Introduction

Chronic kidney disease (CKD) represents a significant global health challenge and is a primary contributor to the overall renal disease burden [1]. The World Health Organization (WHO) categorizes CKD as a major source of worldwide disability, attributing this to its widespread occurrence, enduring nature, and associated comorbidities [2]. Addressing CKD necessitates effective prevention and therapeutic strategies, underpinned by a profound understanding of its biological mechanisms [3]. Although genetic determinants are recognized contributors to CKD’s pathogenesis [4], the disease’s multifaceted nature means many of its biological intricacies remain elusive. Despite advancements in genetics, especially through genome-wide association studies (GWAS), translating genetic insights into tangible biological mechanisms presents a substantial hurdle [5].

Recent strides in omics technologies, notably metabolomics, have enriched our understanding of disease etiologies. Metabolomics, by elucidating altered metabolic pathways and intermediate metabolites, offers unprecedented insights into disease mechanisms [6]. Groundbreaking GWAS studies focused on metabolites have pinpointed disease-relevant loci, proposing that metabolites, as functional intermediaries, elucidate genetic underpinnings of renal disorders [7].

Metabolites, either as end products or intermediates of metabolic processes, hold pivotal roles in human physiology. The recent establishment of the genotype-dependent metabolic phenotypes database, which catalogues genetically determined metabolites (GDMs), through a comprehensive GWAS employing non-targeted metabolomics, sheds light on the intricate interplay between human serum metabolites and genetic variants in renal disease pathogenesis. Yet, a comprehensive understanding of GDMs, particularly in the context of CKD, remains incomplete, highlighting the need for in-depth analyses elucidating the interplay between genetic variations, metabolites, and CKD mechanisms.

Mendelian randomization (MR) serves as a robust epidemiological methodology, utilizing genetic variants as instrumental variables (IV) to establish definitive putatively causal links between exposures and outcomes. Owing to the inherent stability of genotypes, established at conception, MR provides robust estimates, largely impervious to confounding variables and reverse causation [8]. Capitalizing on this strength, MR, over the past decade, has been instrumental in delineating putatively causal relationships between risk exposures and diseases using GWAS summary statistics. The recent expansion of the metabolic spectrum through GWAS has culminated in a comprehensive atlas of GDMs.

In this study, we postulate the potential of this GDM atlas in elucidating the putatively causal relationship between GDMs and CKD. We employ a bidirectinal MR methodology to: (i) ascertain the putatively causal impact of human serum metabolites on CKD, and (ii) discern metabolic pathways that could illuminate CKD’s mechanistic underpinnings.

Materials and methods

Results

Instrumental variables (IVs) selection

For the set of 486 metabolites, the number of selected IVs varied, ranging from 3 to 479, with a median count of 15. Notably, the F statistics, which test the validity of IVs, all surpassed the threshold of 10. The lowest observed F statistic value was 17.411, signifying that the potential for weak instrumental bias is minimal. The full results of IVs selection were given in Additional file. The overarching design of this MR study is depicted in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The overview of the research workflow.

CKD: chronic kidney disease; SNP: single nucleotide polymorphism; MR: mendelian randomization; IVW: inverse variance weighted.

https://doi.org/10.1371/journal.pone.0298729.g001

Sensitivity analysis

To mitigate the potential influence of horizontal pleiotropy on MR estimates, sensitivity analyses were conducted. Fig 2 encapsulates the sensitivity results for the seven metabolite-CKD pairs with significant putatively causal associations. Generally, the putatively causal associations were deemed robust if they achieved statistical significance (p<0.05) in at least two additional MR tests, typically the Weighted Median and the Weighted.Mode tests. The MR-Egger’s intercept term was utilized to identify potential horizontal pleiotropy across all associations. Notably, the associations of proline and X-12510 with CKD were significant in the Egger Intercept test, registering p-values of 4.572E-02 and 4.936E-02 respectively. Scatter (S3 Fig) and leave-one-out plots (S4 Fig) further confirmed the absence of potential outliers for all identified metabolites. We excluded the association between metabolite X-12798 and CKD due to the indication of heterogeneity via Cochran’s Q test (Q_p_val = 1.079E-02). Consequently, four out of the seven pairs—mannose, N-acetylornithine, glycine, and succinylcarnitine with CKD—were identified as robust. A comprehensive report of the sensitivity and pleiotropy analysis is detailed in the supplementary files.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Sensitivity analysis for significant metabolites on CKD.

