3.1 Clinical features of odontogenic cysts/tumor

The specimens in this study included OC, AM, and OKC. Representative imaging data and pathological data are shown in the Fig. 1a, b. We compared the pattern of bone destruction shown on radiographs, according to the result, it was confirmed that AM had the largest extent of destruction, and the unique pattern of destructive activities resulted in the strongest destructive capacity of AM (Fig. 1c). The unique pattern of destructive activities resulted in the strongest destructive capacity of AM (Fig. 1c). TRAP assay is commonly used to detect bone destruction capacity. We confirmed that AM had the strongest capacity for bone destruction, followed by OKC and OC the least (Fig. 1d). Both CTSK and MMP9 can achieve bone destruction through tissue matrix remodeling, and their high expression often predicts strong bone destruction capacity [17]. The IHC also confirmed that the bone destruction capacity of AM and OKC was significantly higher than that of OC. Among them, AM had the highest expression of CTSK (Fig. 1e), while MMP9 was the strongest in OKCs (Fig. 1f). The large deposition of Col1A1 is associated with tissue repair [18]. In Supplementary Fig. 1a, we also observed that AM had a lower expression of Col1A1.

Fig. 1

Differences in the extent of bone destruction in odontogenic cysts/tumor. a Panoramic radiographs in OC, OKC and AM. b H&E stain of OC, OKC and AM. c Statistical plot of the extent of bone destruction. d TRAP stain of OC, OKC and AM. e CTSK expression of OC, OKC and AM tissues. f MMP9 expression of OC, OKC and AM. Data information: All data shown (c–f) represent the means _ SD (n ≥ 3 biological replicates). ANOVA followed by Tukey’s post hoc test (c–f) was used for the statistical analysis. (*P < 0.05; **P < 0.01; ***P < 0.001)

3.2 Metabolic differences among odontogenic cysts/tumor

As shown in Fig. 2a, we performed metabolic profiling by employing the LC–MS/MS method and standardized the obtained data for subsequent analysis. 37 patients were included in this study, and the demographic and clinical characteristics are shown in Supplementary Table 2. A total of 697 metabolites were identified by high-throughput targeted metabolomics assays in tissue samples. As usual [19], the typical total ion chromatograms (TIC) map was obtained and the chemical composition of the metabolites was analyzed and representative TIC map in positive and negative ion modes were shown in Fig. 2b. The metabolites were identified (Supplementary Fig. 1a) and quality control was performe (Supplementary Fig. 1b). First, unsupervised principal component analysis (PCA) was performed after data standardization and showed significant clustering of OC, OKC, and AM tissues (Fig. 2c). Principal component (PC) 1 accounted for 25% of the variance observed among the metabolites associated with the three diseases. The initial three PCs collectively elucidated 47.7% of the variation among these factors. Subsequently, we found differences among the constituent components of each group by PCA (Supplementary Fig. 1c, d, e,). We then showed differences in metabolite composition among different groups (Top 20 were listed) (Fig. 2d). Among the three diseases, OC exhibited the highest lipids, while AM showcased the highest nucleic acid levels, and OKC displayed the most abundant peptides content. And then, the top 50 metabolites of relative abundance were selected for sample clustering (Fig. 2e). We found different metabolite enrichment patterns in different diseases. Examples include N, N-Dimethylglycine and Choline hydroxide, which are highly enriched in OC, Isoleucine and L-Methionine sulfone in OKC, and L-Garnitine and Hypoxanthine in AM. Taken together, the metabolic profile suggests a unique metabolic reprogramming of OKC and AM.

