Research ArticleCancer

Ras and TGF-β signaling enhance cancer progression by promoting the ΔNp63 transcriptional program

See allHide authors and affiliations

Science Signaling  23 Aug 2016:
Vol. 9, Issue 442, pp. ra84
DOI: 10.1126/scisignal.aag3232

Metastatic convergence at ΔNp63

The p53 family of transcription factors, which includes p53, p63, and p73, is implicated in both tumor-suppressive and tumorigenic functions. Activation of the Ras and transforming growth factor–β (TGF-β) signaling pathways are similarly enigmatic with both tumor-suppressing and tumor-promoting activity. Deciphering their roles in various stages of tumor development is critical to developing targeted therapeutics. Vasilaki et al. found that these factors are all connected in the development of some cancers. Activation of Ras or TGF-β signaling stimulated the transcriptional activity of the isoform ΔNp63 in breast or squamous cancer cells by suppressing the abundance of the mutant form of p53, an inhibitor of ΔNp63 and also a feature implicated in early tumorigenesis. Increased abundance of ΔNp63 in breast cancer cells stimulated metastatic behaviors in culture and in mice and correlated with poor prognosis in patients with mutant p53–positive tumors. These findings provide some insight into dichotic, stage-specific signals in tumorigenesis. Identification of this downstream common effector could also offer new therapeutic opportunities for some advanced Ras- or TGF-β–driven tumors.

Abstract

The p53 family of transcription factors includes p63, which is a master regulator of gene expression in epithelial cells. Determining whether p63 is tumor-suppressive or tumorigenic is complicated by isoform-specific and cellular context–dependent protein associations, as well as antagonism from mutant p53. ΔNp63 is an amino-terminal–truncated isoform, that is, the predominant isoform expressed in cancer cells of epithelial origin. In HaCaT keratinocytes, which have mutant p53 and ΔNp63, we found that mutant p53 antagonized ΔNp63 transcriptional activity but that activation of Ras or transforming growth factor–β (TGF-β) signaling pathways reduced the abundance of mutant p53 and strengthened target gene binding and activity of ΔNp63. Among the products of ΔNp63-induced genes was dual-specificity phosphatase 6 (DUSP6), which promoted the degradation of mutant p53, likely by dephosphorylating p53. Knocking down all forms of p63 or DUSP6 and DUSP7 (DUSP6/7) inhibited the basal or TGF-β–induced or epidermal growth factor (which activates Ras)–induced migration and invasion in cultures of p53-mutant breast cancer and squamous skin cancer cells. Alternatively, overexpressing ΔNp63 in the breast cancer cells increased their capacity to colonize various tissues upon intracardiac injection in mice, and this was inhibited by knocking down DUSP6/7 in these ΔNp63-overexpressing cells. High abundance of ΔNp63 in various tumors correlated with poor prognosis in patients, and this correlation was stronger in patients whose tumors also had a mutation in the gene encoding p53. Thus, oncogenic Ras and TGF-β signaling stimulate cancer progression through activation of the ΔNp63 transcriptional program.

INTRODUCTION

The transcription factor p63 is a member of the p53 family of transcription factors and shares a conserved basic domain structure and a high amino acid similarity with p53 and p73, but p63 exerts distinct functions (13). Although the developmental aspects of p63 function as the epidermal master regulator are well defined, the roles of p63 in cancer are unclear. In contrast to the high mutation rate of p53 (4), mutations of p63 are rare in human cancers, indicating that p63 is not a canonical tumor suppressor.

The existence of multiple p63 isoforms that are differentially expressed in the cells and their ability to tetramerize with the other p53 members drastically expands the complexity. The TP63 gene encodes multiple isoforms of p63. The two different promoters (P1 and P2) lead to the generation of two classes of proteins: TAp63, which contains the N-terminal transactivation (TA) domain, and ΔNp63, an N-terminal truncated isoform, which lacks the TA domain. Alternative splicing at the 3′ end of the transcripts results in five different isoforms (α to ε) (2). TAp63 induces cell cycle arrest and cell death after DNA damage, which is commonly observed in the p53 family, and functions as a tumor suppressor at least in mice (3). The ΔNp63 isoforms have been shown to act antagonistically to TAp63 and p53 (3), suggesting that they have oncogenic properties. Consistently, ΔNp63 has been proposed to function as a potent oncogene and tumor-initiating factor in squamous cell carcinomas (SCCs) of diverse origins and in breast cancer cells, modulating proliferation, survival, and migration (1). Thus, p63 has both tumor-suppressive and tumor-promoting properties depending on the isoforms expressed and the cellular context.

In addition, p53 mutation status affects the function of p63. Tumor-derived p53 mutants are stable with higher binding affinity for p63. Interaction between mutant p53 and p63 involves the formation of heteromeric complexes, resulting in decreased transcriptional activity of p63 and neutralization of p63 function (3). Furthermore, Adorno and colleagues proposed that activation of transforming growth factor–β (TGF-β) and Ras–mitogen-activated protein kinase (MAPK) signaling pathways promotes the assembly of a ternary complex involving mutant p53, p63, and SMAD, a transcription factor that mediates TGF-β signaling, and that the ternary complex antagonizes the functions of p63 and empowers TGF-β–induced metastasis (5).

Because TGF-β conveys cytostatic and cell death processes in mature tissues, it functions as a suppressor of epithelial cell tumorigenesis at early stages. On the other hand, accumulating evidences support the notion that TGF-β is converted to an enhancer of tumor invasion and metastasis in more advanced stages of carcinogenesis (6). The Ras-MAPK pathway, where mutations are accumulating in human cancers (4), regulates cell migration and invasion synergistically with TGF-β (6). These facts suggest that clarifying the crosstalk among TGF-β, Ras-MAPK, and p63 will unknot the complex molecular mechanisms underlying their context-dependent regulation of cell migration and invasion.

Here, we performed p63 genome-wide chromatin immunoprecipitation and sequencing (ChIP-seq) analyses in epithelial cells, which have mutant p53 and express predominantly ΔNp63. We found that Ras and TGF-β activation enhanced the strength of p63 binding to its genomic sites through suppressing the abundance of mutant p53. Enhanced p63 activity increased the expression of dual-specificity phosphatases DUSP6 and DUSP7 (hereafter, DUSP6/7), which unexpectedly promoted migration and invasion in cells in culture and in vivo. Using a series of phosphorylation site mutants of a mutant p53 protein, we have found that TGF-β or Ras and DUSP6 differentially modulated the phosphorylation status and stability of mutant p53. Our study therefore illustrates a model of a mutual inhibitory network between TGF-β, oncogenic Ras, ΔNp63, and mutant p53.

RESULTS

DNA binding of p63 is enhanced by activation of Ras and TGF-β signaling at a genome-wide level

We determined the genome-wide landscape of p63 binding using ChIP-seq in HaCaT keratinocytes, which carry the p53-H179Y/R282W mutations, have wild-type Ras, and express predominantly ΔNp63 (5). Through the analysis, 10,160 peaks representing p63 binding sites were identified. A de novo motif-finding method predicted a p63 binding motif (Fig. 1A), which resembled a previously reported p63 motif (7). Around 55% of identified peaks contained at least one motif within 300 base pairs (bp) around the peak summits (Fig. 1B). Moreover, the p63 binding motif was distributed around peak summits in all ChIP-seq data derived from the four biological conditions examined (Fig. 1C), confirming that the p63 ChIP-seq data are valid.

