To further understand the mechanisms employed by TAp63 as a tumour suppressor gene, we studied the effects of TAp63 on a p53^+/− or p53^−/− background^8,16–18. Some human tumours show alterations in both p53 and p63 (refs ^{12, 13}), and point mutant p53 binds and functionally inactivates p63 (refs ^{17, 18}). To study these interactions, we generated six cohorts of 30 mice each: TAp63^−/−;p53^+/− and TAp63^+/−;p53^+/− for comparison with p53^+/− mice, and TAp63^−/−;p53^−/− and TAp63^+/−;p53^−/− for comparison with p53^−/− mice. Unlike p53^+/− and p53^−/− mice, TAp63/p53 compound mutant mice developed a remarkable number of metastatic tumours (Fig. 1h–k, Supplementary Fig. 1c–f and Supplementary Table 1)^8,16. Whereas the TAp63^+/−;p53^+/− mice had a similar tumour spectrum to that of TAp63^−/−;p53^+/− mice (Supplementary Table 1), the sarcomas of the TAp63^+/−;p53^+/− mice were significantly more invasive and metastatic (88%) than those in the TAp63^−/−;p53^+/− cohort (14%) (Fig. 1k). In contrast, the carcinomas in the TAp63^−/−;p53^+/− cohort were more metastatic (44%) than those in the TAp63^+/−;p53^+/− group (29%) (Fig. 1k). We next examined sarcomas (n = 10) and carcinomas (n = 10) from TAp63^+/−;p53^+/− mice and found that the wild-type allele of TAp63 was retained (Fig. 1l). We also examined TAp63^+/−;p53^+/− tumours and TAp63^−/−;p53^+/− tumours for p53 loss of heterozygosity (LOH) and found that one out of ten sarcomas showed LOH of p53 in TAp63^+/−;p53^+/− tumours (Fig. 1m). A slightly higher frequency was noted in tumours from TAp63^−/−;p53^+/− sarcomas (one out of eight) and carcinomas (two out of eight), indicating that there is increased selective pressure to lose the wild-type p53 allele in the complete absence of TAp63 (Fig. 1m). However, the frequency of p53 LOH was lower than the previously reported frequency (ranging from 75–90%) of sarcomas from p53^+/− mice^8,16 and was more reminiscent of the frequency of p53 LOH in mouse models of Li–Fraumeni syndrome¹⁷, further supporting the concept that p53 and p63 interact in tumour suppression. We also analysed TAp63^−/−;p53^−/− and TAp63^+/−;p53^−/− mice and found that highly metastatic carcinomas and sarcomas developed in addition to the thymic lymphomas found in the p53^−/− mice (Fig. 1j and Supplementary Table 1). Sarcomas (n = 5) and carcinomas (n = 5) from these mice retained the wild-type allele of TAp63 (Fig. 1l), providing further evidence that TAp63 is a haploinsufficient tumour suppressor gene.

To understand how heterozygosity of TAp63 led to a more severe tumour phenotype than a complete loss of TAp63, we examined whether senescence in the TAp63^−/− mice, which age prematurely¹⁹, inhibited tumorigenesis or metastasis. We found several examples of TAp63^−/−;p53^+/− mice that developed non-metastatic osteosarcomas together with metastatic carcinomas (Fig. 2a, b and Supplementary Fig. 1e, f). Tumours from these mice, along with sarcomas and carcinomas from TAp63^+/−;p53^+/− mice, were assayed for senescence-associated markers, namely SA-β-gal (Fig. 2c and Supplementary Fig. 2a, b) or PML (encoding promyelocytic leukemia protein), p16^Ink4a and p19^Arf (Fig. 2d). We found that osteosarcomas and rhabdomyosarcomas in the TAp63^−/−;p53^+/− mice expressed high levels of senescence markers, whereas the same tumour types in the TAp63^+/−;p53^+/− mice did not (Fig. 2c, d). None of the carcinomas from TAp63^−/−;p53^+/− or TAp63^+/−;p53^+/− mice was positive for these markers, indicating that TAp63 deficiency has a tissue-specific function in the induction of senescence. These results correlated with the level of aggressiveness and metastatic potential of these tumours.

