Article Text

Original research
Harnessing lipid metabolism modulation for improved immunotherapy outcomes in lung adenocarcinoma
  1. Yang Chen,
  2. Yu Zhou,
  3. Ran Ren,
  4. Yu Chen,
  5. Juan Lei and
  6. Yongsheng Li
  1. Department of Medical Oncology, Chongqing University Cancer Hospital, Chongqing, Chongqing, China
  1. Correspondence to Professor Yongsheng Li; lys{at}cqu.edu.cn

Abstract

Background While anti-programmed cell death protein-1 (PD-1) monotherapy has shown effectiveness in treating lung cancer, its response rate is limited to approximately 20%. Recent research suggests that abnormal lipid metabolism in patients with lung adenocarcinoma may hinder the efficacy of anti-PD-1 monotherapy.

Methods Here, we delved into the patterns of lipid metabolism in patients with The Cancer Genome Atlas (TCGA)-lung adenocarcinoma (LUAD) and their correlation with the immune microenvironment’s cellular infiltration characteristics of the tumor. Furthermore, the lipid metabolism score (LMS) system was constructed, and based on the LMS system, we further performed screening for potential agents targeting lipid metabolism. The mechanism of MK1775 was further validated using RNA sequencing, co-culture technology, and in vivo experiments.

Results We developed an LSM system and identified a potential sensitizing agent, MK1775, which targets lipid metabolism and enhances the effects of anti-PD-1 treatment. Our results demonstrate that MK1775 inhibits tumor progression by influencing lipid crosstalk between tumor cells and tumor-associated macrophages and CD8+T cells, thereby increasing the effectiveness of anti-PD-1 treatment. Further, we found that MK1775 inhibited the phosphatidylinositol 3-kinase(PI3K)/AKT/mammalian target of rapamycin (mTOR) signaling pathway, which on one hand downregulated FASN-mediated synthesis of fatty acids (FAs) to inhibit fatty acid oxidation of tumor-associated macrophages, and on the other hand, promoted IRF-mediated secretion of CXCL10 and CXCL11 to facilitate the infiltration of CD8+ T cells.

Conclusions These findings emphasize the important role of lipid metabolism in shaping the complex tumor microenvironment. By manipulating the intricate intricacies of lipid metabolism within the tumor microenvironment, we can uncover and develop promising strategies to sensitize immunotherapy, potentially revolutionizing cancer treatment approaches.

  • Immune Checkpoint Inhibitor
  • Lung Cancer
  • Macrophage
  • Tumor microenvironment - TME

Data availability statement

No data are available.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • Abnormal lipid metabolism has been demonstrated to promote the formation of a tumor immunosuppressive microenvironment, which in turn limits the efficacy of immunotherapy. It is therefore anticipated that targeting lipid metabolism will reverse the immunosuppression of the tumor microenvironment and sensitize the efficacy of immunotherapy.

WHAT THIS STUDY ADDS

  • A lipid metabolism scoring system was constructed for patients with lung adenocarcinoma, after which drug screening was performed using the Genomics of Drug Sensitivity in Cancer database. The results indicated that MK1775 was the most promising drug for sensitizing the efficacy of anti-PD-1 by targeting lipid metabolism.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • This study represents the first demonstration that MK1775 can enhance the efficacy of anti-PD-1 therapy by targeting lipid metabolism to remodel the tumor immune microenvironment. The combination of MK1775 with anti-PD-1 may be effective in increasing the response rate of anti-PD-1 in patients with lung adenocarcinoma with abnormal lipid metabolism.

Background

Despite significant advancements in treatment, lung cancer continues to be the leading cause of cancer-related deaths.1 2 Anti-programmed cell death protein-1 (PD-1) therapy has emerged as a ray of hope, showing promising results as a first-line treatment option for patients with advanced non-small cell lung cancer (NSCLC).3 However, the response rate to immune checkpoint inhibitors (ICIs) remains modest at only 28%, and the issue of resistance in NSCLC persists as a significant concern.4 Various biomarkers, including programmed death-ligand 1 (PD-L1) expression,5 mutational burden,6 mismatch repair defects,7 infiltrating lymphocytes,8 and inflammatory gene expression,9 have been extensively studied to predict the effectiveness of ICIs. However, the observed outcomes have not met the expected levels of success. To address this challenge, there is an urgent need for more specific biomarkers that can accurately predict the effectiveness of ICIs. The identification of such biomarkers will greatly facilitate the development of sensitization strategies for anti-PD-1 monotherapy.

