ΔNp63 promotes IGF1 signalling through IRS1 in squamous cell carcinoma

Accumulating evidence has proved that deregulation of ΔNp63 expression plays an oncogenic role in head and neck squamous cell carcinomas (HNSCCs). Besides p63, the type 1-insulin-like growth factor (IGF) signalling pathway has been implicated in HNSCC development and progression. Most insulin/IGF1 signalling converges intracellularly onto the protein adaptor insulin receptor substrate-1 (IRS-1) that transmits signals from the receptor to downstream effectors, including the PI3K/AKT and the MAPK kinase pathways, which, ultimately, promote proliferation, invasion, and cell survival. Here we report that p63 directly controls IRS1 transcription and cellular abundance and fosters the PI3K/AKT and MAPK downstream signalling pathways. Inactivation of ΔNp63 expression indeed reduces tumour cell responsiveness to IGF1 stimulation, and inhibits the growth potential of HNSCC cells. In addition, a positive correlation was observed between p63 and IRS1 expression in human HNSCC tissue arrays and in publicly available gene expression data. Our findings indicate that aberrant expression of ΔNp63 in HNSSC may act as an oncogenic stimulus by altering the IGF signalling pathway.


INTRODUCTION
Head and neck squamous cell carcinoma (HNSCC), the sixth most frequent malignancy worldwide, is a heterogeneous disease that develops from the stratified epithelium of the upper aerodigestive tract [1]. Despite recent diagnostic and therapeutic advances, the prognosis and survival of HNSCC patients remain poor. In response to the increase in the number of new HNSCC cases worldwide, further knowledge of tumour biology and the identification of novel clinical biomarkers are needed to improve prognostic stratification and optimise the anti-cancer therapies.
The Insulin-like Growth Factor (IGF) axis is composed of the tyrosine kinase IGF type 1 receptor (IGF1R), its ligands, insulin, IGF1 and IGF2, the adaptor protein Insulin Receptor Substrate 1 (IRS1) and a family of six ligand-binding proteins (IGFBPs) that regulate the bioavailability and half-life of circulating IGF1. Among the IGFBPs, IGFBP3 binds to more than 95% of circulating IGF. The IGF1R binds IGF1 and IGF2 with high affinity, as well as insulin, though with lower affinity. Upon ligation, IGF1R recruits and phosphorylates IRS1. Phosphorylated IRS1 acts as docking site for intracellular adaptor proteins that, ultimately, activate two downstream signalling cascades: the PI3K/AKT and the MAPK pathways, both of which have mitogenic and pro-survival roles.
Reduced responsiveness to IGF1/insulin can occur as a result of diminished total or phosphorylated amounts of IGF1R or IRS1, and results in decreased activation of the PI3K/AKT and MAPK pathways [73].
Deregulation of the IGF axis has been implicated in the development and progression of several human cancers. Elevated serum IGF1 levels, and increased levels or constitutive activation of IGF1R and IRS1 are associated with increased risk of a variety of epithelial cancers, metastasis and therapeutic resistance [74][75][76][77][78][79][80]. Hence, it has been proposed that reduction of IGF signalling in may have a prognostic impact and a therapeutic benefit in some cancer types. In HNSCC, IGF1R overexpression is associated with adverse survival, HPV negativity and high tumour T-stage [76,81,82]. High IGF1 levels and both lower and higher levels of IGFBP3 are predictor risk factors for secondary tumour development in patients with HNSCC [83,84]. In addition, increased IRS1 expression was found in nasopharyngeal carcinoma patients, where it correlated with lymph node metastasis [80].
Notably, the existence of a possible crosstalk between p63 and the IGF system has been reported. The tumour suppressive TAp63 isoforms negatively control Igf1r transcription [85], whereas Igfbp3 is a target of transcriptional repression by ΔNp63 [86]. Here we report, that ΔNp63 affects the transcription and the cellular abundance of Irs1 in HNSCC cells and, that, as a result, p63 down-regulation impairs the activation of the intracellular signalling pathways following IGF1R stimulation.

