ABSTRACT In heterogeneous head and neck cancer (HNC), subtype-specific treatment regimens are currently missing. An integrated analysis of patient HNC subtypes using single-cell sequencing

and proteome profiles reveals an epithelial-mesenchymal transition (EMT) signature within the epithelial cancer-cell population. The EMT signature coincides with PI3K/mTOR inactivation in

the mesenchymal subtype. Conversely, the signature is suppressed in epithelial cells of the basal subtype which exhibits hyperactive PI3K/mTOR signalling. We further identify YBX1

phosphorylation, downstream of the PI3K/mTOR pathway, restraining basal-like cancer cell proliferation. In contrast, YBX1 acts as a safeguard against the proliferation-to-invasion switch in

subtype as a promising therapeutic approach against HNC. SIMILAR CONTENT BEING VIEWED BY OTHERS SINGLE CELL DECIPHERING OF PROGRESSION TRAJECTORIES OF THE TUMOR ECOSYSTEM IN HEAD AND NECK

CANCER Article Open access 22 March 2024 INTEGRATIVE SINGLE CELL TRANSCRIPTOMIC ANALYSIS REVEALS 3P DELETION ASSOCIATED TUMOR MICROENVIRONMENT AND CHEMORESISTANCE IN HEAD AND NECK SQUAMOUS

CELL CARCINOMA Article Open access 10 March 2025 WNT7A PROMOTES TUMORIGENESIS OF HEAD AND NECK SQUAMOUS CELL CARCINOMA VIA ACTIVATING FZD7/JAK1/STAT3 SIGNALING Article Open access 22 January

2024 INTRODUCTION Head and neck cancer (HNC) is the sixth most common cancer worldwide with a poor overall survival rate1. At initial diagnosis, more than 60% of HNC patients present with

advanced-stage disease including local invasion, evidence of metastases to regional lymph nodes, or both2. HNC metastasis is a multi-step process that involves dissemination of cancer cells

from the primary tumour site, nodal metastases, intravascular survival and circulation, seeding and propagation at a secondary organ3. Metastatic HNC is highly resistant to therapy with a

five-year overall patient survival of less than 20% (approximately 10-months median overall survival)4. Therefore, understanding the molecular mechanisms of HNC metastasis is critical to

enhance therapy response and reduce cancer morbidity and mortality5. HNC tumours have been classified into four major molecular subtypes: basal, mesenchymal, classical and atypical whereby

the basal and mesenchymal subtypes represent more than 70% of HNC6. Tumours classified as mesenchymal reflect a basal transcriptomic signature in malignant epithelial cells with a large

stromal component and were described as malignant-basal7. Invasive cells at the leading edge of combined malignant-basal HNC undergo partial epithelial to mesenchymal transition (p-EMT)7.

While the classical and atypical HNC subtypes are able to metastasise at similar rates to the malignant-basal, they did not express p-EMT genes7. The p-EMT program reorganises the cell

cytoskeleton to enable motogenesis at the expense of mitogenesis and manifests by decreased expression of epithelial markers (e.g. CDH1) and increased expression of mesenchymal markers (e.g.

PDPN), required for HNC cell migratory and invasive abilities8,9,10. Moreover, in HNC patients, p-EMT serves as a predictive biomarker of nodal metastases, lymphovascular invasion and

extranodal extension3. While proliferative and invasive cancer cells retain common genetic alterations, discrete molecular processes evolve in proliferating cancer cells to acquire an

invasive potential11,12,13. However, molecular factors responsible for the proliferative to invasive switch remain poorly understood and represent an attractive, yet unexplored therapeutic

opportunity against metastatic cancers. Large-scale genomic and transcriptomic analyses of HNC revealed frequent dysregulation of the phosphatidylinositol 3-kinase (PI3K) signalling pathway

components such as _EGFR_, _PIK3CA_, _AKT_, _PTEN_ and _mTOR_14. Over 40% of HNC patients present with amplification of the _PIK3CA_ gene (encoding the catalytic subunit of the PI3K complex)

and up to 10% with _PIK3CA_ gain-of-function mutations1,15,16. However, compared to _PIK3CA_ wild-type HNC patients, treatment of gain-of-function _PIK3CA_-mutant HNC patients with the PI3K

inhibitor did not provide a therapeutic advantage17. These results suggest that other factors may influence PI3K signalling to confer therapy resistance in HNC. The

transcription/translation Y Box binding protein 1 (YBX1) is a prognostic biomarker for disease-specific survival in HNC and is significantly increased in high-grade HNC cells at the tumour

invasive front18. YBX1 which plays a crucial, albeit controversial, role in cell proliferation and invasion was shown to transcriptionally promotes cell proliferation in the nucleus19 and to

prevent protein translation in the cytoplasm20. In this study, we investigate the relationship between the PI3K signalling and YBX1 and their roles in HNC cell proliferation and invasion.

Our data identify HNC subtype-specific, distinct functions for YBX1 and PI3K-dependent phospho-YBX1 in the regulation of the proliferative to invasive switch, with therapeutic implications

against metastatic HNC. RESULTS THE EMT SIGNATURE INVERSELY CORRELATES WITH PI3K PATHWAY ACTIVATION IN BASAL AND MESENCHYMAL HNC PATIENT SUBTYPES To evaluate the EMT signature score in

patient HNC, single-sample gene set enrichment analysis (ssGSEA) was performed by applying hallmark gene sets to The Cancer Genome Atlas Head-Neck Squamous Cell Carcinoma (TCGA-HNC) cohort

(Fig. S1A). We identified two distinct subgroups of HNC based on EMT gene expression, EMT_Hi_ and EMT_Lo_ (Fig. 1A). ESTIMATE and Stroma scores on bulk RNA-seq show a biased EMT signature in

the mesenchymal subtype (Fig. S1B) with the overall survival, disease-specific survival, and progression-free survival are significantly poorer in EMT_Hi_ compared to EMT_Lo_ patients (Fig.

 S1C). We further conducted multivariate Cox regression survival analyses to adjust for clinically-relevant risk factors including HPV status, age, sex, and treatment (i.e., radiotherapy).

Importantly, the forest plot for overall survival hazard ratios indicates that the EMT signature is an independent indicator of survival outcomes after adjustment (Fig. 1B). Additionally,

mesenchymal tumours displayed a significantly higher EMT score, compared to the basal subtype (Fig. 1C). We further assessed whether any correlation exists between EMT and intracellular

signalling pathways in HNC tumours using TCGA-HNC transcriptomic data (Fig. S1D). A dynamic range of correlations appeared but importantly, a significant negative correlation (−0.3631,

_p_-value < 0.01) was obtained between EMT and PI3K/AKT/mTOR signalling (Fig. 1D). This was further confirmed by differentially expressed proteins using the TCGA-HNC reverse phase protein

array (RPPA) data (Fig. 1E). Interestingly, the G1/S mitotic cell cycle transition pathway was enriched in the basal subtype while positive regulation of cell migration was associated with

the mesenchymal subtype (Fig. 1F). Enrichment for the PI3K signalling pathway was consistent in both subtypes suggesting an important regulation of this pathway by activators and inhibitors.

