Abstract

   Oral squamous cell carcinoma (OSCC) is the most prevalent cancer of the
   mouth, characterised by rapid progression and poor prognosis. Hence, an
   urgent need exists for the development of predictive targets for early
   diagnosis, prognosis determination, and clinical therapy. Dysregulation
   of lymphoid enhancer-binding factor 1 (LEF1), an important
   transcription factor involved in the Wnt-β-catenin pathway, contributes
   to the poor prognosis of OSCC. Herein, we aimed to explore the
   correlation between LEF1 and histone lysine demethylase 4 A (KDM4A).
   Results show that the KDM4A complex is recruited by LEF1 and
   specifically binds the LATS2 promoter region, thereby inhibiting its
   expression, and consequently promoting cell proliferation and impeding
   apoptosis in OSCC. We also established NOD/SCID mouse xenograft models
   using CAL-27 cells to conduct an in vivo analysis of the roles of LEF1
   and KDM4A in tumour growth, and our findings show that cells stably
   suppressing LEF1 or KDM4A have markedly decreased tumour-initiating
   capacity. Overall, the results of this study demonstrate that LEF1
   plays a pivotal role in OSCC development and has potential to serve as
   a target for early diagnosis and treatment of OSCC.

   Subject terms: Oral cancer, Oncogenes, Epigenetics

Introduction

   Oral squamous cell carcinoma (OSCC) is the most prevalent head and neck
   squamous cell carcinoma (HNSCC), with an annual incidence of
   approximately 4 per 100,000 [[38]1, [39]2]. The risk factors of OSCC
   are complex, and the disease is characterised by a lack of symptoms in
   the early stage, strong invasiveness, rapid progression, and poor
   prognosis [[40]3]. Therefore, further elucidation of the molecular
   mechanisms underlying OSCC carcinogenesis and the identification of
   predictive biomarkers and therapeutic targets are crucial.

   The highly conserved Wnt/β-catenin signalling pathway, which plays a
   vital role in embryonic development, adult tissue homeostasis, and
   cancer progression [[41]4], is reportedly activated in OSCC [[42]5].
   Lymphoid enhancer-binding factor 1 (LEF1), an important transcription
   factor involved in the Wnt-β-catenin pathway, activates the
   transcription of many downstream target genes (such as cyclin D1,
   prostaglandin-endoperoxide synthase 2, interleukin-6, and large tumour
   suppressor kinase 2 [LATS2]) by interacting with β-catenin in the
   nucleus, and independently modulates the activation or repression of
   transcription [[43]6, [44]7]. Dysregulation of LEF1 contributes to the
   progression of multiple cancers, including breast, colon, prostate, and
   oesophageal squamous cell carcinoma [[45]8–[46]11]. Meanwhile, asporin
   (ASPN) binds directly to LEF1, promotes LEF1-mediated gene
   transcription independent of β-catenin, and inhibits cell apoptosis in
   gastric cancer [[47]7]. Moreover, LEF1 negatively regulates
   cylindromatosis (CYLD) by binding to the CYLD promoter site and
   inhibiting its transcription [[48]12]. Selenite treatment inhibits LEF1
   recruitment to the CYLD promoter and reduces the average tumour weight
   in colorectal cancer [[49]13]. In OSCC, LEF1 is overexpressed in tumour
   tissues compared to non-tumorous oral mucosa [[50]14]. However, its
   role in regulating cell proliferation and apoptosis during OSCC
   progression remains to be explored.

   Histone modifications play crucial roles in OSCC pathogenesis and
   progression through gene silencing or activation [[51]15, [52]16]. In
   particular, histone lysine demethylases (KDMs) participate in
   transcriptional regulation by affecting the methylation of H3K4, H3K9,
   H3K27, and H3K36, which may lead to changes in the expression of
   tumour-suppressor genes and proto-oncogenes [[53]17–[54]20]. Indeed,
   KDMs have been identified as potential prognostic markers and
   therapeutic targets in OSCC [[55]21]. Lysine-specific demethylase 4A
   (KDM4A), also known as Jumonji domain-containing protein 2A (JMJD2A),
   is a member of the KDM family that regulates gene expression via its
   function as a demethylase and recruits chromatin factors [[56]22,
   [57]23]. Dysregulation of KDM4A expression has been linked to various
   human cancers, including breast, colon, gastric, prostate, and
   non-small cell lung cancers [[58]24–[59]26]. KDM4A acts as a cofactor
   of the androgen receptor (AR) and stimulates AR-dependent transcription
   of proliferative genes via its demethylase activity on H3K9me3 in
   prostate cancer [[60]27]. As a recruiter of chromatin factors, it
   interacts with nuclear receptor co-repressor (N-CoR) to repress
   achaete-scute family bHLH transcription factor 2 (ASCL2) expression
   [[61]28, [62]29]. The level of KDM4A is notably higher in oral cancer
   and primary squamous cell carcinoma tissues than in adjacent epithelium
   [[63]19, [64]20]; however, the molecular mechanisms responsible for its
   overexpression remain unknown.

   LEF1 contributes to the recruitment of the histone modification enzyme
   protein arginine methyltransferase 6 (PRMT6) to the high-mobility group
   (HMG) domain, thus influencing cellular differentiation and
   proliferation [[65]30]. Herein, we aimed to elucidate the molecular
   mechanisms related to LEF1 and KDM4A in OSCC and to identify potential
   targets for the early diagnosis and treatment of OSCC.

Results

Upregulation of LEF1 correlates with OSCC progression

   Although LEF1 overexpression has been reported in OSCC [[66]14], its
   role in OSCC progression remains unknown. Considering that LEF1 is
   associated with histone modification [[67]30, [68]31] and that the KDM4
   subfamily of histone-modifying enzymes is dysregulated in OSCC
   [[69]18], we analysed the expression of LEF1 and KDM4s (KDM4A–D) using
   The Cancer Genome Atlas (TCGA) Head and Neck Cancer dataset.
   RNA-sequencing (RNA-seq) analysis of 566 samples (44 normal and 522
   tumour) suggested that LEF1, KDM4A, and KDM4D were expressed at higher
   levels in tumour tissues than in normal tissues, whereas KDM4B was
   expressed at lower levels in tumour tissues (Fig. [70]1A). Moreover,
   the expression levels of KDM4A and LEF1 were highly correlated in HNSCC
   based on the Pearson correlation coefficient (Fig. [71]1B).

Fig. 1. Upregulation of LEF1 correlates with OSCC progression.

