Cancer stem cells (CSCs) are initiating cells in colorectal cancer (CRC). Colorectal tumours undergo epithelial to mesenchymal transition (EMT)-like processes at the invasive front, enabling invasion and metastasis, and recent studies have linked this process to the acquisition of stem cell-like properties. It is of fundamental importance to understand the molecular events leading to the establishment of cancer initiating cells and how these mechanisms relate to cellular transitions during tumourigenesis. We use an in vitro system to recapitulate changes in CRC cells at the invasive front (mesenchymal-like cells) and central mass (epithelial-like cells) of tumours. We show that the mesoderm inducer BRACHYURY is expressed in a subpopulation of CRC cells that resemble invasive front mesenchymal-like cells, where it acts to impose characteristics of CSCs in a fully reversible manner, suggesting reversible formation and modulation of such cells. BRACHYURY, itself regulated by the oncogene β-catenin, influences NANOG and other ‘stemness’ markers including a panel of markers defining CRC-CSC whose presence has been linked to poor patient prognosis. Similar regulation of NANOG through BRACHYURY was observed in other cells lines, suggesting this might be a pathway common to cancer cells undergoing mesenchymal transition. We suggest that BRACHYURY may regulate NANOG in mesenchymal-like CRC cells to impose a ‘plastic-state’, allowing competence of cells to respond to signals prompting invasion or metastasis.
Colorectal tissue undergoes continuous renewal to maintain normal function. The renewal process is brought about by multipotent, tissue-specific stem cells that give rise to specialised cell types of this tissue.1, 2 Such tissue-specific stem cells are critical for regulating self-renewal and differentiation so that cell proliferation is balanced with turnover. Signalling through components of the Wnt/β-catenin pathway is key in maintaining stem cells in many epithelial tissues, including the intestine.3–6 In addition to its role in maintaining adult stem cells, Wnt signalling is important in the maintenance of pluripotency in mouse and human embryonic stem (ES) cells.7, 8
One of the fundamental downstream effectors of these signals, required for promotion of pluripotency in both mouse and human ES cells, is the homeodomain transcription factor NANOG.9–12 The NANOG gene has been shown to be expressed in tissue-specific stem cells, such as the pancreas,13 and in some cancers such as germ cell tumours14 and Ewings sarcoma.15 Recent studies have also provided evidence that NANOG plays a role in regulating the development of tumours in humans.16
Tissue-specific stem cells may serve as cancer initiating cells when processes that regulate their self-renewal or differentiation go awry: these cells have become known as cancer stem cells (CSCs).17 Although the concept of CSC has been accepted for haematopoetic malignancies for some years, colorectal tumours have only relatively recently been described as having a minor component of cells expressing specific markers associated with a stem cell phenotype.18 The current dogma proposes that colorectal tumours comprise tumour stem cells (derived from crypt stem cells) that can differentiate into epithelial cells, which represent the main tumour mass.19–21 Numerous markers have been identified for CSCs of colorectal cancers (CRCs). The mesenchymal stem cell marker CD166 has been described as being a specific CSC marker for CRC compared to the CD133 marker,22 a finding supported by others.23–25 However, CD133 has been shown to be present on intestinal stem cells that are susceptible to neoplastic transformation21 and the AC133 antibody, that recognises a glycosylated epitope of CD133, is reported to be a good marker for CSCs of colorectal origin.26 CD44 has also been used as a marker for CRC-CSCs, but immunohistochemical analysis of colonic tissue has shown that the zone of CD44 expression extends beyond the stem cell compartment into the region of proliferation.27 More recently, aldehyde dehydrogenase (ALDH1) has proven to be a robust marker of colon CSCs.28