These data indicate that PI3K/mTOR signalling is active in proliferative cells of the basal HNC subtype while an EMT signature in the mesenchymal HNC subtype correlates with the inhibition

of PI3K/mTOR signalling, decrease in cell proliferation, induction of cell migration, and a poor prognosis. SINGLE CELL ANALYSIS OF PATIENT HNC UNCOVERED AN EPITHELIAL-SPECIFIC INVERSE

CORRELATION BETWEEN PARTIAL EMT AND PI3K SIGNALLING To accurately evaluate the EMT status of malignant epithelial cells and its correlation with PI3K signalling outside the biased bulk

RNA-seq (Fig. S1B), we explored transcriptomic heterogeneity at a cellular resolution using single-cell RNA sequencing (scRNA-seq). scRNA-seq profiles of primary and metastatic tumours were

generated from four treatment-naïve patients (GSE140042) (Fig. 2A, B). The single-cell transcriptomes from 12,341 cells were retained after initial quality controls and partitioned into 13

clusters by gene expression levels. Individual clusters were then annotated according to characteristic marker genes (Fig. 2C–E). Malignant epithelial cells (cluster 4) were identified

within the cellular heterogeneity of the tumour using epithelial-specific markers (Fig. 2F). A finer TCGA-HNC subtype classification (Fig. 2G) depicted intra-cellular heterogeneity within

the malignant epithelial cells, demonstrating a dominant basal subtype signature. Within the malignant epithelial cells, we observed enrichment for the EMT signature in G1-arrested cancer

cells while the PI3K/AKT/mTOR signalling was active in cycling cells (S and G2/M phases) (Fig. 2H). This agrees with the TCGA-HNC bulk RNA sequencing data in which the EMT signature

negatively correlates with PI3K-active cycling cells (Fig. 1D). Notably, signatures from these cells differ between primary tumours and matched lymph node metastases (Fig. 2G); lymph node

metastatic cells characteristically lacked the mesenchymal gene signature. Evaluation of the p-EMT program identified key p-EMT genes including extracellular matrix (PDPN), mesenchymal

markers and EMT regulators in metastatic epithelial cells (Fig. 2I, J). Moreover, cells with a p-EMT_Hi_ signature had a decreased proliferation rate (Fig. 2K). We validated the findings

using two independent scRNA-seq datasets (GSE103322: 2215 tumour cells from 7 patients and GSE164690: 13875 tumour cells from 11 patients) to confirm the inversed correlation between EMT and

PI3K-related cell proliferation. The cell cycle analysis shows enrichment for the EMT signature in G1-arrested cancer cells while the PI3K-AKT-mTOR signalling was active in cycling cells

(Fig. S2), consistent with the scRNA-seq analysis of patient samples (GSE140042) and bulk RNA-seq studies. Taken together, our scRNA-seq analysis of epithelial HNC cells identified a p-EMT

state that inversely correlates with PI3K/mTOR signalling and cell cycling at a single epithelial cell level. THE YBX1 FACTOR LINKS THE PI3K SIGNALLING TO PARTIAL EMT IN SPECIFIC HNC

SUBTYPES Six HNC cell lines were classified into molecular subtypes using the transcriptomic signature of subtypes identified in the TCGA-HNC and the Cancer Cell Line Encyclopaedia (CCLE)

databases21. SCC15 and SCC25 corresponded to the mesenchymal and basal like subtypes, respectively, and showed the highest correlation coefficients among the cell lines (Fig. 3A). Analysis

of the PI3K pathway mutation in these cells revealed a missense mutation (putative driver) in the PI3K regulatory subunit (_PIK3R3_) gene in the mesenchymal-like SCC15 cells (Fig. 3B).

Interestingly, _PIK3R3_ is only found to be mutated in mesenchymal patient HNC (Fig. S3)22. Additionally, basal-like SCC25 cells had no PI3K signalling related mutations. To investigate the

direct role of PI3K signalling in HNC, we developed a relevant genetic mouse model by breeding the conditional _Pik3ca__H1047R_ knock-in mouse23 onto a _Grhl3_-deficient background24,25 to

induce the development of PI3K-dependent HNC; GRHL3 being a differentiation factor that prevents HNC26,27 and is downregulated in HNC samples of the TCGA cohort, with a comparable level

between epithelial cells of basal and non-basal subtypes in scRNA-seq (Fig. S4A). Knock-in of an _H1047R_ mutation into one allele of the endogenous _Pik3ca_ gene in keratin-14 positive

epithelial cells using a constitutive K14-Cre recombinase did not show any head and neck phenotype. Interestingly, spontaneous tumours developed in _Pik3ca_H1047R_Grhl3_cKO double mutant

mice in less than 3 months (Fig. S4B). Marked hyperactivation of PI3K signalling was observed in the tumours along with expression of the cell proliferation markers Ccnd1 and pMet (Fig. 

S4C), the epithelial marker Cdh1 and the phosphorylation of Ybx1, but not the p-EMT marker Pdpn (Fig. S4D). This data indicates that PIK3CA-driven tumour cell proliferation does not induce

p-EMT. Transcriptomic analysis of mesenchymal SCC15 cells demonstrated upregulation of mesenchymal markers and downregulation of PI3K pathway-related genes. In contrast, transcriptomic

analysis of basal SCC25 cells indicated an increased mRNA expression of epithelial markers and upregulation of PI3K genes (Fig. 3C). These findings were validated by proteomic RPPA analyses

showing hyperactivation of the PI3K signalling (upper panel) and expression of the proliferative markers (lower panel) in basal SCC25 cells (Fig. 3D) compared to the mesenchymal SCC15 cells

which showed inactivation of the PI3K signalling and expression of the p-EMT marker PDPN (Fig. S5A). Interestingly, SCC15 and SCC25 show an opposite pattern of expression for the EMT markers

N-cadherin and Twist1 (Fig. S5B). To validate our findings in vivo, we established an orthotopic HNC xenograft mouse model by transplanting luciferase-tagged HNC cells into the tongue of

immunocompromised mice. This approach recapitulated the location of primary tumours in patients21 and allowed weekly monitoring of in vivo tumour progression using bioluminescence imaging

over a 6-week period (Fig. 3E). Interestingly, engrafted basal-like cells developed in situ tumours within the tongue while mesenchymal-like cells grew invasive tumours that underwent

regional metastasis to lymph nodes (Figure S5C). Additionally, CDH1 expression (epithelial marker) was observed in basal SCC25-derived tumours whereas mesenchymal SCC15 tumours were positive

for PDPN (p-EMT marker) (Fig. 3F). The PI3K signalling components EGFR, EIF4E and pEIF4E were highly expressed in SCC25 xenografts compared to SCC15 (Figure S5D). This data indicates that

the PI3K-active basal subtype favours tumour growth over invasion while p-EMT induction in the PI3K-inactive mesenchymal subtype confers a metastatic potential in vivo. To identify the

molecular switch linking PI3K inactivation to p-EMT induction, we investigated oncogenic factors that are known to regulate HNC cell proliferation and invasion. We prioritised the YBX1

factor whose overexpression and cytoplasmic localisation were identified at the invasive front of metastatic patients HNC18. Furthermore, YBX1 phosphorylation at Ser102 by the PI3K

signalling was recently shown to positively correlate with the expression of EGFR to facilitate cell proliferation and tumour growth19. Hence, the expression and phosphorylation of YBX1 were

assessed in subtype-specific HNC cells and xenografts. WB analyses showed the highest level of total YBX1 in mesenchymal SCC15 cells and phosphorylated forms of YBX1 in basal SCC25 cells

within the HNC cell lines (Fig. 3G). Additionally, YBX1 was phosphorylated at Ser102 in response to 100 ng/ml of EGF or 20% FBS treatments of SCC25 and this phosphorylation was prevented by

the dual PI3K/mTOR inhibitor BEZ235 (Figs. S5E and S5F). Immunofluorescence (IF) staining detected YBX1 mainly in the nuclear compartment of SCC25 whereas YBX1 was cytoplasmic in invasive

SCC15 cells and phospho-YBX1 localised specifically to the mitosis spindles in dividing basal SCC25 cells (Fig. 3H). YBX1 and phospho-YBX1 IF quantification shows higher nuclear localisation

in SCC25 compared to SCC15 (Fig. 3I) which was further validated using WB on cytoplasmic and nuclear cellular fractions following EGF treatment (Fig. S5G). This data demonstrates that

nuclear YBX1 occurs in PI3K-active proliferative cells while its cytoplasmic counterpart associates with the PI3K-inactive invasive subtype. LOSS OF YBX1 INHIBITS CELL PROLIFERATION IN

