Author contributions: E.E.I. and R.B.-J.: conception and design, collection and/or assembly of data, data analysis and interpretation, and final approval of manuscript; A.S. and B.S.-D.: collection and/or assembly of data, data analysis and interpretation, and final approval of manuscript; S.H., M.A., N.L., and W.F.: collection and/or assembly of data and final approval of manuscript; M.I. and D.B.: provision of study material or patients and final approval of manuscript; A.S.N.: conception and design, financial support, provision of study material or patients, manuscript writing, and final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS July 31, 2012
Embryonic NANOG (NANOG1) is considered as an important regulator of pluripotency while NANOGP8 (NANOG-pseudogene) plays a role in tumorigenesis. Herein, we show NANOG is expressed from both NANOG1 and NANOGP8 in human colorectal cancers (CRC). Enforced NANOG1-expression increases clonogenic potential and tumor formation in xenograft models, although it is expressed only in a small subpopulation of tumor cells and is colocalized with endogenous nuclear β-cateninHigh. Moreover, single NANOG1-CRCs form spherical aggregates, similar to the embryoid body of embryonic stem cells (ESCs), and express higher levels of stem-like Wnt-associated target genes. Furthermore, we show that NANOG1-expression is positively regulated by c-JUN and β-catenin/TCF4. Ectopic expression of c-Jun in murine ApcMin/+-ESCs results in the development of larger xenograft tumors with higher cell density compared to controls. Chromatin immunoprecipitation assays demonstrate that c-JUN binds to the NANOG1-promoter via the octamer M1 DNA element. Collectively, our data suggest that β-Catenin/TCF4 and c-JUN together drive a subpopulation of CRC tumor cells that adopt a stem-like phenotype via the NANOG1-promoter. STEM Cells2012;30:2076–2087
Human colorectal cancer (CRC) cells are thought to have a hierarchical architecture whereby decisions of self-renewal or differentiation arise from integration and reciprocal titration of numerous regulatory networks and important signaling pathways including the Wnt-signaling pathway. In fact, most of CRCs contain a constitutive active form of the canonical Wnt/β-catenin-signaling either with or without mutation(s) of Wnt-components (e.g., Adenomatous polyposis coli [APC] and β-catenin). However, β-catenin shows a heterogeneous pattern of protein expression level and intracellular distribution . In principle, Wnt-signaling promotes self-renewal, stimulates growth, and mediates developmental and carcinogenic intercellular signaling [2–4]. Furthermore, several recent studies have shown that high levels of nuclear β-catenin protein (β-cateninHigh) are associated with normal intestinal stem cells and a small subpopulation of cells within CRCs which may represent cancer stem cells (CSCs) [4–6].
The homeodomain NANOG protein is a transcription factor required for maintaining the undifferentiated state and self-renewal of embryonic stem cells (ESCs) [7, 8]. It is encoded by the human NANOG1 gene (gi 79923) that maps to chromosome 12 and comprises four exons and three introns. There are 10 NANOG-pseudogenes hitherto reported in the human genome but only NANOGP8 has an intact open reading frame (ORF). The NANOGP8 gene (gi 388112) is composed of a single exon and located on chromosome 15  and, although structurally different from NANOG1, it codes for a protein that differs from the NANOG1-encoded protein by one amino acid (at position c.253). The biological activity of both protein products is indistinguishable and both are therefore known as NANOG protein.
In addition to ESCs, NANOG1 has been recently detected in some germ cell tumors, CRC, and leukemia cells [10–12]. In contrast, NANOGP8 has been detected in many types of carcinomas [12–17]. In ESCs, NANOG1 forms a regulatory circuitry with Sox2, Oct4, β-catenin, Tcf3, Stat3, Wnt11, and Myc, controlling self-expression, and the expression of pluripotency-related genes . The threshold expression of NANOG which is required for sustaining ES properties relies on synergy between the Stat3, Wnt/β-catenin, Hedgehog (HH), and Smad signaling pathways under the specific conditions of the stem cells microenvironment [19–21]. However, to date, comprehensive expression patterns of NANOG1 and NANOGP8 in human cancers have not been fully addressed. Recent reports have confirmed that these signals can also contribute to CSC survival and tumorigenesis [4, 16, 19, 22]. This prompted us to further examine NANOG1 function in CRC cells.
