Causal Relationship Between Plasma Lipidome and Six Types of Cancer: A Mendelian Randomization Study

doi:10.21203/rs.3.rs-4075562/v1

Background: The plasma lipidome is intricately associated with cancer. However, the causal relationship between them remains uncertain. Therefore, this study employs Mendelian randomization (MR) based on genetic principles to investigate the potential causal relationship between plasma lipidome and six common types of cancer.

Methods: MR analysis utilizes publicly available genetic data, employing a genome-wide association study (GWAS) of 179 lipid species as exposure and GWAS datasets of six different cancers as outcomes. The inverse variance weighted (IVW) method serves as the primary approach, with MR-Egger regression and weighted median (WM) method employed as supplementary methods for analysis. Additionally, sensitivity analyses including Cochran's Q test, MR-Egger intercept test, and leave-one-out analysis are conducted to assess the reliability and stability of causal relationships. The Steiger test is also utilized to determine the directionality of causal relationships.

Results: The IVW analysis reveals that phosphatidylethanolamine (16:0_20:4) levels and others are implicated as risk factors for hepatic cancer, while sphingomyelin (d40:1) levels and others are identified as protective factors against hepatic cancer. Sterol ester (27:1/20:4) levels and others are associated with increased risk of lung cancer, whereas sterol ester (27:1/18:2) levels and others are associated with decreased risk of lung cancer. Sterol ester (27:1/20:3) levels and others are identified as risk factors for colorectal cancer, whereas phosphatidylcholine (18:2_0:0) levels and others are protective against colorectal cancer. Phosphatidylcholine (16:0_20:4) levels and others are linked to increased risk of esophageal cancer, while phosphatidylcholine (16:0_18:3) levels and others are associated with protection against esophageal cancer. Phosphatidylinositol (18:0_20:4) levels and others are identified as risk factors for thyroid cancer, whereas phosphatidylinositol (16:0_18:2) levels and others are protective against thyroid cancer. Diacylglycerol (18:1_18:2) levels and others are identified as protective factors against breast cancer.

Conclusions: There exists a clear causal relationship between plasma lipidome and six types of cancer. Additionally, it has been observed that the same single-nucleotide polymorphisms (SNPs) serve as instrumental variables (IVs), influencing cancer through the plasma lipidome. This provides further avenues and methodologies for early screening and effective treatment of cancer.

Cancers

Plasma lipidome

Mendelian randomization

Causal inference

Single-nucleotide polymorphism

Cancer represents a significant global public health concern. According to statistics from the American Cancer Society[1], in 2021, cancer accounted for 17% of all deaths, being a primary cause of death for females aged 40–79 and males aged 60–79. Lung cancer remains the leading cause of cancer-related deaths annually. Esophageal cancer ranks as the sixth most common cause of cancer deaths worldwide. During the period of 2015–2019, there was an annual increase of 2%-3% in the incidence rate of hepatic cancer among females, 1%-2% increase in the incidence rate of colorectal cancer among individuals under 55 years old, and a 0.6%-1% increase in the incidence rate of breast cancer. Between 1998 and 2020, the diagnosis rate of thyroid cancer among adolescents aged 15–19 increased by 4%-5% annually. Early detection and prevention efforts significantly reduce cancer incidence rates and prolong patient survival. Therefore, it is essential to identify appropriate methods for early screening and prevention of cancer.

Lipids are essential constituents of cells, participating in vital biological processes such as cell growth, signal transduction, formation of cell barriers, energy storage, and metabolism[2, 3]. Currently, total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), and triglycerides (TG) are commonly used clinical lipid analysis indicators. Plasma lipid concentrations are dynamic, varying with physiological and pathological states, and are correlated with tumor growth and progression[4]. Studies have conducted comparative analyses of plasma lipidome between patients with prostate cancer, renal cancer, breast cancer, and healthy controls, revealing significant differences in plasma lipidome between various cancers and healthy groups[5]. Furthermore, research suggests that plasma lipidome can serve as early screening tools for endometrial cancer, thyroid cancer, prostate cancer, colorectal cancer, gastric cancer, and breast cancer[6–11]. Correspondingly, plasma lipidome can also serve as tools for monitoring the prognosis of anticancer treatments[2, 12]. Therefore, exploring the potential causal relationship between plasma lipidome and cancer is imperative.

Mendelian randomization (MR) is a method that utilizes genetic variation as instrumental variable (IV) to infer causal relationships between exposure and outcome. It follows Mendelian genetic principles where parental alleles are randomly allocated to offspring[13], effectively avoiding potential confounding factors and reverse causality interference commonly encountered in previous observational studies. Moreover, it has been observed that the same genetic variations can influence different outcomes through the same exposure. Currently, there is still room for improvement in utilizing MR to explore the causal relationship between plasma lipidome and cancer. In this study, we delve into the causal relationships between plasma lipidome and hepatic cancer, lung cancer, colorectal cancer, esophageal cancer, thyroid cancer, and breast cancer, aiming to contribute to the early screening and treatment of these six types of cancer.

Study design

We employed MR analysis, utilizing various statistical methods to infer the causal relationship between plasma lipidome and six types of cancer, and conducted sensitivity analyses to assess the stability and reliability of the relationships. This study is based on the following three assumptions: (1) Genetic variations are closely associated with plasma lipidome, demonstrating strong correlation, (2) Genetic variations are not associated with any other confounding factors, (3) Genetic variations are not directly related to cancer. Figure 1 illustrates the specific study design.