p63 controls IRS1 expression levels in HNSCC cells
By exploiting RNA sequencing (RNA-seq) transcripttome profiling, we identified genes regulated by ΔNp63 in normal human epidermal keratinocytes (NHEKs) (E.C. unpublished data). The analysis of RNA-Seq data obtained from p63-depleted cells revealed almost 50% reduction in IRS1 expression levels relatively to control cells (Fig. 1A). We then sought to test whether IRS1 expression is regulated by p63 in HNSCC cells. HNSCC cell lines display moderate/high levels of p63 expression (Fig. 1B, upper panel). In particular, the predominantly expressed isoform of p63 in HNSCC cells is ΔNp63, with TAp63 being undetectable in the majority of the cell lines (Fig. 1C). We observed a consistent correlation between ΔNp63 protein levels and the gene expression pattern of IRS1 in most of the HNSCC cell lines analysed (Fig. 1B, lower panel). Validation of RNA-seq data showed that, following p63 knockdown, IRS1 transcript and protein levels were reduced in NHEK and in a panel of HNSCC cell lines (Fig. 1D).

p63 induces IRS1 expression by binding directly to the regulatory region of the Irs1 gene
Genome-wide profiling of p63 binding sites by Chromatin IP Sequencing (ChIP-seq) analysis of NHEKs [87] revealed peaks of p63 binding to regions downstream the Irs1 locus ( Fig. 2A). The algorithm p63scan identified a putative p63 responsive element (RE) in the most distant enriched peak. To validate direct interaction of p63 with this putative RE, we examined p63 occupancy at the site identified in the ChIP-seq analysis. By ChIP experiments in HNSCC cells, we found binding of p63 to a regulatory region located downstream the Irs1 locus (+148 kbps from the TSS) (Fig. 2B). In addition, by performing luciferase activity reporter assays in H1299 cells, we found that the ΔNp63 isoforms activate a luciferase reporter gene driven by the p63 RE located in the regulatory region of the Irs1 locus (Fig. 2C). Site-specific mutagenesis of the p63 RE almost completely abrogated the transactivating AGING ability of ΔNp63 (Fig. 2C). Overall, these data demonstrate that Irs1 is a direct transcriptional target of ΔNp63.

p63 inactivation impairs cellular sensitivity of HNSCC cells to IGF1/insulin stimulation
We next tested whether knockdown of ΔNp63 affects the level of activated IRS1. To assess whether downregulation of IRS1 in p63-depletd HNSCC cells would impair cellular responsiveness to receptor stimulation, we treated serum starved Fadu cells with both IGF1 and insulin. Stimulation of the IGF1R resulted in reduced levels of phosphorylated (Ser612) IRS1 in p63-depleted relatively to control cells (Fig. 3A). To rule out off-target effects, two independent siRNAs against p63 and one specific ΔNp63 siRNA were employed (Fig. 3A, Fig.  S1A, and S1B).
To examine whether alteration of IRS1 and phospho-IRS1 levels induced by p63 silencing may affect downstream IGF signal transduction, we measured the activation of PI3K/AKT and MAPK downstream signalling pathways in p63-depleted cells. Upon p63 knock-down, we observed desensitization of HNSCC cells to IGF1 stimulation, as assessed by decreased amounts of phospho-AKT (Ser473) and phospho-S6 (Ser235/236) (Fig. 3B). Activation of MAPK (phospho-Erk1/2; Thr202/Tyr204) signalling pathway in response as measured by RNA-Seq analysis of p63-depleted NHEK. Cells were transfected with p63 (sip63#1) or scrambled control (siScr) siRNAs. Pvalue = 0,005. (B) The amount of p63 was measured in NHEK and HNSCC cell lines by western blot analysis (upper panel). IRS1 transcript levels were analysed by RT-qPCR (lower panel). RT-qPCR was performed in duplicate. IRS1 expression was normalized on Tbp housekeeper and plotted relative to NHEK cells (mean ± s.d.). (C) The transcript levels of TAp63 (black box) and ΔNp63 (grey box) were measured in NHEK and HNSCC cell lines by RT-qPCR. RT-qPCR was performed as above. Gene expression was normalized on Tbp housekeeper and plotted relative to NHEK cells (mean ± s.d.). (D) RT-qPCR analysis (upper panels) of two independent experiments performed in duplicates for Irs1 transcripts in NHEK and HNSCC cells transfected with scrambled control (siScr) or p63 (sip63#1) siRNAs. Cells were harvested 48 h after transfection. qRT-PCR was performed as above. Values are normalized to Tbp and plotted relative to control cells (mean ± s.d.). Western Blot analysis for IRS1 and p63 in HNSCC cells transfected as above. Cells were harvested 48 h after transfection. β-actin served as loading control.
AGING to IGF-1 stimulation was also markedly reduced in p63depleted cells (Fig. 3B), further proving that p63 affects cellular sensitivity to IGF1/insulin stimulation through the regulation of IRS1 cellular abundance. Notably, knockdown of IRS1 hampered the proliferation of HNSCC cells, mimicking the effect of p63 inactivation (Fig. 3C). These findings indicate that the p63/IRS1 functional axis positively regulates the growth potential of HNSCC cells.