Herein, we provide evidence that NANOG1 gene expression is a marker for a small (0.5%–2%) subpopulation of CRC cells which, together with high Wnt/β-catenin activity, functionally defines the CRC stem cell population. Chromatin immunoprecipitation (ChIP) assays revealed that both c-Jun and TCF4 bind on the NANOG1 promoter and c-JUN/TCF4 was shown to cooperatively activate the NANOG-1 promoter in a β-catenin-dependent manner. Unlike ESCs, the absence of Sox2 and/or OCT4 response elements caused no repression of NANOG1-promoter activity. The β-catenin/TCF4/c-Jun complex, via NANOG1-induction, controls a subpopulation of CRC cells driving clonogenic units, spherical embryoid body (EB)-like aggregates, and is associated with high expression of Wnt-target genes. Therefore, we propose that this mechanism also underpins the embryonic stem cell phenotype and that colon CSCs can be identified by NANOG1 activity levels.
MATERIALS AND METHODS
Full details of Materials and Methods provided as Supporting Information Material, see STEM CELLS website.
NANOG Protein Is Expressed Heterogeneously in Colorectal Tumors
There is some controversy surrounding the pattern of NANOG expression, its cellular distribution, and its correlation with clinicopathological variables in CRC [15, 23]. We evaluated NANOG expression in a series of primary CRCs from 49 patients. Cores from formalin-fixed paraffin-embedded tumor tissue were arrayed (in triplicate) onto a tissue microarray (TMA) and tested using immunohistochemistry (IHC). Our results showed a heterogeneous nuclear NANOG expression pattern with varying numbers of positive tumor nuclei in the TMA tumors (Fig. 1, red arrowheads). Furthermore, human NANOG cDNA was cloned into a FLAG-tagged mammalian expression vector to confirm our IHC data by immunofluorescence. This demonstrated that the anti-NANOG antibody used in this study could detect both endogenous and exogenous NANOG nuclear protein (Supporting Information Fig. S1A). The specificity of the NANOG-antibody was further confirmed by NANOG-siRNA transfected in mouse ESCs (Supporting Information Fig. S1D). Our analysis (Table 1) indicated that the increased NANOG expression is statistically associated with more advanced stage of the tumor (p =.019) and not with the age of patients (p =.698), gender (p =.754), tumor metastasis to the liver (p =.240), and lymph node metastasis (p =.14). Additionally, in a variety of CRC cell lines (HCT116, SW620, SW480, and DLD1), we observed NANOG protein expression in the nuclei of a subpopulation of cells (3% ± 2.7%) with no evidence of cytoplasmic NANOG expression (Supporting Information Figs. S1A, S1B, S5A and data not shown).
Table 1. Clinicopathological variables and correlation with NANOG expression
The bold value indicates statistical significance, *p ≤.05.
NANOG1 Proximal Promoter Is Sufficient to Recapitulate Endogenous Expression of the NANOG1 Transcript in CRC
Previous reports have showed that the ESC-like transcriptional program, including NANOG, is active in several human tumors and is correlated with increased risk of metastasis and mortality [24, 25]. However, the origin of the NANOG full-length transcript in cancer cells remained unclear. Therefore, we set out to define the origin of NANOG full-length mRNA in CRC cells. To assess NANOG mRNA expression, cDNAs were generated from different CRC cells (HCT116, SW620, and DLD1) and several pairs of primers were used to amplify both NANOG1 and NANOGP8, transcripts (Fig. 2A and Supporting Information Table S1). Our RT-PCR analysis using a forward primer (a) specific to a region upstream of the NANOG1 transcription initiation site revealed a precursor NANOG1 transcript (a + b) (Fig. 2A, lower left panel) compared with oligos in the coding region (b + c), which amplified both NANOG1 and NANOGP8 (Fig. 2A, lower panels). Full-length (918 nt) NANOG cDNA was seen in all tested CRC cell lines; HCT116, SW620, and DLD-1 (Fig. 2B, right, first panel). A human ESC (hESC) was used as a positive control.
As outlined above, it is very difficult to distinguish between NANOGP8 and NANOG1 as their amino acids sequence is almost identical with only one change from “Gln-253” in NANOG1 to “His-253” in NANOGP8  (Fig. 2A). Our subsequent sequencing analysis showed the presence of both NANOGP8 and NANOG1 ORF in the above samples (Fig. 2B, right, second panel). Moreover, sequencing of rescued plasmids showed cloning of NANOG1 in 7.5% (3/40) compared to 92.5% of NANOGP8 (37/40) of bacterial colonies after transformation (data not shown).
