Role of TRAIL in mediating the effect of lipidome on breast cancer: a Mendelian randomization study

Introduction

Breast cancer (BC) is one of the most common cancers among women and becomes the leading cause of cancer-related mortality (1,2) (Figure 1A). Approximately 1 in 8 to 10 women will develop BC during their lifetime. The primary treatments for BC include surgery, endocrine therapy, chemotherapy, radiotherapy, and targeted therapy (3). With significant improvements in survival rates over the past two decades due to early detection and treatment, the 5-year survival rate for early-stage (i.e., stage I or II) BC can reach 90% (4,5). However, for patients diagnosed with metastatic disease (i.e., stage IV), the majority (60%) receive only radiotherapy and/or chemotherapy, with a 5-year survival rate of less than 30% (1,2,4-6). Meanwhile, the incidence of BC continues to rise globally. Within the European Union, the total cost of BC ranks second among all cancers, estimated at approximately 15 billion euros or 0.15% of gross domestic product (GDP) (7). The Lancet Oncology Commission reported that in 2010, BC was the most expensive cancer in the United States, costing 16.5 billion dollars, or 13% of all cancer-related expenditures (8,9). Therefore, it is crucial to explore the pathogenesis of BC, mine possible biomarkers, and seek new treatment methods.

Figure 1 Overview of the study on phosphatidylcholine (16:0_16:0), TRAIL, and breast cancer. (A) Incidence rates were obtained from the Global Cancer Observatory (GCO, IARC, 2022) and are presented as crude incidence per 100,000 population for each country or region. (B) Schematic of the molecular structure of phosphatidylcholine (16:0_16:0) and the domain architecture of TRAIL. (C) Study workflow illustrating the MR analysis of lipidomes and inflammatory factors in breast cancer. The diagram details the screening process, identification of key lipidomes and inflammatory markers, and the proposed mechanistic link between phosphatidylcholine (16:0_16:0), TRAIL, and breast cancer, further validated through molecular docking and molecular dynamics simulations. MR, Mendelian randomization; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand.

Lipidomes are a diverse class of non-polar or amphipathic biomolecules that are synthesized de novo or acquired from the surrounding microenvironment (10). An expanding body of observational and multi-omics evidence implicates dysregulated lipid metabolism and perturbations of the circulating lipidome in BC initiation, progression, and therapeutic response (10). Ferreri et al. (11) reported that alterations in the erythrocyte-membrane lipidome—characterized by lower saturated and higher ω-6 fatty acids—are associated with increased membrane fluidity and heightened systemic inflammation, changes that may facilitate BC initiation and progression. In complementary work, Tokareva et al. (12) showed that lipidomic remodeling correlates with multiple aggressive clinicopathologic features (tumor size, stage/grade, axillary nodal burden, Nottingham Prognostic Index, Ki-67): tumor tissue is enriched for phosphatidylcholine (PC) signatures, whereas peritumoral microenvironmental tissue exhibits shifts in oxidized lipids, sphingolipids, and phosphatidylethanolamines, supporting a role for lipid metabolic reprogramming in disease evolution and risk stratification. Taken together, these data substantiate a multidimensional relationship between the lipidome and BC biology while underscoring substantial heterogeneity across lipid classes, biospecimens, and study designs, with causal direction and mediating pathways yet to be fully resolved.

Furthermore, it is also not clear that there are intermediate factors in the causal relationship between lipidomes and BC. Previous observational studies (13,14) have identified certain inflammatory factors that potentially interact with BC. According to research conducted by Riera-Domingo and colleagues (15), the downregulation of tumor necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL) promotes BC metastasis. Consequently, TRAIL could serve as a plausible intermediary factor linking lipidomes to BC. As a member of the TNF superfamily, TRAIL induces apoptosis by binding its death receptors (DR) TRAIL-R1 (DR4) and TRAIL-R2 (DR5) (16-18), are reported to be upregulated in BC (19,20). Clinically, lower circulating TRAIL has been observed in patients compared with healthy controls—for example, Ibrahim et al. (21) discovered significantly lower serum TRAIL levels in patients compared to controls (P<0.001), with similar observations in renal cell carcinoma, metastatic melanoma, and chronic myeloid leukemia (22-24). Mechanistically, ligand engagement triggers formation of the death-inducing signaling complex via recruitment of Fas-associated death domain protein (FADD) and procaspase-8/-10, leading to activation of executioner caspases 3/6/7 and apoptosis (25-29). Emerging evidence indicates that phosphatidylcholine (PC) biosynthesis is upregulated in cancer cells through increased utilization of fatty acids and their derivatives (30,31). Independently, augmenting TRAIL may enhance engagement of DR4/DR5 on tumor cells and could promote apoptotic signaling, potentially limiting tumor progression and metastasis in certain contexts (32). Additionally, phosphatidylcholine can regulate programmed cell death as a signaling molecule (33). Taken together, these observations suggest a plausible intersection between lipid metabolic states and TRAIL pathway activity in BC, warranting further investigation. However, observational studies are vulnerable to confounding and reverse causation, and the causal direction and mediating mechanisms remain incompletely resolved.

Mendelian randomization (MR) assesses whether observational associations have causal effects (34). It utilizes genetic variations as instrumental variables (IVs) to determine the relationship between risk factors and outcomes (35,36). Compared to conventional observational studies, MR has the capacity to mitigate the impact of confounding factors and reverse causation bias on research findings (37). By emulating randomized controlled trials, MR not only overcomes the limitations of observational studies but also provides additional clinical evidence, thereby enhancing the credibility of causal inferences between lipidomes and BC (38,39).

Hence, in this study, a two-sample MR analysis was firstly performed to investigate the causal effect of lipidomes on BC. Next, we conducted a mediation analysis to explore the intermediary role of inflammatory factors in developing BC associated with lipidomes (Figure 1B,1C). And finally, molecular docking and molecular dynamics simulations were both used to explore the possible TRAIL-mediated carcinogenesis mechanism. We present this article in accordance with the STROBE-MR reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1096/rc).

Methods

Data sources

We used European ancestry data from the publicly available genome-wide association study (GWAS) database, selecting single nucleotide polymorphisms (SNPs) as IVs. For lipidomes IVs, the summary statistics from the GWAS catalog [codes GCST90277238-GCST90277416 (40)] were employed. The BC data, comprising 15,680 cases and 167,189 controls, were sourced from the FinnGen database (version R11, data release date: May 8, 2023; https://finngen.gitbook.io/documentation/).

We also utilized GWAS data from the Olink Target Inflammation panel, which measured 91 circulating inflammatory factors in 14,824 participants of European ancestry (41). Given that all sources are European-ancestry cohorts, extrapolation to non-European populations should be made with caution.

Participants

This MR study did not involve direct recruitment of participants. Instead, summary-level GWAS data were obtained from publicly available sources. BC data were derived from the FinnGen consortium (R11 release), including 15,680 cases and 167,189 controls, all of European ancestry. Data on circulating inflammatory biomarkers were obtained from GWAS summary statistics based on 14,824 individuals of European descent from the Olink Target Inflammation panel. As only publicly available aggregated data were used, no individual-level inclusion or exclusion criteria were applied. Power or sample size calculations were not conducted a priori, given the large sample sizes of the included GWAS datasets, which provided sufficient statistical power for the MR analyses.

IVs selection

The causal impact of exposure on outcomes was carried out using genetic variants as IVs. The approach adopted in this work adhered to the three core assumptions of MR: (I) Genetic variants must be significantly associated with the exposure to ensure validity of the IVs and robustness of the results. (II) The selected SNPs must be independent of any potential confounders and not directly related to the outcome variables. (III) The association between IVs and outcome variables should be fully mediated through the exposure factor, with no alternative pathways (i.e., no pleiotropy). Moreover, to minimize the influence of potential confounders, we systematically queried all candidate SNPs in the PhenoScanner database (http://www.phenoscanner.medschl.cam.ac.uk/). Variants associated with traits that could bias the lipid-BC relationship, such as sex, educational attainment, smoking behavior, body mass index, insulin resistance, age at menarche, alcohol consumption, family history of cancer, or other cancer diagnoses, were excluded. Specifically, rs10758669, rs10406080, rs6440013, rs754388, rs2049045, rs61739285, rs63750417 and rs492602 were removed due to above reasons.

To ensure the validity and reliability of the MR analysis, the following three steps were implemented: (I) We used the “Two-Sample-MR” package, setting the linkage disequilibrium threshold of R² less than 0.001 with a window size of 10,000 kb. (II) Due to sample size limitations, we applied a P value threshold less than 5×10⁻⁸. (III) We deemed an F-statistic greater than 10 to be sufficient to mitigate bias from weak IVs.

