Spatial metabolomics atlas in the progression of oral squamous cell carcinoma

Abstract

Background

Oral squamous cell carcinogenesis is a complex biological process, although some progress has been made in predicting the risk of malignant transformation of oral squamous cell carcinoma (OSCC). We aimed to visualise the dynamic metabolic characteristics of the progression of OSCC using desorption electrospray ionisation mass spectrometry imaging (DESI-MSI).

Methods

Eight matched OSCC samples were analysed using DESI-MSI. MetaboAnalyst database was used to screen for differential metabolism, and perform metabolic pathway analysis. Key pathways and enzymes were validated using immunohistochemical (IHC) techniques in an additional 60 patients with OSCC. Knockdown OSCC cell lines were constructed and functional experiments were carried out to elucidate the effect of key enzyme on the proliferation and migration of OSCC.

Results

Spatial metabolomics revealed 55 differential metabolites between cancer and normal mucosal tissues, 59 between cancer and precancerous tissues, and 52 between precancerous and normal mucosal tissues. We further identified 30 metabolites that were either increased or decreased from normal tissues to precancerous lesions and then to cancer tissues. Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathway analysis confirmed that sphingolipid metabolism and biosynthesis of unsaturated fatty acids are the main metabolic pathways involved in OSCC. Dihydroceramide desaturase (DEGS1) is the key enzyme in sphingolipid metabolism and is involved in ceramide synthesis, which is closely related to tumour carcinogenesis. We further confirmed that high expression of DEGS1 predicts poor overall survival and disease-free survival, and is closely related to the pathological grade of the tumour. In multivariate COX regression analysis, high expression of DEGS1 in tumor cells was an independent risk factor for OS in OSCC patients. Moreover, we demonstrated that DEGS1 knockdown inhibited the proliferation and migration of OSCC cells in vitro.

Conclusion

Our study used DESI-MSI to uncover the progression of OSCC at the spatial metabolomics level, and discovered key metabolic molecules involved in the carcinogenesis of OSCC. Sphingolipid metabolism plays an important role in OSCC carcinogenesis, and its key metabolic enzyme, DEGS1, is expected to become a new therapeutic target for OSCC in the future.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07421-2.

Introduction

Oral squamous cell carcinoma (OSCC) is a prevalent malignant tumour originating from the epithelium of the oral mucosa within the oral and maxillofacial regions, constituting over 85% of head and neck cancers [1]. The primary treatment approach involves surgical intervention complemented by a multidisciplinary approach, including radiotherapy, chemotherapy, and biological therapy. Despite advancements in therapeutic strategies, the 5-year overall survival (OS) rate for OSCC has remained at approximately 60% over the past three decades [2]. The American Cancer Society highlights that the majority of OSCC cases are diagnosed at advanced stages, significantly impacting the prognosis [3]. Notably, approximately 20% of OSCC cases arise from epithelial dysplasia, classified as oral potentially malignant disorders, such as oral leucoplakia and erythroplakia [4]. Consequently, the precise prediction of high-risk oral mucosal precancerous lesions is crucial for the early diagnosis and treatment of OSCC.

Currently, surgical excision biopsy is generally performed in clinical practice for precancerous lesions suspected of malignant transformation, and pathological diagnosis is made based on cell and tissue morphology changes observed under the microscope. However, molecular level changes at the early stages of malignant transformation cannot be distinguished by tissue morphology [5]. A defining characteristic of cancer cells is metabolic reprogramming, which alters metabolic pathways to satisfy the increased energy demands of proliferating cancer cells [6, 7]. Unlike genomics and proteomics, metabolomics directly examines metabolites downstream of gene and protein expression, which is more closely linked to phenotypic changes [8]. Metabolomics involves both qualitative and quantitative analyses of metabolites within organisms, focusing on key metabolic processes, such as carbohydrate, amino acid, and lipid metabolism. The primary objective of this study was to identify altered metabolic pathways and associated biomarkers [9]. In our previous study, we successfully used metabolomics, applying techniques such as gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS), to identify amino acid markers of OSCC [10]. However, conventional metabolomics requires tissue homogenisation, which destroys the spatial context of samples.

In recent years, ambient ionisation mass spectrometry (AIMS) has emerged as a significant tool for the rapid and precise diagnosis of cancer [11]. Unlike traditional LC/GC-MS systems, AIMS does not require routine sample extraction or purification, thereby preserving the spatiotemporal integrity of cells within frozen tissue sections and allowing in-situ visualisation of tissue metabolic characteristics [12]. Among the techniques included under ambient ionisation, desorption electrospray ionisation (DESI) is particularly notable for its minimal sample preparation requirements [13, 14]. DESI-MSI not only captures a wide array of chemical substances but also provides detailed spatial distribution data [15], which has been applied to differentiate cancerous from normal tissues across several cancer types, including ovarian, pancreatic, gastric, brain, and breast cancers [16–20]. This study focused on the dynamic metabolic characteristics of OSCC in-situ, and explored new molecular mechanisms of OSCC carcinogenesis.

