|Home | About | Journals | Submit | Contact Us | Français|
Epithelial stem cells self-renew while maintaining multipotency, but the dependence of stem cell properties on maintenance of the epithelial phenotype is unclear. We previously showed that trophoblast stem (TS) cells lacking the protein kinase MAP3K4 maintain properties of both stemness and epithelial-mesenchymal transition (EMT). Here, we show that MAP3K4 controls the activity of the histone acetyltransferase CBP, and that acetylation of histones H2A and H2B by CBP is required to maintain the epithelial phenotype. Combined loss of MAP3K4/CBP activity represses expression of epithelial genes and causes TS cells to undergo EMT while maintaining their self-renewal and multipotency properties. The expression profile of MAP3K4 deficient TS cells defines an H2B acetylation regulated gene signature that closely overlaps with that of human breast cancer cells. Taken together, our data define an epigenetic switch that maintains the epithelial phenotype in TS cells and reveal previously unrecognized genes potentially contributing to breast cancer.
The transition of epithelial cells to motile mesenchymal cells occurs through a process known as epithelial-mesenchymal transition (EMT), in which epithelial cells lose cell-cell contacts and apical-basal polarity concomitantly with the acquisition of a mesenchymal morphology and invasive properties. Several molecular events are coordinated for the initiation and completion of EMT including loss of the adhesive cell-surface protein E-cadherin, activation of EMT-inducing transcription factors and reorganization of the actin cytoskeleton (Yang and Weinberg, 2008). During development, EMT is responsible for proper formation of the body plan and for differentiation of many tissues and organs. Examples of EMT in mammalian development include implantation, gastrulation and neural crest formation (Thiery et al., 2009). Initiation of placenta formation regulated by trophoectoderm differentiation is the first, and yet most poorly defined developmental EMT.
The commitment of stem cells to specialized cell types requires extensive reprogramming of gene expression, in part, involving epigenetic control of transcription. The first cell fate decision is the formation of the trophoectoderm and the inner cell mass of the blastocyst. Trophoblast stem (TS) cells within the trophoectoderm maintain a self-renewing state in the presence of FGF4 (Tanaka, 1998). For TS cells and most other tissue stem cells, self-renewal is defined as cell division with the maintenance of multipotency (He et al., 2009). Diminished exposure to FGF4 induces TS cells to give rise to multiple differentiated trophoblast lineages, each required for establishment of the placenta. For implantation to occur, TS cells must undergo morphological changes to a more invasive phenotype, thereby exhibiting the functional hallmarks of EMT.
An emerging topic in the EMT field is the intersection between EMT and stemness. We have previously characterized the developmental defects of a genetically engineered mouse with the targeted inactivation of MAP3K4, a serine-threonine kinase important for JNK and p38 activation in response to FGF4 (Abell et al., 2009). In addition to embryonic defects, the MAP3K4 kinase-inactive mouse (KI4) has trophoblast defects resulting in hyperinvasion, defective decidualization, fetal growth restriction and implantation defects (Abell et al., 2009; Abell et al., 2005). TS cells isolated from the conceptuses of KI4 mice (TSKI4 cells) exhibit the hallmarks of EMT, while maintaining TS cell multipotency and are a unique developmental stem cell model to examine parallels between EMT and stemness.
Recently, EMT has been linked to the metastatic progression of cancer and to the acquisition of stem cell properties (Mani et al., 2008). The claudin-low (CL) intrinsic subtype of breast cancer is characterized by its mesenchymal and stem cell-like properties. In concordance with the stem cell-like CD44+/CD24−/low and CD49f+/EpCAM− antigenic phenotypes of breast tumor initiating cells (TICs) and mammary stem cells, gene expression profiling demonstrated that CL tumors have reduced expression of several epithelial differentiation markers, while exhibiting increased expression of certain stemness and mesenchymal markers (Lim et al., 2009; Prat et al., 2010).
Herein, we define an epigenetic mechanism important for the initiation of the first EMT event during development. Using TSKI4 cells uniquely trapped in EMT prior to initiation of the trophoblast differentiation program, we capture the genetic and epigenetic profile of the intersection between the properties of EMT and stemness. Importantly, we identify a molecular mechanism reliant on the histone acetyltransferase CBP that is responsible for controlling epigenetic remodeling during EMT in TSKI4 cells. TS cells lacking CBP (TSshCBP) expression exhibit an EMT similar to TSKI4 cells, which is mediated by the selective loss of H2A and H2B acetylation. Comparison across developmental and cancer EMT models exhibiting stem-like properties demonstrated a highly significant intersection between the gene expression profiles of TSKI4 cells and CL human tumors and cell lines. Repressed genes from the EMT gene signature demonstrated loss of H2BK5Ac in TSKI4, TSshCBP and CL cells. These results highlight the importance of MAP3K4/CBP-mediated acetylation of H2BK5 for maintenance of the TS cell epithelial phenotype.