Abnormal lipid metabolism is a prominent metabolic change observed in cancer. Cancer cells use lipid metabolism to gain energy, synthesize biofilms, and enhance proliferation, invasion, and metastasis. Additionally, this alteration significantly influences the tumor microenvironment and the response to cancer therapy.10 The dysregulation of lipid metabolism in tumor cells results in an immunosuppressive phenotype, as it extensively interacts with the complex tumor immune microenvironment.11 For example, CD36 is essential for lipid uptake in tumor-associated macrophages (TAMs), which can promote tumor growth and contribute to drug resistance in the tumor microenvironment.12 CD36 also plays a crucial role in myeloid-derived suppressor cells by increasing fatty acid uptake and oxidation, which supports their immunosuppressive function.13 Moreover, abnormal lipid metabolism in tumor-infiltrating regulatory T cells significantly enhances their immunosuppressive function.14 Dysregulated lipid metabolism in the tumor microenvironment (TME) inhibits the recruitment of CD8+ T cells and impairs their tumor-killing ability.15 Recent studies have suggested that abnormal lipid metabolism in patients with cancer impeded the efficacy of anti-PD-1 monotherapy.16 Therefore, targeting altered lipid metabolism pathways has emerged as a potentially promising strategy for anticancer treatment.

In this study, we conducted a comprehensive analysis of lipid metabolism patterns in patients with The Cancer Genome Atlas (TCGA)-lung adenocarcinoma (LUAD). We also investigated the correlation between these patterns and the tumor immune microenvironment to develop a lipid metabolism score (LMS) system and to identify potential sensitizer for anti-PD-1 monotherapy. Our study provides novel insights and a scientific foundation for the development of new combination immunotherapy approaches.

Methods

Data for the training and validation sets

The fragments per kilobase of transcript per million fragments mapped (FPKM) expression profiles (log-transformed), single nucleotide variant (SNV) and copy number variation (CNV) mutation information, overall survival (OS), and clinical information (including age, stage, gender, etc) of TCGA-LUAD were obtained using the R package TCGAbiolinks. A total of 497 tumor samples with both expression and survival information were selected for further analysis (table 1).

Table 1

The sample statistics of TCGA-LUAD

For model validation, we downloaded four sets of expression profile data from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/): GSE42127_GPL6884, GSE68465_GPL96, GSE72094_GPL15048, and GSE50081_GPL570. In the GEO data set, the probes were converted into symbols based on the probe correspondence of each platform. Probes that corresponded to multiple genes were removed, and if multiple probes corresponded to the same symbol, the median value was taken.

To extract clinical and transcriptomic data for a group of patients with urothelial carcinoma who were treated with the PD-L1 blocker atezolizumab, we used the R package IMvigor210 Core Biologies. Clinical data and transcriptomic data (GSE135222) for a cohort of patients with NSCLC who were treated with a PD-L1 blocker were downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/geo/), the patient information is attached to online supplemental table S1.

Supplemental material

Consistent clustering analysis based on lipid metabolism genes

We conducted an unsupervised cluster analysis using the R package ConsensusClusterPlus to identify distinct patterns of lipid metabolism. We used the expression profile data of lipid metabolism genes to classify patients for further analysis. To ensure the reliability of our classification, we performed 100 replications.

Analysis of TME cell infiltration

We quantified the ratio of immune cell infiltrates using three different methods and compared the distribution among different sample groups using a Wilcoxon test. The relative abundance of each cell infiltration in the tumor microenvironment was estimated using the single-sample Gene Set Enrichment Analysis (ssGSEA) algorithm. The gene set for each immune cell type in the TME was obtained from published data, which included 28 human immune cell types such as activated CD8+ T cells, activated dendritic cells, and macrophages. The ssGSEA analysis calculated the enrichment score, representing the relative abundance of TME-infiltrating cells in each sample. The CIBERSORT method, in conjunction with the LM22 feature matrix, estimated the proportion of 22 cell phenotypes in the samples. The sum of the proportions of all estimated immune cell types for each sample equaled 1. Additionally, the xCell method in the R package IOBR was used to calculate the proportion of 64 immune cell infiltrates.

Lipid metabolism score

To evaluate the lipid metabolism pattern of individual patients with lung adenocarcinoma, a scoring system called LMS was developed. The process involved constructing the scoring system and establishing its parameters.

In this study, we employed one-way Cox regression analysis to determine the HR and prognostic significance of differential genes between subgroups. Genes with a p value of <0.01 were screened as prognosis-related genes, and principal component analysis (PCA) was performed based on these genes. Principal component 1 and principal component 2 were selected as characteristic scores. This approach allows us to focus on genes with the highest correlation or anti-correlation in the gene set, while reducing the weight of contribution from other unrelated member genes in the set. The following equation illustrates this advantage:

Score=PC1+PC2.

The study classified samples into high and low groups based on the median LSM. The correlation between these groups and overall survival was then analyzed. Prognostic analysis was performed using the Kaplan-Meier method to generate survival curves, and the significance of differences was assessed using the log-rank test. The prediction accuracy of the scoring system was evaluated using the receiver operating characteristic (ROC) curve, and the area under the curve (AUC) was visualized using the R package time ROC.

Prediction of drug sensitivity

The drug IC50 values for each sample in the training set were determined using the calcPhenotype algorithm of the R package oncoPredict, which uses the GDSC V.2 cancer genomics drug sensitivity database (https://www.cancerrxgene.org/).

