Tryptophan metabolism-related gene expression patterns: unveiling prognostic insights and immune landscapes in uveal melanoma

Study population and data collection

UVM mRNA expression profiles were explored in a cohort of 80 patients, with data sourced from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/) and the Gene Expression Omnibus (GEO, GSE22138). Throughout this research, strict adherence to relevant guidelines ensured methodological rigor. Duplicates were removed, and cases lacking clinical outcomes were excluded (processed datasets were shown in Supplementary File 1). For data processing, the R programming language was employed, effectively translating Ensembl ID numbers into gene symbols.

Defining tryptophan metabolism-related genes

In our exploration of TMRGs (Supplementary File 2), we focused on those associated with the “KEGG TRYPTOPHAN METABOLISM” pathway, drawing data from The Molecular Signatures Database (MSigDB) (http://software.broadinstitute.org/gsea/msigdb/index.jsp).

Alterations of TMRGs in UVM

TMRG modifications in the TCGA database, which include changes in gene expression, methylation, and copy number variations (CNVs), were analyzed using the Gene Set Cancer Analysis platform (http://bioinfo.life.hust.edu.cn/GSCA/#/). Our study assessed the impact of these TMRG alterations on disease-free survival (DSS), overall survival (OS), and progression-free survival (PFS) in patients.

Identification of subtypes and biological function enrichment

Using consensus unsupervised clustering analysis, patients were classified into specific molecular subgroups based on the expression patterns of 40 TRMGs. We then assessed the differences in prognosis and survival rates among these subtypes, employing univariate Cox regression and Kaplan-Meier survival analyses with the aid of the ‘survival’ (3.5.7) and ‘survminer’ (0.4.9) R packages. Additionally, using the Gene set variation analysis (GSVA), we determined the variations in molecular subtype enrichment within biological processes, referencing the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) gene sets.

Establishment of the risk model

To construct a precise prognostic signature focusing on minimal RNA processing factors, we applied the least absolute shrinkage and selection operator (LASSO) penalty using R software’s “glmnet” (4.1.8) package. Ten-fold cross-validation determined the optimal penalty parameter (λ) and established regression coefficients for individual variables. The formula used was RiskScore = Σβi × Expi, where i represents gene expression levels and β corresponds to the gene’s Cox regression coefficient. Hazard ratios emerged from both univariate and multivariate Cox regression analyses, visualized through forest plots generated by R’s “forestplot” (3.1.3) package.

Evaluation of the TRMGs prognostic model

Based on the median risk score, patients were classified into either high- or low-risk group. The efficacy of the model was assessed by comparing survival rates between these groups using Kaplan-Meier plots generated via R’s “survival” package. Differences in CRG expression between the groups were illustrated using a heatmap created with the “pheatmap” package. Time-dependent receiver ROC analysis, facilitated by the “survminer” and “timeROC” R packages, assessed the model’s specificity and sensitivity over 1-, 3-, and 5-year periods. The area under the ROC curve (AUC) indicated the model’s predictive accuracy. A nomogram for predicting UVM patient survival over 1-, 3-, and 5- year durations was formulated using the “rms” package, integrating factors like age, pathologic stage, T-stage, gender, and the risk score.

Analysis of immune capacity and drug susceptibility across risk categories

We employed CIBERSORTx to calculate the tumor infiltrating immune cells (Supplementary File 3), and then evaluate correlations with risk scores. The research also identified differential expressions of immune cell subsets between the high- and low-risk groups. To explore variations in drug sensitivity between the risk groups, we utilized the ‘pRRophetic’ (0.5) algorithm and ‘ggpubr’ (0.6.0) packages, computing IC50 values for commonly used immunotherapeutic drugs in cancer treatment.

Immune cell infiltration assessment

The gene markers were retrieved from Immune Cell Abundance Identifier (ImmuCellAI), which was shown in Supplementary File 4. We compared immune cell infiltration patterns across 24 immune cells (ImmuCellAI) between high- and low-risk groups. Grouping 143 UVM patients based on median risk scores, we analyzed data on the platform to understand the dynamics of immune cell abundance and infiltration.

Cell culture

UVM cell lines C918 and M619, sourced from ATCC (VA, USA), were cultivated in DMEM supplemented with 10% fetal bovine serum and incubated at 37°C with 5% CO2. Post 24-hour treatment with 5 μM N,N-Dimethyltryptamine (DMT), one group was established as the DMT group, while the other served as a control (NC group).