Fig. 1 p63 binding to the genome is enhanced by activation of Ras and TGF-β signaling at a genome-wide level.

(A) Sequence logo of a putative p63 binding motif. (B) Summary of the number of p63 binding peaks in each ChIP-seq condition and enrichment of the p63 binding motif. (C) Histogram representation of the location of the p63 binding motif in p63 binding regions. The p63 binding motif was mapped in regions within 500 bp from the peak summits or random control genomic regions. Fold change of the number of the motif relative to control genomic regions was calculated and presented (bin 10). (D) Venn diagram representing overlap between p63 binding sites in control cells (LacZ untreated) and those in TGF-β–treated cells. (E) Venn diagram representing overlap between p63 binding sites in HaCaT cells infected with LacZ, caRas, or dnRas adenovirus in the presence of TGF-β stimulation. (F) Validation of the ChIP-seq data by ChIP-qPCR (quantitative polymerase chain reaction). Results of n = 2 independent experiments are shown by dot plot chart. Horizontal lines in the chart indicate average values. (G) Histogram representation of the effect of TGF-β stimulation and Ras mutant overexpression on the strength of p63 binding to its genomic sites. Fold change of normalized sequence read counts, which represents binding strength of p63, was calculated, and the number of peaks was counted (bin 0.2).

We then evaluated the effects of Ras and TGF-β signaling on p63 binding to the genome. Although our previous ChIP-seq analysis of TGF-β/Smad signaling revealed that binding of Smad2 and Smad3 (Smad2/3) to their target genomic sites reaches a maximum after 1.5 hours of TGF-β treatment in HaCaT cells (8), the genome-wide binding pattern of p63 was not much affected by 1.5 hours of TGF-β stimulation (Fig. 1, B and D); 82% of the p63 binding peaks were unchanged in TGF-β–stimulated compared to TGF-β–unstimulated cells. This suggests that nuclear accumulation of activated Smad proteins does not redirect p63 to different binding sites.

Then, we modified Ras activity using adenovirally expressed constitutively active Ras (caRas; H-Ras G12V) or dominant-negative Ras mutants (dnRas; H-Ras N116Y) (9), in addition to TGF-β treatment. Activation of the Ras pathway did not appreciably change the number of the p63 binding peaks; about 80% of the peaks were shared between the control (LacZ + TGF-β) and caRas-overexpressed (caRas + TGF-β) conditions (Fig. 1, B and E). On the other hand, introduction of dnRas decreased the number of p63 binding peaks, although most of them overlapped with those of LacZ control and caRas-transfected cells (Fig. 1, B and E). This suggests that Ras activation affects the strength of the p63 binding to the genome, rather than redirecting p63 to novel target sites. We investigated several p63 binding peaks and confirmed that activation of Ras and TGF-β signaling enhanced p63 binding to these target genomic sites using ChIP-qPCR (Fig. 1F). In addition, calculation of the fold change of normalized sequence read counts revealed that the distribution was shifted toward the right (meaning that many p63 binding peaks had enhanced binding) after treatment with TGF-β in the presence of caRas, whereas it was shifted toward the left in the presence of dnRas (Fig. 1G). The effect of TGF-β treatment was not strong. In contrast, p63 binding was strongly enhanced by Ras activation; the median of caRas/dnRas was 20.91 (=1.87)–fold change. Thus, at a genome-wide level, p63 binding to the genome is generally enhanced by activation of Ras and TGF-β signaling in HaCaT cells.

Down-regulation of mutant p53 abundance by activation of Ras and TGF-β signaling enhances p63 binding to the genome

It is widely accepted that mutant p53 can physically interact with p63 and neutralizes p63 function (3, 10). We investigated whether activation of Ras and TGF-β signaling strengthens the binding of p63 to the DNA through down-regulation of the mutant p53 abundance. Silencing of mutant p53 in HaCaT cells strongly enhanced p63 binding to the genome (Fig. 2, A and B). The overexpression of caRas in combination with TGF-β stimulation decreased the abundance of mutant p53 protein, whereas overexpression of dnRas increased the abundance of mutant p53 in HaCaT cells (Fig. 2C).

Fig. 2 Down-regulation of mutant p53 by activation of Ras and TGF-β signaling enhances p63 binding to the genome.

(A) ChIP was performed using an antibody against p63α in HaCaT cells transfected with either two different small interfering RNAs (siRNAs) specific for p53 (si-p53 #1 and si-p53 #2) or control siRNA (si-control). The ChIP samples were quantified by qPCR with locus-specific primers (table S3) (data are means ± SD; n = 3 experiments; *P < 0.05 by Student’s t test with Bonferroni correction). (B) Western blot for mutant p53 protein in HaCaT cells treated with either siRNAs specific for p53 or control siRNA. α-Tubulin was used as a loading control. Data are representative of n = 3 experiments. (C) Western blot for mutant p53 and phosphorylated extracellular signal–regulated kinase 1 and 2 (pERK1/2) in HaCaT cells infected with LacZ, caRas, or dnRas adenoviruses and treated, or not, with TGF-β for 1.5 hours. Data are representative of n = 5 experiments. (D) Western blot for mutant p53 in HaCaT cells pretreated, or not, with MG132 (10 μM, 6 hours) and treated with TGF-β for 1.5 hours. Data are representative of n = 3 experiments. (E) qRT-PCR analysis of HaCaT cells infected with control (LacZ), caRas, or dnRas adenoviruses and treated, or not, with TGF-β for 1.5 hours. GAPDH was used as endogenous control, and data were normalized to the control condition (data are means ± SD, n = 3 experiments). (F) Western blot for mutant p53 and pERK1/2 in HaCaT cells pretreated, or not, with the MEK1/2 inhibitor U0126 (10 μM, 24 hours) and treated, or not, with TGF-β for 1.5 hours. Data are representative of n = 3 experiments. (G) Western blot for mutant p53 in HaCaT cells transfected with either siRNAs specific for MDM2 or control siRNA and treated, or not, with TGF-β for 1.5 hours. Data are representative of n = 2 experiments. (H) Western blot for mutant p53 in HaCaT cells pretreated, or not, with the MDM2 inhibitor Nutlin-3 (10 μM, 24 hours) and treated with TGF-β for 1.5 hours. Data are representative of n = 3 experiments. (I) Western blot for mutant p53 in HaCaT cells with increasing amounts of the dominant-negative mutant of MDM2 (dnMDM2) and treated, or not, with TGF-β for 1.5 hours. Data are representative of n = 2 experiments.

