Here, we show that expression of ZNF281/ZBP‐99 is controlled by SNAIL and miR‐34a/b/c in a coherent feed‐forward loop: the epithelial–mesenchymal transition (EMT) inducing factor SNAIL directly induces ZNF281 transcription and represses miR‐34a/b/c, thereby alleviating ZNF281 mRNA from direct down‐regulation by miR‐34. Furthermore, p53 activation resulted in a miR‐34a‐dependent repression of ZNF281. Ectopic ZNF281 expression in colorectal cancer (CRC) cells induced EMT by directly activating SNAIL, and was associated with increased migration/invasion and enhanced β‐catenin activity. Furthermore, ZNF281 induced the stemness markers LGR5 and CD133, and increased sphere formation. Conversely, experimental down‐regulation of ZNF281 resulted in mesenchymal–epithelial transition (MET) and inhibition of migration/invasion, sphere formation and lung metastases in mice. Ectopic c‐MYC induced ZNF281 protein expression in a SNAIL‐dependent manner. Experimental inactivation of ZNF281 prevented EMT induced by c‐MYC or SNAIL. In primary CRC samples, expression of ZNF281 increased during tumour progression and correlated with recurrence. Taken together, these results identify ZNF281 as a component of EMT‐regulating networks, which contribute to metastasis formation in CRC.
The ZNF281/ZBP‐99 protein has characteristics of a transcription factor and contains four Krüppel‐type zinc‐finger domains (Law et al, 1999; Lisowsky et al, 1999). ZNF281 is related to ZBP‐89, which has been implicated in the regulation of cell proliferation (Bai and Merchant, 2001), apoptosis (Bai et al, 2004), differentiation (Li et al, 2006) and tumorigenesis (Law et al, 2006). ZNF281 mediates transcriptional repression and activation (Wang et al, 2008). For example, ZNF281 directly regulates the expression of gastrin and represses ornithine decarboxylase (ODC) by binding to GC‐rich sequences in their promoters (Law et al, 1999; Lisowsky et al, 1999). Previously, we detected a direct interaction between ZNF281 and the c‐MYC oncoprotein (Koch et al, 2007). Moreover, ZNF281 participates in the regulation and maintenance of pluripotency by interacting with transcription factors controlling stemness, such as Nanog, Oct4 and Sox2 (Wang et al, 2006, 2008). Furthermore, ZNF281 directly regulates Nanog expression and contributes to its auto‐regulation by recruiting the NuRD complex in mouse embryonic stem cells (Fidalgo et al, 2011, 2012). The Sox4 transcription factor directly induces ZNF281 transcription (Scharer et al, 2009). Interestingly, Sox4 has been implicated in the regulation of differentiation, proliferation, epithelial–mesenchymal transition (EMT) and shows increased expression in many human cancers (Zhang et al, 2012). After DNA damage, the ZNF281 protein is phosphorylated by ataxia telangiectasia mutated (ATM) and ATM and Rad3‐related (ATR) kinases (Matsuoka et al, 2007). However, other signals regulating ZNF281 activity and expression have remained elusive.
As a morphogenic programme, EMT is involved in the formation of tissue and organs during embryonic development and wound healing. During EMT, epithelial cells acquire mesenchymal features, such as decreased cell–cell contacts and loss of polarity, which promote increased motility and invasiveness. Thereby, EMT contributes to the progression of early‐stage tumours to invasive malignancies (Thiery, 2002; Lee et al, 2006; Hugo et al, 2007). So far only a few transcription factors, such as SNAIL, SLUG, TWIST1/2 and ZEB1/2, are thought to constitute the central regulatory core of EMT (Peinado et al, 2007; Sanchez‐Tillo et al, 2012). EMT has also been shown to promote stemness of cancer cells that may endow tumour initiating cells with traits necessary for metastasis formation, as shown for immortalized human mammary epithelial cells undergoing EMT (Brabletz et al, 2005b; Mani et al, 2008). Accordingly, SNAIL, TWIST and ZEB1 share the ability to induce both stemness and EMT (Mani et al, 2008; Wellner et al, 2009). Furthermore, tumour cells undergoing EMT show accumulation of active β‐catenin in the nucleus (reviewed in Brabletz et al, 2005a).
Recently, microRNAs (miRNAs) have emerged as major regulators of EMT (Brabletz, 2012; Hermeking, 2012). For example, members of the miR‐200 family and miR‐205 promote MET by inhibiting EMT inducing factors like ZEB1 and ZEB2 (Gregory et al, 2008; Park et al, 2008), and miR‐34a/b/c achieve the same effect by downregulating SNAIL expression (Kim et al, 2011; Siemens et al, 2011). Moreover, SNAIL directly represses miR‐34a/b/c transcription (Siemens et al, 2011). The resulting double‐negative feedback loop represents a bistable switch, which can be locked in the mesenchymal state by inactivation of miR‐34 genes by CpG methylation, which is often found in cancer cells (Hermeking, 2012; Siemens et al, 2013). Interestingly, the genes encoding the miR‐200 and miR‐34 families are direct p53 targets and their mediation of MET presumably contributes to tumour suppression by p53 (Hermeking, 2012).
Here, we show that ZNF281 expression is regulated by a feed‐forward loop involving SNAIL and miR‐34a. Taken together, we show that ZNF281 is an integral part of the EMT‐regulating transcriptional network and controls processes relevant to colorectal cancer (CRC) progression, such as migration, invasion, stemness and metastasis.