Just as Wnt/β-catenin signalling has been shown to be involved in maintaining a stem cell phenotype, β-catenin signalling has also been identified as a key factor maintaining a CSC phenotype in epidermal tumours29, 30 and more recently has been shown to serve as a marker for tumour cell clonogenicity and cancer stemness.31
Epithelial to mesenchymal transition (EMT) is a fundamental process during development and in the adult, whereby epithelial cells lose their cell:cell connections and migrate to alternative locations. An EMT-like process also occurs in cancer where the transition from a stable, epithelial phenotype to a migratory, mesenchymal phenotype promotes invasion and metastasis of the tumour.32 Recently, a possible link between cells undergoing EMT and cells with properties of CSCs has been described in epithelial cells.33 It is not easy to study events surrounding EMT during progression of cancer, partly due to the very transient nature of this process in situ. We therefore used CRC cells from the cell line SW480, grown to low and high density, as a model for cells transitioning between a mesenchymal-like and epithelial-like cell state, respectively.34, 35 We show that the founding member of the T-box transcription factor family of genes, BRACHYURY, which encodes a mesoderm inducer, is differentially expressed in low-density (mesenchymal-like) cells. We present novel data demonstrating that BRACHYURY, which we show to be regulated by β-catenin, acts to maintain levels of the pluripotency determinant NANOG in low-density cells and maintains expression of a range of markers characterizing CSCs. We present data to show that NANOG is similarly regulated by BRACHYURY in other cancer cell lines suggesting that this may be a general pathway in tumour cells capable of mesodermal differentiation.
Material and Methods
SW480 cells were maintained in Dulbecco's modified eagle medium (DMEM)/10% fetal bovine serum (FBS). Subconfluent cells were seeded at 5 × 106 cells/T75 flask and grown for 24 hr, and super-confluent cells were seeded at the same density but grown for 4 days. T84 cells were grown in a 1:1 ratio of DMEM/10% FBS and Ham's F-12 media (Invitrogen, Paisley, UK, GIBCO N6658)/10% FBS. T84 cells were seeded at 3.6 × 107 cells/T75 flask and grown for 24 hr for subconfluent cells. NTERA2/D1 cells were cultured in DMEM (Sigma-Aldrich, Dorset, UK/Invitrogen)/10% FBS (Invitrogen) and were seeded at 5 × 105 cells/six-well plate and grown for 24 hr for subconfluent cells. H460 cells were grown in Roswell park memorial institute (RPMI)-1640 American type culture collection (ATCC)/10% FBS.
Quantitative RT-PCR analysis
Total RNA was isolated with TRIzol reagent (Invitrogen), 1 μg total RNA was used as template for cDNA synthesis. Reverse transcription was with oligo(dT) primers (500 μg/ml), using SuperScript III first-strand cDNA synthesis kit (Invitrogen). Control reactions using mock cDNA preparations lacking reverse transcriptase were run in parallel for each analysis to ensure absence of genomic DNA contamination. qPCRs were performed with the 2× Power SYBR Green Mastermix (Applied Biosystems, Warrington, UK, Cat. no. 4368702) using 1.5 μl of cDNA. The samples were loaded in AB7900HT instrument, using manual baseline and manual Ct settings between 3 and 15.
Primers for RT-PCR analysis of NANOG are described in Refs.16 and36. All other primers are shown in Supplementary Information Tables 1 and 2.