Cell viability assay

5 × 10³ cells were seeded onto a 96-well plate and allowed to adhere overnight in a medium containing 10% FBS. After the initial treatment, cell viability was gauged using the Cell Counting Kit-8. Post-DMT treatment, the cells were left to form colonies for seven days. Following this period, the cells were fixed, stained, and colony numbers were quantified using ImageJ software.

Colony formation assay

Following DMF treatment, cells were seeded in 6-well plates at a density of 2 × 10³ cells/well and incubated in DMEM supplemented with 20% FBS for seven days. After this period, cells were fixed using 70% ethanol and stained with a 0.25% crystal violet solution (Thermo-Fisher, PA, USA). Colonies were then quantified using ImageJ software.

Cell apoptosis

Using flow cytometry, apoptosis was analyzed with the Annexin V-FITC cell apoptosis detection kit (Beyotime, Shanghai, China). Detached cells were suspended in 1× binding buffer at a concentration of 1 × 10⁵ cells/mL. Following this, 5 μL of FITC Annexin-V and 5 μL of propidium iodide were added to 100 μL of the cell suspension, which was then incubated in the dark for 15 minutes. After incubation, 400 μL of 1× binding buffer was added. Apoptosis was analyzed using the FACS Calibur flow cytometer with Cell-Quest software (BD Cell Quest Pro Software, BD Biosciences, CA, USA).

Cell invasion assay

Transwell cell culture systems with 8 μM pores in the upper chamber membranes were used for cell invasion assays. The upper chamber membrane was coated with a mixture of Matrigel (Corning, NY, USA) and serum- free DMEM. 1 × 10⁴ cells in suspension were seeded into each upper chamber in serum-free medium, with or without 5 μM DTM. The lower chamber was filled with DMEM medium containing 20% FBS. After a 24-hour incubation at 37°C and 5% CO2, the cells were fixed with 4% formaldehyde and stained with 0.1% crystal violet. Non-invaded cells on the upper membrane were removed using a 70% ethanol-soaked cotton swab. Invaded cells on the lower side of the membrane were visualized and captured with an inverted microscope (Ti-S, Nikon, Tokyo, Japan). Using ImageJ software, the invaded cells were counted, and an average value was calculated from three randomly chosen fields of view, each at 100× magnification.

Scratch wound-healing migration assay

UVM cells were cultured on a six-well plate at a density of 5 × 10⁵ cells per well and incubated until they reached confluence. Using a 10 μL pipette tip, a linear scratch was made across the monolayer. The cells were then treated with either 0 or 12.5 nm of DMT. Images were captured at 0 and 24-hour intervals using an inverted light microscope (Leica DM IL, Wetzlar, Germany). The wound area was analyzed with ImageJ software, and the wound closure rate was calculated using the formula: (initial wound area - residual wound area)/initial wound area ×100%.

Statistical analysis

For comparisons between two groups, the Student’s t-test was employed. For multiple group comparisons, the Wilcoxon test was used. Survival differences were assessed using the log-rank test. A p-value of less than 0.05 was deemed statistically significant. The code script was shown in Supplementary File 5.

Data availability statement

The datasets presented in this study can be found in online repositories.

The difference in survival according to alteration of TMRGs in patients with UVM

In assessing the survival differences linked to TMRGs alterations in UVM patients, a thorough analysis of the mRNA expression of TMRGs in the UVM cohort from TCGA revealed a significant correlation. High expression levels of WARS1, KMO, and IDO1 were associated with shorter DSS, OS, and PFS. Conversely, low expression of OGDHL, HAAO, EHHADH, ALDH9A1, ALDH1B1, and AANAT yielded similar outcomes (Figure 1A). Additionally, an examination of TMRGs methylation levels classified UVM patients into two groups: those with hazard ratios exceeding 1 and those below. It was found that higher methylation of ALDH7A1 and EHHADH, and lower methylation of IDO2, MAOB, INMT, and DDC were indicative of unfavorable DSS, OS, and PFS outcomes (Figure 1B). A subsequent survival analysis using CNV data in UVM confirmed that CNV of TPH1, IDO2, IDO1, EHHADH, and CAT were also associated with poor DSS, OS, and PFS (Figure 1C).

Figure 1. The disparity in survival outcomes in the modifications of TMRGs among patients afflicted with UVM. (A) The distinction in survival outcomes between elevated and diminished TMRG expression profiles; (B) Survival difference between high and low methylation of TMRGs; (C) Survival difference between CNV groups of TMRGs. TMRGs, tryptophan metabolism-related genes. The color from blue to red represents the hazard ratio, size represents statistical significance in bubble plot. The black outline border indicates Cox P ≤ 0.05.

