistent withHT29’s limited capacity for differentiation.The FACS analyses of the CD44±CD24± subpopulations of

cells from these two cell lines, grown inMatrigel, are shown in Fig.4C. For these two cell lines, theCD44+CD24− population has onlyslightly less clonogenic capacity than the CD44+CD24+ sub-population, suggesting that, for these two lines, CD44 is the moreimportant determinant with respect to enriching for CSCs (15).Immunofluorescence staining with the Ki67 antibody, a well

known marker of dividing cells (16, 17), provides a furtherassessment of clonogenicity, as well as indicating the position ofthe most rapidly dividing cells in a colony. As shown in Fig. 5, thenumber of Ki67-positive cells correlates precisely with theaggressiveness of the cell lines and inversely with their propensityto differentiate. The Ki67 cells are in all cases found at the

periphery of the colonies. This finding is a priori where the mostrapidly dividing cells might be expected to be found. This couldbe the case even in differentiating colonies, where the peripheralcells are likely to be progenitor cells that divide more rapidly, butto a more limited extent, than the stem cells that might beexpected to be found nearer the center of a colony.CD44+CD24+ and CD44−CD24− cells from the HCT116 and

HT29 lines cells were injected into NOD/SCID mice, as beforefor SW1222. The results show that the selection does not seemto affect the tumor forming capacity of HCT116 cells, as thereis no significant difference in the tumor sizes between theCD44+CD24+ and CD44−CD24− subpopulations. For HT29, onthe other hand, there is a significant difference in size betweenthese two cell subpopulations, although it is not as great as for theSW1222 cell line (Fig. 6). These results suggest that HCT116contains a high proportion of CSCs that have lost the capacity to

Fig. 3. Characterization of CD44/CD24 selected SW1222 cells. (A) The extreme 0.5 to 1% of each CD44/CD24 subgroup by FACS analysis was taken for furtherexperiments. (B) CD44+CD24+ SW1222 cells are the most clonogenic and form the greatest numbers of megacolonies when grown in Matrigel, compared toother CD44/CD24 subgroups. (C) A single CD44+CD24+ cell can give rise to a megacolony forming multiple differentiated cell types: AUA-1 (anti-EpCam pan-epithelial marker), CDX1 (enterocyte), chromogranin A (enteroendocrine), and PR4D4 (mucin, see Fig. S6). Nuclei stained blue with DAPI. Megacolonies grownin Matrigel for 4 weeks. (D) CD44+CD24+ sorted cells are more tumorigenic in NOD/SCID mice: 200 and 1,000 cells of CD44+CD24+ and CD44−CD24− cells wereinjected into NOD/SCID mice (Upper) and assessed for tumor forming capacity. The results of the 200 and 1,000 cell groups were combined together in a 2 × 2contingency table (Lower), and analyzed statistically using Fisher’s exact test (P = 0.007). (E) Both CD44+CD24+ sorted and unsorted SW1222 cells are able toform well-differentiated adenocarcinomas that are histologically indistinguishable when injected into NOD/SCID mice (H&E stain, magnification 20×). (F)Xenografts derived from SW1222 CD44+CD24+ sorted cells express the differentiation markers chromogranin A (enteroendocrine) and CDX1 (enterocyte).

Yeung et al. PNAS Early Edition | 3 of 6

MED

ICALSC

IENCE

S

Page 4: Cancer stem cells from colorectal cancer-derived cell lines

differentiate. This finding is exactly what might be expected for anaggressive and poorly differentiated tumor and correlates clin-ically with lymph node metastases (18, 19) and reduced survival(20, 21). HT29, on the other hand, is in general similar to SW1222,with a detectable proportion of stem cells with the capacityto differentiate.