p63 and IRS1 expression correlates in HNSCC patients
To examine possible correlations between the expres-sion levels of Np63 and IRS1 in NHSCC primary tumours, clinical NHSCC tumour specimens sample sand related benign controls were examined for p63 and IRS1 staining on tissue microarray slides. Consistent with previous reports [65,66], the majority of the patients (63%) showed high levels of p63 expression ( Fig. 4A; representative staining patterns are seen in Fig. 4C, top panels).
High and moderate tumour cell IRS1 expression was observed in 27% and 10%, respectively of the cases (Fig. 4B, representative staining patterns are seen in Fig. 4C, bottom panels). Significant correlation of p63 and IRS1 expression was observed in 23 out of 60 HNSCC samples (38%, Fig. 4C and Table S1).
To further investigate the expression levels of p63 and IRS1, we analysed publicly available transcriptome sequencing data of 522 HNSCC patients from the TCGA repository [88]. Overall, the transcript levels of p63 and IRS1 are significantly higher in tumour specimens than in normal samples (Fig. 4D, and 4E). On the basis of their expression levels, for each gene, we stratified tumour patients into two distinct groups, displaying either up-regulation or down-regulation relatively to normal subjects. Notably, we found that 66.2% of tumour samples with p63 up-regulation also exhibited high levels of IRS1 expressiA similar correlation was observed in samples with p63 downregulation, in which 58% of the patients concomitantly NHEKs by ChIP-sequencing (ChIP-seq) using 4A4 and H129 anti-p63 antibodies in two normal human primary keratinocyte cell lines (K1 and K2) [87]. (B) ChIP analysis of p63 occupancy at the regulatory regions of the Irs1 gene. ChIP assays were performed in Fadu HNSCC cells using H129 anti-p63 antibody and control IgGs. PCR validation was performed using primers spanning the p63-binding sites located within the genomic regions identified by ChIP-seq assays. (C) Luciferase reporter assays of Irs1 regulatory regions (left panel). The pGL3 reporter vector (30 ng) and the pRL-CMV-Renilla luciferase plasmid (5 ng) were cotransfected with the empty pcDNA-HA vector or plasmids coding ΔNp63α, ΔNp63β, and ΔNp63γ (150 ng) into the p53 null human H1299 cell line. The luciferase activities of cellular extracts were measured 24 h after transfection. Cellular lysates were also analysed by western blot (right panel). Data are presented as mean ± SD and are representative of three independent experiments. AGING displayed low levels of IRS1 transcripts (Fig. 4F, p-value= 4.293e-07). Overall, these findings demonstrate that a statistical significant positive association exists between p63 and IRS1 expression inHNSCC patients. . Forty-eight h after transfection, cells were seeded in 6-cm plates at 500,000/plate and growth was followed until day 6.

