Genetically predicted inflammatory protein mediate the association between lipidome and colon cancer

Abstract

Background

Previous studies show correlations between colon cancer, dyslipidemia, and inflammatory proteins. However, the exact causal relationships are still unclear and no comprehensive analysis exists.

Methods

We used a two-sample Mendelian Randomization (MR) strategy to assess the link between lipidomes and colon cancer risk. Bayesian Weighted Mendelian Randomization (BWMR) and MR-robust adjusted profile score (MR-RAPS) were employed to validate our findings. Furthermore, inverse MR analysis was conducted to investigate the possibility of reverse causation. Through a comprehensive two-step MR analysis, we identified the mediating effect of inflammatory factors. Various sensitivity analyses were carried out to ensure the reliability and robustness of our results.

Results

Our MR analysis highlighted negative correlations between Phosphatidylcholine (O-16:1_18:0) (OR, 95% CI 0.771, (0.610, 0.975)), Phosphatidylethanolamine (O-16:1_22:5) (OR, 95% CI 0.788, (0.664, 0.934)), Phosphatidylethanolamine (O-18:1_18:2) (OR, 95% CI 0.825, (0.708, 0.960)) and colon cancer. In contrast, we found positive correlations with Sterol ester (27:1/18:3) (OR, 95% CI 1.278, (1.068, 1.530)), Phosphatidylcholine (18:0_22:5) (OR, 95% CI 1.322, (1.118, 1.563)), Triacylglycerol (46:1) (OR, 95% CI 1.272, (1.047, 1.546)), Triacylglycerol (46:2) (OR, 95% CI 1.236, (1.034, 1.478)) and colon cancer. Furthermore, mediation analysis revealed that a fraction of the impacts of Phosphatidylcholine (18:0_22:5) and Phosphatidylethanolamine (O-16:1_22:5) on colon cancer were mediated by TNF-related apoptosis-inducing ligand levels.

Conclusion

This study presents evidence supporting potential causal correlations between 7 lipidome levels and colon cancer, based on Mendelian randomization analyses. Additionally, inflammatory factors mediate some of these effects.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-03367-5.

Introduction

Colon cancer remains a prevalent malignancy that presents substantial challenges to public health [1]. Understanding its underlying causes and mechanisms is crucial for enhancing prevention and treatment strategies. A noteworthy aspect of colon cancer is its association with significant alterations in cellular lipidome profiles. These changes impact not only energy balance but also the integrity and functionality of biological membranes. Moreover, lipidome dysregulation marks a common characteristic of cancer cells [2, 3]. Proliferative cancer cells require various lipidomes, such as phospholipidomes, cholesterol, and sphingolipidomes, for membrane synthesis [4–6]. Additionally, lipidomes serve essential roles as energy sources, particularly triglycerides. Research indicates that lipidome levels in colon cancer tissues differ markedly from those in normal colonic mucosa. Increasing evidence points to the possibility that colon cancer induces changes in lipidome levels, which could influence prognosis and recovery outcomes [7, 8]. Investigating the mechanisms behind lipidome alterations during cancer progression may unveil vital insights into colonic cell transformation and present new preventative or therapeutic targets.

Past research has examined the connection between specific lipidome types and colon cancer, demonstrating through both in vitro and in vivo experiments that treatment with sodium palmitate (C16:0) can mitigate cancer-associated fibroblast-induced membrane fluidity alterations, hinder cell invasiveness, and diminish tumor growth and dissemination [9]. Elevated total cholesterol levels correlate with increased colon cancer risk (OR = 1.15 [1.01, 1.31], IVW p value = 0.029), while reverse mendelian randomization analysis suggests that colon cancer itself may raise lipid levels [10].

Furthermore, changes in blood lipid levels closely relate to inflammatory protein levels [11]. Both in vivo and in vitro research suggests that lipidome levels might modulate the pro-carcinogenic inflammatory milieu in the colon [12, 13]. Lipidome characteristics such as low-density lipoprotein cholesterol (LDL-C), apolipoprotein B (ApoB), and triglycerides (TG) are known to escalate the risk of inflammation [14, 15]. Consequently, inflammatory proteins could act as intermediaries between lipidomes and colon cancer.