Data sources for exposure and outcome data

Regarding the exposure factor, we utilized the recent genetic study on the plasma lipidome of 7174 Finnish individuals, which included 179 lipid species[14]. We conducted MR analysis using the genome-wide association study (GWAS) IDs (GCST90277238-GCST90277416) provided by the authors. The authors also identified 495 genome-trait associations at 56 genetic loci (including 8 novel loci). As for the outcome factors, the data for hepatic cancer (including 379 cases and 475,259 controls), lung cancer (including 3,791 cases and 489,012 controls), colorectal cancer (including 6,581 cases and 463,421 controls), esophageal cancer (including 998 cases and 475,308 controls), thyroid cancer (including 1,054 cases and 490,920 controls), and breast cancer (including 17,389 cases and 240,341 controls) were obtained from the IEU GWAS database (https://gwas.mrcieu.ac.uk/), all based on European population demographics. Since all data used in this study were obtained from publicly available databases, no additional ethical approval or informed consent was required. Detailed GWAS information for cancer is provided in Table S1.

Selection of IVs

In this MR study, due to the large sample size of single-nucleotide polymorphisms (SNPs), we initially set the genome-wide significance (GWS) threshold at a P value < 5×10^-8 to identify SNPs associated with the exposure factor. Subsequently, we set a linkage disequilibrium threshold of R² < 0.001 and kb = 10000 to identify SNPs not in linkage disequilibrium with other SNPs. Simultaneously, we extracted data such as beta values and standard errors for each SNP to calculate the F-statistic. SNPs with an F-statistic > 10 were retained to ensure the inclusion of strongly correlated SNPs with the exposure factor and to avoid errors caused by weakly correlated SNPs. The final set of SNPs obtained were then included as IVs in the MR analysis. Detailed information on the IVs used in this study is provided in Table S2.

Statistical analysis

We employed the inverse variance weighted (IVW) method as the primary approach, with MR-Egger regression and weighted median (WM) method utilized as supplementary methods for analysis. Additionally, we conducted sensitivity analyses combining multiple tests to assess the reliability and stability of the causal relationship. Cochran's Q test was used to evaluate the presence of heterogeneity for each SNP. Based on the presence or absence of heterogeneity, different IVW models were chosen. If heterogeneity was absent (i.e., if the P value > 0.05), the fixed-effect inverse variance weighted (IVW-FE) model was utilized. If heterogeneity was present, the multiplicative random effect inverse variance weighted (IVW-MRE) model was employed. The IVW model is nearly always the primary computational method in MR analysis. A P value < 0.05 in the results of IVW analysis indicates statistical significance in the causal relationship between exposure and outcome. Odds ratio (OR) and their corresponding 95% confidence interval (CI) were used to determine whether the exposure factor acted as a risk factor or a protective factor. To reduce bias due to pleiotropy, the MR-Pleiotropy residual sum and outlier (MR-PRESSO) test were utilized to identify and eliminate potential outliers in the SNP data. Furthermore, the MR-Egger intercept test was utilized to assess the presence of horizontal pleiotropy. A P value > 0.05 in this test indicates the absence of horizontal pleiotropy, signifying that there are no other confounding factors influencing the causal relationship, which is the desired outcome. To prevent excessive influence of individual SNPs on horizontal pleiotropy, we also conducted leave-one-out analysis to assess the stability of the causal relationship. By observing the forest plot results, if the CI within the range of the All value does not include 0, it indicates that there is no single SNP dominating the causal relationship. Additionally, the Steiger test was used to determine the directionality of the causal relationship. A P value < 0.05 in this test indicates the presence of a unidirectional causal relationship, thus avoiding the possibility of reverse causality.

The statistical analyses described above were conducted using the "TwoSampleMR" and "ggplot2" packages within R software version 4.2.2.

Causal relationship between plasma lipidome and hepatic cancer

As shown in Figure 2, in the MR analysis of the causal relationship between plasma lipidome and hepatic cancer, phosphatidylethanolamine (16:0_20:4) levels (OR=1.189, 95% CI=1.078-1.313, P<0.001), phosphatidylethanolamine (18:0_18:2) levels (OR=1.277, 95% CI=1.056-1.545, P=0.012), phosphatidylethanolamine (18:0_20:4) levels (OR=1.164, 95% CI=1.034-1.310, P=0.012), and sphingomyelin (d34:2) levels (OR=1.403, 95% CI=1.069-1.842, P=0.015) are associated with an increased risk of hepatic cancer. Conversely, sphingomyelin (d40:1) levels (OR=0.660, 95% CI=0.456-0.955, P=0.028), triacylglycerol (56:3) levels (OR=0.797, 95% CI=0.655-0.971, P=0.024), and triacylglycerol (56:4) levels (OR=0.798, 95% CI=0.639-0.996, P=0.046) are associated with a decreased risk of hepatic cancer. According to the Steiger test, sphingomyelin (d34:2) levels (P=0.149), triacylglycerol (56:3) levels (P=0.087), and triacylglycerol (56:4) levels (P=0.150) exhibit bidirectional causal relationships with hepatic cancer. Apart from these 7 lipid species showing causal relationships with hepatic cancer, the remaining 172 lipid species are not causally associated with hepatic cancer.