   [72]Fig. 1
   [73]Open in a new tab

   A Analysis of LEF1 and KDM4 family expression in HNSCC using TCGA
   database. Control: n = 44; cancer: n = 522. B Relative levels of the
   KDM4 family plotted against LEF1 expression in HNSCC using TCGA
   database. C Analysis of public datasets (TCGA) for the expression of
   LEF1 and KDM4A in the normal epithelium as well as HNSCCs with
   histological grades I, II, III, and IV (44 normal, 63 grade I, 305
   grade II, 125 grade III, and 29 grade IV). D Tissue microarrays showed
   differential expression of LEF1 and KDM4A in the normal epithelium,
   histological grade I OSCC, and histological grade II or III OSCC (9
   normal, 27 grade I, 30 grade II or III). Positively stained cells
   (percentage) in the grouped samples were analysed. Scale bar: 100 μm. E
   Immunohistochemical staining of LEF1 and KDM4A in normal epithelial
   tissues and tumours (n = 4). The percentage of positively stained cells
   (in percentages) in samples were analysed. Scale bar: 100 μm or 25 μm.
   Data represent the mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.

   We then analysed the expression of LEF1 and KDM4A and their correlation
   with histological grades of OSCC (Fig. [74]1C). Comparing the
   expression in the normal epithelium and tumour tissues from
   histological grade I, II, III, or IV revealed that higher histological
   grades corresponded to higher LEF1 and KDM4A mRNA expression. We
   further validated the abnormal expression of LEF1 and KDM4A in a
   clinicopathologically relevant context. At the protein level, we
   compared the expression of LEF1 and KDM4A by immunohistochemical
   staining of tissue microarrays comprising 57 human pathological OSCC
   tissues (27 grade I and 30 grade II or III; Fig. [75]1D) and 9 normal
   epithelium samples, as well as four additional human OSCC samples and
   adjacent normal epithelium samples (Fig. [76]1E). The results further
   confirmed that LEF1 and KDM4A were notably upregulated in tumours and
   that the differential expression of LEF1 and KDM4A was related to OSCC
   histological grade.

LEF1 physically interacts with KDM4A

   LEF1 may function by recruiting epigenetic cofactors to target specific
   genes [[77]30]. To gain a better understanding of the mechanistic
   relationship between LEF1 and KDM4A, we first analysed LEF1 and KDM4A
   localisation in two OSCC cell lines, CAL-27 and SCC-9, using
   immunofluorescence and found that LEF1 colocalised with KDM4A in the
   nucleus (Fig. [78]2A). Since KDM4A relies on N-CoR for transcriptional
   repression [[79]28], we also evaluated whether LEF1 interacted with the
   KDM4A complex (KDM4A and N-CoR) and found that LEF1 colocalised with
   N-CoR in the nucleus (Fig. [80]2A). To reveal their physical
   association, total proteins from CAL-27 and SCC-9 cells were extracted
   and co-immunoprecipitation (co-IP) assays were performed.
   Immunoprecipitants with antibodies against LEF1 proteins were subjected
   to western blotting (WB) with antibodies against KDM4A, showing that
   LEF1 physically interacted with KDM4A (Fig. [81]2B). Similarly,
   immunoprecipitants with antibodies against KDM4A or N-CoR followed by
   WB with antibodies against LEF1 confirmed this interaction (Fig.
   [82]2B). The interaction between LEF1 and KDM4A was also demonstrated
   in HEK293T cells overexpressing Flag-KDM4A and GFP-LEF1 (Fig. [83]2C).
   We then assessed the LEF1 and KDM4A combination in vitro; reciprocal
   Glutathione-S-transferase (GST) pull-down was performed using GST-fused
   LEF1 or KDM4A and in vitro-transcribed/translated KDM4A or LEF1,
   revealing that the LEF1 and KDM4A molecules interacted directly (Fig.
   [84]2D). Collectively, these results suggest an interaction between
   LEF1 and the KDM4A complex.

Fig. 2. Physically interaction of LEF1 with the KDM4A complex.

   [85]Fig. 2
   [86]Open in a new tab

   A Normally cultured CAL-27 and SCC-9 cells were fixed and analysed by
   indirect immunofluorescence using antibodies specific to human LEF1,
   KDM4A, and N-CoR. Nuclear DNA was stained by
   4’,6-diamidino-2-phenylindole (DAPI). Scale bar: 10 μm. B Association
   of LEF1 with KDM4A complex in CAL-27 and SCC-9. Whole-cell lysates were
   prepared, co-IP was performed using antibodies against LEF1, KDM4A, or
   N-CoR, and captured samples were immunoblotted with antibodies against
   the indicated proteins. IgG served as the negative control. C
   Association of LEF1 with KDM4A in HEK293T. Co-IP of exogenous
   EGFP-tagged LEF1 and FLAG-tagged KDM4A in HEK293T was detected. IgG
   served as the negative control. D GST pull-down experiments performed
   using bacterially expressed GST-fusion proteins and in
   vitro-transcribed/translated proteins. E Identification of LEF1 domains
   required for interaction with KDM4A. The Flag-tagged LEF1 or
   Flag-tagged truncation constructs of LEF1 were overexpressed in HEK293T
   cells with EGFP-tagged KDM4A. Whole-cell lysates were prepared, Flag
   and EGFP expression were detected, and co-IP was performed using
   antibodies against Flag; then, captured samples were immunoblotted with
   antibodies against EGFP. Cells with overexpression of pCMV6 plasmid and
   EGFP-tagged LEF1 served as the negative control. F Identification of
   KDM4A domains required for interaction with LEF1. The Flag-tagged KDM4A
   or Flag-tagged truncation constructs of KDM4A were overexpressed in
   HEK293T cells with EGFP-tagged LEF1. Co-IP was performed as indicated
   in E.

   Next, we sought to gain insights into the essential domains required
   for the interaction between LEF1 and KDM4A. LEF1 contains a
   β-catenin-binding site located at amino acid residues 1–62 and a
   DNA-binding HMG domain at the C-terminus (residues 299–367; Fig.
   [87]2E) [[88]6]. KDM4A contains a higher number of structural domains,
   including the JmjN domain (residues 14–56), the catalytic JmjC domain
   (residues 142–308), two plant homeodomain-type zinc fingers (residues
   709–767 and 828–885), and two Tudor domains near the C-terminus
   (residues 897–954 and 955–1011; Fig. [89]2F) [[90]32]. Co-IP was
   further performed with various plasmids in HEK293T cells, and proteins
   were subjected to WB with antibodies against EGFP (enhanced green
   fluorescent protein). Depletion of the LEF1 β-catenin-binding site did
   not interrupt KDM4A binding, and expression of the HMG domain resulted
   in KDM4A interaction (Fig. [91]2E), indicating the function of this
   domain goes beyond DNA-binding [[92]6]. Similarly, the Tudor domain of
   KDM4A not only serves as a reader module that recognises and binds to
   methylated amino acid residues [[93]33] but is also responsible for the
   interaction with LEF1 (Fig. [94]2F). In summary, we verified the
   interaction between LEF1 and the KDM4A complex and identified its
   interaction with the HMG domain of LEF1 and the Tudor domain of KDM4A
   (Fig. [95]2F). Therefore, we believe that LEF1 can guide the
   recruitment of the KDM4A complex to the chromatin in OSCC.