Statistical analysis

The MR analysis was conducted using R software (version 4.2.0, http://www.r-project.org) and the “Two-Sample-MR” package (version 0.5.6). MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO) analysis was performed with the MR-PRESSO package in R. Moreover, the odds ratios (OR) and their associated 95% confidence intervals (CI) were used to evaluate the causal relationships between variables.

Primary analysis

Via the inverse variance weighted (IVW) method, a meta-analysis was conducted on the Wald ratios between the selected SNPs associated with lipidomes and BC, aiming to estimate the internal causal effect (42). Additionally, we supplemented the weighted median to address outliers or pleiotropy in IVs, along with the MR-Egger regression method, the simple mode method, and the weighted mode method to ensure the consistency and robustness for MR results (43). Furthermore, the Bayesian weighted MR (BWMR) method was employed to improve the precision of causal inference by addressing pleiotropy and heterogeneity, while mitigating bias from weak IVs (44).

Mediation analysis

In the mediation analysis, lipidomes, inflammatory factors, and BC were included. Two-step MR analysis was conducted between each pair to investigate if circulating inflammatory factors mediate the causal relationship between lipidomes and BC outcomes. Initially, we assessed the potential causal relationship between liposomal IVs and the mediator variables. Subsequently, we examined the causal impact of circulating inflammatory factors on BC. Unless otherwise stated, all tests were two-sided with a primary significance threshold of P<0.05.

Sensitivity analysis

Various sensitivity analyses and statistical assessments were conducted to evaluate the hypothesis’s validity. The intercept and slope of the MR-Egger method were utilized for pleiotropy assessment, indicating the relationship between IVs and outcomes (45). Cochran’s Q test was also employed to evaluate causal effects, with a value above 0.05 suggesting homogeneity. Statistical heterogeneity and stability were assessed with scatter plots and funnel plots. MR-PRESSO detected pleiotropy and potential outliers in SNPs, whose correction is helpful to the robustness of causal effect estimates (46). To address horizontal pleiotropy resulting from individual SNPs, the leave-one-out analysis was conducted.

Ethics statement

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was based entirely on publicly available summary-level GWAS datasets, including data from the GWAS catalog (GCST90277238–GCST90277416), the FinnGen consortium (version R11), and the Olink Target Inflammation panel. All original studies received approval from their respective institutional ethics committees, and participants provided informed consent. As this analysis used only de-identified, aggregated data from these public resources, no additional ethical approval or informed consent was required.

Structural homology searches

An exhaustive search strategy was employed using Foldseek to cross-compare TRAIL and DR4 with the AlphaFold database, aiming to reveal structural evolutionary characteristics and origins. In the Foldseek algorithm, three-dimensional (3D) protein structures are reduced to one-dimensional 3Di sequences. A structural alphabet (47) was trained using machine learning, where the local structure of each residue is mapped to a symbolic letter, with 20 different characters representing structural features such as α-helices and β-sheets. This encoding method preserves key structural information while converting complex 3D coordinates into linear data analogous to amino acid sequences, facilitating rapid downstream comparisons.

For the generated 3Di sequences, an optimized Many-against-Many sequence searching (MMseqs2) (48) algorithm was applied for pre-filtering, rapidly selecting potential structural homologs. Subsequently, local or global alignment algorithms (e.g., Gotoh-Smith-Waterman or TMalign) were utilized for precise structural comparisons, ensuring high sensitivity and accuracy. Notably, the AlphaFold database, which contains over 200 million protein structures, enables Foldseek to compare billions of structure pairs within just a few hours—several orders of magnitude faster than traditional tools such as DALI and TM-align (49).

Molecular docking

Molecular docking can be used to predict the complex model of protein-protein or protein-small molecules. The essence of molecular docking lies in spatial shape complementarity and energy matching between receptors and ligands. As a classical molecular docking tool, AutoDock based on Lamarckian genetic algorithm has also been widely used in the field of drug design. In the process of conformation search, not only the rotation of small molecules is fully considered, but also the side-chain flexibility of the protein is partially sampled. The affinity function is set as docking scoring (50), by fitting experimental data. The smaller the value, the stronger the receptor-ligand binding potency. AutoDock software package consists of two modules Autogrid and Autodock (51). The former is used to preprocess target proteins computationally, assigning charge and position to the grid. The latter is responsible for the docking of ligands into the gridded target proteins, where affinity grid visualization facilitates the subsequent compound modification.

The molecular docking experiments were performed between DR4_TRAIL complex (PDB ID: 5cir) and phosphatidylcholine (16:0_16:0) with AutoDock 4.2 software package (52). In the actual docking, a cube box with a size of 50×50×50 ų was used, with a lattice spacing of 0.375 Å and a maximum docking frequency of 128 times. Finally, the complex model with the lowest binding free energy in the largest cluster was identified as the near-native conformation. Here, the docking complex model was used for subsequent comparative molecular dynamics simulations, represented by TRAIL_DR4_PL.

Molecular dynamics simulation

Two 200 ns comparative molecular dynamics simulations (53) were performed for the TRAIL_DR4 and TRAIL_DR4_PL systems, adopting AMBER14SB force field and CHARMM TIP3P38 (52) water model. Each set of kinetic systems is placed in a cubic box filled with water molecules, where periodic boundary conditions are defined. In addition, the system charge balance is achieved by adding sodium and chloride ions. Based on the coordinate of phosphatidylcholine (16:0_16:0), the corresponding topological file was generated using Sobtop tool (54). Energy minimization was achieved for all systems using the steepest descent algorithm, with convergence of energy difference between neighboring snapshots being less than 1,000 kJ·mol⁻¹·nm⁻¹, or up to a 20,000 pre-set iterative steps.

The Nose-Hoover thermostat (55) was used to maintain constant temperature during simulations with the coupling constant of 1.0 ps. Berendsen barostat (56) with high numerical stability was selected during the preparation of the investigated systems. The reference pressure was 1 bar, with the coupling constant of 5.0 ps and a compressibility of 4.5×10⁻⁵ bar⁻¹. In the simulation run, the pressure regulator was changed to the Parrinello-Rahman strategy (57) in view of its advantages of better representing the constant number, pressure, temperature ensemble as a whole. For van der Waals interactions, force-based switch function (1.0–1.2 nm) was used, while long-range Coulomb interactions were handled with the particle mesh Ewald method (58). The LINCS algorithm (59) was adopted to constrain the bonds between heavy atoms and hydrogen with a LINCS order of 4 at each iteration step. The integral step used in the simulations was 2 fs, with the reference temperature and box size being respectively set as 310 K and 15×15×15 nm³. All system preparation, trajectories acquisition, and analyses were done with GROMACS v2019.3 (60).

Binding free energy calculation

Binding free energy calculation is one of important methods to study mutual recognition of proteins by various ligands. In this work, the average binding free energies were predicted for the TRAIL_DR4 and TRAIL_DR4_PL systems by molecular mechanics/generalized Born surface area (MM/GBSA) method. Total 20 snapshots were collected from each molecular dynamics trajectory every 1 ns intervals from 1 to 200 ns. The calculation formula of binding free energy (∆G_bind) is as follows:

ΔGbind=ΔH−TΔS=(ΔEVDW+ΔEELE+ΔGGBELE+ΔGGBSUR)−TΔS[1]

where ∆H corresponds to total enthalpy change, and T∆S is the product of absolute temperature and conformational entropy change. In addition, ∆H is composed of four parts: ∆E_VDW, ∆E_ELE, ∆G_GBELE, and ∆G_GBSUR. Specifically, ∆E_VDW indicates intramolecular van der Waals energy under vacuum; ∆E_ELE refers to the electrostatic fraction; ∆G_GBELE and ∆G_GBSUR represent the hydrophilic and hydrophobic parts of solvation binding free energy, respectively.

Weak interaction analysis

Weak interactions in proteins mainly include hydrogen bonds and van der Waals forces, which are critical for the selective binding of ligands and the maintenance of high affinity. In this study, the independent gradient model based on Hirshfeld partitioning (IGMH) (61) method, embedded in the Multiwfn program (62,63), was employed to analyze weak interactions in the TRAIL-DR4 and TRAIL-DR4-PL systems.

IGMH is an independent gradient model based on Hirshfeld atomic partitioning, enabling the effective visualization and analysis of weak interactions both within and between molecular fragments. In biochemical research, it is widely used to investigate hydrogen bonding, van der Waals forces, and π-π stacking between protein residues.

In this study, stable converged conformations obtained from molecular dynamics simulations were used to identify key residues involved in hydrogen bonds and van der Waals interactions during TRAIL-DR4 recognition. These key residues were defined as designated fragments, allowing a comparative analysis of differences in weak interactions between the systems.