Increased Expression of CDX1 in HCT116 Reduces Clonogenicity andInduces Lumen Formation Within Colonies. We have previouslyshown that CDX1 expression regulates the differentiation of crypt

progenitor cells into mature intestinal epithelial cells (22) and, inparticular, directly controls the expression of CK20 and probablyother markers of differentiation of colonic epithelial cells as well(11). To test whether increasing the expression of CDX1 in anonexpressing cell line decreases clonogenicity and increases thecapacity to differentiate, we compared the clonogenic capacity ofthe HCT116 cell line stably transfected with a CDX1-expressingvector with that of the HCT116 cell line transfected with a vectorcontrol (11) (Fig. 7). The CDX1-expressing HCT116 cells aremuch less clonogenic, and also induce lumen formation withincolonies, which is absent from the vector control. This supports therole of CDX1 in controlling colonic epithelial cell differentiationand shows that CDX1 expressionmay enable theCSCs inHCT116to differentiate at least to some extent. The results also help toexplain why loss of CDX1 expression (by methylation) is relativelyfrequently selected for in colorectal cancers (22).

DiscussionThrough a combination of colony formation and surface markerselection, we have shown how CSCs can be recognized in the col-orectal cancer-derived cell line SW1222. Thus, between 20 and 30%of SW1222 cells can form megacolonies containing crypt-likestructures with multiple markers of differentiation when grown inMatrigel. The 3D structure of these megacolonies is strikinglysimilar to the morphology of organoids that can be generated fromnormal intestinal stem cells (23), suggesting that normal and cancerstem cells have similar mechanisms for self-renewal and differ-entiation. The expression of the differentiation marker, CDX1,increases with time in these colonies and cells from these coloniescan self-renew, differentiate, and initiate tumors in immunocom-promised NOD/SCID mice. Putative CSC markers, such as CD44,CD24 and, notably CD133, have been used to enrich for CSCs incolorectal and other tumors, although mainly in fresh tumor

Fig. 4. Cell line clonogenicity. (A) Clonogenicity of HCT116, HT29, and SW1222 in Matrigel. Numbers of lumen-forming and total colonies counted per well.One thousand cells plated in Matrigel per well in 96-well plates, grown for 2 weeks. (B) HCT116 and HT29 colonies grown in Matrigel for 2 weeks stain positivefor AUA1. HCT116 colonies are CDX1-negative, and HT29 colonies weakly express CDX1. Nuclei stained blue with DAPI. Magnification, 20×. (C) TheCD44+CD24+ and CD44+CD24− subpopulations are of comparable clonogenicity for HCT116 and HT29 cells grown in Matrigel. Three hundred cells seeded perwell in 24-well plates.

Fig. 5. Ki-67 staining for HCT116, HT29, and SW1222 colonies grown inMatrigel. Magnification, 20×.

4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.0915135107 Yeung et al.

Page 5: Cancer stem cells from colorectal cancer-derived cell lines

material (3, 4, 24, 25).CD44 siRNAknockdown in colorectal cancercells has been associated with suppression of xenograft growth innudemice (26).However,CD133 is not expressed inSW1222and socannot be a universal marker for CSC enrichment, at least in col-orectal cancer, and the combination of CD44 with CD24 has notbeen examined before in colorectal cancer cell lines nor in primarycolorectal cancer tissue. Theupper 0.5 to 1%ofCD44+CD24+ cellsfrom SW1222 gives rise to the highest proportion of crypt-formingmegacolonies; cells from these colonies can self-renew after severalroundsof selectionandcontinue toproduceall four combinationsofCD44±CD24± cells, as well as be the most efficient subset ofSW1222CD44±CD24± cells in forming tumors inNOD/SCIDmice.This finding is consistent with selection for CD44+CD24+ cells inSW1222 enriching significantly for CSCs in the SW1222 cell line.However, not all CD44+CD24+ cells are clonogenic, and a fewCD44−CD24− cells are also able to form colonies. More specificmarkers, including perhaps aldehyde dehydrogenase 1 (27), areneeded for enriching for CSCs. The more differentiated cells,including goblet and enteroendocrine cells, appear to be locatedtoward the periphery of the megacolonies, and the less differ-entiated cells more toward the center of the colonies, where thereare reduced levels of CDX1, chromogranin A, and mucin. Thisfinding suggests that the center of the megacolony may contain themajor part of the CSC population.Our results with SW1222 are consistent with earlier work on

the HRA-19 colorectal carcinoma cell line, which was shown tobe able to differentiate into multiple lineages, including colum-

