Genetically predicted metabolites mediate the association between immune cells and metabolic dysfunction-associated steatotic liver disease: a mendelian randomization study

Abstract

Background

Existing studies have presented limited and disparate findings on the nexus between immune cells, plasma metabolites, and metabolic dysfunction-associated steatotic liver disease (MASLD). The aim of this study was to investigate the causal relationship between immune cells and MASLD. Additionally, we aimed to identify and quantify the potential mediating role of metabolites.

Methods

A Mendelian randomization (MR) analysis was conducted using two samples of pooled data from genome-wide association studies on MASLD that included 2568 patients and 409,613 control individuals. Additionally, a mediated MR study was employed to quantify the metabolite-mediated immune cell effects on MASLD.

Results

In this study, eight immunophenotypes were linked to the risk of MASLD, and thirty-five metabolites/metabolite ratios were linked to the occurrence of MASLD. Furthermore, a total of six combinations of immunophenotypic and metabolic factors demonstrated effects on the occurrence of MASLD, although the mediating effects of metabolites were not significant.

Conclusion

Our study demonstrated that certain immunophenotypes and metabolite/metabolite ratios have independent causal relationships with MASLD. Furthermore, we identified specific metabolites/metabolite ratios that are associated with an increased risk of MASLD. However, their mediating role in the causal association between immunophenotypes and MASLD was not significant. It is important to consider immune and metabolic disorders among patients with MASLD in clinical practice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12944-024-02245-3.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD) is the most recent term used to describe fatty liver disease that is associated with metabolic syndrome [1]. It is considered the leading cause of chronic liver disease, and there is strong evidence for a genetic predisposition [1]. The pathological features of this condition primarily involve the excessive deposition of triglycerides in the liver. The severity can range from simple fatty liver to metabolic dysfunction-associated steatohepatitis (MASH) [2]. The prevalence of MASLD is rapidly increasing and currently stands at approximately 38% among adults globally [3]. Significantly, the pathophysiological impact of MASLD extends beyond hepatic manifestations, engendering a spectrum of systemic complications. It is intricately linked with a confluence of metabolic cardiovascular risk factors, including but not limited to obesity, insulin resistance, type 2 diabetes mellitus, metabolic syndrome, hypertension and dyslipidemia [4]. Therefore, it is crucial to elucidate its pathogenesis and identify potential targets for intervention.

Notably, immune cells significantly contribute to the pathogenesis of MASLD by promoting liver inflammation and fibrosis through proinflammatory and anti-inflammatory responses [5]. The involvement of diverse immune cells is increasingly recognized as instrumental in the pathogenesis of MASLD and is correlated with the exacerbation of liver conditions such as steatosis and fibrotic changes, thereby highlighting the potential immunological underpinnings of MASLD severity [6]. Previous research has indicated that activated immune cells can produce cytokines that increase oxidative stress levels, leading to chronic liver inflammation and fibrosis [7]. Hepatic inflammation in patients with MASLD is mainly mediated by immune cells that produce inflammatory mediators, such as tumour necrosis factor (TNF), which further induce lipid loading and sensitise these hepatocytes to cytokine-mediated cell death [6]. In addition to the pivotal role of immune cells, metabolic disorders such as insulin resistance, heightened inflammatory response, and increased oxidative stress levels are also characteristic features of MASLD [7]. This intricate interplay between immune cells and metabolic dysfunction underscores the potential of metabolic disorders to act as a mediator in the relationship between immune cells and MASLD [8]. Consequently, the interaction between immune cells and metabolic pathways represents a crucial axis in the pathogenesis of MASLD, and understanding the interplay between these two components is essential for a comprehensive comprehension of the disease mechanism.

To investigate the causal relationship between immune cells and MASLD and evaluate the potential mediating role of metabolites, this study employed a two-step Mendelian randomization (MR) approach. MR is a causal inference method that uses genetic variables as instrumental variables to estimate the impact of exposure factors on outcome variables from genome-wide association studies (GWAS) [9]. This research method not only effectively reduces the interference of non-measurement errors and confounding factors, but also solves the problem of reverse causality by utilising the fundamental principle of Mendelian inheritance law of heredity. This study employed a two-step MR approach to investigate the causal relationship between immune cells and MASLD and to evaluate the extent to which metabolites mediate this relationship.