In the presence of FGF4, TS cells maintain self-renewal as defined by the maintenance of cell division with multipotency (Niwa et al., 2005; Tanaka, 1998). TSKI4 cells lack MAP3K4 activity but are self-renewing and multipotent. When cultured in the presence of FGF4, TSKI4 cells expressed the TS cell markers Cdx2, Eomes, Esrrβ and FGFR2 at levels similar to TSWT cells (Figures 1A–D). Removal of FGF4 promoted the differentiation of TSKI4 cells into all trophoblast lineages similar to TSWT cells with the loss of expression of stem cell markers (Figures 1A–D) and the gain of expression of trophoblast lineage markers including the giant cell marker PLI, the spongiotrophoblast marker Tpbpα, and the syncytiotrophoblast marker Gcm1 (Figures 1E–G). Developmental potency of these stem cells was established through the injection of GFP-labeled TSWT or TSKI4 cells into wild-type blastocysts. Both TSWT and TSKI4 cells produced chimeric conceptuses with similar frequencies (Figure 1H) and contributed selectively to the trophoblast lineages including the extraembryonic ectoderm, ectoplacental cone, and giant cells (Figures 1I–N and Figures S1A–H). These cells were not found in the embryo or in any ICM-derived extraembryonic tissue (Figures 1I–N).
TSKI4 cells also exhibit the molecular and cellular characteristics of EMT. TSWT cells grew with an epithelial morphology (Figure 1O), with strong peripheral E-cadherin (Figure 1P) and cortical actin expression at cell-cell contacts (Figures 1Q and 1R). In contrast, TSKI4 cells exhibited a polarized front-back end, mesenchymal morphology (Figure 1S). E-cadherin was significantly reduced in TSKI4 cells (Figures 1T). Also, filamentous actin was observed at the leading edge of TSKI4 cells (Figures 1U and 1V). Reduction of total E-cadherin in TSKI4 cells relative to TSWT cells was further validated by Western blotting (Figure 1W). Real-time quantitative RT-PCR (qRT-PCR) showed a modest increase in the expression of the mesenchymal markers vimentin and N-cadherin in TSKI4 cells, and Western blotting revealed a 2.9-fold and 2.2-fold increase in vimentin and N-cadherin protein expression respectively (Figures S1I and S1J). Differentiation of both TSWT and TSKI4 cells also increased vimentin and N-cadherin protein expression (Figure S1J). TSKI4 cells and 4-day differentiated TSWT cells exhibited a fourteen to sixteen-fold increase in invasiveness compared to TSWT cells (Figure 1X). Hyperinvasiveness of the TSKI4 cells was also seen in vivo. Defective decidualization is induced by excessive trophoblast invasion (Norwitz et al., 2001). Compared to injection of TSWT cells (Figure 1Y), injection of TSKI4 cells into wild-type blastocysts resulted in defective decidualization consistent with the hyperinvasiveness of TSKI4 cells (Figures 1Z-BB). Cumulatively, these data demonstrate that TSKI4 cells are self-renewing stem cells with properties of EMT including loss of E-cadherin and gain of invasiveness.
Removal of FGF4 promotes the differentiation of TS cells and increased invasiveness of trophoblasts through Matrigel-coated transwells (Figure 2A). The invasive population was largest at four days post FGF4 withdrawal (TINV) (Figure 2A). Morphologically, TSWT cells cultured in FGF4 grew in tight epithelial sheets with actin localized around the cell periphery (Figures 2B and 2C). In contrast, TINV cells isolated from the bottom of Matrigel-coated transwells exhibited mesenchymal cell characteristics with prominent actin stress fibers and filamentous actin localized to the front end of polarized cells (Figures 2D and 2E) and loss of E-cadherin (Figures 2F and 2G). E-cadherin was more significantly reduced in TINV cells compared to four day differentiated trophoblasts (TDIFF) not selected for invasiveness (Figures 2F and 2G). Expression of the mesenchymal marker fibronectin was significantly increased and vimentin was also expressed in TINV cells (Figure S2A). These changes in invasiveness, morphology, and E-cadherin expression indicated that the TINV cells have undergone EMT.