Antibodies and reagents

MK1775 (molecular formula C27H32N8O2), docetaxel (molecular formula C43H53NO14) and AZD7762 (molecular formula C17H19FN4O2S) were purchased from Selleck (#S1525, #S1148, #S1532 China). Antibody of PD-L1 for Flow Cytometry (#124308) was obtained from BioLegend, Anti-PD-L1 (#60475), Anti-IRF1 (#8478), Anti-FASN (#3180), Anti-mTOR (#2983), Anti-p-mTOR (#5536), Anti-AKT (#4691), Anti-p-AKT (#4060) for western blot were obtained from Cell Signaling Technology (Danvers, Massachusetts, USA). Cell counting kit-8 (CCK-8) were purchased from APExBIO (Catalog No. K1018, USA). Anti-IRX5 (#H00010265-M01), Anti-DKK1 (#H00022943-M11), Anti-ANLN (#NBP2-52908) and Anti-CCNA2 (#NBP2-34311) for immunohistochemistry were obtained from Novus Biologicals.

Cell culture

The human and mice lung cancer cell lines NCI-H1299 and LLC were sourced from the Chinese Academy of Sciences (Shanghai, China). The cells were cultured in Roswell Park Memorial Institute-1640 (RPMI-1640) and Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS) at 37°C in a 95% air and 5% CO2 atmosphere. Hypoxic conditions were achieved by exposing the cells to 1% O2, 5% CO2, and 94% N2.

CCK-8 assays

After 24 hours of transient transfection, the cells were digested, centrifuged, and counted. A cell suspension was prepared, with each group having three secondary wells containing 3,000 cells per well. The cells were then plated in 96-well plates. Subsequently, 100 µL of the cell suspension was added to each well and incubated in the incubator for 3 hours. Once the cells had adhered, MK1775 (10 µM) was added, followed by the rapid addition of 20 µL of CCK-8 solution to each well after 24 hours. The cells were further incubated for 2 hours and the optical density at 450 nm was measured.17

Flow cytometry assay

Fresh tumor tissues were mechanically dispersed and digested using type IV collagenase (1 mg/mL; catalog no. C5138, Sigma-Aldrich) for 30 min at 37°C. After dissociation, the tumor suspensions were filtered and washed with cold phosphate-buffered saline (PBS). The resulting single-cell suspension was collected for further flow cytometry analysis. The cells were then labeled with the indicated antibodies for 30 min at 4°C. Flow cytometry was performed on CytoFLEX LX platforms, and the results were analyzed using FlowJo Software V.10.8.1.

Xenograft experiments

Four-week-old male C57BL/6 mice were purchased from Beijing Vital River Laboratory Animal Technology (Beijing, China). The animals were raised in a pathogen-free environment at the Animal Laboratory of Chongqing University Cancer Hospital. All in vivo experiments were conducted with ethical approval from the Ethics Review Committee of Chongqing University Cancer Hospital (Ethical approval number: CZLS2023239-A) and followed the guidance of the Institutional Animal Care and Use Committee of Chongqing University. When the tumor diameters reached 15 mm, the mice were euthanized by CO2 inhalation. To induce tumor growth, LLC cells (3×106) were injected subcutaneously into the right scapular region of 16 mice, suspended in 150 µL PBS. Once the tumor size reached approximately 50 mm3, the mice were randomly divided into four groups, four mice in one group. The control group received 0.9% saline orally. The monotherapy group was administered MK1775 orally at a dosage of 10 mg/kg daily for 2 weeks and PD-1 intraperitoneally at a dosage of 200 µg once every 3 days for a total of five times. The combination therapy group received both monotherapy treatments. Tumor measurements were taken weekly using a caliper, and body mass was also measured. Tumor volumes were calculated using the formula V=(width2×length)/2. During the execution of the experiment, Yang Chen was responsible for the allocation, Yu Zhou was responsible for the experimental operation, and Ran Ren performed the result evaluation and data analysis. The experimental data were collected in three separate experiments and expressed as means±SD.

Oil Red O stain

The cell culture medium was removed and the cells were washed with PBS. Subsequently, they were fixed with isopropyl alcohol, stained with Oil Red O for 20 min, and observed under a microscope after hematoxylin re-staining.

Immunohistochemistry and multicolor fluorescence Immunohistochemistry

In a previous study, we have provided a detailed description of the methods used for preparing tumor samples, as well as the techniques employed for calculating staining intensity and area.18 The histological score was determined by multiplying the staining intensity and area. Final scores were established through the joint evaluation of two pathologists. A score of 0 was classified as negative (−), while weakly positive (+) was defined as 1–4 points and strongly positive (++) was represented by 6–12 points. On the other hand, mIHC (multiplex immunohistochemistry) is a technique that combines immunohistochemistry with additional multichannel fluorescent secondary antibody labeling. This is followed by tissue scanning for observation and analysis.