Methods

Research design

In this study, a two-step MR analysis was conducted to explore the causal relationship between immune cells and MASLD and to investigate whether metabolites act as mediators of the influence of immune cells on MASLD (Fig. 1). The overall effect of immune cells on MASLD can be divided into direct and indirect effects. In phase 1, we performed two-sample MR using aggregate GWAS data to assess the total effect of 731 immune cell traits on MASLD while assessing reverse causality between MASLD and each immune cell trait to ensure that no bidirectional effect could compromise the validity of the mediation model (Fig. 1A). In phase 2, the mediating MR was applied to assess the mediating role of 1,400 plasma metabolites (1091 plasma metabolites and 309 plasma metabolite ratios) in the causal relationship between immune cells and MASLD and to quantify the metabolite-mediated mediating effect (Fig. 1B). Finally, after adjusting for potential metabolite effects, the total effect was subtracted from the mediated effect to determine the direct effect of immune cells on the MASLD. The MR study strictly followed the STROBE-MR guidelines and followed the basic assumptions of MR as follows: 1) instrumental variables (IVs) are closely related to exposure factors; 2) IVs are not associated with confounding factors in the relationship between exposure and outcome; and 3) IVs affect outcomes only through exposure, not through any other direct or indirect pathway [10].

Fig. 1.

Schematic representations of the associations investigated in the current study. A The total effect between immunophenotypes and MASLD. B The total effect was decomposed into indirect effect and direct effect using a two-step approach

GWAS summary data sources

The GWAS Catalog provides publicly available summary statistics for immune cells (accession numbers GCST0001391-GCST0002121). This study, which included a cohort of 3,757 Sardinians, identified approximately 22 million genetic variants that impact 731 immune cell traits. These traits included 118 absolute cells (ACs), 389 median fluorescence intensities (MFIs) representing surface antigen levels, 192 relative cell masses (RCs), and 32 structural features (MPs) [11]. Flow cytometry was utilized to determine the 731 immunophenotypes. Peripheral blood from donors was collected in heparin tubes, stained with antibodies, and processed using flow cytometry. The data were acquired using two standardized BD FACSCanto II flow cytometry machines and analysed with BD FACSDiva software (BD Biosciences).

The GWAS Catalog provides aggregate-level GWAS statistics for 1400 plasma metabolites obtained from the Canadian Longitudinal Study on Aging (CLSA) (access numbers GCST90199621-GCST90201020). Additionally, independent GWAS data for 1,091 blood metabolites and 309 plasma metabolites in 8192 individuals of European ancestry were established in the study [12]. This metabolomics study focused on 8,299 individuals of European ancestry in the CLSA. The circulating plasma metabolite data for these individuals were subjected to data processing and quality control for genome-wide genotyping. Identifying the genetic determinants that influence the ratio of substrate to product can offer valuable insights into biological processes that may otherwise be overlooked when focusing solely on individual metabolites. Therefore, this study investigated the ratio of metabolic substrates to metabolites. The fastGWA tool of GCTA 1.93.2 beta was used to perform GWASs for linear regression of metabolites and metabolite ratios. The analysis was adjusted for age, sex, number of hours since the last meal or drink, and genotyping batches.

The non-alcoholic fatty liver disease (NAFLD) database can be used under the MASLD concept, as suggested by experts [13]. Additionally, multiple studies have demonstrated that 98%-99% of individuals who meet the NAFLD criteria also meet the MASLD criteria [14, 15]. As a result, the aggregated GWAS statistics for MASLD in this study were derived from the FinnGen Biobank, which included 412,181 population samples of European ancestry (Case/Control: 2568/409613). The FinnGen Biobank is a public‒private partnership project dedicated to the collection and analysis of more than 500,000 individual biological samples in Finland (https://www.finngen.fi/en, visited on February 20, 2024) [16].