To induce EMT in TS cells, we ectopically expressed a defined inducer of EMT, the transcription factor Snail, in TSWT cells (TSSnail) (Cano et al., 2000). TSSnail cells exhibited a polarized, mesenchymal morphology with increased filamentous actin and mesenchymal markers vimentin and N-cadherin (Figures 2H–J and Figure S2B). A ten-fold increase in invasiveness compared to TSWT cells and loss of E-cadherin showed Snail overexpression induced TS cell EMT (Figures 2K and 2L). TSSnail cells expressed TS stem cell markers and were able to differentiate into trophoblast lineages upon removal of FGF4 (Figures S2C and S2D). To identify transcription factors responsible for inducing EMT in TS cells, we compared the expression levels of eight transcription factors known to regulate EMT in other systems. FGF4 withdrawal from TSWT cells induced Snail message at four days of differentiation and in TINV cells suggesting that Snail is important for TS cell EMT (Figures 2M and Figure S2E). The EMT-inducing transcription factors Slug and Twist were elevated in undifferentiated TSKI4 cells, but were not induced with differentiation or invasion of TSWT cells (Figures 2N and 2O and Figure S2E). Upregulation of Lef1 and Ets1 were also detected in TSKI4 and TSSnail cells (Figures 2P and 2Q). Similar to Snail, Lef1 was induced at four days of TSWT cell differentiation and in TINV cells (Figure 2P and Figure S2E). Other known transcriptional regulators of EMT like Zeb1, Foxc2, and Gsc were not induced with EMT of TINV or TSKI4 cells (Figures S2F–H).
To identify genes related to the acquisition of invasiveness in trophoblast EMT, we measured gene expression changes in TINV cells compared to TSWT cells using genome-wide Agilent microarrays. With a Benjamini-Hochberg p-value <0.05, 6641 genes exhibited a two-fold change in expression in TINV cells (Table S1). Since TINV cells were differentiated for four days, a component of the 6641 changed genes reflected differentiation-specific changes unrelated to the acquisition of invasiveness. When gene expression changes of TDIFF cells compared to TSWT cells were measured, 5706 genes exhibited altered expression by a minimum of two-fold (Table S1). Direct comparison of gene expression changes in TINV and TDIFF cells revealed that 2359 genes had significantly different expression measured as a minimum 1.5-fold change with 80% of these genes being changed only in TINV cells (Table S2). Gene ontology (GO) analysis of the 2359 genes showed significant enrichment in canonical KEGG pathways regulating focal adhesions, actin cytoskeleton, and adherens junctions (Figure S3A). These findings define an invasive gene signature for trophoblast EMT.
Unlike TINV cells that have undergone EMT but have lost stemness, TSKI4 cells are self-renewing stem cells in EMT. TSKI4 cells serve as a useful model to distinguish between genes that mediate invasiveness and EMT. Gene expression profiling defined changes in TSKI4 cells compared to TSWT cells, identifying 1083 significantly upregulated genes and 702 significantly downregulated genes by more than two-fold (Benjamini-Hochberg p-value <0.05) (Table S1). GO analysis of downregulated genes showed a significant enrichment in nine canonical signaling pathways (Figure S3B). The top three enriched signaling pathways included focal adhesions, ECM-interactions, and regulation of the actin cytoskeleton, the same pathways that were identified in the TINV signature.
Unsupervised hierarchical clustering of TSKI4, TINV and TDIFF cells showed a common gene set for TINV and TDIFF cells indicated by a node correlation coefficient of 0.82 (Figure 3A). This strong correlation reflects the significant changes in gene expression due to FGF4 removal and loss of stemness in TINV and TDIFF cells, as compared to TSKI4 cells, showing that TSKI4 cells are not differentiated cells. Venn diagrams of upregulated (Figure 3B) and downregulated (Figure 3C) genes highlight the significant intersection between TDIFF and TINV cells, while demonstrating the limited intersection of 416 genes between TSKI4 and TINV cells. Significantly changed genes were categorized according to biological function, and 25% of these genes regulate cell adhesion and motility including Lamb2, Fn1 and RhoB (Figure 3D and Figure S3C). In addition, genes were identified whose importance to invasion and EMT has not been previously defined, including the enrichment for genes regulating RNA splicing, transcription, translation, and protein degradation (Figure S3C). However, these genes must be functionally tested to prove their role in invasion and EMT. Using qRT-PCR, we validated the expression changes of several genes showing similar changes in TSKI4 and TINV cells (Figures S3D and S3E).