SNAIL regulates ZNF281 expression
We previously identified an interaction between c‐MYC and ZNF281 proteins in a systematic analysis of c‐MYC‐associated protein complexes (Koch et al, 2007). ZNF281 was among the proteins represented by the highest number of mass‐spectrometric sequence reads, indicating that it is associated with a large fraction of cellular c‐MYC and presumably represents a significant regulator or effector of c‐MYC. However, so far it is largely unknown how ZNF281 expression itself is regulated and whether it participates in regulatory pathways, which might by relevant for c‐MYC function and tumour biology. In order to identify upstream regulators of ZNF281, we inspected the ZNF281 promoter sequence for binding sites of transcription factors, which might hint towards cancer‐relevant functions of ZNF281. Thereby, we identified several E‐Box motifs (CACCTG) in the ZNF281 promoter, which represent putative SNAIL binding sites (SBSs; Figure 1A). SNAIL is a known regulator of EMT (Mauhin et al, 1993; Batlle et al, 2000) and, similar to ZNF281, a zinc‐finger‐containing transcription factor (Nieto, 2002; Sanchez‐Tillo et al, 2012; see also comparison in Supplementary Figure S1). Two of the SBSs were located ∼500 and ∼700 bp upstream of the transcription start site (TSS; Figure 1B). SBS2 and SBS4 are conserved between the human and mouse ZNF281 promoters, indicating functional relevance (Figure 1B). When SNAIL was ectopically expressed in DLD‐1 CRC cells using a Doxycycline (DOX) inducible episomal vector system an increase in the SNAIL occupancy of the ZNF281 promoter was detected by chromatin immunoprecipitation (ChIP) analysis at SBS2 and SBS3, whereas SBS1, 4 and 5 did not display increased binding of SNAIL (Figure 1C). Also endogeneous SNAIL protein selectively occupied SBS2 and SBS3 in SW620 CRC cells (Supplementary Figure S2A). Furthermore, ectopic SNAIL enhanced the expression of ZNF281 at the protein and the mRNA level in DLD‐1 cells (Figure 1D and E). SNAIL also induced ZNF281 expression in SKBR3 breast cancer and MiaPaCa2 pancreatic cancer cells (Supplementary Figure S2B). Therefore, the induction of ZNF281 by SNAIL is not restricted to a specific cell type. In order to determine whether ZNF281 is induced by SNAIL via SBS motifs, a region encompassing ∼2 kbp upstream of the ZNF281 transcriptional start site was subjected to a dual reporter assay (Figure 1F). Indeed, the wild‐type reporter was induced by SNAIL, whereas mutation of SBS2 abolished and SBS3 mutation decreased the responsiveness to SNAIL. Also a reporter with combined mutation of SBS2 and SBS3 resulted in complete loss of responsiveness to SNAIL. A CDH1 promoter reporter was repressed by SNAIL in an SBS‐dependent manner in this assay. ZNF281 displayed the highest expression level in CRC cell lines with mesenchymal features, such as Colo320, SW480 and SW620, whereas HT29, DLD‐1 and HCT‐15 cells, which display an epithelial phenotype, showed comparatively low ZNF281 expression levels (Figure 1G). ZNF281 expression positively correlated with SNAIL and Vimentin and inversely with E‐cadherin expression (Figure 1G). Moreover, analysis of publicly available mRNA expression profiles obtained from seven CRC cell lines (COLO205, HCC2998, HCT116, HCT15, HT29, KM12, SW620) within the NCI‐60 panel (Shoemaker, 2006) confirmed a significant correlation between ZNF281 and the mesenchymal markers SNAIL, Vimentin and Fibronectin‐1 (Supplementary Table S1). Taken together, these results suggested that the induction of ZNF281 by SNAIL may be an important component of the EMT programme induced by SNAIL. Indeed, when ZNF281 was downregulated using two different siRNAs the induction of EMT by SNAIL was prevented in DLD‐1 cells (Figure 1H and I; Supplementary Figure S2C). Also the loss of E‐cadherin from the outer membrane, which is typical for EMT, was prevented by simultaneous siRNA‐mediated downregulation of ZNF281. Therefore, ZNF281 is required for SNAIL‐induced EMT.
miR‐34a directly regulates ZNF281 expression
The differences between the pronounced increase in ZNF281 protein levels and minor increase in mRNA levels after ectopic SNAIL expression suggested the possibility of an additional translational regulation mediated by miRNAs. Inspection of the ZNF281 3′‐UTR using the TargetSCAN and Miranda algorithms (John et al, 2004; Grimson et al, 2007) revealed a conserved miR‐34 seed‐matching sequence (Figure 2A). Since we had previously shown that the miR‐34a and miR‐34b/c genes are directly repressed by SNAIL (Siemens et al, 2011), we hypothesized that at least part of the increase in ZNF281 expression observed after SNAIL induction might be due to a repression of miR‐34 genes. Indeed, ectopic miR‐34a expression resulted in the downregulation of endogeneous ZNF281 expression at the protein and mRNA level in SW480 CRC cells (Figure 2B and C). This was also observed in MiaPaCa2 pancreatic cancer cells (Supplementary Figure S3A and B). Therefore, the regulation of ZNF281 by miR‐34a is not restricted to CRC cells. Furthermore, reporter constructs containing the complete 3′‐UTR of ZNF281 (720 bp) or a 77‐bp fragment including the seed‐matching sequence were repressed by co‐transfection of pre‐miR‐34a, but not when the seed‐matching sequence was mutated, demonstrating that it mediates repression by miR‐34a (Figure 2D and E). The induction of ZNF281 by SNAIL was prevented by concomitant transfection of pre‐miR‐34a (Figure 2F). Therefore, the previously documented repression of the miR‐34a gene by SNAIL (Siemens et al, 2011) is presumably necessary for the SNAIL‐mediated increase in ZNF281 expression. In summary, these results demonstrate that ZNF281 is directly regulated by miR‐34a and that SNAIL induces ZNF281, at least in part, by repressing miR‐34a.