Results

We first conducted MR analysis on GWAS data comprising 179 lipidomes and BC, identifying six lipids associated with the disease. The corresponding codes and lipidome were both summarized in Table S1. Next, reverse MR analysis was performed to explore the causal relationship between these six lipidomes and BC, revealing that five of them factually lacked a reverse causal association with the disease. And then, we conducted a two-sample MR analysis involving these five lipids and 91 inflammation factors. The results show that 15 inflammation factors might be causally linked to lipids. Finally, we fully examined these 15 inflammation factors concerning BC, discovering that one of them might have a causal relationship with the disease. Hence, we posit that this inflammation factor could mediate the causal association between lipids and BC. The flow chart is summarized in Figure 1C.

In the meantime, to ensure the genetic similarity between the two sample groups and address the uncertainty regarding potential sample overlap, the exposure and outcome datasets both consist of individuals of European ancestry, ensuring genetic variant-exposure associations are comparable across samples. Moreover, the summary statistics were generated using similar quality control measures and analysis pipelines, supporting the consistency of genetic effects. The exposure and outcome datasets were obtained from publicly available consortia (GWAS catalog, FinnGen, and Olink panel), and to the best of our knowledge, there is no known overlap between participants in the exposure and outcome samples. However, due to the use of summary-level data from public databases, precise information on individual overlap is unavailable.

MR analysis reveals a positive cause‑effect relationship between lipidome and BC

Initial investigations into the impact of 179 lipidomes on the overall BC risk were conducted through two-sample MR analyses. The SNPs were summarized in Table S2. We then utilized IVW, MR-Egger, simple mode, and weighted median methods to estimate the causal relationship between the lipidome and BC. The IVW method as the primary results showed that sterol ester (27:1/15:0) levels [OR of 0.8739 (95% CI: 0.8125–0.9398), P<0.001], phosphatidylcholine (16:0_16:0) levels [OR of 1.0764 (95% CI: 1.0124–1.1443), P=0.02], phosphatidylcholine (16:1_18:1) levels [OR of 0.9362 (95% CI: 0.8863–0.9888), P=0.02], phosphatidylcholine (18:1_20:2) levels [OR of 0.9607 (95% CI: 0.9231–0.9998), P=0.049], phosphatidylcholine (O-18:0_14:0) levels [OR of 1.0913 (95% CI: 1.0185–1.1692), P=0.01], and phosphatidylinositol (18:1_20:4) levels [OR of 0.9115 (95% CI: 0.8637–0.9620), P<0.001] might be causally related to BC.

To further clarify the accuracy of the MR results, we conducted further validation using the BWMR approach. Five positive results have been found: sterol ester (27:1/15:0) levels (P<0.001), phosphatidylcholine (16:0_16:0) levels (P=0.01), phosphatidylcholine (16:1_18:1) levels (P=0.02), phosphatidylcholine (O-18:0_14:0) levels (P=0.02), and phosphatidylinositol (18:1_20:4) levels (P=0.001). The results are shown in Table 1, Figures 2,3.

Figure 2 Forest plot to visualize the causal effects of TRAIL with phosphatidylcholine (16:0_16:0) and breast cancer. BC, breast cancer; CI, confidence interval; IVM, inverse variance weighting; nSNP, number of single nucleotide polymorphism; OR, odds ratio; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand.

Figure 3 Forest plot to visualize the causal effect of TRAIL with lipidome and breast cancer. (A) Represented the causal effect of Phosphatidylcholine (16:0_16:0) on breast cancer. (B) Represented the causal effect of Phosphatidylcholine (16:0_16:0) on TRAIL. (C) Represented the causal effect of TRAIL on breast cancer. The funnel plots show the inverse variance weighted MR estimate of TRAIL with lipidome and breast cancer. (D) Represented the funnel plot of Phosphatidylcholine (16:0_16:0) on breast cancer. (E) Represented the funnel plot of Phosphatidylcholine (16:0_16:0) on TRAIL. (F) Represented the funnel plot of TRAIL on breast cancer. The sensitivity-analysis plot to visualize the causal effect of TRAIL with lipidome and breast cancer. (G) Represented the sensitivity-analysis plot of Phosphatidylcholine (16:0_16:0) on breast cancer. (H) Represented the sensitivity-analysis plot of Phosphatidylcholine (16:0_16:0) on TRAIL. (I) Represented the sensitivity-analysis plot of TRAIL on breast cancer. IVM, inverse variance weighting; MR, Mendelian randomization; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand.

Then, reverse MR analysis was employed to investigate the potential presence of reverse causality. As shown from Table S3, the predicted results indicate there is no reverse causal effect between the above lipidomes and BC except for the phosphatidylcholine (18:1_20:2) levels [OR of 1.0573 (95% CI: 1.0121–1.1046), P=0.01]. Thus, we used the rest of the five lipidomes to explore the mediators causing BC.

MR analysis reveals TRAIL as a mediator between lipidome and BC

To establish the specific association between lipoproteins and BC, we further explored whether the 91 inflammatory factors are the mediators via the IVW method. The SNPs were summarized in Table S4. As for the relationship between above five positive lipidome [i.e., sterol ester (27:1/15:0) levels, phosphatidylcholine (16:0_16:0) levels, phosphatidylcholine (16:1_18:1) levels, phosphatidylcholine (O-18:0_14:0) levels, and phosphatidylinositol (18:1_20:4) levels] and 91 inflammatory factors, total of 15 significant positive results were identified. Specifically summarized in Table S5, they include interleukin-22 receptor subunit alpha-1 levels (P=0.01), adenosine deaminase levels (P=0.01), C-C motif chemokine 23 levels (CCL23, P=0.043), natural killer cell receptor 2B4 levels (P=0.006), T-cell surface glycoprotein CD5 levels (P=0.041), CUB domain-containing protein 1 levels (CDCP1, P=0.03), C-X-C motif chemokine 1 levels (CXCL1, P=0.03), interleukin-18 levels (P=0.03), latency-associated peptide transforming growth factor beta 1 levels (P=0.046), monocyte chemoattractant protein-4 levels (P=0.043), TNF ligand superfamily member-14 levels (P=0.03), TRAIL levels (P=0.02), eotaxin levels (P=0.02), CDCP1 (P=0.01), cystatin D levels (P=0.043) and TNF-beta levels (P=0.02).

Referring to the SNPs summarized in Table S6, we then analyzed the relationship between the above 15 inflammatory factors and BC. The results revealed that TRAIL was identified as a significant factor [OR of 0.8739 (95% CI: 0.8125–0.9398), P<0.001], which was a potential mediator bridging lipoproteins to BC (indirect effect: 0.0041). In summary, our study suggests that elevated lipoprotein levels may lead to a decrease in TRAIL levels, consequently correlating with an increased risk of BC.

The statistical results show good reliability and robustness

The sensitivity analyses were conducted to assess the robustness and resilience of conclusions against potential biases (Table S7). The leave-one-out analysis indicated that the effect estimates remained consistent when each SNP was removed individually, confirming the robustness of the statistical results. The Cochran’s Q test demonstrated non-significant heterogeneity among the effects of SNPs. Additionally, the MR-Egger analysis did not detect heterogeneity, and the MR-PRESSO test did not identify pleiotropy. Scatter plots and funnel plots are provided in Figure 3D-3F, also showing no apparent heterogeneity in the results.

Structural evolution conserves molecular recognition between TRAIL and DR4

As a member of the TNF superfamily, TRAIL induces apoptosis through interaction with its receptors. The TRAIL receptor system exhibits considerable complexity, comprising both pro-apoptotic receptors, such as DR4 and DR5, which contain a fully functional death domain, and decoy receptors, including TRAIL receptor 3 (TRAILR3, also termed decoy receptor 1, DcR1) and TRAIL receptor 4 (TRAILR4, also termed decoy receptor 2, DcR2), which lack a complete death domain and are incapable of transducing apoptotic signals. Among them, TRAILR4, as a key regulatory receptor, is hypothesized to modulate TRAIL-DR4-mediated apoptotic signaling by competitively sequestering TRAIL, thereby attenuating apoptosis. Mechanistically, TRAIL binding to DR4 promotes receptor trimerization, subsequently initiating downstream apoptotic signaling cascades—a process of pivotal significance in tumor suppression and immune modulation (64).

Foldseek was employed to search for distantly related homologs of TRAIL and DR4 in bacteria. In the viral taxonomic classification system, five bacterial genus proteins were identified from different hierarchical taxa under the Bamfordvirae megavirus clade (Figure 4). Specifically, the evolutionary path from higher to lower taxonomic levels follows this order: Bamfordvirae (kingdom), Nucleocytoviricota (phylum), Megaviricetes (class), Pimascovirales (order, infecting invertebrates and vertebrates), and Iridoviridae (family). Homologs with an E-value below 1e−5 were considered significant matches. These findings suggest that both TRAIL and DR4 have structurally similar homologous proteins in bacteria, which may have independently evolved analogous binding interfaces through convergent evolution.

