Landscape of the oncogenic role of fatty acid synthase in human tumors

Gene mapping and protein structure analysis

Based on the UCSC genome browser on human Dec. 2013 (GRCh38/hg38) assembly [14], information related to the genome location of the FASN gene was obtained. Further, the “HomoloGene” function of the NCBI was used to conduct conserved functional domain analysis of the FASN protein in different species. The phylogenetic status across diverse types of species were evaluated with the “COBALT” function of NCBI.

Gene expression analysis of TIMER2

We obtained the difference in FASN expression using the “Gene_DE” module of tumor immune estimation resource, version 2 (TIMER2). In tumors without contrast tissues or enough contrast group, the Gene Expression Profiling Interactive Analysis, version 2 (GEPIA2) web server [15] was utilized from the Genotype-Tissue Expression (GTEx) (P = 0.01, log2 fold change = 1).

We evaluated violin plots of FASN in various pathological stages across tumors using the “Pathological Stage Plot” module (GEPIA2). The log2 [TPM (Transcripts per million) +1] transformed expression results with a log-scale test were employed in the plots.

The Clinical Proteomic Tumor Analysis Consortium module [16] was used to determine the expression level of the total protein or phosphoprotein (S207, S724, S725, T976, S1174, S1411, S2198, and T2204 sites) of FASN (NP_004595.4) between cancers and corresponding tissue, respectively. Six tumors were screened: breast cancer, ovarian cancer, colon cancer, clear cell renal cell carcinoma (RCC), uterine corpus endometrial carcinoma (UCEC), and lung adenocarcinoma (LUAD).

Gene expression analysis using human protein atlas

The Human Protein Atlas (HPA) database was utilized to obtain the expression data of FASN in different cells, tissues, and plasma. The data for plasma samples were estimated via mass spectrometry-based proteomics in the HPA database. “Low specificity” was defined by “Normalized expression ≥1 in at least one tissue/cell type, but not elevated in any tissue/cell type.”

Immunohistochemistry (IHC) images of FASN in five pairs of normal and tumor tissues, including breast invasive carcinoma (BRCA), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), colon adenocarcinoma (COAD), liver hepatocellular carcinoma (LIHC), and prostate adenocarcinoma (PRAD) were downloaded from the HPA and analyzed.

Gene expression analysis using Oncomine

We collected the different expression results of FASN between the tumor group and the corresponding normal group (P = 0.05, fold change = 1.5). Pooling analysis was performed across five comparisons. The median rank for FASN across each of the analyses, the P-value for the median-ranked analysis, and the legends of the enrolled studies were obtained.

Survival prognosis analysis

We utilized the Kaplan-Meier “Survival Map” module (GEPIA2) [15] to obtain the overall survival (OS) and disease-free survival (DFS) map of FASN. The expression threshold for splitting the high/low expression groups was 50%. The Kaplan-Meier survival analysis module of GEPIA2 was used to obtain survival plots using the log-rank test.

The interface of the Kaplan-Meier plotter was used to pool the different GEO datasets for a series of analyses of OS, distant metastasis-free survival (DMFS), relapse-free survival (RFS), post-progression survival (PPS), first progression (FP), disease-specific survival (DSS), and progression-free survival (PFS). The cases of lung, ovarian, lung, gastric, and liver cancers were split into two groups by setting “autoselect best cutoff.” The hazard ratio (HR), 95% confidence intervals, and log-rank P-value were calculated, and Kaplan-Meier survival plots were generated. Clinical factors (e.g., histology, sex, smoking history, or chemotherapy) were also selected for a series of subgroup analyses.

Based on STATA 15.1 software (StataCorp LLC, College Station, TX, USA), we used the Kaplan-Meier function to perform a meta-analysis to pool the survival data of FASN.

Genetic alteration analysis

We collected the genetic alteration features, alteration frequency, mutation type, copy number alteration (CNA), mutated site information, and three-dimensional (3D) structure from the cBioPortal web [17, 18]. The OS, DFS, and PFS for the tumors with or without FASN genetic alterations were collected. Log-rank P-value was obtained and Kaplan-Meier analysis was performed.

Correlation of FASN and tumor mutational burden (TMB)/microsatellite instability (MSI)

TMB and MSI were obtained from the article, The Immune Landscape of Cancer and the Landscape of Microsatellite Instability Across 39 Cancer Types. The rank sum test detected two sets of data (p = 0.05). Spearman correlation was used to compare TMB/MSI and FASN gene expression.