p53 represses ZNF281 via miR‐34a
Since the miR‐34 genes represent direct p53 targets, we asked whether p53 represses ZNF281 expression via inducing miR‐34a. Indeed, ectopic expression of p53 resulted in a decrease in ZNF281 at the protein and mRNA level (Figure 3A and B). As expected, miR‐34a/b/c levels were increased upon p53 activation, which is likely to mediate the decrease in ZNF281 protein expression (Figure 3C and D). Since miR‐34b/c is expressed at least at 10‐fold lower levels in CRC and CRC cell lines compared to miR‐34a (Toyota et al, 2008; Siemens et al, 2013) we focussed on miR‐34a in the further analyses. The recovery of ZNF281 mRNA expression by 72 h of ectopic p53 expression is presumably due to the declining expression of ectopic p53 and therefore reduced pri‐miR‐34 induction at this time point (Figure 3A, B and D). Nonetheless, ZNF281 protein was still downregulated 72 h after activation of p53 (Figure 3A). In order to determine whether downregulation of ZNF281 is a result of reduced SNAIL expression caused by direct interaction of SNAIL with p53 (Lim et al, 2010) or due to p53‐induced miR‐34, we directly interfered with miR‐34a function using antagomirs. Indeed, miR‐34a‐specific antagomirs largely abolished the downregulation of ZNF281 after p53 induction, whereas a control antagomir did not affect the repression of ZNF281 by p53 (Figure 3E). The remaining minor repression of ZNF281 may be due to p53‐induced miR‐34b and ‐c, which are presumably not affected by the miR‐34a‐specific antagomir used here. Additionally, we analysed the expression of ZNF281 in HCT116 p53+/+ cells and an isogenic clone with homozygous deletion of p53 resembling p53 inactivation in tumours. HCT116 p53+/+ cells expressed lower endogeneous levels of ZNF281 protein and mRNA than p53‐deficient cells (Figure 3F and G). As previously described (Siemens et al, 2011), the expression of the SNAIL protein was elevated in the HCT116 p53−/− cells (Figure 3F). When SNAIL was downregulated using a SNAIL‐specific siRNA, the expression of ZNF281 protein was only decreased to a minor extent in p53‐deficient HCT116 cells (Figure 3H). Therefore, the increase in ZNF281 expression is presumably mainly due to the decrease in miR‐34a expression in p53‐deficient cells (Figure 3G). Taken together, these results show that miR‐34a represents an important mediator for the repression of ZNF281 by p53.
ZNF281 induces EMT, migration and invasion
Since ZNF281 expression was induced by SNAIL and required for SNAIL‐induced EMT, we determined whether ectopic expression of ZNF281 is sufficient to promote EMT. For this purpose, a pool of DLD‐1 cells harbouring an episomal pRTR construct that allows the DOX‐inducible expression of ZNF281 was generated. After addition of DOX >90% of the cells were positive for eGFP, which is expressed from a bidirectional promoter also driving the expression of ZNF281 (Supplementary Figure S4A). After induction of ectopic ZNF281 expression DLD‐1 cells changed from an epithelial morphology (dense islands of cobblestone‐shaped cells) to a mesenchymal morphology with spindle‐shaped cells forming protrusions and displaying a scattered growth pattern (Figure 4A). This was reminiscent of the effect of ectopic SNAIL expression observed in DLD‐1 cells before (Siemens et al, 2011). Also molecular markers of EMT were regulated by the expression of ZNF281 (Figure 4B). The distinct membrane‐bound expression of E‐cadherin in DLD‐1 cells was lost upon ZNF281 activation. Furthermore, ZNF281‐expressing cells displayed an increased cytoplasmic expression of the mesenchymal marker Vimentin. In addition, F‐actin, which forms stress fibres (Moreno‐Bueno et al, 2009), was relocated from the membrane to the cytoplasm.
Subsequently, we determined whether ectopic ZNF281 expression influences cellular migration and invasion, since EMT has been previously linked to increased migration and invasion (reviewed in Christiansen and Rajasekaran, 2006). In a wound‐healing assay, ectopic ZNF281 expression resulted in a minor, but reproducible increase in the closure of a scratch in a confluent layer of DLD‐1 cells compared to controls (Figure 4C; Supplementary Figure S4H). When migration and invasion were examined in Boyden‐chamber assays the effect of ectopic ZNF281 expression was more pronounced (Figure 4D and E). Furthermore, ectopic expression of ZNF281 significantly enhanced the ability of DLD‐1 cells to form colonies in soft agar (Figure 4F). The addition of DOX to DLD‐1 cells harbouring an empty vector control did not result in EMT‐related morphological changes or significant effects in the above‐mentioned assays (Supplementary Figure S4E–H). The effects of ectopic ZNF281 were not due to increased proliferation, since ZNF281 activation had a slight anti‐proliferative effect (Supplementary Figure S5), which has also been described for other EMT‐TFs, such as SNAIL (Peinado et al, 2007). Taken together, these results show that ectopic expression of ZNF281 is sufficient to mediate EMT and enhances migration, invasion and anchorage‐independent growth.