Knockdown of LEF1 or KDM4A suppresses proliferation while promoting CAL-27
and SCC-9 cell apoptosis in vitro

   Abnormal expression of LEF1 has been implicated in the proliferation
   and apoptosis of several cancer cells, and LEF1-mediated gene
   transcription may go beyond the Wnt-β-catenin pathway [[96]7, [97]34].
   Therefore, we investigated the role of LEF1 and KDM4A in the
   proliferation and apoptosis of oral squamous cells. First,
   5-ethynyl-2′-deoxyuridine (EdU) staining revealed that LEF1 or KDM4A
   knockdown was associated with a notable decrease in the proportion of
   EdU-labelled cells (Fig. [98]3A), indicating impaired DNA replication.
   The Cell Counting Kit-8 (CCK8) results further revealed that cell
   viability decreased with LEF1 or KDM4A knockdown (Fig. [99]3B). These
   findings demonstrated that LEF1 and KDM4A promoted the proliferation of
   oral squamous cells in vitro.

Fig. 3. Knockdown of LEF1 or KDM4A suppresses proliferation but promotes
apoptosis in CAL-27 and SCC-9 cells in vitro.

   [100]Fig. 3
   [101]Open in a new tab

   A Twenty-four hour after transfection of LEF1 or KDM4A siRNA, the
   CAL-27 or SCC-9 cells were incubated with EdU for 3 h. A fluorescence
   microscope was used to detect EdU. Scale bar: 100 μm. B Cell viability
   of CAL-27 and SCC-9 48 h after transfection, as assessed by CCK-8. C
   Expression of apoptosis markers measured using RT-PCR in
   LEF1/KDM4A-depleted CAL-27 and SCC-9 cells. Results are expressed as
   fold-change relative to control, and GAPDH was used as negative
   control. D Expression of apoptosis markers was measured using western
   blotting (WB) in LEF1/KDM4A-depleted CAL-27 and SCC-9 cells. GAPDH
   served as a loading control. E Apoptosis rate of LEF1/KDM4A-depleted
   CAL-27 cells assessed by flow cytometry. F Cell cycle analysis of
   LEF1/KDM4A-depleted CAL-27 cells assessed by flow cytometry. Data are
   shown as means ± SD from three independent experiments. *p < 0.05,
   **p < 0.01, ***p < 0.001.

   Next, we measured the expression of apoptosis markers in CAL-27 and
   SCC-9 cells after LEF1 or KDM4A knockdown. KDM4A or LEF1 knockdown in
   CAL-27 and SCC-9 cells resulted in increased expression of apoptosis
   markers, including BAX, FAS, caspase 8, and PARP1, at the mRNA (Fig.
   [102]3C) and protein (Fig. [103]3D) levels. Additionally, flow
   cytometry showed that depletion of LEF1 or KDM4A induced cell cycle
   arrest in the G1 phase in CAL-27 cells (Fig. [104]3E); this depletion
   contributed to apoptosis (Fig. [105]3F) in both cell lines. Hence, LEF1
   and KDM4A promote proliferation and inhibit apoptosis in vitro, which
   is potentially associated with the regulation of LEF1/KDM4A target
   genes.

Transcriptome identification of the LEF1/KDM4A complex transcription targets

   To confirm that LEF1 mediates KDM4A recruitment to target genes, we
   identified genes coregulated by LEF1/KDM4A. Whole-transcriptome
   clustering analysis revealed that LEF1 knockdown caused the expression
   of 2468 genes to exhibit a >1.5 fold-change, while KDM4A knockdown
   induced differential expression of 2478 genes (Q value < 0.05). Among
   these genes, 401 were co-upregulated, and 281 were co-downregulated in
   the siLEF1 and siKDM4A groups (Fig. [106]4A–[107]B). The genes
   coregulated and regulated by LEF1 or KDM4A are presented in
   Differentially Expressed Genes Identified by RNA-sequencing analysis.

Fig. 4. Transcriptome analysis of LEF1 and KDM4A.

   [108]Fig. 4
   [109]Open in a new tab

   A Heatmap of coregulated gene expression after LEF1/KDM4A knockdown.
   Representative genes, GO terms, and KEGG pathways of upregulated or
   downregulated genes are also shown. LEF1 or KDM4A was knocked down in
   CAL-27 cells by using siRNAs. Three independent samples were separately
   subjected to RNA-seq analysis. B Box plot showing the fragments per
   kilobase of transcript per million mapped reads (FPKM) of gene
   expression distribution of upregulated or downregulated genes. For each
   sample, the target groups were separately compared to three control
   samples. C Pathway analysis of LEF1/KDM4A-regulated target genes
   arranged into functional groups. D Verification of RNA-seq results
   through qPCR analysis of the indicated genes in CAL-27 cells. Results
   are represented as fold-change compared to the control, with GAPDH used
   as the internal reference. E Heatmap of differentially expressed genes
   from the Hippo signalling pathway. F Verification of important genes in
   the Hippo signalling pathway through qPCR analysis in CAL-27 cells.
   Results are represented as fold-change compared to the control, with
   GAPDH used as the internal reference. Data represent mean ± SD of three
   independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001.

   Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment
   analysis was performed to further explore the signalling cascades
   downstream of LEF1 and KDM4A (Fig. [110]4C). Although knockdown of LEF1
   or KDM4A in CAL-27 cells led to the altered expression of several
   crucial genes at the transcriptional level (Fig. [111]4D), we focused
   on the Hippo signalling pathway, which is associated with cell
   proliferation, cycle arrest, and apoptosis [[112]35]. To explore the
   importance of this cascade in LEF1/KDM4A regulation, we generated a
   heatmap showing the involvement of the Hippo signalling pathway in
   LEF1/KDM4A depletion by analysing ten differentially expressed genes,
   which were then validated by real-time reverse transcription-polymerase
   chain reaction (RT-PCR; Fig. [113]4E–F). Based on the results, we
   propose that LEF1 recruits the KDM4A complex to alter the expression of
   a series of genes through transcriptional activation or repression.