Cells were lysed in 200 mM NaCl, 50 mM Tris (pH 7.4), 0.5% Triton X-100, protease inhibitors. Membranes were blocked with 10% dry milk/phosphate buffered saline (PBS) and incubated with primary antibody in PBS-0.1% Tween (monoclonal antibodies) or PBS-0.5% Tween (polyclonal antibodies) overnight/4°C. Secondary antibodies were added for 1 hr/room temperature. Primary antibodies include CD133 (Abcam, Cambridge, UK, cat no. ab19898: antibody recognizes glycosylated CD133, used at 1:500 dilution), anti-Brachyury (R&D systems, Abingdon, UK, cat no. AF2085, used at 1:1,000 dilution), anti-Nanog (Abcam, cat no. ab21624, used at 1:300 dilution), anti-Lamin B (Santa Cruz, Wembley, UK, cat no. sc-6217, used at 1:1,000 dilution), CD166 (R&D systems, cat no. MAB6561, used at 1:250 dilution), anti-β-catenin (Santa Cruz, cat no. sc-7199, used at 1:1,000 dilution), anti-Lamin A antibody (Abcam, cat no.ab40579, used at 1:750 dilution). Secondary antibodies include donkey anti-mouse IgG, Jackson Immuno Research (Newmarket, UK) (cat no. 715-035-150, used at 1:20,000 dilution), donkey anti-rabbit IgG (Jackson Immuno Research, cat no. 711-035-152, used at 1:25,000 dilution), rabbit anti-goat IgG (Sigma, cat no. A5420, used at 1:10,000 dilution).
Immunostaining and confocal microscopy
Cells were fixed with 4% paraformaldehyde, 10 min, permeabilised with 0.2% triton-X-100/PBS, 20 min, and blocked using 5% FBS/PBS, 1 hr, at room temperature. Primary antibodies include CD133 (Abcam, cat no. ab19898, used at 1:150 dilution), anti-Brachyury (Abcam, ab57480, used at 1:250 dilution), anti-Nanog (Abcam, cat. no. ab21624, used at 1:150 dilution), anti-Nanog (Abcam, cat no. ab62734, used at 1:150 dilution), CD166 (R&D systems, cat no. MAB6561, used at 1:200 dilution), NFκB p65 (Abcam, ab32536, used at 1:1000 dilution). Secondary antibodies include goat anti-mouse AlexaFluor 568 (Invitrogen, used at 1:400); goat anti-rabbit AlexaFluor 488 (Invitrogen, used at 1:200). Cells were imaged using a Zeiss LSM 510 confocal microscope; similar conditions were applied to all images.
Brachyury siRNA (cat no. SI00738255) and β-catenin siRNA (SI02662478) was from Qiagen, Crawley, UK. SW480, T84 and NTERA2/D1 cells were transfected with 5 nM of siRNA duplex using Hiperfect transfection reagent (Qiagen; cat no. 301705) at a ratio 1:1.
Differential gene expression in SW480 cells grown to low and high density
An in vitro system that recapitulates dynamic changes that occur in CRC cells during EMT-like processes would be of great use in defining determinants of these changes and would contribute to the analysis of mesenchymal-epithelial transition (MET)-like redifferentiation processes occurring during metastasis. As mentioned, SW480 CRC cells grown to low and high density resemble cells at the invasive front and central mass of the tumour, respectively, reversibly undergoing EMT/MET.35, 37, 38 We further validated this system of EMT through NFkB relocalisation as described in Supporting Information Figure 1. As transient activation of genes involved in EMT occurs at the invasive front of epithelial tumours, we used this in vitro system to determine whether a selected group of genes, involved in mesoderm induction during normal development, are differentially expressed in low- and high-density SW480 cells. BRACHYURY, the product of which is one of the main mesoderm inducers during normal development,39 was differentially expressed in a density-dependent manner in SW480 cells, with increased expression observed in low-density cells (Fig. 1a). Density-dependent differential expression of BRACHYURY was shown to be fully reversible on prolonged passaging of the cells with newly plated, low-density cells expressing BRACHYURY to high levels compared to cells grown to high density. This reversible change in gene expression is therefore not likely to be due to progressive genetic changes. Furthermore, identical expression patterns were observed in early passage SW480 cells (P15) compared to high passage cells (up to P50), again suggesting that this is not due to fixed genetic changes acquired during the passaging of cells.
Endogenous expression of BRACHYURY in CRC has recently been reported in SW480 cells and other tumour types,40, 41 further validating our observations of this system. Based on our initial findings, we carried out quantitative RT-PCR (qRT-PCR) on a number of additional genes to determine the relative changes in expression levels (Fig. 1a). We observed expression changes that might be associated with MET from low to high cell density, including a decrease in expression of SNAIL, SLUG, ZEB1 and BRACHYURY and an increase in expression of E-CADHERIN.