Transcriptional regulation of EMT markers by ZNF281
Next, we determined whether ZNF281 also induces changes in the expression of genes previously implicated in the transcriptional programme of EMT. After activation of ectopic ZNF281 expression in DLD‐1 cells, an upregulation of SNAIL was observed at the protein and mRNA level (Figure 5A and B). In addition, the mesenchymal markers SLUG, ZEB1 and Fibronectin‐1 were induced after ectopic ZNF281 expression in DLD‐1 cells (Figure 5B). In line with the indirect immunofluorescence results shown in Figure 4B, E‐cadherin/CDH‐1 was repressed at the protein level after induction of ZNF281 (Figure 5A), whereas expression of CDH1 mRNA was not significantly affected by ZNF281 (Figure 5C). However, when ZNF281 was expressed in HT29 CRC cells E‐cadherin was repressed at both the protein and mRNA level (Supplementary Figure S6A and C). Therefore, the regulation of EMT markers by ZNF281 is at least partially dependent on the cellular context. Other epithelial markers, such as OCLN and CLDN‐7, were repressed at the mRNA level in DLD‐1 and HT29 cells (Figure 5C; Supplementary Figure S6C), which is characteristic for EMT (Ikenouchi et al, 2003; Martinez‐Estrada et al, 2006). Furthermore, ectopic ZNF281 expression resulted in the downregulation of a number of additional epithelial marker genes encoding components of tight junctions (ZO‐1/3, CLDN‐1) and adherens junctions (CDH‐3), as well as desmosomes (PKP2, DSP) (Figure 5D), as previously shown for ZEB2 (Vandewalle et al, 2005).
Since SNAIL is a potent inducer of EMT, we determined whether the upregulation of SNAIL by ZNF281 is mediated by direct occupancy of the SNAIL promoter. ZNF281 is known to occupy GC‐rich DNA sequences, as previously shown for the ODC1 promoter (Law et al, 1999; Lisowsky et al, 1999). Similar GC‐rich sequences are present in the SNAIL promoter (Figure 5E). When these regions were analysed by a ZNF281‐specific ChIP, occupancy by ectopic and endogeneous ZNF281 was detected in the vicinity of the SNAIL TSS (Figure 5F; Supplementary Figure S7A). The highest occupancy by ZNF281 protein was detected in a region encompassing the SNAIL promoter itself and ∼600 bp upstream. A set of deletion constructs of the human SNAIL promoter (Barbera et al, 2004) was used to determine the regions mediating the regulation by ZNF281 in a reporter assay. The activation of the SNAIL promoter was most dominant for the −869/+59 reporter (Figure 5G), which was in line with the dominant binding of ectopic ZNF281 to a region ∼600 bp upstream of the TSS in the ChIP assay (Figure 5F). Unexpectedly, the 1558/+92 construct resulted in a weaker response to ZNF281, which may be due to repressive elements or binding sites for other transcription‐ or co‐factors in the region between −869 and −1558, bp. Also, the decreasing activity of further truncations indicates that the predominant region of ZNF281 binding is located between 514 and 869 bp upstream of the SNAIL TSS. Another, less effective binding region is presumably located closer to the TSS as also suggested by the ChIP results (Figure 5F; Supplementary Figure S7A). Occupancy by endogeneous and ectopic ZNF281 was also detected in the promoter regions of CDH1, OCLN and CLDN‐7 (Figure 5H; Supplementary Figure S7B).
When SNAIL was downregulated by RNA interference (Figure 5I and J) in DLD‐1 cells ectopically expressing ZNF281 morphological changes associated with EMT were not observed (Figure 5J). Moreover, E‐cadherin persisted at the cell membrane, whereas co‐transfection of a control siRNA did not prevent its relocalization (Figure 5J). Therefore, ZNF281‐induced EMT is mediated, at least in part, by SNAIL. In summary, these analyses show that ZNF281 directly regulates the expression of a subset of EMT regulators and effectors. The requirement of SNAIL for ZNF281‐induced EMT suggests that at least some of these regulations are mediated and/or enhanced via the induction of SNAIL. Furthermore, ZNF281 directly induces SNAIL, which itself activates ZNF281 expression, thereby forming a positive feedback loop, which may enforce and stabilize the process of EMT.