Figure 4 Structure of the viral proteins of TRAIL (A-E) and death receptor 4 (F-J) under the different hierarchical taxonomic units of the virus. The e value is a key indicator to measure the statistical significance of structural alignment results, and pLDDT is used to represent the confidence of predicted structures. e, Expect value; pLDDT, predicted local distance difference test; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand.

By analyzing the qStart-qEnd and tStart-tEnd fields in the sequence alignment results, the interaction domains of TRAIL and DR4 were mapped. Notably, the C-terminal β-strand of TRAIL (190F-198E) and the cysteine-rich domain 2 (CRD2, 148C-171K) and cysteine-rich domain 3 (CRD3, 189C-211C) domains of DR4 exhibit corresponding matching regions in bacterial homologs, indicating structural conservation. This suggests that the TRAIL-DR4 binding mechanism has been evolutionarily preserved (Figure 4).

Phosphatidylcholine potentially disrupts TRAIL-mediated apoptotic signaling

According to the study by Valley (65), experiments involving cellular fractionation and Western blot analysis demonstrated the regulatory role and dependence of TRAIL receptor behavior on the lipid microenvironment, particularly membrane cholesterol. Cholesterol was found to be essential for TRAIL receptor dimerization and the formation of higher-order receptor networks, as its depletion inhibited DR5 ligand binding and functional activation. Using immunofluorescence microscopy, Vanamee and Faustman (66) visually illustrated the redistribution of DR4 and DR5 into lipid rafts. Furthermore, through immunoprecipitation assays, they demonstrated that this redistribution facilitates the formation of the death-inducing signaling complex, thereby enhancing apoptosis. Collectively, the relocalization of TRAIL receptors into lipid rafts amplifies TRAIL-induced apoptosis. Given the established interaction between TRAIL, its receptors, and lipid rafts, there is a strong indication that TRAIL receptor activity may be intrinsically linked to phospholipids. Based on this rationale, it is plausible that phosphatidylcholine (16:0_16:0) modulates BC risk by influencing the TRAIL signaling pathway. According to molecular simulation results by Putra et al. (67), caffeic acid (−6.4 kcal/mol) and homovanillic acid (−6.6 kcal/mol) exhibit the highest binding affinities for TNF alpha and TRAIL, respectively. Although caffeic acid and homovanillic acid are classified as phenolic compounds, they share considerable structural similarity with phospholipids. Both molecules contain hydrophilic polar functional groups (phosphate or hydroxyl groups) capable of forming hydrogen bonds or ionic interactions with water. Based on these findings, molecular docking was employed to construct a ternary complex model involving TRAILR4, TRAIL, and DR4, providing a theoretical framework for elucidating the regulatory mechanisms of this pathway. As shown in Figure 5, phosphatidylcholine (16:0_16:0) stably binds within the active pocket formed by S241, I242, Y243, and R227, engaging in hydrogen bond interactions.

Figure 5 The TRAIL_DR4_PL ternary complex model and its key residues, where green dashed lines are used to indicate hydrogen bonding. DR4, death receptor 4; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand.

Trajectory convergence being the prerequisite of molecular recognition research

To further investigate the interaction between phosphatidylcholine (16:0_16:0) and the TRAIL-DR4 complex, comparative molecular dynamics simulations of 200 ns were performed for two complex systems: TRAIL_DR4 and TRAIL_DR4_PL.

Figure 6A illustrates the changes in potential energy over the simulation time for both TRAIL_DR4 and TRAIL_DR4_PL systems. Overall, the potential energy of the systems rapidly reached equilibrium, indicating a stable simulation process and reliable environmental variable settings. The root mean square deviation (RMSD) was calculated by measuring the binding differences relative to the first snapshot, effectively evaluating whether the simulation trajectory had reached equilibrium (Figure 6B). After 60 ns, the RMSD values of both systems exhibited similar fluctuation trends, achieving a well-converged state. The relatively higher RMSD value of TRAIL_DR4 after stabilization may be attributed to significant conformational changes in DR4 upon binding, indicating an induced-fit binding process. This observation aligns with the findings of Vanamee and Faustman (66), who reported that TRAIL binding leads to dynamic rearrangement of the CRD2 and CRD3 domains of DR4, resulting in the exposure of the death domain (246L-315G).

Figure 6 Conformation convergence parameters for molecular dynamics simulations of the TRAIL_DR4 and TRAIL_DR4_PL systems. The variation of potential energy (A) and RMSD (B) with time; (C) RMSF distribution of TRAIL and DR4 at the residue level; (D) RMSF correlation of TRAIL_DR4 with the TRAIL_DR4_PL systems. CRD, cysteine-rich domain; DR4, death receptor 4; RMSD, root mean square deviation; RMSF, root mean square fluctuation; THD, tumor necrosis factor homologous domain; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand.

Since the RMSD values remained relatively stable after 60 ns, the molecular dynamics trajectories from this period were used for subsequent molecular recognition studies. Another key parameter for macroscopic molecular dynamics simulation analysis, the root mean square fluctuation (RMSF), provides residue-level displacement data throughout the simulation. A higher RMSF value indicates greater conformational flexibility of the residues. The RMSF distributions of TRAIL_DR4 and TRAIL_DR4_PL were highly similar, with high flexibility observed in the CRD2 and CRD3 functional domains of DR4 and the β2 strand (130R-138S) in the TNF homologous domain of TRAIL (Figure 6C). This result is consistent with the mutagenesis study by Truneh et al. (68), which demonstrated that structural rearrangement of CRD2 and CRD3 in DR4 is essential for TRAIL-induced apoptosis. Compared to TRAIL_DR4, the binding of phospholipids (i.e., in TRAIL_DR4_PL) significantly reduced the flexibility of the CRD2 and CRD3 domains.

In summary, the binding of phosphatidylcholine (16:0_16:0) did not significantly affect the overall flexibility distribution of TRAIL and DR4. The TRAIL_DR4 complex exhibited minimal fluctuations during the simulation, maintaining a relatively rigid structure. It is speculated that TRAIL and DR4 form a relatively stable weak interaction, which may be partially disrupted upon phosphatidylcholine (16:0_16:0) binding. Furthermore, a strong correlation was observed between the RMSF values of Cα atoms within the system, further confirming the reliability of the molecular dynamics trajectories (Figure 6D).

Phosphatidylcholine (16:0_16:0) binding significantly weakens H-bonds between TRAIL and DR4

Hydrogen bonding is a key determinant of receptor-ligand specificity and a crucial force in maintaining complex stability. To investigate this, the hydrogen bond interactions between TRAIL and DR4 were analyzed based on molecular dynamics trajectories from two systems. Unexpectedly, the total number of hydrogen bonds in the larger TRAIL_DR4_PL system (~109) was lower than that in the TRAIL_DR4 system (~155). Li and Shu (69) validated the interaction between DR4 and TRAIL using co-immunoprecipitation. Similarly, Ramamurthy et al. (70) confirmed the core binding residues R191T and D173D within the DR4-TRAIL complex through chromatographic purification and gel analysis of the corresponding fractions. Additionally, K71D and D267T/269T formed extra salt bridges. Upon phosphatidylcholine (16:0_16:0) binding, both the number and occupancy of critical hydrogen bonds between TRAIL and DR4 decreased (Table 2). This further supports the notion that phosphatidylcholine (16:0_16:0) partially disrupts the molecular recognition between TRAIL and DR4, ultimately impairing the downstream apoptotic signaling pathway in BC cells.

Binding of phosphatidylcholine (16:0_16:0) lowers the free energy of TRAIL-DR4 interaction

Binding free energy is a crucial physicochemical parameter for evaluating receptor-ligand recognition and serves as a key criterion in computer-aided drug design for selecting lead compounds. Table 3 presents the binding free energy between TRAIL and DR4 predicted using the MM/GBSA method, along with the contributions of individual energy components. Under identical computational conditions, the lowest binding free energy (−17.79 kcal/mol) was observed in the absence of phosphatidylcholine, indicating that TRAIL spontaneously binds to DR4 to activate the apoptotic signaling pathway. The slightly higher binding free energy in the TRAIL_DR4_PL system (−15.84 kcal/mol) suggests that the apoptotic function is partially suppressed in the presence of phospholipids.

As listed in Table 3, nonpolar interactions (VDW_IN + VDW_GB) serve as the primary driving force for TRAIL-DR4 recognition. As previously mentioned, phosphatidylcholine binding weakens the hydrogen bond interactions between TRAIL and DR4. The changes in molecular recognition between the TRAIL_DR4 and TRAIL_DR4_PL systems are further analyzed in detail in the subsequent discussion of weak interactions.