The GWAS data used in our study were from participants of European descent. This study utilized publicly available GWAS data for two-sample Mendelian randomization (MR) analysis; thus, no additional approval was necessary. The details of all GWAS datasets included in our study are presented in Table 1. The GWAS data on exposures and outcomes utilised in this paper originate from disparate geographical regions, thereby precluding any overlap in the underlying populations.

Table 1.

Information on the GWAS datasets used in the two-step MR study

Trait	Consortium or cohort study	Ethnicity	Sample size	Year of publication	PMID
Immunophenotypes	Sardinians cohort	European	3,757	2020	32,929,287
Plasma metabolites	CLSA cohort	European	8,299	2023	36,635,386
MASLD	FinnGen Biobank	European	412,181	2023	/

Abbreviation: GWAS Genome-wide association studies, MR Mendelian randomization, CLSA Canadian Longitudinal Study on Aging, MASLD Metabolic dysfunction-associated steatotic liver disease

Selection of IVs

To identify the most significant IVs for each immune cell and metabolite/metabolite ratio, we utilized the recently established significance level of 1 × 10^–5 [17]. To ensure data quality, we employed a clumping program from PLINK software (version 1.90) to eliminate chained unbalanced single nucleotide polymorphisms (SNPs) (r² < 0.001, distance range 10,000 kb). For candidate IVs in MASLD, only those with a significance level of 5 × 10^–8 were selected [18]. Due to the limited number of SNPs that met the aforementioned thresholds, these SNPs were classified based on linkage disequilibrium (r² < 0.01, distance range = 5000 kb). Besides, when MASLD was as exposure factor, we used ldlink (https://ldlink.nih.gov/?tab=ldtrait, accessed 26 June 2024) to minimize the impact of confounding factors associated with immune traits [19]. Linkage disequilibrium (LD) was confidently estimated based on European samples from the 1000 Genomes Project [20]. If a specific exposed SNP is absent in the resulting dataset, a proxy SNP is confidently used by the LD tag. Individuals with palindromes and ambiguous SNPs were confidently excluded from the MR analysis. The F statistic is defined as follows: F = [(N-K-1)/K]/[r²/(1-r²)], where K is the number of genetic variants, N is the sample size [21]. Weak IVs (F statistic < 10) were removed.

Statistical methods

Mediation MR analysis

During phase 1, we confidently estimated the total effect of each immune cell trait on MASLD using two-sample MR. We expressed each estimate as β₀ (a in Fig. 1A). Additionally, we thoroughly assessed the reverse causality (b in Fig. 1A) between the MASLD score and each immune cell trait to ensure that the validity of the mediation model was not affected by any bidirectional effect. In Phase 2, we utilized mediating MR to accurately quantify the mediating effects of 1400 plasma metabolites (1091 plasma metabolites and 309 metabolite ratios) on the causal relationship between immune cells and MASLD (Fig. 1B). First, we estimated the causal effect of each plasma metabolite on MASLD risk (β₂, d in Fig. 1B). Second, we determined the causal effect of immune traits on each plasma metabolite using two-sample MR (β₁, c in Fig. 1B). The 95% confidence intervals (CIs) were calculated using the propagation of error method [22]. To calculate the proportion of the direct effect of each immune trait on MASLD, we subtracted the intermediate effect (β₁ × β₂) from the total effect (β₀) [23].