Causal relationship between plasma lipidome and lung cancer

In the MR analysis of the causal relationship between plasma lipidome and lung cancer, sterol ester (27:1/20:4) levels (OR=1.083, 95% CI=1.024-1.146, P=0.005), sterol ester (27:1/20:5) levels (OR=1.162, 95% CI=1.077-1.254, P<0.001), phosphatidylcholine (20:4_0:0) levels (OR=1.111, 95% CI=1.051-1.176, P<0.001), phosphatidylcholine (16:0_20:4) levels (OR=1.093, 95% CI=1.011-1.181, P=0.025), phosphatidylcholine (16:0_22:5) levels (OR=1.146, 95% CI=1.037-1.265, P=0.007), phosphatidylcholine (17:0_20:4) levels (OR=1.110, 95% CI=1.048-1.175, P<0.001), phosphatidylcholine (18:0_20:4) levels (OR=1.109, 95% CI=1.056-1.164, P<0.001), phosphatidylcholine (18:0_20:5) levels (OR=1.157, 95% CI=0.760-0.892, P<0.001), phosphatidylcholine (18:1_20:4) levels (OR=1.111, 95% CI=1.047-1.178, P<0.001), phosphatidylcholine (O-16:0_20:4) levels (OR=1.180, 95% CI=1.089-1.277, P<0.001), phosphatidylcholine (O-16:1_20:4) levels (OR=1.155, 95% CI=1.077-1.239, P<0.001), sphingomyelin (d32:1) levels (OR=1.139, 95% CI=1.013-1.280, P=0.029), sphingomyelin (d36:2) levels (OR=1.311, 95% CI=1.108-1.552, P=0.002), sphingomyelin (d38:2) levels (OR=1.241, 95% CI=1.045-1.472, P=0.013), and sphingomyelin (d40:2) levels (OR=1.240, 95% CI=1.076-1.430, P=0.003) were identified as risk factors for lung cancer. Conversely, sterol ester (27:1/18:2) levels (OR=0.873, 95% CI=0.763-0.998, P=0.046), diacylglycerol (18:1_18:1) levels (OR=0.878, 95% CI=0.787-0.980, P=0.021), phosphatidylcholine (18:2_0:0) levels (OR=0.644, 95% CI=0.525-0.791, P<0.001), phosphatidylcholine (15:0_18:2) levels (OR=0.823, 95% CI=0.760-0.892, P<0.001), phosphatidylcholine (16:0_18:2) levels (OR=0.863, 95% CI=0.801-0.931, P<0.001), phosphatidylcholine (16:0_18:3) levels (OR=0.740, 95% CI=0.561-0.976, P=0.033), phosphatidylcholine (16:0_20:2) levels (OR=0.876, 95% CI=0.768-0.999, P=0.049), phosphatidylcholine (16:1_18:1) levels (OR=0.807, 95% CI=0.703-0.927, P=0.002), phosphatidylcholine (16:1_18:2) levels (OR=0.856, 95% CI=0.796-0.921, P<0.001), phosphatidylcholine (17:0_18:2) levels (OR=0.816, 95% CI=0.718-0.927, P=0.002), phosphatidylcholine (18:0_18:2) levels (OR=0.844, 95% CI=0.720-0.989, P=0.036), phosphatidylcholine (18:1_18:1) levels (OR=0.774, 95% CI=0.668-0.897, P<0.001), phosphatidylcholine (18:1_18:2) levels (OR=0.847, 95% CI=0.787-0.911, P<0.001), phosphatidylcholine (18:1_20:2) levels (OR=0.882, 95% CI=0.821-0.947, P<0.001), phosphatidylcholine (18:2_18:2) levels (OR=0.803, 95% CI=0.679-0.950, P=0.011), phosphatidylethanolamine (18:0_18:2) levels (OR=0.915, 95% CI=0.854-0.981, P=0.013), phosphatidylinositol (18:1_18:2) levels (OR=0.768, 95% CI=0.606-0.973, P=0.029), sphingomyelin (d34:0) levels (OR=0.842, 95% CI=0.741-0.958, P=0.008), triacylglycerol (51:3) levels (OR=0.893, 95% CI=0.803-0.994, P=0.039), triacylglycerol (52:3) levels (OR=0.911, 95% CI=0.831-0.999, P=0.048), triacylglycerol (52:4) levels (OR=0.903, 95% CI=0.821-0.992, P=0.033), triacylglycerol (53:3) levels (OR=0.863, 95% CI=0.750-0.993, P=0.040), triacylglycerol (54:3) levels (OR=0.836, 95% CI=0.699-0.999, P=0.049), and triacylglycerol (54:4) levels (OR=0.888, 95% CI=0.805-0.978, P=0.016) were identified as protective factors against lung cancer. In the Steiger test, diacylglycerol (18:1_18:1) levels (P=0.087), phosphatidylcholine (18:2_0:0) levels (P=0.381), phosphatidylcholine (16:0_18:3) levels (P=0.429), phosphatidylcholine (16:1_18:1) levels (P=0.136), phosphatidylcholine (18:1_18:1) levels (P=0.071), phosphatidylinositol (18:1_18:2) levels (P=0.197), sphingomyelin (d36:2) levels (P=0.069), and triacylglycerol (51:3) levels (P=0.085) exhibited bidirectional causal relationships with lung cancer. Apart from these 39 lipid species showing causal relationships with lung cancer, the remaining 140 lipid species were found to have no causal relationship with lung cancer. The MR forest plot is depicted in Figure S1, while the leave-one-out analyses are illustrated in Figure S2.