DISCUSSION
It has been originally hypothesized that, ΔNp63 mainly exerts its oncogenic functions by acting as a dominant negative repressor of the tumour suppressive members of the p53 family, including TAp63. As a consequence of preventing access to their DNA binding sites, ΔNp63 would impinge on the transcription of genes involved in cell cycle control and cell death. In addition, emerging evidence has unveiled key tumour-related signalling pathways that are transcriptionally regulated by ΔNp63, in a p53 independent manner [89]. For instance, by acting in concert with the chromatin remodelling factor ACTL6A, p63 controls chromatin accessibility and functions as a direct transcriptional repressor of the Hippo/ YAP regulator WWC1 in SCC [66]. Further-more, ΔNp63 controls a transcriptional program com-prising the hyaluronic acid (HA) synthase HAS3 and two hyaluronidase genes, HYAL-1 and HYAL-3, thus sustaining the pro-tumorigenic HA metabolism and signalling [59]. Abraham and collaborators [90] have recently identified components of the transforming growth factor-β signalling and the RHOA GTPase as targets and mediators of ΔNp63-dependent cell proliferation in SCCs.
Our data demonstrate that IRS1 is a direct target of transcriptional activation by ΔNp63 in HNSCC cells. Coherently, ΔNp63 and IRS1 expression patterns are positively correlated in primary tumours, suggesting that the interaction of p63 and IRS1 might contribute to the pathogenesis of HNSCC. More broadly, in ΔNp63 overexpressing SSC tumours, unbalanced expression of the TAp63/ΔNp63 isoforms may lead to enhanced Igfr1/Irs1 transcriptional activation, resulting in augmented protein abundance of IGFR1 and IRS1, which may increase sensitivity of cancer cells to growth factor stimulation ( Figure 5). In addition, overexpression of ΔNp63 prevents expression of Igfbp3 thus, ultimately, enhancing circulating IGF-1 (Fig. 5). Thus, the existence of a crosstalk between p63 and the IGF1 system may represent a mechanism by which tumours that overexpress ΔNp63 escape apoptosis and acquire a proliferative advantage.
Elevated levels of IRS1 have been reported to contribute to cancer development and progression [78,80,91,92]. Notably, high expression of IRS1 in breast cancer cells was positively correlated with aberrant phosphorylation of AKT, which was significantly associated with lymph node metastasis [93]. Coherently, we observed reduced ligand-stimulated activation of AKT in p63-depleted cells, implying that reducing the cellular abundance of IRS1 could be a strategy to diminish the mitogenic potential of the IGF system. Acting as an adaptor protein that conveys signals originating from different receptors to multiple downstream signalling molecules, IRS1 represents a potentially relevant predictive clinical biomarkers for cancers susceptible to IGF-IR targeting.
Several reports have showed a correlation between the IGF pathway and HNSCC clinical parameters [80][81][82][83][84]. On the other hand, IGF pathway-related proteins have not been implemented as cancer biomarkers yet, due to the existence of contradictory findings on their prognostic impact. Indeed, negative correlations between levels of IGF1/IGF1R and clinical outcomes have been also reported in HNSCC [94]. Thus, in addition to unveiling alterations within the IGF system in tumours, the identification of alterations outside the IGF axis (e.g. upstream regulators), which may affect the IGF signalling, would be relevant to establish additional predictive markers for patient stratification and clinical management.
The controversial clinical data on the prognostic impact of the IGF system in HNSCC may reflect the high histopathological heterogeneity within this disease that includes tumours arising from various anatomical sites. Whether or not deregulation of the IGF signalling network and the positive regulation of IRS1 by ΔNp63 may have a prognostic significance for HNSCC are still relevant questions in the field, and further studies are needed to clarify these issues.