BASAL-LIKE HNC CELLS WITH ACTIVE PI3K SIGNALLING To evaluate the oncogenic function of YBX1 in the basal subtype, we employed a doxycycline (DOX)-induced CRISPR-Cas9 system with single guide

RNAs against _YBX1_. Q-PCR and WB analyses were used to assess YBX1 knockdown efficiency (~80%) (Fig. 4A). Interestingly, YBX1 knockdown decreased the 3D-growth of SCC25 basal-like cells in

an ultra-low attachment condition with a significant reduction in the number and size of colonies (Fig. 4B). Furthermore, diminished cell proliferation correlated with reduced SCC25

invasion in a transwell invasion assay (Fig. 4C). Whole transcriptome analysis of _YBX1_-knockdown SCC25 (–YBX1) compared to SCC25 parental (+YBX1) cells showed enrichment for genes involved

in the suppression of G2/M checkpoint and ribosomal biogenesis, and activation of apoptosis and anti-tumour inflammatory responses (Fig. 4D). Importantly, YBX1 downregulation resulted in a

significant loss of the E2F, YY1, and MYC/MAX transcription factor target genes. The gene signature which correlates with cell proliferation and survival was enriched in +YBX1 cells (Fig. 

4E). Moreover, SCC25 –YBX1 cells treated with EGF showed strong activation of PI3K/mTOR signalling, suggesting that activation of PI3K signalling is a compensatory mechanism in basal-like

cells (Fig. 4F). We next established orthotopic SCC25 (–/+YBX1) xenografts. YBX1 downregulation strikingly inhibited tumour growth as measured by bioluminescence imaging, the average

intensity radiance and tumour weight (Fig. 4G). Residual tumours from SCC25 –YBX1 were confirmed negative for YBX1 by immunohistochemistry (IHC) staining (Fig. 4H). NSG mice bearing SCC25

(−/+YBX1) xenografts were administered daily the PI3K/mTOR inhibitor BEZ235 (35 mg/kg) by oral gavage over a 4-week period. Loss of YBX1 reduced tumour growth of SCC25 which was further

reduced in response to BEZ235 treatment (Figs. S6A and S6B). Inactivation of the ribosomal protein S6 shown by loss of p-RPS6 compared to total RPS6 confirmed the efficacy of BEZ235

treatment in SCC25 xenografts (Fig. S6C). This data validates the oncogenic function of YBX1 in promoting tumour development of basal HNC cells with active PI3K signalling. CYTOPLASMIC YBX1

INDUCES P-EMT IN THE ABSENCE OF PI3K SIGNALLING The function of YBX1 was evaluated in mesenchymal-like SCC15 cells with inactive PI3K signalling. _YBX1_ knockdown efficiency (~80%) in SCC15

(–YBX1) was determined by Q-PCR and WB (Fig. 5A). Compared to SCC15 (+YBX1), SCC15 (–YBX1) cells proliferated more and grew into bigger colonies in an ultra-low attachment condition (Fig. 

5B) and were significantly more invasive in the transwell invasion assay (Fig. 5C). RNA sequencing of SCC15 –YBX1 demonstrated enrichment for genes related to PI3K/mTOR signalling and

translation initiation (Fig. 5D). Furthermore, downregulation of negative PI3K regulators (e.g. TSC2) and upregulation of proliferation suppressors (e.g. CDKN1A) were observed in SCC15 –YBX1

cells, independent of EGF treatment (Fig. 5E), indicating a reduced PI3K-dependent proliferation. Next, we established orthotopic SCC15 (–/+YBX1) xenografts to investigate the role of YBX1

in metastasis. Unlike SCC25, SCC15 –YBX1 xenografts exhibited a faster growth rate and a significantly higher metastatic potential compared to SCC15 (+YBX1) (Fig. 5F and Figs. S6A and S6B).

However, tumour weight did not significantly differ between –YBX1 and +YBX1 xenografts (Fig. 5F). IHC and IF staining of YBX1 in GFP-labelled invasive SCC15 cells demonstrated YBX1

localisation to the invasive front of primary tumours (Fig. 5G). Moreover, YBX1 was strongly expressed in the lymph node metastatic lesions. NSG mice bearing SCC15 (−/+YBX1) xenografts were

administered daily the PI3K/mTOR inhibitor BEZ235 (35 mg/kg) by oral gavage over a 4-week period. While loss of YBX1 induced a growth advantage for SCC15 xenografts, SCC15 –YBX1 developed

therapy resistance to BEZ235 treatment with increased metastasis to lymph nodes (Figs. S6A and S6B) while p-RPS6 was completely loss (Fig. S6C). Taken together, these data indicate that YBX1

is a suppressor of metastasis in mesenchymal HNC with inactive PI3K signalling. Furthermore, we overexpressed a myc-tagged YBX1 in invasive SCC15 cells. Transfected cells (57.4%) were

confirmed for YBX1 overexpression (OE) using flow cytometry (Fig. S7A) and by anti-MYC and anti-YBX1 western blots (Fig. S7B). The transwell invasion assay shows OE cells with nuclear YBX1

expression in green (IF for the MYC tag) are retained in the inner membrane (non-invasive) whether SCC15 cells with cytoplasmic YBX1 are detected in the outer membrane (invasive). The

quantification of YBX1 OE demonstrates decreased nuclear to cytoplasmic ratio in invasive cells (Fig. S7C). While YBX1 is mainly shown in the cytoplasm of SCC15 cells (Fig. 3H), this data

indicates that OE of YBX1 in SCC15 can localise to the nucleus to inhibit cell invasion. THE PI3K-PHOSPHO-YBX1 AXIS IN BASAL AND MESENCHYMAL HNC SUBTYPES PREDICTS PATIENT PROGNOSIS

Considering the high rate of _PIK3CA_ mutations in HNC, we evaluated the clinical significance of the PI3K-phospho-YBX1 signalling in patients. RPPA data from basal and mesenchymal HNC

subtypes were extracted from the TCGA-HNC dataset (_n_ = 58). Compared to basal HNC, mesenchymal tumours showed a lower YBX1 phosphorylation rate, but no significant difference in the

expression of total YBX1 (Fig. 6A). In addition, phospho-YBX1 expression level, but not total YBX1, was found to be inversely correlated with the EMT score by a Spearman’s correlation

analysis (Fig. 6B). An unsupervised clustering of basal and mesenchymal patient HNC subtypes identified an inverse correlation of the PI3K-phospho-YBX1 axis with the EMT markers (Fig. 6C).

Using the TCGA-HNC and matched RPPA data, expression of phospho-YBX1 predicted survival outcomes of HNC patients from all subtypes (_n_ = 337). Overall survival rates of patients with high

expression of phospho-YBX1 (p-YBX1_Hi_) were significantly and independently improved when compared to p-YBX1_Lo_ patients (Fig. 6D). In addition, the p-YBX1_Hi_ HNC tumours were more

differentiated with lower histological grade, disease stage and lymph node involvement compared to p-YBX1_Lo_ HNC patients (Fig. 6E). It should be noted that total YBX1 expression did not

significantly differ between p-YBX1_Lo_ and p-YBX1_Hi_ tumours (Fig. 6F), underscoring the impact of YBX1 phosphorylation on patient outcomes. Furthermore, a detailed analysis of p-YBX1_Hi_

tumours identified upregulation of the PI3K signalling pathway as shown in the volcano plot (Fig. 6G). The data was validated in patient HNC tissue microarrays by IHC staining and

demonstrated a mutually exclusive expression pattern for phospho-YBX1 and PDPN, particularly at the invasive front of HNC (Fig. 6H and Figs. S8A–C). Overall, our data establish the

PI3K-phospho-YBX1 axis as an oncogenic driver of tumour growth in patient HNC basal subtype and the YBX1 factor as a suppressor of metastasis in the mesenchymal subtype (Fig. 7). DISCUSSION