To complement these experiments, we also investigated the expression pattern of NANOG mRNA in 12 CRC tumor tissues, hESCs, and fibroblast cells. The G → A transversion creates a unique AlwN1 site in NANOGP8 cDNA (ACGGAGACTGTC → ACAGAGACTGTC) . Our analysis through a standard RT-PCR and followed by the AlwN1 restriction enzyme digestion showed that NANOGP8 is detectable in all CRCs tested but NANOG1 mRNA is detectable in some not all CRCs (Supporting Information Fig. S2D). Furthermore, our real-time RT-PCR study also shows that both NANOG mRNAs are heterogeneously expressed in all CRCs cells and tumor tissues (Fig. 2C). NANOG1 is an important gene that was thought to be exclusively expressed in pluripotent stem cells and is involved in stem cell proliferation and self-renewal. It is possible that the NANOG1-expressing subpopulation CRC cells may represent the CSCs and this prompted us to further evaluate mechanisms of NANOG1 gene expression in CRC cells.
Recent data have suggested that the NANOG1 proximal promoter is sufficient to recapitulate endogenous expression of NANOG1 mRNA in human and mouse ESCs [26, 27]. To determine whether endogenous expression of NANOG1 mRNA could also be functionally driven from NANOG1-promoter in CRC cells, we developed a variety of reporter constructs with both mouse and/or human promoter sequences (mouse, −2,828 + 190 bp and −332 + 50 bp; human, −380 + 40 bp; relative to the transcription start site) driving the expression of eGFP (Fig. 2D, top panel). Consistent with experiments in mouse ESCs (Supporting Information Fig. S2B), both mouse (−332 + 50 bp) and human (−380 + 40 bp) NANOG1 promoters resulted in green fluorescent protein (GFP) expression in 0.5%–2% cells in several CRC cell lines (HCT116, SW620, SW480, and DLD1) (Figs. 2D, 5A and data not shown), whereas a backbone vector, pC2-eGFP, as a control construct, was highly expressed (35% ± 10%) (data not shown). Furthermore, fluorescence-activated cell sorting (FACS) analysis also revealed a small subpopulation of NANOG1-promoter-GFP expressing cells (0.2%–3%) (Fig. 6B, 6F). Most importantly, NANOG1-promoter-GFP expressing cells (Fig. 2D, green arrowheads) were positive for endogenous NANOG (Fig. 2D, yellow arrowheads). These data indicate that this region of the NANOG1-promoter is transcriptionally active and is sufficient to activate the endogenous expression of NANOG1 mRNA. Second, those regulatory DNA elements located in a ≈350-bp region upstream of NANOG are essential for NANOG1 promoter activity in the CRC cells.
NANOG Expression in CRC Cells Requires cis-Regulatory Octamer M1-DNA Element
In order to identify the important cis-regulatory elements located approximately in the ≈350-bp region upstream of the NANOG1 gene, we mutated specific DNA transcription factor binding sites that are crucial for NANOG expression in ESCs (Fig. 3). The mouse or human NANOG1-minimal promoter was fused to firefly-Luciferase gene (Fig. 3A). Consistent with the data obtained using NANOG1-GFP construct, the wild-type (wt) fragment induced Luciferase expression in all CRC cell lines (HCT116, SW480, and DLD1), human embryonic kidney (HEK293), and mouse ESCs (Fig. 3B and Supporting Information Fig. S2C, data not shown). In contrast, the control vector (pGL3-Basic) gave no activity. Unexpectedly, mutation of Sox2/Oct4-binding elements either individually (Octmut or Soxmut) or together (Octmut/Soxmut) did not affect NANOG1-promoter activity in the CRC cells (Fig. 3B). In contrast, these mutations significantly decreased the NANOG1-promoter activity in embryonic HEK293 cells compared to the wild-type promoter (Fig. 3B).
The mouse and human NANOG1-promoter contains a third potential cis-regulatory element known as M1 (comprising the sequence: GTCTGGGT), which is required for the transcriptional regulation of NANOG1 gene in ESCs and Glioblastoma stem cells [19, 27]. A three bases substitutions of the M1 element (GTCTGGGT to GTAGTGGT) (Fig. 3A) led to a significant reduction of Luciferase activity in both CRC and HEK293 cells (p ≤.05) (Fig. 3B). Furthermore, mutation of the M1 element of the NANOG1-promoter-GFP construct (M1mut) resulted in loss of GFP expression in CRC cells (data not shown). In addition, the overexpression of HH-downstream transcriptional regulators Gli1 and Gli2, which have been suggested to regulate NANOG expression in Glioblastoma stem cells , did not activate the NANOG promoter in CRC cells (Supporting Information Fig. S3A and data not shown). Hence, our data suggested that efficient endogenous NANOG1 mRNA expression in CRC cells may require previously undefined factors that bind to the M1 DNA-binding element.