We next sought to address how activation of Ras and TGF-β signaling down-regulates the mutant p53 abundance in HaCaT cells. Application of the proteasome inhibitor MG132 rescued the caRas- and TGF-β–induced down-regulation of the abundance of mutant p53 protein, although we did not detect a significant change in the amount of TP53 mRNA (which encodes p53 protein) under the same conditions (Fig. 2, D and E). This indicates that the down-regulation of mutant p53 abundance occurs mainly at the posttranscriptional, rather than the transcriptional, level. The degradation of mutant p53 was restored when cells were treated with U0126, a selective inhibitor of MAPK kinase 1 and 2 (MEK1/2) (Fig. 2F). Because activation of Ras and TGF-β signaling induces and activates MDM2 (11), one of the principal ubiquitin ligases responsible for targeting p53 for degradation, we performed loss-of-function experiments for MDM2. Treatment with siRNA against MDM2, or an MDM2 inhibitor, Nutlin-3, showed only weak effects (Fig. 2, G and H). On the other hand, overexpression of the C438A MDM2 mutant, which lacks ubiquitin ligase activity and acts in a dominant-negative manner by interfering with the effect of MDM2 and other E3 ligases, partially rescued the caRas- and TGF-β–induced degradation of mutant p53 (Fig. 2I). These findings are consistent with those shown in previous reports that mutant p53 protein is not as sensitive to MDM2-mediated degradation as is wild-type p53 and that multiple ubiquitin ligases are involved in mutant p53 degradation (12).

Ras and TGF-β regulate distinct sets of genes in a p63-dependent manner

To determine how p63 function is affected by the caRas- and TGF-β–induced degradation of mutant p53, we combined ChIP-seq and microarray expression data in HaCaT cells. We first identified the genes in which the binding of p63 was increased by caRas and TGF-β, and then we categorized these genes based on the effects of Ras and TGF-β signaling, as well as of mutant p53 depletion, on their mRNA expression profile. As indicated in Fig. 3A, two subgroups were isolated; in subgroup 1, p63 functions as an activator, whereas in subgroup 2, p63 functions as a repressor. Distinct gene ontologies (GOs) were enriched in these two groups; genes in subgroup 1 were associated with cell motility and participate in epidermal growth factor (EGF)–EGF receptor (EGFR)–MAPK signaling, whereas immune response and cell proliferation were overrepresented GO terms in subgroup 2 (Fig. 3B). The most highly regulated genes from each subgroup are presented in Fig. 3C, and all the genes from both subgroups are shown in tables S1 and S2. Two genes from each subgroup were validated using quantitative reverse transcription PCR (qRT-PCR) (Fig. 3, D and E).

Fig. 3 Identification of direct target genes of p63 whose expression is regulated by Ras and TGF-β signals.

(A) Categorization of 4778 genes identified through ChIP-seq analysis to which p63 binding was enhanced by both Ras and TGF-β signaling activation. Subgroup 1 (in which p63 functions as an activator) contains genes whose mRNA expression was up-regulated by p63, caRas, and TGF-β signaling and down-regulated by mutant p53 (p53 mut). Subgroup 2 (in which p63 functions as a repressor) contains genes whose mRNA expression was down-regulated by p63, caRas, and TGF-β signaling and up-regulated by mutant p53. (B) GO enrichment analysis of each subgroup of p63 direct target genes. Top five GOs are presented. (C) Representation of the most highly regulated genes from each subgroup based on their sensitivity to activation of Ras and depletion of p53. The values from the caRas/dnRas and si-control/si-p53 ratios are shown. (D and E) Validation of the microarray data by qRT-PCR analysis. HaCaT cells were infected with LacZ, caRas, or dnRas and treated, or not, with TGF-β for 24 hours (D) or transfected with siRNAs specific for either p53, p63, or control siRNA (E). Data are means ± SD; n = 3 experiments; *P < 0.05, **P < 0.01, ***P < 0.001 by Student’s t test with Bonferroni correction.

The antibody used in the ChIP-seq analyses recognizes both TAp63α and ΔNp63α, whereas HaCaT cells predominantly express ΔNp63. To confirm whether ΔNp63 functioned as both a transcriptional activator and a repressor, we designed and used the siRNA specific for the ΔNp63 isoforms. Knockdown of the ΔNp63 isoforms resulted in the same effects as the total p63 siRNA on the genes in each subgroup (fig. S1, A and B). Thus, the genes identified in our selection are downstream targets of Ras and TGF-β pathways, whose expression is modulated in a ΔNp63-dependent manner.

p63 is necessary for the up-regulation of DUSP6 and DUSP7 by active Ras and TGF-β signals

Because GO analysis suggested that genes in subgroup 1 are involved in EGF-EGFR signaling (Fig. 3B), we focused on DUSP6 (also known as MKP3/PYST1) and DUSP7 (MKP-X/PYST2) (Fig. 3C). DUSP6 and DUSP7, together with DUSP9 (MKP4/PYST3) whose abundance is mainly restricted to stem cells, belong to a subfamily of dual-specificity phosphatases; all three members inactivate ERK1 and ERK2 (ERK1/2) (13).

The binding of p63 to the DUSP6 and DUSP7 loci was enhanced by caRas and TGF-β stimulation and decreased by dnRas in HaCaT cells (Fig. 4, A and B). On the other hand, knockdown of mutant p53 enhanced the binding of p63 to the DUSP6 and DUSP7 loci (Fig. 4B). In accordance with the p63 binding, the abundance of DUSP6 and DUSP7 was increased at both the mRNA and protein levels by caRas expression and TGF-β stimulation, whereas overexpression of dnRas suppressed each in HaCaT cells (Fig. 4, C and D). In addition, knockdown of p63 reduced the abundance of DUSP6 and DUSP7 (at both the mRNA and protein levels) in HaCaT cells, whereas knockdown of mutant p53 increased each (Fig. 4, C and E). Double knockdown of mutant p53 and p63 prevented the effects of depletion of mutant p53 (Fig. 4, C and E). Moreover, introduction of a DNA binding domain mutant of p53, p53-R175H, attenuated p63 binding to the DUSP6 locus, whereas overexpression of ΔNp63α enhanced the amounts of DUSP6 and DUSP7 in HaCaT cells (Fig. 4, F and G). We introduced siRNA against p53, which targets the 3′ untranslated region (3′UTR) of the endogenous transcript, and performed rescue with wild-type and mutant (R175H) p53 complementary DNA (cDNA) constructs lacking the 3′UTR (Fig. 4H). Knockdown of p53 induced the abundance of DUSP6, which was canceled by introduction of p53-R175H but not of wild-type p53; rather, wild-type p53 induced the abundance of DUSP6 in the control conditions. Thus, mutant p53, which lacks DNA binding capacities, antagonizes ΔNp63 in HaCaT cells, and down-regulation of mutant p53 results in enhanced binding of ΔNp63 to the genome.

Fig. 4 p63 is necessary for the up-regulation of DUSP6 and DUSP7 by active Ras and TGF-β signals.