ZNF281 regulates β‐catenin localization and activity
Interestingly, ectopic expression of ZNF281 in DLD‐1 cells resulted in the translocation of β‐catenin from the cell membrane to the nucleus (Figure 6A), which is another characteristic of EMT (Brabletz et al, 2005b). Although, β‐catenin mRNA and protein levels remained unchanged upon ZNF281 expression (Figure 6B and C), a significant increase in the transcriptional activity of β‐catenin/TCF4 was observed after ZNF281 expression in DLD‐1 and SW480 cells in a reporter assay (Figure 6D). Axin2 negatively regulates the WNT/β‐catenin/TCF4 signalling pathway by promoting phosphorylation and degradation of β‐catenin/TCF4 via a multi‐protein complex including APC and GSK3β (Lustig et al, 2002). Therefore, decreased levels of inhibitory Axin2 might result in the translocation of β‐catenin. Indeed, Axin2 mRNA was repressed upon ectopic expression of ZNF281 (Figure 6E). Furthermore, we detected increased binding of ZNF281 at the Axin2 promoter, which harbours GC‐rich regions representing potential binding sites for ZNF281 (Figure 6F). Therefore, it is conceivable that the direct repression of Axin2 by ZNF281 contributes to the increased activity of TCF4/β‐catenin. In addition, the loss of E‐cadherin from the cell membrane may contribute to the activation of TCF4/β‐catenin after ZNF281 activation (see Discussion). In line with increased β‐catenin/TCF4 activity LGR5 and CD133, which are known β‐catenin target genes (Katoh and Katoh, 2007; Fan et al, 2010; Glinka et al, 2011; Carmon et al, 2012), were induced after ectopic expression of ZNF281 in DLD‐1 cells (Figure 6E). qChIP analysis revealed ZNF281 occupancy at the LGR5 promoter, but not at the CD133 promoter (Figure 6F). Since LGR5 and CD133 represent markers for cancer stem cells (Barker et al, 2007; Zhu et al, 2009; Munoz et al, 2012), the activation of ZNF281 may be accompanied by the acquisition of stem‐cell traits. Indeed, ectopic ZNF281 expression significantly enhanced the formation of colono‐spheres by non‐adherent DLD‐1 cells (Figure 6G and H). Taken together, ZNF281 enhances β‐catenin activity and stemness of tumour cells. Both effects may contribute to metastasis formation.
Requirement of ZNF281 for EMT, migration and invasion
Next, we thought to determine whether ZNF281 is not only sufficient for the induction of EMT but also necessary for the maintenance of an EMT‐like state in the CRC cell line SW480, which displays mesenchymal features such as low E‐cadherin and high SNAIL expression (Figure 1G), as well as enhanced migration, invasion and metastasis. In line with our previous findings, SW480 cells displayed high levels of endogenous ZNF281 when compared to the more epithelial cell lines DLD‐1 and HT29 (Figure 1G). Therefore, we generated cell pools with DOX‐inducible expression of a ZNF281‐specific miRNA or the respective control driven by an episomal pRTS vector. These cell pools showed ectopic expression of an inducible, co‐expressed mRFP marker in ∼90% of the cells after addition of DOX (Supplementary Figures S8A, S9A and S11A). After induction of the ZNF281‐specific miRNAs, endogenous ZNF281 was repressed at the mRNA and protein level (Figure 7A and B; Supplementary Figure S9B and D). Simultaneously, mesenchymal markers, such as SNAIL and Vimentin, were repressed, whereas the epithelial marker E‐cadherin was induced at the protein level (Figure 7B; Supplementary Figure S9D). Furthermore, SW480 cells lost their mesenchymal morphology and gained epithelial phenotypes with an increase in cell–cell contacts (Figure 7C; Supplementary Figure S9C) and diminished expression of Vimentin (Figure 7B and C; Supplementary Figure S9D). Moreover, downregulation of ZNF281 in SW480 cells resulted in decreased migration in a scratch and a Boyden‐chamber assay (Figure 7D and E; Supplementary Figure S8E and F), as well as diminished invasion in a Matrigel‐transwell assay (Figure 7E; Supplementary Figure S9F). The downregulation of ZNF281 had no significant effects on cell proliferation, cell‐cycle distribution and apoptosis (Supplementary Figure S10). In addition, ZNF281 downregulation resulted in a decrease in colony formation in soft agar (Figure 7F) and a reduction in sphere formation (Figure 7G). Similar effects were observed after expression of the other ZNF281‐specific miRNA (Supplementary Figure S9G and H), whereas a non‐specific control miRNA did not result in significant effects in any of the assays described above (Supplementary Figure S11B–E). Taken together, downregulation of ZNF281 induces a MET of SW480 cells, which is associated with the loss of migratory and invasive capacities, as well as reduced stemness. Therefore, expression of ZNF281 is not only sufficient for induction of EMT but presumably also required to maintain a mesenchymal state in CRC cell lines. Furthermore, ZNF281 is not only sufficient to induce SNAIL, but also required for its continued expression.
Requirement of ZNF281 for c‐MYC‐induced EMT
We recently observed that ectopic c‐MYC expression effectively induces EMT in DLD‐1 cells, which was accompanied by an activation of SNAIL expression, mediated, to a large extent, by AP4 (Jackstadt et al, 2013). Since we had detected an interaction between c‐MYC and ZNF281 before (Koch et al, 2007), we asked whether ZNF281 is required for c‐MYC‐induced EMT. Interestingly, c‐MYC activation resulted in an induction of ZNF281 expression at the protein and mRNA level (Figure 8A and B). This was presumably indirect since MYC binding sites were not identified in the ZNF281 promoter region. Interestingly, ectopic c‐MYC expression and concomitant transfection of a SNAIL‐specific siRNA diminished the induction of ZNF281 (Figure 8C). When c‐MYC was ectopically expressed in DLD‐1 cells in the presence of siRNAs directed against ZNF281, the repression of E‐cadherin and also the induction of SNAIL was less pronounced than with co‐transfection of control siRNA (Figure 8D). In addition, siRNA‐mediated downregulation of ZNF281 prevented the adoption of a mesenchymal phenotype after activation of c‐MYC in DLD‐1 cells (Figure 8E). After activation of ectopic c‐MYC and concomitant transfection of ZNF281‐specific siRNAs, E‐cadherin remained at the membrane, whereas co‐transfection of a control siRNA did not interfere with the c‐MYC‐induced loss of membranous E‐cadherin (Figure 8E). Taken together, these results show that c‐MYC‐induced EMT is mediated by ZNF281.