Phosphatidylcholine (16:0_16:0) weakens R191T-D173D and K171D-D267T/D269T interactions

By mapping weak interaction regions, not only can molecular recognition changes be intuitively revealed, but potential differences in reactivity and activation energy can also be inferred. Here, interaction-guided molecular hotspots analysis was employed to investigate weak receptor-ligand interactions, such as hydrogen bonding and van der Waals interactions, in the equilibrated molecular dynamics trajectories of the TRAIL_DR4 and TRAIL_DR4_PL systems. Based on the experimental findings of Ramamurthy et al. (70), particular attention was given to the weak interactions between the critical TRAIL residues R191^T/D267^T/D269^T and the key DR4 residues K171^D/D173^D (Figure 7).

Figure 7 Weak interaction of key residues in the TRAIL_DR4 (A) and TRAIL_DR4_PL (B) systems, where DR4 and TRAIL are colored in red and black, respectively. DR4, death receptor 4; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand.

In the TRAIL_DR4 system, a strong hydrogen bond network was observed, as indicated by the formation of blue isosurfaces between R191^T and D173^D, as well as between K171^D and D267^T/D269^T. However, in the TRAIL_DR4_PL system, the hydrogen bonding interaction between D173^D and R191^T was completely disrupted, while the interaction between K171^D and D267^T/D269^T was significantly weakened. This weakening was accompanied by a transition from hydrogen bonding to van der Waals interactions, as evidenced by the isosurface color shift from blue to green.

These results suggest that the incorporation of phosphatidylcholine disrupts hydrogen bonds and salt bridges formed between the core β-strand of TRAIL and the CRD2 domain of the DR4 receptor, thereby impairing their molecular recognition.

Discussion

To the best of our knowledge, this is among the first MR study exploring the causal effects among lipidome, inflammation factors, and BC. This study comprehensively evaluated the causal impacts of 179 lipidomes on BC and provided genetic evidence suggesting their positive association by employing a two-step MR analysis. In fact, this effect was mediated through TRAIL, and the statistical robustness was confirmed with sensitivity analyses. According to the MR results, these lipidomes may elevate the risk of BC by decreasing TRAIL levels, consequently diminishing its protective effect against the disease.

In this work, a positive causal relationship was highlighted between phosphatidylcholine (16:0_16:0) and BC, which is consistent with the previous MR data by Cao et al. (71). Some researchers suggested that the accumulation of choline metabolites (precursors or breakdown products of phosphatidylcholine) in cancer cells, may favor cancer cell survival, proliferation, and immune regulation (72). As a primary source of phospholipids in mammalian cell membranes (approximately 50%), phosphatidylcholine consists of a glycerol backbone, two fatty-acyl chains, and a phosphocholine headgroup (73,74). Inhibiting the biosynthesis of phosphatidylcholine—and of other lipids required for membrane biogenesis, including lipid-raft formation—can effectively suppress tumor growth (75). Collectively, these observations raise the hypothesis that modest adjustments to phosphatidylcholine-related lipid metabolism could be linked to improved outcomes. However, epidemiologic metabolomics studies have reported inverse associations for several other phosphatidylcholine species with BC risk (e.g., PC C36:3, aa C36:3, ae C34:2, ae C36:2, ae C38:2 with OR<0.86; and PC aa C32:2, aa C36:0, aa C36:1) (76,77). These discrepancies underscore species-specific heterogeneity within the phosphatidylcholine class. The observed positive association for PC (16:0/16:0)—and the question of whether its selective targeting could improve prognosis or clinical outcomes—warrants further investigation to elucidate the underlying biology.

TRAIL has been identified as a potential intermediary factor. The utilization of TRAIL agonists in BC therapy is currently under extensive investigation (78). Currently, the utilization of TRAIL agonists in BC therapy is under extensive investigation, and drugs based on the principle of TRAIL are in clinical practice (75). In preclinical research, a study by Holland and colleagues (79) highlights that Dulanermin can effectively inhibit human BC MDA-MB-231 cells. Dulanermin is a recombinant human apolipoprotein two ligand/TRAIL (rhApo2L/TRAIL/dulanermin) that preferentially induces apoptosis in cancer cells. In a separate preclinical study (79), co-administration with a RANKL inhibitor further reduced skeletal tumor burden in murine models of BC bone metastasis. Clinically, however, the value of TRAIL agonists in BC remains unestablished: in metastatic triple-negative BC (mTNBC), adding the DR5 agonist tigatuzumab to nab-paclitaxel did not improve objective response rate or progression-free survival overall, although a small number of durable responses were observed (80), underscoring the need for biomarker-guided patient stratification and rational combination strategies to overcome resistance. Now, a clinical study on the combined treatment of metastatic BC with TRAIL-R2 and HER2 is currently in the recruitment phase (NCT06251544), whose results have received wide attention. Accordingly, whether enhancing TRAIL signaling can reduce BC risk or improve clinical outcomes remains to be determined.

Beyond lipid blockade and TRAIL agonism, it is pertinent to ask whether standard adjuvant therapies impinge on the relationships among PC (16:0/16:0), TRAIL, and BC outcomes. Although therapy-stratified MR analyses were not feasible using summary-level data, several biologically plausible mechanisms suggest potential intersections. First, alterations in the phosphatidylcholine (PC) profile have been linked to endocrine therapy response and, in certain contexts, to TRAIL sensitivity. Li et al. (81) reported that loss of the mitochondrial transacylase tafazzin disrupts cardiolipin homeostasis, accompanied by elevated lysophosphatidylcholine and its derivative lysophosphatidic acid, which suppresses estrogen receptor (ER)-α expression and nuclear localization, thereby promoting tamoxifen resistance. Similarly, Piggott et al. (82) demonstrated that endocrine-resistant ER-positive BC cells exhibit acquired TRAIL sensitivity, driven by JNK-dependent phosphorylation of the E3 ligase ITCH, which facilitates cellular FLICE-inhibitory protein (c-FLIP) degradation and lowers the apoptotic threshold to TRAIL. Second, chemotherapy has been shown to reprogram choline-phospholipid metabolism at an early stage. In BC models, the glycerophosphocholine (GPC)-to-phosphocholine ratio is more sensitive and broadly applicable than total choline as a pharmacodynamic indicator, though its molecular underpinnings remain to be elucidated (72,83). Cross-tumor evidence also points to chemotherapy-induced TRAIL sensitization: cisplatin or 5-fluorouracil pretreatment enhanced the antitumor efficacy of MSC-TRAIL in CD133⁺ non-small cell lung cancer (NSCLC) models (84), while oxaliplatin pretreatment increased the cytotoxic and pro-apoptotic effects of soluble or membrane-bound TRAIL in oxaliplatin-resistant colorectal cancer cells (85). Taken together, these observations provide a coherent biological rationale that adjuvant therapies may influence PC (16:0/16:0), TRAIL, and BC outcomes, and potentially modulate their interrelationships. Nonetheless, these hypotheses require further investigation and validation in BC cohorts and functional experiments.

According to the MR study in this work, there is a negative causal relationship between phosphatidylcholine (16:0_16:0) and TRAIL. As previously discussed, TRAIL is a critical apoptotic signaling molecule, with its receptors DR4 and DR5 playing essential roles in oncogenesis and immune regulation. The functionality of TRAIL receptors is not solely dictated by ligand-binding affinity but is also subject to modulation by the lipid microenvironment. Notably, TRAIL receptors have been shown to localize within lipid rafts, a process that facilitates receptor dimerization and the formation of signaling complexes (65,86). Lipid rafts, which are highly dynamic membrane microdomains composed of cholesterol, phospholipids, and sphingolipids, are increasingly recognized as pivotal regulators of apoptotic signaling (87). Emerging evidence indicates that certain small molecules can potentiate TRAIL-induced apoptosis by altering the spatial distribution of DR4 and DR5 within these membrane domains (87-91). Alterations in the lipid milieu—such as elevated ceramide levels or increased membrane fluidity—may further influence TRAIL receptor organization and downstream signaling by modulating lipid raft integrity, membrane dynamics, and the surrounding microenvironment, thereby serving as key determinants of apoptotic regulation (92). Furthermore, in the context of bone metabolism, TRAIL has been shown to suppress receptor activator of nuclear factor kappa-B (RANK)-mediated lipid raft signaling through its receptors, underscoring the potential role of TRAIL receptor-membrane lipid interactions in orchestrating downstream signal transduction (93).