Causal relationship between plasma lipidome and colorectal cancer

In the MR analysis examining the causal association between plasma lipidome and colorectal cancer, the following lipid species were identified as risk factors for colorectal cancer: sterol ester (27:1/16:0) levels (OR=1.198, 95% CI=1.063-1.350, P=0.003), sterol ester (27:1/20:3) levels (OR=1.151, 95% CI=1.034-1.281, P=0.010), sterol ester (27:1/20:4) levels (OR=1.056, 95% CI=1.002-1.114, P=0.043), sterol ester (27:1/20:5) levels (OR=1.153, 95% CI=1.088-1.223, P<0.001), phosphatidylcholine (20:4_0:0) levels (OR=1.099, 95% CI=1.011-1.194, P=0.027), phosphatidylcholine (16:0_20:4) levels (OR=1.093, 95% CI=1.040-1.149, P<0.001), phosphatidylcholine (16:0_22:5) levels (OR=1.149, 95% CI=1.085-1.218, P<0.001), phosphatidylcholine (18:0_20:5) levels (OR=1.149, 95% CI=1.084-1.217, P<0.001), phosphatidylcholine (18:1_20:3) levels (OR=1.163, 95% CI=1.025-1.320, P=0.019), phosphatidylcholine (O-16:1_20:4) levels (OR=1.128, 95% CI=1.047-1.216, P=0.002), phosphatidylcholine (O-18:1_20:4) levels (OR=1.186, 95% CI=1.098-1.280, P<0.001), and sphingomyelin (d34:2) levels (OR=1.219, 95% CI=1.023-1.453, P=0.027). Conversely, the following lipid species were identified as protective factors against colorectal cancer: phosphatidylcholine (18:2_0:0) levels (OR=0.744, 95% CI=0.589-0.939, P=0.013), phosphatidylcholine (15:0_18:2) levels (OR=0.885, 95% CI=0.790-0.992, P=0.036), phosphatidylcholine (16:1_18:1) levels (OR=0.857, 95% CI=0.752-0.976, P=0.020), phosphatidylcholine (16:1_18:2) levels (OR=0.894, 95% CI=0.822-0.972, P=0.008), and phosphatidylcholine (18:1_20:2) levels (OR=0.869, 95% CI=0.825-0.915, P<0.001). In the Steiger test, phosphatidylcholine (18:2_0:0) levels (P=0.381), phosphatidylcholine (16:1_18:1) levels (P=0.137), phosphatidylcholine (18:1_20:3) levels (P=0.193), and sphingomyelin (d34:2) levels (P=0.150) exhibited bidirectional causal relationships with colorectal cancer. Apart from these 17 lipid species showing causal relationships with colorectal cancer, the remaining 162 lipid species were found to have no causal relationship with colorectal cancer. The MR forest plot is depicted in Figure S3, while the leave-one-out analyses are illustrated in Figure S4.

Causal relationship between plasma lipidome and esophageal cancer

As shown in Figure 3, in the MR analysis investigating the causal relationship between plasma lipidome and esophageal cancer, the following lipid species were identified as risk factors for esophageal cancer: sterol ester (27:1/16:1) levels (OR=1.709, 95% CI=1.001-2.917, P=0.049), phosphatidylcholine (16:0_20:4) levels (OR=1.111, 95% CI=1.009-1.224, P=0.032), phosphatidylcholine (18:1_20:4) levels (OR=1.120, 95% CI=1.004-1.250, P=0.043), phosphatidylcholine (O-16:0_20:4) levels (OR=1.210, 95% CI=1.046-1.401, P=0.010), phosphatidylcholine (O-18:1_20:4) levels (OR=1.220, 95% CI=1.023-1.455, P=0.027), sphingomyelin (d34:2) levels (OR=1.411, 95% CI=1.101-1.809, P=0.007), sphingomyelin (d38:2) levels (OR=1.329, 95% CI=1.019-1.734, P=0.036), triacylglycerol (56:5) levels (OR=1.386, 95% CI=1.133-1.697, P=0.002), and triacylglycerol (56:8) levels (OR=1.411, 95% CI=1.071-1.859, P=0.014). Conversely, the following lipid species were identified as protective factors against esophageal cancer: phosphatidylcholine (16:0_18:3) levels (OR=0.567, 95% CI=0.339-0.947, P=0.030), phosphatidylcholine (18:0_20:3) levels (OR=0.786, 95% CI=0.623-0.991, P=0.041), phosphatidylcholine (18:1_20:2) levels (OR=0.866, 95% CI=0.765-0.981, P=0.023), and phosphatidylinositol (18:0_20:3) levels (OR=0.819, 95% CI=0.721-0.932, P=0.002). In the Steiger test, sterol ester (27:1/16:1) levels (P=0.504), phosphatidylcholine (16:0_18:3) levels (P=0.429), phosphatidylcholine (18:0_20:3) levels (P=0.070), sphingomyelin (d34:2) levels (P=0.149), triacylglycerol (56:5) levels (P=0.154), and triacylglycerol (56:8) levels (P=0.180) exhibited bidirectional causal relationships with esophageal cancer. Apart from these 13 lipid species showing causal relationships with esophageal cancer, the remaining 166 lipid species were found to have no causal relationship with esophageal cancer.

Causal relationship between plasma lipidome and thyroid cancer

As depicted in Figure 4, in the MR analysis exploring the causal relationship between plasma lipidome and thyroid cancer, phosphatidylinositol (18:0_20:4) levels (OR=1.195, 95% CI=1.021-1.399, P=0.026), sphingomyelin (d36:1) levels (OR=1.373, 95% CI=1.003-1.880, P=0.048), sphingomyelin (d38:1) levels (OR=1.241, 95% CI=1.037-1.484, P=0.018), and triacylglycerol (56:5) levels (OR=1.358, 95% CI=1.038-1.775, P=0.026) were identified as risk factors for thyroid cancer. Phosphatidylinositol (16:0_18:2) levels (OR=0.766, 95% CI=0.595-0.985, P=0.038) were recognized as protective factors against thyroid cancer. In the Steiger test, triacylglycerol (56:5) levels (P=0.154) exhibited bidirectional causal relationships with thyroid cancer. Apart from these 5 lipid species showing causal relationships with thyroid cancer, the remaining 174 lipid species were found to have no causal relationship with thyroid cancer.