Oxygen consumption rates

Cells were seeded at 2×104 cells/well in 96-well plates for 3–4 hours to allow adherence to the plate. Then, they were treated with or without MK1775 overnight, the cells were changed to unbuffered assay media, and incubated in a non-CO2 incubator at 37°C for 1 hour. Oxygen consumption rates (OCR) were measured using an XF96 extracellular flux analyzer (Seahorse Bioscience) after the sequential addition of oligomycin diluted in Seahorse XFp Base media, FCCP diluted in Seahorse XFp Base media, and antimycin/rotenone diluted in Seahorse XFp Base media.

LC-MS/MS-based lipidomics

Cell lipids were extracted for liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis. The extraction process involved using a UPLC I-Class system (Waters, Milford, Massachusetts, USA) equipped with an AB SCIEX Instruments 6500 Q-TRAP mass spectrometer (Applied Biosystems, Foster City, California, USA) in a negative ionization mode. The mobile phase consisted of a mixture of methanol, water, and acetic acid (60:40:0.01, v/v/v), which was gradually adjusted to a ratio of 85:15:0.01 over a period of 30 min, and then to 100:0:0.01 over the next 5 min, with a flow rate of 200 mL/min. The instrument control and data acquisition were carried out using Analyst V.1.6 software (Applied Biosystems). For lipid identification, a minimum of six diagnostic ions and retention time were used. Quantification was based on the peak area of multiple reaction monitoring transitions and a linear calibration curve was established for each compound.

Bulk RNA sequencing and bioinformatic analysis

Control and MK1775 treated cells were prepared for the experiment. RNA samples were extracted and subjected to final data acquisition, which included sample testing and quality control. Strand-specific library construction was performed using the ribosomal RNA removal method, followed by Illumina PE150 sequencing after pooling the libraries based on the effective concentration and data output requirements. These procedures were carried out by PANOMIX Biotech. The obtained results were analyzed using RStudio and the Reactome data set.

Statistical analysis

All bioinformatics analyses were performed using R V.4.1.2. The Wilcoxon rank-sum test was employed to compare differences between two groups of samples for significance analysis of various values, including expression, infiltration ratio, and various eigenvalues. Kruskal-Wallis test was used to compare differences between multiple groups of samples. For the sake of presentation, p values>0.05 were denoted as ‘ns’, p values<0.05 were denoted as ‘*’, p values<0.01 were denoted as ‘**’, p values<0.001 were denoted as ‘***’, and p values<0.0001 were denoted as ‘****’. The survival curves were generated using the Kaplan-Meier method for prognostic analysis, and the significance of differences was determined by log-rank tests. To map the chromosomal distribution of lipid metabolism genes in 23 pairs of chromosomes, the R package RCircos was used. For mutation mapping presentation, the R package maftools was used to illustrate the mutational landscape of lipid metabolism genes. The experimental data were collected in three separate experiments and expressed as means±SD. Differences between groups were calculated using Student’s two-tailed t-test methods. All analyses were performed using RStudio or GraphPad Prism V.8 software, and a p value<0.05 was considered statistically significant.

Results

Construction of a scoring system for lipid metabolism based on lipid metabolic profiles of LUAD

97 lipid metabolism genes were obtained from previous studies.19 The expression differences between normal and tumor samples were compared, as well as among different clinical characteristic subgroups of tumor samples. The genes were ranked according to significance p value from smallest to largest, and the top 20 were selected for display. Genes such as LPGAT1, LYPLA2, and MBOAT7 were found to be significantly overexpressed in the tumor samples (online supplemental figure S1A). Additionally, genes like LPGAT1, ETNPPL, and SELENOI showed significant differential expression across different stages (online supplemental figure S1B). The mutation frequencies of lipid metabolism genes were generally low, with DGKB having the highest mutation frequency at 6%, while the remaining genes had a mutation rate of approximately 1% (online supplemental figure S1C). Our analysis did not reveal consistent insertion and deletion ratios among different lipid metabolism genes. Among the top 20 genes, LPCAT1 showed the highest insertion ratio, whereas CEPT1 had the highest deletion ratio (online supplemental figure S1D). Our findings suggest that mutations in lipid metabolism genes currently do not play a significant role in the reprogramming of tumor lipid metabolism. Nevertheless, unsupervised clustering identified two distinct lipid metabolism patterns among patients. Furthermore, gene expression in patients with different lipid metabolism patterns showed correlations with immunophenotyping (online supplemental figure S2). To investigate the biological behavior of different lipid metabolism patterns, we conducted gene set variation analysis (GSVA) enrichment analysis. Our analysis revealed that 50 pathways in the HALLMARK data set were significantly enriched in both subpopulations, although with distinct patterns. Specifically, Pattern2 exhibited enrichment in pathways such as HALLMARK DNA REPAIR, HALLMARK E2F TARGETS, HALLMARK HYPOXIA, and several others (figure 1A). In addition, we investigated the tumor microenvironmental profile of two lipid metabolism patterns. Notably, we observed significant differences in most immune cell infiltration between the subpopulations (figure 1B,C). Furthermore, Pattern1 exhibited lower stromal scores and total ESTIMATE scores compared with Pattern2 (figure 1D–F), while tumor purity was significantly higher in Pattern2 (figure 1G). However, immune scores did not show significant differences between the two subpopulations (figure 1E).