A Mendelian randomization (MR) approach was employed to address the challenges posed by acquired confounders and reverse causation. This MR investigation examined the relationship between specific exposures, as predicted by genetic variants, and their associated outcomes. Given that an individual’s genetic makeup remains unchanged post-conception, utilizing genetic variants as instrumental variables (IVs) significantly enhances the robustness of the findings. Additionally, leveraging publicly accessible genetic datasets, this approach can be broadly implemented across various studies. This framework enables for the application of MR to explore causal relationships between lipidomes and colon cancer. In this study, we conducted a bidirectional MR analysis complemented by mediation analysis, utilizing summary statistics from the most extensive and recent genome-wide association studies (GWAS) on lipidomes, inflammatory proteins, and colon cancer to elucidate their interconnections.

Methods

GWAS data sources

The dataset for plasma lipidome stems from both univariate and multivariate genome-wide analyses involving 179 lipids (Figure S1). This analysis included 7174 Finnish participants from the GeneRISK cohort [16]. For colon cancer, the dataset comes from the UK Biobank, featuring 456,348 individuals of British ancestry, alongside 11,842,647 SNPs [17].The GWAS for inflammatory proteins encompasses 11 cohorts with 14,824 individuals of European ancestry. This research also leverages genome-wide protein quantitative trait locus (pQTL) studies focused on 91 plasma proteins [18], detailed information was provided in Table S2. The GWAS data utilized in this study was derived from diverse tissues, thus guaranteeing no sample duplication.

Instrument variables selection

In the process of selecting IVs, a two-step approach was employed. Initially, single nucleotide polymorphisms (SNPs) associated with lipidomes and inflammatory proteins were chosen based on a significance threshold of P < 1 × 10⁻⁵, consistent with previous MR investigations [13, 18]. Subsequently, SNPs related to colon cancer were identified using the GWAS threshold of P < 5 × 10⁻⁸. Independent SNPs (kb = 10,000, r² < 0.001) were then filtered based on the linkage disequilibrium (LD) structure within European populations. Prior to the MR analysis, exposure SNPs directly correlated with the outcome were excluded by referencing the PhenoScanner website (http://www.phenoscanner.medschl.cam.ac.uk/). SNPs with F-statistics < 10 was also removed to address potential weak instrument bias.

Univariate MR analysis

The schematic summary we analyzed is illustrated in Fig. 1. Initially, we conducted a two-sample bidirectional MR study to explore the reciprocal causation between the plasma lipidomes and colon cancer. The univariate MR analysis was conducted utilizing the “Two Sample MR” package in R software version 4.2.0. To examine the causal relationship between urate levels and infertility, four distinct methodologies were applied: inverse variance weighted (IVW), MR-Egger, weighted median, and weighted model. Also, to control for false positive results during multiple hypothesis testing, we adopted a false discovery rate (FDR) to control for statistical bias due to multiple comparisons. The MR-Egger intercept and MR-PRESSO test were employed to evaluate the presence of horizontal pleiotropy. In our quest to ensure that our research findings are as robust as possible and to minimize the risk of false positives (Type I errors) in our discoveries, we’ve integrated the MR-robust adjusted profile score (MR-RAPS) method [19]. This method is facilitated by the mr.raps toolkit (version 0.4.1). Additional analyses were carried out to ensure the reliability of the results. The BWMR method was utilized to account for the uncertainty arising from weak effects due to polygenicity. This approach also tackled issues of pleiotropy violations by identifying outliers through Bayesian weighting. If the MR analysis indicates variability associated with the outcome surpassing that of the exposure, it could potentially lead to misleading correlations. Furthermore, a reverse Mendelian randomization analysis was performed to assess the potential causal association between colorectal cancer and lipidomes, aligning with the forward MR analysis methodology.

Fig. 1.

The design of MR study. The ‘×’ means that genetic variants are not associated with confounders or cannot be directly involved in outcome but via the expo-sure pathway. The ‘√’ means that genetic variants are highly correlated with exposure. Solid paths are significant; dashed paths should not exist in the MR study. SNP single nucleotide polymorphism;

Mediation analysis

To explore the potential involvement of inflammatory proteins in mediating the causal association between plasma lipidome and colon cancer, we performed a mediation MR analysis utilizing a two-step methodology (Fig. 1) [35]. First, we examined the causal effect of lipidomes on the mediator (β1). Subsequently, we assessed the causal effect of the mediator on colorectal cancer (β2). The significance of the mediation effect (β1*β2) and the proportion of the mediation effect relative to the total effect were determined using the delta method.