Methods

Participants and tumour specimens

Eight participants were recruited from the Nanjing Stomatological Hospital of Nanjing Medical University and the Jiangsu University Affiliated Hospital. The study was approved by the Medical Ethics Committee, and all participants provided informed consent. The inclusion criteria targeted patients with primary OSCC who underwent radical surgery at the aforementioned hospitals. For DESI-MSI analysis, only superficial mucosal edges were collected. Each sample comprised OSCC tumour tissue along with a contiguous mucosal margin of at least 1.5 cm. Post-collection, tissues were snap-frozen in liquid nitrogen within 30 min and stored at -80 °C.

Histopathological examination of tissues

Two parallel frozen tissue sections, labelled #1 and #2, were subjected to either histological evaluation or DESI-MSI analysis. Following sectioning, all samples were stained with haematoxylin-eosin (HE), which facilitated the histopathological identification of areas containing OSCC, dysplastic epithelium (precancerous lesions), and normal mucosal tissue within the samples. After the standard HE staining protocol, two pathologists delineated the specific areas of squamous cell carcinoma, precancerous lesions, and normal mucosal tissues on the slides. The distinct morphologies of tumour tissue, dysplastic epithelium, and normal mucosal tissue, were differentiated under microscopic observation, ensuring accurate analysis and comparison across sections.

DESI-MSI analysis

The DESI-MSI setup involved a two-dimensional DESI source (Prosolia, Indianapolis, IN, USA) integrated with an LTQ Orbitrap Elite mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) for the molecular imaging of tissues. DESI-MSI was conducted in the negative ion mode, scanning a mass-to-charge (m/z) range of 200–1,000, with a spatial resolution of 200 μm, tailored for imaging experiments. The system’s DESI sprayer, featuring two coaxial capillaries, facilitated the delivery of the spray solvent and nebulising gas across the tissue sections. The employed spray solvent was a 1:1 volume-to-volume mix of dimethylformamide and acetonitrile, with a flow rate set at 2.0 µL/min, a high voltage of -4 kV, and a nebulising gas pressure of 1.0 MPa. Further technical details on the DESI geometric principles and specifics concerning the equipment are provided in the supplementary information and based on previous publications [21, 22]. To process the raw files from DESI-MSI, they were first converted into.cdf format, and subsequently loaded into a specialised imaging software developed by Massimager company. This software is adept at ion imaging reconstruction and facilitates multivariate analysis, allowing for the integration of ion images with optical images stained with HE.

Identification of metabolites

The extracted admixture ions were identified using the free Human Metabolome Database (www.hmdb.ca) and LIPID MAPS (www.lipidmaps.org). Using the standard molecular weight in the database with a mass accuracy error of < 5 ppm, combined with the high-resolution MS isotope peak, the element name of the m/z admixture ion was finally determined.

Kyoto encyclopedia of genes and genomes (KEGG) gene set enrichment analysis

Differentially expressed metabolites, either progressively increasing or decreasing from normal tissue through precancerous lesions to cancerous tissue, were analysed using KEGG (https://www.kegg.jp/) metabolic pathway analysis to identify crucial metabolic pathways involved in carcinogenesis.

Immunohistochemistry (IHC) and quantification

60 participants were recruited from the Jiangsu University Affiliated Hospital, and all participants provided informed consent. A total of 60 patients with OSCC with cancerous and matched normal tissues were used to validate the expression of dihydroceramide desaturase1 (DEGS1), which was analysed by IHC staining (Bioss, bs-4057R, Beijing, China). Tissue slides fixed in formalin and embedded in paraffin were dewaxed using xylene and rehydrated using an ethanol series. Antigen retrieval was performed in a pressure cooker, followed by blocking of endogenous peroxidase activity using a 3% hydrogen peroxide solution. After three washes with phosphate-buffered saline, the slides were incubated overnight at 4 °C with a rabbit anti-DEGS1 antibody at a 1:200 dilution. The secondary antibody was applied and incubated at 37 °C for 20 min. Detection was achieved using the diaminobenzidine (DAB) chromogen. Protein expression was assessed based on staining intensity and the percentage of positive cells, with intensity scored from 0 (negative) to 3 (high intensity) and positive percentage from 0 (no staining) to 4 (over 75% cells). The final score was the product of the two metrics. Scoring was performed independently by two pathologists who were blinded to the patients’ clinical details and outcomes. DEGS1 expression was categorised as “low” if below the mean and “high” if equal to or above the mean.

