Serum metabolomics analysis reveals metabolite profile and key biomarkers of idiopathic membranous nephropathy

Patients

We recruited 25 patients diagnosed with idiopathic membranous nephropathy from the First Affiliated Hospital of Jinan University. The ages of all IMN patients ranged from 40 to 73 years. Likewise, our study also recruited 15 age-matched people for healthy controls. All IMN patients underwent renal biopsy to confirm the diagnosis. The inclusion and exclusion criteria of our study were shown in the Supplemental Materials. Our research was approved by the Helsinki Declaration of ethical principles for medical research about human subjects and the ethics committee of the First Affiliated Hospital of Jinan University (KY-2021-018). We collected the serum samples after obtaining the written informed consent of all participants.

Metabolomics analysis

Statistical analysis

Our study performed the statistical analysis by GraphPad Prism 9.0 software, using two-tailed Student’s t-tests for measurement data. A p-value less than 0.05 indicates a statistically significant difference. We applied Spearman correlation analysis to analyze correlations between DEMs and clinical parameters. The correlation analysis and correlation network were visualized with the OmicStudio tools.

Basic characteristics of participants

To investigate the effects of serum metabolites in IMN patients, our study performed metabolomic analysis for 40 serum samples based on LC-MS/MS system from the IMN group and HC group. There were no significant differences in the general information, such as age, sex, and BMI. We summarized specimen information in Table 1. Statistical analysis of clinical specimen information revealed that some clinical indicators were not statistically different, such as alanine transaminase (ALT), aspartate transaminase (AST), creatinine, high-density lipoprotein (HDL), and triglyceride (TG) (p > 0.05). There were significant differences in serum albumin, cholesterol, low-density, lipoprotein (LDL), 24-h proteinuria, uric acid (UA), blood urea nitrogen (BUN), IgG, and Cys-C between the two groups (p < 0.05). For IMN patients, levels of serum albumin and lgG were significantly decreased; however, levels of cholesterol, LDL, BUN, 24-h proteinuria, and UA were significantly increased. This was consistent with the clinical presentation of membranous nephropathy, and IMN patients often present with profuse proteinuria, hypoproteinemia, oedema, and hyperlipidemia.

Community metabolomics profile

We recruited 25 IMN patients and 15 healthy controls to identify serum metabolites based on the UHPLC-QE-MS technique. Firstly, a series of data management was processed, such as filtering individual peaks to remove noise, filtering on individual peaks, missing value recording, and normalization. The results of the PCA score plot illustrated that the samples were largely within the 95% confidence interval (Fig. 1C). We also analyzed the results using the statistical method of OPLS-DA and permutation test of the OPLS-DA model, which can demonstrate the separation feature between the IMN group and HC group relatively well, with good repeatability and stability and no overfitting (Figs. 1A and 1D). We visualized the screening results for DEMs in a volcano plot (Fig. 1B). As a result, 215 metabolites were defined as DEMs. Moreover, 104 metabolites were identified as up-regulated types (log2(FC) > 0, p < 0.05), and 111 DEMs were down-regulated (log2(FC) < 0, p < 0.05).

Identification of metabolites in serum samples

In our study, we performed the metabolomic analysis in serum metabolites from IMN patients and healthy controls based on the UHPLC-QE-MS technique. As a result, 652 metabolites were successfully identified, and 215 metabolites were significantly altered. The identified metabolites can be categorized as follows: lipids and lipid-like molecules, organoheterocyclic compounds, organic acids and derivatives, benzenoids, organic oxygen compounds, phenylpropanoids and polyketides, and nucleosides nucleotides and analogues. The result showed that lipids and lipid-like molecules accounted for 31.44% of identified metabolites and 37.67% of differentially expressed metabolites. Likewise, organic acids and derivatives for 22.24% of all identified metabolites and 21.40% in DEMs (Figs. 2B and 2C). To some extent, it indicated the effects of lipid metabolic disorders in patients with idiopathic membranous nephropathy.

Metabolic KEGG pathway enrichment analysis

In addition, we performed statistical analysis for the identified metabolites using the online graphing tool MetaboAnalyst 5.0, and the data were normalized by sum and set auto-scaling. Unsupervised clustering analysis of the membranous nephropathy serum metabolites shows that ethylbenzene, 1-(1-Pyrrolidinyl)-2-butanone, LysoPE (16:0/0:0), dimethylethanolamine, alpha-methylstyrene, deoxyribose 5-phosphate cysteine-S-sulfate, N-ethylglycine, 2-methyl-3-oxopropanoic acid D-2,3-dihydroxypropanoic acid hypoxanthine, and 2-methylglutaric acid were significantly expressed in disease group (Fig. 2A). These metabolites were significantly up-regulated in IMN group. However, dehydroepiandrosterone sulfate, androsterone sulfate, norambreinolide, 3,4-dimethyl-5-pentyl-2-furanheptanoic acid, 13-heptadecyn-1-ol, and 6-deoxodolichosterone were observably clustered in the control group.