Sensitivity analysis

Heterogeneity among individual SNP estimates was examined using the Cochran Q-test. Statistically significant results indicated substantial heterogeneity in the analysis. Additionally, the MR-Egger intercept test was employed to assess horizontal pleiotropy of SNPs.

Result

The association between lipidomes and colon cancer

Our study aimed to investigate the associations between selected instrumental variables and colon cancer using univariate MR methods. After screening 179 lipidomes with selected IVs, we identified potential causal connections. The IVW analysis suggested a potential causal link between 10 lipidomes and colon cancer, with P < 0.05. Following a robust RAPS analysis, 3 lipidomes were further excluded due to lack of statistical significance, ultimately identifying 7 lipidomes with a demonstrable causal relationship to colon cancer. Noteworthy findings included Phosphatidylcholine (O-16:1_18:0), Phosphatidylethanolamine (O-16:1_22:5), and Phosphatidylethanolamine (O-18:1_18:2) showing a negative association with colon cancer. Conversely, Sterol ester (27:1/18:3), and various Triacylglycerols displayed positive correlations. Phosphatidylcholine (18:0_22:5) exhibited the highest impact on colon cancer risk, with an odds ratio of 1.322 (95% CI 1.118, 1.563, P = 0.001, P_fdr=0.012) (Table 1; Fig. 2, Table S3). We also utilized additional MR methods, including MR Egger, Weighted median, Simple mode, and Weighted mode, to ensure robust analysis (Table 1). Consistency between these methods and IVW enhanced the reliability of the identified relationships (Table 1). BWMR further supported the IVW outcomes (Figure S1). Sensitivity analyses confirmed the robustness of our findings, indicating no evidence of pleiotropy (Table 1).

Table 1.

MR results of the causal effect of lipidomes on colon cancer

Exposures	No. of SNPs	Method	OR (95% CI)	P value		Heterogeneity test		Pleiotropy test	MR-Presso Global-test
						Cochran’s Q (I2)	P value	Egger Intercept	P value
Sterol ester (27:1/18:3)	18	MR Egger	1.076 (0.732, 1.581)	0.715		15.256	0.506	0.029	0.087
	18	IVW	1.278 (1.068, 1.530)	0.007		16.236	0.507
	18	Weighted median	1.331 (1.032, 1.717)	0.029
	18	Simple mode	1.323 (0.876, 2.003)	0.204
	18	Weighted mode	1.345 (0.939, 1.928)	0.124
	18	MR-RAPS	1.269 (1.070, 1.538)	0.005
Phosphatidylcholine (18:0_22:5)	27	MR Egger	1.357 (0.954, 1.930)	0.102		32.712	0.138	− 0.004	0.051
	27	IVW	1.322 (1.118, 1.563)	0.001		32.749	0.170
	27	Weighted median	1.200 (0.950, 1.515)	0.126
	27	Simple mode	1.414 (0.883, 2.266)	0.161
	27	Weighted mode	1.238 (0.970, 1.580)	0.098
	27	MR-RAPS	1.323 (1.182, 1.591)	0.002
Phosphatidylcholine (O-16:1_18:0)	16	MR Egger	0.945 (0.573, 1.558)	0.829		17.379	0.237	− 0.029	0.073
	16	IVW	0.771 (0.610, 0.975)	0.030		18.394	0.242
	16	Weighted median	0.792 (0.577, 1.087)	0.148
	16	Simple mode	0.676 (0.368,1.241)	0.226
	16	Weighted mode	0.730 (0.405,1.318)	0.313
		MR-RAPS	0.766 (0.613, 0.972)	0.017
Phosphatidylethanolamine (O-16:1_22:5)	17	MR Egger	0.770 (0.568, 1.043)	0.112		16.786	0.332	0.006	0.066
	17	IVW	0.788 (0.664, 0.934)	0.006		16.823	0.397
	17	Weighted median	0.772 (0.613, 0.971)	0.027
	17	Simple mode	0.816 (0.588, 1.131)	0.240
	17	Weighted mode	0.797 (0.588, 1.081)	0.164
	17	MR-RAPS	0.784 (0.668, 0.932)	0.002
Phosphatidylethanolamine (O-18:1_18:2)	28	MR Egger	0.809 (0.589, 1.112)	0.203		23.341	0.614	0.003	0.039
	28	IVW	0.825 (0.708, 0.960)	0.013		23.359	0.666
	28	Weighted median	0.786 (0.623, 0.990)	0.041
	28	Simple mode	0.756 (0.511, 1.119)	0.174
	28	MR-RAPS	0.828 (0.710, 0.965)	0.017
Triacylglycerol (46:1)	20	MR Egger	1.209 (0.831, 1.758)	0.335		21.808	0.241	0.008	0.075
	20	IVW	1.272 (1.047, 1.546)	0.016		21.928	0.288
	20	Weighted median	1.251 (0.950, 1.646)	0.111
	20	Simple mode	1.191 (0.784, 1.811)	0.422
	20	Weighted mode	1.285 (0.907, 1.820)	0.174
	20	MR-RAPS	1.271 (1.045, 1.542)	0.036
Triacylglycerol (46:2)	21	MR Egger	0.979 (0.672, 1.425)	0.911		16.226	0.642	0.041	0.049
	21	IVW	1.236 (1.034, 1.478)	0.020		18.141	0.578
	21	Weighted median	1.167 (0.914, 1.492)	0.216
	21	Simple mode	1.276 (0.846, 1.926)	0.259
	21	Weighted mode	0.996 (0.667, 1.487)	0.984
	21	MR-RAPS	1.238 (1.144, 1.339)	0.015