Induction of differentiation by FGF4 removal from TSWT cells was accompanied by reduced acetylation of all core histones, specifically at H2AK5, H2BK5/K12/K15/K20, H3K9, and H4K8 (Figure 4A). Trimethylation of H3K4 and H3K9 was unchanged with differentiation (data not shown). The differentiation-induced loss of histone acetylation suggests the importance of histone acetylation in maintaining the undifferentiated epithelial state of TS cells. We therefore examined histone modifications in TSKI4 cells. Figure 4B shows the loss of acetylation at H2AK5, H2BK5, H2BK12, and H2BK15 in TSKI4 cells compared to TSWT cells. H3K9Ac and H4K8Ac were unaffected demonstrating that loss of histone acetylation was selective for H2A and H2B. Examination of histone methylation showed that trimethylation of H3K4 and H3K9 was not altered, suggesting that loss of H2A and H2B acetylation (H2A/H2BAc) in TSKI4 cells occurs independently of changes in histone methylation (Figure 4C). Examination of TSSnail cells that exhibit properties of stemness and EMT similar to TSKI4 cells revealed the selective loss of H2A/H2BAc with no change in H3 and H4 acetylation (Figure 4D). These data show the association of loss of H2A/H2BAc with changes characteristic of EMT while maintaining stemness in TSKI4 cells.
We used ChIP-seq to identify genes associated with the epigenetic mark H2BK5Ac during TS cell reprogramming important for the induction of EMT. After obtaining DNA samples by ChIP using a highly specific H2BK5Ac antibody (Figures S4A and S4B), we generated a total of 44 and 27 million Illumina sequence reads for TSWT and TSKI4 cells, respectively, consistent with Western blotting data (Figure 4). This genome-wide analysis of the read-tag distribution demonstrated that H2BK5Ac is significantly enriched near the transcription start sites (TSS) of 13625 genes in TS cells (Table S3 and Figure 5A). This H2BK5Ac peak location profile near the TSS is consistent with published studies in CD4+ T cells (Wang et al., 2008).
Limiting the analysis to the well-annotated RefSeq gene set (NCBI37/mm9), we compared the read tag density between TSWT and TSKI4 cells within 1kb upstream and downstream of the TSS. We normalized read-tag counts based upon the ratio of genome-wide total mapped reads between TSWT and TSKI4 cells. We removed background regions by filtering genes whose read tag counts did not significantly exceed cell-type specific background noise (<20 read tags per 2kb region). After removing duplicate genes, 4163 genes were significantly different in H2BK5Ac between TSWT and TSKI4 cells, as determined by the exact rate ratio test with Benjamini-Hochberg adjusted p-value ≤ 0.05. From the 4163 genes with a significant change in H2BK5Ac, 3515 genes had a significant loss of H2BK5Ac, while 648 genes had a significant gain of H2BK5Ac in TSKI4 compared to TSWT cells (Figures 5B and 5C).
Changes in H2BK5Ac were visualized by normalized read tag density plots. We demonstrated the dramatic loss of H2BK5Ac in TSKI4 cells for a select set of genes including Acsl6, Dbndd2, Itgav, Krt19 and Trim54 (Figures 5D and Figure S4C). These are examples of genes with a highly significant loss of H2BK5Ac (i.e. Benjamini-Hochberg p-values <10−18) and occur in the top 3% of affected genes (Table S3). Loss of H2BK5Ac in TSKI4 compared to TSWT cells was confirmed by ChIP-qRT-PCR (Figure 5E) and correlated with the loss of gene expression (Table S2 and Figure S3E). Furthermore, Acsl6, Itgav, Lamb2 and Trim54 demonstrated a similar loss of H2BK5Ac in TINV cells by ChIP-qRT-PCR, suggesting the importance of these genes in regulating the EMT program that occurs during trophoblast differentiation (Figure 5E). Although the majority of genes had a loss of H2BK5Ac in TSKI4 compared to TSWT cells, 648 genes had an increase in H2BK5Ac density in TSKI4 cells. Normalized density plots of Dkk3 and Mycn highlight two genes with a significant increase in H2BK5Ac (i.e. Benjamini-Hochberg p-values < 10−12) and a coordinate increase in gene expression in TSKI4 cells (Figure 5F, Table S1 and Table S3).
Consistent with the maintenance of TS cell multipotency, enrichment of H2BK5Ac occurs at the promoters of the TS cell markers Cdx2, Eomes, and Fgfr2 (Figure 5G). As indicated from normalized promoter density plots, Eomes and Esrrβ demonstrated unchanged levels of H2BK5Ac between TSWT and TSKI4 cells, while Cdx2 and Fgfr2 demonstrate a 50% decrease in H2BK5Ac density (Figure 5G). H2BK5Ac levels were validated by H2BK5Ac ChIP-qRT-PCR for the Eomes and Cdx2 promoters (Figure S4D).