(A) The genomic loci of DUSP6 and DUSP7 are illustrated along with the p63 binding peaks obtained from the ChIP-seq analysis. (B) ChIP was performed using an antibody against p63α in HaCaT cells infected with LacZ, caRas, or dnRas adenoviruses for 48 hours and treated, or not, with TGF-β for 1.5 hours (left) or transfected with either control or siRNAs targeting p53 (right). qPCR was performed to quantify the fold enrichment of p63 at the DUSP6 or DUSP7 gene loci. Data are means ± SD; n = 3 experiments; *P < 0.05, **P < 0.01 by Student’s t test with Bonferroni correction. (C) qRT-PCR analysis of HaCaT cells infected with LacZ, caRas, or dnRas adenoviruses and treated, or not, with TGF-β for 24 hours (left) or transfected with siRNAs specific for p53 and/or p63 or control siRNA, as indicated (right). GAPDH was used as endogenous control, and data were normalized to the control condition. Data are means ± SD; n = 3 experiments; *P < 0.05, **P < 0.01, ***P < 0.001 by Student’s t test with Bonferroni correction. (D) Western blot for DUSP6 and DUSP7 in HaCaT cells infected with LacZ, caRas, or dnRas adenoviruses and treated, or not, with TGF-β for 24 hours. Data are representative of n = 4 experiments. (E) Western blot for DUSP6 in HaCaT cells transfected with siRNAs specific for p53 and/or p63 or control siRNA, as indicated. Data are representative of n = 3 experiments. (F) ChIP was performed using an antibody against p63α in HaCaT cells co-infected, or not, with the adenovirus expressing p53-R175H and treated, or not, with TGF-β for 1.5 hours. qPCR was performed to quantify the fold enrichment of p63 at the DUSP6 locus relative to LacZ + TGF-β condition. Data are means ± SD; n = 2 experiments; *P < 0.05 by Student’s t test. (G) Western blot for DUSP6 and DUSP7 in HaCaT cells transfected with vector expressing ΔNp63α or control. Data are representative of n = 3 experiments. (H) Western blot for DUSP6 in HaCaT cells transfected with siRNA specific for endogenous p53 (si-p53 3′UTR) and infected with adenoviruses expressing p53-R175H or wild-type p53, as indicated. Data are representative of n = 3 experiments. (I) DUSP6 reporter assays in HaCaT cells cotransfected with the wild-type (wt) or mutant (mut) DUSP6 reporter plasmids and vectors expressing ΔNp63α or control, as indicated (left), or cotransfected with siRNA specific for p53 or control siRNA and vectors expressing ΔNp63α or control (right), as indicated. Relative luciferase activity values were normalized to the control condition. Data are means ± SD; n = 3 experiments; *P < 0.05, **P < 0.01, ***P < 0.001 by Student’s t test with Bonferroni correction.

To generalize our findings, we used the skin squamous carcinoma (SCC) cell line A431, which harbors p53-R273H (10) and amplified wild-type EGFR, and predominantly expresses ΔNp63. The expression of DUSP6 and DUSP7 was modulated by activation of Ras and TGF-β signaling and depletion of mutant p53, whereas the abundance of mutant p53 protein was also down-regulated by activation of Ras and TGF-β signaling in A431 cells (fig. S2, A to D). We also used Ras-transformed MCF10A-MII breast epithelial cells expressing wild-type p53. In contrast to the cells with mutations in the TP53 gene, knockdown of endogenous wild-type p53 suppressed the abundance of DUSP6 in MCF10A-MII cells at the mRNA and protein levels (fig. S2, E and F), whereas overexpression of p53-R175H abolished the abundance of DUSP6 (fig. S2G). We conclude that induction of the DUSP6 gene depends on the DNA binding properties of p53 and p63.

To further characterize functions of the ΔNp63 isoforms, we performed luciferase reporter assays. The DUSP6 promoter carries a putative p63 binding site at around −140 bp from the transcription start site; this region was cloned into luciferase constructs (DUSP6-Luc) (Fig. 4I and fig. S2H). Overexpression of ΔNp63α led to increased luciferase activity of the DUSP6-Luc in HaCaT and A431 cells. Mutation of the four most conserved base pairs in the p63 DNA binding motif attenuated the ΔNp63α-induced luciferase activity in HaCaT and A431 cells. Knockdown of mutant p53 increased the luciferase activity in both the basal condition and after overexpression of ΔNp63α in HaCaT and A431 cells. Our observations support the notion that ΔNp63 directly regulates DUSP6 by binding to and activating the DUSP6 promoter. Our data also confirmed that both wild-type p53 and ΔNp63α can up-regulate DUSP6 (14, 15).

ΔNp63 and its target genes, DUSP6 and DUSP7, aggravate the cancer cell metastasis in vivo

To explore the potential role of the ΔNp63 and its target genes, DUSP6 and DUSP7, in migration and invasion in vivo, we used both gain-of-function and loss-of-function experiments using A431 cells with ΔNp63α, as well as the highly bone metastatic triple-negative breast cancer cell line MDA-MB-231-D (16), which harbors p53-R280K and constitutively active mutations in Ras-MAPK pathways, that is, K-Ras-G13D/BRaf-G464V, and mainly expresses the TAp63 isoform (5, 10).

First, we stably expressed ΔNp63α or green fluorescent protein (GFP) as a control in MDA-MB-231-D cells using lentivirus. The ΔNp63α-expressing cells (designated 231D-ΔNp63) showed enhanced migration compared to the control cells (231D-GFP) (Fig. 5A and fig. S3, A and B), although no noticeable difference was observed on the proliferation of these cells (fig. S3, C and D). Silencing of both DUSP6 and DUSP7 expression led to decreased motility of the cells, and the effect was more prominent in the 231D-ΔNp63 cells (Fig. 5A). Thus, our ΔNp63 gain-of-function experiments in cultured cells suggest that ΔNp63 and its target genes, that is, DUSP6 and DUSP7, promote cellular migration without affecting cell proliferation. We also performed three-dimensional (3D) cell invasion assays using A431 cells to evaluate the role of ΔNp63. Ablation of ΔNp63 or DUSP6/7 expression significantly perturbed not only the TGF-β– and EGF-induced 2D migration (Fig. 5B) but also the EGF- and TGF-β–induced 3D collagen invasion of A431 cells (Fig. 5C and fig. S3E). This ΔNp63 loss-of-function experiment also suggests an important role of ΔNp63 in enhancing the motile and invasive phenotype of these cells.

Fig. 5 ΔNp63 and its target genes, DUSP6 and DUSP7, promote metastasis in vivo.

(A) Wound-healing assay of MDA-MB-231-D cells stably expressing ΔNp63α or GFP as control and transfected with control siRNA or with siRNAs targeting DUSP6 and DUSP7 (DUSP6/7). Images were taken at time 0 hour and after 24 hours of incubation and analyzed by TScratch software. Relative values are means ± SD from n = 3 experiments. ***P < 0.001 by Student’s t test with Bonferroni correction. (B) Migration assays of A431 cells transfected with control siRNA or with siRNAs specific for p63, DUSP6, and DUSP7 in the presence or absence (−) of TGF-β or EGF stimulation. The relative values are means ± SD from n = 3 experiments. ***P < 0.001 by Student’s t test with Bonferroni correction. (C) 3D invasion assay of spheroids of A431 cells, transfected with control siRNA or with siRNAs against p63 or DUSP6 and DUSP7, embedded in collagen in the presence or absence (−) of TGF-β and/or EGF as indicated. Data are representative of two independent experiments. Results are means ± SD from n > 6 spheroids. *P < 0.05, ***P < 0.001 by Student’s t test with Bonferroni correction. (D and E) MDA-MB-231-D cells stably expressing ΔNp63α or GFP and transfected with control shRNA or shRNAs targeting DUSP6 and DUSP7 (sh-DUSP6/7) were intracardially injected into nude mice [number as indicated (n)]. Representative bioluminescent images of the mice used in the study on day 35 (D) and Kaplan-Meier analysis of overall survival of the mice (E) are shown. Survival analysis was performed using the log-rank test. In (D), mice died before image acquisition due to anesthesia (GFP-sh-control, n = 1) or tumor progression (ΔNp63-sh-control, n = 3). cpm, counts per minute.