DNA methylation analysis

We utilized the MEXPRESS web with the query “FASN” to obtain the DNA methylation status of FASN of various probes (e.g., cg06234966, cg24715260) in glioblastoma (GBM). The beta value of each sample, and the Benjamini-Hochberg-adjusted P-value and Pearson correlation coefficient (R) values were obtained.

We used the GSE50923 dataset [19] to assess the methylation status in 54 GBM tissues and 24 normal brain tissues. Briefly, the “minfi” R package and boxplot were employed to perform normalization. The mean methylation level of FASN and the normalized beta value of each sample at the selected methylation probes (cg03386722 and cg23244421) were visualized using the ggviolin function of the ggpubr package, with the setting, stat_compare_means (paired = T, method = “wilcox.test”).

Phosphorylation feature prediction

We used the open-access PhosphoNET database to obtain the predicted phosphorylation features of the S207, S724, S725, T976, S1174, S1411, S2198, and T2204 sites by searching the protein name “FASN.”

Immune infiltration analysis

The TIMER2 web server was used to explore the association between FASN expression and immune infiltrates. Immune cells of CD8⁺ T-cells, CD4⁺ T-cells, cancer-associated fibroblasts, and NK cells were screened. Purity-adjusted Spearman’s rank correlation test was used to obtain P-values and partial correlation (cor) values.

FASN-related gene enrichment analysis

The STRING website was utilized to obtain FASN-binding proteins, as demonstrated by experiments. The low confidence score was set as 0.150. Further, maximum number of interactors ≤50, physical interaction, and evidence were the basis of the interaction types.

The top 100 FASN-correlated targeting genes were obtained using the “Similar Gene Detection” module (GEPIA2). The pairwise gene correlation of FASN and the selected genes was obtained using the “correlation analysis” module. Moreover, a heatmap of the screened genes was obtained with the “Gene_Corr” module. Purity-adjusted Spearman’s rank correlation test was used to obtain P-values and partial correlation (cor) values.

We conducted Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis with the two sets of data. Database for Annotation, Visualization, and Integrated Discovery (DAVID) was utilized to acquire the functional annotation results. The enriched pathways were finally visualized with the “tidyr” and “ggplot2” R packages. Gene ontology (GO) enrichment analysis was also conducted with the “clusterProfiler" R package (Two-tailed P < 0.05).

Data availability

The datasets analyzed in this study can be found in the online dataset. Requests for further access to the dataset can be directed to hxl950513@126.com.

Highlights

A pan-cancer analysis of FASN. FASN is differentially associated with the prognosis of different tumor cases. The link between FASN and CD8⁺ T-cells, CD4⁺ T-cells, cancer-associated fibroblast infiltration, and NK cells. Enhanced phosphorylation level of S207 in several tumors, such as colon cancer. Hormone receptor binding- and fatty acid metabolic process-associated issues are involved in the cancerous role of FASN.

Gene mapping and protein structure analysis

In this study, we investigated the oncogenic function of FASN (NM_004104 or NP_004595.4, Supplementary Figure 1A). The protein structure of FASN is conserved among various species (e.g., H. sapiens, G. gallus, C. elegans) and generally involves the cond_enzymes (cl09938) domain and acyl_transf_1 (cl08282) domain (Supplementary Figure 1B). A phylogenetic tree (Supplementary Figure 2) revealed the evolutionary relationship of the FASN protein across various species.

Gene expression analysis

The expression level of FASN in various cells/nontumor tissues/plasma was determined. FASN had the highest expression in the adipose tissue, followed by the breast and liver (Figure 2A). FASN was identified in all tissues, except granulocytes, and displayed tissue-enhanced (adipose tissue) RNA tissue specificity. Low RNA blood cell type specificity was demonstrated in various blood cells (Figure 2B). The protein density of FASN in the plasma was 1.6 μg/L, which might be due to the biological external leakage of the intracellular compound (Figure 2C).

We analyzed the expression status of FASN in cancers. The expression level of FASN in the tissues of bladder urothelial carcinoma (BLCA), COAD, LIHC, PRAD, READ (Rectum adenocarcinoma), Stomach adenocarcinoma (STAD), Uterine Corpus Endometrial Carcinoma (UCEC) (P < 0.001), CESC, esophageal carcinoma (ESCA), head and neck squamous cell carcinoma (HNSC), kidney renal papillary cell carcinoma (KIRP) (P < 0.01), and kidney renal clear cell carcinoma (KIRC) (P < 0.05) was greater than that in comparable tissues (Figure 3A).