nar, goblet, and enteroendocrine cells, from a single originalclone (28). Both sets of observations are consistent with evidencethat normal colonic crypts in both mice (29) and humans (30) areclonal, and so have normal stem cells that can give rise to all ofthe three types of differentiated cells found in colorectal crypts.These tumor-derived cell lines are therefore recapitulating, in aderanged manner, the situation in a normal crypt.The results for SW1222 have been extended to two other cell

lines, HCT116 and HT29, which represent different extents ofdifferentiation, or the lack of it. HCT116 is a highly aggressive cellline that shows no ability to differentiate and does not expressCDX1; HT29 has an intermediate capacity to differentiate. Com-paredwithHCT116, SW1222 retains the ability to differentiate andexpress high levels of CDX1, consistent with the hypothesis thatSW1222 contains a lower proportion of undifferentiated cancerstem cells. This result confers a less aggressive phenotype, asdemonstrated by a lower clonogenic and tumorigenic potentialthan HCT116. HCT116 cells do not significantly separate intodifferent types of colony-forming cells, nor intodifferent categorieswith respect to the ability to form tumors in NOD/SCIDmice. Thisfinding is consistent with the HCT116 cell line containing almostonlyCSCs, as also found byKai et al. (31).HT29, however, behavesin an analogous way to SW1222, although with a higher proportionof clonogenic cells and a lesser propensity to differentiate. Therelative numbers of dividing cells in the three lines are proportionalboth to their clonogenicity and tumor-formation capacity.It is now becoming quite clear that the proportion of CSCs in a

tumor can vary from very high and with little capacity to differ-entiate, as in HCT116, to much lower with much more differ-entiation, as inSW1222.Thepossibility thatCSCsmay in somecasesconstitute a relatively high proportion of the cells in a cancer issupported by recent data on highly efficient tumor formation fromprimary cells frommelanoma patients in NOD/SCID Il2rg−/−mice(32). Melanomas are, however, notoriously aggressive tumors.Theoretical models (33) can readily explain substantial varia-

tion in the proportion of CSCs versus differentiated cells in acancer. The ratio between CSCs and differentiated cells can beshown, on reasonable assumptions, to depend on the rate ofdifferentiation relative to the turnover rate of the CSCs. This isexactly as might be expected, namely that the less the propensityto differentiate, in general, the more aggressive the tumor andthe higher the proportion of CSCs in it. The CSC field has beendogged by the assumption that CSCs are necessarily a smallproportion of the cells in a cancer, which is clearly not the case.It was suggested by Wong et al. (22) that selection for reduced

expression of CDX1 in colorectal cancers (by methylation) islikely to be associated with a reduced capacity of the CSCs todifferentiate and so contributing to an increased rate of growth.Following this suggestion, Chan et al. (11) provided evidencethat CDX1 expression is a key switch for intestinal epithelialdifferentiation, probably mainly of columnar cells, so thatstudying the patterns of expression of CDX1 should enablebetter characterization of CSCs both in colorectal cancer celllines and in primary colorectal cancer. The data presented in thisarticle corroborate that suggestion, especially in showing thatforced expression of CDX1 in HCT116 leads to reduced clono-genicity and, most significantly, a substantial proportion of crypt-forming colonies. This finding is consistent with the control ofcolumnar cell differentiation by CDX1 and directly explains theadvantage of selection for CDX1 down-regulation.In preliminary further studies we have extended the classi-

fication of our cell-line panel into crypt colony-forming versusnot, and are using this classification for the search for furthermarkers for stem cell characterization and enrichment. We willalso explore developing our colony-forming assays using themore sophisticated cell culture conditions described by Ootaniet al. (34) and for the application to fresh tumor tissue.

p<0.001 p=0.27* Harvested day 48 ** Harvested day 49

Rank Cell phenotype Size, mm31 CD44+CD24+ *2 CD44+CD24+ **3 CD44+CD24+ 3684 CD44+CD24+ 9.45 CD44-CD24- 06 CD44-CD24- 07 CD44-CD24- 08 CD44-CD24- 0

200 cells SW1222 Day 56

Rank Cell phenotype Size, mm31 CD44+CD24+ 1412 CD44+CD24+ 1103 CD44+CD24+ 1024 CD44+CD24+ 625 CD44-CD24- 9.46 CD44-CD24- 07 CD44-CD24- 08 CD44-CD24- 0