OR: odds ratio; CKD: chronic kidney disease.

https://doi.org/10.1371/journal.pone.0298729.g002

Causal effects of CKD on metabolites

For CKD, 12 SNPs were chosen as instrumental variables (IVs). Upon application of the Cochran’s Q test, ensuring the absence of heterogeneity (Q_p_val>0.05), the IVW-FE analysis revealed a non-significant putatively causal effect of CKD on the metabolite mannose, with an OR of 1.00 (95% CI: 0.98–1.02) and p-value of 8.878E-01. Comparable non-significant results were obtained for CKD’s effect on glycine, with an OR of 0.99 (95% CI: 0.97–1.01) and p-value of 4.801E-01, even when using MR mixture models. In contrast, significant putatively causal effects were observed for CKD on N-acetylornithine (OR = 1.05, 95% CI: 1.01–1.09, p = 8.724E-03) and succinylcarnitine (OR = 1.02, 95% CI: 1.00–1.04, p = 3.643E-02) using IVW-FE, although these results were not consistently mirrored in other MR models(Fig 3). S5 and S6 Figs present the scatter plots with the regression line and funnel plots.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Sensitivity analysis for CKD on the four metabolites selected.

CKD: chronic kidney disease.

https://doi.org/10.1371/journal.pone.0298729.g003

Discussion

In this bidirectional Mendelian randomization study, we employed a rigorous, unbiased approach to investigate the putatively causal relationship between metabolites and CKD, leveraging the GWAS dataset from CKDGen. Out of 486 metabolites examined, seven were found to be relevant to CKD risk when probed using genetic variants. Notably, after a comprehensive sensitivity analysis, four of these metabolites emerged as particularly significant. Reverse Mendelian analysis further validated the robust putatively causal relationship of two metabolites, mannose and glycine, with CKD. Additionally, our pathway enrichment analysis highlighted five significant metabolic pathways: “Methionine Metabolism,” “Arginine and Proline Metabolism,” “Betaine Metabolism,” “Fatty Acid Biosynthesis,” and “Glycine and Serine Metabolism”—indicating their primary involvement in CKD’s pathogenesis.

To the best of our knowledge, this is the inaugural bidirectional MR study that synergizes genomics with metabolomics to discern causality in CKD. Our findings spotlight a suite of serum metabolites that exhibit a relationship with CKD, with mannose and glycine standing out due to their pronounced impact on the disease.

D-Mannose, an isomer of glucose, is not recognized as an essential nutrient. Previous evidence from prospective cohorts indicates a significant correlation between elevated mannose levels and high risk of kidney failure requiring replacement therapy [19]. Concurrently, mannose has been employed as a biomarker to elucidate the observed association between the consumption of ultra-processed food and chronic kidney disease (CKD) [20]. These findings align with our own. However, the putatively causal relationship between mannose and CKD could be multifaceted. Certain research implicates mannose as a key factor driving γδ T cell infiltration in the kidney [21]. Kidney surface-associated mannose structures have been implicated in triggering γδ T cell recognition, instigating a pathogenic interleukin-17a (IL-17a)-mediated autoimmune response [22]. IL-17A, for instance, is implicated in the immunopathogenesis of crescentic glomerulonephritis (GN) by promoting neutrophil recruitment [23], and it plays a pivotal role in MPO-ANCA GN pathogenesis by fostering the development of MPO-specific αβ T cells [24]. Moreover, reports suggest that mannose, through induction of energy expenditure via cellular ATP depletion, selectively modifies the synthesis of rapidly turning over glycosylated proteins such as PGs. This effect results in kidney size diminution, ureteric bud branching disorder, and subsequent kidney malformation [25]. Mannose is the primary monosaccharide involved in N-linked glycosylation, a form of post-translational modification. Exposure of mannose-glycosylated polysaccharides in lupus nephritis (LN) takes place on the surface of renal cells, thus enhancing the recognition by polysaccharide recognition receptors expressed on immune cells. Consequently, mannose glycosylation levels detected in renal biopsy specimens from LN patients have been validated as predictive markers for CKD progression [26]. Despite mannose’s potential as a promising biomarker for CKD, further investigations are warranted to elucidate the underlying mechanisms.

