Elevation of fatty acid desaturase 2 in esophageal adenocarcinoma increases polyunsaturated lipids and may exacerbate bile acid-induced DNA damage

Abstract

Background: The risk of esophageal adenocarcinoma (EAC) is associated with gastro-esophageal reflux disease (GERD) and obesity. Lipid metabolism-targeted therapies decrease the risk of progressing from Barrett's esophagus (BE) to EAC, but the precise lipid metabolic changes and their roles in genotoxicity during EAC development are yet to be established.

Methods: Esophageal biopsies from the normal epithelium (NE), BE, and EAC, were analyzed using concurrent lipidomics and proteomics (n = 30) followed by orthogonal validation on independent samples using RNAseq transcriptomics (n = 22) and immunohistochemistry (IHC, n = 80). The EAC cell line FLO-1 was treated with FADS2 selective inhibitor SC26196, and/or bile acid cocktail, followed by immunofluorescence staining for γH2AX.

Results: Metabolism-focused Reactome analysis of the proteomics data revealed enrichment of fatty acid metabolism, ketone body metabolism, and biosynthesis of specialized pro-resolving mediators in EAC pathogenesis. Lipidomics revealed progressive alterations (NE-BE-EAC) in glycerophospholipid synthesis with decreasing triglycerides and increasing phosphatidylcholine and phosphatidylethanolamine, and sphingolipid synthesis with decreasing dihydroceramide and increasing ceramides. Furthermore, a progressive increase in lipids with C20 fatty acids and polyunsaturated lipids with ≥4 double bonds were also observed. Integration with transcriptome data identified candidate enzymes for IHC validation: Δ4-Desaturase, Sphingolipid 1 (DEGS1) which desaturates dihydroceramide to ceramide, and Δ5 and Δ6-Desaturases (fatty acid desaturases, FADS1 and FADS2), responsible for polyunsaturation. All three enzymes showed significant increases from BE through dysplasia to EAC, but transcript levels of DEGS1 were decreased suggesting post-translational regulation. Finally, the FADS2 selective inhibitor SC26196 significantly reduced polyunsaturated lipids with three and four double bonds and reduced bile acid-induced DNA double-strand breaks in FLO-1 cells in vitro.

Conclusions: Integrated multiomics revealed sphingolipid and phospholipid metabolism rewiring during EAC development. FADS2 inhibition and reduction of the high polyunsaturated lipids effectively protected EAC cells from bile acid-induced DNA damage in vitro, potentially through reduced lipid peroxidation.

Keywords: Barrett's esophagus; FADS2; esophageal adenocarcinoma; lipid desaturation; lipid metabolism.

FIGURE 1

Workflow for patient sample multiomic analyses and immunohistochemistry validation. Three different cohorts were used for multiomics analyses and validation of selected proteins. Concurrent lipidomics and proteomics were conducted on 10 biopsies each from the normal epithelium (NE), Barrett's esophagus (BE), and esophageal adenocarcinoma (EAC). Data from three samples were excluded from further analyses after quality control by tissue morphology or principal component analysis (PCA, Figure S1). The second set of biopsies was used for RNAseq transcriptomic analysis and used to support pathway interpretation. Finally, tissue expression levels of selected proteins in lipid metabolism were validated by immunohistochemistry conducted on a tissue microarray generated from 84 archival esophageal biopsies. MTBE, Methyl tert‐butyl ether; DDA, data‐dependent acquisition; IHC, immunohistochemistry; JIVE, Joint and Individual Variation Explained.

FIGURE 2

Overview of omics data . Principal component analysis (PCA) of patient biopsy lipidomics data (n = 586 lipids) (A) and proteomics data (n = 3387 proteins) (B) after exclusions revealed distinct clusters. Fuzzy C‐means clustering was used to divide the lipid and protein datasets into three clusters; L1–L3 for lipids (C) and P1–P3 for proteins (D), demonstrating a progressive change from the normal esophagus (NE, n = 10) through Barrett's esophagus (BE, n = 8) to esophageal adenocarcinoma (EAC, n = 9). (E) Joint and Individual Variation Explained (JIVE) analysis of the lipidomics and proteomics data revealed a single significant component (PC1) which divides the samples into the progression continuum of NE, BE, EAC shown in a bar graph. The proteins and lipids in PC1 are shown in Table S3, and the top JIVE components are shown in (F).