Using mRNA-seq to quantitatively compare the changes in H2BK5Ac with the changes in gene expression for TSWT and TSKI4 cells, there was a modest positive correlation between changes in gene expression and H2BK5Ac, as determined by a Pearson correlation coefficient of 0.62 (p-value <10−16) (Figure 5H). This finding was further supported by GO analysis of genes both significantly downregulated and hypoacetylated showing shared gene changes important for maintenance of focal adhesions, the actin cytoskeleton, and extracellular matrix interactions (Figure S3B and Figure S4E). Collectively, these results highlight the importance of H2BK5Ac in regulating the active gene transcription program of TS cells, whereby loss of H2BK5Ac results in repression of genes critical to maintenance of the epithelial phenotype.
Recent studies have shown that the CL subtype of triple negative breast cancer exhibits both mesenchymal and stem-like properties (Prat et al., 2010). Compared to the four other breast tumor subtypes (i.e. luminal A, luminal B, HER2-enriched and basal-like), CL tumors have the lowest expression of epithelial differentiation markers CD24, EpCAM, KRT7/19 and the cell adhesion proteins CLDN3/4/7 and CDH1, while exhibiting highest expression of the mesenchymal markers VIM, N-cadherin, SNAI2 and TWIST1 (Prat et al., 2010). Hierarchical clustering of 22 genes characteristic of EMT and stemness from gene expression data of TSKI4 cells and 52 breast cancer cells lines reported by Neve et al. revealed that TSKI4 cells clustered most closely with the CL breast cancer subtype (Figure 6A) (Neve et al., 2006). Similar to the CL cell lines, TSKI4 cells exhibited an increase in the mesenchymal markers VIM, CDH2, SNAI2 and TWIST1 with loss of the epithelial differentiation and cell adhesion markers CD24, KRT7/8/19 and CLDN4 (Figure 6A). Next, we examined the intersection between genes with unique Entrez identifiers from gene array data of CL cell lines compared to gene array data of TSKI4 cells. The intersection between upregulated and downregulated genes in the CL cell lines compared to TSKI4 cells was determined to be significant with 62 upregulated (p-value < 0.005) and 31 downregulated (p-value < 0.01) overlapping genes (Figure 6B). This overlapping gene set was plotted on the basis of log2 ratio values from TSKI4 cells to demonstrate gene expression changes of the intersecting TSKI4/CL cell EMT gene signature (Figure 6C). Genes important for the induction of the mesenchymal phenotype, such as CDH2, DKK1, MET, PDGFRβ, SNAI2, TIMP2, THY1, TWIST1 and VIM were significantly upregulated, while genes important for maintenance of the epithelial phenotype and cell adhesion, such as AIM1, BCAM, KRT7, KRT19 and RAB25, were repressed (Figure 6C). In addition to these known regulators of EMT, this significant genetic intersection between two distinct EMT models with stem cell characteristics highlights a gene set important for both EMT and stemness.
TS cell EMT models, TSKI4 and TSSnail, demonstrated selective loss of histone H2A/H2BAc (Figures 4D). By H2BK5Ac ChIP-qRT-PCR, we examined the levels of H2BK5Ac on 32 downregulated genes that have overlapping gene expression profiles between CL SUM159 and TSKI4 cells and are known to have a significant loss of H2BK5Ac in TSKI4 cells by ChIP-seq (Figures 6C and Table S3). Of the 32 genes tested by H2BK5Ac ChIP-qRT-PCR, 75 percent of these genes were validated to have a loss of H2BK5Ac and a coordinate loss of gene expression in TSKI4 compared to TSWT cells (Figure 6D and Figures S5A–C). Furthermore, 81 percent of these genes had a similar loss of H2BK5Ac in TSSnail cells (Figure 6D and Figures S5B and S5C). Due to the significant genetic intersection between CL cell lines and the TSKI4 EMT model, we determined the levels of H2BK5Ac on the promoters of the same 32 overlapping genes in the CL SUM159 cell line. Importantly, 80 percent of these genes had a loss of H2BK5Ac in SUM159 cells compared to human mammary epithelial cells (HMECs), as determined by ChIP-qRT-PCR (Figure 6E and Figures S5D and S5E). These results suggest that loss of H2BK5Ac represses genes with an important role in maintenance of the epithelial phenotype, thereby contributing to the progression of two distinct EMT programs.