The LEF1/KDM4A complex regulates OSCC carcinogenesis by suppressing LATS2
expression

   To explore whether the selected genes are transcriptional targets of
   LEF1 and KDM4A, chromatin IP (ChIP) experiments were performed in
   CAL-27 cells. Different levels of LEF1 and KDM4A enrichment were
   observed in the promoter regions of the selected genes implicated in
   tumour suppression or promotion (Fig. [114]5A–[115]B). Among these
   genes from the Hippo signalling pathway, the DNA segments of the LATS2
   promoter region were highly enriched; LATS2 encodes a vital
   serine/threonine protein kinase belonging to the large tumour
   suppressor family that influences mitosis initiation [[116]36]. Further
   results showed that the significant enrichments of LEF1 were mapped to
   three regions of the LATS2 promoter at upstream ~1500 to ~800 promoter
   regions. Meanwhile, KDM4A was bound to the upstream ~1500 to ~1000
   promoter regions. Accordingly, the upstream ~1500 to ~1000 regions of
   the LATS2 promoter might represent the primary region where the
   LEF1/KDM4A complex negatively regulates LATS2 expression (Fig.
   [117]5C–[118]D). qChIP assays showed that suppression of LEF1 or KDM4A
   expression resulted in a significant reduction in the recruitment of
   LEF1 and KDM4A to the LATS2 promoter. (Fig. [119]5E). Since KDM4A
   mediates the demethylation of H3K9 and H3K36, and is more efficient in
   demethylating tri- versus dimethylated H3K9/H3K36 [[120]32], we
   analysed the levels of H3K36me3 and H3K9me3 in the LATS2 promoter of
   KDM4A-depleted CAL-27 cells. qChIP assays showed that suppression of
   KDM4A expression resulted in a significant increase in the enrichment
   of H3K36me3 on the LATS2 promoter (Fig. [121]5F).

Fig. 5. The LEF1/KDM4A complex regulates OSCC carcinogenesis by suppressing
LATS2 expression.

   [122]Fig. 5
   [123]Open in a new tab

   Verification of RNA-seq results through qChIP analysis of genes in A
   Fig. [124]4D and B Fig. [125]4F in CAL-27 cells. Results are expressed
   as fold-change relative to IgG control. GAPDH was used as an internal
   standard. C Primer pairs including #1 to #10 synthesised to cover the
   promoter region of LATS2. qChIP-based promoter-walk experiments were
   performed using CAL-27 cells; the enrichment of LEF1 or KDM4A was
   mapped to three or two regions of the LATS2 promoter, respectively. D
   ChIP analysis on CAL-27 cells with antibodies against LEF1 and KDM4A at
   the LATS2 promoter. E qChIP analysis of LEF1 and KDM4A recruitment to
   the LATS2 promoter in CAL-27 cells after transfection with control
   shRNA (shSCR) or shRNAs targeting LEF1 or KDM4A. IgG served as a
   negative control. F qChIP analysis of H3K9me3 and H3K36me3 enrichment
   on the LATS2 promoter in CAL-27 cells after transfection with control
   shRNA (shSCR) or shRNA targeting KDM4A. IgG served as a negative
   control. G Expression of LATS2 measured using western blotting in
   LEF1/KDM4A-depleted CAL-27 and SCC-9 cells. GAPDH served as a loading
   control. H CAL-27 cells treated with siRNAs were assessed by flow
   cytometry to detect cell cycle distribution, revealing that LATS2
   knockdown in CAL-27 cells causes altered cell cycle distribution. I
   Immunohistochemical staining of LATS2 in normal epithelial tissue and
   tumours (n = 4). Scale bar: 100 μm or 25 μm. J Analysis of public
   datasets (TCGA) for LATS2 expression in the normal epithelium and
   HNSCCs with histological grades I, II, III, and IV (44 normal, 63 grade
   I, 305 grade II, 125 grade III, and 29 grade IV). Data represent the
   mean ± SD of three independent experiments. *p < 0.05, **p < 0.01,
   ***p < 0.001.

   To better understand the relationship between LEF1, KDM4A, and LATS2,
   we assessed LATS2 expression following LEF1 or KDM4A knockdown by
   RT-PCR (Fig. [126]4D) and WB (Fig. [127]5G) in CAL-27 and SCC-9 cells.
   The expression of LATS2 was markedly increased. We then generated
   LATS2-knockdown CAL-27 cells and conducted a subsequent rescue
   experiment to determine whether LATS2 is downstream of LEF1/KDM4A.
   Following LEF1 or KDM4A knockdown, the cell cycle was stalled at the G1
   phase. However, in the presence of simultaneous LATS2 knockdown, the
   proportion of cells in S phase increased (Fig. [128]5H). Hence, LEF1
   might recruit the KDM4A complex to suppress the levels of LATS2,
   further confirming the role of LEF1 and KDM4A in cell proliferation and
   apoptosis in vitro.

   To extend our observations to a clinicopathological level,
   immunohistochemical staining was performed to assess LATS2 abundance in
   the tissues collected from four OSCC patients. LATS2 was expressed at
   low levels (Fig. [129]5I). Finally, to verify our observations, we
   analysed the expression of LATS2 and its correlation with OSCC
   histological grade in patients using the TCGA Head and Neck Cancer
   dataset. LATS2 expression decreased with an increase in histological
   grade (Fig. [130]5J). These results verified that LATS2 is negatively
   regulated by LEF1/KDM4A and the LEF1-KDM4A-LATS2 axis plays a vital
   role in OSCC.

LEF1 and KDM4A promote the growth of oral squamous tumour xenografts

   To investigate the role of LEF1 and KDM4A in oral squamous cell
   proliferation and apoptosis in vivo, we established mouse xenograft
   models of CAL-27 cells. The efficiency of the lentivirus-mediated
   knockdown in CAL-27 cells was verified using RT-PCR (Fig. [131]6A).
   Tumour growth was visualised by bioluminescence imaging four weeks
   after subcutaneous injection (Fig. [132]6B). Cells stably suppressing
   LEF1 or KDM4A had markedly decreased tumour-initiating capacity, with
   tumours from the control shRNA (shSCR) group notably larger than those
   from the other two groups. The tumour volume of each group was
   calculated and analysed (Fig. [133]6C), demonstrating that LEF1/KDM4A
   promoted the growth of established tumour xenografts. The expression of
   the indicated proteins in tumour specimens was analysed by WB,
   revealing higher LATS2 expression in the LEF1- or KDM4A-knockdown
   groups (Fig. [134]6D). These experiments verified that LEF1 recruits
   KDM4A to promote cell proliferation in vivo.