β-Catenin/TCF4/c-Jun-Mediated Transcriptional Regulation of NANOG1
Previous computational analysis of the 5′upstream region sequences flanking mouse and human NANOG promoter has predicted several Tcf transcription factor binding sites [28, 29]. The essential M1 regulatory DNA-element is located in a ≈350-bp region upstream of both human and mouse NANOG1 and contains a cognate sequence (AGTCTGGGT) corresponding to that of the AP1 core response element-binding protein (TA/GAGTCT/AG/C) . We therefore hypothesized that mechanistically AP1 transcription factors together with β-catenin/TCF4 act on the NANOG1 promoter in CRC cells.
First, we examined the biological function of c-Jun, an AP1 transcription factor, in the regulation of the NANOG1 promoter. C-Jun, in association with β-catenin/TCF4, has an important function in regulating intestinal and CRC stem cells [31, 32]. Our data showed that the transfection of CRC cell lines (HCT116 and SW620) using a constitutively active CMV-promoter to drive c-Jun expression caused a significant induction in the endogenous NANOG1-promoter (wt) (Fig. 3C and data not shown). We next investigated the roles of β-catenin/TCF4 in the transcriptional regulation of NANOG1. The β-catenin and TCF4 expression constructs (FLAG-β-cateninS33Y and HA-hTCF4) were cotransfected into HCT116 and SW620 cells along with the various NANOG1-Luc constructs (wt, Octmut, Soxmut, and Octmut/Soxmut). Strikingly, all NANOG1-repoter constructs showed similar levels of Luciferase activity in the CRC cells when both β-cateninS33Y and TCF4 were overexpressed (Fig. 3C and Supporting Information Fig. S3B left panels and S3C, data not shown). However, the highest level of Luciferase activity was detected in CRC cells when the mouse long NANOG1-Luc construct (−2,828 bp) was used (data not shown). Our data also showed that the expression of dominant-negative TCF4 (ΔN-TCF4) efficiently blocked the transcription of Luciferase gene, both in short (−332 bp) (Fig. 3C, lane six vs. lane nine) and long (−2,828 bp) (data not shown) NANOG1-reporters. Our data on the Luciferase activity in response to β-catenin/TCF4 overexpression showed enhanced c-Jun mediated transcription, whereas the M1mut remained inactive irrespective of any cotransfected expression constructs (Fig. 3C).
Quantification of mRNA by real-time PCR suggested that ectopic expression of β-catenin and TCF4 in both HCT116 and SW620 cells led to increased NANOG expression (Fig. 4A, left panel and data not shown). To complement these studies, we knocked down β-catenin with a small hairpin RNA (shRNA) construct  and TCF4 with TCF4-specific duplex siRNA (Supporting Information Fig. S3B, right panel), and NANOG mRNAs were compared with cells transfected with sequence scrambled controls siRNA (c-jun control). Knocking down of c-jun, β-catenin, and TCF4 decreased both endogenous NANOG and c-jun mRNA levels (Fig. 4A, left panel). To further investigate the significance of c-Jun and β-catenin/TCF4 complex in NANOG regulation, we examined endogenous protein levels by Western blotting using mouse ESCs lacking c-jun (mES c-jun−/−). Deletion of c-jun gene abolished expression of both Nanog and Tcf4 mRNA and proteins in mES c-jun−/− compared to c-jun+/+ cells (Fig. 4A, middle and right panels).
Our data led us to hypothesize that c-Jun may mediate a stable M1-DNA-protein complex with β-catenin/TCF4 (when bound to Tcf-DNA binding sites) to stimulate NANOG1 expression. To examine this, c-Jun binding to the M1-DNA element was tested using a modified ChIP assay  using a protocol given in details in Figure 4B, left panel. In this procedure, Dynabeads M-280 Streptavidin was used to capture biotinylated NANOG1-minimal-promoter-DNA-protein complex. Biotinylated NANOG1-promoter (nucleotides −300 to −59) was transfected in c-jun−/− or c-jun+/+ mouse embryonic fibroblast and DNA-binding proteins were fixed as previously described . Specific binding of c-Jun protein onto the DNA fragments was detected by immunoblotting (Fig. 4B, right panel). To support this, ChIP analysis  of protein bound to the M1 based on quantitative-PCR showed efficient binding of both c-Jun and TCF4 proteins to AP1 (nucleotides −300 to −59) and TCF (nucleotides −1,230 to −900) sites in the NANOG1 promoter (Fig. 4C, left panel). Notably, the proximal TCF binding sites were also detectable in c-Jun ChIP and vice versa (Fig. 4C, left panel). This may suggest that c-Jun and TCF4 can bind to both AP1 and TCF sides via interaction to each other . Reporter gene activation by c-JUN and TCF4 overexpression (Fig. 3C) was greatly reduced by siRNA-mediated knockdown of β-catenin (Supporting Information Fig. S3C, lower panel), suggesting that c-JUN/TCF4-mediated transcriptional activation is dependent on β-catenin in CRC cells (Fig. 4C, right panel).
