Immune cells as mediators in lipids-colorectal cancer risk: A Mendelian randomization study

Abstract

The occurrence and progression of colorectal cancer (CRC) are closely associated with abnormal lipid metabolism and immune regulation in the tumor microenvironment. Lipids are not only essential components of cell membranes but also influence CRC progression by modulating signaling pathways and inflammatory responses. In recent years, Mendelian randomization (MR) analysis, a causal inference method utilizing genetic variants as instrumental variables, has provided a novel approach to investigate the relationship between lipids and CRC risk. Immune cells in the CRC microenvironment play a dual role, either promoting antitumor immunity or accelerating tumor immune escape. However, how lipids mediate this causal pathway through immune cells remains unclear. This study aims to investigate the impact of the lipids on CRC risk and determine the potential involvement of immune cells as mediators. Utilizing single nucleotide polymorphisms as instrumental variables, a 2-sample bidirectional MR was conducted to explore the potential causal relationship between lipids and CRC. Subsequently, A 2-step MR analysis was implemented to determine the effect value of immune cell features as mediators. Heterogeneity was assessed using the Cochran Q test. MR-Egger intercept and MR-Pleiotropy RESidual Sum and Outlier Global Test were employed for evaluation pleiotropy. Excluding a lipid with reverse causal effects, 5 lipids were demonstrated to exert causal effects on CRC risk, including phosphatidylcholine (20:4_0:0), phosphatidylethanolamine (O-18:2:20:4), sphingomyelin (d38:1), sphingomyelin (d 40:2), triacylglycerol (49:2). CD80 on CD62L+ myeloid DC (mediator effect β = −0.0050), CD45 on CD33br HLA DR+ (mediator effect β = −0.0041) and Basophil %CD33dim HLA DR− CD66b− (mediator effect β = −0.0145) were involved as mediating factors. The results of our study indicated causal connections between CRC risk and multiple lipids with immune cells involved as mediators.

1. Introduction

Colorectal cancer (CRC), which ranks as the second leading cause of cancer-related mortality, is projected to impose an even greater burden on global healthcare systems by 2030, with its incidence anticipated to escalate by 60%.^[1] CRC predominantly affects individuals lacking a family history of the disease and without a genetic predisposition, with the majority of cases being sporadic. The majority of sporadic malignancies originate from benign precancerous polypoid alterations and undergo years of cumulative genetic and epigenetic modifications.^[2] Conventional treatments encompass surgery, radiation and chemotherapy. The standard chemotherapy armamentarium commonly includes 5-FU, platinum-based drugs, and irinotecan with its diverse derivatives. Recent progressions in targeted therapy and immunotherapy have instigated a paradigm shift in medical care. Nevertheless, issues such as drug resistance and a low immune response rate frequently hamper their efficacy, emphasizing the urgent need for the exploration of novel therapeutic targets.

Lipids (collectively referred to as the lipidome when considering all lipid species) constitute an essential and highly heterogeneous class of molecules that play a crucial role in cell structure, cell signaling, and bioenergetics.^[3] Based on the distinct molecular structure of lipids, they can be classified into 4 categories: glycerolipids, glycerophospholipids, sphingolipids, and sterols. CRC cells demonstrate abnormal lipid metabolism, primarily manifested by the upregulation of fat production, involving de novo fatty acid synthesis and triglyceride synthesis, augmented lipid uptake and abundance, and a general reliance on fatty acids.^[4,5] However, to date, the influence of lipid subtypes on CRC remains incompletely understood.

Multiple previous studies have indicated that lipidome induces multiple immune cell responses implicated in disease.^[6] The engagement of various immune cell subtypes plays a significant role in the progression, metastasis, and drug resistance of CRC. Hence, we hypothesized that immune cells act as mediators mediating the effects of the lipidome on CRC.

Randomized controlled studies are often limited by ethical considerations and uncontrollable external factors. Nevertheless, MR offers an alternative methodology by utilizing genetic variations as surrogates for exposure to draw causal inferences.^[7] This method provides the advantages of circumventing bias due to unobservable confounders, reverse causation, or measurement errors. With the recent released of summary-level data from genome-wide association studies on the lipidome, we were presented with the opportunity to systematically identify specific lipids that might contribute to CRC. Furthermore, by conducting mediation analysis, we were able to identify the factors that mediate the relationship between exposure and outcomes, facilitating targeted interventions to mitigate the effects of exposure by addressing these mediators.^[8]

In this research, we utilized a 2-sample MR to investigate the individual causal association between the lipidome and CRC, with a particular emphasis on the mediating role of immune cell subtypes in the progression of CRC. Our aims were to identify potential targets for intervention and enhance clinical practice.