Previously, heterogeneous HNCs have been stratified into four distinct gene-expression subtypes28. Basal, mesenchymal, atypical and classical subtypes were characterised by somatic

mutations, copy number alterations, gene expression and DNA methylation profiles15,29. Recently, HNC tumours were suggested to be refined into three subtypes (malignant-basal, classical, and

atypical), with the mesenchymal subtype reflecting malignant-basal tumours with a large stromal component7. Our current data agrees with the HNC malignant-basal subtyping analysis and

further adds a PI3K/mTOR-dependent sub-classification for potential targeted therapy stratification. Mutations in components of the PI3K pathway are among the most frequent in HNC22 with an

increased burden in metastatic disease5. The clinical trials using PI3K signalling inhibitors have shown substantial differences in therapy response against solid cancers including

HNC14,30,31,32. In line with this, our data indicate that the mesenchymal subtype is molecularly programmed to endure PI3K/mTOR inhibition and suggest the inclusion of HNC molecular

subtyping in future trials to predict therapy response. Moreover, since basal-like SCC25 cells are highly sensitive to multiple PI3K inhibitors33, chemotherapy (Cisplatin/5-FU)34,35 and

Cetuximab36, this study provides mechanistic insights into therapy resistance37 and the downstream effectors in the PI3K-inactive mesenchymal subtype. Considering the differential response

of basal and mesenchymal subtypes to Cisplatin/5-FU, HNC patient of the basal subtype are proposed to respond better than those of the mesenchymal subtype. Our data are also relevant to the

induction of EMT in resistant Cetuximab-treated HNC in the clinic38. Because the proliferation-invasion switch is the basis of therapy resistance, metastasis and cancer mortality39,40, we

showed phosphorylation of YBX1 in basal-like HNC cells with active PI3K signalling confers an oncogenic role for this factor, while unphosphorylated YBX1 acts as a suppressor of metastasis

in mesenchymal-like cells with inactive PI3K signalling. This subtype-specific function of YBX1 is PI3K-dependent and underscores a pivotal role for YBX1 in fine-tuning the cancer cell

fitness with a critical impact on HNC patient outcomes. Genetic alterations are rarely detected at the _YBX1_ locus, with only 2 amplifications identified in 2 HNC samples and a missense

mutation (R279) in 1 sample (TCGA-HNC patient database, _n_ = 529) (Fig. S8D). Moreover, YBX1 was shown to be regulated at the post-translational level as the main determinant of its

function41. PI3K-mediated phosphorylation of YBX1 at Ser102 induces its translocation to the nucleus and results in the transcription of E2F, YY1 and MYC/MAX proliferative target genes

(Figs. 4D, E)41,42. It remains to be tested whether other phosphorylations of YBX1 such as phospho-S20943 affect the YBX1 function in tumour growth and invasion44, and in _PIK3CA_ mutant HNC

as seen in _JAK2_ mutant hematopoietic cancers45. In the absence of PI3K signalling, YBX1 binds to mRNAs of p-EMT markers in the cytoplasm20. Reduced levels of cytoplasmic YBX1 promote

translation of p-EMT transcripts while high levels compete with eIF4E to inhibit the initiation of translation46. This function is evident in the mesenchymal-like SCC15 with inactive

PI3K/mTOR/eIF4E and expression of cytoplasmic YBX1. Interestingly, _YBX1_ knockdown in these cells activates the initiation of translation and results in increased colony growth and

invasion. Therefore, YBX1-dependent translational regulation of p-EMT factors constitutes a vulnerability for therapeutic reversion of the metastatic phenotype47. Our data highlights the

role of epithelial factors as predominant regulators of the EMT switch and cancer cell fitness. It is equally important to consider the role of the tumour microenvironment in future studies,

particularly the interactions between immune and malignant epithelial cells through secreted proteins48,49. Moreover, the immune checkpoint PD-1/PD-L1 inhibitors Nivolumab and Pembrolizumab

are FDA-approved for patients with recurrent or metastatic HNC and, interestingly, YBX1 has been shown to regulate the expression of PD-L1 within an immunosuppressive microenvironment50,

establishing this factor as an important therapeutic target against both epithelial cancer cells and the tumour microenvironment. In conclusion, this study discovered a mutually exclusive

interplay between PI3K-mediated cell proliferation and p-EMT-initiated HNC invasion in basal and mesenchymal HNC subtypes and identified PI3K-dependent phosphorylation of YBX1 as a limiting

factor of the switch to invasion. Our data also associate the subtype-specific PI3K-phospho-YBX1 axis with the patient survival outcomes. Future studies aiming at activating phospho-YBX1 may

prove effective in shifting the mesenchymal phenotype towards a basal subtype and would constitute a promising therapeutic avenue against metastatic HNC. METHODS RNA-SEQ ANALYSIS The total

RNA quantity was measured using Qubit RNA HS (Thermo Fisher Scientific). 500 ng of total RNA was used for library preparation according to standard protocols (QuantSeq 3’mRNA-seq FWD,

Lexogen). Indexed libraries were pooled and sequenced on a NextSeq500 (Illumina). A total of 5–15 million single-end 75 bp reads were generated per sample. Sequence reads were trimmed and

aligned to hg38 genome using Cutadapt and HISAT2 packages. Gene counts were obtained from featureCounts. Expression normalisation and filtration were performed using Limma. All analysis

packages were operated within the Galaxy suite environment (version 4.0). Gene set enrichment analysis (GSEA_v4.3.2) was used for functional enrichment analysis of annotated terms from GO,

KEGG and Hallmark using the human genome (GRCh38) as background. BIOINFORMATIC DATASETS AND ANALYSES Publicly available datasets were retrieved from TCGA Research Network using cBioportal

(cbioportal.org). Gene and protein expression and patient survival information for 337 HNC samples were analysed. Sample purity was calculated using the tumour purity algorithm ESTIMATE.

Gene mutation and expression profiling of the cell lines were retrieved from CCLE using depmap portal (depmap.org). Publicly available scRNA-seq data (“GSE140042”), (“GSE103322”),

(“GSE164690”) were retrieved from Gene Expression Omnibus (GEO) (ncbi.nlm.nih.gov/geo). Single-sample gene set enrichment analysis (ssGSEA) module on GenePattern (cloud.genepattern.org) was

used with EMT and PI3K gene sets to generate the EMT and PI3K signature scores (HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION, HALLMARK_PI3K_AKT_MTOR_SIGNALING). SINGLE CELL ANALYSIS

Droplet-based scRNA-seq data (GSE140042 and GSE164690) on the 10X Genomics Chromium platform were processed in RStudio using the Seurat package (version 4.1.0). Only HNC samples with

2000–4000 sequenced cells were considered for downstream analyses while others were excluded. Cells with more than 200 RNA features were retained and RNA features detected in more than 2

cells were considered for scRNA-seq analyses. Low-quality barcodes and unique molecular identifiers (UMIs) were filtered, mapped to human genome (GRCh38) and batch normalised using the Cell

Ranger pipeline v3.0.2 (10X Genomics). The cell-gene matrix of UMI counts was then imported to Seurat. Genes expressed in >2 cells and cells with at least 200 genes were retained. Seurat

objects were subsequently normalised, scaled, and integrated using CCA algorithm. A dimensional reduction matrix and clustering were obtained by aligning the CCA subspaces. Differentially

expressed genes in individual clusters were identified and used for cell type assignment in each cluster by the CellMarker human cell markers. The cell-cycle score was calculated based on

the expression of S and G2/M phase markers, and cells expressing neither marker were identified in the G1 phase. EMT, p-EMT and PI3K ssGSEA scores were achieved using escape (version 1.3.3).