Cell culture and reagents

The human OSCC cell lines CAL33, CAL27, SCC9, OSCC3, and the immortalised human keratinocyte cell line (HaCaT) were cultured in Dulbecco’s Modified Eagle Medium, high glucose (DMEM-H), supplemented with 10% foetal bovine serum and 1% penicillin–streptomycin. The cells were cultured in a standard incubator at 37℃ with 5% CO₂. All cell culture reagents were obtained from Gibco (Thermo Fisher Scientific, Waltham, MA, USA).

RNA extraction and quantitative real-time PCR (qRT-PCR) analyses

RNA was extracted using TRIzol reagent following the manufacturer’s instructions, and RNA concentration and purity were measured. Total RNA was then reverse-transcribed into cDNA using the MightyScript Plus First Strand cDNA Synthesis Master Mix (Novoprotein, Cat. No.: E047 Shanghai, China). Gene expression levels were quantified using NovoStart^® SYBR qPCR SuperMix Plus (Novoprotein, Cat. No.: E096 Shanghai, China). The following specific primer sequences were used: forward primer 5–TGTGGAATCGCTGGTTTGGAATG–3 and reverse primer 5–GGAATATCTACATCGACGCCATCAG–3 for human DEGS1; forward primer 5–CATGTACGTTGCTATCCAGGC–3 and reverse primer 5– CTCCTTAATGTCACGCACGAT–3 for human β-actin. Gene expression was normalised to β-actin, and the 2^−ΔΔCT method was employed to quantify relative changes in gene expression.

RNA interference

We initially performed a transient transfection of Cal33 cells using a small interfering RNA (siRNA) kit (Zebrafish Biotech Co., Ltd. Nanjing, China) targeted at DEGS1, which included three different siRNAs and a control. Following the 24-h transfection period, DEGS1 mRNA expression levels were assessed using qPCR [23]. The target sequences for SiRNAs DEGS1 were as follows: siRNA1, 5–GGUCAUGAAACUUACUCAUAU–3; SiRNA2, 5–CUUCAAUGUGGGUUAUCAUAA–3; siRNA3, 5–GGCAAAGUAUCCAGAGAUA–3.

Cell counting Kit-8 (CCK8) assay

Cells were seeded in 96-well plates at an initial density of 3,000 cells/well. Cell proliferation was monitored at 0, 24, 48, and 72 h post-inoculation. Optical density values at 450 nm were recorded to assess the proliferation rates. Statistical analyses were performed using GraphPad Prism 8 software.

Wound healing assay

Cells were seeded in 6-well plates and grown to 100% confluency. Following overnight starvation in serum-free DMEM, wounds were created using a micropipette (200 µL) tip. The cells were then washed to remove dislodged cells and debris. The same area of the wound was photographed at 0 and 24 h post-wounding to assess cell migration.

Statistical analysis

To assess the variable data distribution across the tumour (T), precancerous lesion (Pre), and normal (N) regions, we compared the mean intensity of each feature marker using a nonparametric hypothesis test. The statistical significance of these markers was established using the Mann–Whitney U test. HE staining of frozen sections from large OSCC specimens with continuous mucosal incisal margins was initially performed. These sections, obtained from bedside surgeries, were histopathologically examined to include eight samples with complete T, Pre, and N margins for statistical analysis. MS data were generated for each tissue type, and differential metabolites between T vs. Pre, Pre vs. N, and T vs. N were identified through univariate analysis (Fold change, Mann–Whitney U test). Concurrently, multivariate analyses, including hierarchical clustering and discriminative orthogonal projection to latent structure-discriminant analysis (OPLS-DA), were conducted. The diagnostic performance of the characteristic lipids was evaluated using the area under the receiver operating characteristic curve. The Kruskal–Wallis test was used to determine the significance of metabolites across multiple groups, leading to the identification of metabolite changes during the progression from normal mucosa to precancerous lesions to carcinoma. Key metabolic pathways in OSCC carcinogenesis were delineated using KEGG pathway analysis. Differences in DEGS1 mRNA expression between tumour and adjacent normal tissues were analysed using a paired T-test, and Pearson’s chi-square test was used to assess the correlation with clinicopathological features in patients with OSCC. Kaplan–Meier analysis was employed to examine patient survival, and OS and disease-free survival (DFS) were evaluated using log-rank tests. An independent prognostic analysis was conducted using univariate COX and multivariate COX analyses. A p-value < 0.05 was considered statistically significant. Statistical analyses and graphical representations were performed using GraphPad Prism (version 8.0) and SPSS (version 25.0).