Fig. 2

Metabolic profiling of odontogenic cysts/tumor. a High-throughput targeted metabolome flow diagrams. b The chemical composition of the metabolites was analyzed and representative total ion chromatograms (TIC) in positive and negative ion modes. c PCA of OC, OKC and AM. d metabolite composition of OC, OKC and AM. e Cluster heat map analysis of OC, OKC and AM

Further, we explored the differences of metabolic profiles. Firstly, orthogonal partial least squares discriminant analysis (oPLS-DA) showed a clear separation between AM and OC (R2X = 0.332, R2Y = 0.984, Q2 = 0.494) (Supplementary Fig. 2a). The validity of the oPLSDA model was confirmed by permutation tests of 100 interactions (Supplementary Fig. 2b). Differential expression analysis identified significant differences in 13 metabolites, of which 8 metabolites were depleted and 5 metabolites were increased (Supplementary Fig. 2c, Supplementary Table 2). The top 5 metabolites of AM metabolism changes were the Nicotinamide D-ribonucleotide, L-Cysteine, L-Cystathionine, Carnosine and S-Sulfo-L-cysteine. Based on the differential metabolites, we identified 9 differential metabolic pathways in AM, including taurine and hypotaurine metabolism, nicotinate-nicotinamide metabolism, and glycine, serine and threonine metabolism (Supplementary Fig. 2d, Supplementary Table 3).

Similarly, oPLS-DA with permutation test confirmed the difference between OKC and OC (R2X = 0.361, R2Y = 0.968, Q2 = 0.573) (Supplementary Fig. 2e, & f). And subsequently, results showed that 6 metabolites decreased and 4 metabolites increased in OKC (Supplementary Fig. 2 g, Supplementary Table 4). Based on the differential metabolites of Diphosphoric acid, Nicotinamide D-ribonucleotide, Urocanate, Propane-1,2,3-tricarboxylate, and other metabolites, we found 10 differential metabolic pathways, including citrate cycle, pyrimidine metabolism, and galactose metabolism (Supplementary Fig. 2 h, Supplementary Table 5).

Inflammatory response has a certain impact on metabolism (He et al. 2020), and metabolic differences exist between healthy pulp and periapical lesions (Altaie et al. 2021). Therefore, we aimed to explore the metabolic compositions of NOC and IOC. The significantly clustered NOC and IOC also had different metabolite compositions, suggesting distinct metabolic processes between the two types of lesions (Supplementary Fig. 3a, b, and c). 8 metabolites were absent and 14 metabolites (Supplementary Fig. 3f) were increased in the IOC group with significant differences showed by oPLS-DA (R2X = 0.38, R2Y = 0.938, Q2 = 0.209) (Supplementary Fig. 3d, e, Supplementary Table 6). Histidine metabolism, primary bile acid biosynthesis, caffeine metabolism caused by vitamin C, hydantoin-5-propionate, glycocholate, glycochenodeoxycho-late and taurochenodeoxycholate were significantly different metabolic pathways in IOC (Supplementary Fig. 3f, and g, Supplementary Table 7). The same approach was used in the difference analysis between non- inflammatory OKC (NOKC) and inflammatory OKC (IOKC). The initial three PCs collectively elucidated 59.2% of the variation among these factors (Supplementary Fig. 4a). Different metabolites were shown in Supplementary Fig. 4b, and c. Similarly, PCA and permutation test were performed in NOKC and IOKC (Supplementary Fig. 2d, & e). And subsequently, results showed that 8 metabolites decreased and 13 metabolites increased in IOKC (Supplementary Fig. 4f, Supplementary Table 7). Based on the differential metabolites, we found 8 differential metabolic pathways (Supplementary Fig. 4 g, Supplementary Table 8).