For TCGA subtype assignment, differentially expressed genes (log fold-change > 2 and _p_ < 0.01 by limma) in each of the four subtypes were identified, and single cells were scored by

the four subtype gene sets using SingleR (version 1.8.1). HNC CELL LINES AND CELL CULTURE The normal oral epithelial cell line OKF6 was purchased from the Harvard Skin Disease Research

Centre (HSDRC, Boston MA). The oral cancer cell lines SCC9 (CRL-1629), SCC15 (CRL-1623), SCC25 (CRL-1628), and CAL27 (CRL-2095), A253 (HTB-41) and FaDu (HTB-43) were purchased from the

American Type Culture Collection (ATCC, Manassas, VA). All cell lines were authenticated and validated by short tandem repeat (STR) profiling and tested negative for mycoplasma

contamination33. OKF6 and SCC25 were cultured in keratinocyte serum-free medium (K-SFM, GibcoTM) supplemented with growth factors (25 μg/mL BPE, 0.2 ng/mL EGF and 0.3 mM CaCl2) and 1%

penicillin-streptomycin (P/S). SCC9, SCC15, and CAL27 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, GibcoTM) with 10% foetal bovine serum (FBS) and 1% P/S. All cell lines

were cultured at 37 °C in a 5% CO2 humidified incubator and maintained at less than 25 passages. To activate the PI3K signalling, cells were grown in normal media and treated with EGF (100 

ng/mL) for 30 minutes. To inhibit the PI3K signalling, cells were grown in normal media and treated with 100 nM BEZ235 (HY-50673, MedChemExpress) for 6 hours. The media was removed, and

cells were washed with phosphate-buffered saline (PBS), then lysed in RIPA buffer for western blot analyses. GENERATION OF LENTIVIRAL-TRANSDUCED STABLE CELL LINES Lentiviral expression

constructs were mixed with the packaging plasmids pMDLgRRE, pRSV-Rev, pCMV-VSVG (Addgene) and transfected into lenti-X293T cells (Takara Bio, #632180) using Lipofectamine 2000 (Thermo Fisher

Scientific) according to the manufacturer’s protocol. Virus-containing conditioned media was collected 48–72 h after transfection and filtered using a 0.45 µm filter. Viral conditioned

media was supplemented with polybrene and applied to sub-confluent cells. Conditioned media was replaced with fresh cell culture media 6 h post transduction. Double-positive GFP and

mCherry-expressing cells were sorted using a FACS Calibur cell sorter 72 h post transduction and then expanded in culture. Doxycycline (Dox) Hyclate (Sigma Aldrich) was added (1 µg/ml) to

the culture media to induce CRISPR-Cas9-mediated gene deletion for downstream analysis. Gifted plasmids were used for the inducible sgRNA construct pFgh1UTG and the FUCas9Cherry. The sgRNA

targeting sequences for YBX1 (5’-GTAATGGCTTTTGTAGGGTG-3’ and 5’-GTTTGACACCGTTCATTGCA-3) were designed using the web-based tool CHOPCHOP (http://chopchop.cbu.unib.no/). The knockout

efficiency was validated by quantitative polymerase chain reaction (Q-PCR) using human YBX1 primers (forward: 5’-AAGTGATGGAGGGTGCTGAC-3’ and reverse: 5’- TTCTTCATTGCCGTCCTCTC-3’) and

housekeeping GAPDH primers (forward:5’- ACCCAGAAGACTGTGGATGG −3’ and reverse: 5’- CAGTGAGCTTCCCGTTCAG −3’). CLONOGENICITY AND INVASION ASSAYS Clonogenicity was assayed by seeding 500 cells

of SCC15 and SCC25 at day 0 onto ultra-low attachment plates (Corning) and cultured in normal growth media. After 7 days, the images of spheroids were recorded on an EVOS live cell imaging

microscope. The number of colonies was counted using Image J (version1.53). For the invasion assay, 24-well transwell plates with 8 μm pore size (Corning Costar, USA) were used. The

polycarbonate membrane was coated with Matrigel (Corning, USA) diluted in a coating buffer at 1:9 ratio and allowed to set for 2 hours. Next, cells were starved for 24 hours and seeded at

100,000 cells per Matrigel-coated well in a serum-free medium (top chamber) while complete culture media was added into the bottom chamber. The cells were incubated for 25 h at 37 °C in a

humidified 5%CO2 incubator and then fixed in 4% formalin at 4 °C overnight. Invasive cells in the transwell membranes were permeabilised with triton X-100 and stained by 1 ng/mL DAPI dye in

methanol. Images were acquired using a BX-51 Olympus microscope and SPOT software 5.0. PROTEIN EXTRACTION, WESTERN BLOTTING (WB) AND IMMUNOHISTOCHEMISTRY (IHC) Cells were lysed in RIPA

buffer with 1X protease inhibitor (Roche), and protein concentrations were determined using the DC protein assay kit (Bio-Rad) against BSA (Bovine serum albumin) standards according to the

manufacturer’s protocol. Absorbance readings were made on a Cytation 3 cell imaging multi-mode reader (Agilent, USA). For WB analysis, 30 µg total protein was denatured (95 °C, 2 min) in

Laemmli Buffer and resolved on 10% polyacrylamide gels. Proteins were transferred to PVDF membranes (Millipore), blocked in 5% slim milk powder or BSA in TBS-T (Tris Buffered Saline with

Tween) buffer for 2 hours. Membranes were then incubated with the primary antibodies listed in Supplementary Table 1 in TBS-T with 5% (w/v) BSA or 5% (w/v) milk powder overnight at 4 °C.

Membranes were then washed in TBS-T and incubated with HRP-conjugated secondary antibodies. Membranes were washed and developed using Clarity Western ECL Blotting Substrate (Bio-Rad) and

imaged on a ChemiDoc MP Imaging System (Bio-Rad). β-actin and tubulin were used as protein loading controls. For IHC, whole mouse tongues were fixed in formalin and paraffin embedded. Tumour

sections (5 µM) were deparaffinized, antigen was retrieved in sodium citrate and the endogenous peroxidase activity was blocked using 3% H2O2. The sections were then blocked with 5% normal

goat serum and incubated with primary antibodies overnight followed by biotin-linked secondary antibodies incubated for 1 hour at room temperature. Chromogenic colour was developed using

3,3-diaminobenzidine (DAB). Slides were then counter-stained in Hematoxylin and a coverslip was mounted. No primary or no secondary antibodies were included in our IHC experiments as

controls for the antibody staining specificity. Knockout cells were additional negative controls to validate the YBX1 antibody specificity. The patient HNC tissue arrays contained 123

patient tumours that were stained in serial sections for phospho-YBX1 and PDPN. All microscopy images were acquired using a BX-51 Olympus microscope and SPOT software 5.0. The

semi-quantification of the nuclear or cytoplasmic localisation in SCC15 and SCC25 was performed using CellProfiler 4.2.4. The semi-quantification of pYBX1 and PDPN region of interests (ROIs)

in serial sections of the HNC tissue arrays was analysed on the HALO quantitative image analysis platform version 2.0 (Indica Labs) using random forest. GENETIC MOUSE MODELS AND ORTHOTOPIC

XENOGRAFTS All animal studies including breeding, experiments and euthanasia protocols were performed in accordance with the National Code of Practice for the Care and Use of Animals for

Scientific Purposes and the Peter MacCallum Cancer Centre Animal Ethics guidelines and were approved by the Institutional Animal Experimentation Ethics Committee (E587 and E632). Mice were

housed under 12 hours light-dark cycles at 20-26 C and 20-70% humidity, in ventilated cages and constant access to food and water. The genetic _Pik3ca_23 and _Grhl3_24 mouse models were

characterised previously. Both animal strains were established on a C57BL/6 background and males and females equally used at 3 months old. The animals were monitored according to the Peter