To confirm the roles of ΔNp63 and its target genes, DUSP6 and DUSP7, in invasiveness in vivo, we injected 231D-ΔNp63 or 231D-GFP cells expressing either control short hairpin RNA (shRNA) or shRNAs targeting DUSP6 and DUSP7 into the left ventricle of the heart of nude mice (16). ΔNp63α promoted metastasis, as displayed by the increased number and size of metastatic nodules (Fig. 5D). In contrast, depletion of DUSP6/7 attenuated the effect of ΔNp63α (Fig. 5D). In line with this, mice injected with 231D-ΔNp63-sh-control cells showed a much shorter survival than those injected with 231D-GFP-sh-control or 231D-ΔNp63-sh-DUSP6/7 cells (Fig. 5E). Together, the data suggest that ΔNp63 plays a crucial prometastatic role in cancer cells with a mutant p53 genetic background.

High expression of TP63 is associated with poor prognosis in patients with ΔNp63-dominant cancer

To determine the prognostic significance of ΔNp63, we analyzed patient data sets from The Cancer Genome Atlas (TCGA) for head and neck SCC (HNSCC); overexpression of ΔNp63 was observed in 81 to 100% of the cases (2). In the HNSCC patients, overall survival was significantly different among patient groups separated by mutation status of the TP53 gene and mRNA expression of TP63 (Fig. 6A); mutations in TP53 and high expression of TP63 were associated with poor prognosis, whereas high expression of TP63 in the wild-type TP53 background was associated with better prognosis (Fig. 6A and fig. S4A). On the other hand, this trend was not observed in breast cancer cells that express both TAp63 and ΔNp63 isoforms depending on the subtypes (Fig. 6B and fig. S4B) (2). Patients with high TP63 expression in the mutant TP53 background experienced relatively low survival in the basal-like subtype, which mainly expresses ΔNp63, although the difference was not significant, possibly because of the small number of patients (Fig. 6B).

Fig. 6 High expression of TP63 is associated with poor prognosis in patients with ΔNp63-high cancer.

(A and B) Kaplan-Meier analysis of overall survival in the HNSCC data set (A) and breast cancer data set (B) of TCGA (44, 45), based on p53 mutation status and TP63 expression in the primary tumor of the patients. Survival analysis was performed using the log-rank test. (C) Kaplan-Meier analysis of overall survival in the HNSCC data set of TCGA. The HNSCC patients with mutant p53 were classified on the basis of TP63 expression and pERK1/2 amounts in the primary tumor. Survival analysis was performed using the log-rank test. (D) DUSP6, DUSP7, and TP63 mRNA expression in the primary tumor of the HNSCC patients with mutations in the TP53 gene. Z-scored expression values of mRNA were obtained with cBioPortal (***P < 0.001; n.s., not significant; Welch t test with Bonferroni correction) (42, 43). (E) Abundance of p53 and pERK1/2 in the primary tumor of the HNSCC patients with mutations in the TP53 gene. Relative abundance was obtained using cBioPortal (*P < 0.05, ***P < 0.001, Welch t test with Bonferroni correction).

We also evaluated the effects of Ras-MAPK activation. In HNSCC patients with mutations in TP53, the amounts of phosphorylated ERK1/2 (pERK1/2) clearly separated the outcome of the TP63-high patient group (Fig. 6C). In accordance with our analysis, the “TP63-high, pERK1/2-high” cohort had lower p53 protein expression values compared with the “TP63-low, pERK1/2-low” cohort (Fig. 6, D and E). That is, in the mutant p53 and ΔNp63 genetic background, patients with low mutant p53 protein had the poorest prognosis (Fig. 6, C and E). Consistent with our results in cultured cells, expression of DUSP6/7 appears to be TP63-dependent in the pERK1/2-high cohort (cohort 3 versus cohort 4 in Fig. 6D). However, DUSP6 and DUSP7 were expressed at comparable degrees in the pERK1/2-low cohort, although they are well-established targets of pERK1/2, suggesting that other signaling pathways may contribute to DUSP6/7 induction in cancer patients in vivo.

DUSP6 modulates the phosphorylation status and stability of mutant p53 protein

Our data clearly showed that Ras and TGF-β signaling aggravate cancer progression through activation of the transcriptional program of the ΔNp63 isoforms through degrading mutant p53. One remaining question is the role of DUSP6 and DUSP7 in cancer progression. Because DUSP6/7 have been established as ERK1/2 MAPK-specific phosphatases, it is possible that DUSP6/7 induction is part of a negative feedback mechanism. However, recent reports suggest the existence of non-MAPK substrates of DUSP6 (13), which may convey tumor-promoting properties as observed in our in vitro and in vivo experiments (Fig. 5 and fig. S3). Because another member of the DUSP family of phosphatases, DUSP26, dephosphorylates p53 and inhibits p53 tumor suppressor functions in neuroblastoma cells (17), we sought to address whether mutant p53 protein is a target of DUSP6/7.

First, depletion of DUSP6 attenuated the binding of p63 to its genomic sites (Fig. 7A). We then investigated whether mutant p53 and DUSP6 physically interact in HaCaT cells by coimmunoprecipitation; we detected physical interaction between the two endogenous proteins (Fig. 7, B and C), consistent with the possibility that DUSP6/7 dephosphorylate mutant p53. However, both phosphorylation mediated by ERK1/2 and dephosphorylation by DUSP6/7 promote degradation of mutant p53 protein, generating conflicting conclusions. To gain insights into the role of the phosphorylation status of mutant p53 for its stability, we introduced a nonphosphorylatable alanine residue in each of the well-characterized serine phosphorylation sites in p53-R273H mutant (18). Exogenous p53-R273H was degraded in the presence of active Ras and TGF-β signaling, although depletion of DUSP6 stabilized it (fig. S5, A and B). Among the phosphorylation site mutants of p53-R273H, S20A was resistant to caRas- and TGF-β–induced degradation, whereas S9A and S37A were resistant to si-DUSP6–induced stabilization (fig. S5, A to C). Consistently, cycloheximide chase analyses revealed that the S9A mutation enhanced degradation of p53-R273H protein, whereas the S20A mutation stabilized it (Fig. 7, D and E). Thus, the ΔNp63-DUSP6/7 module controls the phosphorylation status of the p53 protein, promoting the release of ΔNp63 protein from mutant p53.

Fig. 7 DUSP6 modulates the phosphorylation status and stability of mutant p53 protein.

(A) ChIP was performed using an antibody against p63α in HaCaT cells, transfected with siRNAs specific for DUSP6 or control siRNA, infected with LacZ or caRas adenoviruses for 48 hours, and treated, or not, with TGF-β for 1.5 hours. qPCR was performed to quantify the fold enrichment of p63 at the indicated gene loci. Data are representative of n = 2 independent experiments. (B and C) Endogenous coimmunoprecipitation of p53 and DUSP6 in HaCaT cells after transfection with control, p53-specific, or DUSP6-specific siRNA. Data are representative of n = 3 experiments. (D and E) Stability of p53-R273H protein and its phosphorylation site mutants. Cycloheximide (CHX) chase experiments were performed with A431 cells stably expressing hemagglutinin (HA)–tagged p53-S9A-R273H (D) and p53-S20A-R273H protein (E). Because p53-S20A-R273H protein is stable, the cells were infected with caRas adenoviruses and treated with TGF-β for 24 hours (E).