MR Mendelian randomization, SNPs single nucleotide polymorphisms, IVW inverse variance weighted, OR odds ratio, CI confidence interval

Fig. 2.

Forest plot of the causal effects of lipidomes with colon cancer

In the reverse MR analysis, where we examined lipid levels as outcomes and colon cancer as the exposure, no significant causal associations were observed (Table S4).

Inflammatory protein screening

Utilizing the IVW method, we conducted an analysis to identify inflammatory protein that may have causal relationships with nine specific lipidomes. Our findings indicate that out of the 91 inflammatory proteins examined, 22 showed statistically significant potential causal associations with lipidomes (P < 0.05) (Table S5). Among the 22 identified inflammatory proteins, we pinpointed two that are specifically linked to colon cancer. TNF-related apoptosis-inducing ligand (OR 0.855, 95% CI 0.743–0.984, P = 0.029) exhibited a negative correlation with colon cancer (Table 3). Notably, both the IVW tests and MR-Egger regression Q statistics did not reveal any significant heterogeneity in the results (P > 0.05). Moreover, the MR-Egger intercept result (P > 0.05) suggests an absence of horizontal pleiotropy in the data (Table 2).

Table 3.

MR analyses of the causal effects of lipidome levels on colon cancer through inflammatory protein

Exposure— mediator—outcome	Total effect	Direct effect A	Direct effect B	Mediation effect
	β1	β2	β3	β	Proportion mediated
Phosphatidylcholine (18:0_22:5)—TNF-related apoptosis-inducing ligand levels—Colon cancer	0.279	− 0.095	− 0.157	0.015	0.053
Phosphatidylethanolamine (O-16:1_22:5)—TNF-related apoptosis-inducing ligand levels— Colon cancer	− 0.193	0.086	− 0.157	− 0.013	0.070

‘Total effect’ indicates the effect of lipidome levels on colon cancer, ‘Direct effect A’ indicates the effect of lipidome levels on inflammatory proteins, ‘Direct effect B’ indicates the effects of inflammatory proteins on colon cancer and ‘Mediation effect’ indicates the effect of lipidome levels on colon cancer through inflammatory proteins. Total effect, Direct effect A and Direct effect B were derived by IVW; mediation effect was derived by using the product of coefficients method. All statistical tests were two-sided

Table 2.