2. Materials and Methods

2.1. Study design

MR analysis relies on meeting 3 key assumptions: strong correlation between instrumental variables (IVs) and exposure factors, independence of the single nucleotide polymorphisms (SNPs) from confounding factors, and the SNPs can affect the outcome exclusively through exposure factors. The experimental design was illustrated in Figure 1, and all data used in the experiments were approved by the institutional review committee.

Figure 1.

The design of the study was presented. Explored the causal relationship between lipdome and CRC and identified the mediated immune cells involved. CRC = colorectal cancer.

2.2. Data sources of GWASs

The IVs for lipids were obtained from a genome-wide analysis of 179 lipids in human plasma, which was previously published. The study involved 7174 individuals from the GeneRISK cohort in Finland and focused on 13 lipid classes, including 4 major lipid categories: glycerolides, glycerolipids, sphingolipids, and sterols. Genotyping and interpolation were carried out using Illumina’s HumanCoreExome BeadChip, with genotype calling done using GenomeStudio and zCall by the Finnish Institute of Molecular Medicine.^[9] CRC GWAS summary data from the FinnGen R10 release. The study utilized the “Colorectal cancer” phenotype, with a total of 6847 cases and 3,14,193 controls included.^[10] The GWAS catalog contains summary statistics for 731 immune cell subtypes (from GCST0001391 to GCST0002121) derived from 5,63,085 individuals of European descent. Imputation was carried out on a genome-wide level using a Sardinian sequence-based reference panel consisting of 3514 individuals and the Minimac58 software on pre-phased genotypes.^[11]

2.3. Acquisition of IVs

Given the genome-wide significance of SNPs for each trait and an ample number of SNPs, a significance threshold of P < 1 × 10⁻⁵ was utilized to identify potential IVs. To ensure SNP independence, linkage disequilibrium was calculated, and clumping was performed (r² < 0.001, kb < 10,000) based on data from the 1000 Genomes Project.^[12] Palindromes and fuzzy SNPs were excluded from IVs according to the Effect Attributable to the Factor.^[13] In addition, the F-value of each SNP was calculated, setting a threshold F > 10 to retain strong IVs. $F=\frac{R^2\times (\mathrm{N}-1-K)}{(1-R^2)\times K}$, where R² represents the proportion of exposure variance explained by SNPs, N represents the sample size, and K represents the number of SNPs. The SNPs for each trait were shown in Table S1, Supplemental Digital Content, https://links.lww.com/MD/P982.

2.5. Primary analysis

We conducted a bidirectional MR study to investigate the causal relationship between lipids and CRC. The IVW method is employed under the assumption that all SNPs are valid IVs, and meta-analyses are conducted to combine Wald ratios of causal effects for each SNP.^[14] Consequently, it is deemed the most reliable estimation technique. We chose the inverse variance weighted (IVW) random effects model for assessment, utilizing MR-Egger regression, weighted median, and Bayesian weighted Mendelian randomization (BWMR) as supplements to MR analysis. When over half of the weight in the analysis is derived from valid genetic instruments, the weighted median method can offer a reliable estimate.^[15] MR-Egger regression, while considering the existence of intercept terms, offers an MR estimate.^[16] BWMR addresses the issue of violating the IV hypothesis caused by pleiotropy by employing Bayesian weighted outlier detection.^[17] The Cochran Q test assessed heterogeneity, while the intercept value of MR-Egger regression and the “MR-PRESSO Global Test” evaluated pleiotropy. The “MR-PRESSO outlier test” was used to identify and remove abnormal SNPS in order to estimate the corrected results.

2.6. Two-step intermediary MR analysis

In our study, we utilized a 2-step mediation analysis to determine if immune cells act as a mediator in the pathway from lipids to CRC. The total effect of lipids on CRC is broken down into: the direct effect of lipids on CRC and the indirect effects of lipids on CRC through immune cells (a × b; total effect = direct effect + indirect effect).