Results

Clinical characterisation of OSCC tissues and DESI-MSI analysis

The flowchart of the study is shown schematically in Fig. 1. Eight paired human OSCC tissue samples comprising tumour tissue (T), precancerous lesions (Pre), and normal tissue (N) were collected prospectively and analysed using DESI-MSI (Fig. S1). The clinical information of the 8 participants, including patients with early and advanced OSCC, is shown in Table 1. All cases were confirmed using HE staining to distinguish different tissue areas (Fig. 2A-C). As shown in Fig. 2D-F, the ion abundance (m/z) varied in different tissue regions. Preliminary mass spectrometry analysis revealed that within the mass-to-charge ratio range of 700–900, tumour and precancerous tissues presented differentiated signal patterns compared to normal tissues. This suggests that there may be characteristic molecular changes related to this m/z range among different OSCC progression stages, which could potentially be related to changes in lipid metabolism.

Fig. 1.

The overall flow chart of the study. A, Sample collection and DESI-MSI analysis(n = 8). B, Metabolomics data analysis. C, Validation analysis

Table 1.

Clinical case data were included in this study(n = 8)

Number	Gender	Age	Position	Pathological grading	LNM(1+, 0-)	pTNM staging
1	Male	66	Tongue	I-II	1	pT3N2bM0
2	Male	56	Tongue	I-II	0	pT1N0M0
3	Female	80	Tongue	I-II	0	pT2N0M0
4	Male	58	Tongue	I-II	1	pT3N2bM0
5	Female	74	Cheek	I-II	1	pT3N2bM0
6	Male	74	Cheek	I-II	1	pT2N2bM0
7	Male	73	Oropharynx	I-II	1	pT2N1M0
8	Male	56	Oropharynx	I-II	1	pT4aN2bM0

Fig. 2.

Representative HE staining images and mass spectral profiles of different regions of the same patient. A, D Normal tissues; B, E precancerous tissues; C, F Cancerous tissues

Screening differential metabolites in normal, precancerous, and cancerous tissues

Eight paired OSCC patients were analysed using DESI-MSI, and 474 ions were detected. The metabolite names of the ions were identified, and most of them were lipid molecules (Tables S1).

OPLS-DA was used to investigate the overall metabolic differences between tumour, precancerous, and normal tissues in patients with OSCC. The results showed significant differences in the tissue metabolic profiles between the groups (Fig. 3A-C).To discern the most critical metabolites across different groups, we evaluated their expression patterns using the variable importance for the projection (VIP) scores from the OPLS-DA model, focusing on their impact on sample classification and discrimination. Metabolites significant in both multivariate (VIP > 1) and univariate (p < 0.05) analyses were highlighted, revealing 55 significant metabolites between tumour and normal tissues (Fig. 3D, Table S2), 59 between tumour and precancerous tissues (Fig. 3E, Table S3), and 52 between precancerous and normal tissues (Fig. 3F, Table S4).

Fig. 3.

The pairwise differences in metabolites were obtained through t-test for 8 patients. A Scores plot segregated cancer and normal tissues by OPLS-DA. B Scores plot segregated cancer and precancerous tissues by OPLS-DA. C Scores plot segregated cancer and precancerous tissues by OPLS-DA. D Volcano plot between cancer and normal tissues. E Volcano plot between cancer and precancerous tissues. F Volcano plot between precancerous tissues and normal tissues. G-I ROC analysis of for distinguish cancer and normal tissues. J-L ROC analysis for distinguish cancer and precancerous tissues

Hierarchical clustering was applied to all metabolomic data to detect significant metabolic shifts among the different tissue groups. The intragroup similarities of the samples and their metabolic profiles were visualised as heatmaps in the negative ion mode (Fig. S2A-B). Our findings indicated pronounced metabolic dysregulation in cancer tissues compared to normal or precancerous tissues, characterised by an overall increase in metabolites, with few exhibiting decreased levels.

Receiver operating characteristic curve analysis further assessed the discriminative power of these metabolites for OSCC and precancerous lesions, revealing metabolites with AUC values > 0.70 in tumour vs. normal (Fig. 3G-I, Table S5) and tumour vs. precancerous comparisons (Fig. 3J-L, Table S6).

Spatial metabolic changes during the normal tissues-to‐premalignant (moderate to severe dysplasia)‐to‐OSCC transition