Figure 1

Lipid metabolism score construction using bioinformatics. (A) HALLMARK functional pathway activity heatmap; (B–C) 28 immune cell infiltration score distribution, (B) represents adaptive immunity, (C) represents innate immunity; (D–G) immune stroma score and tumor purity distribution. (H) KEGG functional enrichment analysis of differential genes in lipid metabolism pattern; (I) OS curve of lipid metabolism genomic phenotypes; (J) heatmap of expression distribution of lipid metabolism pattern differential genes; (K) expression of lipid metabolism genes in different genomic phenotype samples. ALK, anaplastic lymphoma kinase; ECM, extracellular matrix; EGFR, epidermal growth factor receptor; IL, interleukin; KRAS, kirsten rat sarcoma; MDSC, myeloid-derived suppressor cell, KEGG, kyoto encyclopedia of genes and genomes; OS, overall survival; WT,wild type.

To explore the biological relevance of genes linked to different lipid metabolism patterns, we employed the limma R package to pinpoint 594 differential genes (DEGs) across 50 pathways that showed significant enrichment in both lipid metabolism patterns. Subsequently, kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis revealed significant enrichment of DEGs in pathways such as extracellular matrix (ECM)-receptor interaction and cell cycle (figure 1H). We then performed unsupervised clustering analysis using the 594 DEGs to classify tumor samples into distinct genomic subtypes, geneCluster1 and geneCluster2. The results aligned with the clustering grouping of lipid metabolism patterns, unveiling two distinct lipid metabolism genomic phenotypes, geneCluster1 and geneCluster2. Notably, geneCluster2 exhibited poorer overall survival (figure 1I). Moreover, we observed significant differences in the distribution of clinical characteristics, including age, stage, and gender, among patients with different genomic subtypes (figure 1J). We further compared the differences in lipid metabolism genes between the two genomic phenotypes and found that most of the lipid metabolism genes were significantly differentially expressed. The top 20 genes were arranged in ascending order of significance based on their p value (figure 1K).

One-way Cox’s analysis was conducted on 594 DEGs, resulting in the identification of 137 genes associated with prognosis (p<0.01). Subsequently, these genes were subjected to PCA to calculate the LMS system . Based on the median value of the score (−0.04613714), the samples were categorized into high and low score groups. The high score group had significantly shorter OS time compared with the low score group (figure 2A). The AUCs of the samples at 1, 3, and 5 years were 0.694, 0.644, and 0.623, respectively, indicating that the score provided good characterization. The LMS system was validated in the validation set, and the results showed that the AUCs of the GSE72094_GPL15048 data set samples were 0.696, 0.671, and 0.774 at 1, 3, and 5 years, respectively. The AUCs of the GSE42127_GPL6884 data set samples were 0.891 and 0.774 at 1, 3, and 5 years, respectively. The AUCs for the GSE50081_GPL570 data set were 0.675, 0.722, and 0.775 for 1, 3, and 5 years, respectively. All four validation sets showed consistency (figure 2B–D). The AUCs of the GSE68465_GPL96 data set samples were 0.700, 0.674, and 0.620 at 1, 3, and 5 years, respectively (online supplemental figure S3A). One-way and multi-way Cox regression analyses were conducted to assess the prognostic value of the model in comparison to other clinical factors. After adjusting for other confounding factors, single and multifactor Cox analyses revealed that the LMS remained an independent prognostic factor (figure 2E–H, online supplemental figure S3B–D). Our LMS system is not affected by important clinical observables such as epidermal growth factor receptor (EGFR), anaplastic lymphoma kinase (ALK), and kirsten rat sarcoma (KRAS) mutations (online supplemental figure S4). The study examined the correlation between LMS and the tumor immune microenvironment (online supplemental figure S5). Furthermore, immunohistochemistry staining and prognostic analysis were conducted on key molecules of LMS using 170 lung adenocarcinoma tissues. The findings revealed that DKK1, ANLN, and CCNA2 exhibited high expression levels in the tumors and were strongly linked to a poor prognosis, whereas IRX5 demonstrated low expression levels in the tumors and was significantly associated with a better prognosis (online supplemental figure S6).

Figure 2

Prognostic analysis of the LMS system. (A) TCGA training set; (B) validation set GSE72094_GPL15048; (C) validation set GSE42127_GPL6884; (D) validation set GSE50081_GPL570; (E) prognostic independence of the LMS system test of TCGA training set; (F–H) prognostic independence of the LMS system test of the validation set GSE68465_GPL96, validation set GSE72094_GPL15048, validation set GSE42127_GPL6884 and validation set GSE50081_GPL570. AUC, area under the curve; KRAS,kirsten rat sarcoma; LMS, lipid metabolism score; LUAD,lung adenocarcinoma; TCGA, the cancer genome atlas.