Establishment of Trp subtypes and analysis of biological pathways

To elucidate the role of TMRGs in UVM progression, we analyzed various genomic factors. We classified patients from TCGA and GSE22138 into two distinct clusters based on the consensus matrix (Figure 2A). Notably, cluster A exhibited a significantly prolonged OS compared to cluster B (Figure 2B). A GSVA in GO revealed enrichment in metabolic processes (such as NADH, ADP, ribonucleoside diphosphate, and nucleoside diphosphate) as well as vesicle and granule pathways in cluster A (Figure 2C). Similarly, a GSVA in KEGG showed that cluster A was rich in diverse pathways, including cancer (glioma and bladder cancer) and metabolic pathways (like linoleic acid, glyoxylate and dicarboxylate, inositol phosphate, and galactose) (Figure 2D). Delving deeper with ssGSEA, we observed greater immune cell infiltration in cluster A, characterized by the prevalence of immune cell types such as Activated. B.cell, Activated (Figure 2E). CD4.T.cell, Activated.CD8.T.cell, and others. Our findings suggest that TMRGs may play a pivotal role in UVM progression, potentially by regulating immune-related and metabolic processes and pathways.

Figure 2. Trp-related dual classifications and the enrichment of pertinent biological functions. (A) Harmonized matrix heatmap delineating a pair of distinct clusters; (B) Survival analysis among the two subtypes; P value was calculated based on Log-rank test; (C) Differences in the PCA analysis among the two clusters; (D) Disparities in expression levels of TMRGs related to clinical characteristics within the two unique subtypes; (E) Disparities in the immune infiltration between the two distinct clusters. Abbreviations: Trp: tryptophan metabolism; KEGG: Kyoto Encyclopedia of Genes and Genomes; GO: Gene Ontology.

Identification of gene subtypes

Through comparative analysis of the two Trp subtypes, we identified 994 differentially expressed genes (DEGs) at their intersection. Univariate Cox regression analysis was employed to assess the prognostic significance of these DEGs, leading to the selection of 284 specific genes. Leveraging the insights from the Trp subtypes, we utilized the consensus clustering algorithm to stratify patients into three genomic subtypes based on these 284 prognostic genes, which we termed gene subtypes A, B, and C (Figure 3A). Clinical analysis showed that cluster B had a higher expression of TMRGs compared to cluster C (Figure 3B). Survival analysis underscored the superior survival benefits for individuals in gene cluster A (Figure 3C). A comprehensive review of TMRGs expression levels across the gene subtypes revealed marked differences in genes such as AANAT, ACAT2, and ECHS1, among others (Figure 3D).

Figure 3. Construction of gene subtypes. (A) Harmonized matrix heatmap delineating a pair of distinct gene assemblages; (B) Disparate expression of TMRGs amid clinical characteristics within the duet of gene subclasses; (C) Survival curves for OS of the two gene subtypes; (D) Variations in the manifestation of 40 TMRGs between the pair of gene subclasses.

Construction of the prognostic model of TMRGs in UVM

We subjected UVM patients to LASSO and multivariate Cox regression analyses, identifying five key model genes with their associated coefficients (Figure 4A, 4B). The resulting formula was: risk score = (−0.168 × expression of ATP13A3) + (0.060 × expression of CHST11) + (0.218 × expression of TPST2) + (0.085 × expression of TAPBPL) + (0.045 × expression of MRPL24). Using this risk score, we stratified the patient cohort into high-risk and low-risk groups based on the median signature risk score. A Sankey diagram illustrated the distribution and characteristics of patients, highlighting the majority in the high-risk group who sadly succumbed (Figure 4C). Notably, the risk score distribution varied among TMG and gene clusters: gene cluster C had the highest risk score, while TMRGs cluster A was associated with a higher risk score (Figure 4D, 4E). Moreover, significant differences in TMRGs expression levels, including AADAT, ACAT2, and ECHS1 among others, emerged between the high- risk and low-risk groups (Figure 4F).

Figure 4. Construction of the prognostic model for TMRGs in UVM. (A) TMRGs profiles based on LASSO coefficients; (B) LASSO coefficient values of the TMGRs in UVM; (C) Sankey diagrams of different status, risk score, gene cluster and TMG-associated cluster; (D) Risk score in the three gene clusters; (E) Risk score in the two TMG-associated cluster. (F) Differences in the expression of 40 TMRGs between different risk subgroups.