Glycine, ubiquitous in all living organisms, can be synthesized in humans from threonine, choline, or hydroxyproline via organ interconversion metabolism primarily in the liver and kidneys. Despite its simple structure, glycine plays intricate roles within the human body. Interestingly, administration of glycine to rats demonstrated no observable effect on either the rate of glomerular filtration or renal blood flow [27]. Conversely, subsequent investigations suggested that glycine significantly augments renal hemodynamics through as-yet-undefined mechanisms, with secondary increases in excretion [28]. While there is evidence suggesting altered glycine metabolism in CKD disease models [29], the positive or negative impact of glycine on the kidney remains under debate. Historical proposals suggested a potential nephrotoxic effect of glycine [30], and recent findings indicate that glycine exacerbates ischemia-reperfusion-induced kidney injury via NMDA receptor agonism [31]. Notably, one of the facets of ischemia-reperfusion injury is the onset of oxidative stress. Yet, certain studies suggest glycine might inhibit oxidative stress onset by promoting the nuclear translocation of Nrf1, thus enhancing Glo2 function [32], which could mitigate kidney injury in models induced by lead [33] or cobalt chloride [34]. Furthermore, CKD development has been linked to ferroptosis [35], a process that glycine can potentiate via SAM-mediated GPX4 promoter methylation [36]. Intriguingly, glycine also contributes to the synthesis of glutathione (GSH), an antioxidant implicated in ferroptosis inhibition [37]. Moreover, glycine demonstrates a robust association with CKD, presenting potential predictive value for significant health outcomes that may impact CKD.

It’s worth noting that the kidneys contribute to glucose synthesis from amino acids via gluconeogenesis, regulate acid-base balance through inter-organ metabolism of glutamine, and synthesize amino acids like arginine, tyrosine, and glycine. Renal dysfunction can lead to an imbalance in these metabolic processes, promoting muscle protein breakdown, inflammation, mitochondrial abnormalities, immune response defects, and cardiovascular diseases [38]. Another study elaborates on how the progressive loss of kidney function in CKD patients is associated with several complications, including metabolic acidosis, malnutrition, and protein-energy wasting. Altered kidney amino acid and protein metabolism could play a key role in the homeostasis of various vasoactive compounds and hormones in patients with advanced disease [39]. Based on these insights, it’s evident that the findings of our study, which identified a strong association between certain amino acids (like proline and glycine) and CKD, might be reflective of the kidney’s diminished capacity to metabolize and excrete these amino acids. This metabolic impairment, rather than the amino acids themselves, could contribute to their increased serum levels in CKD.

In this study, the metabolic pathway analysis showed that “Methionine Metabolism”, “Arginine and Proline Metabolism”, “Betaine Metabolism”, “Fatty Acid Biosynthesis” and “Glycine and Serine Metabolism” pathways are mainly associated with CKD.

Notably, Methionine, an aliphatic and sulfur-containing essential amino acid, has been implicated in lipid metabolism, the activation of endogenous antioxidant enzymes such as methionine sulfoxide reductase A, and glutathione biosynthesis [40]. An escalation in ROS levels in the kidneys was noted in the MetO group and the Met+MetO group, denoting heightened oxidative stress levels [41]. Other research has noted an increase in tHcy and a concomitant decrease in Met levels in the blood as CKD progresses. In particular, a decline in GFR, especially in the presence of proteinuria, is associated with reduced Met levels [42]. The upregulation of NNMT expression in the kidneys has been linked to disturbances in methionine metabolism and consequent renal fibrosis [43]. Moreover, methionine serves as a precursor of homocysteine and cystathionine. Hyperhomocysteinemia (HHcy) and raised homocystinuria levels are closely associated with chronic renal failure [44]. Taken together, these findings strongly suggest a critical role for methionine metabolism in the biological mechanisms underlying CKD.