200 cells HT29 Day 50

Rank Cell phenotype Size, mm31 CD44+CD24+ 4022 CD44+CD24+ 3603 CD44-CD24- 2264 CD44-CD24- 2015 CD44+CD24+ 1846 CD44+CD24+ 1787 CD44-CD24- 818 CD44+CD24+ 379 CD44-CD24- 0

200 cells HCT116 Day 39

Fig. 6. Xenograft data ranked by size of tumor. SW1222 CD44+CD24+ cells aremore tumourigenic than CD44−CD24− cells (Fig. 3D). Tumors derived fromHT29CD44+CD24+ cells are significantly larger than those derived from CD44−CD24−

cells (unpaired t test, P < 0.001), but that is not the case for the HCT116 cell line(P = 0.27). Shaded rows highlight individual mice that developed tumors.

Fig. 7. HCT116 stable-transfectant clonogenicity. (A) Forced CDX1 expres-sion in HCT116 colonies reduces clonogenicity and results in the develop-ment of lumen-forming colonies. (B) HCT116 vector control cell line producesno lumen-forming colonies. (C) Forced CDX1 expression induces lumen for-mation within HCT116 colonies.

Yeung et al. PNAS Early Edition | 5 of 6

MED

ICALSC

IENCE

S

Page 6: Cancer stem cells from colorectal cancer-derived cell lines

Wehaveobtainedpreliminary evidence, using ourCSCs from thecell lines, that hypoxia inhibits differentiation and so maintainsstemness, as has been shown in neuroblastoma tumors (35). Thereare clearly many advantages to working with CSCs from cell lines,both for basic investigation of the properties of CSCs and throughthose findings, identification of novel targets for drug development,aswell as for thepossible development of relatively high-throughputdrug-response assays for CSCs. We suggest that the data presentedin this article provide a basis for these important developments.

Materials and MethodsDetailed discussion can be found in the SI Materials and Methods.

Cell Lines. HCT116 and HT29 colorectal cancer cell lines were originallyobtained from the American Type Culture Collection (ATCC). The SW1222 celllinewas a generous gift fromMeenhardHerlyn,Wistar Institute, Philadelphia.All three cell lines were cultured in complete DMEM supplemented with 10%FBS and 1%penicillin/streptomycin (Invitrogen). Cells were incubated at 37 °Cin a humidified 10% CO2 environment and were grown to 50 to 80% con-fluence before the next passage or further experiments. Cellular suspensionswere obtained through trypsinization (0.5% trypsin for 2–5 min) and singlecells were obtained by passing cells through a 20-μm filter (Celltrics; Partec).The stably transfected HCT116-CDX1 and vector-control cell lines were pre-viously generated in our laboratory and maintained with geneticin selectionpressure (11). All colony counts are based on the counts from three wells.

Immunohistochemistry and Confocal Immunofluorescence. Immunohis-tochemistry was performed according to standard procedures. Dilutions ofantibodyusedwere: anti-CDX1mAb (in-house; 1:100), anti-KRT20mAb (Dako;1:100), AUA1 anti-EpCammAb (in-house; 1:100), PR4D4mAb (in-house; 1:10),chromogranin A pAb (Dako; 1:500) and anti-Ki67 mAb (Dako; 1:100)

Fluorescent-Activated Cell Sorting. Fluorescent-activated cell sorting wasperformed using standard protocols. Cells were labeled with mouse anti-human CD44-PE and mouse anti-human CD24-FITC (Caltag; Invitrogen). Cellswere analyzed using a DakoCytomation Cyan machine, or sorted using a

MoFlo cell sorter (Beckman Coulter). The top or bottom 0.5 to 1% of eachCD44/CD24 subpopulation was collected for further experiments.

Matrigel Colony Morphology Assay. Single cell suspensions were obtainedusing filtration and resuspended in a 1:1 mixture of Matrigel (BD Biosciences)and medium, and plated in 96-well plates in a limiting dilution. After 4 weeksof incubation, wells with single colonies were identified and colonies wereretrieved using Matrigel Recovery Solution (BD Biosciences) as per manu-facturer’s protocol. Viability of the cells was checked using Trypan Bluebefore being used for further experiments.