Fig. 6. LEF1 and KDM4A promote the growth of oral squamous tumour xenografts.

   [135]Fig. 6
   [136]Open in a new tab

   Knockdown efficiency of shRNAs targeting LEF1 or KDM4A. B CAL-27 cells
   infected with lentiviruses carrying the indicated shRNAs were
   inoculated subcutaneously into the 6-week-old male NOD/SCID mice
   (n = 4). Primary tumours were quantified through bioluminescence
   imaging at 4 weeks after initial implantation. Representative in vivo
   bioluminescence images are shown, and tumour specimens were examined
   using in vitro bioluminescence measurements. C Tumour specimens were
   examined by in vitro measurements. D WB was used to verify the
   efficiency of protein knockdown and the expression of LATS2 in tumours.
   Data represent the mean ± SD of three independent experiments.
   *p < 0.05, **p < 0.01, ***p < 0.001.

Discussion

   The transcription factor LEF1 belongs to the T-cell factor (TCF)/LEF
   family and is known for its role in cancer cell migration, invasion,
   proliferation, and viability [[137]37, [138]38]. However, the role of
   epigenetic modification in LEF1 regulation of cancer progression is
   poorly understood. Here, we demonstrated that LEF1 expression was
   positively correlated with that of the histone H3K9/K36 demethylase
   KDM4A in OSCC. The LEF1-KDM4A complex promotes cell proliferation and
   inhibits apoptosis by inhibiting LATS2 transcription. Decreased LATS2
   expression reverses the phenotypic alteration of OSCC cells induced by
   LEF1 or KDM4A depletion. Our proposed mechanism of action for the
   LEF1-KDM4A complex in regulating OSCC cell proliferation and apoptosis
   in OSCC is shown in Fig. [139]7.

Fig. 7. A proposed mechanism of action of the LEF1-KDM4A complex in
regulating OSCC cell proliferation and apoptosis.

   [140]Fig. 7
   [141]Open in a new tab

   Created with BioRender ([142]https://app.biorender.com).

   The Wnt/β-catenin signalling pathway plays a vital role in cancer
   development [[143]4]. As a key transcription factor of this pathway,
   LEF1 alters the expression of downstream genes involved in cell
   proliferation, organ development, and cellular immunity [[144]39].
   Indeed, LEF1 functions as an oncogenic molecule in various tumours
   [[145]9, [146]11–[147]14]; our results demonstrated that its expression
   was upregulated in OSCC and promoted cancer in vitro and in vivo. KDM4
   family members, serving as histone-modifying enzymes, play a role in
   transcriptional regulation by interacting with transcription factors.
   For example, TCF4, another key transcription factor in Wnt signalling
   cascades, interacts with KDM4C, promoting HP1g removal and
   transcriptional activation in the nucleus [[148]40]. Among the KDM4
   family, KDM4A is the best explored, and its abnormal upregulation is
   closely related to OSCC [[149]23]. Therefore, we investigated whether
   LEF1 cooperates with KDM4 family members to regulate OSCC. As LEF1 had
   the highest correlation coefficient with KDM4A, we focused on
   elucidating the relationship between this pair. LEF1 promotes the
   transcription of downstream target genes of the β-catenin in the
   Wnt/β-catenin signalling pathway [[150]41, [151]42], and also functions
   independently of β-catenin with intrinsic histone deacetylase activity
   or by interacting with histone-modifying enzymes [[152]30, [153]31].
   Here, we found that LEF1 interacts with the KDM4A/N-CoR transcriptional
   repression complex in a β-catenin-independent manner. These findings
   suggest that LEF1 is closely associated with histone modification and
   that its role in OSCC may depend on different interactions with
   specific molecules.

   KDM4A, a member of the JMJD2 family, is dysregulated in several cancer
   types and plays an important role in proliferation, invasion, and
   metastasis through complex mechanisms [[154]43–[155]45]. KDM4A
   primarily demethylates H3K9me2/me3, with a lower rate for H3K36me2/me3.
   The demethylation of these two distinct methylation markers by KDM4A
   reflects divergent functions. That is, while a transcriptional
   repression function is generally assumed for H3K9me3, H3K36me2/3 is
   associated with active transcription, the detailed mechanism for which
   remains unclear. Nevertheless, in single-celled organisms, PH
   domain-containing protein in retina 1 (PHR1) expression is reduced by
   KDM4A with a lower level of H3K36 methylation at the PHR1 promoter,
   suggesting that transcriptional repression is strongly influenced by
   histone demethylase activity [[156]46]. In mammals, enrichment of
   H3K36me3 represents a universal signature of gene activity, which is
   correlated with SET domain 2 (SETD2) and recruitment of other
   chromatin-associated proteins that mediate a range of cellular
   processes, including transcription elongation, heterochromatin
   formation, mRNA splicing, and DNA repair [[157]47]. Moreover,
   enrichment of H3K36me3 at the promoter regions is associated with
   transcriptional activation. SET and MYND domain-containing Protein 5
   (SMYD5), which is recruited to chromatin by RNA polymerase II, acts as
   a methyltransferase that catalyses H3K36me3, resulting in H3K36me3
   enrichment at promoters. In Smyd5 KO cells, the upregulated genes are
   associated with high H3K36me3 levels, while the downregulated genes are
   associated with low H3K36me3 levels [[158]48]. Furthermore, Jie et al.
   showed that the abnormally high expression of KDM4C in lung cancer
   inhibits CXCL10 transcription by reducing the enrichment of H3K36me3 at
   the CXCL10 promoter region [[159]49]. Transcriptional stimulation or
   repression of KDM4A may be associated with other molecules, including
   N-CoR, histone deacetylases, or transcription factors, such as p53
   tumour suppressor, in different biological processes [[160]28, [161]32,
   [162]50]. According to our study, KDM4A functions with N-CoR to repress
   transcription. Evidence from a multicentre cohort tissue microarray
   study showed that KDM4A could serve as a prognostic marker for OSCC
   [[163]20]. Further, targeting KDM4A induces DNA replication stress and
   activates antitumor immunity in HNSCCs, primarily comprising different
   OSCCs [[164]23]. Our results showed that LEF1 recruits KDM4A and N-CoR,
   functioning in a coordinated manner to transcriptionally inhibit the
   tumour-suppressor gene LATS2 and promote OSCC progression.