Causal relationship between plasma lipidome and breast cancer

As depicted in Figure 5, in the MR analysis examining the causal relationship between plasma lipidome and breast cancer, diacylglycerol (18:1_18:2) levels (OR=0.932, 95% CI=0.877-0.991, P=0.025) and sphingomyelin (d34:0) levels (OR=0.919, 95% CI=0.857-0.985, P=0.018) were identified as protective factors against breast cancer. In the Steiger test, these two lipid species exhibited unidirectional causal relationships with breast cancer. Apart from these 2 lipid species showing causal relationships with breast cancer, the remaining 177 lipid species were found to have no causal relationship with breast cancer.

Sensitivity analyses

As shown in Table 1 and Table S3, there is heterogeneity in the causal relationship between phosphatidylethanolamine (18:0_18:2) levels and hepatic cancer. Similarly, heterogeneity exists in the causal relationship between phosphatidylcholine (18:0_18:2) levels and triacylglycerol (54:3) levels with lung cancer. Moreover, sterol ester (27:1/16:0) levels, sterol ester (27:1/20:4) levels, phosphatidylcholine (20:4_0:0) levels, phosphatidylcholine (15:0_18:2) levels, phosphatidylcholine (16:1_18:1) levels, phosphatidylcholine (16:1_18:2) levels, and sphingomyelin (d34:2) levels exhibit heterogeneity in their causal relationship with colorectal cancer. The IVW method results indicate that phosphatidylcholine (17:0_20:4) levels, phosphatidylcholine (18:0_20:4) levels, and phosphatidylcholine (18:1_20:4) levels are risk factors for colorectal cancer. However, the MR-Egger intercept test indicates that these three sets of causal relationships exhibit horizontal pleiotropy, and the P-values from MR-Egger regression and WM method are > 0.05 (Figure S3). Therefore, the causal relationship between these three lipid species and colorectal cancer is excluded[15]. The leave-one-out analyses results demonstrate that no single SNP dominates the causal relationship (Figure 2-5, S2, S4).

Table 1 The results of the Cochran's Q test, Egger intercept test, and Steiger test were used to assess the causal relationship between plasma lipidome and four types of cancer.

Outcome	Exposure	Cochran's Q test P-value	Egger intercept test P-value	Steiger test P-value
Hepatic cancer	Phosphatidylethanolamine (16:0_20:4) levels	0.682	0.499	<0.001
	Phosphatidylethanolamine (18:0_18:2) levels	0.034	0.605	<0.001
	Phosphatidylethanolamine (18:0_20:4) levels	0.363	0.299	<0.001
	Sphingomyelin (d34:2) levels	0.368	0.172	0.149
	Sphingomyelin (d40:1) levels	0.730	0.495	0.003
	Triacylglycerol (56:3) levels	0.697	0.975	0.087
	Triacylglycerol (56:4) levels	0.634	0.763	0.150
Esophageal cancer	Sterol ester (27:1/16:1) levels	0.984	0.890	0.504
	Phosphatidylcholine (16:0_18:3) levels	0.074	0.268	0.429
	Phosphatidylcholine (16:0_20:4) levels	0.629	0.320	<0.001
	Phosphatidylcholine (18:0_20:3) levels	0.425	0.416	0.070
	Phosphatidylcholine (18:1_20:2) levels	0.878	0.763	<0.001
	Phosphatidylcholine (18:1_20:4) levels	0.634	0.204	<0.001
	Phosphatidylcholine (O-16:0_20:4) levels	0.996	0.968	<0.001
	Phosphatidylcholine (O-18:1_20:4) levels	0.867	0.554	<0.001
	Phosphatidylinositol (18:0_20:3) levels	0.772	0.835	<0.001
	Sphingomyelin (d34:2) levels	0.521	0.924	0.149
	Sphingomyelin (d38:2) levels	0.669	0.535	0.002
	Triacylglycerol (56:5) levels	0.988	0.909	0.154
	Triacylglycerol (56:8) levels	0.278	0.903	0.180
Thyroid cancer	Phosphatidylinositol (16:0_18:2) levels	0.541	0.467	0.010
	Phosphatidylinositol (18:0_20:4) levels	0.400	0.080	<0.001
	Sphingomyelin (d36:1) levels	0.890	0.208	0.047
	Sphingomyelin (d38:1) levels	0.978	0.398	<0.001
	Triacylglycerol (56:5) levels	0.759	0.979	0.154
Breast cancer	Diacylglycerol (18:1_18:2) levels	0.746	0.501	0.039
	Sphingomyelin (d34:0) levels	0.598	0.305	0.016

In this study, 179 lipid species and six types of cancer were selected as exposure factors and outcome factors respectively. Following MR analysis, it was discovered that 7 lipid species were associated with hepatic cancer, 39 with lung cancer, 17 with colorectal cancer, 13 with esophageal cancer, 5 with thyroid cancer, and 2 with breast cancer. The respective risk factors and protective factors for each were determined. Some causal relationships exhibited heterogeneity, possibly attributed to differences in sample populations, sequencing methods, among other factors.