DISCUSSION

Here, we have performed genome-wide analyses of ΔNp63 binding sites in HaCaT keratinocytes through ChIP-seq and have revealed a mechanism regulating ΔNp63 function through activation of Ras and TGF-β signaling pathways. ΔNp63 is frequently overexpressed in a variety of human malignancies, especially in SCC. A subset of tumor-derived p53 mutants acquires novel functions that promote development of a broader spectrum of tumors and metastases (10); however, it is worth noting that mutant p53 does not affect overall survival in mouse models (19, 20). Although one of the mechanisms describing how mutant p53 proteins exert their novel functions is to interact with and neutralize p63 function (3, 10), the role of ΔNp63 in this mutant p53 “gain-of-function” model has not been established. Our data suggest that mutant p53 antagonizes and alleviates the migratory, invasive, and metastatic capacity of ΔNp63, although ΔNp63 has been shown to have weaker binding affinity for mutant p53 than TAp63 (21, 22). Moreover, oncogenic signals by Ras and TGF-β activation release ΔNp63 from mutant p53, resulting in induction of a transcriptional program, which aggravates cellular motility and invasiveness. Consistently, analysis of TCGA data sets confirmed that high expression of TP63 in ΔNp63-expressing HNSCC results in poor prognosis in patients with mutant p53.

In our analysis, ΔNp63 functions as an activator for a group of target genes (subgroup 1 in Fig. 3). The GO analysis suggested that these target genes are associated with the EGF-EGFR signaling pathway. Consistently, our analyses of TCGA data sets showed that high pERK1/2 and high TP63 are associated with poor survival (Fig. 6C). Among the target genes, we focused on DUSP6 and DUSP7, which were induced by active Ras and TGF-β signals through p63. DUSP6/7 have been shown to participate in negative feedback loops of the Ras-MAPK pathway by inactivating ERK1/2 (13). Our findings support the notion that DUSP6/7 have important effects also on non-MAPK substrates; DUSP6 interacts with mutant p53 and leads to its destabilization, tipping the balance toward ΔNp63 release. Because of the unstable nature of wild-type p53 protein, we were unable to determine whether DUSP6 also targets wild-type p53 or not. In Caenorhabditis elegans, loss of lip-1, a worm ortholog of DUSP6/7/9, results in enhanced p53/CEP-1–dependent DNA damage–induced apoptosis (23). This suggests that there are evolutionarily conserved genetic interactions between the DUSP family of phosphatases and p53 family members and that induction of DUSP expression by DNA binding of p53 and p63 may have originally developed as a negative feedback mechanism for the control of p53 function. In the context of mutant p53 and ΔNp63, however, induction of DUSP6/7/9 empowers the ΔNp63 through degradation of mutant p53.

Our detailed analyses of mutation of phosphorylation sites of mutant p53 suggest that the target sites of phosphorylation/dephosphorylation by active Ras and TGF-β signals and by DUSP6 are different. It is widely accepted that phosphorylation of the N-terminal region of p53 affects the interaction between p53 and MDM2, resulting in stabilization of p53 protein. It has been reported that mutant p53 is not as sensitive to MDM2 as wild-type p53 and that multiple ubiquitin ligases are involved in mutant p53 degradation (12). This would explain how the mechanisms of p53 dephosphorylation by DUSP6 control the degradation of mutant p53. On the other hand, Ser20 of mutant p53 has been reported to be less phosphorylated in several tumor cell lines (24). In addition, caspase 8/10–associated RING proteins (CARPs) have been identified as E3 ubiquitin ligases, which promote MDM2-independent degradation of phospho-p53 (Ser20) (25). We have not fully addressed the roles of each E3 ligase, that is, MDM2 and CARPs, but our data support the notion that the ΔNp63-DUSP6/7 axis controls the phosphorylation status of mutant p53 protein and stabilizes ΔNp63 function by releasing it from mutant p53 protein.

Thus, our data imply that complete loss of p53 results in more migratory, invasive, and metastatic phenotypes than mutant p53 when ΔNp63 isoforms are predominant. A recent report suggests that primary mouse ovarian surface epithelium cells with the p53-null genotype are more migratory and invasive than mutant p53 cells (26); in the ovary, germ cells (oocytes) mainly express TAp63, whereas epidermis expresses ΔNp63 (27, 28). Therefore, our results suggest that the differential expression of p63 isoforms can alter the gain-of-function phenotype of mutant p53.

MATERIALS AND METHODS

Cell culture

Human immortalized keratinocyte HaCaT cells and human embryonic kidney (HEK) 293T cells were cultured as described previously (29). Skin SCC A431 cells were obtained from RIKEN and were maintained in RPMI supplemented with 2 mM l-glutamine (Sigma), 10% fetal bovine serum (FBS) (HyClone), and 1% penicillin-streptomycin (Sigma). The highly bone metastatic human breast cancer cell line MDA-MB-231-D that was derived from the parental MDA-MB-231 line, as described previously (16), was maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Sigma) supplemented with 10% (v/v) FBS and 1% penicillin-streptomycin. The MDA-MB-231-D-ΔNp63α cells were produced after lentiviral infection of MDA-MB-231-D-Luc cells with a lentivirus expressing ΔNp63α. Recombinant lentiviral ΔNp63α vector was generated, as described previously (30). The human Ras-transformed breast cancer cell line MCF10A-MII was cultured as described previously (31).

Production and amplification of adenoviruses

The construction of adenoviral vectors was performed with the ViraPower Adenoviral Expression System from Invitrogen (Life Technologies), following the manufacturer’s instructions. The viral constructs were introduced into HEK293T cells through transfection with Lipofectamine 2000 (Life Technologies). The titer of the adenoviral stocks was determined by a fluorescence-forming assay using antibody against hexon (Millipore) and fluorescence microscopy. The infection capability of the viruses was measured in fluorescence-forming unit. The multiplicity of infection used in each assay was 30, unless differently indicated in the figure.

Reagents and antibodies

Recombinant human TGF-β1 and EGF were purchased from PeproTech. The MEK1/2 inhibitor U0126 and the proteasome inhibitor MG132 were purchased from Millipore. Cycloheximide and Nutlin-3 were purchased from Sigma-Aldrich, and puromycin was purchased from InvivoGen. The concentration of antibiotics used in this study was 0.5 μg/ml for puromycin, unless indicated otherwise. Antibodies against the following proteins were used: FLAG (M2, F3165, Sigma-Aldrich), HA (ab-hatag, InvivoGen), HA (clone Y-11, sc-805, Santa Cruz Biotechnology), α-tubulin (clone TU-02, sc-8035 and clone B-5-1-2, sc-23948, Santa Cruz Biotechnology), p53 (clone DO-1, sc-126 and clone FL-393, sc-6243, Santa Cruz Biotechnology), p63α (clone H-129, Santa Cruz Biotechnology), p44/42 MAPK (ERK1/2) (clone 137F5, #4695, Cell Signaling), phospho-p44/42 MAPK (ERK1/2) (#4370, Cell Signaling), MDM2 (HDM2-323, sc-56154, Santa Cruz Biotechnology), DUSP6 (clone EPR129Y, ab76310, Abcam), DUSP7 (ab100921, Abcam), normal mouse immunoglobulin G1 (IgG1) (MB002, R&D Systems), and normal rabbit IgG (SouthernBiotech).