Screening of potential therapeutic strategies based on lipid metabolism score

We examined the predictive ability of LMS on patient prognosis in an immunocompetent cohort. While the survival curves for GSE135222 did not show statistical significance, both cohorts, GSE135222 and IMvigor210, exhibited consistency with the overall cohort and demonstrated a worse prognosis (figure 3A–F), indicating a potential association with a poor response to immunotherapy. These results suggest that the LMS could be valuable in guiding future clinical use. To predict the Genomics of Drug Sensitivity in Cancer (GDSC) database drug IC50 values for the training set samples, we used the R package oncoPredict in this study. By combining the GDSC database drug information with the expression profiles of the training set, we analyzed the spearman correlation between the LMS and each drug log2 (IC50). Our analysis revealed significant differences in the drug log2 (IC50) among different LMS (online supplemental figure S7A–D). Moreover, we identified MK1775, docetaxel, and AZD7762 as the top three drugs with significant negative correlations (p<0.05, figure 3G–H). These findings suggest that these drugs may have potential as antitumor agents to modulate tumor lipid metabolism.

Figure 3

Prognostic efficacy of LMS in an immune cohort and drug sensitivity. (A) K-M survival curves for different score subgroups of GSE135222 cohort; (B) LMS distribution for different response groups of GSE135222 cohort; (C) immunotherapy response distribution for different score subgroups; (D) K-M survival curves for different score subgroups of IMvigor210 cohort; (E) LMS distribution for different response groups of IMvigor210 cohort; (F) immunotherapy response distribution for different score subgroups of IMvigor210 cohort; (G–H) LMS-related drug sensitivity screening. LK-M,Kaplan Meier; MS, lipid metabolism score; PFS, progression-free survival.

MK1775 suppresses tumor proliferation by inhibiting tumor lipid metabolism and enhancing anti-PD-1 efficacy

Compared with the other two drugs, MK1775 was ranked as the top drug (figure 3G). The most notable changes in Oil Red O staining were observed following MK1775 treatment (online supplemental figure S7E,F, figure 4A,B), indicating a strong correlation between MK1775 and LMS. This led us to further investigate MK1775 both in vitro and in vivo. In vitro experiments revealed that MK1775 upregulated the expression of PD-L1 on the surface of tumor cells (figure 4C), suggesting a potential association with the efficacy of PD-1 monoclonal antibody. Furthermore, in vivo experiments confirmed that the inhibitory effect of MK1775 on tumor proliferation was significantly enhanced when combined with PD-1 monoclonal antibody (figure 4D,E). Notably, there was no significant difference in body weight between the control and experimental groups (figure 4F). H&E staining results indicated that PD-1 monoclonal antibody exhibited some nephrotoxicity, which was alleviated with MK1775 treatment. Additionally, MK1775 did not show significant hepatotoxicity or nephrotoxicity (figure 4G). Both MK1775 and PD-1 monoclonal antibody effectively inhibited lipid accumulation and Ki67 expression in tumors, as shown by the results of Oil Red O staining and immunohistochemistry. Notably, when used together, their inhibitory effect was significantly enhanced. Furthermore, immunohistochemistry results indicated that MK1775 upregulated PD-L1 expression in tumors (figure 4H).

Figure 4

MK1775 suppresses tumor proliferation and enhances anti-PD-1 efficacy. (A) Cell counting kit-8 assay was performed to assess the cytotoxic effects of MK1775 on H1299 and LLC cells. (B) Oil Red O experiment was conducted to detect lipid deposition in lung adenocarcinoma cells. (C) Flow cytometry analysis was used to detect the expression of PD-L1 on the surface of tumor cells. (D) Lipidomics analysis was performed to detect fatty acids, triglycerides, cholesterol, and sphingolipids. (E) Subcutaneous tumor models in C57BL/6 mice were used to evaluate the antitumor effects of MK1775 and the synergistic effects in combination with PD-1 monoclonal antibody. (F) Graph of changes in mouse body weight was plotted. (G) H&E staining was used to observe the drug side effects of MK1775 and PD-1 monoclonal antibody. (H) Histological analysis of subcutaneous tumor tissue was performed using H&E, Oil Red O and immunohistochemistry staining. PD-1, programmed cell death protein-1; PD-L1, programmed death-ligand 1.

MK1775 inhibits lipid crosstalk between tumor cells and TAMs via suppressing the PI3K/AKT/mTOR signaling in tumor cells

The results of tumor tissue multiplex immunofluorescence showed a significant reduction in the number of tumor cells and TAMs in the MK1775-treated group. There was no significant difference in the number of TAMs in the PD-1-conjugated group compared with the single-agent group, but the number of tumor cells was further reduced (figure 5A). To investigate whether MK1775 promoted lipid crosstalk between tumors and TAMs and sensitized the efficacy of anti-PD-1 monoclonal, we conducted lipid mass spectrometry on MK1775-treated H1299 cells. The results indicated that MK1775 significantly downregulated lipids including FFA, TAG, 24-OH-cholesterol, and sphingomyelin (figure 5B,C). Additionally, we performed RNA sequencing analysis on MK1775-treated H1299 cells, and the downregulated differential gene enrichment analysis revealed that the most enriched genes were related to the PI3K-AKT signaling pathway (figure 5D,E). Confirmation through western blot assay showed that MK1775 inhibited the phosphorylation activity of PI3K and downstream mTOR, AKT, as well as the expression of FASN. Furthermore, PI3K activators were able to partially reverse the inhibitory effect of MK1775 (figure 5F).