To further explore the role of the DEMs in membranous nephropathy, we conducted KEGG pathway enrichment analysis for the 215 significantly expressed metabolites. We respectively showed the top 25 KEGG pathways, such as histidine metabolism, purine metabolism, arginine and proline metabolism, and aminoacyl-tRNA biosynthesis (Fig. 3A). Compared with the other pathways, the histidine metabolism pathway was listed as the most significant pathway. L-glutamic acid, urocanic acid, L-histidine, imidazole-4-acetaldehyde, 3-methylhistidine, and formiminoglutamic acid were identified in the histidine metabolism pathway. Similarly, the purine metabolism pathway also drew our attention. Hypoxanthine, inosine, guanine, guanosine, and uric acid were enriched in the purine metabolism pathway. The results showed that idiopathic membranous nephropathy patients presented lower levels of imidazole-4-acetaldehyde, guanine, and guanosine compared to the control group. L-glutamic acid, urocanic acid, L-histidine, 3-methylhistidine, formiminoglutamic acid, hypoxanthine, and inosine were identified as up-regulated types in the IMN group. And the corresponding content trend changes of DEMs were displayed in the radar chart. The corresponding metabolites were shown in Fig. 3B, including oleic acid (log2(FC)= −0.66), palmitoleic acid (log2(FC) = −1.38), uracil (log2(FC) = −1.14), hypoxanthine (log2(FC) = 0.73), thiiraneacetonitrile (log2(FC)= − 1.21), 1-methylnicotinamide (log2(FC)= − 1.38), and N6-methyladenosine (log2(FC) = 0.26). The up-regulated DEMs were remarkably enriched in histidine metabolism, arginine biosynthesis, arginine and proline metabolism, D-glutamine and D-glutamate metabolism KEGG pathways in the IMN group (Fig. 3C). Moreover, the down-regulated DEMs were significantly enriched in nicotinate and nicotinamide metabolism, linoleic acid metabolism, pantothenate and CoA biosynthesis, citrate cycle (TCA cycle), lipoic acid metabolism, and biosynthesis of unsaturated fatty acids (Fig. 3D). The results indicated that a series of changes for serum metabolites and metabolic recoding occurred in the blood circulation of IMN patients. Some studies suggested that metabolic changes altered the epigenetic landscape in the disease processes and occurred metabolic reprogramming at multiple levels (Wang & Lei, 2018).

Diagnostic performance of metabolites

In addition, we performed univariate ROC curve analysis for 215 differentially expressed metabolites using the biomarker analysis module of MetaboAnalyst 5.0. The area under the ROC curve (AUC) was used to identify the sensitivity and specificity of biomarkers, with AUC values greater than 0.7 indicating a good predictive model (Mandrekar, 2010; Răchieriu et al., 2021). Based on ROC curve analysis for individual biomarkers, the following five metabolites were brought to our attention with potential diagnostic significance, including dehydroepiandrosterone sulfate (AUC = 0.995, log2(FC) = −3.934), ethylbenzene (AUC = 0.949, log2(FC) = 0.551), hypoxanthine (AUC = 0.936, log2(FC) = 0.730), urocanic acid (AUC = 0.917, log2(FC) = 1.423), and L-glutamic acid (AUC = 0.877, log2(FC) = 0.616) (Figs. 4A–4E). Violin plots illustrated a statistically significant difference in these metabolites’ expression between in IMN group and the HC group (Figs. 4H–4L). In particular, DHEAS was significantly downregulated in IMN, since the studies showed that dehydroepiandrosterone sulfate was higher in males than in females (Rainey et al., 2002), we compared DHEAS for different genders separately between the IMN group and HC group. DHEAS was significantly down-regulated in membranous nephropathy in either men or women (p < 0.001) (Figs. 4F and 4G). In our study, we focused on histidine metabolism, steroid hormone biosynthesis, and purine metabolism pathways. L-glutamic acid, urocanic acid, imidazole-4-acetaldehyde, imidazoleacetic acid, and formiminoglutamic acid were enriched in the histidine metabolism pathway. Hypoxanthine, inosine, guanine, and guanosine were identified in the purine metabolism pathway (Table 2). It showed that the metabolites have a promising future as diagnostic markers for predicting diseases and will need more studies to confirm it (Bajaj et al., 2021; van der Velpen et al., 2019). Therefore, we inferred that serum metabolites were significantly altered in IMN patients, thus indicating alterations in metabolic pathways leading to some phenotypic changes, which have positive clinical implementation for exploring the pathogenic mechanisms and characteristics of idiopathic membranous nephropathy.