Reversible β-catenin translocation to the nucleus is a hallmark of malignant CRC cells undergoing EMT.42 Low-density SW480 cells show nuclear localization of β-catenin, low E-CADHERIN levels (cytoplasmic location), nuclear NFκB (Supporting Information Fig. 1) and gene expression patterns with mesenchymal phenotype (Fig. 1a). In contrast, high-density SW480 cells have cytoplasmic β-catenin and NFκB localization, high E-CADHERIN levels and gene expression patterns consistent with an epithelial phenotype. Our observations add further support to the use of SW480 cells grown to low and high densities as a model for cells transitioning between invasive, mesenchymal-like cells and differentiated epithelial-like cells, respectively, in CRCs.
NANOG expression is altered in response to changes in BRACHYURY in low-density SW480 cells
Invasive regions of tumours are characterized by the loss of membrane attachments and activation of transcription factors involved in EMT. Therefore, to further clarify the role of endogenous levels of BRACHYURY in CRC, we studied the expression patterns of a repertoire of genes known to be involved in these processes that might be regulated by BRACHYURY. These included known targets of BRACHYURY such as NANOG,43WNT11, ORTHOPEDIA, IL2, IL15, epithelial markers of cell polarity such as Crumbs (CRB3), PALS1, EVA1, a marker for intestinal epithelium, CDX2, regulators of EMT such as ZEB1 and ZEB2 and E-CADHERIN. siRNA was used to mediate knockdown of BRACHYURY in low-density SW480 cells, and expression levels were assessed by qRT-PCR (Fig. 1b). These experiments revealed modest changes in the expression of EMT regulators (ZEB1 and ZEB2, SNAIL, SLUG and E-CADHERIN) and no sizeable change in expression of cell polarity markers (CRB3, PALS1 and EVA1). These data confirmed our earlier observations (on β-catenin localisation) that a reduction in endogenous BRACHYURY expression alone is not sufficient to reverse EMT in these cells. However, we did note a consistent and relatively large (73%) reduction in expression of NANOG (and other BRACHYURY target genes, not addressed further here) as a result of BRACHYURY knockdown in low-density SW480 cells.
NANOG is an important determinant of pluripotency in ES, embryonal carcinoma (EC) and germ cells and has been shown to be present in a number of different cancer cell types where it is involved in tumour progression.16 The regulation of BRACHYURY and subsequent effect on NANOG was therefore investigated further.
Expression of NANOG pseudogenes in the cancer cell lines SW480, T84 and H460
Characterisation of NANOG expression has been somewhat confused and complicated by the presence of pseudogenes of NANOG in some cancer cell lines.44, 45, 16, 36 The processed pseudogene, NANOGP816 and the nonprocessed pseudogene NANOG236 have been shown to be preferentially expressed in some cancer cell lines where they have the potential to generate protein products that function in tumour development, behaving similarly to NANOG1.16, 36 We set out to define the origin of NANOG mRNA in SW480 cells (and other BRACHYURY expressing cell types T84, H460, NTERA/D1)40, 41 using a panel of primers specific for different NANOG pseudogenes (results shown in Supporting Information Table 3). PCR products were confirmed by sequencing. Our results demonstrate that SW480 cells and H460 express NANOG2 and do not appear to express either NANOG1 or NANOGP8. T84 cells do not express NANOGP8, but were found to express NANOG1 and NANOG2. The other BRACHYURY+ cell line used in our study, NTERA2/D1, has previously been shown to express both NANOG1 and NANOG2 and the protein product may be derived from one or other of the transcripts of these two genes.36 Our data for NANOG expression in NTERA2/D1 cells are in compliance with previously published data as we also observed expression of NANOG1 and NANOG2 in these cells (data not shown). The size of the NANOG protein present in SW480 cells was identical to that in NTERA2/D1 and T84 cells, running at approximately 34 kDa. However, NTERA2/D1 also displayed a less prominent species of NANOG1 at around 48 kDa, reported to be due to post-translational modification of NANOG136 (data not shown). Based on the fact that only NANOG1, NANOG2 and NANOGP8 are able to be translated into functional proteins, we suggest that the NANOG protein present in SW480 cells is derived from NANOG2 mRNA.