Given the established correlation between TRAIL receptor family members and lipid metabolism (94), DR4 is postulated to play a role in lipid dysregulation-associated pathologies, such as atherosclerosis (95). In hematologic malignancies and NSCLC, the localization of DR4 within lipid rafts has been shown to be critical for TRAIL-induced apoptosis (88-91). Apoptotic death receptors can translocate into lipid rafts, triggering downstream apoptotic signaling, where lipid rafts serve as platforms for adaptor proteins [e.g., FADD and TNFRSF1A-associated death domain protein (TRADD)] and effector proteins (e.g., pro-CASP8 and pro-CASP10) involved in the extrinsic apoptotic pathway (96). Within this context, we further posit that increased phosphatidylcholine (16:0_16:0) could perturb lipid-raft architecture and death-receptor clustering, indirectly altering TRAIL–DR4/DR5 engagement, with consequent modulation of apoptosis.

In terms of potential binding modes, TRAIL interacts with DR4 through its TNF homologous domain, which consists of β-sheet regions (V97-Q103, R130-R138, E190-D198), forming stable hydrogen bonds and hydrophobic interactions with CRD2 (C148-K171) and CRD3 (C189-C211) of DR4 (86). Based on the crystal structure of the TRAIL-DR5 complex reported by Mongkolsapaya et al. (97), it is reasonable to infer that DR4 may adopt a similar binding pattern. Upon binding, DR4 undergoes conformational changes, may facilitate downstream signal transduction and initiate the apoptotic signaling pathway.

Under this mechanistic framework, we conducted methodologically simplified molecular docking and molecular dynamics simulations to explore the putative impact of phosphatidylcholine (16:0_16:0) on TRAIL-DR4 interactions. The results demonstrated that, in the presence of phosphatidylcholine (16:0_16:0), both the number and occupancy of hydrogen bonds between TRAIL and DR4 were reduced, while the flexibility of the CRD2 and CRD3 domains was significantly diminished. Moreover, binding free energy analysis revealed a substantial increase in the presence of phosphatidylcholine (16:0_16:0), suggesting that it may attenuate the DR4 interaction, thereby suppressing TRAIL-mediated apoptotic signaling in BC cells. Notably, this mechanism provides a plausible explanation for the observed positive causal relationship between phosphatidylcholine (16:0_16:0) and BC. This study not only offers novel insights into the lipid-mediated regulation of TRAIL receptors but also suggests that phosphatidylcholine (16:0_16:0) may serve as a potential biomarker for BC, providing a theoretical foundation for future therapeutic strategies.

Taken together, these results provide preliminary support for a mechanism whereby increased PC (16:0_16:0) may weaken TRAIL-DR4/DR5 engagement and dampen extrinsic apoptotic signaling, but they do not establish causality; targeted validation in membrane-reconstitution or lipid-raft-perturbation cell systems and in vivo models is required, including studies that disentangle direct conformational effects from alternative mechanisms such as changes in receptor expression, glycosylation, or endocytic trafficking.

It is worth noting that there are several limitations in this study. First, our analyses exclusively used European-ancestry GWAS resources (GWAS catalog, FinnGen, Olink panels). Population differences in linkage disequilibrium structure, allele frequencies, lipid metabolism, and BC epidemiology may lead to reduced generalizability of the causal estimates beyond European populations. Accordingly, the applicability of our findings to non-European ancestries remains to be established and is beyond the scope and data availability of the present study. Second, despite the efforts to identify and eliminate outliers, it is impossible to exclude the potential impact of horizontal pleiotropy. Third, summary-level statistical data are utilized, which greatly excludes other cancer risk confounding factors, while preventing further exploration of causal relationships in subgroup analyses. Moreover, the data do not allow assessment across different exposure periods or varying exposure levels, which may further limit the generalizability of our findings. Specifically for BC, subtype-specific analyses could not be performed. Given the etiologic and prognostic heterogeneity among ER-positive, HER2-positive, and triple-negative BC. This is a purely computational, hypothesis-generating study; all proposed mechanisms and causal inferences require experimental validation in appropriate in vitro, in vivo, and clinical settings.

Conclusions

This study establishes a positive causal relationship between phosphatidylcholine (16:0_16:0) and BC and suggests that TRAIL may act as a mediator in this association by influencing apoptotic pathways. Through molecular docking and molecular dynamics simulations, it is found that phosphatidylcholine (16:0_16:0) may interfere with the binding of TRAIL to its receptor DR4, reducing hydrogen bond interactions and increasing binding free energy, thereby affecting the stability and function of the DR4 complex. This disruption may lead to impaired apoptotic signaling, providing a potential mechanistic explanation for the observed causal relationship. Given the current limited evidence, further experimental validation is necessary to elucidate the precise molecular mechanisms. This work not only explored lipid-mediated regulation of TRAIL signaling, having theoretical significance in apoptosis, but also identified phosphatidylcholine (16:0_16:0) as a possible biomarker for BC development, showing certain clinical application value.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE-MR reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1096/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1096/prf