Mac Animal Ethics guidelines and maximal tumour size and burden were not exceeded. The patient samples were collected under the University of Western Australia Human Research Ethics Approval

number RA/4/1/8562 and made into tissue arrays. Informed consent was obtained by all participants. Established cell lines derived from patients with HPV-negative (SCC15 and SCC25) HNC were

used to generate orthotopic xenografts in NSG mice (6–8 weeks of age). HNC cells (_n_ = 10,000) were transduced with a lentivirus containing the luciferase (pCDH-EF1a-eFFly-mCherry, Addgene

plasmid #104833) reporter and mCherry genes. The cells were then sorted and cultured to increase cell numbers. The cells were then resuspended in media containing Matrigel and culturing

media (1:1) and orthotopically injected into the mouse tongue. Tumour growth was monitored weekly by bioluminescence imaging. The experimental mice were injected intraperitoneally with

D-Luciferin (150 mg Luciferin/kg body weight prepared in PBS) and then anaesthetised using 4-unit isoflurane. 5 minutes following luciferin injection, bioluminescence images were captured at

1 second exposure with the IVIS Spectrum In Vivo Imaging System. The mice were culled at the experimental endpoint, or earlier if they showed signs of distress, and their xenografts were

collected for downstream analyses. REVERSE PHASE PROTEIN ARRAY (RPPA) Protein lysates from human HNC cells were extracted using the RPPA lysis buffer and quantified by the Pierce Coomassie

Plus (Bradford) Assay Kit (ThermoFisher Scientific, Cat No. 23236). Sample lysates were run at the Host and Tumour Profiling Unit (HTPU, Cancer Research UK, Edinburgh, UK). RPPA Relative

Fluorescence Intensity (RFI) values were calculated by the Zeptoview software. A weighted linear regression through the dilution series was used to calculate the sample fluorescence

intensity value which was then normalised to the reference BSA grid to account for intra-array spatial variation. Each intensity value was corrected to the background signal and the

secondary antibody controls to validate the RFI value for each sample/antibody combination. ‘0’ accounted for RFI values where the primary antibody signal was lower than the signal emitted

by the secondary antibody alone. 1e-9 was added to all RFI values for statistical analysis. STATISTICAL ANALYSES AND REPRODUCIBILITY Statistical significance was assessed using the unpaired

Student’s t test, one-way ANOVA, Spearman’s correlation test, two-sided Wald test for multivariate Cox regression analysis or log-rank test for survival analysis using Prism 9 (GraphPad).

Statistical analyses for the RPPA, RNA sequencing and single-cell RNA sequencing were carried out using R version 4.0.5. Packages used for single-cell RNA sequencing analysis included Cell

Ranger pipeline v.3.0.2, Seurat_4.1.0; SummarizedExperiment_1.20.0; clustree_0.4.4; escape_1.3.3; SingleR_1.8.1. Packages used for RNA sequencing analysis included limma_3.46.0, GSEA_v4.3.2

and Galaxy suie environemtn (version 4.0, including Cutapdapt, HISAT2, featureCounts and Limma). Packages used for plotting included tidyverse_1.3.1, ComplexHeatmap_2.8.0 and ggpubr_0.4.0.

The flow cytometry data was analyzed using FACSDiva 9.0. SPOT software 5.0, CellProfiler_4.2.4 and HALO quantitative image analysis platform 2.0 were used for IHC image analysis. The results

are presented as mean ± SEM. The statistical values were considered significant at *_p_-value < 0.05, **_p_-value < 0.01, ***_p_-value < 0.001 and **** _p_-value < 0.0001.

Experiments were repeated at least once. Replicates were reproducible. Schematic illustrations in Fig. 3E, Fig. 7 and supplementary Fig. 4B were created using Biorender.com. REPORTING

SUMMARY Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. DATA AVAILABILITY The data that support this study are available

in this paper and stored in GEO database as SuperSeries GSE226357 (“GSE226357”), including RNAseq raw data (“GSE226355”), and RPPA raw data (“GSE226356”) or from the corresponding author

([email protected]) upon reasonable request. Publicly available data from TCGA, Broad Institute and Stanford University were used. TCGA genetic, transcriptomic, proteomic, and

clinical data were downloaded from the cBioportal data portal (https://www.cbioportal.org/; Head and Neck Squamous Cell Carcinoma (TCGA, PanCancer Atlas)). Genetic and transcriptomic data

from Broad Institute were downloaded from depmap portal (https://depmap.org/portal/; CCLE_expression.csv; CCLE_mutation.csv). The single-cell RNA sequencing data (“GSE140042”),

(“GSE103322”), (“GSE164690”) were downloaded from GEO database (https://www.ncbi.nlm.nih.gov/gds). Raw data for bulk mRNA-seq on mouse tissues and human cancer cell lines and RPPA data from

different conditions have been deposited on the Figshare repository. Raw data on mouse tissues and human cancer cell lines from different conditions are available on figshare; for bulk

mRNA-seq (“20024246”) and for RPPA (“20024258”). Public single-cell RNA-seq data on patient samples were downloaded from GSE140042, GSE103322 and GSE164690. The processed single-cell RNA-seq

data is available on figshare (“20033024”). Additional data are available as supplementary materials and source data as Source Data files. Source data are provided with this paper. CODE

AVAILABILITY All software algorithms used for analysis are available for download from public repositories. All code used to generate figures in the manuscript are made available in the

following Github repository: https://github.com/DaridoLab-HNSC/YBX1-integration-of-oncogenic-PI3K-signalling.git. REFERENCES * Johnson, D. E. et al. Head and neck squamous cell carcinoma.

_Nat. Rev. Dis. Primers_ 6, 92 (2020). Article  PubMed  PubMed Central  Google Scholar  * Chow, L. Q. M. Head and neck cancer. _N. Engl. J. Med._ 382, 60–72 (2020). Article  CAS  PubMed 

Google Scholar  * Lambert, A. W., Pattabiraman, D. R. & Weinberg, R. A. Emerging biological principles of metastasis. _Cell._ 168, 670–91 (2017). Article  CAS  PubMed  PubMed Central 

Google Scholar  * Duprez, F. et al. Distant metastases in head and neck cancer. _Head & Neck_ 39, 1733–43 (2017). Article  Google Scholar  * Nguyen, B. et al. Genomic characterization of

metastatic patterns from prospective clinical sequencing of 25,000 patients. _Cell._ 185, 563–75 e11 (2022). Article  CAS  PubMed  PubMed Central  Google Scholar  * Farah, C. S. Molecular

landscape of head and neck cancer and implications for therapy. _Ann. Transl. Med._ 9, 915 (2021). Article  CAS  PubMed  PubMed Central  Google Scholar  * Puram, V. S. et al. Single-cell

transcriptomic analysis of primary and metastatic tumor ecosystems in head and neck. _Cancer Cell._ 171, 1611–24.e24 (2017). CAS  Google Scholar  * Varga, J. & Greten, F. R. Cell

plasticity in epithelial homeostasis and tumorigenesis. _Nat. Cell Biol._ 19, 1133–41. (2017). Article  CAS  PubMed  Google Scholar  * Lamouille, S., Xu, J. & Derynck, R. Molecular

mechanisms of epithelial–mesenchymal transition. _Nat. Rev. Mol. Cell Biol._ 15, 178 (2014). Article  CAS  PubMed  PubMed Central  Google Scholar  * Darido, C. et al. Stage-dependent

therapeutic efficacy in PI3K/mTOR-driven squamous cell carcinoma of the skin. _Cell Death Differ._ 25, 1146–59. (2018). Article  CAS  PubMed  Google Scholar  * Matus, D. Q. et al. Invasive

cell fate requires g1 cell-cycle arrest and histone deacetylase-mediated changes in gene expression developmental cell article invasive cell fate requires g1 cell-cycle arrest and histone

deacetylase-mediated changes in gene expression. _Developmental Cell_ 35, 162–74 (2015). Article  CAS  PubMed  PubMed Central  Google Scholar  * Lattmann, E., Deng, T. & Hajnal, A. To

divide or invade: a look behind the scenes of the proliferation-invasion interplay in the caenorhabditis elegans anchor cell. _Front. Cell Dev. Biol._ 8, 1727 (2021). Article  Google Scholar