Kruskal–Wallis test was used to elucidate the dynamic metabolic changes from normal mucosa to premalignant lesions to carcinoma, which identified 57 metabolites of statistical significance (Fig. 4A, Table S7). To investigate the spatial characteristics of the differential metabolites, the 57 metabolites were subjected to K-means cluster analysis. Four distinct clusters of metabolites were identified (Fig. 4B). Many metabolites displayed consistent changes over time, such as those in clusters 1, 2, and 3. In particular, cluster 1 metabolites (e.g., PS [18:0/18:1(11Z)], 3-hydroxyicos-11-enoic acid) and cluster 3 (e.g., C16-ceramide, 3-O-Sulfogalactosylceramide [d18:1/18:0]) metabolites showed a steady increase in upregulation throughout the normal tissues-to-precancer tissues‐to‐cancer tissues transformation, whereas the regulation of cluster 2 metabolites (e.g., 5-beta-cholanic acid) diminished over time. Some metabolites were significantly upregulated only at specific stages. For instance, metabolites in cluster 4 were significantly upregulated only at the precancer stage. It is worth noting that the different mass spectrometry characteristics of adrenic acid were simultaneously identified in both Cluster 1 and Cluster 2, which represent opposite trends. Specifically, the ion with a mass-to-charge ratio (m/z) of (X331.2637) showed a significant increase in Cluster 1, while the ion with m/z of (X367.2403) showed a significant decrease in Cluster 2 (Fig. 4B). This suggests that adrenic acid may be involved in a variety of functionally distinct biological processes during carcinogenesis. By conducting metabolic pathway analysis on each of the four metabolite clusters, it was found that cluster 1 and cluster 2 were significantly enriched in the same metabolic pathways, mainly including Ferroptosis and biosynthesis of unsaturated fatty acids (Fig. S3A-B). The metabolites in cluster 3 mainly participate in sphingolipid metabolism, fatty acid biosynthesis, and the biosynthesis of unsaturated fatty acids (Fig.S3C). However, for metabolites in cluster 4, which have high heterogeneity, no significant pathway enrichment signals were observed. Then, among the 57 metabolites, 30 metabolites that exhibited an increasing or decreasing trend from normal tissues to Dys and then to cancer tissues, were further selected (Table 2). Heatmaps illustrated the distribution of these changes across samples (Fig. 4A).

Fig. 4.

Differential metabolites associated with OSCC progression. A Heat map analysis of increasing or decreasing metabolites from normal tissue to cancerous tissue, p < 0.05. B A K-means clustering analysis was conducted on 57 metabolites with statistical significance. C The characteristic metabolite PS (18:0/18:1(11Z)) was showed by DESI-MSI

Table 2.

Differential metabolites that increase or decrease from normal-precancerous-cancerous tissues

m/z	Name
281.2463	Oleic acid
301.2145	Eicosapentaenoic acid
305.2459	20-Hydroxy-6Z,15Z-eicosadienoic acid
307.2612	3-hydroxyicos-11-enoic acid
309.2774	11Z-Eicosenoic acid
327.2295	Dihydrotestosterone propionate
329.2456	Docosapentaenoic acid
331.2637	Adrenic acid
445.3137	Menahydroquinone-4
535.469	DG(14:0/0:0/17:0)
535.4685	DG(14:0/0:0/17:0)
537.4861	FAHFA(18:0/7-O-16:0)
561.4843	DG(15:0/18:1(11Z)/0:0)
562.4874	Cer(d17:1/18:1(12Z)-O(9 S,10R))
564.5031	N-(2R-Hydroxyhexadecanoyl)-2 S-amino-9-methyl-4E,8E-octadecadiene-1,3R-diol
565.5157	FAHFA(18:0/8-O-18:0)
572.4758	C16-Ceramide
572.4769	C16-Ceramide
788.539	3-O-Sulfogalactosylceramide (d18:1/18:0)
788.5478	PS(18:0/18:1(11Z))
281.2533	NA
282.2497	NA
306.2488	NA
536.4722	NA
626.5311	NA
626.5375	NA
395.2724	5beta-Cholanic acid
337.1884	NA
368.2449	NA
529.4456	NA

These metabolites are potential diagnostic markers in clinical settings, where early molecular changes are indistinct in cell morphological assessments. This is supported by the consistent correlation between the molecular findings of the 30 metabolites and the pathological outcomes of HE-stained sections in subsequent analyses of the 8 frozen slices (Fig. 4C, Fig.S4). These results establish the clinical relevance of these metabolites in distinguishing normal, precancerous, and cancerous tissues, underscoring their potential for enhancing diagnostic precision.

Sphingolipid metabolism May be involved in the progression of OSCC

Dysregulated lipid metabolism is one of the most prominent metabolic alterations observed in cancer. Abnormal lipid metabolism regulates the metabolic flux of tumour cells and extensively affects various malignant biological processes, such as cell survival and proliferation. Therefore, we conducted metabolic pathway analyses of the differential metabolites to identify the key metabolic pathways involved in OSCC carcinogenesis. KEGG pathway analysis of the selected differential metabolites highlighted significant pathways, such as sphingolipid metabolism and the biosynthesis of unsaturated fatty acids. Among them, the sphingolipid metabolism was the most active in tumour tissues (Fig. 5A), indicating a potential link between lipid metabolism and the OSCC. Sphingolipid metabolism, known for its complexity, plays a pivotal role in OSCC pathogenesis (Fig. 5B), and may be involved in OSCC progression.

Fig. 5.