   LATS2, a serine/threonine kinase in the Hippo signalling pathway,
   participates in multiple cellular processes, including proliferation,
   apoptosis, morphogenesis, and differentiation [[165]36, [166]51,
   [167]52]. It plays a critical role in centrosome duplication, mitotic
   fidelity, and genomic stability and negatively regulates the G1/S
   transition by reducing cyclin E/CDK2 kinase activity. Here, we observed
   dysregulation of Hippo following LEF1 and KDM4A knockdown and noticed
   that LATS2 regulated multiple downstream targets associated with cell
   cycle and apoptosis [[168]53, [169]54]. After LEF1 or KDM4A knockdown,
   the cell cycle was stalled at the G1 phase [[170]18, [171]55], which
   may be due to increased LATS2 expression. Meanwhile, qChIP results
   showed that H3K36me3 levels increased at the LATS2 promoter with
   suppression of KDM4A expression. However, H3K9me3 abundance did not
   markedly change following KDM4A knockdown. Accordingly, we hypothesised
   that less deposition of H3K9me3 occurred at the LATS2 promoter in
   CAL-27 cells. In line with our hypothesis, previous research showed
   that loss of KDM4A/C mediates an increase in glial fibrillary acidic
   protein (GFAP) expression, which correlates with increased H3K36me3 and
   RNA polymerase II recruitment in transcribed regions, as well as the
   promoter, without impacting H3K9me3 in the GFAP promoter [[172]56]. We
   suspect that LEF1 recruits KDM4A/N-CoR and may recruit other epigenetic
   modification complexes to maintain the homeostasis of histone
   modifications at target promoters, thereby transcriptionally
   suppressing tumour-suppressor genes such as LATS2 and inhibiting the
   progression of OSCC. Additionally, altering the level of KDM4A-induced
   H3K36me3 demethylation may be a part of the process. However, the
   specific chromatin factors recruited by the LEF1/KDM4A/N-CoR complex
   and their functions, and the role of KDM4A as a demethylase in this
   process, require further investigation.

   Although we have proposed a mechanism of action for the LEF1-KDM4A
   complex in regulating cell proliferation and apoptosis in OSCC, it
   remains to be explored whether other molecules from the Wnt/β-catenin
   pathway interact with other KDM family members and if the LEF1-KDM4A
   complex asserts its effect through altering the methylation level of
   H3K9 or H3K36.

   Certain limitations were noted in this study regarding the clinical
   samples, particularly those from poorly differentiated neoplasm. Our
   study emphasises the LEF1-KDM4A axis, while the association of the KDM4
   family and Wnt/β-catenin pathway is not fully discussed, and other
   downstream genes of the axis, save for LATS2 remain to be explored.
   Moreover, it is unclear whether our findings can be applied to other
   solid cancers. Finally, the results of this study require validation in
   clinical settings via the development of agents capable of targeting
   LEF1-KDM4A-LATS2.

   In conclusion, this study revealed that LEF1 recruits KDM4A/N-CoR to
   suppress LATS2 expression, providing new insights into the functioning
   of LEF1 in cooperation with histone-modifying enzymes in OSCC. Our data
   provide a molecular basis for understanding the pathophysiological
   function of LEF1 and support the hypothesis that the LEF1-KDM4A-LATS2
   axis could be a potential therapeutic target for OSCC.

Materials and methods

   An extended description of the used materials as well as tables listing
   the employed antibodies is provided in [173]Supplementary Fig. and
   Tables.

Cell culture and transfection

   The CAL-27 cells used in this study were obtained from Procell Life
   Science & Technology Co., Ltd. (Wuhan, China); the SCC-9 cells were
   purchased from Shanghai EK-Bioscience Biotechnology Co., Ltd.
   (Shanghai, China). Both cell lines were STR-authenticated and tested
   for mycoplasma contamination. CAL-27 cells were cultured in Dulbecco’s
   modified Eagle’s medium (DMEM; VivaCell, Shanghai, China) supplemented
   with 10% foetal bovine serum (Gibco, Grand Island, NY, USA) and 1%
   antibiotics (Gibco). SCC-9 cells were cultured in DMEM/Nutrient Mixture
   F-12 (DMEM/F12) supplemented with 10% foetal bovine serum, 1%
   antibiotics, and 400 ng/mL hydrocortisone (Complete Growth Medium from
   Procell). Small interfering RNA (siRNAs) were transfected at a working
   concentration of 100 nM using Lipofectamine® RNAiMAX Reagent
   (Invitrogen, Carlsbad, CA, USA), according to the manufacturer’s
   instructions. Stable cell lines expressing shLEF1 or shKDM4A were
   generated by transfecting LV2-shLEF1 or LV2-shKDM4A into CAL-27 cells
   using the transfection reagent polybrene (Gene Pharma, Shanghai,
   China), followed by selection with 2 µg/ml puromycin (Sigma-Aldrich,
   St. Louis, MO, USA). The siRNA and small hairpin RNA (shRNA) sequences
   designed by Gene Pharma are listed in Table S[174]1 of Supplementary
   Figures and Tables.

Patient recruitment

   The patients were diagnosed and received treatment at the School and
   Hospital of Stomatology, Shandong University between May 2021 and July
   2022. Patients were included if they had histologically confirmed OSCC
   and had not receive related treatment prior to tumour removal. Patients
   were excluded if they had a history of cutaneous HNSCC, a second
   primary squamous cell carcinoma at a mucosal site outside the oral
   cavity, or other malignant tumours or serious diseases of vital organs.
   There were no samples excluded. Four samples of tumour tissue and
   corresponding adjacent normal tissue, which were confirmed by
   pathologists, were collected with the permission of the ethics
   committee of the Stomatological Hospital of Shandong University (No.
   20220703) and with patients' informed consent. No blinding method was
   used for sample collection. Associated clinical information is provided
   in Table S[175]2 of Supplementary Figures and Tables.

Immunohistochemistry (IHC) and immunofluorescence staining

   For IHC, microarrays of human oral cancer tumour tissues and adjacent
   normal tissues (HOraC080PG01) were purchased from Shanghai Outdo
   Biotech Co., Ltd. (Shanghai, China). Pathological information of the
   tissues is provided in Table S[176]3 of Supplementary Figures and
   Tables. For the IHC assay, the microarrays or other paraffin section
   samples were dewaxed with dimethylbenzene and dehydrated using an
   alcohol gradient (100%, 90%, and 75%). After antigen retrieval with
   citrate buffer, the samples were blocked with 3% H[2]O[2] and incubated
   with donkey serum. Slides were incubated with antibodies overnight at
   4 °C as described in Table S[177]4 of Supplementary Figures and Tables,
   followed by incubation with the appropriate horseradish peroxidase
   (HRP)-conjugated immunoglobulin G (IgG) polyclonal antibody for 30 min
   at 37 °C. Diaminobenzidine (DAB) and haematoxylin from Beijing Zhong
   Shan Golden Bridge Biological Technology Co., Ltd. (Beijing, China)
   were used to stain the slides. Specimens were assessed based on the
   positive stained percentage as determined by two independent observers
   who were not informed of the experiments, using an OLYMPUS-BX530
   microscope (Tokyo, Japan).