  * Kohrman, A. Q. & Matus, D. Q. Divide or conquer: cell cycle regulation of invasive behavior. _Trends in Cell Biol._ 27, 12 (2017). Article  CAS  Google Scholar  * Tan, Bai, Saintigny

& Darido mTOR signalling in head and neck cancer: heads up. _Cells_ 8, 333 (2019). Article  CAS  PubMed  PubMed Central  Google Scholar  * Lawrence, M. S. et al. Comprehensive genomic

characterization of head and neck squamous cell carcinomas. _Nature_ 517, 576–82 (2015). Article  ADS  CAS  Google Scholar  * Stransky, N. et al. The mutational landscape of head and neck

squamous cell carcinoma. _Science_ 333, 1157–60 (2011). Article  ADS  CAS  PubMed  PubMed Central  Google Scholar  * Hong, D. S. et al. A multicenter phase I trial of PX-866, an oral

irreversible phosphatidylinositol 3-kinase inhibitor, in patients with advanced solid tumors. _Clin. Cancer Res._ 18, 4173–82 (2012). Article  CAS  PubMed  Google Scholar  * Kolk, A. et al.

Expression of Y-box-binding protein YB-1 allows stratification into long- and short-term survivors of head and neck cancer patients. _Br. J. Cancer_ 105, 1864 (2011). Article  CAS  PubMed 

PubMed Central  Google Scholar  * Alkrekchi, A., Wang, W., Rana, P. S., Markovic, V. & Sossey-Alaoui, K. A comprehensive review of the functions of YB-1 in cancer stemness, metastasis

and drug resistance. _Cellular Signalling_ 85, 110073 (2021). Article  CAS  PubMed Central  Google Scholar  * Sutherland, B. W. et al. Akt phosphorylates the Y-box binding protein 1 at

Ser102 located in the cold shock domain and affects the anchorage-independent growth of breast cancer cells. _Oncogene._ 24, 4281–92 (2005). Article  CAS  PubMed  Google Scholar  * Yu, K. et

al. Comprehensive transcriptomic analysis of cell lines as models of primary tumors across 22 tumor types. _Nat. Commun._ 10, 3574 (2019). Article  ADS  CAS  PubMed  PubMed Central  Google

Scholar  * Lui, V. W. Y. et al. Frequent mutation of the PI3K pathway in head and neck cancer defines predictive biomarkers. _Cancer Discov._ 3, 761–9 (2013). Article  CAS  PubMed  PubMed

Central  Google Scholar  * Kinross, K. M. et al. Ubiquitous expression of the Pik3caH1047R mutation promotes hypoglycemia, hypoinsulinemia, and organomegaly. _FASEB J._ 29, 1426–34 (2015).

Article  CAS  PubMed  Google Scholar  * Darido, C. et al. Targeting of the tumor suppressor GRHL3 by a miR-21-dependent proto-oncogenic network results in PTEN loss and tumorigenesis.

_Cancer Cell_ 20, 635–48 (2011). Article  CAS  PubMed  Google Scholar  * Goldie, S. J. et al. Loss of GRHL3 leads to TARC/CCL17-mediated keratinocyte proliferation in the epidermis. _Cell

Death Dis._ 9, 1072 (2018). Article  PubMed  PubMed Central  Google Scholar  * Georgy, S. R. et al. Identification of a novel proto-oncogenic network in head and neck squamous cell

carcinoma. _J. Natl Cancer Inst._ 107, djv152 (2015). Article  PubMed  PubMed Central  Google Scholar  * Bai, Y. et al. The balance between differentiation and terminal differentiation

maintains oral epithelial homeostasis. _Cancers (Basel)_ 13, 5123 (2021). Article  CAS  PubMed  Google Scholar  * Chung, C. H. et al. Molecular classification of head and neck squamous cell

carcinomas using patterns of gene expression. _Cancer Cell_ 5, 489–500 (2004). Article  CAS  PubMed  Google Scholar  * Walter, V. et al. Molecular subtypes in head and neck cancer exhibit

distinct patterns of chromosomal gain and loss of canonical cancer genes. _PLoS One_ 8, e56823 (2013). Article  ADS  CAS  PubMed  PubMed Central  Google Scholar  * Castel, P., Toska, E.,

Engelman, J. A. & Scaltriti, M. The present and future of PI3K inhibitors for cancer therapy. _Nat. Cancer_ 2, 587–97 (2021). Article  CAS  PubMed  PubMed Central  Google Scholar  *

Vanhaesebroeck, B., Perry, M. W. D., Brown, J. R., André, F. & Okkenhaug, K. PI3K inhibitors are finally coming of age. _Nat. Rev. Drug Discov._ 20, 741–69 (2021). Article  CAS  PubMed 

PubMed Central  Google Scholar  * Jimeno, A. et al. A randomized, phase II trial of cetuximab with or without PX-866, an irreversible oral phosphatidylinositol 3-kinase inhibitor, in

patients with relapsed or metastatic head and neck squamous cell cancer. _Ann. Oncol._ 26, 556–61 (2015). Article  CAS  PubMed  Google Scholar  * Bai, Y., Zhao, Z., Boath, J., van Denderen,

B. J. & Darido, C. The functional GRHL3-filaggrin axis maintains a tumor differentiation potential and influences drug sensitivity. _Mol. Ther._ 29, 2571–82 (2021). Article  CAS  PubMed

  PubMed Central  Google Scholar  * Chen, J. P., Chang, J. Y. & Hong, R. L. Different treatment response in several head and neck squamous cell carcinoma cell lines possibly reflecting

underlying molecular signatures. _Ann. Oncol._ 29, ix101–ix2 (2018). Article  Google Scholar  * Lima, D. G. et al. Combined therapy with cisplatin and 5-AZA-2CdR modifies methylation and

expression of DNA repair genes in oral squamous cell carcinoma. _Int. J. Clin. Exp. Pathol._ 15, 131–44 (2022). ADS  CAS  PubMed  PubMed Central  Google Scholar  * Kjaer, I. et al. Cetuximab

resistance in squamous carcinomas of the upper aerodigestive tract is driven by receptor tyrosine kinase plasticity: potential for mab mixtures. _Mol. Cancer Ther._ 15, 1614–26 (2016).

Article  CAS  PubMed  Google Scholar  * Jung, K., Kang, H. & Mehra, R. Targeting phosphoinositide 3-kinase (PI3K) in head and neck squamous cell carcinoma (HNSCC). _Cancers Head Neck_ 3,

3 (2018). Article  PubMed  PubMed Central  Google Scholar  * Schmitz, S., Bindea, G., Albu, R. I., Mlecnik, B. & Machiels, J. P. Cetuximab promotes epithelial to mesenchymal transition

and cancer associated fibroblasts in patients with head and neck cancer. _Oncotarget._ 6, 34288–99 (2015). Article  PubMed  PubMed Central  Google Scholar  * Thiery, J. P., Acloque, H.,

Huang, R. Y. J. & Nieto, M. A. Epithelial-mesenchymal transitions in development and disease. _Cell_ 139, 871–90 (2009). Article  CAS  PubMed  Google Scholar  * Brabletz, T., Kalluri,

R., Nieto, M. A. & Weinberg, R. A. EMT in cancer. _Nat. Rev. Cancer_ 18, 128–34 (2018). Article  CAS  PubMed  Google Scholar  * Zhang, J. et al. Structural basis of DNA binding to human

YB-1 cold shock domain regulated by phosphorylation. _Nucleic Acids Res._ 48, 9361–71 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Lyabin, D. N., Eliseeva, I. A. &

Ovchinnikov, L. P. YB-1 protein: functions and regulation. _Wiley Interdiscip. Rev. RNA_ 5, 95–110 (2014). Article  CAS  PubMed  Google Scholar  * Sogorina, E. M. et al. YB-1 phosphorylation

at serine 209 inhibits its nuclear translocation. _Int. J. Mol. Sci._ 23, 428 (2021). Article  PubMed  PubMed Central  Google Scholar  * Ishay-Ronen, D. et al. Gain Fat—Lose Metastasis:

converting invasive breast cancer cells into adipocytes inhibits cancer metastasis. _Cancer Cell_ 35, 17–32.e6 (2019). Article  CAS  PubMed  Google Scholar  * Jayavelu, A. K. et al. Splicing

factor YBX1 mediates persistence of JAK2-mutated neoplasms. _Nature_ 588, 157–63 (2020). Article  ADS  PubMed  Google Scholar  * Wang, Y., Arribas-Layton, M., Chen, Y., Lykke-Andersen, J.