To elucidate the physiological relevance of the c-Jun/TCF4 interaction with NANOG1 promoter in tumors, we initially examined whether the elevated NANOG level in CRC cells was influenced by canonical β-catenin/c-Jun signaling in vivo. IHC of intestinal tissues from ApcMin/+ mice showed that mouse Nanog protein level is elevated in a subpopulation of epithelial cells (Supporting Information Fig. S4A, a and b). However, Nanog was hardly detectable in normal intestine of the wild-type mouse (Supporting Information Fig. S4A, panels c and d). This data further supported the hypothesis that activated Wnt-signaling (through APC mutation) may induce Nanog expression in the epithelia of ApcMin/+ mouse model. Stable downregulation of Nanog has been attempted in both murine ESCs and CRC cells using NANOG shRNA lentiviral vector plasmid based on puromycin selection (Santa-Cruz, California, http://www.scbt.com/; sc-44833-SH) without success probably because of loss of cell renewal in the clones lacking NANOG expression or cell differentiation as many other previous reports also suggested [34, 35]. This necessitated other methods to investigate NANOG regulation by Wnt/c-Jun in mouse ESCs. Therefore, we generated mouse ESCs harboring the Apc mutation (ApcMin) to act as a model replicating the molecular mechanism(s) of both activated Wnt/c-Jun signaling and Nanog1 simultaneously.
An in vivo teratoma analysis was performed by injecting undifferentiated ApcMin/+-ESCs with constitutive stable expression of wild-type c-Jun (ApcMin/c-Junwt) (Supporting Information Fig. S4B, left panel) and the transcriptionally inactive form of c-Jun which is lacking the transactivation domain (ApcMin/c-JunΔTA) (Supporting Information Fig. S4B, left panel) subcutaneously into nonobese diabetic/severe combined immunodeficient (ND/SCID) mice to induce teratomas formation [36, 37]. Morphometric analysis of teratomas was achieved by counting tumor cells in an area of 100 μm, (n = 10, ×20 objective) using AxioVision digital image processing software. Quantitative analysis showed that ApcMin/c-Junwt had approximately 1.7-fold higher cell number/density on average than that of ApcMin/+-ESCs expressing GFP (Supporting Information Fig. S4B, right panel), whereas the level of cell death/apoptosis was comparable between ApcMin/c-Junwt and ApcMin/GFP (data not shown). ApcMin/c-JunΔTA mouse ESCs showed similar induction of cell growth to ApcMin/GFP cells (Supporting Information Fig. S4B, middle panel). These data demonstrate that high levels of all components of activated Wnt-signaling, that is, nuclear β-catenin and TCF4 along with c-Jun are required for optimal expression of NANOG1 in CRC cells. Thus, the β-catenin/TCF4 complex may supply the endogenous AP1 transcription factor c-Jun to induce NANOG1-promoter activity in CRC cells. Taken together, our data support our hypothesis that c-Jun and TCF4/β-catenin interact on the NANOG1 promoter in vivo and in a β-catenin-dependent manner (Fig. 4C, right panel).
NANOG-Expressing Cells Form Spheres and Clonogenic Tumor Cells
Despite the presence of mutation(s) in canonical β-catenin-associated genes (APC or CTNNB1, etc.), it has been shown previously that nuclear β-catenin is highly expressed only in a minority of CRC cells and could be used to define the CSCs [1, 4]. Therefore, we conducted colocalization studies between the β-catenin and NANOG and we also used MUC2 as an epithelial differential marker, in different CRC cells (HCT116 and SW620). Our data demonstrated that only CRC cells with the highest nuclear β-catenin level are colocalized with NANOG (Supporting Information Fig. S5A), while cytoplasmic MUC2 was significantly decreased in these cells (Supporting Information Fig. S5B). Furthermore, we examined whether there is a positive correlation between phosphorylated c-JUN and c-JUN expression levels and NANOG-expressing cells. Immunofluorescent assays detected that the highest expressing cells for both c-JUN and phospho-c-JUN cells were positive for NANOG1 (using NANOG1-GFP construct) (Fig. 5A). HCT116 and SW620 cells are both well-described CRC cell lines which also express NANOG full-length at both the mRNA and protein levels (Fig. 2). Thus, these cell lines were chosen to test further the differentiation capacity of cells expressing NANOG1 using EBs and teratoma xenografts assays, as previously described .