Sensitivity analysis

In the present investigation, MR analyses were carried out to estimate causal relationships with the results articulated through odds ratios (ORs) and regression coefficients (β values). We used two-sample MR with variance weighting (IVW) of the random effects model as the main analysis method [24]. To corroborate the reliability of the IVW-derived estimates, we conducted an array of sensitivity analyses encompassing several methodologies: MR‒Egger, the weighted median, the simple mode, and the weighted mode [25]. The MR‒Egger method is ingeniously designed to yield a consistent estimation of causal influence, albeit under the premise of weak IV assumptions [26]. The weighted median estimator, in contrast, assures consistent estimates provided that a minimum of 50% of the information in the analysis originates from valid IVs [27]. We evaluated heterogeneity among SNPs by computing Q statistics and their associated P (P > 0.05 indicated that there was no heterogeneity in the included IVs) [21]. Additionally, horizontal pleiotropy, which occurs when genetic variants influence the outcome through pathways other than the exposure of interest, was scrutinized through the MR‒Egger intercept test and the MR-PRESSO method [21]. To further solidify our findings, we conducted a leave-one-out sensitivity analysis, which involves sequentially excluding each SNP to determine its influence on the aggregate causal estimate. Visual representations, such as funnel plots and scatter plots, were generated to intuitively display any indications of horizontal pleiotropy. R software (version 4.3.2, http://www.r-project.org) was used for all MR analyses in this study. The R packages "TwoSampleMR (version 0.5.10)", "MRPRESSO" and "Mendelian Randomization" were utilized for all MR analyses. Furthermore, in light of the potential for an increased overall false positive rate in multiple comparisons, we applied a false discovery rate (FDR) correction to the primary P_IVW using the Benjamini–Hochberg procedure (FDR_IVW < 0.1) [28]. The corrected P calculated by the remaining four MR methods were not employed as the primary results, in order to minimise the occurrence of false negatives. In the case of P that do not necessitate correction for multiple tests, a value of P < 0.05 represents the threshold of significance.

Results

Causal effects of immune cells on MASLD

Following the aforementioned steps, we obtained six SNPs in MASLD as IVs that were significant in terms of the genome-wide threshold. One SNP (rs2267595) was excluded from the study due to its association with monocyte count (Supplementary Table 1). As a result, five MASLD-associated SNPs were included in this analysis (Supplementary Table 2). To clarify the relationship between the 731 immunophenotypes and MASLD, we used two-sample Mendelian randomization, with IVW as the main analysis method. A total of eight immunophenotypes were significantly associated with MASLD (FDR_IVW < 0.1), as depicted in Fig. 2. Stratification across immunophenotypic clusters demonstrated that associations spanned multiple cellular subsets. Among the four types of immune cells analyzed for their protective effect on MASLD, CD19 on the sw me cell subset was found to have the most significant protective effect on MASLD (OR = 0.89, 95% CI = 0.82–0.96). Conversely, 4 immune cell phenotypes were identified as risk factors for MASLD, with CD64 on CD14 + CD16 + monocytes conferring the highest risk (OR = 1.23, 95% CI = 1.04–1.45). Further rigorousness of our findings is substantiated by sensitivity analyses employing 4 alternative MR methods which affirm the stability and robustness of the observed causal associations (Supplementary Table 3). Additionally, MR‒Egger intercept analysis mitigated the concern of horizontal pleiotropy (Supplementary Table 4). Complementary scatter and funnel plot visualizations further corroborated the consistency and reliability of the associations (Supplementary Figs. 1–2). Finally, exploratory analyses utilizing reverse MR demonstrated the absence of reverse causal influence with 7 of the identified immunophenotypes. However, it was determined that MASLD is associated with CD64 on CD14 + CD16 + monocyte (reverse FDR_IVW = 0.026), and thus it was excluded from subsequent analyses (Supplementary Table 5).

Fig. 2.

Forest plot showing the causal associations between immune cell traits and MASLD. Abbreviation: MASLD, metabolic dysfunction-associated steatotic liver disease; IVW, inverse variance weighting; OR, odds ratio; CI, confidence interval