2.7. Statistical analysis

The analysis was carried out in R version 3.2, with a causal relationship considered significant if the P value < .05.

3. Results

3.1. Associations of lipids with CRC risk

We utilized IVW, MR-Egger, and weighted median to assess the causal relationship between lipids and CRC risk, with IVW serving as the primary method of evaluation (Fig. 2). The outcomes demonstrated that phosphatidylcholine (20:4_0:0; IVW: OR = 1.13, 95% CI: 1.08–1.19, P = 6.51E–07), phosphatidylethanolamine (O-18:2_20:4; IVW: OR = 1.14, 95% CI: 1.03–1.25, P = .01), sphingomyelin (d38:1; IVW: OR = 1.09, 95% CI: 1.02–1.16, P = .01), sphingomyelin (d40:1; IVW: OR = 1.07, 95% CI: 1.01–1.30, P = .02), sphingomyelin (d40:2; IVW: OR = 1.10, 95% CI: 1.02–1.17, P = .01) was associated with an increased risk of CRC. Triacylglycerol (49:2; IVW: OR = 0.89, 95% CI: 0.80–1.00, P = .04) reduced the risk of CRC. We performed heterogeneity and sensitivity analyses. MR-PRESSO Global test indicated that phosphatidylethanolamine (O-18:2:20:4) MR analysis highlights the pleiotropic effects. (Table S2, Supplemental Digital Content, https://links.lww.com/MD/P982). The MR analysis was performed again after the Outlier SNP was eliminated. phosphatidylethanolamine (O-18:2:20:4; IVW: OR = 1.10, 95% CI: 1.00–1.20, P = .04) is still causally associated with CRC. In order to ensure the reliability of our MR Analysis, we use the BWMR analysis method as a supplement, and the results remain consistent (Table S3, Supplemental Digital Content, https://links.lww.com/MD/P982). Subsequently, we used MR to analyze the reverse causality of lipids in CRC. Genetic prediction showed that CRC was negatively correlated with sphingomyelin (d40:1; IVW: OR = 0.94, 95% CI: 0.88–1.00, P = .04) and had no correlation with other lipids (Table S4, Supplemental Digital Content, https://links.lww.com/MD/P982).

Figure 2.

Mendelian randomization analysis was conducted to investigate the causal effect of lipids on colorectal cancer. The forest plots display the odds ratio and 95% confidence intervals for various analysis methods.

3.2. Associations of immune cells with CRC

The composition of immune cells in CRC tumors was diverse, with 6 immune traits being associated with CRC. We utilized SNPs as IVs to study the causal regulation of immune cell traits on CRC (Fig. 3). CD20 on B cell (IVW: OR = 1.06, 95% CI: 1.01–1.11, P = .02) increased the risk of CRC. FSC-A on CD14+ monocyte (IVW: OR = 0.96, 95% CI: 0.92–1.00, P = .03), CD80 on CD62L+ myeloid DC (IVW: OR = 0.97, 95% CI: 0.94–0.99, P = .02), CD45 on CD33br HLA DR+ (IVW: OR = 0.97, 95% CI: 0.94–1.00, P = .04), Basophil %CD33dim HLA DR− CD66b− (IVW: OR = 0.94, 95% CI: 0.91–0.98, P = 3.83E-03), CD33 on CD33br HLA DR+ CD14− (IVW: OR = 0.97, 95% CI: 0.95–1.00, P = .02) reduced the risk of CRC. Cochran’s Q statistic detected pleiotropy, MR-Egger intercept and MR-PRESSO Global test detected pleiotropy, and there was no statistical significance in P value (P < .05; Table S2, Supplemental Digital Content, https://links.lww.com/MD/P982).

Figure 3.

MR analysis of the causal effect of immune cells traits on colorectal cancer. MR = Mendelian randomization.