Cancer has become one of the most formidable challenges in modern medicine. Increasing evidence suggests that plasma lipidome can serve as tools for early cancer screening and diagnosis. The sustained supply of lipids is essential for the growth of cancer cells, and lipid dependency has been observed in many types of cancer. Lipids dynamically interact with cancer through influencing cell growth, signaling pathway transmission, membrane function integrity, among other mechanisms. The wasting and muscle atrophy seen in cancer cachexia patients are also closely associated with lipids. Phosphatidylethanolamine, as a crucial constituent of the plasma lipidome, holds significant implications in distinguishing the benign or malignant nature of solitary pulmonary nodules, serving as a potential biomarker for diagnosing lung cancer[16]. Phosphatidylethanolamine (18:2e/18:2), via the PI3K-Akt signaling pathway, regulates the expression level of LPAR6, mitigating cancer cachexia-associated fat loss, preventing muscle wasting, and excessive emaciation in patients[17]. Phosphatidylcholine plays a critical role in lipid-dependent signaling pathways and membrane structural composition. Studies have indicated that phosphatidylcholine (18:1/18:1) can mediate lipopolysaccharide-induced inflammation in HepG2 cells and decrease its levels through exercise, establishing an unfavorable tumor microenvironment for hepatic cancer[18]. Tumor suppressors also inhibit cancer occurrence and progression by affecting lipid biological processes. P53 influences choline metabolism and lipid droplet growth via phosphatidylcholine, exerting a certain inhibitory effect on cancer[19]. Phosphatidylcholine can further utilize its lipid properties to encapsulate botanical drugs, thereby enhancing the therapeutic efficacy of pharmaceuticals[20]. Following Wnt3a treatment, diacylglycerol and ceramide are downregulated on the membrane of hepatic cancer cells. Both can influence the Wnt signaling pathway, and their significant depletion can impede tumor cell growth and proliferation[21]. Phosphatidylethanolamine (34:2), phosphatidylcholine (34:0), phosphatidylcholine (37:1), and triacylglycerol (56:4) play intermediary roles in the relationship between bisphenol A and colorectal cancer, offering better insights into the causative factors of colorectal cancer development[22]. Further studies have shown that lipid profiles, including phosphatidylcholines and sphingolipids, are weakly negatively correlated with the risk of colorectal cancer[23]. Research has found that phosphatidylcholine (32:0), sphingomyelin (d18:0/18:1(9Z)), phosphatidylethanolamine (36:2), and phosphatidylethanolamine (36:3) levels in the blood samples of early gastric cancer patients are higher than those in healthy individuals. Moreover, after surgical resection of the lesion, the levels of phosphatidylcholine (32:0), sphingomyelin (d18:0/18:1(9Z)), and phosphatidylethanolamine (36:2) significantly decrease in the patients' bodies[24], demonstrating that measuring these lipid contents can assist in the diagnosis of early gastric cancer. Cholesterol ester, as one of the lipids, can also serve as a potential biomarker for the prognosis of endometrial cancer[25]. Lipid metabolism dysfunction is also closely associated with cancer. Research indicates that melanoma patients exhibit higher levels of sphingomyelins, dihydroceramides, and ceramides. Furthermore, resistance to BRAF inhibitors and MEK inhibitors in melanoma patients is also linked to lipid metabolism dysfunction[26]. In addition to the mentioned lipids, many lipid species causally related to cancer remain inadequately studied, warranting further exploration to broaden avenues for early cancer screening and treatment.

Moreover, we have found causal relationships between Phosphatidylethanolamine (18:0_18:2) levels and both hepatic and lung cancer. Sphingomyelin (d34:0) levels are causally related to lung and breast cancer. Phosphatidylcholine (O-18:1_20:4) levels are causally related to colorectal and esophageal cancer. Triacylglycerol (56:5) levels are causally related to esophageal and thyroid cancer. Phosphatidylcholine (16:0_18:3) levels, Phosphatidylcholine (18:1_20:4) levels, Phosphatidylcholine (O-16:0_20:4) levels, Sphingomyelin (d38:2) levels are causally related to lung and esophageal cancer. Sphingomyelin (d34:2) levels are causally related to hepatic, colorectal, and esophageal cancer. Sterol ester (27:1/20:4) levels, Sterol ester (27:1/20:5) levels, Phosphatidylcholine (18:2_0:0) levels, Phosphatidylcholine (20:4_0:0) levels, Phosphatidylcholine (15:0_18:2) levels, Phosphatidylcholine (16:0_22:5) levels, Phosphatidylcholine (16:1_18:1) levels, Phosphatidylcholine (16:1_18:2) levels, Phosphatidylcholine (18:0_20:5) levels, Phosphatidylcholine (O-16:1_20:4) levels are causally related to lung and colorectal cancer. Phosphatidylcholine (16:0_20:4) levels, Phosphatidylcholine (18:1_20:2) levels are causally related to lung, colorectal, and esophageal cancer. Additionally, the causal relationships between the same lipid species and different cancers are influenced by the same SNPs. This provides a pathway for us to identify common genetic mutation sites across various cancers and explore lipid-targeted therapies for cancer treatment.

Our study has several strengths. Firstly, we comprehensively utilized MR analysis to examine the causal relationships between plasma lipidome and six different cancers, thus avoiding potential confounding factors and reverse causality issues commonly encountered in observational studies. Secondly, lipid samples were based on a population of 7174 Finnish individuals, while the six different cancer samples were obtained from the IEU GWAS database (https://gwas.mrcieu.ac.uk/), all of which were derived from European populations, ensuring large sample sizes and similar genetic backgrounds. Lastly, we employed rigorous criteria including strict P-value thresholds, linkage disequilibrium thresholds, and F-value thresholds for IV selection, ensuring strong correlation between SNPs and lipid samples. We also utilized various analytical methods and sensitivity tests, and removed causality relationships affected by horizontal pleiotropy, ensuring the stability and reliability of the causal relationships.