Role of ZNF281 in metastasis formation
Since EMT and the resulting cellular properties have been implicated in the metastatic process (Valastyan and Weinberg, 2011), we asked whether inactivation of ZNF281 in the highly metastatic CRC line SW620 would influence metastasis formation in a xenograph mouse model. Therefore, we generated SW620 cells stably expressing luciferase2 to monitor the development of metastases over time in a non‐invasive manner (Supplementary Figure S13). Transfection of SW620‐Luc2 cells with two different ZNF281‐specific siRNAs resulted in a pronounced downregulation of ZNF281 expression, which was accompanied by a decrease in Vimentin and SNAIL protein (Figure 9A), in line with SNAIL being a target gene of ZNF281 (Figure 5A). Subsequently, these cells were injected into the tail vein of NOD/SCID mice. Within 4 weeks, mice injected with control siRNA‐transfected cells gave rise to luminescence signals in the lung indicating metastases, whereas mice injected with cells transfected with ZNF281‐specific siRNAs did not show luminescence signals until 7–8 weeks after injection (Figure 9B and C). Nine weeks after injection luminescent metastases were easily detectable in mice, which had received control siRNA‐treated cells, whereas cells treated with ZNF281‐specific siRNAs only rarely gave rise to small metastases as evidenced by weak luminescence signals. At this time point, lungs displayed macroscopically visible metastases in the control group, whereas lungs from mice injected with cells transfected with ZNF281‐specific siRNAs were devoid of macroscopically visible metastases (Figure 9C). Haematoxylin and eosin (H&E) staining revealed the presence of metastases in the lungs of the control siRNA group, whereas the knock‐down of ZNF281 largely prevented the colonization of SW620 cells in the lung (Figure 9C and D). Histological examination of the lungs revealed a significant decrease in the total number of metastatic nodules upon inhibition of ZNF281 (Figure 9D). In conclusion, ZNF281 is therefore necessary for metastatic colonization of CRC cells in this in vivo model.
ZNF281 is upregulated in human colon and breast cancer
In order to evaluate whether the pro‐metastatic functions of ZNF281 are reflected in enhanced expression of ZNF281 during progression of CRC and other carcinomas, we analysed the Oncomine database (Rhodes et al, 2004). In 11 out of 12 tumour entities ZNF281 expression was found to be up‐regulated in tumour versus normal tissue (Supplementary Figure S14A). In primary tumour samples of two colorectal and two breast cancer cohorts, cancer‐specific upregulation of ZNF281 was consistently found (Supplementary Figure S14B–E). In addition, an increased ZNF281 expression in the primary tumour of CRC patients was associated with recurrence, and therefore presumably metastasis, 3 years after removal of the primary tumour (Supplementary Figure 14F).
Here, we could show that ZNF281 is an integral part of the regulatory network that controls the transition between epithelial and mesenchymal states in CRC cells (see schematic model in Figure 9E). The results imply that ZNF281 expression is induced by a coherent feed‐forward loop consisting of SNAIL and miR‐34a. Furthermore, ZNF281 expression is sufficient to elicit an EMT and necessary to maintain a mesenchymal state in CRC cell lines. This effect was mediated by the direct activation of SNAIL and repression of epithelial marker genes and effectors.
As shown here, ZNF281 is regulated in a coherent feed‐forward loop, involving SNAIL, which directly binds to the ZNF281 promoter and induces its transcription. Besides being mainly a transcriptional repressor, direct induction of target genes by SNAIL has been shown before (Guaita et al, 2002; De Craene et al, 2005; Vetter et al, 2010). The regulatory loop identified here involves miR‐34a, which is repressed by SNAIL in a negative feedback loop (Kim et al, 2011; Siemens et al, 2011) and itself targets ZNF281. We recently demonstrated that increased SNAIL expression inversely correlates with miR‐34a expression in a cohort of 94 colon cancer patients and was associated with liver metastasis (Siemens et al, 2013). Therefore, the increased expression of ZNF281 in colorectal and other tumour types identified in public data sets may be caused by miR‐34a down‐regulation due to cancer‐specific CpG methylation of miR‐34a and/or p53 inactivation.
Ectopic ZNF281 directly induced SNAIL transcription. However, this resulted in varying degrees of EMT effector regulations in the two different CRC cell lines DLD‐1 and HT29. Nonetheless, ZNF281 expression consistently resulted in the loss of intercellular adhesions and enhanced migration and invasion in those two CRC cell lines. The subtle differences in the transcriptional response of different CRC cell lines to ZNF281 activation may be due to variations in related signalling pathways and therefore varying degrees of permissiveness for EMT or plasticity of the respective cells. We also provided evidence that ZNF281 is induced by SNAIL in other types of carcinomas, such as pancreatic and breast carcinomas. Furthermore, miR‐34a repressed ZNF281 in a pancreatic cancer cell line. Therefore, the regulations described here may also be important for the regulation of EMT and metastasis in other carcinomas besides CRC.