To corroborate these findings, HCT116 cells stably expressing either GFP (HCT116GFP) or NANOG-GFP (HCT116NANOG/GFP) (Supporting Information Fig. S6A) were injected subcutaneously into the flanks of 15 ND/SCID mice, with different cell numbers (2 × 105, 2 × 103, and 500) per injection. All xenograft tumors were analyzed histologically and biochemically (Supporting Information Fig. S6B). No variation in the incidence of xenograft tumors was observed on injecting 2 × 105 cells and tumor arose in all mice. However, injecting 2 × 103 GFP-expressing cells resulted in a delay in tumor formation (tumors formed in only two mice), while the 500 cells failed to form any tumors. In contrast, all mice injected with 500 NANOG-GFP-expressing cells formed tumors (Fig. 5B). Histological analysis of the 2 × 105-xenograft tumors stained with hematoxylin and eosin showed no obvious histological differences between GFP and NANOG tumors (Fig. 5C, a,b vs. c,d). However, immunohistochemical staining of tumors for proliferation marker, Ki67, indicated more cycling cells in the NANOG tumors (Fig. 5C, e,f vs. g,h and Supporting Information Fig. S6C), whereas E-cadherin levels were lower in HCT116NANOG/GFP than that in HCT116GFP tumor cells (Fig. 5C, i,j vs. k,l). Intriguingly, costained NANOG/β-CateninHigh cells indicated that the positively labeled NANOG cells are allocated into a colony unit (Fig. 5D, c and d) and colocalized with high nuclear β-catenin (Fig. 5D, e and f) and low MUC2 expression (Supporting Information Fig. S5C). Also, NANOG+ cells are determined by low 4′,6-diamidino-2-phenylindole (DAPI) stain intensity (Fig. 5D, a and b). Interestingly, the maximum number of clonogenic NANOG+ cells in HCT116GFP xenograft was three (Fig. 5D, top panel), whereas the number of cells increased to up to 19–20 cells per colony in HCT116NANOG/GFP tumor (Fig. 5D and Supporting Information Fig. S6D). To demonstrate that the cell cycle time is different between NANOG+ clonogenic and nonclonogenic tumor cells in vitro, in a separate tissue culture studies, we observed that ectopic expression of NANOG increased cell number at G2/M transition (Fig. 6A, top panel and Supporting Information Fig. S6E) and also HCT116NANOG/GFP cells were more efficient in an in vitro wound healing assay (Fig. 6A, low panel). This provided further support for the hypothesis that NANOG+ clonogenic (NANOG+) and nonclonogenic (NANOG-) tumor cells have different biological behaviors, and that potentially these differences also exist between CSCs and non-stem cells .
To further evaluate whether the clonogenic phenotype of HCT116NANOG/GFP and HCT116GFP xenografts were biologically reflective of endogenous NANOG1 expression to CSC, colony forming was assessed by the single-cell-derived cultures for both SW620 and HCT116, using FACS and single-cell plating for NANOG1-promoter-GFP expressing cell (Fig. 6B). NANOG1-promoter-GFP expressing cells aggregated and formed foci with the appearance of the sphere-like structures of EB (Fig. 6C). In agreement with our xenograft tissue analysis, ectopic expression of NANOG protein also shifted CRC cells into the formation of sphere body-like structures that were not seen in pC2-eGFP controls (Fig. 6D, left panel vs. right panels). Only typical clump-like cell aggregates were formed by cells ectopically expressing GFP alone (Fig. 6D).
CRC NANOG1-eGFP+ Cells Exhibit High Levels of Stem-Cell-Associated Wnt-Target Genes
To further explore the important links between NANOG1 transcription and activated Wnt-signaling in NANOG1-promoter-GFP expressing CRC cells (and therefore NANOG1), we assessed Wnt-target genes expression by RT-PCR. HCT116 cells were transfected with the dual CMV-promoter-RFP:NANOG1-promoter-GFP reporter construct (Fig. 6E) and then followed by FACS (Fig. 6F) analysis. The CMV-promoter directs the dsRFP (dsRed-Fluorescent Protein) reporter to monitor the transfection efficiency and the GFP reporter indicates the NANOG1-promoter activity (Fig. 6E, low panels). Once again, the dual CMV-promoter-RFP:NANOG1-promoter-GFP reporter construct have further confirmed that the proximal NANOG1-promoter is indeed transcriptionally active albeit only in a small subpopulation of the transfected cells (0.5%–2%) (Fig. 6F and Supporting Information Fig. S6G, S6H). The CMV-RFP+:NANOG1-GFP+ cells were clearly shift to the right above of CMV-RFP+:GFP− cells (Fig. 6F). To exclude variation of endogenous NANOG expression, total CMV-derived RFP+ cells were initially selected, then NANOG1-promoter-GFP+ cells were sorted from the CMV-RFP+ population (Fig. 6F).