Causal effects of plasma metabolites with MASLD

Among the 1400 plasma metabolites, there are thirty-five metabolite/metabolite ratios were found to be correlated with MASLD (FDR_IVW < 0.1), as shown in Fig. 3. Of these, seventeen displayed inverse correlations with MASLD, among which the compound denoted as X − 25,790 levels demonstrated the most pronounced protective effect (OR = 0.76, 95% CI = 0.61–0.93). Notably, the association between elevated cysteine-glutathione disulfide (CySSG) levels and reduced risk of MASLD was highly statistically significant (P_IVW = 0.004), with an OR of 0.81 (95%CI = 0.71, 0.92). Conversely, eighteen metabolites were positively correlated with MASLD risk, with taurochenodeoxycholate levels exhibiting the greatest increase in risk (OR = 1.31, 95% CI = 1.10–1.55). The metabolites were further investigated according to specific metabolic domains. Among the factors related to amino acid metabolism, the level of histidine emerged as a protective factor against MASLD (OR = 0.82, 95% CI = 0.71–0.95), and the histidine/asparagine ratio was identified as a risk factor (OR = 1.17, 95% CI = 1.06–1.30). Notably, for factors related to nucleotide metabolism, inosine levels were associated with an increased risk for MASLD (OR = 1.20, 95% CI = 1.05–1.36). For factors related to fatty acid metabolism, the concentration of eicosenedioate (C20:1-DC) was also identified as a risk factor for MASLD (OR = 1.18, 95% CI = 1.05–1.33). The robustness and reliability of these causal inferences are further supported by sensitivity analyses utilizing four alternate MR techniques (Supplementary Table 6). Moreover, intercept analysis from MR‒Egger regression indicated the absence of horizontal pleiotropy (Supplementary Table 7).

Fig. 3.

Forest plot showing the causal associations between blood metabolite/metabolite ratios and MASLD. Abbreviation: MASLD, metabolic dysfunction-associated steatotic liver disease; IVW, inverse variance weighting; OR, odds ratio; CI, confidence interval

Causal effects of immune cells with plasma metabolites

According to the MR analyses that investigated both immune cell subsets and plasma metabolite profiles, significant yet modest correlations were detected between six pairs of immunophenotypes and 11 corresponding plasma metabolite/metabolite ratios (FDR_IVW < 0.1, Fig. 4). Among these associations, HLA DR + NK %NK counts presented the most notable positive correlation with plasma levels of dopamine 4 − sulfate levels (OR = 1.05, 95% CI = 1.00–1.10–1.55). Conversely, CD14 + CD16 − monocyte AC exhibited the most substantial negative correlation with the levels of eicosenedioate (C20:1-DC) (OR = 0.96, 95% CI = 0.93–0.99). These findings are supported by sensitivity analyses that used four alternative MR approaches (Supplementary Table 8). Additionally, MR-Egger intercept analysis was performed to address concerns about horizontal pleiotropy (Supplementary Table 9).

Fig. 4.

Forest plot showing the causal associations between immune cell traits and blood metabolite/metabolite ratios. Abbreviation: IVW, inverse variance weighting; OR, odds ratio; CI, confidence interval

Associations between immune cells and MASLD mediated by plasma metabolites

To investigate the potential causal associations among immunophenotypes, metabolite/metabolite ratios and MASLD, we employed a two-step MR to discern the mediatory role of metabolite/metabolite ratios in the etiopathogenesis of MASLD influenced by immunophenotypic variations. During the intermediate MR analysis, we identified a correlation between three distinct immune cell types (CD14 + CD16- monocyte AC, CD3- lymphocyte AC, CD45RA- CD28- CD8br %CD8br, HLA DR + NK %NK, monocyte AC) and four metabolites (MTA, C20:1 − DC, histidine, histidine to asparagine ratio, succinimide, carnitine to ergothioneine ratio). These variables formed six paired immunophenotypic-metabolite/metabolite ratio combinations that were associated with the prevalence of MASLD (Table 2). Interestingly, our analysis revealed that these six metabolites presumably function as intermediaries in causal effects related to specific immune cell phenotypes, but these mediating effects were not statistically significant.

Table 2.

Associations between immune cell traits and MASLD mediated by blood metabolite/metabolite ratios 

Immune cell traits	Blood metabolites/ metabolite ratios	Outcome	Mediated effect	P
Monocyte AC	Eicosenedioate (C20:1-DC) levels	MASLD	-0.00627(-0.0258, 0.0133)	0.530
CD14 + CD16- monocyte AC	Histidine to asparagine ratio	MASLD	-0.00569(-0.0219, 0.0105)	0.491
CD14 + CD16- monocyte AC	Eicosenedioate (C20:1-DC) levels	MASLD	-0.00528(-0.0248, 0.0143)	0.597
CD45RA + CD8br %T cell	Succinimide levels	MASLD	-0.0026(-0.0109, 0.00567)	0.538
CD14 + CD16- monocyte AC	Carnitine to ergothioneine ratio	MASLD	-0.00728(-0.0327, 0.0181)	0.574
CD14 + CD16- monocyte AC	Histidine to asparagine ratio	MASLD	-0.00569(-0.0219, 0.0105)	0.491

Abbreviation: MASLD Metabolic dysfunction associated steatotic liver disease

Discussion

This MR study aimed to investigate the independent causal effects of 731 immunophenotypes on MASLD and to determine whether these causal effects were mediated by 1400 plasma metabolites/metabolite ratios. The rigorous analytical framework employed herein identified eight genetically inferred immunophenotypes as having a consequential causal relationship with the modified risk profile of MASLD. Concurrently, thirty-five genetically predicted plasma metabolites/ metabolite ratios demonstrated a similar causal association with the altered risk of MASLD. After exhaustive research, we identified six paired immunophenotypic-metabolite/metabolite ratio combinations associated with the risk of MASLD. However, the contribution of these plasma metabolites/metabolite ratios as a mediator in this biological cascade was not statistically significant. These findings provide support for the independent causal role of circulating immune cells and plasma metabolites in MASLD.

MASLD is the most common chronic liver disease globally, affecting more than 38% of the adult population [3]. Despite its metabolic foundations, MASLD exhibits pronounced immunoinflammatory effects. The pathogenesis of MASLD is characterised by a series of immune responses, including fibrosis and inflammation, which are triggered through crosstalk between hepatocytes and different immune cell subpopulations, such as T-cells, monocyte-macrophages, neutrophils and hepatic stellate cells [29]. Prolonged and excessive activation of the immune and inflammatory response weakens its protective effects, causing immune dysfunction and metabolic disorders and metabolic inflammation. Recent investigations have highlighted the role of immune mediators from both resident liver parenchymal cells and a cohort of resident and infiltrating immune cells in precipitating the activation of innate and adaptive immune responses, thereby promoting the onset and progression of MASLD [30]. Nevertheless, despite the diversity of known immune cell phenotypes, comprehensive studies examining specific immune cell subsets related to MASLD are lacking.

By utilizing genetic variants as IVs, our investigation provides novel insights demonstrating that eight genetically predicted immunophenotypes exhibit a substantial association with MASLD. We discerned that certain immune cell populations within peripheral blood, including monocytes, T/NK cells, B cells, as determined at the genetic level, have a positive causal association with MASLD. Previous studies have analysed single-cell RNA sequencing in the livers of NAFLD mice, demonstrating that with the development of liver fat deposition and inflammation, there is a relative increase in the number of infiltrating immune cells. These include monocyte-macrophages, dendritic cells, T cells, neutrophils, and B cells [31]. Our study demonstrated that three types of monocytes, distinguished by the CD64 marker on the surface of CD14 + CD16 + monocytes, CD14 + CD16- monocytes AC and Monocyte AC, exhibited a correlation with the risk factors for MASLD. It is notable that monocyte or monocyte-derived macrophage is scarce in healthy livers. Following dietary induction, there is an influx of monocyte or monocyte-derived macrophage into liver tissue, accumulating around injured hepatocytes.

Furthermore, our study identified four genetically determined T/NKT cells (CD45RA + CD8br %T cell, CD28- CD127- CD25 +  + CD8br %T cell, CD28- CD127- CD25 +  + CD8br %CD8br and HLA DR + NK %NK) and B cells (CD19 on sw mem) with MASLD. With the exception of the CD45RA + CD8br %T cell, which is a risk factor for MASLD, the remaining cells are protective factors for MASLD. However, there is a paucity of current research on the underlying mechanisms by which these particular T cells and B cell exert their influence on MASLD. A number of studies have demonstrated that NKT cells play a role in the development of diet-induced obesity and metabolic dysfunction [32, 33]. In the context of diet-induced NAFLD, the Tim-3/Gal-9 pathway is implicated in the depletion and subsequent proliferation of NKT cells [34]. Correspondingly, mice lacking NKT cells exhibited elevated rates of weight gain and hepatic steatosis [35].

The role of immunity and inflammation in MASLD patientsis complex and multifaceted. In general, inflammatory triggers originate from both hepatic sources, including lipid overload, lipotoxicity, and oxidative stress, and extrahepatic factors, such as the gut–liver axis, adipose tissue, and skeletal muscle, collectively contributing to distinct immune-mediated pathomechanisms in MASLD [36]. Hepatocyte steatosis generates lipid toxicity, which in turn causes oxidative stress and mitochondrial dysfunction. The resulting cellular distress leads to the release of proinflammatory mediators from hepatocytes. This induces robust immune cell activation and infiltration, thereby perpetuating hepatocyte injury and fostering insulin resistance [6, 37]. The association between specific immunophenotypes and an altered risk of MASLD revealed in our study underscore the relevance of immune regulatory mechanisms in the etiology of fatty liver disease. These findings potentially guiding targeted interventions to modulate immune cell function and ameliorate disease pathology.

Moreover, our analysis elucidated the independent effects of thirty-five distinct plasma metabolites/metabolite ratios, each of which exhibited a significant correlation with the MASLD. Among the aforementioned associations, there was a statistically significant correlation between elevated levels of genetically determined CySSG and a reduced risk of MASLD. This finding indicate that CySSG may be a key player in the development of MASLD. CySSG, a sulfhydryl-modified prodrug of glutathione, is found in mammalian plasma. It consists of two isoforms, L-CySSG and D-CySSG [38]. L-CySSG has been shown to be effective in reducing acetaminophen-induced liver injury in mice by releasing glutathione and the key amino acid for de novo glutathione biosynthesis (cysteine), through disulphide bond breakage [39–41]. In terms of therapeutic potential, L-CySSG is chemically more stable than glutathione and is expected to be used as a prophylactic agent to terminate hepatotoxicity by combining with potentially hepatotoxic drugs [42]. Alternatively, it may be used as a dietary supplement for patients with MASLD. In another recent case–control study, a significant protective effect of CySSG against NAFLD was also demonstrated [43–45].

Our study examined risk factors positively linked to MASLD development and revealed that elevated genetically determined plasma levels of taurochenodeoxycholate are positively associated with MASLD. This finding is notable, as such elevations have previously been noted predominantly in Eastern populations when contrasted with their Western counterparts [46]. Nonetheless, studies have suggested that taurochenodeoxycholate may attenuate the progression of MASLD by improving intestinal inflammation and barrier integrity, which adds complexity to our understanding of the role of bile acid metabolites in MASLD pathophysiology [47, 48]. Furthermore, despite inosine being traditionally perceived as hepatoprotective and implicated in human energy metabolism and protein synthesis, our investigation indicated that elevated plasma inosine levels are a risk factor for MASLD. This association may be attributable to the notion that excess inosine can promote the absorption of exogenous fatty acids and their subsequent esterification into triglycerides. This process is potentially mediated by increased phosphorylation of adenosine 5’-monophosphate-activated protein kinase within hepatic tissue, resulting in lipid accumulation in the liver [49]. Additional metabolites implicated as risk factors for MASLD of interest to us include phenylacetate and eicosenedioate (C20:1-DC). The literature on phenylacetate and eicosenedioate (C20:1-DC) is sparse; our study is the first to report that genetically determined plasma phenylacetate levels are indicative of an elevated risk for MASLD. Phenylacetate, recognized as a microbial metabolite, has been associated with lipid accumulation within the liver and can cause MASLD [50]. Notably, a significant proportion of the plasma metabolites identified are influenced by the gut microbiota, underscoring the intricate interaction between the intestinal microbiota and MASLD pathogenesis [43, 45, 51]. The current research on eicosenedioate (C20:1-DC) is focused on its potential to chemically modify human glucose-dependent insulinotropic polypeptide and insulin, thereby delaying the half-life of the drug and reducing the frequency of administration [52, 53]. Although the treatment of T2DM has the potential to improve fat deposition, it is important to consider the potential role of eicosenedioate (C20:1-DC) in liver fat deposition.