3.3 WGCNA and differential analysis confirmed L-cysteine as the characteristic metabolite of OKC

WGCNA was used to seek the main representative modules for each group. Firstly, we searched for the outlier metabolites in the metabolites by hierarchical clustering method, and finally we found that all the metabolites were not outliers (Supplementary Fig. 5a). Using the WGCNA method, we divided the metabolites into 10 modules (Fig. 3a, b). Considering the effect of chronic inflammation on destructive activities, we divided OC into non- inflammatory OC (NOC) and inflammatory OC (IOC). Among them, NOC has 7 samples and IOC has 9 samples. The MEgreen is negatively correlated with AM (R = − 0.36, p = 0.03). The MEpink is negatively correlated with OKC (R = − 0.35, p = 0.04). As a result, the red module containing 35 metabolites became the representative module of AM (Supplementary Fig. 5b & Supplementary Table 8), and the pink module containing 13 metabolites was used for OKC (Supplementary Fig. 5c). Then, differential metabolites in each module were screened (Supplementary Table 9). We found that cysteine might be an important metabolite contributing to the development of AM, and Uridine diphosphate (UDP) played the similar role in OKC. In addition, the relationships between cysteine and AM, OKC and UDP were investigated. Similar to the aforementioned process, the obviously divergent AM and OKC had 29 differential metabolites (R2X = 0.396, R2Y = 0.944, Q2 = 0.455) (Fig. 3c, d). 12 of these metabolites were decreased and 17 metabolites were increased (Fig. 3e & Supplementary Table 8). In the comparison between OKC and AM, cysteine and methionine metabolism were the most significantly different pathways (Fig. 3f). In addition, up-regulation of 3-Sulfinoalanine, N-Acetyl-D-glucosamine 1-phosphate, D-Lactate, D-2-Aminobutyrate, L-Cysteine, and L-Cystathionine and down-regulation of Enterodiol, 15(S)-HETE, Urocanic acid, Palmitoylethanolamide, and Tricarballylate were also observed in the AM tissues. Meanwhile, the contents of cystine (Fig. 3g) and cysteine (Fig. 3h) gradually increased in OC, OKC and AM groups. According to the KEGG analysis of cysteine and methionine metabolism in OKC and AM, we traced back to CTH (Fig. 4a). Meanwhile, CTH-regulated cysteine metabolism played an essential role in the progression of OKC and AM. The difference in cysteine metabolic flux may lead to differences in biological properties between the two diseases. The expressions of CTH in the OKC, AM and OC were detected by IHC. As shown in the Fig. 3I, CTH also showed a similar trend as cysteine, which showed a gradually increase trend in OC, OKC and AM. Furthermore, the expression of CTH was positively correlated with the expression of CTSK (Fig. 3j). Thus, we hypothesized that the cysteine metabolism regulated by the CTH was the possible reason of different invasion ability of OC, OKC, and AM.

Fig. 3

WGCNA and differential analysis confirmed L-cysteine as the characteristic metabolite of OKC. Cluster tree analysis a and module analysis b among OC, OKC and AM. c PCA of AM and OKC. d Substitution test of AM and OKC. e Differential metabolite analysis of AM and OKC. f KEGG analysis of AM and OKC. L-cystathionine g and L-cysteine h analysis of OC, OKC and AM. i CTH expression of OC, OKC and AM. j Correlation analysis of CTH and CTSK. Data information: All data shown (g–i) represent the means _ SD (n ≥ 3 biological replicates). ANOVA followed by Tukey’s post hoc test (g–i) was used for the statistical analysis. (*P < 0.05; **P < 0.01; ***P < 0.001)

Fig. 4

L-cysteine induced ferroptosis tolerance is a key factor in destructive activities in odontogenic cysts/tumor. a KEGG map of L-cysteine pathway. b Western Blot of HaCaT after si-CTH and pathway metabolite difference after treatment of si-CTH. c GSEA and KEGG analysis of OKC and AM. d SLC7A11 expression of OC, OKC and AM tissues. e Correlation analysis of CTH and SLC7A11. EdU stain (f) and statistical chart (g) of HaCaT after treatment. GSH (h), MDA (i), intracellular ferrous ions (j) of HaCaT after treatment. Data information: All data shown (b, d, and f–j) represent the means _ SD (n ≥ 3 biological replicates). Student’s t-test (b) and ANOVA followed by Tukey’s post hoc test (d, and f–j) was used for the statistical analysis. (*P < 0.05; **P < 0.01; ***P < 0.001)