Our RT-PCR (Supporting Information Table S1) analysis showed that stem-cell Wnt-associated genes, including SOX2, c-MYC, and β-catenin (CTNNB1), and the intestinal stem cell marker, LGR5 , were predominantly upregulated, whereas MUC2 (Mucin2) associated with epithelial differentiation and DNMT-3β (DNA-(cytosine-5-)-methyltransferase-3-β)  associated with an epigenetic methylation modification were clearly repressed in NANOG1-promoter-GFP expressing cells (Fig. 6G). Furthermore, we analyzed the DNA methylation status of cytosine nucleotides using bisulfite sequencing in the NANOG1-promoter [41, 42]. The result revealed that the NANOG1-promoter was less methylated in CMV-RFP+:NANOG1-GFP+ cells, whereas these nucleotides were efficiently methylated in CMV-RFP+:GFP− cells (Fig. 6H and Supporting Information Fig. S6F).
Here, through a detailed analysis of human NANOG transcription and its activity, we have found that the embryonic NANOG1 gene is expressed in a subpopulation of cells in CRC cancer cell lines which comprises 0.5%–2% of the total cell population. NANOG1-reporter constructs provided evidence that nuclear NANOG1 activity could functionally and mechanistically identify the colorectal CSC population in vivo and in vitro. Our data derived from primary tumors clearly showed more stained tumor cells using an anti-NANOG antibody (Fig. 1) than the 0.5%–2% cells observed using variety of NANOG1-promoter reporter constructs (Figs. 2 and 6). Our quantitative data show that the majority of the expressed NANOG mRNA comes from NANOGP8 (and possibly other pseudogenes) and it is interesting to note that Ishiguro et al. have similarly  reported a differential expression of NANOG1 and NANOGP8 mRNA in colon cancer cells. The differential expression is probably true also for protein expression but the high degree of sequence homology between the two protein products makes comparison difficult. An antibody that recognizes only the NANOG1-specific gene product may potentially also be used to identify CSCs in CRC. Beyond the objectives of our study, we noticed there were no significant differences in the number of GFP-positive cells and/or luciferase activity, derived from NANOG1-promoter, respectively, between SW620 and HCT116 cells. However, HCT116 is a colorectal carcinoma cell line and SW620 is from a lymph node metastasis and we have chosen both of them for most of our studies.
Our data suggested that β-catenin/TCF4:c-Jun factors are required as coregulators for NANOG1 transcription activity through the AP1 binding M1 element and TCF DNA-binding domains present in the NANOG1-promoter in CRC cells (Figs. 3 and 4). The M1 domain comprises a conserved sequence important for NANOG1 expression [26, 27] while there is no M1-DNA binding element within 2,000 bp upstream of NANOGP8 (5′UTR through to −2,000 bp relative to AUG) (data not shown). This sequence was initially identified distal downstream from the site of transcription initiation for the promoters of, for example, ornithine decarboxylase , adenosine deaminase , and hypoxanthine-phosphoribosyl transferase . Tan et al. recently found that M1 element is also present and functionally involved within the pluripotent factor, undifferentiated transcription factor 1 3′ enhancer . Notably, more recently, Po et al. demonstrated that downstream effectors in the HH pathway, Gli1 and Gli2, directly bind to M1-domain and thus activate NANOG expression in medulloblastoma stem cells [16, 19]. Earlier studies demonstrating possible proteins bound to the M1 elements have shown a high affinity to AP1 factors . To pursue this further in CRC cells, we repeated our NANOG1-promoter assays with and without the altered M1 element (M1mut) along with overexpression of Gli1 and Gli2 constructs using both human and mouse NANOG1 reporter systems. While the M1 element was shown to be essential for NANOG1 transcription, co-overexpression of both proteins (Gli-1 or Gli-2) did not affect the NANOG1-promoter activities in CRC cells (SW620 and HCT116) (Supporting Information Fig. S3A). However, the role of the HH pathway in CRC cells has remained controversial with recent data indicating both synergistic and opposing interactions between the HH and Wnt pathways in colonic epithelial cell proliferation [47–49].