Finally, our study elucidated the interconnections between immunophenotypes and metabolite/metabolite ratios. Although we observed associations between certain immunophenotypes and metabolite/metabolite ratio combinations with respect to MASLD, our findings indicate that these metabolites do not substantially mediate the causal association between immune cells and MASLD. For instance, despite taurochenodeoxycholate levels being identified as a notable risk factor for MASLD, they did not affect the relationship between the immunophenotype and MASLD. This suggests that the impact of the immunophenotype on MASLD could be either direct or potentially mediated by other currently unknown metabolic pathways or molecular mechanisms [54]. Although these metabolites did not demonstrate significant mediating effects in our analysis, their presence may still be indicative of a contributory role in linking immune regulation with metabolic dysregulation. Such insights underscore the multifactorial nature of MASLD and the necessity for comprehensive investigations to reveal the multiple factors and interactions that govern its pathogenesis.

The use of the MR approach in this study leverages genetic variants as instrumental variables to mitigate confounding effects and the issue of reverse causation, thereby providing more robust causal inferences. It is important to acknowledge the limitations of this study. First, the associations between genetically predicted immunophenotypes and MASLD may be subject to gene‒environment interactions. Second, the study did not include all possible immune cell phenotypes and plasma metabolites, meaning that the role of other, unstudied immune cells and metabolites in MASLD cannot be discounted. Third, the sample population primarily consisted of individuals from specific ethnic groups, which may limit the generalizability of the findings. To enhance the applicability of the results, further validation through independent cohorts and multiethnic population samples is necessary.

Conclusion

In conclusion, the present study employed MR analysis to elucidate the independent causal associations between specific immunophenotypes and metabolite/metabolite ratios and the risk of MASLD. Although the mediating role of certain metabolites in the causal pathway linking the immunophenotype to MASLD was not substantiated, the relationships identified herein provide a foundation for subsequent research endeavours. Further research is required to elucidate the biological causation, which may involve an expanded array of immune cell subsets and metabolites. By collaboratively integrating genetic, metabolomic, and immune datasets in a collaborative manner, the paradigm of personalized medicine in the context of MAFLD could be advanced.

Abbreviations

MASLD

Metabolic dysfunction-associated steatotic liver disease

MASH

Metabolic dysfunction-associated steatohepatitis

NAFLD

Non-alcoholic fatty liver disease

MR

Mendelian randomization

GWAS

Genome Wide Association Studies

IVs

Instrumental variables

LD

Linkage disequilibrium

CI

Confidence Interval

IVW

Inverse Variance Weighting

OR

Odds Ratios

SNPs

Single Nucleotide Polymorphisms

CySSG

Cysteine-glutathione disulfide

MTA

5-Methylthioadenosine

Authors’ contributions

Dan Ye contributed to conceptualization,data curation, methodology, formal analysis, software, writing – original draft preparation and visualization. Jiaofeng Wang, Jiaheng Shi, Yiming Ma contributed to resources, formal analysis, software, writing – original draft preparation, data curation and visualization. Jie Chen, Xiaona Hu contributed to resources, writing – review & editing, funding acquisition, validation, supervision, and project administration. Zhijun Bao contributed to funding acquisition, writing – review, validation, and supervision.

Funding

This research was funded by the National Key R&D Program of China (2018YFC2002000, 2018YFC2002002), National Natural Science Foundation of China (82071581), Shanghai Municipal Health Commission (JKKPYL-2022–15), Key Discipline Projects of Huadong Hospital (LCZX220), Shanghai Outstanding Young Medical Personnel Training Program (Excellence Project of Shanghai Municipal Health Commission, 20224Z0009) and Key Specialized Diseases Construction of Huadong Hospital (ZDZB2225).

Availability of data and materials

The study analyzed publicly available datasets. Please refer to the GWAS Summary Data Sources above for the relevant data links.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study used only publicly available GWAS data. The ethics approval and consent to participate can be found in the original GWAS study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.