Live and Dead Cell Staining in Matrigel. Cells were grown in a 1:1 mixture ofMatrigel and medium at a density of 500 cells per 50 μL in each chamber ofthe Nunc 8-well slides. Colonies were grown for up to 4 weeks. Cells wereincubated with both fluorescein diacetate (Fluka) and propidium iodide (BDPharmagen), and then visualized under fluorescence.

In Vivo Tumorigenic Assays. Single cells were resuspended in a 1:1 mixture ofMatrigel and serum-free medium (DMEMwith 1% penicillin/streptomycin). A100-μL suspension containing between 200 and 1,000 cells was injected s.c.into the flanks of 6- to 10-week-old female NOD/SCID immunodeficientmice, obtained from the John Radcliffe Hospital Biomedical Services, Oxford,United Kingdom. Animal studies were conducted according to the Universityof Oxford institutional guidelines and within the limits of the Project Licenseissued by the Home Office, United Kingdom.

Statistical Analysis. All data are shown as mean ± SEM. For 2 × 2 contingencytables, Fisher’s exact test was used.

ACKNOWLEDGMENTS. We thank Kevin Clark and Craig Waugh for theirexpertise on cell sorting, Joy Winter for processing histological specimens,Fillitsia Ntouroupi for her help with immunofluorescent microscopy, PeterThomas for his help on the confocal microscope, Professor Zhanfeng Cui forthe use of the multiphoton microscope, and Sally Hill, Karla Watson, OsamaAl-assar, and Thomas Brunner for their help with animal work. This work wasmainly funded by a Cancer Research U.K. program grant (to W.F.B. asPrincipal Investigator), by a Rhodes Scholarship (S.C.G.), and by support fromthe Jacqueline Seroussi Memorial Foundation for Cancer Research. T.M.Y. isa Royal College of Surgeons of England Research Fellow.

1. Reya T, Morrison SJ, Clarke MF, Weissman IL (2001) Stem cells, cancer, and cancer stemcells. Nature 414:105–111.

2. Bonnet D, Dick JE (1997) Human acute myeloid leukemia is organized as a hierarchythat originates from a primitive hematopoietic cell. Nat Med 3:730–737.

3. O’Brien CA, Pollett A, Gallinger S, Dick JE (2007) A human colon cancer cell capable ofinitiating tumour growth in immunodeficient mice. Nature 445:106–110.

4. Ricci-Vitiani L, et al. (2007) Identification and expansion of human colon-cancer-initiating cells. Nature 445:111–115.

5. Liu Y, Bodmer WF (2006) Analysis of P53 mutations and their expression in 56colorectal cancer cell lines. Proc Natl Acad Sci USA 103:976–981.

6. Volchenboum SL, et al. (2009) Comparison of primary neuroblastoma tumors andderivative early-passage cell lines using genome-wide single nucleotidepolymorphism array analysis. Cancer Res 69:4143–4149.

7. Douglas EJ, et al. (2004) Array comparative genomic hybridization analysis ofcolorectal cancer cell lines and primary carcinomas. Cancer Res 64:4817–4825.

8. Willson JK, Bittner GN, Oberley TD, Meisner LF, Weese JL (1987) Cell culture of humancolon adenomas and carcinomas. Cancer Res 47:2704–2713.

9. Richman PI, Bodmer WF (1988) Control of differentiation in human colorectalcarcinoma cell lines: epithelial-mesenchymal interactions. J Pathol 156:197–211.

10. Pignatelli M, Bodmer WF (1988) Genetics and biochemistry of collagen binding-triggered glandular differentiation in a human colon carcinoma cell line. Proc NatlAcad Sci USA 85:5561–5565.

11. Chan CW, et al. (2009) Gastrointestinal differentiation marker Cytokeratin 20 isregulated by homeobox gene CDX1. Proc Natl Acad Sci USA 106:1936–1941.

12. Demers MJ, et al. (2009) Intestinal epithelial cancer cell anoikis resistance: EGFR-mediated sustained activation of Src overrides Fak-dependent signaling to MEK/Erkand/or PI3-K/Akt-1. J Cell Biochem 107:639–654.