By incorporating the GTEx dataset, the expression level between the tumor-normal pair tissues was further identified in BLCA, CESC, COAD, lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), LIHC, ovarian serous cystadenocarcinoma (OV), pancreatic adenocarcinoma (PAAD), PRAD, READ, testicular germ cell tumors (TGCT), thymoma (THYM), UCEC, and uterine carcinosarcoma (UCS) (Figure 3B, P < 0.05). However, a significant difference was not found for other tumors, such as ESCA, HNSC, LGG (brain lower grade glioma), or SARC (sarcoma) (Figure 4A).

Increased FASN expression was found in colon cancer and UCEC (Figure 3C, P < 0.001) relative to comparable tissues from the CPTAC dataset. The results of the pooling analysis (Figure 5) further affirmed that FASN expression was higher in bladder cancer, colorectal cancer, lymphoma, ovarian cancer, and prostate cancer tissues (P < 0.001) than in corresponding normal controls. A correlation was also found between FASN expression and the pathological stages of CESC, KICH, KIRC, KIRP, OV, READ, and thyroid carcinoma (THCA) (Figure 3D, P < 0.05); however, no correlation was found for other tumors (Figure 4B–4E).

Immunohistochemistry analysis

The IHC results were obtained and the results of FASN gene expression in tumor tissues were compared with those in normal tissues. The gene expression data were found to be consistent with those of IHC. Negative or low IHC staining was obtained in the normal breast, cervix, colon, liver, and prostate tissue, while medium or strong staining was obtained in the cancer tissue (Figure 6).

Survival analysis data

We explored the relationship between FASN expression and the outcome of patients in the high and low expression groups. As shown in Figure 7A, high expression of FASN was related to poor OS outcomes in ACC (p = 0.015), BLCA (p = 0.0044), CESC (p = 0.004), HNSC (p = 0.045), KIRC (p = 3.9e-06), KIRP (P = 0.017), MESO (Mesothelioma) (P = 0.0043), LGG (P = 0.0035), and SARC (P = 0.018). DFS analysis (Figure 7B) further revealed a relationship between high FASN expression and poor outcome of ACC (p = 0.015), CESC (p = 0.023), LGG (p = 0.028), and KIRP (p = 0.02).

The Kaplan-Meier plotter tool revealed that FASN was associated with poor RFS (P = 0.0093) outcome for ovarian cancer (Figure 8A) and poor OS (P = 1e-08), FP (P = 1.2e-06), and PPS (P = 0.019) prognosis for lung cancer (Figure 8B). Additionally, a high FASN level was found to have a significant effect on the poor OS (P = 1.6e-16), FP (P = 1.8e-13), and PPS (P < 1e-16) outcomes for gastric cancer (Figure 8C). In contrast, a significant relationship between low expression of FASN and poor OS (P = 0.018), poor RFS (P = 3.5e-06), poor DMFS (P = 5.7e-05), and poor PPS (P = 0.022) outcomes was found for breast cancer (Figure 8D). In fact, a high expression of FASN was associated with poor OS, RFS, DMFS, and PPS (Supplementary Table 1, P < 0.05) for breast cancer patients with HER2 negative. Although we carried out further studies of liver cancer, a significant relationship between FASN and OS, PFS, RFS, and DSS outcomes was not found (Figure 8E, P < 0.05). Nonetheless, the results of the meta-analysis (Figure 9) confirmed the significant relationship between FASN expression and the outcomes for breast cancer, liver cancer, lung cancer, and gastric cancer (P < 0.05) but not for ovarian cancer (P = 0.998). Subgroup analyses were also performed with the clinical characteristics and definite results were obtained (Supplementary Tables 1–5). Overall, the findings revealed that FASN is dissimilarly associated with the outcome of cancer cases.

Genetic alteration analysis

The genetic alteration level of FASN has been observed in various cancers. The highest alteration frequency of FASN (>11%) was found in patients with Skin Cutaneous Melanoma with “mutation” (Figure 10A). The CNA data of liver HCC showed a >5% alteration incidence (Figure 10A). However, the value in thymoma cases (~1% genetic alteration frequency) was not detected (Figure 10A). Figure 10B shows the types, sites, and number of FASN genetic alterations. The missense mutation of FASN is the principal type of genetic alteration. L1353Sfs*20 alteration (four cases of uterine endometrioid carcinoma and one case of stomach adenocarcinoma) between Methyltransf_12 and ADH_zinc_N could induce a translation mutation from L (leucine) to S (serine) at the 1353 site of the FASN protein (Figure 10B). The L1353 locus was identified in the 3D protein structure (Figure 10C). Moreover, the relationship between genetic alterations and clinical outcomes of cancers was investigated.