& Sen, G. L. DDX6 orchestrates mammalian progenitor function through the mrna degradation and translation pathways. _Mol. Cell_ 60, 118–30 (2015). Article  CAS  PubMed  PubMed Central 

Google Scholar  * Yang, J. et al. Guidelines and definitions for research on epithelial–mesenchymal transition. _Nat. Rev. Mol. Cell Biol._ 21, 341–52 (2020). Article  PubMed  PubMed Central

  Google Scholar  * Joyce, J. A. & Pollard, J. W. Microenvironmental regulation of metastasis. _Nat. Rev. Cancer_ 9, 239–52 (2009). Article  CAS  PubMed  Google Scholar  * Gocheva, V. et

al. IL-4 induces cathepsin protease activity in tumor-associated macrophages to promote cancer growth and invasion. _Genes Dev._ 24, 241–55 (2010). Article  CAS  PubMed  PubMed Central 

Google Scholar  * Azoury, S. C., Gilmore, R. C. & Shukla, V. Molecularly targeted agents and immunotherapy for the treatment of head and neck squamous cell cancer (HNSCC). _Discov. Med._

21, 507–16 (2016). PubMed  Google Scholar  Download references ACKNOWLEDGEMENTS We thank members of the Darido laboratory including Imran Khan, Adelle Marrazzo, Sarah Fourari and Jade Tran

for technical and general assistance with experiments and figures. The authors would like to acknowledge the assistance of David Goode from the Computational Cancer Biology Programme at the

Peter MacCallum Cancer Centre. We also acknowledge the assistance from the Peter MacCallum Cancer Centre Core Facilities including the Bioinformatics Core Facility, the Victorian Centre for

Functional Genomics, the Molecular Genomics Centre, the Centre for Advanced Histology & Microscopy and the Animal Facility. We thank Oliver Bissinger for the intellectual contribution

and support of the co-author CG. This research was supported by a grant from the Australian National Health and Medical Research Council (NHMRC, APP1106697) to SJM and CD, and a Victorian

Cancer Agency mid-career Fellowship (MCRF16017) to CD. AUTHOR INFORMATION AUTHORS AND AFFILIATIONS * Peter MacCallum Cancer Centre, 305 Grattan St, Melbourne, VIC, 3000, Australia Yuchen

Bai, Ginevra Chincarini, Clare Slaney, Jarryd Boath, Luc Furic, Wayne A. Phillips, Steven A. Stacker & Charbel Darido * Department of Oral and Maxillofacial Surgery, Technische

Universität München, Fakultät für Medizin, Klinikum rechts der Isar, Ismaningerstraße 22, 81675, Munich, Germany Carolin Gotz * Department of Oral and Maxillofacial Surgery, Medizinische

Universität Innsbruck, Anichstraße 35, 6020, Innsbruck, Austria Carolin Gotz * Sun Yat-sen University Cancer Center, Yuexiu District, Guangzhou, Guangdong Province, China Zixuan Zhao * The

Sir Peter MacCallum Department of Oncology, The University of Melbourne, Parkville, VIC, 3010, Australia Clare Slaney, Luc Furic, Wayne A. Phillips, Steven A. Stacker & Charbel Darido *

Cancer Program, Biomedicine Discovery Institute and Department of Anatomy and Developmental Biology, Monash University, Clayton, VIC, 3800, Australia Luc Furic * Department of

Histopathology, Peter MacCallum Cancer Centre, Melbourne, VIC, 3000, Australia Christopher Angel * Department of Medicine, Central Clinical School, Monash University, 99 Commercial Road,

Melbourne, VIC, 3004, Australia Stephen M. Jane * Australian Centre for Oral Oncology Research & Education; Fiona Stanley Hospital; Hollywood Private Hospital; Australian Clinical Labs,

CQ University, Perth, WA, 6009, Australia Camile S. Farah Authors * Yuchen Bai View author publications You can also search for this author inPubMed Google Scholar * Carolin Gotz View author

publications You can also search for this author inPubMed Google Scholar * Ginevra Chincarini View author publications You can also search for this author inPubMed Google Scholar * Zixuan

Zhao View author publications You can also search for this author inPubMed Google Scholar * Clare Slaney View author publications You can also search for this author inPubMed Google Scholar

* Jarryd Boath View author publications You can also search for this author inPubMed Google Scholar * Luc Furic View author publications You can also search for this author inPubMed Google

Scholar * Christopher Angel View author publications You can also search for this author inPubMed Google Scholar * Stephen M. Jane View author publications You can also search for this

author inPubMed Google Scholar * Wayne A. Phillips View author publications You can also search for this author inPubMed Google Scholar * Steven A. Stacker View author publications You can

also search for this author inPubMed Google Scholar * Camile S. Farah View author publications You can also search for this author inPubMed Google Scholar * Charbel Darido View author

publications You can also search for this author inPubMed Google Scholar CONTRIBUTIONS Conceptualisation: Y.B., C.G., C.D.; Data curation: Y.B., L.F., C.D.; Formal analysis: Y.B., C.G.,

G.C., Z.Z., C.S., C.D.; Investigation: Y.B., C.G., G.C., Z.Z., J.B., C.A.; Methodology: Y.B., L.F., S.M.J., W.A.P., S.S., C.S.F., C.D.; Resources: C.S., L.F., S.M.J., W.A.P., S.S., C.S.F.,

C.D.; Supervision: C.S., C.S.F., C.D.; Writing–original draft: Y.B., C.G., C.D.; Writing–review and editing: Y.B., C.G., G.C., Z.Z., C.S., J.B., L.F., C.A., S.M.J., W.A.P., S.S., C.S.F.,

C.D.; Funding acquisition: S.M.J. and C.D. The funders had no role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the

decision to submit the manuscript for publication. CORRESPONDING AUTHOR Correspondence to Charbel Darido. ETHICS DECLARATIONS COMPETING INTERESTS The authors declare no competing interests.

PEER REVIEW PEER REVIEW INFORMATION _Nature Communications_ thanks Jochen Hess and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer

report are available. ADDITIONAL INFORMATION PUBLISHER’S NOTE Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

SUPPLEMENTARY INFORMATION SUPPLEMENTARY INFORMATION PEER REVIEW FILE REPORTING SUMMARY SOURCE DATA SOURCE DATA RIGHTS AND PERMISSIONS OPEN ACCESS This article is licensed under a Creative

Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the

original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in

the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended

use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit

http://creativecommons.org/licenses/by/4.0/. Reprints and permissions ABOUT THIS ARTICLE CITE THIS ARTICLE Bai, Y., Gotz, C., Chincarini, G. _et al._ YBX1 integration of oncogenic PI3K/mTOR

signalling regulates the fitness of malignant epithelial cells. _Nat Commun_ 14, 1591 (2023). https://doi.org/10.1038/s41467-023-37161-0 Download citation * Received: 25 May 2022 * Accepted:

06 March 2023 * Published: 22 March 2023 * DOI: https://doi.org/10.1038/s41467-023-37161-0 SHARE THIS ARTICLE Anyone you share the following link with will be able to read this content: Get

shareable link Sorry, a shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative