Colorectal carcinoma (CRC) is a common malignant tumor of the digestive tract. The prevalence of CRC is steadily rising, especially among individuals under 50 years old
[1-3]. The exact causes and development of CRC remain incompletely elucidated, although they likely involve a complex interaction of genetic and environmental factors.
The inflammatory response is a host immune reaction initiated by tissues in response to danger signals during infection or injury. Recent studies
[4-5] indicate that inflammatory cytokines are closely linked to the development, therapeutic response, and prognosis of CRC. Interleukin (IL)-6 is a significant predictor of CRC metastasis
[4]. Elevated levels of C-reactive protein (CRP), IL-6, monocyte chemoattractant protein-1 (MCP-1), and adiponectin are significantly associated with a higher risk of all-cause mortality
[5]. Nevertheless, epidemiological data on the associations between circulating inflammatory cytokines and CRC risk remains inconsistent
[6-7]. Emerging evidence indicates that metabolic dysregulation is a hallmark of cancer development, with numerous serum metabolites closely linked to the cancerous process
[8]. Serum metabolites have the potential to act as diagnostic biomarkers for the detection of CRC, as indicated by limited data from observational epidemiological studies
[9-10]. Serum glycochenodeoxycholate, a bile acid metabolite, is positively associated with colorectal cancer in women
[9-10]. Four specific metabolites [including N-phenylacetylasparticacid, tyrosyl-gamma-glutamate, tyrosine-serine (Tyr-Ser), and sphingosine] in serum samples of CRC patients returned to healthy levels after surgery
[11].
Previous research
[4-7] has largely relied on observational studies and simple experimental models. Inherent limitations, such as measurement errors, reverse causality, and various confounding factors, cannot be overlooked in these studies. Furthermore, measuring individual cytokines and metabolites in large populations is time-consuming and expensive. Consequently, their association with CRC remains unclear due to the scarcity of studies investigating their causal relationship.
These limitations necessitate an alternative approach. Mendelian randomization (MR) is an effective approach that leverages genetic variants from large, publicly available genome-wide association studies (GWAS) as instrumental variables (IVs) to evaluate the causal effects of exposures on outcomes
[12]. Since genetic variants and alleles are randomly distributed during the development of a fertilized egg, MR analyses can significantly reduce the effects of confounding and reverse causation. The outcomes of MR are analogous to those of randomized clinical trials (RCTs) regarding the establishment of causality. MR is particularly important when conducting RCTs is challenging or unethical.
This study employed a bidirectional two-sample MR approach and large-scale datasets to investigate the causal relationship among inflammatory cytokines, serum metabolites, and CRC. We aimed to uncover the functional roles of key biomarkers in cellular processes and identify potential therapeutic targets.
1 Materials and methods
1.1 Study design
We conducted a two-sample MR analysis utilizing publicly available summary data from GWAS focused on inflammatory cytokines, serum metabolites, and CRC. Single-nucleotide polymorphisms (SNPs) associated with the levels of 91 different inflammatory cytokines were utilized as IVs to evaluate potential causal associations with CRC risk. The inflammatory cytokines significantly associated with CRC were subsequently considered exposures to investigate their causal relationships with 1 400 serum metabolites (
Figure 1). We adhered to the recommendations of the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) guidelines to conduct our MR analysis
[13]. All members of the GWAS cohort were of European descent, and the demographic characteristics of the exposure and outcome groups were non-overlapping. Ethical approval and informed consent were obtained for all the original studies. MR relies on 3 key assumptions
[14]: 1) There must be a strong correlation between the IVs and the exposure. 2) There should be no direct association between the IVs and any confounders affecting the exposure-outcome relationship. 3) The IVs should influence the outcome solely through their effect on the exposure (
Figure 2).
1.2 Sources of data
The summary statistics for 91 inflammatory cytokines were obtained from the largest available GWAS, which included 14 824 individuals of European ancestry from multiple countries, including Sweden, Denmark, the United Kingdom, Germany, Estonia, and Croatia. These data were derived from genome-wide protein quantitative trait locus (pQTL) studies. The GWAS data can be accessed from the GWAS Catalog (GCST90274758-GCST90274848)
[15].
The GWAS summary statistics for CRC were obtained from FinnGen’s Meta-analysis (R10), which included 321 040 individuals of European ancestry, consisting of 6 847 cases and 314 193 controls
[16]. The FinnGen cohort primarily comprises adults aged 18 years and older, and the database is updated annually. It is widely recognized as a leading resource for genetic research. Detailed information is available on their official website.
The GWAS summary data for 1 400 serum metabolites were sourced from the GWAS Catalog (GCST90199621-GCST90201020), comprising 8 299 individuals of European ancestry from the Canadian Longitudinal Study of Aging (CLSA). The CLSA collected physiological, biological, and medical data from over 50 000 Canadians aged 45 to 85
[17]. In that study, the metabolite data were meticulously screened to ensure that all serum metabolites originated from individuals of European ancestry. Comprehensive quality control and correlation analyses were conducted, as detailed in the original study
[18].
1.3 Screening of genetic instruments
Based on recent study
[19], the following screening guidelines are deemed reliable: SNPs associated with inflammatory cytokines from the GWAS were selected using a locus-wide significance threshold of
P<1×10
-5. Additionally, linkage disequilibrium (LD) among these SNPs was minimized using the
r2 statistic, with a threshold of
r2<0.1 and a window size of 500 kb
[20]. No weak IVs were identified, as the
F-statistic exceeded 10.
1.4 MR analyses
A total of 5 691 exposure-outcome pairs were analyzed using MR to evaluate the causal relationships among 91 inflammatory cytokines, CRC, and 1 400 serum metabolites. We primarily employed the random-effects inverse variance weighted (IVW) method as our analytical approach to summarize the relationships effectively. Among all MR methods, the IVW method is regarded as the most powerful and statistically robust
[21]. However, the IVW method assumes a consistent level of pleiotropy, relying on the validity of all IVs
[22]. This assumption can lead to bias if a significant proportion of IVs exhibit horizontal pleiotropy. To address this issue, we conducted sensitivity analyses using the Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO), MR Egger, and weighted median methods. The weighted median method is suitable when less than 50% of the IVs exhibit horizontal pleiotropy
[23], whereas the MR Egger method is applicable when more than 50% of IVs demonstrate horizontal pleiotropy
[24]. Additionally, the MR Egger intercept method was utilized to evaluate horizontal pleiotropy, with analytical conclusions deemed unreliable if
P was less than 0.05. Furthermore, we utilized the MR-PRESSO method, which measures the extent of horizontal pleiotropy, removes outliers, evaluates calibration outcomes, and assesses whether there are significant differences between the precalibrated and calibrated results
[25]. Considering that the biological effects of the SNPs were statistically significant, we decided against using PhenoScanner to exclude effects related to other potential confounders. The heterogeneity of the IVs was evaluated using Cochran’s
Q test
[26]. Odds ratio (
OR) and 95% confidence interval (
CI) were calculated. Statistical analysis was conducted with a significance threshold of
P<0.05. Analyses were performed using R version 4.3.2, along with R packages including “TwoSampleMR” and “MR-PRESSO”.
2 Results
2.1 Effect of inflammatory cytokines on CRC
Of the 91 MR analyses results (91 inflammatory cytokines analyzed in relation to CRC), 4 inflammatory cytokines were significantly associated with CRC (all
P<0.05,
Figure 3). Among these, axin-1 (
AXIN1) (
OR=0.841, 95%
CI 0.714 to 0.991) and Fms-related tyrosine kinase 3 ligand (
Flt3L) (
OR=0.916, 95%
CI 0.844 to 0.994) were identified as protective factors against CRC. In contrast, Delta/Notch-like epidermal growth factor-related receptor (
DNER) (
OR=1.119, 95%
CI 1.009 to 1.241) and vascular endothelial growth factor A (
VEGF-A) (
OR=1.078, 95%
CI 1.011 to 1.150) were determined to be risk factors.
2.2 Effect of inflammatory cytokines on serum metabolites
These 4 inflammatory cytokines were subsequently analyzed through 5 600 MR analyses involving 1 400 serum metabolites (4 exposures ×1 400 outcomes). After screening the MR results for significant correlations (
PIVW<0.05) and confirming the absence of horizontal pleiotropy (
PMR Egger intercept>0.05), a total of 144 serum metabolites were identified as significantly correlated with the inflammatory cytokines (
Figure 4 and supplementary
Figure 1;
https://doi.org/10.57760/sciencedb.30758). Among these, pyridoxal, N-carbamoyl alanine, glyco-beta-muricholate, and glycosyl ceramide (d18:1/23:1, d17:1/24:1) exhibited the strongest positive correlations with each inflammatory cytokine (
β: 0.08 to 0.28). Conversely, 4 serum metabolites showed the most significant negative correlations (
β:-0.24 to -0.10), including sphingomyelin (d18:2/24:2), 6-bromotryptophan, the adenosine 5’-diphosphate (ADP) to glycerol ratio, and 2-butenoylglycine.
Furthermore, each of the 4 serum metabolites was simultaneously associated with 2 inflammatory cytokines. These metabolites include gamma-glutamyl glycine (βAXIN1=-0.19, βDNER=0.10), 2-hydroxydecanoate (βAXIN1=0.18, βVEGF-A=-0.06), the glycine to alanine ratio (βAXIN1=-0.17, βDNER=0.09), and the 4-methyl-2-oxopentanoate to 3-methyl-2-oxobutyrate ratio (βAXIN1=0.18, βFlt3L=-0.06).
According to the sensitivity analysis, the overall direction of most relationships remained consistent across the different MR methods. Horizontal pleiotropy was not supported in any of the MR analyses, as indicated by the MR Egger intercept. The MR-PRESSO test identified no outliers, and the Cochran’s Q test showed no heterogeneity in the majority of MR analyses.
A total of 144 metabolites identified in the present study span 7 core biological classes, with their distribution patterns and key functional properties consistent with their well-established roles in cellular metabolism, signal transduction, and structural maintenance. Detailed classifications and functional annotations are as follows. 1) Amino acids and derivatives. As the most abundant class, this category encompasses canonical amino acids (e.g., glycine, tryptophan, 2-methylserine), amino acid conjugates (e.g., N-acetylputrescine, N-acetyl-2-aminooctanoate), and structurally modified amino acid derivatives (e.g., N6-methyllysine, S-methylcysteine). These metabolites serve as fundamental building blocks for protein biosynthesis, critical neurotransmitters in neural communication, and essential precursors for bioactive signaling molecules. Additionally, these metabolites play pivotal roles in cellular nitrogen metabolism and homeostasis. 2) Lipids and lipid derivatives. This class includes sphingolipids [e.g., sphingomyelin (d18:2/24:2)], glycerolipids (e.g., oleoyl-linoleoyl-glycerol), fatty acid derivatives (e.g., 9,10-DiHOME, 3-hydroxydecanoate), and bile acids (e.g., cholate, glyco-beta-muricholate). These metabolites serve as key components for maintaining the integrity and fluidity of cell membranes, efficient energy storage reservoirs, and mediating lipid-based signal transduction pathways. Notably, bile acids are indispensable for emulsification and absorption of dietary lipids in the gastrointestinal tract. 3) Carbohydrates and carbohydrate derivatives. This class includes comprising monosaccharides (e.g., glucose, mannose, xylose), sugar alcohols (e.g., glycerol), and glycoside conjugates [e.g., gamma-carboxyethyl hydroxychroma (CEHC) glucuronide]. These metabolites serve as primary energy sources to fuel cellular processes, participate in cell-cell recognition and adhesion events via glycan-mediated interactions, and act as substrates for glycosylation modifications of proteins and lipids, which are critical for regulating biomolecular stability and functionality. 4) Nucleotides and nucleotide derivatives. This class includes encompassing nucleosides (e.g., uridine, cytidine) and structurally modified nucleotides [e.g., flavin adenine dinucleotide (FAD), ADP]. These metabolites serve as essential precursors for de novo synthesis and repair of DNA and RNA, mediate intracellular energy transfer (e.g., ATP/ADP cycle) and function as coenzymes or prosthetic groups in various enzymatic reactions (e.g., FAD in redox reactions). 