MR analyses of the causal effects between inflammatory protein and colon cancer

Exposure	Method	No. of SNPs	Odds ratio(95%CI)	P	Heterogeneity test		Pleiotropy test
					Cochran’s Q (I2)	P value	Egger Intercept
TNF-related apoptosis-inducing ligand	MR Egger	35	0.924 (0.737, 1.158)	0.497	26.964	0.110	− 0.016
	Weighted median	35	0.842 (0.680, 1.044)	0.117
	IVW	35	0.855 (0.743, 0.984)	0.029	27.701	0.116
	Simple mode	35	0.845 (0.573, 1.247)	0.402
	Weighted mode	35	0.862 (0.707, 1.052)	0.154

The mediating role of inflammatory protein in colon cancer

In our initial analysis, we have established a causal relationship between 7 lipidomes and colon cancer. We have conducted a preliminary examination of TNF-related apoptosis-inducing ligand levels that may link lipids to colon cancer. To elucidate the mechanisms by which these inflammatory proteins mediate this association, we have performed a two-step MR analysis.

In the first step, we have assessed the causal effects of lipidome levels on potential mediating inflammatory proteins. We have found that high levels of Phosphatidylcholine (18:0_22:5) (IVW β = −0.095, 95% CI −0.128 to −0.062, P = 0.004) and Phosphatidylethanolamine (O-16:1_22:5) (IVW β = 0.086, 95% CI 0.048 to 0.124, P = 0.023) are associated with reduced levels of TNF-related apoptosis-inducing ligand. (Table 3; Fig. 3).

Fig. 3.

MR analyses show causal effects of lipidome levels on colon cancer through inflammatory protein

In the second step, we have investigated how these inflammatory proteins causally influence colon cancer. Notably, we have observed a significant impact of TNF-related apoptosis-inducing ligand levels (IVW β = −0.157, 95% CI −0.228 to −0.086, P = 0.029) on the cancer outcome (Table 3; Fig. 3).

Lastly, we have examined how lipids indirectly affect colon cancer through the inflammatory protein. The mediation effect of Phosphatidylcholine (18:0_22:5) was 0.015 (95% CI 0.008–0.023, P = 0.012), accounting for a 5% mediation proportion. For Phosphatidylethanolamine (O-16:1_22:5), the indirect effect was − 0.013 with a 7% mediation proportion. (Table 3; Fig. 3).

Discussion

Our study contributes significant insights into the impact of different lipidomes on the incidence of colon cancer. We have identified 7 lipid species that may have causal links to the development of colon cancer. Through a two-step mediation analysis, we have uncovered potential underlying mechanisms. Notably, our study reveals TNF-related apoptosis-inducing ligand that are associated with colon cancer, Phosphatidylcholine (18:0_22:5) appears to promote colon cancer by reducing levels of TNF-related apoptosis-inducing ligand, while Phosphatidylethanolamine (O-16:1_22:5) demonstrates protective qualities by increasing these levels.

Our study has identified a robust causal link between certain lipid species and colon cancer, with 7 out of 179 lipids exhibiting a significant association with the disease. Extensive documentation supports the notion that distinct lipid profiles, alongside conventional risk factors, may play a pivotal role in the etiology of colon cancer [20–22]. Our investigation has unveiled a significant positive correlation between specific triglycerides and the development of colon cancer. Hypertriglyceridemia may lead to an increase in pro-inflammatory cytokines such as IL-6 and TNF-α, while simultaneously reducing anti-inflammatory cytokines like IL-10. This inflammatory response could result in DNA damage and impact the growth, apoptosis, and proliferation of colon tumor cells [23]. The negative correlation between phosphatidylcholine and colon cancer is attributed to its anti-inflammatory effects in the colon [24]. These findings suggests that certain lipid molecules are posited to contribute to inflammation, which are pivotal process in the pathogenesis of colon cancer, by engaging specific molecular pathways. Therapy that modulates lipid metabolism presents innovative avenues for the prevention, treatment, and management of tolerance in colon cancer. Evidence suggests that statin utilization is correlated with a diminished risk of colorectal cancer incidence [25]. Moreover, the targeted delivery of rotenone lipids to colon cancer cells has been shown to markedly restore chemosensitivity to FOLFOX, thereby potentiating its antitumor efficacy [26]. Additionally, the targeted inhibition of lipid metabolism genes, such as CYP19A1, has been demonstrated to enhance the response to immunotherapy in colon cancer [27]. Our study underscores the potential for employing targeted lipid modulation strategies, predicated on lipid profiles that exhibit a causal relationship with colon cancer, to retard the progression of the disease. This strategy is in line with the current trend of transitioning to precision medicine, which emphasizes the customization of treatment based on individual genetic background, environmental factors, and lifestyle differences. Furthermore, the exploration of lipid-modulating drug targets underscores the significant potential for tailored interventions in lipid management within the realm of personalized medicine.