RNA interference and oligonucleotides

Stealth siRNAs specific for p53 (HSS110905 and HSS186390), p63 (HSS189462) and MDM2 (HSS142911), and control siRNA (12935-112), were purchased from Life Technologies. The ON-target plus siRNAs pools against DUSP6 (J-003964-06) and DUSP7 (J-003567-07), or control siRNA pool (D-001810-10), as well as the ON-target plus custom siRNA specific for ΔNp63 (sequence GGACAGCAGCATTGATCAA), were purchased from Dharmacon Thermo Scientific. siRNAs were introduced into cells with the siLentFect reagent (Bio-Rad) according to the manufacturer’s instructions. The final concentration of the siRNAs used was 20 nM.

Lentiviral shRNA expression vectors

H1 promoter–driven shRNA expression vectors were constructed as described previously (32). The following target sequences were used for each shRNA: sh-DUSP6, 5′-AAGACGGTGGCGTGGCTCA-3′; sh-DUSP7, 5′-GGCCTATCTGATGCAGAAG-3′.

ChIP and data analysis

Chromatin isolation, sonication, and immunoprecipitation using antibody against p63α were performed as described previously (8). HaCaT cells (8 × 106) were infected with LacZ, caRas, or dnRas adenoviruses for 48 hours and stimulated with TGF-β (5 ng/ml) for 1.5 hours in the presence of 10% FBS. High-throughput sequencing of the ChIP fragments was performed using Illumina Genome Analyzer IIx (Illumina) following the manufacturer’s protocols. Sequence reads were aligned against the human reference genome (NCBI Build 36, hg18) using ELAND (Illumina). Peaks were called using MACS v2 (two-sample analysis with a P value cutoff of 1 × 10−5) (33). Assigning a binding site to the nearest gene within 100 kb from a peak was performed using CisGenome v1.2 (34). Normalized sequence read counts for each p63 binding peak were calculated using the Perl script (35). De novo motif prediction was performed by MEME-ChIP with default settings (http://meme-suite.org/tools/meme-chip) (36). The logo plots were generated using the seqLogo package in R (http://bioconductor.org/packages/release/bioc/html/seqLogo.html). The frequency of the p63 binding motif in p63 binding regions (motif counts per sequences) was computed with the likelihood ratio greater than or equal to 500 (default value of CisGenome). GO enrichment analysis was performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID v6.7; http://david.abcc.ncifcrf.gov) (37).

RNA isolation

The extraction of total RNA was performed with the RNeasy kit (Qiagen) according to the manufacturer’s instructions.

Microarray analysis and identification of putative p63 target genes

HaCaT cells, cultured in 10% FBS, were infected with adenoviruses expressing LacZ, caRas, or dnRas and treated with TGF-β (5 ng/ml) for 24 hours. To determine the regulation by mutant p53, we transfected HaCaT cells with control or p53-specific siRNA. Gene expression profiling was performed with a GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix), as described previously (8). The gene expression profiles of HaCaT keratinocytes treated with p63 siRNA (GSE4975) (38) were retrieved from the NCBI Gene Expression Omnibus (GEO) database (http://ncbi.nlm.nih.gov/geo/). To isolate p63 direct target genes, we set several thresholds. There was at least one p63 binding region within 100 kbp from the gene body, and the p63 binding to the region was strong in the control LacZ-untreated condition (normalized sequence read counts for each p63 binding was greater than or equal to 1.3 reads per million mapped reads). In addition, the p63 binding was enhanced by both Ras activation and TGF-β treatment (caRas + TGF-β)/(dnRas + TGF-β) ≥ 1.2. We identified 4778 putative p63 direct target genes. We further filtered genes whose expression was modulated by Ras, TGF-β, and mutant p53. For subgroup 1 (p63 as an activator), we set the following parameters: sic/sip63 ≥ 1.2, caRas/dnRas ≥ 1.2, (caRas + TGF-β)/caRas ≥ 1.0, and sic/sip53 ≤ 0.714 (1/1.4). For subgroup 2 (p63 as a repressor), we set the following parameters: sic/sip63 ≤ 0.833 (1/1.2), caRas/dnRas ≤ 0.833 (1/1.2), (caRas+TGF-β)/caRas ≤ 1.0, and sic/sip53 ≥ 1.4.

ChIP-qPCR and qRT-PCR

The qPCR was conducted with iQ SYBR Green Supermix (Bio-Rad), and the reactions were carried out in triplicate in the CFX96 Real-Time PCR Detection System (Bio-Rad). Primer sequences are given in table S3. The amount of immunoprecipitated DNA was calculated relative to the input. The fold enrichment corresponds to the p63α enrichment (% input value) in each locus divided by that in the negative control regions. The mRNA expression of each gene was normalized to GAPDH expression values.

Plasmid construction

The ΔNp63α plasmid vector (#26979) (39) was obtained from Addgene. The coding region of the mouse H-Ras was amplified by PCR, and their sequences were confirmed. Plasmids for p53-R273H and MDM2 were provided by Y. Inoue (Nagoya City University, Nagoya, Japan). For stable cell line establishment, cDNAs were cloned into an episomal expression vector pPyCAG-IRES-Puro, which contains polyoma ori and can be propagated episomally in cells (40). Mutations were introduced by site-directed mutagenesis using PCR with specific primers shown in table S4. All constructs were confirmed by sequencing.

The p63 binding region in the DUSP6 promoter that was identified in the ChIP-seq analysis was amplified from human genomic DNA by PCR and cloned into a modified pGL4 reporter plasmid (Promega) driven by the minimal adenoviral major late promoter (30). Mutagenesis of the four most conserved nucleotides of the p63 binding motif was performed with PCR using the primers listed in table S4.

Luciferase assays

Cells were transiently transfected with the luciferase reporter constructs along with the Renilla luciferase reporter vector pGL4.74[hRluc/TK] (Promega) as an internal control, and the expression vectors pcDNA3 and pcDNA3 ΔNp63α. For the siRNA and DNA transfection experiments, cells were transfected with p53 siRNA or siRNA control, and the DNA plasmids at the same time with Lipofectamine 2000 according to the manufacturer’s protocol. Cells were harvested 48 hours after transfection and assayed for luciferase activity using the Firefly & Renilla Luciferase Kit (Biotium) and the EnSpire plate reader (PerkinElmer).

Coimmunoprecipitation

HaCaT cells transfected with control siRNA or with siRNA-targeting p53 or DUSP6 were lysed in lysis buffer [20 mM tris-HCl (pH 7.5), 150 mM NaCl, 10% glycerol, and 1% Triton X-100] and incubated overnight with protein A or anti-mouse IgG Dynabeads (Life Technologies) that had been preincubated with antibody against DUSP6 or p53 or IgG control in phosphate-buffered saline (PBS) supplemented with 0.5% bovine serum albumin. The beads were washed three times with lysis buffer, and the immunoprecipitated proteins were eluted in 2× SDS buffer and subjected to SDS–polyacrylamide gel electrophoresis.