Figure 10D shows that MESO patients with altered FASN had better outcomes for OS (P = 0.0248), but not for PFS (P = 0.209) while STAD patients had better OS (P = 0.602) and PFS (P = 0.392) than patients without FASN alteration.

TMB/MSI analysis

The correlation between FASN expression and TMB/MSI was determined. An adverse significant association was found between FASN and TMB for STAD, HNSC, UCEC, LUAD, LIHC, KICH, READ, THYM, PRAD, ACC, SARC, and BLCA, while a significant association was found for BRCA, LGG, and THCA (Figure 11). With respect to MSI, DLBC, READ, PCPG, THCA, HNSC, and BRCA were found to have a significant negative correlation, while CESC, LUAD, UCEC, KIRC, SARC (Sarcoma), GBM, TGCT, STAD, UVM, and LUSC had a significant positive correlation (Figure 12). More studies are required to explain this phenomenon.

DNA methylation analysis

We investigated the correlation between FASN DNA methylation and tumorigenesis in various cancers. With respect to GBM, we obtained a significant adverse relationship between DNA methylation and gene expression at probes of the non-promoter region, such as cg06234966 (P < 0.01, R = 0.322) (Figure 13). The GSE50923 dataset was used to confirm these results. In light of the standard methylation result (Figure 14A), we neglected the difference between GBM tissues and normal tissues (Figure 14B, P = 2.06e-06). Furthermore, we observed a decrease in the methylation level of FASN in cancer tissue compared to normal tissue for the selected probes (Figure 14C, cg03386722, P = 1.34e-06, Figure 14D, cg23244421, P = 0.03).

Protein phosphorylation analysis

FASN phosphorylation levels between normal tissues and the corresponding tumor group were obtained, and six types of cancers (breast cancer, clear cell RCC, LUAD, ovarian cancer, colon cancer, and UCEC) were analyzed. Figure 15A summarizes the FASN phosphorylation sites among these cancers.

The S207 sites within the ketoacyl-synt domain of FASN showed significant difference phosphorylation levels in tumor tissue compared to normal tissue (Figure 15B–15D, Figure 15F–15H, all P < 0.05), followed by the T2204 sites between PP-binding and thioesterase domain for colon cancer (Figure 15B, P = 5.8e-10), UCEC (Figure 15C, P = 8.0e-04) and breast cancer (Figure 15E, P = 1.9e-03). The PhosphoNET database revealed that FASN phosphorylation of S207 in the molecular responses to pathway activation was experimentally supported by previous investigations [20, 21] (Supplementary Table 6). Such finding aided in the exploration of the cancer role of S207 phosphorylation in tumors.

Immune infiltration analysis

Tumor-infiltrating immune cells can enhance the development, progression, or metastasis of cancers [22, 23]. A statistically significant adverse relationship was found between CD8⁺ T cells and FASN expression in HNSC, KIRC, OV, and SARC (Figure 16), depending on all/most algorithms. Further, a statistically significant relationship was presumed between FASN expression and cancer-associated fibroblasts for CESC, KIRC, KIRP, OV, and UVM; however, a negative correlation was presumed for LGG, PRAD, and STAD (Figure 17). A negative correlation was also found between FASN expression and CD4+ T cells and NK cells in BLCA, BRCA, and THCA, while a positive correlation was found in HNSC (Supplementary Table 7).

Enrichment analysis of FASN-related partners

We screened out FASN-binding proteins and FASN-expression-correlated genes for functional enrichment analyses. We collected 51 FASN-binding proteins using the STRING tool (Figure 18A). FASN expression was positively associated with five genes. The top 100 FASN-correlated targeting genes included glycerol-3-phosphate acyltransferase, mitochondrial (GPAM) (R = 0.64), stearoyl-CoA desaturase (SCD) (R = 0.62), cell death inducing DFFA-like effector c (CIDEC) (R = 0.61), diacylglycerol O-acyltransferase 2 (DGAT2) (R = 0.58), and pyruvate carboxylase (PC) (R = 0.32) (P < 0.001) (Figure 18B). The heatmap revealed a positive correlation between FASN and the five genes across cancer types (Figure 18C). PC was the only intersection member of the two groups (Figure 18D).

The KEGG analysis resulted presented in Figure 18E revealed that “metabolic pathways” and “PPAR signaling pathway” might play a role in the effect of FASN on tumorigenesis. GO analysis further suggested that these genes were mainly enriched in lipid metabolism, such as triglyceride lipase activity, lipid droplets, fatty acid metabolic processes, regulation of lipid metabolic processes, and others (Figures 18F and 19).