5) Organic acids and their derivatives. This category includes tricarboxylic acid (TCA) cycle intermediates (e.g., succinate, methylsuccinate), carboxylic acids (e.g., 5-hydroxyhexanoate, 2-hydroxydecanoate), and acylcarnitine derivatives (e.g., indoleacetoylcarnitine). These metabolites are central to mitochondrial energy metabolism through their involvement in the TCA cycle, facilitating intracellular fatty acid transport across organelle membranes (acylcarnitines). They also contribute to cellular pH regulation and acid-base homeostasis. 6) Vitamins and cofactors. This class includes comprising B-group vitamins (e.g., pantothenate, pyridoxal) and vitamin-derived bioactive molecules (e.g., thyroxine, cortisone). These metabolites act as essential coenzymes or cofactors to modulate the activity of metabolic enzymes (B-group vitamins) or regulate physiological processes through hormone-mediated signaling pathways (e.g., thyroxine in metabolic rate regulation, cortisone in anti-inflammatory responses). 7) Miscellaneous metabolites. This minor class consists of unclassified or specialized biomolecules, including X-series unknown metabolites (e.g., X-12013, X-19299) and xenobiotic compounds (e.g., ethylparaben sulfate). This category may include environmental contaminants, uncharacterized metabolic intermediates, or species-specific specialized metabolites, the biological functions of which require further investigation.
3 Discussion
This study offered novel insights into the causal relationships of inflammatory cytokines and serum metabolites with CRC risk by conducting MR analysis on 5 691 exposure-outcome associations.
We identified significant associations between 4 inflammatory cytokines and CRC. We identified
AXIN1 and
Flt3L as protective factors, corresponding to a 16% (
OR=0.84) and 8% (
OR=0.92) reduction in CRC odds, respectively. This finding is consistent with several experimental studies. AXIN1 dampens Wnt signaling, which is frequently hyperactivated through somatic mutations in CRC, and simultaneously limits tumor growth via interferon-γ/Th1-mediated immunity
[27-28]. Flt3L expands dendritic cell pools and enhances tumor-antigen cross-presentation, with higher circulating levels reported in patients exhibiting histological responses to neoadjuvant therapy and prolonged survival after resection of metastatic disease
[29-31]. Conversely, 2 inflammatory cytokines conferred increased risk.
VEGF-A has been linked to an increased likelihood of CRC (
OR=1.08). This finding aligns with previous studies
[32] in the field. VEGF-A promotes CRC not only through its role in angiogenesis and increased vascular permeability but also by significantly enhancing the motility and invasiveness of CRC cells
[32]. We also found that higher levels of
DNER were associated with an increased incidence of CRC (
OR=1.12). However, this relationship has not been extensively studied. DNER is a non-canonical Notch ligand that lacks the delta-serrate-Lag (DSL)-2 motif. Notch signaling plays a crucial role in numerous aspects of cancer biology, including angiogenesis, tumor immunity, and the maintenance of cancer stem-like cells.
DNER has been identified as an oncogene that enhances proliferation, migration, and invasion across various cancer types. Furthermore,
DNER promotes the development of gastric cancer cells and may serve as a marker of increased carcinogenesis and mortality in patients with gastric adenocarcinoma
[33]. Given the detrimental effects of DNER on gastric cancer, future studies may investigate its potential as a prognostic or therapeutic target for CRC.
Given that AXIN1 and Flt3L have been identified as protective factors for CRC, whereas DNER and VEGF-A are associated with an increased risk, we hypothesized that for exposure-outcome pairs where serum metabolites have β values in the same direction (either both positive or both negative), the serum metabolites involving these different types of inflammatory cytokines would be expected to have opposite causal effects on CRC. Using this indirect approach, we identified 72 serum metabolites as protective factors for CRC, 71 as risk factors, and 1 with an indeterminate effect. The causal relationships identified between specific serum metabolites, and CRC offer potential for developing biomarkers for early detection and risk assessment.