13. Choi SR, et al. (2000) Biological properties and expression of mucins in 5-fluorouracilresistant HT29 human colon cancer cells. Int J Oncol 17:141–147.

14. Chakrabarty S (1992) Regulation of human colon-carcinoma cell adhesion toextracellular matrix by transforming growth factor beta 1. Int J Cancer 50:968–973.

15. Li C, et al. (2007) Identificationofpancreatic cancer stemcells.CancerRes67:1030–1037.16. Cattoretti G, et al. (1992) Monoclonal antibodies against recombinant parts of the Ki-

67 antigen (MIB 1 and MIB 3) detect proliferating cells in microwave-processedformalin-fixed paraffin sections. J Pathol 168:357–363.

17. Scholzen T, Gerdes J (2000) The Ki-67 protein: from the known and the unknown. JCell Physiol 182:311–322.

18. Chung CK, Zaino RJ, Stryker JA (1982) Colorectal carcinoma: evaluation of histologicgrade and factors influencing prognosis. J Surg Oncol 21:143–148.

19. Berti Riboli E, et al. (1983) Colorectal cancer: relationship of histologic grading todisease prognosis. Tumori 69:581–584.

20. Halvorsen TB, Seim E (1988) Degree of differentiation in colorectal adenocarcinomas:a multivariate analysis of the influence on survival. J Clin Pathol 41:532–537.

21. Bjerkeset T,Morild I,Mørk S, SøreideO (1987) Tumor characteristics in colorectal cancerand their relationship to treatment and prognosis. Dis Colon Rectum 30:934–938.

22. Wong NA, et al. (2004) Loss of CDX1 expression in colorectal carcinoma: promotermethylation, mutation, and loss of heterozygosity analyses of 37 cell lines. Proc NatlAcad Sci USA 101:574–579.

23. Sato T, et al. (2009) Single Lgr5 stem cells build crypt-villus structures in vitro withouta mesenchymal niche. Nature 459:262–265.

24. Vermeulen L, et al. (2008) Single-cell cloning of colon cancer stem cells reveals a multi-lineage differentiation capacity. Proc Natl Acad Sci USA 105:13427–13432.

25. Haraguchi N, et al. (2008) CD133+CD44+ population efficiently enriches colon cancerinitiating cells. Ann Surg Oncol 15:2927–2933.

26. Subramaniam V, Vincent IR, Gilakjan M, Jothy S (2007) Suppression of human coloncancer tumors in nude mice by siRNA CD44 gene therapy. Exp Mol Pathol 83:332–340.

27. Huang EH, et al. (2009) Aldehyde dehydrogenase 1 is a marker for normal andmalignant human colonic stem cells (SC) and tracks SC overpopulation during colontumorigenesis. Cancer Res 69:3382–3389.

28. Kirkland SC (1988) Clonal origin of columnar, mucous, and endocrine cell lineages inhuman colorectal epithelium. Cancer 61:1359–1363.

29. Ponder BA, et al. (1985) Derivation of mouse intestinal crypts from single progenitorcells. Nature 313:689–691.

30. Novelli MR, et al. (1996) Polyclonal origin of colonic adenomas in an XO/XY patientwith FAP. Science 272:1187–1190.

31. Kai K, et al. (2009) Maintenance of HCT116 colon cancer cell line conforms to astochastic model but not a cancer stem cell model. Cancer Sci 100:2275–2282.

32. Quintana E, et al. (2008) Efficient tumour formation by single human melanoma cells.Nature 456:593–598.

33. Johnston MD, Edwards CM, Bodmer WF, Maini PK, Chapman SJ (2007) Examples ofmathematical modeling: tales from the crypt. Cell Cycle 6:2106–2112.

34. Ootani A, et al. (2009) Sustained in vitro intestinal epithelial culture within a Wnt-dependent stem cell niche. Nat Med 15:701–706.

35. Jögi A, et al. (2002) Hypoxia alters gene expression in human neuroblastoma cells towardan immature and neural crest-like phenotype. Proc Natl Acad Sci USA 99:7021–7026.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.0915135107 Yeung et al.