Funding: This work was supported by the Sichuan Science and Technology Program (Grant No. 2025ZDZX0012) and the National Guidance Fund on Developing Local Science and Technology for Sichuan Province (China) (No. 2023ZYD0167).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1096/coif). All authors report that this work was supported by the Sichuan Science and Technology Program (Grant No. 2025ZDZX0012) and the National Guidance Fund on Developing Local Science and Technology for Sichuan Province (China, No. 2023ZYD0167). The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Nolan E, Lindeman GJ, Visvader JE. Deciphering breast cancer: from biology to the clinic. Cell 2023;186:1708-28. [Crossref] [PubMed]
2. Miller KD, Nogueira L, Devasia T, et al. Cancer treatment and survivorship statistics, 2022. CA Cancer J Clin 2022;72:409-36. [Crossref] [PubMed]
3. Harbeck N, Gnant M. Breast cancer. Lancet 2017;389:1134-50. [Crossref] [PubMed]
4. Fernández Y, Cueva J, Palomo AG, et al. Novel therapeutic approaches to the treatment of metastatic breast cancer. Cancer Treat Rev 2010;36:33-42. [Crossref] [PubMed]
5. Burguin A, Diorio C, Durocher F. Breast Cancer Treatments: Updates and New Challenges. J Pers Med 2021;11:808. [Crossref] [PubMed]
6. Weaver KE, Forsythe LP, Reeve BB, et al. Mental and physical health-related quality of life among U.S. cancer survivors: population estimates from the 2010 National Health Interview Survey. Cancer Epidemiol Biomarkers Prev 2012;21:2108-17. [Crossref] [PubMed]
7. Luengo-Fernandez R, Leal J, Gray A, et al. Economic burden of cancer across the European Union: a population-based cost analysis. Lancet Oncol 2013;14:1165-74. [Crossref] [PubMed]
8. Sullivan R, Peppercorn J, Sikora K, et al. Delivering affordable cancer care in high-income countries. Lancet Oncol 2011;12:933-80. [Crossref] [PubMed]
9. Hassett MJ, Elkin EB. What does breast cancer treatment cost and what is it worth? Hematol Oncol Clin North Am 2013;27:829-41. ix. [Crossref] [PubMed]
10. Ward AV, Anderson SM, Sartorius CA. Advances in Analyzing the Breast Cancer Lipidome and Its Relevance to Disease Progression and Treatment. J Mammary Gland Biol Neoplasia 2021;26:399-417. [Crossref] [PubMed]
11. Ferreri C, Ferreri R, Filippone A, et al. Use of membrane lipidome, body weight, and composition in stratification of early breast cancer patients. NPJ Breast Cancer 2025;11:66. [Crossref] [PubMed]
12. Tokareva AO, Starodubtseva NL, Chagovets VV, et al. Lipidomic markers of tumor progress in breast cancer patients. Biomed Khim 2022;68:144-52. [Crossref] [PubMed]
13. Cruceriu D, Baldasici O, Balacescu O, et al. The dual role of tumor necrosis factor-alpha (TNF-α) in breast cancer: molecular insights and therapeutic approaches. Cell Oncol (Dordr) 2020;43:1-18. [Crossref] [PubMed]
14. Zhu M, Ma Z, Zhang X, et al. C-reactive protein and cancer risk: a pan-cancer study of prospective cohort and Mendelian randomization analysis. BMC Med 2022;20:301. [Crossref] [PubMed]
15. Riera-Domingo C, Leite-Gomes E, Charatsidou I, et al. Breast tumors interfere with endothelial TRAIL at the premetastatic niche to promote cancer cell seeding. Sci Adv 2023;9:eadd5028. [Crossref] [PubMed]
16. Bertsch U, Röder C, Kalthoff H, et al. Compartmentalization of TNF-related apoptosis-inducing ligand (TRAIL) death receptor functions: emerging role of nuclear TRAIL-R2. Cell Death Dis 2014;5:e1390. [Crossref] [PubMed]
17. Emery JG, McDonnell P, Burke MB, et al. Osteoprotegerin is a receptor for the cytotoxic ligand TRAIL. J Biol Chem 1998;273:14363-7. [Crossref] [PubMed]
18. Kimberley FC, Screaton GR. Following a TRAIL: update on a ligand and its five receptors. Cell Res 2004;14:359-72. [Crossref] [PubMed]
19. Griffith TS, Lynch DH. TRAIL: a molecule with multiple receptors and control mechanisms. Curr Opin Immunol 1998;10:559-63. [Crossref] [PubMed]
20. Ashkenazi A, Dixit VM. Apoptosis control by death and decoy receptors. Curr Opin Cell Biol 1999;11:255-60. [Crossref] [PubMed]
21. Ibrahim FAR, Abouelenein MS, Sakr OSA, et al. TRAIL and TRAIL Receptors as Prognostic Markers in Breast Cancer Patients. American Journal of Molecular Biology 2019;9:204-17.
22. Toiyama D, Takaha N, Shinnoh M, et al. Significance of serum tumor necrosis factor-related apoptosis-inducing ligand as a prognostic biomarker for renal cell carcinoma. Mol Clin Oncol 2013;1:69-74. [Crossref] [PubMed]
23. Tecchio C, Huber V, Scapini P, et al. IFNalpha-stimulated neutrophils and monocytes release a soluble form of TNF-related apoptosis-inducing ligand (TRAIL/Apo-2 ligand) displaying apoptotic activity on leukemic cells. Blood 2004;103:3837-44. [Crossref] [PubMed]
24. Tanaka H, Ito T, Kyo T, et al. Treatment with IFNalpha in vivo up-regulates serum-soluble TNF-related apoptosis inducing ligand (sTRAIL) levels and TRAIL mRNA expressions in neutrophils in chronic myelogenous leukemia patients. Eur J Haematol 2007;78:389-98. [Crossref] [PubMed]
25. Vanden Berghe T, van Loo G, Saelens X, et al. Differential signaling to apoptotic and necrotic cell death by Fas-associated death domain protein FADD. J Biol Chem 2004;279:7925-33. [Crossref] [PubMed]
26. Barnhart BC, Alappat EC, Peter ME. The CD95 type I/type II model. Semin Immunol 2003;15:185-93. [Crossref] [PubMed]
27. Micheau O, Tschopp J. Induction of TNF receptor I-mediated apoptosis via two sequential signaling complexes. Cell 2003;114:181-90. [Crossref] [PubMed]
28. Rowinsky EK. Targeted induction of apoptosis in cancer management: the emerging role of tumor necrosis factor-related apoptosis-inducing ligand receptor activating agents. J Clin Oncol 2005;23:9394-407. [Crossref] [PubMed]
29. Varfolomeev EE, Ashkenazi A. Tumor necrosis factor: an apoptosis JuNKie? Cell 2004;116:491-7. [Crossref] [PubMed]
30. Cornell R, Vance DE. Translocation of CTP: phosphocholine cytidylyltransferase from cytosol to membranes in HeLa cells: stimulation by fatty acid, fatty alcohol, mono- and diacylglycerol. Biochim Biophys Acta 1987;919:26-36. [Crossref] [PubMed]
31. Pelech SL, Pritchard PH, Brindley DN, et al. Fatty acids promote translocation of CTP:phosphocholine cytidylyltransferase to the endoplasmic reticulum and stimulate rat hepatic phosphatidylcholine synthesis. J Biol Chem 1983;258:6782-8.
32. Ridgway ND. The role of phosphatidylcholine and choline metabolites to cell proliferation and survival. Crit Rev Biochem Mol Biol 2013;48:20-38. [Crossref] [PubMed]
33. Lin Z, Long F, Kang R, et al. The lipid basis of cell death and autophagy. Autophagy 2024;20:469-88. [Crossref] [PubMed]
34. Smith GD, Ebrahim S. 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol 2003;32:1-22. [Crossref] [PubMed]
35. Emdin CA, Khera AV, Kathiresan S. Mendelian Randomization. JAMA 2017;318:1925-6. [Crossref] [PubMed]
36. Burgess S, Davey Smith G, Davies NM, et al. Guidelines for performing Mendelian randomization investigations: update for summer 2023. Wellcome Open Res 2019;4:186. [Crossref] [PubMed]
37. Zhao Y, Quan E, Zeng T, et al. Type 1 diabetes, its complications, and non-ischemic cardiomyopathy: a mendelian randomization study of European ancestry. Cardiovasc Diabetol 2024;23:31. [Crossref] [PubMed]
38. Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ 2018;362:k601. [Crossref] [PubMed]
39. Higbee DH, Granell R, Sanderson E, et al. Lung function and cardiovascular disease: a two-sample Mendelian randomisation study. Eur Respir J 2021;58:2003196. [Crossref] [PubMed]
40. Ottensmann L, Tabassum R, Ruotsalainen SE, et al. Genome-wide association analysis of plasma lipidome identifies 495 genetic associations. Nat Commun 2023;14:6934. [Crossref] [PubMed]
41. Zhao JH, Stacey D, Eriksson N, et al. Genetics of circulating inflammatory proteins identifies drivers of immune-mediated disease risk and therapeutic targets. Nat Immunol 2023;24:1540-51. [Crossref] [PubMed]
42. Zhao J, Ming J, Hu X, et al. Bayesian weighted Mendelian randomization for causal inference based on summary statistics. Bioinformatics 2020;36:1501-8. [Crossref] [PubMed]
43. Carter AR, Sanderson E, Hammerton G, et al. Mendelian randomisation for mediation analysis: current methods and challenges for implementation. Eur J Epidemiol 2021;36:465-78. [Crossref] [PubMed]
44. Bowden J, Davey Smith G, Haycock PC, et al. Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator. Genet Epidemiol 2016;40:304-14. [Crossref] [PubMed]
45. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 2015;44:512-25. [Crossref] [PubMed]
46. Verbanck M, Chen CY, Neale B, et al. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 2018;50:693-8. [Crossref] [PubMed]
47. Ma J, Wang S. Algorithms, applications, and challenges of protein structure alignment. Adv Protein Chem Struct Biol 2014;94:121-75. [Crossref] [PubMed]
48. van Kempen M, Kim SS, Tumescheit C, et al. Fast and accurate protein structure search with Foldseek. Nat Biotechnol 2024;42:243-6. [Crossref] [PubMed]
49. Luty BA, Wasserman ZR, Stouten PFW, et al. A molecular mechanics/grid method for evaluation of ligand–receptor interactions. J Comput Chem 1995;16:454-64.