Metabolic pathway analysis of OSCC carcinogenesis. A Pathway enrichment analysis of OSCC by KEGG, p < 0.05. B Schematic diagram of sphingolipid metabolic pathways. C C16-ceramide relative abundance in normal, precancer and cancer, *** for p < 0.001, ** for p < 0.01, * for p < 0.05. D DEGS1 catalyzes the synthesis of ceramide from dihydroceramide

We analysed the expression of key metabolites involved in the sphingolipid metabolism pathways in different tissue regions. Our results indicated that C16-ceramides progressively increase from normal to precancerous to cancerous tissues (Fig. 5C). This metabolite change suggests that the ceramide metabolic pathway is aberrantly activated in tumour tissues of patients with OSCC. Ceramide is produced from dihydroceramide, catalysed by DEGS1 (Fig. 5D). DEGS1, a desaturase/hydroxylase superfamily member, is the only DEGS present in human cells [24] that catalyses the final step in the de novo synthesis of ceramide. Currently, the role of DEGS1 in OSCC remains unclarified. However, given its key role in regulating the ceramide-dihydroceramide metabolic axis and pro-tumour effects in other cancers, DEGS1 may be a promising therapeutic target in OSCC.

DEGS1 is a potential biomarker of OSCC

To further assess the diagnostic and therapeutic significance of DEGS1 in OSCC, we compared the expression levels of DEGS1 between head and neck squamous cell carcinoma (HNSCC) tumours and normal samples using data from the TIMER2 database (http://timer.comp-genomics.org/). The results showed that the mRNA level of DEGS1 was significantly upregulated in HNSCC tumour tissues compared to normal tissues (Fig. 6A). We further analysed the expression profiles of DEGS1 across tissue specimens and observed differential expression levels between normal and tumour tissues. The results showed that DEGS1 was highly expressed in OSCC tumour tissues (p < 0.05) (Fig. 6B). We further validated DEGS1 using clinical samples; IHC analysis confirmed elevated DEGS1 expression in cancer versus normal tissues (Fig. 6C, D). DEGS1 expression was positively correlated with tumour differentiation (Table 3), and its high expression was associated with poor OS and DFS (Fig. 6E, F). Univariate COX regression analysis showed that high expression of DEGS1 in tumor cells, T stage, and lymph node metastasis were risk factors for OS and DFS in patients (Fig. 6G, H). In multivariate COX regression analysis, high expression of DEGS1 in tumor cells was an independent risk factor for OS in OSCC patients, while lymph node metastasis was an independent risk factor for DFS in OSCC patients (Fig. 6I, J).

Fig. 6.

DEGS1 promote OSCC progression. A DEGS1 expression levels was analyzed by TIMER2 online database. B Expression level of DEGS1 mRNA in OSCC tissues. C, D DEGS1 expression was detected by IHC in OSCC tissue specimens. E, F DEGS1 expression correlated with OSCC prognosis. G Expression level of DEGS1 mRNA in OSCC cell lines. H-J Knockdown expression of DEGS1 inhibited OSCC proliferation and migration ability. *** for p < 0.001, ** for p < 0.01, * for p < 0.05

Table 3.

Correlation between clinicopathology of OSCC patients and DEGS1 expression

Clinical Variables	Low n (%)	High n (%)	χ2	p Value
Gender			0.049	0.825
Male	15(60.0)	10(40.0)
Female	20(57.1)	15(42.9)
Age			0.034	0.853
< 60	12(60.0)	8(40.0)
≥ 60	23(57.5)	17(42.5)
T stage			3.704	0.054
I–II	20(71.4)	8(28.6)
III–IV	15(46.9)	17((53.1)
Lymph node metastasis			0.17	0.895
No	16(59.3)	11(40.7)
Yes	19(57.6)	14(42.4)
Differentiation			5.346	0.021*
Well	32(65.3)	17(34.7)
Moderate/poor	3(27.3)	8(72.7)

χ2, Pearson’s chi-squared test. * represents that differences were considered statistically significant with p < 0.05

In this study, we performed in vitro functional validation of the key gene, DEGS1. To investigate the function of DEGS1 in OSCC, we first compared the expression of DEGS1 in OSCC cell lines using RT-qPCR (Fig. 6K). The efficiency of DEGS1 gene knockdown in Cal33 cell lines was verified using RT-qPCR (Fig. 6L). Silencing of the DEGS1 gene using siRNA significantly inhibited Cal33 cell growth (Fig. 6M). Further experiments revealed that DEGS1 knockdown markedly reduced cell migration (Fig. 6N). Taken together, these results suggest that DEGS1 may serve as a potential biomarker for OSCC, and the mechanism by which it promotes cancer development deserves further study.

Discussion

OSCC is one of the most common malignant tumours of the head and neck, and poor prognosis of OSCC often stems from patients presenting at advanced clinical stages, emphasising the need for early detection, diagnosis, and treatment collectively referred to as the “three early stages”. The significance of early diagnosis in improving patient outcomes cannot be overstated. Current techniques used for the diagnosis of OSCC, such as Stokes shift spectroscopy, autofluorescence imaging, toluidine blue staining, genetic screening, and in vivo digital biopsy via confocal microscopy, although valuable, do not provide detailed molecular insights [25–28]. As an in-situ detection technique, DESI-MSI not only differentiates among normal, precancerous, and tumour tissues but also elucidates the metabolites driving these distinctions. Recent advances in molecular technologies, notably spatial metabolomics, have proven instrumental in various applications, including cancer diagnosis [29].

In a previous study, Dória et al. applied DESI-MSI to ovarian cancer samples to explore its diagnostic applications. The results showed that based on their lipidomic profile, tumour tissue and the surrounding stroma, as well as different epithelial ovarian carcinoma types, can be characterised and differentiated using DESI-MSI [30]. Santoro et al. applied DESI to identify lipidomic profiles characteristic of distinct molecular subtypes of breast cancer, including precancerous lesions, underscoring its potential to elucidate the mechanisms underlying breast cancer development [31].

OSCC carcinogenesis is a complex biological process, and significant progress has been made in predicting the risk of OSCC malignant transformation. Studies on the carcinogenesis of OSCC primarily focus on the genome, transcriptome, and proteome. For example, Syed et al. explored tissue-based metabolomics to identify clinically relevant metabolomic biomarkers for potential malignant oral diseases, aiming for early tumour diagnosis [32]. However, this lacks the spatial information of tissue samples and thus cannot achieve precise localization of cancer tissues. Therefore, achieving the spatial localization of OSCC metabolites could facilitate early detection, potentially preventing their progression [33]. In addition, the heterogeneity of tumour cells necessitates precise spatial localisation of tissue cells [34]. The latest research by Yuan et al. employed spatial multi-omics technology to reveal the spatial transcriptomics and spatial metabolomics characteristics of OSCC derived from oral submucosal fibrosis (OSF). Compared with previous studies, this research was able to precisely locate metabolic activities to specific cell regions, revealing the complex metabolism within the tumor microenvironment (TME), which is something that traditional omics cannot achieve. Moreover, this research revealed the metabolic reprogramming during the malignant progression of OSCC derived from OSF, and discovered abnormal expression of polyamine metabolism reprogramming and polyamine metabolism enzymes [35]. Cheng et al. utilized spatial multi-omics technology to investigate the internal metabolic heterogeneity of OSCC. They found that the tumor margins of OSCC have a high metabolic level. Moreover, the lower expression of Ras-related glycolysis inhibitors and calcium channel regulators (RRAD) at the tumor margins will activate glucose metabolism, thereby promoting the progression of OSCC [36]. However, the continuous progress process of OSCC has not yet been studied. Therefore, the progression from normal to precancerous (moderate to severe dysplasia) to tumour tissues in OSCC requires further spatial-level verification and study.

This study conducted a comparative analysis of metabolite evolution across different stages of OSCC, from normal mucosa to precancerous (moderate to severe dysplasia) to cancerous tissues, both morphologically and metabolically, to identify key differential metabolites that could assist in clinicopathological assessments for accurately diagnosing high-risk precancerous oral mucosal lesions. Using DESI-MSI, we analysed frozen sections from 8 OSCC specimens, identifying 55 distinct metabolites between tumour and normal tissues, 59 between tumour and precancerous (moderate to severe dysplasia) tissues, and 52 between precancerous (moderate to severe dysplasia) and normal tissues. Further comparative screening across these three groups revealed 57 significant metabolites, these 57 differential metabolites can be classified into 4 clusters. Although the trends of cluster 1 and cluster 2 were completely opposite, the analysis results were both significantly and jointly enriched in the two core metabolic pathways: Ferroptosis and biosynthesis of unsaturated fatty acids. This result indicates that during the carcinogenesis process of oral squamous cell carcinoma, complex and precise reprogramming occurred within these pathways, rather than simple overall activation or inhibition. There is a profound and complex connection between ferroptosis and the biosynthesis of unsaturated fatty acids [37]. Adrenal acid, as a special type of unsaturated fatty acid, plays a crucial role in accelerating cell death during ferroptosis. Ferroptosis is a form of metabolic cell death induced by oxidative stress. During tumor progression, the accumulation of lipids, excessive production of reactive oxygen species, and dysregulation of the cellular antioxidant system all contribute to the promotion of ferroptosis. However, in the later stage of tumor progression, cancer cells will upregulate some anti-ferroptosis pathways to adapt to the harsh environment, gain survival advantages, and thereby promote tumor invasion, metastasis and resistance to treatment [38]. Cluster 3 mainly involves sphingolipid metabolism, fatty acid biosynthesis, and the biosynthesis of unsaturated fatty acids. However, the increased and decreased metabolites in cluster 4 failed to form statistically significant and consistent pathway enrichment. Therefore, we speculate that the metabolic network they are involved in has undergone a complex and non-coordinated reconfiguration, rather than a simple overall activation or inhibition. We further selected 30 of these 57 significant metabolites that showed consistent changes from normal tissues to precancerous lesions to cancer tissues. Among them, 26 increased, and 4 gradually decreased from normal tissues to cancerous tissues. These results suggest that metabolites play an important role in OSCC carcinogenesis.

High-resolution (HR)-MS analysis using databases such as the Human Metabolome Database (HMDB) and Lipid Atlas confirmed that many of these metabolites are fatty acids, essential for cancer cell proliferation through lipid molecular signalling (Table 2). Pathway analysis highlighted lipid metabolism as the dominant pathway and sphingolipid metabolism as crucial. Notably, C16-ceramide levels were found to increase from normal to cancerous tissues, consistent with previous findings in HNSCC, where C16-ceramide is elevated in tumour tissues [39]. Ceramide functions centrally in sphingolipid metabolism as the simplest form and a precursor for more complex sphingolipids. Differential roles of ceramides have also been noted; for instance, C16-ceramide mediates pro-apoptotic functions, whereas C18-ceramide is associated with pro-survival effects in HNSCC, suggesting that ceramides containing different fatty acid chain lengths have different functions [40]. Studies have shown that knocking down CerS6 leads to a decline in C16-ceramide, which mediates activating transcription factor 6 (ATF6) activation and apoptosis [41]. Ceramide, the core component of sphingolipid metabolism, is synthesised from dihydroceramide by the catalytic activity of DEGS1. Although the role of DEGS1 in OSCC remains unclear, we investigated the role of DEGS1 in OSCC progression based on its core position in sphingolipid metabolism.

This study showed that DEGS1 is highly expressed in OSCC tissues and cells. DEGS1 knockdown markedly reduced the proliferation and migration of OSCC cells, consistent with the findings that targeting DEGS1 can suppress cancer cell growth. For instance, sphingosine kinases (SK1 and SK2), which catalyse the formation of sphingosine 1-phosphate, have been implicated in cancer cell dynamics. Inhibition of these enzymes, as demonstrated by McNaughton et al., leads to proteasome-mediated degradation of SK1 and suppression of DEGS1, thereby arresting cell growth [42]. Furthermore, Breen et al. found that DEGS1 knockdown altered sphingoid levels and enhanced apoptosis in human HNSCC cells following photodynamic therapy [23]. Our findings indicate that high DEGS1 expression in OSCC tissues correlates positively with tumour differentiation and predicts poor OS and DFS, underscoring its significance as a biomarker in OSCC. It has been confirmed that inhibiting the DEGS1 enzyme will regulate the level of cyclin D1 through NF-κB-dependent mechanisms, induce the accumulation of dhCer, low phosphorylation of Rb, and cell cycle arrest at the G0/G1 phase, thereby inhibiting cell growth [24]. Gagliostro et al. also found that the accumulation of dihydroceramide caused by inhibition of DEGS1 activity was associated with the expression of cyclin D1, the delay of cell cycle G1/S transition, the induction of ER stress, and the increase of autophagy [43]. The study found that DEGS1 facilitates the in vitro migration of esophageal cancer cell lines and induces tumor metastasis in nude mice, possibly due to the activation of NF-κB which leads to an increase in the expression of cyclin D1 [44]. These studies have shown that by inhibiting DEGS1 to increase the level of dihydroceramide and thereby enhancing autophagy, blocking the cell cycle at the G0/G1 phase, and inhibiting metastasis, this represents a promising target for cancer treatment.DESI-MSI offers a rapid, real-time assay that enhances the histopathological analysis of frozen sections. Unlike traditional mass spectrometry techniques, such as GC-MS and ultra-high-performance liquid chromatography-mass spectrometry (UHPLC-MS), DESI-MSI stands out as a potent adjunct to existing pathological diagnosis techniques for OSCC because of its direct tissue analysis capabilities. However, future studies should include a larger number of OSCC specimens to validate these results. In addition, the OSCC microenvironment is complex, which limits our findings.

In conclusion, our study uncovered the process of OSCC Progression at the spatial metabolomics level using DESI-MSI. Our findings suggest that lipid metabolites play an important role in OSCC progression. Sphingolipid metabolism is a key pathway in the development of OSCC, and its key metabolic enzyme, DEGS1, is expected to become a new therapeutic target for OSCC in the future.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

HtZ, QgH and XhY conceived and designed the experiments. HtZ and XwS performed the experiments. GfL, WL and YjF summarized and analyzed the data. HtZ, CwS, WH, ZjS and XhY contributed to writing and revising the paper. All authors read and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (82173380).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Medical Ethics Committee of the Affiliated Hospital of Jiangsu University.

Consent for publication

Not applicable.

Conflict of interest

The authors declare that they have no conflict of interest.