Genomic instability in tissues of TAp63^−/− mice has been shown to be high¹⁹. To examine whether there is a similarly high level of genomic instability in tumours from TAp63^−/−;p53^+/− mice and whether this is correlated with tumour aggressiveness and metastasis, we performed γ-H2AX immunostaining and metaphase spreads on tumours from TAp63 mutant mice (Fig. 2e, f and Supplementary Fig. 2c–g). We did indeed find that TAp63^−/− carcinomas showed high levels of γ-H2AX, polyploid cells and chromosomal aberrations, suggesting that carcinomas lacking TAp63 overcome senescence in epithelial tissues as a result of the acquisition of further genetic alterations. Although a decrease in TAp63 expression promotes metastasis in both tumour types, senescence is triggered in sarcomas, thus inhibiting progression, whereas in carcinomas, genomic instability ensues, permitting further genetic events favouring tumour progression.

Given the high level of metastasis in TAp63^−/− mice, we examined whether TAp63^−/− mouse embryonic fibroblasts (MEFs) had an enhanced invasive ability. TAp63^−/− MEFs showed a 1.8-fold increase in invasion (Fig. 2g). To assess whether this finding could be generalized to human tumours, we examined primary human head and neck squamous-cell carcinomas (HNSCCs) (cell lines 10A, 17A and 22A) and matched metastatic lesions with markedly low levels of TAp63 (Fig. 2h) (cell lines 10B, 17B and 22B)²⁰. Tumours with low levels of TAp63 showed an increased invasive ability (Fig. 2i), indicating that TAp63 is a critical regulator of cancer metastasis.

Our findings that TAp63^+/− tumours are more aggressive than those from TAp63^−/− mice are reminiscent of the phenotype of Dicer conditional knock out (Dicer1^fl/+) mice intercrossed to the Kras^LSL-G12D lung cancer model²¹, which indicate that Dicer is haploinsufficient for tumour suppression⁴, in a similar manner to our previous observations for TAp63. Additionally, Dicer1^−/− mice show embryonic lethality (before embryonic day E8.5) similar to that in TAp63^−/− mice on an enriched C57BL/6 background¹⁹; 35% of TAp63^−/− embryos die between E6.5 and E8.5. Given these similarities, we examined whether messenger RNA levels of Dicer are dysregulated in tumours and cells lacking TAp63. TAp63^−/− osteosarcomas, lung adenocarcinomas and mammary adenocarcinomas (n = 6) and MEFs expressed significantly lower levels of Dicer than p53^−/− tumours and MEFs (Fig. 3a, b and Supplementary Fig. 3), suggesting that TAp63 is required for the transcription of Dicer. Similarly, we found that metastatic human HNSCCs (n = 46), lung adenocarcinomas and squamous cell carcinomas (n = 92) and mammary adenocarcinomas (n = 43) with low levels of TAp63 had low Dicer expression (Fig. 3c, Supplementary Fig. 4 and Supplementary Tables 2 and 3). These data indicate a critical role for TAp63 in the regulation of Dicer in metastasis in human cancer. Consistent with this is the observation that Dicer and Drosha levels have been shown to be low in human cancer²⁰.

To determine whether Dicer is a transcriptional target of TAp63, we performed chromatin immunoprecipitation (ChIP) analysis for a site matching the p63 consensus binding site (Supplementary Table 4). We found that p63, and not p53, bound to the Dicer promoter (Fig. 3d and Supplementary Fig. 5a), suggesting that Dicer is a direct transcriptional target of p63. In addition, all isoforms of TAp63 (α, β and γ) were able to transactivate a Dicer–luciferase reporter gene with the p63-binding site (Fig. 3e), further indicating that TAp63 regulates Dicer transcriptionally.

To understand whether TAp63 controls metastasis through transcriptional regulation of Dicer, we re-expressed Dicer in TAp63^−/− MEFs (Fig. 3f). These cells lost their ability to invade (Fig. 3g). Conversely, wild-type MEFs expressing a Dicer short hairpin RNA (shRNA)²² (Fig. 3f) showed increased invasiveness (Fig. 3h). These data indicate that low levels of Dicer lead to increased cell invasion, similar to that observed in TAp63-deficient cells.