3.3. Associations of lipids with immune cells

MR analysis revealed that 4 lipids were positively causally associated with immune cell traits (Fig. 4). Gene-predicted phosphatidylethanolamine (O-18:2_20:4)) increased the risk of CD80 on CD62L+ myeloid DC (IVW: OR = 1.16, 95% CI: 1.01–1.33, P = .03). sphingomyelin (d38:1) increased the risk of CD45 on CD33br HLA DR+ (IVW: OR = 1.14, 95% CI: 1.01–1.29, P = .04). triacylglycerol (49:2) increased the risk of Basophil %CD33dim HLA DR− CD66b− (IVW: OR = 1.28, 95% CI: 1.02–1.61, P = .03). The analysis of MR-Egger and Weighted median were consistent. Gene-predicted Phosphatidylcholine (20:4_0:0) increased the risk of CD20 on B cell (IVW: OR = 1.11, 95% CI: 1.00–1.22, P = .04) and decreased the risk of FSC-A on CD14+ monocyte (IVW: OR = 0.90, 95% CI: 0.80–1.00, P = .05) by using the IVW method. in addition, triacylglycerol (49:2) increased the risk of CD33 on CD33br HLA DR+ CD14− (IVW: OR = 1.28, 95% CI: 1.01–1.62, P = .04). But, the OR statistics of MR-Egger contradict the results of IVW. No heterogeneity or pleiotropy was detected in all (Table S2, Supplemental Digital Content, https://links.lww.com/MD/P982).

Figure 4.

Mendelian randomization analysis of the causal effect of lipids on immune cells traits.

3.4. The immune cells mediated the causal effect of lipids on CRC

In our analysis, immune cells were identified as mediators playing a role in the causal effects of lipids on CRC (Fig. 5). phosphatidylethanolamine (O-18:2_20:4) and sphingomyelin (d38:1) increased the risk of CRC, CD80 on CD62L+ myeloid DC (mediator effect β = −0.0050) and CD45 on CD33br HLA DR+ (mediator effect β = −0.0041) played negative regulatory roles as their corresponding mediators. triacylglycerol (49:2) decreased the risk of CRC. triacylglycerol (49:2) enhanced basophil %CD33dim HLA-DR-CD66b expression, which was subsequently associated with a reduced risk of CRC (mediator effect β = −0.0145).

Figure 5.

β total: total causal effect of lipids on CRC, β1: the effect of lipids on immune cell traits, β2: the effect of immune cell traits on CRC, mediator effect β: Mediated effects of lipids on CRC mediated by immune cell traits. CRC = colorectal cancer, MR = Mendelian randomization.

4. Discussion

A comprehensive MR analysis identified 6 lipids that were causally associated with CRC risk. Only sphingomyelin (d40:1) exhibited potential reverse causality. Further analysis disclosed that phosphatidylethanolamine (O-18:2:20:4), sphingomyelin (d38:1), and triacylglycerol (49:2) positively modulate specific immune cell phenotypes, thereby reducing the risk of CRC. Specifically, these lipids were found to exert an influence on CD80 on CD62L+ myeloid DC, CD45 on CD33br HLA DR+, and Basophil %CD33dim HLA DR-CD66B immune cell phenotypes.

Lipid metabolism is intensified at various stages of cancer progression, providing tumor cells with energy, modifying cell signaling and epigenetics, and adjusting cell membrane composition to facilitate cancer cell metastasis. LPCAT1, an enzyme accountable for converting lysophosphatidylcholine to phosphatidylcholine, has been discovered to augment the growth rate of the SW480 colon cancer cell line upon overexpressed.^[18] Phosphatidylethanolamine (PE), being the second richest phospholipid component, was found on the surface of mammalian cell membranes.^[19] Through exosomal lipomic sequencing, it was uncovered that there is a significant elevation in phosphatidylethanolamine (36:2) expression in CRC patients and cells when compared to normal and control cells.^[20] Sphingomyelin metabolism was associated with the development of kidney cancer, breast cancer, giant cell tumor of bone, and CRC.^[21–24] Carboxylesterase 1 enhanced triacylglycerol breakdown and facilitated the advancement of aggressive CRC.^[25] Nevertheless, the influence of phosphatidylcholine, phosphatidylethanolamine, sphingomyelin, and triacylglycerol on CRC remains unreported. We identified novel lipids implicated in colorectal carcinogenesis through the employment of MR. By utilizing this approach, we were able to minimize the effects of confounding variables and reverse causation.