FIGURE 3

Metabolism‐focussed pathway enrichment analysis of the tissue proteomics data using Reactome. Larger circles indicate more significant enrichment, color indicates normalized enrichment score (NES). Names of significant (p < .05) gene sets are highlighted in bold

FIGURE 4

Overview of lipidomics data using lipid set enrichment analysis. (A) Lipid class enrichment comparing normal esophagus (NE, n = 10) with Barrett's esophagus (BE, n = 8) and esophageal adenocarcinoma (EAC, n = 9). Lipid set enrichment analysis was performed on log2 fold‐changes and data was centered and scaled prior to visualization. (B) Individual results for lipid classes altered in EAC compared to NE. Lipids belonging to the corresponding lipid class are highlighted in blue. (C) Unsaturation enrichment for lipids with two fatty acids (including Cer‐NDS, Cer‐NS, PC, PG, ether‐PC, ether‐PE, PS, SM FAHFA, PE, PI) demonstrates an increase in lipids with unsaturation of 4 and 5, and a decrease in saturated lipids (D) Chain length analysis for lipids with two fatty acids. NES: normalized enrichment score. *p.adj < .05

FIGURE 5

Integrated multiomic data highlight altered lipid desaturation and antioxidant pathways during progression from the normal esophagus (NE) to Barrett's esophagus (BE) to esophageal adenocarcinoma (EAC). The proteomic, lipidomic, and transcriptomic data from independent cohorts are shown with heatmaps representing the median of the mean‐centered data

FIGURE 6

Ceramide desaturation is altered in esophageal adenocarcinoma (EAC) progression. (A) Lipidomics data highlight a progressive increase in unsaturated ceramide species in EAC. (B) Changes in sphinganine, DHCer, Cer, and sphingosine appear to be unrelated to changes in DEGS1 and DEGS2 transcript levels (C) across normal epithelium (NE), Barrett's esophagus (BE), and esophageal adenocarcinoma (EAC) tissues. Data are represented using boxplots, with Mann‐Whitney U test or one‐way ANOVA with Benjamini‐Hochberg adjustment, with significance (p < .05) from NE represented by shading in the boxplots. Lipid data: NE n = 10, BE n = 8, EAC n = 9. Transcript data: NE = 7, BE = 9, EAC n = 8. (D) Biopsies were stained and scored for DEGS1 protein on a 4‐point scale (0‐3): with representative areas to highlight normal squamous epithelium (NSE), non‐dysplastic Barrett's esophagus (BE), low‐grade dysplasia (LGD), high‐grade dysplasia (HGD), HGD with intraepithelial carcinoma (HGD + IEC) in 1.5 mm cores indicated with black shapes. Scoring of the immunohistochemistry data is shown in (E) (n in brackets). Ordinal logistic fit provided significant changes between BE and LGD (p = .002)

FIGURE 7

Lipid unsaturation changes with esophageal adenocarcinoma (EAC) progression, linked to desaturase proteins. (A) RNA expression of enzymes involved in lipid desaturation, taken from an independent cohort of patient biopsies representing normal esophagus (NE, n = 7), Barrett's esophagus (BE, n = 9), and esophageal adenocarcinoma (EAC, n = 8). Data are represented using boxplots, with Mann‐Whitney U test or one‐way ANOVA with Benjamini‐Hochberg adjustment, with significance (p < .05) from NE represented by shading in the boxplots. (B) Scoring of immunohistochemistry for FADS1 and FADS2 was performed on a 4‐point scale (0–3) with representative areas (indicated with black shapes) to highlight normal squamous epithelium (NSE), non‐dysplastic Barrett's esophagus (BE), low‐grade dysplasia (LGD), high‐grade dysplasia (HGD), HGD with intraepithelial carcinoma (HGD + IEC) in 1.5 mm cores Pooled scores (n shown in brackets) are shown in (C). Effect likelihood ratio tests provided p values of < .0001 for both FADS1 and FADS2. Ordinal logistic fit provided no significant sequential changes during disease progression for FADS1, and significant changes between NSE and BE (p = .024), and LGD and HGD for FADS2 (p = .042)

FIGURE 8

Fatty acid desaturases (FADS) inhibition alters lipid profile and DNA damage in FLO‐1 EAC cells. FLO‐1 EAC cell line was treated with FADS2 inhibitor SC26196 for 48 hours before treatment with a bile acid cocktail (BAC, 1000 μM final, including an equimolar mixture of sodium salts of taurocholic acid, deoxycholic acid, glycodeoxycholic acid, glycocholic acid, and glycochenodeoxycholic acid) at pH 4 for 20 min, camptothecin (CPT, 1 μM, pH 7) for 60 min, or left untreated. (A) Dose titration of SC26196 for reduction of lipids with three or four double bonds. 500 nM SC26196 provides 80% of maximal response. (B) 500 nM SC26196 decreases lipids with three or four double bonds in lipids with two chains compared to vehicle in FLO‐1 cells, and (C) in lipids with two or four double bonds in single‐chain lipids. (D) Changes in lipid class were also induced by 48 h exposure to 500 nM SC26196. *, p < .05 change from control. (E) Immunofluorescence staining of γH2AX foci (green) to visualize dsDNA damage, with the nuclei stained blue (DAPI). (F) The mean foci per cell was increased with BAC and 1 μM CPT, but SC26196 treatment decreases the mean foci per cell in BAC and vehicle. (G) Similarly, the % of cells showing more than five foci per cell was increased by BAC and CPT, while FADS inhibition by SC26196 decreased the number of positive cells in vehicle and BAC. Data from biological quadruplicates and technical duplicates. ***, p = .001