Our results demonstrate that ZNF281 represents a new EMT‐promoting transcription factor. Notably, ZNF281 is structurally related to the zinc‐finger containing transcription factors SNAIL, SLUG and ZEB1/2 (as indicated in Supplementary Figure S1). The current literature suggests an extensive crosstalk between EMT‐inducing transcription factors and miRNAs during the establishment and maintenance of the mesenchymal phenotype of cells (reviewed in Sanchez‐Tillo et al, 2012 and De Craene and Berx, 2013). Our results further extend this network by adding reciprocal connections between SNAIL, ZNF281 and miR‐34a.
We have previously shown that the oncoprotein c‐MYC binds to ZNF281 (Koch et al, 2007). Here, we found that ectopic expression of c‐MYC increased the ZNF281 expression. The absence of c‐MYC binding sites in the promoter region of ZNF281 implies the existence of alternative regulatory mechanisms. For example, c‐MYC has been shown to directly downregulate miR‐34a in B‐cell lymphoma (Chang et al, 2008). Furthermore, c‐MYC‐induced SNAIL could repress miR‐34, which would contribute to increased ZNF281 expression. Alternatively, the association of c‐MYC protein with ZNF281 may inhibit the turnover of ZNF281. Our results further imply that ZNF281 represents a necessary mediator of SNAIL‐ and c‐MYC‐induced EMT. Therefore, ZNF281 might be an important mediator of tumour progression in tumour entities with deregulation of c‐MYC. It has been shown before that the EMT process yields cells with properties of stem cells (Polyak and Weinberg, 2009; Valastyan and Weinberg, 2011). For example, activation of the EMT inducers SNAIL and ZEB1 promotes the formation of tumour initiating cells with stem‐cell properties (Mani et al, 2008; Wellner et al, 2009; Dang et al, 2011; Hwang et al, 2011). The ZNF281‐induced sphere formation may also involve the previously reported links between ZNF281 and the stemness regulating transcription factors Nanog, Oct4 and Sox2 (Wang et al, 2006, 2008). Recently, it was shown that miR‐34a and its target Notch1 form a bimodal switch which determines whether CRC stem cells (CCSCs) divide symmetrically or asymmetrically and thereby give rise to differentiated cells (Bu et al, 2013). The latter cells showed increased miR‐34a and decreased Notch1 expression, whereas CCSC displayed the reverse expression patterns. It is conceivable that ZNF281 as another miR‐34a target related to stemness may also contribute to fate determination by miR‐34a. The enhancement of β‐catenin activity by ZNF281 may contribute to ZNF281‐induced stemness, since WNT signalling is necessary to maintain stem cells in the intestinal crypts (Pinto and Clevers, 2005; Brabletz et al, 2005b; Scoville et al, 2008). Interestingly, ZNF281 directly binds to the β‐catenin promoter in human multipotent stem cells (hMSCs) (Seo et al, 2013). In addition, we found that ZNF281 directly represses Axin2, a negative regulator of the WNT pathway (Lustig et al, 2002). Thereby, ZNF281 presumably interrupts the negative feedback regulation of β‐catenin by Axin2 and allows β‐catenin to accumulate in the nucleus and activate target genes. However, ZNF281 may also promote β‐catenin/TCF4 activity by mediating the loss of E‐cadherin expression, which is known to inhibit nuclear localization and activity of β‐catenin by recruiting it to the cell membrane (Sadot et al, 1998; Orsulic et al, 1999). Furthermore, our results are consistent with the reported correlation of ZNF281 and β‐catenin expression in hMSCs (Seo et al, 2013). We found that ZNF281 induces expression of the prognostic stem‐cell markers CD133 and LGR5. Colon tumours were shown to contain a subpopulation of CD133‐positive cells with the ability to initiate tumour growth (Horst et al, 2008). Furthermore, high CD133 expression was associated with poor survival of CRC patients. Moreover, the combination of CD133 and the nuclear localization of β‐catenin identified cases of low‐stage CRC with a high risk for tumour progression (Horst et al, 2009). Since downregulation of ZNF281 prevented the formation of lung metastasis of a CRC cell line in a xenograft mouse model, it seems likely that enhancement of EMT and/or stemness by ZNF281 are important functions of ZNF281 during CRC progression. We also observed inhibition of metastasis formation after experimental downregulation of SNAIL in a similar assay (Jackstadt et al, 2013). Therefore, the effect of ZNF281 downregulation might be mediated, at least in part, by decreased expression of SNAIL and the concomitant loss of mesenchymal properties. Furthermore, the downregulation of ZNF281 by p53 via miR‐34a indicates that limiting ZNF281 function is critical for tumour suppression by p53. The increased expression of ZNF281 mRNA in primary colorectal and breast carcinomas also points towards cancer promoting effects of ZNF281. Furthermore, enhanced ZNF281 expression correlated with recurrence 3 years after removal of the primary colorectal tumour, suggesting that detection of elevated ZNF281 expression in primary tumours may have a prognostic value.
Materials and methods
The cell lines HCT‐15, HEK293T, HT29, MiaPaCa2, SKBR3, SW480 and SW620, as well as human diploid fibroblasts (HDFs) were kept in DMEM. DLD‐1, HCT116 p53−/− and HCT116 p53 +/+ cells were cultured in McCoys medium and Colo320 in RPMI medium. Media were supplemented with 10% FCS (Invitrogen) and 1% Penicillin/Streptavidin (Invitrogen) and cells were kept at 5% CO2 and 37 °C. For Hek293T cells, 5% FCS was used. All oligonucleotides (Ambion—Applied Biosystems) were transfected using HiPerfect (Qiagen).