3.4 L-cysteine induced ferroptosis tolerance is a key factor in destructive activities of odontogenic cysts/tumor

According to the KEGG pathway, the hydrogen sulfide may be produced during the process of catalyzing cysteine to the pyruvate (Fig. 4a). As a signaling molecule, hydrogen sulfide may play a role in downstream biological events. Previous studies have shown that CTH have a strong relationship with ferroptosis [20]. Thus, we further explored the reason of jaw destruction caused by CTH. After the silence of CTH in HaCaT cells, the production of hydrogen sulfide and the pyruvate was reduced, and the cysteine was accumulated (Fig. 4b). Figure 4c showed a higher expression of genes related to reactive oxygen species (ROS) and an increased GSH metabolism in the OKC and AM tissues via GSEA and KEGG. These findings imply a higher oxidative homeostasis in OKC. Further, IHC results of OKC and AM tissues displayed elevated expressions of SLC7A11 (Fig. 4d), which were related to the glutathione metabolism and ferroptosis (Fig. 4e). Both the applications of erastine and the silence of CTH in HaCaT cells resulted in the low viability of cells (Fig. 4f, g). Also, it was found that the GSH was increased when the erastine was applied and CTH was silenced (Fig. 4h). And lipid peroxidation, a hallmark of ferroptosis, was increased in cells (Fig. 4i). At the same time, the intracellular ferrous iron was accumulated (Fig. 4j). After the silence of CTH, the cells became more susceptible to the erastine (Fig. 4h–j). Next, we verified the role of cysteine in ferroptosis. As expected, exogenous L-cysteine could rescue the viability of cells and intracellular levels of GSH, lipid peroxidation and ferrous iron induced by erastin (Fig. 4h–j). To confirm that the cell death caused by CTH or the cysteine metabolism was ferroptosis, the cell apoptosis assay was conducted, the results showed that the cell death induced by absence of CTH was not apoptosis (Supplementary Fig. 5d), necrosis and autophagy (Supplementary Fig. 5e and f). Our findings confirmed that the demise of HaCaT cell line induced by CTH silencing was due to ferroptosis. In conclusion, cysteine metabolism regulated by the high expression of CTH might efficiently promote the resistance of ferroptosis in OKC and AM.

3.5 NF-κB pathway is involved in L-cysteine-induced ferroptosis tolerance

At first, we found that the expression level of NF-κB was significantly increased in OKC and AM (Fig. 5a). There is also a certain correlation between the content of L-cysteine and the expression of NF-κB (Fig. 5b). The immunofluorescence staining displayed that the expression and nuclear translocation of NF-κB were increased after treated by L-cysteine (Fig. 5c, d). But after silencing CTH, the expression of NF-κB and the proportion of cells with NF-κB nuclear translocation decreased (Fig. 5c, d). Besides, a previous study has shown that the activation of NF-κB can regulate the expression of CTSK [21]. To further investigated the relationship between CTH and bone resorption, the expression of CTH in HaCaT cells was knocked down. And after that, the expressions of CTSK and MMP9 were reduced (Fig. 5e). All in all, the results showed that the mechanism of invasion in OKC and AM might be attributed to the activation of NF-κB via the high expression of CTH.

Fig. 5

L-cysteine induced ferroptosis tolerance is associated with NF-κb pathway. (a) NF-κB expression among OC, OKC and AM. (b) Correlation analysis of NF-κB and CTH. Fluorescence (c) and statistical graphs (d) Nuclear localization of NF-κB. Correlation analysis of NF-κB and CTSK (e) and MMP9 (f). (g) TRAP of HaCaT after si-CTH. (h) NSF-1 expression among OC, OKC and AM. (i) Correlation analysis of NFS-1 and CTH. Data information: All data shown (a, c, d, g and h) represent the means _ SD (n ≥ 3 biological replicates). Student’s t-test (g) and ANOVA followed by Tukey’s post hoc test (a, c, d and h) was used for the statistical analysis. (*P < 0.05; **P < 0.01; ***P < 0.001)

K. Suzuki et al. has reported that hydrogen sulfide plays an important role in the regulation of ferroptosis [10]. Moreover, we found with knockdown of CTH, the expression of nitrogen fixation 1 homolog (NFS-1) was decreased (Fig. 5f). As previous studies have shown, NFS-1, which can be activated by hydrogen sulfide, could help with biogenesis of intracellular Fe-S cluster in order to consume iron ion in the free state [22, 23]. In our research, the expression of NFS-1 was increased in the OKC and AM (Fig. 5g), and there was a positive correlation between the expression of CTH and NFS-1 (Fig. 5h, i). The increase of NFS-1 regulated by CTH might be an in-depth reason for the tolerance of ferroptosis in OKC and AM. In conclusion, we demonstrated that the production of hydrogen sulfide catalyzed by CTH was the main reason for the tolerance of ferroptosis and osteoclast ability in OKC and AM.