Here, we showed that c-Jun induced NANOG1 expression and the transcriptional cooperation between TCF4 and c-Jun is mediated by activated β-catenin and is dependent on the presence of both the TCF and M1 (AP1) NANOG1 promoter elements (Fig. 4). This suggests that the previously reported c-Jun-TCF4 interaction  may stimulate transcription by recruiting β-catenin to the proximal TCF and AP1 elements close to the transcriptional start site (Fig. 4C, right panel). Despite low level of Wnt-activity, HEK293 cells contain constitutively high JNK (c-jun N-terminal kinase) activity that induces c-Jun phosphorylation and subsequently augments the AP1 transactivation activity through c-jun autoregulation mechanism [32, 50]. Conversely, in CRC cells, c-jun is upregulated through the nuclear β-catenin/TCF4 transcription machinery, therefore M1 (AP1) transcriptional response activity is more dependent on β-catenin activity [31, 51].
OCT4/SOX2 cofactors are key DNA-binding partners for mediating NANOG activity in ESCs. However, they appear not to regulate the NANOG1-promoter in CRC cells. SOX2 expression was increased in NANOG1+ cells but was not correlated with OCT4 (Fig. 6G). Furthermore, thorough analysis may be required to address this; however, these conflicting results may be due to recently reported atypical expression of OCT4 gene in CRC cells [52, 53].
Because Wnt-signaling components and nuclear β-catenin are the major regulators of colon CSCs, our finding that β-catenin/TCF4:c-JUN complex is a key nuclear mediator for NANOG1 activity provides a novel molecular insight into the mechanism by which Wnt-signaling activity dictates CRC cell determination (Fig. 6I). As outlined above, this may suggest that all cancer cells with APC and β-catenin mutations should coexpress NANOG1 but staining on ApcMin/+ tissues and of CRC cells revealed that it was not the case, and indeed samples harboring the above mutations contain neither homogenous nuclear β-catenin nor NANOG expression. Therefore, APC or β-catenin mutated cells contain heterogeneous nuclear β-catenin and the NANOG+/β-cateninHigh-CRC population cells with the highest level of Wnt-activity may define colon CSCs. Indeed, the NANOG+/β-cateninHigh-CSCs population identity was further supported by formation of EB-like spheres and enhanced clonogenicity of NANOG+ cells as well as the low intensity of stained DAPI DNA-staining. Further analysis, beyond the current scope of this article, would be required to study why all the cells were not positive for NANOG staining both in xenograft and amplified colonies derived from HCT116 and SW620 cells ectopically expressing NANOG1 (Fig. 5D and data not shown). The low intensity of DAPI may reflect a quiescent status and suggests slow-cycling CSCs [54, 55].
Additionally, a variety of pathways have been implicated in modulating negatively and/or positively nuclear β-catenin and NANOG expression, including TGFβ, FGF, Notch, BMP, and HH signaling pathways [21, 56, 57]. Many of these factors, including NANOG activity, can also be regulated by the CSCs microenvironment and/or perhaps in a cell type-specific manner. Therefore, the NANOG+ CSCs microenvironment may drive tumor growth and even selectively support a subset of tumor cells. Understanding the interactions and functions of a NANOG+/β-cateninHigh-CSCs population within the context of the tumor microenvironment could be critical to design targeted therapeutics. Furthermore, our findings of NANOG1 activity and NANOG1-reporter system allow to experimentally select cells with many properties of epithelial CSCs that has several potentially important implications. The gene expression profiles of NANOG+ cells, for instance, can be used to identify proteins enriched in this population, facilitating purification, and detection of incipient cancer.
In conclusion, we have uncovered a transcriptional strategy with nuclear β-catenin mediating the actions of TCF4/c-Jun on NANOG1 activity; both for activation and repression of specific target genes that underlines NANOG1 control in cell-fate determination in colon carcinogenesis. We have revealed the unexpected expression of NANOG1 in a subpopulation of CRC cells with stem cell properties (Fig. 6I).
We are grateful to C. Denning, P. Robson, T. Tada, and H. Sasaki for providing essential reagents. We thank R. Muraleedharan, Tarek M. A. Elsaba, E. Nye, and C. Anbalagan for technical advice and reading the manuscript. E. Ibrahim thanks the Ministry of Higher Education of Egypt for generously funding a Ph.D. studentship. This article is dedicated to the memory of our colleague Sue Watson who passed away on 29 November 2011. This work was supported by the Medical Research Council Grant G0700763 to A.S.N.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.