The pathological significance of the TSKI4/CL association was further emphasized by comparing the gene expression profile of TSKI4 cells to the five intrinsic molecular subtypes of breast tumors cataloged in the UNC337 data set (Prat et al., 2010). Tumors from the CL breast cancer subtype showed highest expression of the TSKI4 gene expression signature compared to the basal-like, HER2-enriched, luminal A and luminal B breast tumors (Figure 6F). Further analysis of the overlapping gene expression profiles in CL human tumors and TSKI4 cells demonstrated a highly significant intersection between the gene array profiles of TSKI4 cells and CL human tumors (p-value < 0.0001 for upregulated genes) (Figure 6G). Although the intersection between the downregulated genes from TSKI4 cells and CL human tumors consists of only 13 genes, approximately 50 percent of these genes AIM1, IRX5, KRT7, KRT19, RAB25 and SCYL3 were present in the intersecting TSKI4/CL EMT gene signature; these same genes also exhibited a coordinate loss of H2BK5Ac in TSKI4 and CL cells (Figure 6D). These findings highlight the importance of H2BK5Ac on genes whose repression is important for EMT in both developmental stem cell and metastatic human tumor models of EMT with stem cell properties.
Previously, we showed the requirement of MAP3K4 kinase activity for neural tube, skeletal, and placental development (Abell et al., 2009; Abell et al., 2005). We systematically examined the MAP3K4 signaling network for genes whose targeted disruption resulted in phenotypes similar to that of KI4 mice (Table S5). Strikingly, the genes overlapping most closely with the developmental defects observed with loss of MAP3K4 kinase activity were the histone acetyltransferase (HAT) CBP and its closely related family member p300. This phenotypic overlap suggested that the loss of H2A/H2BAc in TSKI4 cells may be related to altered CBP and/or p300 HAT activity. Nuclear extracts isolated from TSKI4 cells have significantly diminished HAT activity relative to TSWT cells (Figure 7A). Total CBP and p300 protein expression was unchanged in TSKI4 cells (Figure S6A). MAP3K4-dependent JNK phosphorylation of CBP and p300 increased HAT activity, which was blocked by the JNK inhibitor SP600125 (Figure 7B and Figure S6B). TSKI4 cells have a strongly diminished JNK activity (Abell et al., 2009), consistent with MAP3K4-dependent JNK activation regulating CBP/p300 HAT activity. Co-expression of MAP3K4 and JNK resulted in a 17.8 and 8.3-fold increase in the phosphorylation of CBP and p300, respectively (Figure 7C and Figure S6C). CBP/p300 phosphorylation was JNK dependent, as p38 activation did not significantly alter phosphorylation of CBP or p300 (Figure 7C and Figure S6C). To determine if CBP or p300 regulate endogenous TS cell functions, we infected TSWT cells with shRNAs to either CBP or p300. We achieved greater than 85% knockdown of CBP or p300 with three to four individual shRNAs for each (Figures 7D and 7E and data not shown). Unlike control virus infected cells (Figure 7F), loss of CBP resulted in a dramatic change in morphology with TSshCBP cells exhibiting a front-back end polarized morphology (Figures 7G and 7H). In contrast, cells with loss of p300 maintained a normal epithelial morphology (data not shown). Compared to control virus infected cells, stemness markers in TSshCBP cells were unchanged for Eomes and FGFR2 and decreased by 25% for Cdx2 and Esrrβ (Figure S6E). Further, expression of the mesenchymal marker vimentin was increased at both the level of message and protein in TSshCBP cells (Figure S6F and data not shown). Most importantly, TSshCBP cells exhibited a five to fifteen-fold increase in invasiveness as compared to control virus infected cells (Figure 7I). Changes in morphology and the expression of mesenchymal markers combined with increases in invasiveness suggest that loss of CBP in TSWT cells induces EMT. Examination of histone acetylation in TSshCBP cells revealed the selective and specific loss of H2A/H2BAc (Figure 7J). In contrast, loss of p300 resulted in a reduction in H3 and H4 acetylation, but did not affect H2A/H2BAc (Figure 7K). These data strongly suggest that CBP is the primary HAT that regulates H2A/H2BAc in TS cells and that the loss of H2A/H2BAc is sufficient to induce EMT in TS cells.