Brachyury expression in low-density SW480 cells is regulated by β-catenin
β-catenin is a key oncogene product involved in the progression of CRC, identified in the nucleus of cells at the invasive front of colorectal tumours and invading cells in the surrounding stromal tissue2, 29, 30 a finding that is reproduced in our model. As BRACHYURY has been shown to be a direct target of Wnt in the mouse,46 we used siRNA-mediated reduction of β-catenin to determine whether β-catenin regulates BRACHYURY protein levels and expression of its target genes in low-density SW480 cells. Reduced levels of β-catenin indeed lead to a decrease in levels of BRACHYURY, NANOG and the marker of CRC-SCs, CD133 (see comments in sections below for specificity of this marker) as determined by Western blot analysis (Fig. 2 and densitometry, Supporting Information Fig. 2). These data suggest that β-catenin acts upstream of BRACHYURY which then regulates NANOG and the CSC marker CD133 (also known as Prominin 1). Interestingly, activation of β-catenin signalling in the mouse small intestine has recently been shown to result in an increased population of cells at the base of the crypt expressing CD133/Prominin 1.21 Our results suggest that BRACHYURY plays a role in transducing the signal from β-catenin to influence the levels of CD133 in CRC cells.
Cells expressing BRACHYURY and NANOG also display the colorectal cancer stem cell markers, CD166, CD133, CD44
As BRACHYURY influences expression of the pluripotency gene NANOG in SW480 CRC cells, we were interested to determine whether cells that were positive both for BRACHYURY and NANOG also displayed markers of CRC-CSCs. Specifically defining a population of cells within a colorectal tumour as being CSCs currently relies on definition by a number of markers. CSCs have recently been identified in the colon as being included in a CD133+ population.18–21 However, neither all tumours display this marker nor is every CD133+ cell necessarily a CSC.22–24 Recently, CD133 has been shown to label both mouse SC of the small intestine and early transit amplifying progenitor cells.25 This is in compliance with the CSC population potentially being contained within, but not exclusively marked by CD133. In a different study, CD133 was shown to mark mouse small intestinal stem cells, but not colon stem cells.21 Further studies have shown that the antibody AC133 detects the glycosylated form of CD133 which may be specific for CSCs versus the unglycosylated form of CD133 which is also expressed in differentiated cells.26 CD166 is potentially a good marker for CRC-SCs.22 Nearly 60% of CRCs are shown to have strong cytoplasmic staining, and 30% have membrane staining of CD166, and expression of CD166 in CRC is associated with poor prognosis.47 CD4422 and ALDH128 are also described as markers for CRC-CSCs; although again, the domain of localization of CD44 may extend beyond the SC compartment at the base of the crypt and into the proliferative compartment.28 However, the presence of multiple markers should provide a good indication of the ‘stemness’ of the CRC cells. We therefore studied the presence of these CRC-SC markers in SW480 cells and correlated this to BRACHYURY and NANOG levels using specific antibodies. Figures 3a,3c,3e,3g and 3i shows that at low density, all SW480 cells are positive for NANOG (nucleus), BRACHYURY (nucleus), CD166, CD133 and CD44 (cell surface). In contrast, the majority of cells grown to high density were not positive for NANOG, BRACHYURY, CD166, CD133 or CD44 antibodies (see DIC images, Figs. 3b,3d,3f,3h and 3j for representation of total cell numbers and Supporting Information Fig. 3 for graphical representation). It should be noted that we observed rare, high-density cells that were positive for both NANOG and BRACHYURY (Figs. 3b,3d,3f,3h and 3j) and on occasion, we also observed cells that were negative for NANOG but positive for BRACHYURY. However, NANOG positive cells were always BRACHYURY positive, substantiating our finding that BRACHYURY regulates NANOG levels but not vice versa. The presence of CD166 was closely linked with BRACHYURY and NANOG. Most of the rare high-density cells that were positive for CD166 were also positive for NANOG, although a few cells did display CD166 and not BRACHYURY or NANOG. However, NANOG positive cells were always positive for CD166, further suggesting that NANOG regulates CD166 levels. CD44 staining showed similar distribution to CD166. In contrast, CD133 staining was sometimes observed in high-density cells in the absence of either NANOG or BRACHYURY; consistent with published data suggesting that CD133 may have a broader specificity.
Brachyury is required to maintain the CSC (CD166+/CD133+/CD44+/ALDH1+) phenotype of CRC cells
Whilst at low density all cells stained positive for BRACHYURY, NANOG, CD166 and other CSC markers, high-density cells rarely stained positive for these proteins. Western blotting (Fig. 4a) confirms this observation demonstrating the progressive decrease in total levels of NANOG, BRACHYURY, CD166, CD44 and CD133 as cells undergo transition from low to high density (see also Supporting Information Fig. 4a for densitometry measurements).
The requirement for BRACHYURY to maintain levels of NANOG and markers of the CSC phenotype was determined using siRNA-mediated knockdown of BRACHYURY in low-density SW480 cells. Figure 4b shows that when levels of BRACHYURY are decreased, levels of NANOG are reduced similarly. Likewise, levels of CD166, CD133, CD44 and ALDH1 are decreased after BRACHYURY knockdown. Whilst our data (IF) show that CD133+ cell staining is not restricted to NANOG+ or BRACHYURY+ cells (data not shown), nevertheless, levels of CD133 are somewhat reduced upon BRACHYURY knockdown (Fig. 4b, see also Supporting Information Fig. 4b for densitometry measurements).
This observation is consistent with the suggestion that the CSC population is contained within a subpopulation of CD133 cells. In contrast, CD166, CD44 and ALDH1 demonstrate more reliable efficacy as CSC markers in our experiments in as much as expression is linked more closely to the presence of the pluripotency marker, NANOG. Taken together, our data suggest that BRACHYURY is required to maintain the levels of NANOG and the CSC markers CD166, CD44, CD133 and ALDH1 in low-density, mesenchymal-like, SW480 cells.
Expression of pluripotency genes is reduced after BRACHYURY knockdown in the BRACHYURY+/NANOG+ mesenchymal-like cell population of SW480 cells
‘Stemness’ can be defined by the expression of so called ‘pluripotency genes’. In further determining whether the population of low-density, mesenchymal-like SW480 cells (defined as BRACHYURY+/NANOG+) display further characteristics of pluripotent cells, and therefore cancer initiating cells, we looked at expression of key genes involved in pluripotency and how this is affected after knockdown of BRACHYURY. Specifically we studied expression of SOX2, OCT4 and FOXD3 in low- and high-density SW480 cells and in response to knockdown of BRACHYURY. Figures 5a and 5b shows that the expression of both OCT4 and SOX2 is higher in low-density cells versus high-density cells (with 50% and 78% reduction in expression levels, respectively). There is no large change in expression levels of FOXD3 as cells undergo transition from low to high density. Knockdown of BRACHYURY in low-density cells results in reduction of expression of SOX2 (67% reduction compared to noninterfering controls) and FOXD3 (69% reduction) (Figs. 5c and 5d), but the decrease in levels of the pluripotency genes SOX2, FOXD3 and NANOG is not consistent with our observations for OCT4 expression, which remains unaltered after BRACHYURY knockdown, suggesting that a reduction in levels of BRACHYURY and NANOG alone are not sufficient to regulate levels of OCT4. OCT4 has a number of processed but nonfunctional pseudogenes, and our expression data for OCT4 may be falsely elevated due to contributions from pseudogenes. [Western blotting shows a decrease in levels of OCT4 (A isoform) after knockdown of BRACHYURY in NTERA2/D1 cells; in compliance with the decrease in gene expression data for SOX2, FOXD3 and NANOG (data not shown). It is likely that regulation of OCT4 is complex and may be cell line dependent.]
Our results demonstrate that the expression of key pluripotency genes is dependent upon the presence of BRACHYURY in mesenchymal-like CRC cells, in compliance with our data demonstrating the involvement of BRACHYURY in regulating markers of CSCs.
BRACHYURY also regulates NANOG in the CRC cell line T84 and the embryonal carcinoma cell line NTERA2/D1
We were interested to determine whether other BRACHYURY expressing cancer cell lines were similarly able to mediate BRACHYURY-dependent expression of NANOG. We therefore studied three cell lines previously shown to express BRACHYURY, the CRC cell line T8440 and the human EC cell line NTERA2/D1.40, 41 RNAi-induced knock down of BRACHYURY in these cell lines resulted in downregulation of NANOG, as seen in SW480 cells (Fig. 6). CD166 levels also decreased in response to BRACHYURY knockdown in T84 cells, again similar to the response of SW480 cells. NTERA2/D1 cells do not express the CRC-CSC marker CD166. Interestingly, we also observed the downregulation of the pluripotency marker OCT4 in NTERA2/D1 cells, suggesting that BRACHYURY may have an effect on the pluripotency of these EC cells (data not shown). These results lead us to suggest that the regulation of NANOG by BRACHYURY may generally occur in cell types that are competent for mesoderm differentiation.
Little is known about the molecular mechanisms used by tumour cells to establish a population of cancer initiating cells that are now widely accepted to exist. We present novel data to demonstrate that in some tumour cell types, BRACHYURY plays an important role in inducing NANOG in a population of cells that display markers specific for CSCs. Furthermore, we show that the expression of BRACHYURY is reversible dependent upon cell growth conditions and is dependent upon signalling from the oncogene product, β-catenin, which can result in CRC under aberrant cellular conditions, but which is also important in maintaining a stem cell program and defining CSCs in the colon.31
BRACHYURY was originally shown to have a central role in mouse mesoderm development but expression has also been associated with disease such as chordomas and in the EC cell line NTERA2/D1, where it is expressed in the absence of mesodermal differentiation.48, 49 Recently, BRACHYURY was described as a potential candidate for T-cell mediated cancer immunotherapy.40 However, to date only limited studies have been performed to link the presence of BRACHYURY with cancer. The importance of BRACHYURY in cancer progression has recently been highlighted in studies by Fernando et al.41 who demonstrated that BRACHYURY is overexpressed in various human tumour tissues and cancer cell lines, often in a tumour stage-specific manner, and associated with later stage aggressive tumours of the lung. Overexpression studies with exogenous BRACHYURY in PANC-1 cells demonstrated that BRACHYURY induces a mesenchymal-like, migratory phenotype and is important in establishment of lung metastasis in vivo.
Whilst we observed modest, but reproducible decreases in expression of EMT regulatory genes (Fig. 1a) and a modest increase in levels of E-CADHERIN after RNAi induced knockdown of endogenous BRACHYURY in SW480 cells (in agreement with RNAi induced knockdown studies by Fernando and co-workers in lung and prostate cell lines), we were not able to identify any other overt phenotypic changes marking EMT in CRC cells after knockdown of endogenous BRACHYURY (such as relocalization of β-catenin, changes in expression and localization of E-CADHERIN) (data not shown). We note that the levels of BRACHYURY in the overexpression studies of Fernando and co-workers were increased more than 35 times the normal levels and suggest that the difference in ability to induce a mesenchymal phenotype may be due to differences in the epithelial cancer cell lines used but may also reflect a change in the balance of regulatory pathway components that may occur due to ectopic overexpression of BRACHYURY. In compliance with knockdown studies by Fernando and co-workers, we conclude that BRACHYURY alone is not sufficient to induce EMT in CRC cells.
Low-density, mesenchymal-like SW480 cells differentially express BRACHYURY; expression changes are reversible on passaging and occur in low and high passage number cells indicating that this is not a randomly acquired, fixed genetic change that occurs only in high passage number cells.
As mentioned, β-catenin modulates signalling pathways regulating ‘stemness’, and has recently been shown to maintain a discrete population of CSCs in epidermal tumours29, 30 as well as defining CSCs in the colon.31 Our results demonstrate a possible mechanism whereby β-catenin signals through BRACHYURY to maintain a population of cells with stem-like character, as determined by expression of NANOG and markers of CSCs.
Whilst NANOG is not absolutely required for self-renewal of ESC, it is found to be essential for development of primordial germ cells, where it is postulated to have a role in resetting the epigenetic state.50 It is possible that the presence of NANOG in low-density SW480 cells serves to deter differentiation. Changes in the media or growth conditions over time then lead to changes in expression profiles that favour differentiation with a corresponding decrease in levels of NANOG. Furthermore, NANOG may play a role in altering the epigenetic state of low-density SW480 cells to either protect an ‘immature cell state’ (through high expression) or alternatively to allow differentiation (through low expression). If expressed specifically in invading cancer cells, this may endow such cells the ability to alter their epigenetic state/phenotype to respond to their new environment.
Interestingly, we were able to show that all the BRACHYURY+ cell lines studied here express the NANOG2 pseudogene (NTERA2/D1 and T84 also express NANOG2), the protein product of which is functionally identical and indistinguishable from NANOG1. Surprisingly, none of these cell lines expressed NANOGP8, which has been described as being the source of the predominant form of NANOG in many cancer cell lines,16 raising the intriguing possibility that BRACHYURY may be involved in specifically regulating expression from the NANOG2 locus.
Although initial tumour-generating events in CRC may be of a defined nature, the phenotype of the cells alters with progression of the tumour, irrespective of genetic mutations, implying that epigenetic changes are involved, or alternatively, that the cells respond to extrinsic cues.37 Accordingly, our work shows that SW480 cancer cells can undergo transition reversibly between two states, one state bearing characteristics/markers of CSCs and other bearing epithelial-like character. The expression of NANOG2 in SW480 cells might therefore permit a cell to express stem cell markers (such as CD166) and even to have stem cell character. It is interesting to note that we were also able to detect expression of other genes associated with the undifferentiated state such as OCT4, SOX2 and FOXD3 in SW480 cells at low density. The close relationship between cells with properties of CSCs, EMT, invasion and the microenvironment has recently been demonstrated in other models.33 Our work complies with this and when considered in combination with the work of Fernando and co-workers41 suggests that BRACHYURY may be an important factor contributing to the ability of cells to acquire CSC character and undergo EMT.
It is likely that not all tumours/cell lines express BRACHYURY and those tumours that do not express BRACHYURY may have evolved alternative routes to activate functional forms of NANOG to allow establishment of cells with CSC characteristics.
Alternatively, tumours not expressing BRACHYURY and NANOG might be less aggressive; BRACHYURY has been associated with late stage, aggressive lung tumours.41 It would be interesting to determine the generality of the presence and localisation pattern of NANOG and BRACHYURY within a larger panel of colorectal tumours and tissue and to correlate such data with the presence of CSCs and patient outcomes.
We gratefully acknowledge support from our funding bodies: Debalina Sarkar and Melanie Davies were supported by grants from the Tenovus Cancer Charity and GH and CRF; Brian Shields is supported by Cancer Research Wales; Jurgen Muller received support from the NWCRF; Jane Wakeman is supported by the NWCRF.