Figure 5

MK1775 inhibits lipid crosstalk between tumor cells and tumor-associated macrophages via suppressing the PI3K/AKT/mTOR signaling in tumor cells. (A) Multiple fluorescent immunohistochemical staining to demonstrate the TME in the control group and each treatment group. (B–C) Relative quantitative comparison of lipid mass spectrometry assays. (D–E) RNA sequencing analysis of control and MK1775 treatment groups. (F) Western blot assay of the PI3K/AKT signaling pathway in H1299 and LLC cells treated with MK1775 and/or 740 Y-P. AKT, protein kinase B; ECM, extracellular matrix; KEGG, kyoto encyclopedia of genes and genomes; mTOR, mechanistic target of rapamycin; PD-1, programmed cell death protein-1; PI3K, phosphoinositide 3-kinase; SM, sphingomyelin; TME, tumor microenvironment.

MK1775 remodeling the TME by downregulating PI3K/AKT/mTOR pathway-mediated fatty acid synthesis

In vivo experiments demonstrated that the tumor suppressive effect of MK1775 was partially reversed by overexpression of Fasn (figure 6A). Flow cytometry analysis of tumor-infiltrating immune cells revealed that MK1775 significantly increased the levels of CD8+ T cells, while overexpression of Fasn led to a significant decrease in CD8+ T cells. Additionally, MK1775 significantly decreased the abundance of TAMs, whereas overexpression of Fasn resulted in an upregulation of TAM accumulation in the tumors. Furthermore, the effects of MK1775 on both cytotoxic T lymphocytes (CTLs) and TAMs could be partially reversed by Fasn (figure 6B–D). ELISA assay demonstrated that MK1775 significantly decreased the expression of CXCL10, CXCL11, IFN-γ, and IL-2 in the tumors (figure 6E). In our study, we analyzed the distribution of CD8+ T cells and TAMs through multicolor staining. Our findings further supported the notion that MK1775 facilitated the influx of CD8+ T cells while decreasing TAMs (figure 6F).

Figure 6

MK1775 remodeling the TME by downregulating PI3K/AKT/mTOR pathway-mediated fatty acid synthesis. (A) Effects of MK1775 and Fasn on the proliferative viability of subcutaneous tumors observed by in vivo imaging in mouse subcutaneous tumor model. (B–D) Flow cytometry assay detected the effect of MK1775 and Fasn on the number of tumor-infiltrating CD8+ T cells and TAMs. (E) Elisa assay for IL-2, CXCL10, and CXCL11 in tumor tissues of each group after treatment with MK1775 and/or Fasn. (F) Multiple fluorescent immunohistochemical staining to demonstrate the distribution of CD8+ T cells and macrophages in subcutaneous tumors. (G) Confocal microscopic observation of lipid peroxidation levels in primary macrophages cultured in tumor cell supernatants treated with MK1775 and/or Fasn. (H) Seahorse detected OCR levels in primary macrophages after co-culture. (I) ELISA assay for CXCL10, and CXCL11 in LLC cell supernatant of each group after treatment with MK1775 and/or PI3K activator 740Y-P. (J–K) RT-qPCR and western blot for mRNA and protein levels of IRF1, respectively. AKT, protein kinase B; mTOR, mechanistic target of rapamycin; mRNA, messenger RNA; OCR, oxygen consumption rates; PI3K, phosphoinositide 3-kinase; RT-qPCR,reverse transcription quantitative real-time PCR; TAM, tumor-associated macrophages; TME, tumor microenvironment;

In our in vitro experiments, we used BODIPY-C11 to label M2-type macrophages that were co-cultured with LLC cells treated with MK1775 and/or overexpressed Fasn. We observed that MK1775 suppressed fatty acid oxidation (FAO) in TAMs, but the overexpression of Fasn partially reversed this effect (figure 6G). Furthermore, MK1775 inhibited mitochondrial oxidation (OCR) while increasing the FAO level of TAMs. However, this effect was partially reversed after overexpression of Fasn (figure 6H). Additionally, our in vitro ELISA assay demonstrated that MK1775 promoted the secretion of CXCL10 and CXCL11 in tumor cells. However, this secretion was inhibited by PI3K activation, and MK1775 reversed this inhibition (figure 6I). Reverse-transcription quantitative real-time PCR (RT-qPCR) and western blot experiments confirmed that MK1775 promoted the expression of IRF1, a transcription factor of CXCL10 and CXCL11. Interestingly, PI3K activation inhibited IRF1, but MK1775 partially restored its expression (figure 6J,K). Together these findings demonstrate that MK1775 suppresses FAO in TAMs and modulates immune cell composition in the TME.