The Kaplan-Meier survival curve demonstrated that the low-risk patient group had a significantly longer OS than the high-risk group (Figure 5A). Time-dependent ROC curves for 1-year, 3-year, and 5-year OS yielded AUCs of 0.724, 0.795, and 0.841, respectively (Figure 5B). A heatmap visualized the expression patterns of the five model genes across different risk score groups (Figure 5C). As risk scores increased, the distribution of survival status showed a rising mortality rate (Figure 5D, 5E). The calibration plot highlighted a strong alignment between observed and predicted outcomes (Figure 5F). Cox analysis results identified both gender and risk score as independent prognostic factors for UVM patients (Figure 5G). Collectively, these results underscore the reliability of the risk score as a prognostic tool for UVM.

Figure 5. The prognostic signature’s risk score and its correlation with survival outcomes within the UVM cohort. (A) Survival analysis of the different risk subgroups; (B) ROC curves illustrating the predictive accuracy of 1-, 3-, and 5-year survival rates based on varying risk scores; (C) Disparate expression of model genes in the different risk subgroups. (D) Distribution of the risk score in the different risk subgroups. (E) The survival status in the different risk subgroups; (F) Nomogram model for individual patient survival probabilities at 1-, 3-, and 5-year intervals for those diagnosed with UVM; (G) Calibration curves depicting the nomogram’s accuracy in relation to observed OS at 1-, 3-, and 5-year milestones.

Immune infiltration and drug susceptibility in UVM

Upon a thorough examination of immune scores, we discerned that the high-risk group demonstrated a pronounced increase in their stromal, immune, and ESTIMATE scores (Figure 6A). To further elucidate differences in immune infiltration, we compared the low- and high-risk groups (Figure 6B). Notably, high- risk patients showed a significant presence of certain immune cells, notably resting dendritic cells (Figure 6C) and M1 macrophages (Figure 6D). Conversely, the low-risk group was characterized by a higher abundance of T cells CD4 memory resting (Figure 6E), regulatory T cells (Tregs) (Figure 6F), naive B cells (Figure 6G), and plasma cells (Figure 6H).

Figure 6. Prognostic TMRGs intervene in immune infiltration in UVM. (A) Correlations between risk score and immune, stromal and ESTIMATE scores; (B) Heatmap of the immune cells, five differential gene and the risk score; (C–H) Correlation between the risk score and the abundance of immune cells, including (C) Dendritic cells resting, (D) Macrophages M1, (E) T cells CD4 memory resting, (F) T cells regulatory (Tregs), (G) B cells naive, (H) Plasma cells.

Our study also sought to assess the responsiveness of patients in different risk score groups to common immunotherapeutic drugs. Interestingly, the high-risk group predominantly showed lower IC50 values for several drugs, namely AMG.706, temsirolimus, SL.0101.1, PF.4708671, RO.3306, WZ.1.84, and X17.AAG (Figure 7A–7G). On the other hand, the low-risk group displayed lower IC50 values for vinorelbine, etoposide, and FH535 (Figure 7H–7J). From this comprehensive data analysis, we deduced that high-risk patients tend to have a heightened responsiveness to immunotherapy.

Figure 7. Drug sensitivity for TMRGs in UVM. (A–I) Correlations between the risk score and AMG. 706 (A), temsirolimus (B), SL.0101.1 (C), PF.4708671 (D), RO.3306 (E), WZ.1.84 (F), X17.AAG (G), Vinorelbine (H), Etoposide (I), and FH535 (J) sensitivity.

The effect of Trp on tumorigenicity of EC cell

In our efforts to understand the effects of Trp on the malignant behaviors of UVM cells, we utilized DMT as a Trp inhibitor and evaluated its impact on cell proliferation, apoptosis, migration, and invasion. Our results from the CCK-8 assay (Figure 8A) and colony formation assay (Figure 8B) indicated that inhibiting Trp effectively suppressed UVM cell proliferation. Flow cytometry analyses further revealed that Trp inhibition significantly enhanced apoptosis in UVM cells (Figure 8C). Additionally, we noted a marked reduction in cell migration distance upon Trp inhibition (Figure 8D). Moreover, the cells treated with the Trp inhibitor displayed a reduced propensity to invade Matrigel compared to the control cells (Figure 8E). In essence, Trp plays a pivotal role in the tumorigenicity of UVM cells.

Figure 8. The effect of TPM on UVM cells. (A, B) Cell proliferation detected using CCK-8 assay (A) and colony formation assay (B); (C) Cell apoptosis detected using flow cytometry; (D) Cell migration measured using wound healing test; (E) Cell invasive ability detected using Transwell assay. Abbreviation: TPM: tryptophan metabolism.