In this study, 4 serum metabolites demonstrated genetically determined and directionally consistent associations with CRC risk. Pyridoxal (
OR<1) and the primary bile acid glyco-β-muricholate (
OR<1) were both protective, whereas N-carbamoyl alanine and the glycosylated ceramide d18:1/23:1‒d17:1/24:1 (both
OR>1) conferred increased risk. Previous studies
[27-33]. have explored the expression and mechanisms of these metabolites in CRC. Pyridoxal, an active form of vitamin B6-suppresses angiogenesis and tumor initiation in experimental models, and higher circulating levels have been linked to better quality of life and improved survival in CRC patients
[34-35]. Glyco-β-muricholate, a comparatively under-studied bile acid, joins ursodeoxycholic acid and deoxycholic acid as a candidate tumor-inhibitory metabolite that may counteract the proliferative signaling typically attributed to the bile acid pool
[36-38]. Conversely, N-carbamoyl alanine, a peptide-like inhibitor of the antioxidant enzyme peroxiredoxin-2, could potentiate hydrogen-peroxide-mediated DNA double-strand breaks and thereby accelerate colorectal carcinogenesis
[39]. The ceramide derivative d18:1/23:1‒d17:1/24:1 marks dysregulated sphingolipid metabolism: While free ceramide induces mitochondrial apoptosis and tumor suppression, its glycosylated form (generated by ceramide glycosyltransferase) promotes tumor progression and is associated with adverse prognosis
[40-41]. Collectively, these findings highlight vitamin B6 levels, bile-acid composition, oxidative-stress buffering, and sphingolipid glycosylation as metabolically actionable pathways in CRC etiology, and nominate pyridoxal, glyco-β-muricholate, N-carbamoyl alanine, and glycosyl ceramide (d18:1/23:1‒d17:1/24:1) as candidate biomarkers or therapeutic targets warranting mechanistic and clinical follow-up.
Moreover, our MR analysis revealed significant associations between the ratio of 4-methyl-2-oxopentanoate to 3-methyl-2-oxobutyrate and both AXIN1 (β = 0.18) and Flt3L (β = -0.06), respectively. Although AXIN1 and Flt3L appear to exert similar effects on CRC, their opposing β values for this serum metabolite suggest that our MR analysis could not definitively ascertain the impact of this specific serum metabolite on CRC. This outcome may arise from potential associations with other diseases or limitations in our statistical approach, both of which necessitate further investigation.
Our study possesses several notable strengths. First, this study represents the largest MR analysis to evaluate the causal relationships between inflammatory cytokines, serum metabolites, and CRC risk. Thanks to its prospective design, MR is exceptionally effective at controlling for potential confounders and minimizing reverse causation. Additionally, employing MR mitigates biases typically encountered in biomarker measurement studies, thereby facilitating a more robust exploration of the relationships among serum metabolites, inflammatory cytokines, and CRC.
However, several limitations of the study should be acknowledged. First, our MR analysis concentrated on 1 400 serum metabolites and 91 inflammatory cytokines, which may restrict the generalizability of our findings. Second, the participants were predominantly from developed European populations in the UKB and FinnGen databases, which may limit the generalizability of our findings to less developed regions and other ethnic groups. Future validation in diverse ancestry biobanks (e.g., All of Us Research Program, China Kadoorie Biobank) is essential as these resources accumulate sufficient case numbers. Moreover, the absence of gender-disaggregated analysis raises the possibility that gender could have influenced the outcomes. While MR studies are valuable for elucidating causation, they may not completely capture the intricacies of the relationship between exposure and outcome. Finally, although our MR analysis indicated significant associations, it is essential to further investigate these causal relationships through experimental and clinical research.
In conclusion, our findings from these extensive MR analyses offer a more thorough understanding of the causal relationships of inflammatory cytokines and serum metabolites with CRC risk. These findings can serve as a foundation for further exploration of the underlying mechanisms and potential clinical applications in the prevention and treatment of CRC. Future studies could investigate the inflammatory cytokines and serum metabolites closely associated with CRC in this study, either individually or in combination.
the Natural Science Foundation of Hunan Province(2022JJ30987)
the Key Research and Development Project of Hunan Province(2024JK2107)
,
京公网安备11010202008535号
PDF
(1365KB)