Our investigation demonstrates that the inflammatory protein, TNF-related apoptosis-inducing ligand, play a significant role in the initiation and progression of colon cancer through lipidome interactions. Infections and inflammation contribute to approximately 25% of carcinogenic factors [28]. Prolonged inflammation leads to the generation of ROS/RNS, causing DNA damage and promoting cancer development. Inflammatory components are crucial in this regulatory process [29]. Colon cancer, especially in patients with inflammatory bowel disease (IBD), exemplifies inflammation-induced carcinogenesis [30]. TNF-related apoptosis-inducing ligand, a member of the TNF superfamily, can induce apoptosis in tumor cells by binding to its receptor [29], demonstrating confirmed tumor-suppressive activity in colon cancer [31, 32]. Lipids may influence colorectal cancer progression by modulating the production of pro-inflammatory cytokines, such as IL-6 and TNF-α, which create a tumor-promoting microenvironment [33, 34]. Specifically, certain lipid species can activate signaling pathways like NF-κB, leading to chronic inflammation that is strongly associated with cancer development [35]. Additionally, altered lipid metabolism may enhance the recruitment and activation of immune cells, further amplifying inflammatory responses that facilitate tumor growth and metastasis [36, 37]. In hepatocellular carcinoma, the consumption of phosphatidylcholine appears to influence TNF-related apoptosis-inducing ligand-induced apoptosis [38]. The intricate interplay between lipidome levels and inflammatory processes remains an area for further exploration, particularly given the diverse roles of various lipid types and inflammatory mediators. Elucidating this dynamic could yield novel insights into the mechanisms underlying colon cancer, potentially paving the way for innovative interventions aimed at addressing lipid dysregulation and inflammation.

Several notable limitations are present in our study. The standard threshold for genome-wide significance for SNPs is typically set at P < 5 × 10⁻⁸. However, in our analysis of lipidome, we utilized a more lenient threshold of P < 1 × 10⁻⁵. While this may enhance the detection of statistically significant SNPs, it also raises the risk of false positives. To address this concern, we have employed rigorous statistical methodologies, including MR-RAPS and BWMR, to minimize the impact of spurious associations. Additionally, our focus on populations of European descent limits the exploration of genetic variations among diverse ethnic groups, countries, and regions. Variations in dietary habits across different regions may introduce relevant confounding factors in the lipidome. The modest mediation effects (5% and 7%) may be influenced by limited sample size or measurement variability, warranting validation in larger cohorts. The mediation role of TRAIL was inferred statistically and requires experimental validation (e.g., in vitro or in vivo models). Our analysis was restricted to a limited number of SNPs, highlighting the necessity for more extensive GWAS studies to strengthen the reliability of Mendelian randomization endeavors.

Conclusion

In conclusion, this study presents a comprehensive investigation into the causal relationships between lipidome, inflammatory protein, and colon cancer based on Mendelian randomization analyses. These results underscore the importance of elucidating the mechanisms connecting lipidomes and colon cancer, offering new insights for the role of inflammatory proteins and lipidome in colon cancer diagnostic program and treatment.

Author contributions

SJW mainly designed and performed analysis, verifying data and wrote the manuscript; JXY, ST, GYN performed experiments and analysis. All authors have read, provided critical feedback on intellectual content and approved the final manuscript.

Funding

This study was supported by the Natural Science Foundation of Zhejiang Province (LQ23H030001).

Data availability

The data underlying this article are available in the article and in its online supplementary material.

Declarations

Ethics approval and consent to participate

The manuscript does not contain clinical studies or patient data.

Competing interests

The authors declare no competing interests.