Genes downregulated in the TSKI4/CL EMT gene signature were similarly decreased in TSshCBP cells (Figure S6G). Because TSshCBP cells exhibit the selective loss of H2A/H2BAc acetylation similar to TSKI4 and TSSnail cells, (Figure 4D and Figure 7J), we used H2BK5Ac ChIP-qRT-PCR to measure H2BK5Ac on downregulated genes from the intersecting TSKI4/CL gene profile (Figure S6G). Of these genes, 72 percent demonstrated a loss of H2BK5Ac in TSshCBP cells compared to control virus-infected cells (Figure S6H). These data show the coordinate loss of H2BK5Ac and gene expression in CL, TSKI4, TSSnail, and TSshCBP cells. Together, these findings show the importance of CBP-mediated H2BK5Ac in maintaining the epigenetic landscape important for the epithelial phenotype of TS cells.
We have shown MAP3K4 dependent activation of JNK in response to FGF4 controls CBP activity for the maintenance of the TS cell epithelial phenotype. Loss of MAP3K4 kinase activity in TSKI4 cells results in gain of EMT properties including reduced E-cadherin, and morphological changes characteristic of mesenchymal cells and increased invasiveness. TSKI4 cells also retain stemness defined by self-renewal with the maintenance of multipotency. These properties of TSKI4 cells show a functional separation of FGF4 dependent control of epithelial maintenance and stemness, with MAP3K4 signaling being critical for the epithelial phenotype.
TSKI4 cells exhibit the selective loss of H2A/H2BAc, whereas histone H3 and H4 acetylation was largely unaffected. Loss of H2BK5Ac is restricted to a select set of genes in TSKI4 cells whose expression is significantly reduced. Epithelial maintenance was also disrupted by CBP knockdown, causing the loss of H2A/H2BAc similar to that observed with TSKI4 cells. Loss of CBP expression induced a phenotype similar to TSKI4, including gain of invasiveness and EMT characteristics while maintaining stemness. Consistent with the novel role of CBP and H2BK5Ac in regulation of gene expression profiles important for the epithelial phenotype, H3K4me3 and H3K9me3 are unchanged in TSKI4 cells. Additionally, H3K27me3 has been shown as unimportant in TS cell differentiation (Rugg-Gunn et al., 2010). Thus, histone acetylation by CBP is a primary mechanism for maintenance of the epithelial phenotype of TS cells, whereby loss of H2BK5Ac results in EMT. This finding is consistent with the role for CBP in maintaining hematopoietic stem cell self-renewal (Rebel et al., 2002). In addition to direct inhibition of CBP HAT activity, it is possible that a secondary mechanism of regulation exists to target loss of H2A/H2BAc to select gene promoters, whereby changes in CBP phosphorylation controls interactions with transcriptional regulators of EMT (He et al., 2009).
Ectopic expression of Snail has been used to induce EMT in different cell types, and overexpression of Snail in HMECs induced a mesenchymal phenotype with expression of specific stem cell markers (Mani et al., 2008). This phenotype is reminiscent of TSKI4 cells, which induce EMT while maintaining stemness. Stable expression of Snail in TS cells resulted in the selective loss of H2A/H2BAc and properties of EMT and stemness, similar to TSKI4 and TSshCBP cells. ChIP-qRT-PCR studies showed loss of H2BK5Ac on an overlapping set of genes for TSKI4, TSshCBP, and TSSnail cells, defining each as a unique model system for the epigenetic control of EMT in a self-renewing primary cell.
In contrast to TSKI4, TSshCBP, and TSSnail cells, TINV cells have completed EMT, having fully lost their epithelial morphology, as evidenced by the increase in invasiveness and gain in the mesenchymal morphology associated with filamentous actin and increased expression of the mesenchymal marker fibronectin. TINV cells do not self-renew and have lost acetylation of all four core histones. Since TINV cells have completed EMT, TINV gene expression profiles probably lack the EMT initiators, instead showing the expression of EMT executors (Thiery et al., 2009). Compared to TINV cells, TSKI4, TSshCBP, and TSSnail cells are in an intermediate state of EMT, where they are not fully mesenchymal but exhibit properties of both EMT and stemness. TSKI4 cells are uniquely trapped in this intermediate EMT state prior to complete acquisition of the mesenchymal phenotype, which can still be induced by the removal of FGF4.
TS cell EMT shares many key properties with neural crest and cancer cell EMTs including loss of E-cadherin, gain of front-to-back polarity, and increased invasiveness (Yang and Weinberg, 2008). However, there are differences in marker expression among these different EMT models indicative of cell type and stage-specific EMT. For example, mesenchymal markers such as vimentin and N-cadherin are differentially expressed in these different EMT models, with Ncadherin repression being required for neural crest EMT (Yang and Weinberg, 2008). Lamb2 is increased in hepatocyte EMT, but reduced in neural crest, CL, TINV, TSKI4 and TSSnail EMTs. Fibronectin is elevated in breast and gastric cancers and in TINV cells but reduced in TSKI4 and TSSnail cells. The critical property for each EMT model is increased invasiveness (Kalluri and Weinberg, 2009).
Finally, TSKI4 and CL human breast cancer cells share properties of stemness and EMT with a common gene expression profile also found in patient CL tumors. Intersecting genes with loss of expression had a correlative loss of H2BK5Ac in both TSKI4 and CL cells. Some of these genes have defined roles in maintenance of the epithelial phenotype such as Aim1, Rab25 and Galnt3 (Maupin et al., 2010; Ray et al., 1997), but most of the shared genes in the TSKI4/CL intersecting list have not been characterized for their role in epithelial maintenance or EMT and should be analyzed in different EMT models. Discovery of how H2A/H2BAc controls maintenance of the epithelial TS cell phenotype provides unique insight into how signaling networks controlling tissue stem cell EMT can be used to define previously unrecognized genes contributing to cancer cell EMT. This discovery may lead to defining novel gene targets or combinations of targets whose inhibition can be used to selectively inhibit TICs.
TSWT and TSKI4 cells of normal karyotype were isolated from heterozygote crosses of mice with a targeted mutation of MAP3K4 (K1361R) as previously described (Abell et al., 2009). TS cells were cultured without feeders in 30% TS media (RPMI 1640, 20% fetal bovine serum, 1% penicillin and streptomycin, 1% L-glutamine, 1% sodium pyruvate and 100 µM β-mercaptoethanol) and 70% MEF conditioned TS cell media, supplemented with FGF4 (37.5 ng/ml) and Heparin (1 mg/ml). For differentiation, TS cells were cultured in TS media only. Invasion assays, isolation of TINV cells, HEK293 cell culture, transfection, and plasmids are specified in Supplemental Experimental Procedures.
Whole cell and nuclear lysates were isolated as previously described (Abell et al., 2009). For histone lysates, cells were lysed on ice in buffer containing PBS, 1% Triton-X and 1 mM PMSF. Pellets were spun at 2000 rpm for 10 min at 4° C and extracted overnight in 0.2 N HCl with shaking at 4° C. Western blots were performed with the antibodies specified in Supplemental Experimental Procedures.
HAT assays and kinase assays were performed as described in Supplemental Experimental Procedures.
Cells were fixed for 10 min in 1% formaldehyde, sonicated (VCX130 Ultrasonicator) and immunoprecipitated with 5 µg anti-H2BK5Ac and Protein A dynabeads (Invitrogen) (Wang et al., 2008). Crosslinking was reversed by overnight incubation at 65°C. DNA was purified with the MinElute PCR purification kit (Qiagen). Library preparation for Illumina ChIP-seq was performed according to manufacturer’s instructions (Illumina). Illumina Solexa GA II was used to produce ~36 bp sequence reads, which were mapped to the mouse genome using Mapping and Alignment with Quality (MAQ) software in conjunction with EpiCenter for comparative analysis as described in Supplemental Experimental Procedures. PCR conditions and primers used for ChIP-seq validation are described in Supplemental Experimental Procedures.
Total RNA was isolated using RNeasy Plus minikit (Qiagen). cDNA was synthesized from 3 µg RNA using High-Capacity reverse transcription kit (Applied Biosystems). Applied BIosystems 7500 RT-PCR system with inventoried TaqMan probes was used to quantify gene expression levels normalized to β-actin expression. Agilent Gene Expression Arrays, Comparative Expression and GO Pathway Analysis were performed as described in Supplemental Experimental Procedures.
Immunostaining was performed as described in Supplemental Experimental Procedures.
GLJ is supported by NIH grants GM30324 and DK37871 and the University Cancer Research Fund for support of the deep Sequencing Genomics Facility. NVJ is supported by NIH training grant GM007040. WH and LL are supported by the Intramural Research Program of NIH, NIEHS (ES-101765). The authors thank Noah Sciaky for managing databases, Kim Kluckman and the UNC Animal Models Core for production of chimeras, Betsy Clarke for graphical work, Zefang Wang for ChIP-seq discussions and Tso-Pang Yao (Duke University) for CBP/p300 constructs and helpful discussions. We thank Terry Magnuson, Bill Snider, Bryan Richardson and Jon Zawistowski for careful reading of the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
The accession number for the microarray data is GSE27883.
Supplemental data includes Supplemental Experimental Procedures, five tables, and six figures.
The authors have no conflict of interest.