   For immunofluorescence, the cells were fixed with 4% paraformaldehyde
   for 10 min, permeabilised with 0.2% Triton X-100 in phosphate-buffered
   saline (PBS) for 10 min and incubated with 0.1% Triton X-100 and 0.4%
   bovine serum albumin (BSA) in PBS at room temperature. Cells were then
   incubated at 4 °C overnight with the following primary antibodies:
   rabbit polyclonal anti-JMJD2A, mouse monoclonal anti-LEF1, or rabbit
   polyclonal anti-N-CoR. Imaging was performed using a TCS SP8 Leica
   confocal laser scanning microscope and images were acquired using the
   Leica Application Suite X (LAS X, Leica, Wetzlar, Germany).

Proliferation assay and cell viability assay

   CAL-27 and SCC-9 cell proliferation was measured with the Click-iT™ EdU
   Cell Proliferation Kit for Imaging (Invitrogen) and CCK8 (MedChem
   Express, Monmouth Junction, NJ, USA), according to the manufacturer’s
   instructions. For the EdU assay, CAL-27 and SCC-9 cells were seeded at
   a density of 1 × 10^5 cells/well in 12-well plates with glass slides in
   each well. Cells were then treated with siRNAs which were diluted to a
   working concentration of 100 nM and transfected using Lipofectamine®
   RNAiMAX Reagent (Invitrogen). The medium was replaced with normal
   medium after 8 h, and the cells were subsequently cultured for an
   additional 24 h. EdU solution was added to the medium for the last 3 h
   of the 24 h incubation period. EdU incorporation was detected using a
   fluorescence microscope (OLYMPUS-BX530).

   For the CCK8 assay, 5 × 10^3 cells treated with siRNAs were seeded in a
   96-well plate 2 days before reagent addition, and after a 2 h
   incubation, proliferation was measured using a colorimetric assay with
   SpectraMax i3x (Molecular Devices. San Jose, CA, USA).

Flow cytometry

   Cell cycle analysis was performed using a Cell Cycle and Apoptosis
   Analysis Kit (US Everbright® Inc., Suzhou, China). Briefly, cultured
   cells were trypsinised to produce single cells, which were then fixed
   with 70% ethanol at −20 °C overnight. Cells were stained with propidium
   iodide for cell cycle analysis, which was performed using a cell
   analyser (BD LSRFortessa, BD Biosciences, San Jose, CA, USA). Data were
   collected and analysed using ModFit LT software (Verity Software House,
   ME, USA).

   Apoptotic cells were quantified using the Annexin V-FITC Apoptosis
   Detection Kit (Beyotime, Shanghai, China). Cultured cells were digested
   with trypsin to produce single cells, which were then stained with
   annexin V-FITC and propidium iodide, according to the manufacturer’s
   instructions. Apoptosis analysis was performed using BD LSRFortessa
   analyser. Data were collected and analysed using FlowJo software (BD
   Biosciences).

RNA-sequencing analysis

   siRNA was used to knockdown LEF1 and KDM4A expression in CAL-27 cells;
   three independent samples and controls were used in these experiments.
   Total RNA was extracted and purified using oligo(dT)-attached magnetic
   beads, and RNA-sequencing was conducted on an Illumina NovaSeq 6000
   (Illumina, San Diego, CA, USA) by Xiuyue Biol (Jinan, China).
   Differentially expressed genes between each cell group with a Q
   value < 0.05 and fold-change >1.5 were identified. The datasets
   supporting this research are available in the National Center for
   Biotechnology Information Gene Expression Omnibus (NCBI GEO) repository
   with the accession code [178]GSE230372.

IP and WB

   For IP assays, cells were washed twice with cold PBS, and extracts were
   prepared by incubating cells in lysis buffer (50 mM Tris-HCl, pH 7.4,
   150 mM NaCl, 2 mM EDTA, 0.3% NP-40, and protease inhibitor cocktail
   [MedChem Express]) for 30 min at 4 °C, followed by centrifugation at
   12,000 × g for 10 min. Next, 500 μg protein samples were incubated with
   the appropriate primary antibodies or normal rabbit/mouse IgG at 4 °C
   for 12 h with constant rotation and then mixed with 10 µL Dynabeads
   Protein G (Invitrogen) for 2 h at 4 °C. After washing the beads three
   times with cell-lysis buffer (50 mM Tris-HCl, pH = 8.0, 150 mM NaCl,
   2 mM EDTA, 0.1% NP-40), the captured immune complexes were stored at
   −20 °C.

   Whole-cell lysates were separated using 8–10% sodium dodecyl
   sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) Bis-Tris gel,
   transferred to polyvinylidene fluoride membrane (Merck Millipore,
   Billerica, MA, USA), blocked in 5% milk, and blotted with antibodies
   (Table S[179]4 of Supplementary Figures and Tables). The stained bands
   were detected using enhanced chemiluminescence (Merck Millipore)
   according to the manufacturer’s instructions.

GST pull-down experiments

   GST-fusion constructs were produced in Escherichia coli BL21 cells
   (Vazyme Biotech Co., Ltd., Nanjing, China), and bacterial lysates were
   obtained by sonicating the cells at 60 W in cold PBS supplemented with
   a protease inhibitor (MedChem Express). In vitro transcription and
   translation experiments were performed using rabbit reticulocyte
   lysates (Promega, Madison, WI, USA) and plasmid vectors expressing LEF1
   or KDM4A according to the manufacturer’s instructions to produce the
   target protein. For GST pull-down assays, 3–10 μl of the in vitro
   transcription/translation product was mixed with ~10 μg of the
   appropriate pGEX-4T-3-fusion protein and incubated in binding buffer
   (0.8% BSA in PBS containing 1% protease inhibitor). Subsequently, 30 μL
   of glutathione-agarose beads (GE Healthcare Little Chalfont,
   Buckinghamshire, UK) were added to the reaction solution and mixed by
   rotation at 4 °C for 2 h, washed five times with binding buffer, and
   then resuspended in 30 μl of 2× SDS-PAGE loading buffer. The proteins
   were detected by WB using specific antibodies.