50. Morris GM, Goodsell DS, Halliday RS, et al. Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. J Comput Chem 1998;19:1639-62.
51. Friedrich R, Steinmetzer T, Huber R, et al. The methyl group of N(alpha)(Me)Arg-containing peptides disturbs the active-site geometry of thrombin, impairing efficient cleavage. J Mol Biol 2002;316:869-74. [Crossref] [PubMed]
52. Jorgensen WL, Chandrasekhar J, Madura JD, et al. Comparison of simple potential functions for simulating liquid water. J Chem Phys 1983;79:926-35.
53. Van Der Spoel D, Lindahl E, Hess B, Groenhof G, Mark AE, Berendsen HJ. GROMACS: fast, flexible, and free. J Comput Chem 2005;26:1701-18. [Crossref] [PubMed]
54. Tian Lu, Sobtop, Version [1.0(dev5)]. (accessed on 15 January 2025). Available online: http://sobereva.com/soft/Sobtop
55. Nosé S. A unified formulation of the constant temperature molecular dynamics methods. J Chem Phys 1984;81:511-9.
56. Hoover WG. Canonical dynamics: Equilibrium phase-space distributions. Phys Rev A Gen Phys 1985;31:1695-7. [Crossref] [PubMed]
57. Parrinello M, Rahman A. Polymorphic transitions in single crystals: A new molecular dynamics method. J Appl Phys 1981;52:7182-90.
58. Darden T, York D, Pedersen L. Particle mesh Ewald: An N·log(N) method for Ewald sums in large systems. J Chem Phys 1993;98:10089-92.
59. Hess B, Bekker H, Berendsen HJC, et al. LINCS: A linear constraint solver for molecular simulations. J Comput Chem 1997;18:1463-72.
60. Lindahl E, Abraham MJ, Hess B, et al. GROMACS 2019 Source code. Zenodo 2018. Available online: https://doi.org/10.5281/zenodo.2424363
61. Lu T, Chen Q. Independent gradient model based on Hirshfeld partition: A new method for visual study of interactions in chemical systems. J Comput Chem 2022;43:539-55. [Crossref] [PubMed]
62. Lu T, Chen F. Multiwfn: a multifunctional wavefunction analyzer. J Comput Chem 2012;33:580-92. [Crossref] [PubMed]
63. Lu T. A comprehensive electron wavefunction analysis toolbox for chemists, Multiwfn. J Chem Phys 2024;161:082503. [Crossref] [PubMed]
64. Ashkenazi A, Pai RC, Fong S, et al. Safety and antitumor activity of recombinant soluble Apo2 ligand. J Clin Invest 1999;104:155-62. [Crossref] [PubMed]
65. Valley CC. Ligand binding and receptor network formation in the tumor necrosis factor superfamily. 2012. Retrieved from the University Digital Conservancy. Available online: https://hdl.handle.net/11299/135828
66. Vanamee ÉS, Faustman DL. Structural principles of tumor necrosis factor superfamily signaling. Sci Signal 2018;11:eaao4910. [Crossref] [PubMed]
67. Putra WE, Salma WO, Rifa'i M. Anti-inflammatory activity of sambucus plant bioactive compounds against TNF-α and TRAIL as solution to overcome inflammation associated diseases: The insight from bioinformatics study. Nat Prod Sci 2019;25:215-21.
68. Truneh A, Sharma S, Silverman C, et al. Temperature-sensitive differential affinity of TRAIL for its receptors. DR5 is the highest affinity receptor. J Biol Chem 2000;275:23319-25. [Crossref] [PubMed]
69. Li L, Shu H. Molecular mechanisms of TRAIL-induced apoptosis of cancer cells. Chinese Science Bulletin 2001;46:707-8.
70. Ramamurthy V, Yamniuk AP, Lawrence EJ, et al. The structure of the death receptor 4-TNF-related apoptosis-inducing ligand (DR4-TRAIL) complex. Acta Crystallogr F Struct Biol Commun 2015;71:1273-81. [Crossref] [PubMed]
71. Cao Y, Ai M, Liu C. The impact of lipidome on breast cancer: a Mendelian randomization study. Lipids Health Dis 2024;23:109. [Crossref] [PubMed]
72. Saito RF, Andrade LNS, Bustos SO, et al. Phosphatidylcholine-Derived Lipid Mediators: The Crosstalk Between Cancer Cells and Immune Cells. Front Immunol 2022;13:768606. [Crossref] [PubMed]
73. Cui Z, Houweling M. Phosphatidylcholine and cell death. Biochim Biophys Acta 2002;1585:87-96. [Crossref] [PubMed]
74. Zheng L, Xie C, Zheng J, et al. An imbalanced ratio between PC(16:0/16:0) and LPC(16:0) revealed by lipidomics supports the role of the Lands cycle in ischemic brain injury. J Biol Chem 2021;296:100151. [Crossref] [PubMed]
75. Kendrick JE, Estes JM, Straughn JM Jr, et al. Tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) and its therapeutic potential in breast and gynecologic cancers. Gynecol Oncol 2007;106:614-21. [Crossref] [PubMed]
76. His M, Viallon V, Dossus L, et al. Prospective analysis of circulating metabolites and breast cancer in EPIC. BMC Med 2019;17:178. [Crossref] [PubMed]
77. Breeur M, Ferrari P, Dossus L, et al. Pan-cancer analysis of pre-diagnostic blood metabolite concentrations in the European Prospective Investigation into Cancer and Nutrition. BMC Med 2022;20:351. [Crossref] [PubMed]
78. von Karstedt S, Montinaro A, Walczak H. Exploring the TRAILs less travelled: TRAIL in cancer biology and therapy. Nat Rev Cancer 2017;17:352-66. [Crossref] [PubMed]
79. Holland PM, Miller R, Jones J, et al. Combined therapy with the RANKL inhibitor RANK-Fc and rhApo2L/TRAIL/dulanermin reduces bone lesions and skeletal tumor burden in a model of breast cancer skeletal metastasis. Cancer Biol Ther 2010;9:539-50. [Crossref] [PubMed]
80. Forero-Torres A, Varley KE, Abramson VG, et al. TBCRC 019: A Phase II Trial of Nanoparticle Albumin-Bound Paclitaxel with or without the Anti-Death Receptor 5 Monoclonal Antibody Tigatuzumab in Patients with Triple-Negative Breast Cancer. Clin Cancer Res 2015;21:2722-9. [Crossref] [PubMed]
81. Li X, Zhang Y, Zhang T, et al. Tafazzin mediates tamoxifen resistance by regulating cellular phospholipid composition in ER-positive breast cancer. Cancer Gene Ther 2024;31:69-81. [Crossref] [PubMed]
82. Piggott L, Silva A, Robinson T, et al. Acquired Resistance of ER-Positive Breast Cancer to Endocrine Treatment Confers an Adaptive Sensitivity to TRAIL through Posttranslational Downregulation of c-FLIP. Clin Cancer Res 2018;24:2452-63. [Crossref] [PubMed]
83. Tressler CM, Sonkar K, Cheng M, et al. Molecular effects of clinically relevant chemotherapeutic agents on choline phospholipid metabolism in triple negative breast cancer cells. Transl Oncol 2025;53:102311. [Crossref] [PubMed]
84. Fakiruddin KS, Lim MN, Nordin N, et al. Chemo-Sensitization of CD133+ Cancer Stem Cell Enhances the Effect of Mesenchymal Stem Cell Expressing TRAIL in Non-Small Cell Lung Cancer Cell Lines. Biology (Basel) 2021;10:1103. [Crossref] [PubMed]
85. Quiroz-Reyes AG, Delgado-González P, Islas JF, et al. Oxaliplatin Enhances the Apoptotic Effect of Mesenchymal Stem Cells, Delivering Soluble TRAIL in Chemoresistant Colorectal Cancer. Pharmaceuticals (Basel) 2023;16:1448. [Crossref] [PubMed]
86. Psahoulia FH, Drosopoulos KG, Doubravska L, et al. Quercetin enhances TRAIL-mediated apoptosis in colon cancer cells by inducing the accumulation of death receptors in lipid rafts. Mol Cancer Ther 2007;6:2591-9. [Crossref] [PubMed]
87. Sezgin E, Levental I, Mayor S, et al. The mystery of membrane organization: composition, regulation and roles of lipid rafts. Nat Rev Mol Cell Biol 2017;18:361-74. [Crossref] [PubMed]
88. Ouyang W, Yang C, Zhang S, et al. Absence of death receptor translocation into lipid rafts in acquired TRAIL-resistant NSCLC cells. Int J Oncol 2013;42:699-711. [Crossref] [PubMed]
89. Song JH, Tse MC, Bellail A, et al. Lipid rafts and nonrafts mediate tumor necrosis factor related apoptosis-inducing ligand induced apoptotic and nonapoptotic signals in non small cell lung carcinoma cells. Cancer Res 2007;67:6946-55. [Crossref] [PubMed]
90. Marconi M, Ascione B, Ciarlo L, et al. Constitutive localization of DR4 in lipid rafts is mandatory for TRAIL-induced apoptosis in B-cell hematologic malignancies. Cell Death Dis 2013;4:e863. [Crossref] [PubMed]
91. Naval J, de Miguel D, Gallego-Lleyda A, et al. Importance of TRAIL Molecular Anatomy in Receptor Oligomerization and Signaling. Implications for Cancer Therapy. Cancers (Basel) 2019;11:444. [Crossref] [PubMed]
92. Moulin M, Carpentier S, Levade T, et al. Potential roles of membrane fluidity and ceramide in hyperthermia and alcohol stimulation of TRAIL apoptosis. Apoptosis 2007;12:1703-20. [Crossref] [PubMed]
93. Liao HJ, Tsai HF, Wu CS, et al. TRAIL inhibits RANK signaling and suppresses osteoclast activation via inhibiting lipid raft assembly and TRAF6 recruitment. Cell Death Dis 2019;10:77. [Crossref] [PubMed]
94. Figarska SM, Gustafsson S, Sundström J, et al. Associations of Circulating Protein Levels With Lipid Fractions in the General Population. Arterioscler Thromb Vasc Biol 2018;38:2505-18. [Crossref] [PubMed]
95. Michowitz Y, Goldstein E, Roth A, et al. The involvement of tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) in atherosclerosis. J Am Coll Cardiol 2005;45:1018-24. [Crossref] [PubMed]
96. Larsen JB, Jensen MB, Bhatia VK, et al. Membrane curvature enables N-Ras lipid anchor sorting to liquid-ordered membrane phases. Nat Chem Biol 2015;11:192-4. [Crossref] [PubMed]
97. Mongkolsapaya J, Grimes JM, Chen N, et al. Structure of the TRAIL-DR5 complex reveals mechanisms conferring specificity in apoptotic initiation. Nat Struct Biol 1999;6:1048-53. [Crossref] [PubMed]