Discussion

Lipid metabolic reprogramming plays a crucial role in the initiation and progression of cancer. In this study, we established an LMS system using TCGA expression profiles and genes related to lipid metabolism in lung adenocarcinoma. Additionally, we demonstrated the effectiveness of MK1775 in inhibiting the interaction between tumor cells and macrophages by suppressing the PI3K/AKT pathway. This, in turn, enhanced the sensitivity of PD-1 both in vitro and in vivo. Consequently, our findings provide a solid scientific foundation for the development of combination immunotherapy strategies aimed at improving the overall response rate of immunotherapy in patients.

Our LMS system has identified two distinct immune microenvironments and metabolic patterns that correspond to different responses to immunotherapy and prognoses. Previous studies have shown that lipid metabolism produces numerous metabolites that can regulate gene expression and activate immune checkpoints through various pathways.20 Lipids play a crucial role in signaling as second messengers or hormones.21 In a previous study, a prognostic profile was generated using six fatty acid metabolism-related genes that were associated with overall survival in breast cancer through Lasso Cox analysis. This highlights the potential prognostic implications of fatty acid metabolism.22 Furthermore, lipid metabolic features have been significantly associated with prognosis in bladder cancer,23 hepatocellular carcinoma,24 colorectal cancer,25 and gastric cancer.26 The characterization of lipid metabolic profiles not only predicts the prognosis of patients with various tumors but also provides a reference for drug development.27 However, the current drug development targeting lipid metabolism in cancer cells is inadequate.

To address this challenge, we first distinguished the TCGA-LUAD lipid metabolism pattern using 97 known lipid metabolism genes. Then, we pathway-enriched samples from both lipid metabolism patterns, performed unsupervised clustering of 594 DEGs between the two patterns, and constructed LMS by the 137 DEGs with prognostic value. Spearman correlation was used to correlate the LMS with the log2 (IC50) of each drug. Potential immunotherapeutic sensitizing drugs targeting lipid metabolism were screened, suggesting the possibility that they may enhance immune infiltration by modulating lipid metabolism, thereby increasing the sensitivity of immunotherapy. Our analysis indicates that MK1775 is the most effective drug for tumor suppression while also promoting immune infiltration.

Previous studies have shown that MK1775 effectively inhibits the proliferation and stemness characteristics of multiple myeloma stem-like cells by inducing Mus81-Eme1 endonuclease-mediated DNA damage and apoptosis during the S-phase cycle.28 Furthermore, MK1775 has been found to increase the sensitivity of KRAS-mutated NSCLC to sorafenib,29 and has shown preclinical synergy with gemcitabine in clinical trials for refractory recurrent ovarian cancer.30 While the impact of MK1775 on lipid metabolism modulation has not been previously documented, our research reveals that combining MK1775 with anti-PD-1 treatment effectively regulates lipids and eliminates tumors in tumor cells. Moreover, previous studies have shown that TAMs’ lipid metabolism directly influences their immunosuppressive function.11 Of interest, our findings demonstrate that MK1775 can further hinder the lipid communication between tumor cells and TAMs, leading to a suppression of TAMs’ FAO level. As a result, MK1775 emerges as a potential candidate for targeting lipid metabolism to enhance the efficacy of PD-1.

In conclusion, lipid metabolic reprogramming is a critical characteristic of cancer. Our current study developed an LMS system to accurately forecast the therapeutic response of ICIs. Using the LMS system, we discovered that MK1775 has the potential to be an effective agent for combination therapy in patients who do not respond to or are resistant to PD-1 monotherapy. Our findings provide insight into a new strategy that uses modulation of lipid metabolism to enhance the effectiveness of cancer immunotherapy.

Supplemental material

Data availability statement

No data are available.

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.

Acknowledgments

We would like to express our gratitude to Professor Hongfei Yan for her invaluable guidance and support throughout the validation process of the Lipid Metabolism Score model.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors Conception and design: Yang Chen, YL. Development of methods: Yang Chen, YZ, JL, RR, YL. Acquisition of data (provision of animals, acquisition and management of patients, provision of facilities, etc): Yang Chen, YL. Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Yang Chen, YZ, RR, YL. Writing, reviewing, and/or revising manuscripts: Yang Chen, YL. Administrative, technical, or material support (ie, reporting or organizing data, constructing databases): Yang Chen, YL. Research supervision and guarantor: YL.

  • Funding The work was supported by National Natural Science Foundation of China Youth Science Foundation Program (No. 82303505). Chongqing Postdoctoral Special Funding (No. 2022CQBSHTB3069), the Major International (Regional) Joint Research Program of the National Natural Science Foundation of China (No. 81920108027), Senior Medical Talents Program of Chongqing for Young and Middle-aged (2020GDRC005). Founding of Chongqing University Innovation Group, and Funding for Chongqing Young and Middle-Aged Medical Excellence Team. Research Capacity Enhancement Project of Chongqing University Cancer Hospital (Y121).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.