Our findings suggested that the presence of CD80 on CD62L+ myeloid DC can lower the risk of CRC. Dendritic cells (DCs) serve as antigen-presenting cells, with myeloid/classical DCs (cDCs) and plasmacytoid DCs being the 2 main subtypes. cDC1s are particularly adept at processing and presenting intracellular antigens, playing a crucial role in shaping antitumor immune responses by presenting tumor-associated antigens to CD8 T lymphocytes for recognition by major histocompatibility complex class I (MHC I).^[26] cDC2s efficiently present major histocompatibility complex class II recognition to CD4 T cells, thereby promoting Th1, Th2, and Th17 polarization.^[27] DC cells express both PD-1 and CD80 as PD-L1 receptors, and by inhibiting PD-L1, B7.1/CD28 interactions can be promoted to boost T cell activation and generate a potent antitumor immune reaction.^[28]

CD45 on CD33br HLA DR+ and Basophil %CD33dim HLA DR-CD66b, which are part of the myeloid cell panel, have been shown to reduce the risk of CRC. Previous studies have shown that increased CD45 expression in primary tumor cells of CRC leads to poorer prognosis in patients receiving chemoradiotherapy, which contradicts our findings. This discrepancy may be attributed to differences in the cell types expressing CD45 molecules in CRC.^[29] In a group of ovarian cancer patients, higher levels of circulating basophils and basophils with enhanced stimulatory abilities in vitro were associated with improved patient survival outcomes.^[30] This aligns with our findings.

According to the IVW calculation method, Phosphatidylcholine (20:4_0:0) was positively correlated with CD20 on B cells and negatively correlated with FSC-A on CD14+ monocytes. However, the other 2 statistical methods, MR-Egger and weighted median, showed inconsistent directions, so we are uncertain whether there is a causal relationship between them. Arachidonic acid (AA) is often located at the sn-2 position of phosphatidylcholine. Based on the association between human peripheral blood and human methylation, AA was found to be positively correlated with DNA methylation of the PDK4 gene, and PDK4 negatively regulates the expression of MS4A1, the gene encoding CD20.^[31–33] Therefore, we speculate that phosphatidylcholine may promote CD20 expression by regulating PDK4 gene DNA methylation through AA. FSC-A can be used to measure cell size, but the relationship between phosphatidylcholine and FSC-A on CD14+ monocytes has not been reported, and whether there is a correlation remains unknown. Our study also found that sphingomyelin (d38:1) was positively correlated with CD45 on CD33br HLA DR+ cells. CD33br HLA DR+ cells are a type of myeloid-derived immune cell with antigen-presenting function. The relationship between sphingomyelin and the expression level of CD45 on their cell surface has not been reported previously. Our study is the first to identify a regulatory relationship between them through MR analysis.

In addition, there have been many previous MR analyses on the relationship between lipids and CRC, but these studies only included 4 to 5 types of lipids to investigate the causal relationship between lipids and CRC.^[34–36] Wu et al’s study included a larger number of lipids (179 in total), which is the same as our study.^[37] That study also used Bayesian colocalization analysis to examine the relationship between lipids and CRC, which we did not include, but it did not perform mediation analysis. Compared to previous studies, our study added mediation analysis to explore the mediating effect of immune cells in the regulation of CRC risk by lipids.

Nevertheless, our study has certain limitations. We did not categorize colon cancer patients based on age or sex. Additionally, in the mediation analysis, only immune cells were considered, and other potential mediating factors influencing the causal relationship between lipids and CRC were not excluded. Therefore, the validity of these findings requires additional experimental confirmation.

Author contributions

Conceptualization: Fang Yuan.

Data curation: Xu Zhang, Liping Wang, Jian Zhang, Fang Yuan.

Formal analysis: Jun Feng.

Funding acquisition: Xu Zhang, Fang Yuan.

Investigation: Xu Zhang, Xuchu Duan, Jun Feng.

Methodology: Xu Zhang, Xuchu Duan.

Project administration: Cuiping Zhang.

Resources: Liping Wang.

Software: Fang Yuan.

Writing – original draft: Xu Zhang, Fang Yuan.

Writing – review & editing: Xuchu Duan, Liping Wang, Jun Feng, Cuiping Zhang, Fang Yuan.

Supplementary Material

medi-104-e44604-s001.xlsx ^{(150.7KB, xlsx)}