DLD‐1 and SW480 cells harbouring a DOX‐inducible ZNF281 allele or respective miRNAs were cultured for the indicated periods in the presence of DOX (100 ng/ml for pRTR vectors and 1 μg/ml for pRTS vector systems) or left untreated before applying the wound. Mitomycin C (10 ng/ml) was applied 2 h before scratching using a Culture‐Insert (IBIDI). To remove Mitomycin C and detached cells, cells were washed twice with HBSS and medium containing DOX (100 ng/ml or 1 μg/ml) was added where indicated. Cells were allowed to close the wound for the indicated periods and images were captured on an Axiovert Observer Z.1 microscope connected to an AxioCam MRm camera using the Axiovision software (Zeiss) at the respective time points.
Migration and invasion analysis in Boyden chambers
DLD‐1 and SW480 cells harbouring DOX‐inducible pRTR or pRTS vectors were cultured for the indicated periods in the presence or absence of DOX (pRTR: 100 ng/ml; pRTS: 1 μg/ml). Cells were deprived of serum (0.1%) for 48 h before the analysis. To analyse migration, 5 × 104 cells were seeded in the upper chamber (8.0 μm pore size; Corning) in serum‐free medium. To analyse invasion, membranes were coated with Matrigel (BD Bioscience) at a dilution of 3.3 mg/ml in medium without serum. After coating 7 × 104 cells were seeded on Matrigel in the upper chamber. As a chemo‐attractant, 10% FCS was used in the lower chamber. Cultures were maintained for 48 h, then non‐motile cells at the top of the filter were removed and the cells in the lower chamber were fixed with methanol and stained with DAPI. Either the number of cells per well or five different fields per condition were counted by microscopy. Relative invasion/migration was calculated in relation to the control.
Sphere formation assay
Cells were separated by treatment with trypsin after addition of DOX for 48 h. For each triplicate, 1 × 105 cells were seeded into a well of a 6‐well plate coated with attachment preventing poly(2‐hydroxyethyl‐metacrylate) (PolyHEMA, Sigma) in 5 ml sphere medium (Yu et al, 2007). After 7 days, the resulting spheres were documented by phase‐contrast microscopy at × 100 magnification and spheres were dissociated into single cells using a 0.05% trypsin‐EDTA solution. For quantification, 1 × 104 cells/well were seeded in Yu medium containing 1% methyl cellulose (Sigma) into PolyHEMA coated 96‐well plates in the presence or absence of DOX (n=6). The number of colonies larger than 50 μm in diameter were counted after 7 days.
In vivo lung metastasis assay
72 h after siRNA transfection 4 × 106 SW620‐Luc2 cells were resuspended in PBS in a total volume of 0.2 ml and injected into the lateral tail vein of a 6‐ to 8‐week‐old age‐matched male NOD/SCID mouse using a 25‐gauge needle. For monitoring of the injected cells, mice were injected intraperitoneally with d‐luciferin (150 mg/kg) and imaged under anaesthesia with the IVIS Illumina System (Caliper Life Sciences) 30 min after tail vein injection in order to have a reference point. The acquisition time was set to 2 min and imaging was repeated once a week to monitor the seeding and outgrowth of the cells. After 9 weeks complete lungs were resected and photographed. For H&E stainings, lungs were fixed with 4% paraformaldehyde and 5 μm sections were stained with heamatoxy‐lin and eosin. The number of metastases was determined microscopically. Mice were kept under IVC conditions and experiments were performed with permission of the Bavarian state (file number: 55.2‐1‐54‐2532.8‐188‐11).
Unless noted otherwise, each experiment was carried out in triplicates. A Student's t‐test (unpaired, two‐tailed) was used for calculation of significant differences between two groups of samples, with P<0.05 considered as significant. Asterisks generally indicate: *P<0.05, **P<0.01 and ***P<0.001. For correlation analyses, the SPSS software package 19 (SPSS Inc.) was used. Spearman rank correlation test was applied to the NCI‐60 data in order to correlate mRNA expression with the expression of other mRNAs.
Additional detailed Materials and methods are available online. Oligonucleotides used for ChIP analysis and for cloning and mutagenesis are listed in Supplementary Tables S2 and S3, respectively. Expression plasmids are listed in Supplementary Table S4, primers used for qPCR in Supplementary Table S5, and antibodies used for IF, ChIP and WB analysis are listed in Supplementary Table S6.
Conflict of Interest
The authors declare that they have no conflict of interest.
Source data for Suppl. Figure S2
Source data for Suppl. Figure S3
Source data for Suppl. Figure S4
Source data for Suppl. Figure S6
Source data for Suppl. Figure S9
We thank Juanita Merchant, Christopher Contag, Antonio Garcia de Herreros and Lionel Larue for kindly providing plasmids, Ru Zhang for antibody testing and plasmids, Markus Kaller for plasmids and technical advice, Ursula Götz for assistance, Matjaz Rokavec for technical advice and Stefanie Jaitner for NOD/SCID mice. This work was supported by a grant of the Deutsche Forschungsgemeinschaft (DFG; He2701/9‐1) to Heiko Hermeking.
- Copyright © 2013 European Molecular Biology Organization