Cell proliferation assays

MDA-MB-231-D cells stably transfected with lentiviruses expressing GFP (control) or ΔNp63α were transfected with control siRNA or with siRNAs targeting DUSP6 and DUSP7 (DUSP6/7). Cell counting was performed after 24, 48, and 72 hours after seeding with a Beckman Coulter cell counter. For measuring nuclear 5′-ethynyl-2′-deoxyuridine (EdU) incorporation, cells were seeded and cultured for 72 hours, after EdU was added; incubation was prolonged for 4 hours before the assay was performed. EdU-positive cells were recorded using the Click-iT EdU Microplate assay (Life Technologies) and an EnSpire plate reader (PerkinElmer).

Cell scratch assay

MDA-MB-231-D cells (500,000 cells per well in a six-well plate) overexpressing ΔNp63α or GFP as a control were transfected with control siRNA or with siRNA against DUSP6 and DUSP7 in the presence of 3% FBS. Two days after transfection, a wound was made, and the cells were allowed to migrate for another 24 hours. Pictures were taken immediately after the scratch and after 24 hours of migration with an Axiovert 40 CFL microscope (Zeiss). The data were analyzed with the TScratch software (41), and the relative percentage of open wound area was quantified.

Chemotaxis assay

A431 cells were transfected with control siRNA or with siRNAs specific for p63 or DUSP6 and DUSP7. Forty-eight hours later, cells (25,000 per well) were seeded on the upper side of a ChemoTx 96-well plate (101-8; Neuroprobe). Before seeding, the filter was coated on both sides with collagen I (Advanced BioMatrix) for 1 hour at room temperature. Culture medium with 3% FBS, with or without TGF-β (5 ng/ml) or EGF (10 ng/ml), was added on the lower side of the chamber. Cells were incubated at 37°C overnight. On the following day, the nonmigrating cells were scraped off and the upper side of the filter was washed with PBS. The cells that had migrated to the lower side of the chamber were fixed with 96% ethanol, stained with Giemsa, and quantified using an EnSpire plate reader (PerkinElmer).

3D spheroid invasion assays

The 3D invasion assay was performed essentially as described (31). Spheroids of A431 cells, 48 hours after transfection with control siRNA or with siRNAs specific for p63 or DUSP6 and DUSP7, were embedded in collagen in the presence or absence of TGF-β and/or EGF, as indicated. Representative pictures of spheroids were taken just after embedding into collagen, as well as 24 or 48 hours after their embedding (pictures of p63 siRNA after 24 hours of invasion and pictures of DUSP6/7 siRNA after 48 hours of invasion are shown in fig. S3E). Relative invasion was quantified as the mean area that the spheroids transfected with the individual siRNA occupy 24 hours after being embedded in collagen.

Intracardiac experimental metastasis model and survival studies in vivo

Female BALB/c nu/nu mice (4 weeks old) were obtained from Sankyo Laboratory. MDA-MB-231-D ΔNp63α and GFP-expressing cells infected with lentiviral sh-control or sh-DUSP6/7 expression vectors (100,000 cells in 200 μl of sterile DMEM, containing 10% FBS) were injected into the left ventricle of the heart with a 26-gauge needle under anesthesia with diethyl ether, as described previously (16). All animal experiments were reviewed and approved by the Animal Ethics Committee of The University of Tokyo and performed in accordance with the institutional guidelines. The survival of mice was analyzed by Kaplan-Meier plot. The log-rank test was used to compare two groups. P < 0.05 was considered statistically significant.

Statistical analysis

The difference between experimental groups of equal variance was analyzed using Student’s t test with Bonferroni correction, with *P < 0.05, **P < 0.01, ***P < 0.001 being considered significant. Essentially all experiments were performed at least three times independently, and similar results were obtained. For the analysis of patient data sets from TCGA, all statistical tests were performed using R software (version 3.2.5, https://www.r-project.org/). Z-scored expression values of mRNA and relative protein/phosphoprotein values (RPPA) were obtained from cBioPortal in July 2015 (42, 43). Patients were divided into low and high expressers using the median values of mRNA expression or RPPA. The overall survival was estimated with the Kaplan-Meier method, and the differences between groups were evaluated by the log-rank test, using the cmprsk package for R. P values were calculated using pairwise Welch t test with Bonferroni correction (*P < 0.05, **P < 0.01, ***P < 0.001).

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/9/442/ra84/DC1

Fig. S1. Knockdown of the ΔN isoforms of p63 results in the same effects as knockdown of all p63 isoforms.

Fig. S2. p63 is necessary for the up-regulation of DUSP6 and DUSP7 by Ras and TGF-β signaling.

Fig. S3. ΔNp63 and its target genes, DUSP6 and DUSP7, promote metastatic behavior in cultured cancer cells.

Fig. S4. Correlation of mutant status of p53 and expression of TP63 with survival in cancer patients.

Fig. S5. Stability of p53-R273H protein and its phosphorylation site mutants.

Table S1. List of all the genes in subgroup 1 (p63 as an activator).

Table S2. List of all the genes in subgroup 2 (p63 as a repressor).

Table S3. Primer sequences used for ChIP-qPCR and qRT-PCR.

Table S4. Primer sequences used for cloning and mutagenesis.

REFERENCES AND NOTES

Acknowledgments: We thank Y. Inoue (Nagoya City University, Nagoya, Japan) for providing the plasmids for p53 mutants and MDM2 and J. Lennartsson (Ludwig Cancer Research, Uppsala, Sweden) for providing reagents. We thank S. Tsutsumi, H. Mano, T. Ueno, S. Kojima (University of Tokyo), and all the members of Ludwig Cancer Research in Uppsala for helpful comments and discussion. Funding: This research was supported by the Ludwig Institute for Cancer Research; the Swedish Cancer Foundation (grant 100452 to K.M.); the Swedish Research Council (grant 2015-02757 to C.-H.H.); KAKENHI [Grants-in-Aid for Scientific Research on Innovative Areas (Integrative Research on Cancer Microenvironment Network; 22112002 to K.M.)]; Scientific Research (S) (grant 20221009 to H.A. and grant 15H05774 to K.M.) and (C) (grant 24501311 to D.K.) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (MEXT); and the Ministry of Health, Labor, and Welfare of Japan (D.K.). This study was performed in part as a research program of the Project for Development of Innovative Research on Cancer Therapeutics (P-Direct), MEXT. M.M. is supported by the Kanae Foundation for Research Abroad and the ITO Genboku and SAGARA Chian Memorial Scholarship. Author contributions: E.V., M.M., D.K., K.M., and C.-H.H. conceived the study. E.V., M.M., A. Mizutani, Y.H., S.E., A.S., N.K., and J.C. carried out the cell culture and in vivo experiments. M.M. and D.K. performed the bioinformatics analysis. E.V., M.M., D.K., A. Moustakas, K.M., and C.-H.H. wrote the article. The ChIP-seq and microarray data were obtained in the laboratory of H.A. A.-K.O., A. Moustakas, K.M., and C.-H.H. supervised the overall study. All authors have read and approved the final manuscript. Competing interests: The authors declare that they have no competing financial interests. Data and materials availability: Raw sequencing data with peak calling results and microarray data are available at GEO (GSE60814).

Stay Connected to Science Signaling

Navigate This Article