However, there are still several limitations to consider. Firstly, the participants in this study were predominantly Finnish or European individuals, limiting the generalizability of the results to other populations outside of Europe. Secondly, there were relatively few cases of hepatic cancer compared to other cancers. Additionally, despite the use of multiple sensitivity analyses in this study, it is still not possible to completely eliminate confounding bias. Finally, this study was limited to data analysis and further research is needed to explore and experimentally validate these findings.

This study utilized MR to investigate the potential causal relationships between plasma lipidome and six types of cancer, identifying respective risk and protective factors and distinguishing between unidirectional and bidirectional relationships. It was also found that the same lipid species exhibited causal relationships with different cancers, influenced by identical genetic variations. These findings enhance our understanding of the relationship between plasma lipidome and cancer, providing insights for early screening and targeted cancer therapy.

MR, mendelian randomization; GWAS, genome-wide association study; IVW, inverse variance weighting; WM, weighted median; SNP, single-nucleotide polymorphism; IV, instrumental variable; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TG, triglycerides; IVW-FE, fixed-effect inverse variance weighted; IVW-MRE, multiplicative random effect inverse variance weighted; OR, odds ratio; CI, confidence interval; GWS, genome-wide significance; MR-PRESSO, MR-Pleiotropy residual sum and outlier

Acknowledgments

The study’s feasibility owes gratitude to the creators of the IEU GWAS database and the authors who diligently uploaded the data, as their valuable contributions have been instrumental.

Authors’ contributions

Conceptualization, JT and JZ; Data curation, HC; Formal analysis, RY; Funding acquisition, PZ; Investigation, HC; Methodology, JT and JZ; Project administration, PZ; Resources, HC; Software, RY; Supervision, RY and PZ; Validation, RY; Visualization, JT and JZ; Writing – original draft, JT and JZ; Writing – review & editing, JT, JZ, RY, HC and PZ. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by National Natural Science Foundation of China (82074425, 82074425), Natural Science Foundation of Hunan Province, China (2023JJ40400, 2023JJ30364, 2023JJ30448, 2021JJ40310), Key Project of Hunan Provincial Administration of Traditional Chinese Medicine (A2023042), Hunan Provincial Traditional Chinese Medicine Research Program (D2022010), Innovation Project for Postgraduate Students at Hunan University of Chinese Medi-cine (2022CX43), Hunan Province Science and Technology Top Leading Talent Project (2020173, D2022064), Hunan Provincial Graduate Research and Innovation Project (CX20230835), Graduate Human University of Traditional Chinese Innovation (2022CX43, 2022CX195, 2023CX26).

Availability of data and materials

The datasets generated and/or analyzed during the current study are available in the IEU open GWAS project (https://gwas.mrcieu.ac.uk/).

Ethics approval and consent to participate

Not applicable as all analyses in this study were performed using published genome-wide association study summary statistics. Ethical approval and informed consent have been obtained.

Consent for publication

Not applicable as all analyses in this study were performed using published genome-wide association study summary statistics.

Competing interests

The authors declare no conflict of interest.

Author details

¹School of Traditional Chinese Medicine, Hunan University of Chinese Medicine, Changsha 410208, China. ²Hunan Provincial Hospital of Integrated Traditional Chinese and Western & Cancer Research Institute of Hunan Academy of Traditional Chinese Medicine, Hunan Academy of Chinese Medicine, Changsha 410006, China.

Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin. 2024;74(1):12–49.
Yang K, Han X, Lipidomics. Techniques, Applications, and Outcomes Related to Biomedical Sciences. Trends Biochem Sci. 2016;41(11):954–69.
Skotland T, Kavaliauskiene S, Sandvig K. The role of lipid species in membranes and cancer-related changes. Cancer Metastasis Rev. 2020;39(2):343–60.
Wolrab D, Jirásko R, Chocholoušková M, Peterka O, Holčapek M, Oncolipidomics. Mass spectrometric quantitation of lipids in cancer research. TRAC Trends Anal Chem. 2019;120:115480.
Wolrab D, Jirásko R, Peterka O, Idkowiak J, Chocholoušková M, Vaňková Z, et al. Plasma lipidomic profiles of kidney, breast and prostate cancer patients differ from healthy controls. Sci Rep. 2021;11(1):20322.
Shafiee MN, Ortori CA, Barrett DA, Mongan NP, Abu J, Atiomo W. Lipidomic Biomarkers in Polycystic Ovary Syndrome and Endometrial Cancer. Int J Mol Sci. 2020;21(13):4753.
Jiang N, Zhang Z, Chen X, Zhang G, Wang Y, Pan L, et al. Plasma Lipidomics Profiling Reveals Biomarkers for Papillary Thyroid Cancer Diagnosis. Front Cell Dev Biol. 2021;9:682269.
Buszewska-Forajta M, Pomastowski P, Monedeiro F, Walczak-Skierska J, Markuszewski M, Matuszewski M, et al. Lipidomics as a Diagnostic Tool for Prostate Cancer. Cancers (Basel). 2021;13(9):2000.
Yang C, Zhou S, Zhu J, Sheng H, Mao W, Fu Z, et al. Plasma lipid-based machine learning models provides a potential diagnostic tool for colorectal cancer patients. Clin Chim Acta. 2022;536:191–9.
Liu ZC, Wu WH, Huang S, Li ZW, Li X, Shui GH, et al. Plasma lipids signify the progression of precancerous gastric lesions to gastric cancer: a prospective targeted lipidomics study. Theranostics. 2022;12(10):4671–83.
Liu L, Kawashima M, Sugimoto M, Sonomura K, Pu F, Li W, et al. Discovery of lipid profiles in plasma-derived extracellular vesicles as biomarkers for breast cancer diagnosis. Cancer Sci. 2023;114(10):4020–31.
Costantini S, Di Gennaro E, Capone F, De Stefano A, Nasti G, Vitagliano C, et al. Plasma metabolomics, lipidomics and cytokinomics profiling predict disease recurrence in metastatic colorectal cancer patients undergoing liver resection. Front Oncol. 2022;12:1110104.
Sanderson E, Glymour MM, Holmes MV, Kang H, Morrison J, Munafò MR, et al. Mendelian randomization. Nat Rev Methods Primers. 2022;2:6.
Ottensmann L, Tabassum R, Ruotsalainen SE, Gerl MJ, Klose C, Widén E, et al. Genome-wide association analysis of plasma lipidome identifies 495 genetic associations. Nat Commun. 2023;14(1):6934.
Burgess S, Thompson SG. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol. 2017;32(5):377–89.
Fahrmann JF, Grapov D, DeFelice BC, Taylor S, Kim K, Kelly K, et al. Serum phosphatidylethanolamine levels distinguish benign from malignant solitary pulmonary nodules and represent a potential diagnostic biomarker for lung cancer. Cancer Biomark. 2016;16(4):609–17.
Ding Z, Han J, Huang Q, Liu X, Sun D, Sui X, et al. Phosphatidylethanolamine (18:2e/18:2) may inhibit adipose tissue wasting in patients with cancer cachexia by increasing lysophosphatidic acid receptor 6. Nutrition. 2024;120:112356.
Zhang X, Xia Y, Cao L, Ji B, Ge Z, Zheng Q, et al. PC 18:1/18:1 mediates the anti-inflammatory effects of exercise and remodels tumor microenvironment of hepatocellular carcinoma. Life Sci. 2024;336:122335.
Xu X, Wang J, Xu L, Li P, Jiang P. p53 suppresses lipid droplet-fueled tumorigenesis through phosphatidylcholine. J Clin Invest. 2024;134(4):e171788.
Kapoor DU, Gaur M, Parihar A, Prajapati BG, Singh S, Patel RJ. Phosphatidylcholine (PCL) fortified nano-phytopharmaceuticals for improvement of therapeutic efficacy. EXCLI J. 2023;22:880–903.
Azbazdar Y, Demirci Y, Heger G, Ipekgil D, Karabicici M, Ozhan G. Comparative membrane lipidomics of hepatocellular carcinoma cells reveals diacylglycerol and ceramide as key regulators of Wnt/β-catenin signaling and tumor growth. Mol Oncol. 2023;17(11):2314–36.
Hong X, Wang G, Liu X, Wu M, Zhang X, Hua X, et al. Lipidomic biomarkers: Potential mediators of associations between urinary bisphenol A exposure and colorectal cancer. J Hazard Mater. 2022;427:127863.
Harewood R, Rothwell JA, Bešević J, Viallon V, Achaintre D, Gicquiau A, et al. Association between pre-diagnostic circulating lipid metabolites and colorectal cancer risk: a nested case-control study in the European Prospective Investigation into Cancer and Nutrition (EPIC). EBioMedicine. 2024;101:105024.
Zou L, Guo L, Zhu C, Lai Z, Li Z, Yang A. Serum phospholipids are potential biomarkers for the early diagnosis of gastric cancer. Clin Chim Acta. 2021;519:276–84.
Ayyagari VN, Li M, Diaz-Sylvester P, Groesch K, Wilson T, Pasman Z, et al. Evaluation of sterol–o–acyl transferase 1 and cholesterol ester levels in plasma, peritoneal fluid and tumor tissue of patients with endometrial cancer: A pilot study. Oncol Lett. 2023;25(6):231.
Dei Cas M, Ciniselli CM, Vergani E, Ciusani E, Aloisi M, Duroni V, et al. Alterations in Plasma Lipid Profiles Associated with Melanoma and Therapy Resistance. Int J Mol Sci. 2024;25(3):1558.

No competing interests reported.

FigureS1.tif
Additional file 1: Figure S1. Forest plot showing the causal relationship between plasma lipidome and lung cancer using the inverse variance weighted (IVW) method, MR-Egger regression, and weighted median (WM) method.
FigureS2.tif
Additional file 2: Figure S2. The leave-one-out analyses of the causal relationship between plasma lipidome and lung cancer.
FigureS3.tif
Additional file 3: Figure S3. Forest plot showing the causal relationship between plasma lipidome and colorectal cancer using the IVW method, MR-Egger regression, and WM method.
FigureS4.tif
Additional file 4: Figure S4. The leave-one-out analyses of the causal relationship between plasma lipidome and colorectal cancer.
TableS1.xlsx
Additional file 5: Table S1: The GWAS datasets for six types of cancer.
TableS2.xlsx
Additional file 6: Table S2: All instrumental variables used in Mendelian randomization analysis.
TableS3.xlsx
Additional file 7: Table S3: The results of the Cochran's Q test, Egger intercept test, and Steiger test were used to assess the causal relationship between plasma lipidome and two types of cancer.

Causal Relationship Between Plasma Lipidome and Six Types of Cancer: A Mendelian Randomization Study

Status:

Version 1

Abstract

Figures

Introduction

Materials and methods

Study design

Data sources for exposure and outcome data

Selection of IVs

Statistical analysis

Results

Discussions

Conclusions

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1