Cells and culture conditions
Neonatal normal human epidermal keratinocytes (NHEKs, Life Technologies) were cultured in EpiLife medium with human keratinocyte growth supplements added (Life Technologies). FaDu (pharynx squamous cell carcinoma), SSC-090 (oral squamous cell carcinoma) and SSC-154 (tongue squamous cell carcinoma) cells were grown in Eagle's minimum essential medium (EMEM, Lonza, Basel, Switzerland); A253 cells (submaxillary salivary gland carcinoma) were cultured in McCoy's medium (Gibco, Invitrogen); SCC-9 cells (tongue squamous cell carcinoma) were cultured in1:1 and Igfbp3 genes, respectively. In HNSCC cells overexpressing ΔNp63, the transcription of Igf1r would be stimulated as a result of the unbalanced ratio between the TA/ΔN p63 proteins. Aberrant accumulation of IGF1R and its docking protein IRS1 would enhance signalling activation in response to receptor stimulation. On the other hand, reduced expression levels of Igfbp3 would increase the availability of circulating IGF1 that could further potentiate receptor activation.

RNA sequencing
RNA sequencing was performed as previously described [95]. Briefly, total RNA was extracted using a mirVana miRNA isolation kit (Thermo Fisher). rRNA was removed from each RNA extraction before proceeding with RNA seq library construction. Sequencing was performed on a SOLiD sequencer 5500XL (Applied Biosystems) with 75-base-pair single-end reads by Genomnia s.r.l. (Milan, Italy). Sequencing reads in SOLID "xsq" format were mapped to the hg19 genome built and analysed with the Lifetech Lifescope 2.5.1 Whole Transcriptomic analysis pipeline with the Integromics Seqsolve software and proprietary Genomnia procedures.

Chromatin immunoprecipitation assay
Fadu cells were used for ChIP assay. Cells were collected, fixed in 1% formaldehyde, and subjected to sonication for DNA shearing. The ChIP assay was performed with an anti-p63 antibody (H129, Santa Cruz Biotechnology) or unspecific immunoglobulin G (IgG) (Invitrogen) using a ChIP assay Kit (Invitrogen). PCR validation was performed using primers spanning the p63-binding sites located within the genomic regions identified by ChIP-seq assays.

Luciferase assay
For luciferase assays, a total of 1.2×10 6 H1299 cells were seeded in 12-well dishes 24 h before transfection. In total, 30 ng of pGL3 reporter vector, 5 ng of pRL-CMV-Renilla luciferase vector (Promega) and 150 ng of HA-ΔNp63α expression vectors or empty pcDNA-HA vector (as a control) were cotransfected using the Effectene transfection reagent according to the manufacturer's instructions (Qiagen). The luciferase activities of cellular extracts were measured 24 h after trans-AGING fection using a Dual Luciferase Reporter Assay System (Promega). The light emission was measured over 10 sec using a Lumat LB9507 luminometer (EG&GBerthold). The transfection efficiency was normalized to Renilla luciferase activity. The p63 RE was mutated by site-directed mutagenesis using the forward primer: 5'-ATAAGGGCCTTCCTGTTCCGA GGCAGCCGTGTGAACCACC-3' and reverse primer: 5' -GGTGGTTCACACGGCTGCCTCGGAACAGGA AGGCCCTTAT-3'.

Computational methods
Publicly available gene expression data from TCGA were downloaded from the Genomic Data Commons (GDC) Data Portal and pre-processed via the TCGAbiolinks R package [96]. Harmonized and normalized RPKM data from GDC was downloaded for Head and Neck Squamous Cell Carcinomas (TCGA-HNSC).
RPKM distributions for each of the genes of interest have been analyzed either for tumor and normal samples in the HNSC cohort. A discrimination cutpoint between tumor and normal expression distributions either for TP63 and IRS1 has been detected using OptimalCutpoints R package [97]. The identified optimal cutoffs represent the RPKM values that maxi-mize the separation of gene expression distributions in tumor and normal samples. Data mining, statistical tests (Wilcoxon and Fisher's exact tests) and plots on TCGA gene expression data were performed in R.

Statistical analysis
The significance of differences between two experimental groups was calculated using the two-tailed Student's t-test. Values with P<0.05 were considered significant.

CONFLICTS OF INTEREST
The authors of this manuscript declare no conflicts of interest.