LEF1 or KDM4A cDNA expression

   LEF1 cDNA (obtained by reverse transcription of CAL 27 cell RNA) and
   KDM4A cDNA (obtained by reverse transcription of plasmid from Addgene;
   Cambridge, Massachusetts, USA) were cloned into FLAG-tagged or
   EGFP-tagged plasmid (Addgene) using restriction endonuclease (Thermo
   Fisher Scientific, Waltham, MA, USA) and One Step Cloning Kit (Vazyme,
   Nanjing, China) according to manufacturer’s instructions.

   Primers were designed (see Table S[180]5 of Supplementary Figures and
   Tables) and DNA sequences were synthesised by Sangon Biotech (Shanghai,
   China). Linearised double-stranded DNA was transformed into competent
   E. coli cells by heat shock and cultured overnight, positive clones
   were identified by PCR and clone identification was done by the Beijing
   Genomics Institute (Beijing, China). The sequenced bacterial solution
   was cultured, and plasmids were extracted using the EndoFree Maxi
   Plasmid Kit (TIANGEN, Beijing, China). Purified plasmids with the
   target sequences were transfected into cells using polyethylenimine
   (PEI, Warrington, Pennsylvania, USA).

RT-PCR and gel electrophoresis

   Total cellular RNA was extracted using TRIzol reagent (Invitrogen)
   followed by application of chloroform, isopropanol, and 75% ethanol in
   diethylpyrocarbonate treated water. cDNA was prepared using RevertAid
   First Strand cDNA Synthesis Kit (Thermo Scientific, MA, USA). PCR
   amplification was carried out using primers diluted to 10 µM and SYBR
   Green Real-time PCR Master Mix (TOYOBO, Osaka, Japan), and the thermal
   cycling protocol was set according to the manufacturer’s instructions:
   95 °C for 1 min, followed by 40 cycles of 95 °C for 15 s, 60 °C (or
   optimised temperature) for 15 s, and 72 °C for 45 s, followed by
   melting curve analysis. Relative quantitation was performed based on
   SYBR green fluorescence using an ABI QuantStudio Real-Time PCR System
   (Applied Biosystems), and the results were obtained using the
   comparative Ct method (2^-ΔΔCt) with glyceraldehyde 3-phosphate
   dehydrogenase (GAPDH) as an internal control. The primers used are
   listed in the Table S[181]6 of Supplementary Figures and Tables.

Quantitative ChIP (qChIP) assays

   qChIP assays were performed using CAL-27 cells as described previously
   [[182]57]. Briefly, 1 × 10^7 cells from a 10 cm culture dish were
   cross-linked with 1% formaldehyde, sonicated, pre-cleared, and
   incubated with 2 μg of primary antibody against normal rabbit IgG
   (control), normal mouse IgG (control), LEF1, or KDM4A. The complex was
   washed with low- and high-salt buffers, and DNA was extracted for qChIP
   assays. The primers used for qChIP are listed in the Table S[183]7 of
   Supplementary Figures and Tables.

   PCR products were separated by agarose gel electrophoresis with 6×
   loading dye (Beyotime), and DNA was visualised using an ultraviolet
   transilluminator (Tanon, Shanghai, China). The primers used are listed
   in Table S[184]7 of Supplementary Figures and Tables.

In vivo hyperplasia

   Twelve 6-week-old male NOD/SCID mice were purchased from Vital River
   Laboratories (Beijing, China) and housed at appropriate ventilation
   conditions with alternating 12 h light/12 h dark cycles at 25 °C.
   Specific Pathogen Free (SPF) feed and bedding were adopted, and the
   cages and drinking bottles used were replaced, cleaned, and disinfected
   every 3 days. The mice were randomly divided into three groups; the
   four mice in the same group received the same treatment. There were no
   animals excluded.

   CAL-27 cells engineered to stably express firefly luciferase were
   infected with lentiviruses carrying only the vector, vector + shLEF1,
   or vector + shKDM4A. The xenograft models were generated via
   subcutaneous injection of 0.2 ml of CAL-27 cells (5 × 10^6
   cells/0.2 ml) in the right lateral intrascapular areas, and xenograft
   growth was monitored every 2 days. For bioluminescence imaging, the
   mice were abdominally injected with 200 mg/g d-luciferin (PerkinElmer,
   MA, USA) in PBS. Ten minutes after injection, the mice were
   anaesthetised with gaseous isoflurane (RWD, Shenzhen, China), and a
   charge-coupled device camera (IVIS kinetic; PerkinElmer) was used to
   image bioluminescence using the following settings: field of view, D;
   binning, medium; open emission filter; 4/f stop; and imaging time,
   auto. Bioluminescence was manually defined based on the relative
   optical intensity. Photon flux was normalised to the background,
   defined as the relative light intensity plotted from fluorescein-free
   mice. After four weeks, the tumours were removed for measurement and
   WB. No blinding method was used for animal experiments. There were no
   animal exclusion criteria. Animal handling procedures were approved by
   the ethics committee of the Hospital of Stomatology, Shandong
   University (No. 20220704).

Statistical analysis

   Data analysis was performed with GraphPad Prism 8.0 software (GraphPad
   Software Inc., San Diego, CA, USA). The Shapiro–Wilk test was performed
   to assess if values fit the normal distribution. Two-tailed unpaired
   t-test or one-way analysis of variance (ANOVA) were used for normally
   distributed variables. F-testing was used to test homogeneity of
   variance. If equal variance was not assumed, unpaired t-tests with
   Welch’ s correction or Welch ANOVA tests were used. Differences among
   non-normally distributed variables were analysed using the Mann–Whitney
   U-test or Kruskal–Wallis H-test. Correlation coefficients between
   different genes were calculated using Pearson’s correlation analysis.
   Results were reported as the mean ± standard deviation (SD) unless
   otherwise noted. No statistical methods were used to predetermine the
   sample size. No blinding was used during experiments and outcome
   analysis. Tumour datasets were downloaded from
   [185]http://xena.ucsc.edu/, and the number of samples was indicated for
   each figure.

Supplementary information

   [186]41419_2023_6024_MOESM1_ESM.xlsx^ (655.8KB, xlsx)

   Differentially Expressed Genes Identified by RNA-sequencing analysis
   [187]Supplementary Figures and Tables^ (440.3KB, docx)
   [188]Original Data File^ (1.2MB, docx)
   [189]aj-checklist^ (1.8MB, pdf)

Acknowledgements