Evidence has underscored the role of proximal convoluted tubules as the primary site for endogenous arginine biosynthesis [45]. Studies have highlighted variations in arginine and proline metabolism across a range of contexts, such as a Garidi-5-induced rat model of subacute hepato-renal toxicity [46], a chromium exposure-induced injury model [47], an adenine-induced chronic kidney disease model [48], and in instances of diabetic nephropathy [49].

Betaine, a critical osmolyte, accumulates in the basolateral plasma membrane of medullary epithelial cells via the betaine/gamma-aminobutyric acid transporter (BGT1) [50]. Its synthesis is augmented under hyperosmotic conditions [51]. Notably, serum osmolality has emerged as an independent risk factor for CKD onset [52]. In a retrospective longitudinal investigation, a higher baseline urinary osmolality was found to be closely linked to an elevated risk of end-stage kidney disease [53].

Ectopic lipid accumulation within the kidneys has emerged as a characteristic feature of metabolic conditions precipitating chronic kidney disease (CKD) [54]. Persistent disturbances in fatty acid metabolism have been observed in murine CKD models. A heightened fatty acid biosynthesis, coupled with downregulated triglyceride and diglyceride catabolic pathways, correlates with a notable increase in serum lipid levels [55]. Accumulation of fatty acids in the kidneys is purportedly deleterious to podocytes, precipitating cell apoptosis and subsequent glomerulosclerosis [56]. Lipidomic investigations have illuminated the pivotal role of the structural allocation of free fatty acids between fatty acid oxidation and complex lipid fatty acid composition in driving CKD progression [57].

In response to reduced urinary excretion due to diminished GFR, plasma levels of D-serine increase. Consequently, plasma D-serine levels are integral to kidney disease detection, with its concentration potentially escalating alongside CKD progression [58]. Notably, a combination of plasma and urinary D-serine levels serves to differentiate CKD origins, particularly pertinent in the context of lupus nephritis [59]. Collectively, these findings suggest a potential relevance of these metabolites’ biosynthesis and metabolism to the biological mechanisms underpinning CKD.

However, this study faces several limitations: First, the observed associations in our study may not necessarily indicate a direct causative relationship between these amino acids and CKD. Instead, they could represent a secondary effect of CKD-related metabolic dysregulation. This perspective is crucial for understanding the complex interplay between kidney function, amino acid metabolism, and CKD progression. Future research should aim to unravel these intricate relationships and explore potential therapeutic interventions that target metabolic disturbances in CKD. Second, while our analyses revealed potential causal relationships between certain metabolites and CKD, future studies should include more stringent statistical correction methods, such as the False Discovery Rate (FDR) adjustment, to properly address the issue of multiple testing. Such correction is not only crucial for reducing false positives but also enhances our understanding of which findings are truly biologically significant. Third, The efficacy of the instrumental variables (IVs) is closely tied to the GWASs’ sample size. A larger dataset would bolster accuracy. While Mendelian randomization is a potent tool for discerning causality between human blood metabolites and CKD, experimental data-based studies are essential for validation. The reliability of our MR analysis is largely contingent on the instrumental variables’ explanatory power on exposure, emphasizing the need for a more expansive sample size. Last, for pathway analysis, due to data constraints, we included metabolites with P values > 0.10. While we identified CKD-risk-contributing metabolites, their roles in CKD’s pathogenic mechanism warrant further exploration.

Conclusions

In our comprehensive bidirectional MR analysis, we identified mannose and glycine as metabolites with a strong putatively causal association with CKD. We further spotlighted five metabolic pathways that may underpin CKD’s etiology. These insights pave the way for harnessing select metabolites as diagnostic biomarkers and for devising targeted treatment strategies. Nevertheless, further research is paramount to corroborate our findings.

Supporting information

Acknowledgments

We thank all the consortium studies for making the summary association statistics data publicly available.We also thank Yi Yu for his help in coding analysis.