To further explore the connection between TAp63 and Dicer in metastasis, we modulated levels of TAp63 in MEFs by expressing TAp63γ in TAp63^−/− MEFs and human HNSCCs (Supplementary Figs 5b and 6a) and measured Dicer expression and invasion potential. Levels of Dicer increased to wild-type levels in TAp63^−/− cells expressing TAp63γ (Fig. 3i and Supplementary Fig. 6b), which resulted in decreased invasion potential (Fig. 3j and Supplementary Fig. 6c). We also acutely deleted TAp63 after the introduction of adenovirus-Cre in TAp63^fl/fl MEFs (Supplementary Fig. 5c). These TAp63^Δ/Δ MEFs showed concomitant downregulation of Dicer (Fig. 3k) and a twofold increase in cellular invasion (Fig. 3l), indicating that regulation of Dicer by TAp63 is critical in the suppression of invasion.

Because Dicer expression is low in the absence of TAp63, we assessed the processing of miRNAs that have a recognized role in lung and mammary adenocarcinoma metastasis^1–3,23, namely miR-10b, miR-200b, miR200c, miR34a and miR-130b, in TAp63^−/− MEFs and found that their processing is defective (Fig. 3m). To examine whether modulating levels of TAp63 in cells would affect miRNA processing, we scored for processing of miR-10b in TAp63^−/− MEFs reconstituted with TAp63γ and in TAp63-ablated (TAp63^Δ/Δ) MEFs. When TAp63 was re-expressed, we observed a rescue of miRNA processing, and when TAp63 was acutely deleted, we observed a diminution of miRNA processing (Fig. 3n). The precursor of miR-130b (pre-miR-130b) was present at low levels in TAp63^−/− MEFs (Fig. 3o), indicating that the primary transcript of miR-130b may be expressed at low levels in the absence of TAp63.

We found that levels of miR-130b were downregulated in TAp63^−/− MEFs, suggesting a role for TAp63 in its regulation (Fig. 4a). We reasoned that other miRNAs, such as miR-34a, which have been shown to be a transcriptional target of p53 (ref. ²³), may also be a target of TAp63. Indeed, miR-130b and miR-34a were downregulated in the absence of TAp63 (Fig. 4a–c and Supplementary Fig. 7a). miR130b was also low in high-grade lung adenocarcinomas and HNSCCs with low levels of TAp63 and Dicer (Supplementary Table 3). To assess whether p63 binds to the promoters of these miRNAs, we performed a ChIP analysis with identified putative p63-binding sites (Supplementary Table 4). We found a significant level of p63 binding at both the miR-130b and miR-34a promoters (Fig. 4d and Supplementary Fig. 7b–d). To determine whether p63 can directly transactivate miR-130b and miR-34a, we performed luciferase assays (Fig. 4e and Supplementary Fig. 7e) and found that TAp63 transactivated the miR130b reporter. We modulated the levels of TAp63 by overexpression or acute deletion in MEFs and found that levels of miR-130b correlated with levels of TAp63 (Fig. 4f, g). Taken together, these data indicate that miR-130b is a direct target of TAp63.

To understand whether regulation of miR-130b by TAp63 has a function in the metastatic phenotype, we modulated levels of miR-130b in MEFs (Supplementary Fig. 7f–h) and assessed invasion (Fig. 4h, i). Wild-type MEFs with low levels of miR-130b showed increased invasion (Fig. 4h), and TAp63^−/− MEFs expressing both miR-130b and Dicer showed decreased invasion (Fig. 4i) comparable to that of wild-type MEFs. This suggests that coordinate regulation of both Dicer and miR-130b by TAp63 is critical in suppressing metastasis.

We have shown that TAp63 is a suppressor of tumorigenesis and metastasis and that its complete inactivation results in tissue-specific antagonistic pleiotropy with respect to tumour progression including genomic instability and activation of senescence. Both the spectrum and the highly metastatic nature of tumours in TAp63-deficient mice are consistent with a role for TAp63 as a master regulator of metastasis. Furthermore, we have shown a role for transcriptional regulation by TAp63 of Dicer and miR-130b in metastasis. Given the diverse biological roles of both Dicer and TAp63, a complete understanding of the regulation of Dicer by TAp63 in other processes is a key question for the future.

01