Metabolite origin analysis

We performed host-microbiome origin analysis of differentially expressed metabolites using MetOrigin, aiming to effectively explore the interaction between gut microbes and serum metabolites. The sources of DEMs in the membrane nephropathy group and the control group could be classified as host, microbiota, co-metabolism, and other sources (food, drugs, environment, etc.). A total of seven DEMs were from the host only, 12 metabolites were from microbiota, 67 metabolites were from co-metabolism, and 129 metabolites were from others (Figs. 5A and 5C). Our study also performed metabolite pathway enrichment analysis (MPEA) of metabolic pathways for DEMs, which showed eight, seven and 47 metabolic pathways associated with the host, microbiota, and co-metabolism, respectively (Fig. 5D). Among the co-metabolized metabolic pathways, the DEMs were mainly involved in the histidine metabolism. The metabolites were mainly involved in the ethylbenzene degradation pathway for the microbiota (Fig. 5B). Spearman’s correlation analysis was used in metabolites and traceable gut microbes and presented in the Biosankey network. Therefore, we focused on histidine metabolism and ethylbenzene degradation pathways. For DEMs of co-metabolite, the 11 metabolic reactions were mainly contained in the histidine metabolism pathway (R00525, R01168, R02150, R02285, R02286, R02287, R02288, R02914, R03286, R04065, R10330). For example, in histidine metabolism (R01168), some DEMs were identified in this metabolic reaction, including histidine ammonia-lyase (4.3.1.3), L-histidine (C00135), and urocanic acid (C00785). The biosankey network revealed that the phylum proteobacteria, phylum actinobacteria, phylum bacteroidetes, phylum firmicutes, class gammaproteobacteria, class alphaproteobacteria, class actinomycetia, and class bacilli were closely related to these DEMs implicated in the histidine metabolism (R01168) (Fig. S1). Almost all of these metabolic reactions showed the same trend. In addition, in the microbiome, metabolites were mainly enriched in the BIO-ko00642 ethylbenzene degradation pathway, involving four metabolic reactions (R05424, R05425, R05440, R05745), with the main relevant metabolites being ethylbenzene (C07111), styrene (C07083), ferredoxin-NAD(P)+ reductase (naphthalene dioxygenase ferredoxin-specific) (1.18.1.7), and naphthalene 1,2-dioxygenase ferredoxin component. The metabolic reaction R05424 was shown in Fig. 5E. The results indicated that these differentially expressed metabolites were mainly from phylum proteobacteria, phylum actinobacteria, class beta proteobacteria, class alphaproteobacteria, class gammaproteobacteria, family burkholderiaceae, family alcaligenaceae, family comamonadaceae, and family pseudomonadaceae, with the same trend in the other three metabolic reactions (Fig. 5E). MetOrigin performed a functional analysis based on metabolites traceability, integrated correlation results of statistical and biological significance, and finally obtained biomarkers with high accuracy.

Correlation network analysis

To better elucidate the interactions of differently expressed metabolites, we screened 12 significantly expressed metabolites under the criteria of VIP > 1.5 and |log2FC| > 3. The DEMs were subjected to Spearman correlation analysis with clinical indicators. The results showed that dehydroepiandrosterone sulfate (r = 0.63, p < 0.0001), 24-epibrassinolide (r = 0.64, p < 0.0001), gamma-tocopheryl quinone (r = 0.69, p < 0.0001), and androsterone sulfate (r = 0.53, p < 0.001) were positively correlated with the level of serum albumin. This positive correlation also existed in these metabolites with IgG. Among them, dehydroepiandrosterone sulfate was negatively correlated with LDL (r = −0.42, p < 0.01), UTP (r = −0.68, p < 0.0001), i.e., 24 h proteinuria, and cholesterol (r = −0.55, p < 0.001) (Fig. 6A). In addition, for KEGG pathway analysis, we selected six pathways with significant enrichment of DEMs and performed spearman correlation analysis in the metabolites identified on these pathways. We found that several key pathways were closely linked, including histidine metabolism, purine metabolism, arginine and proline metabolism, pantothenate and CoA biosynthesis, pyrimidine metabolism, and steroid hormone biosynthesis KEGG pathways (Fig. 6B). Therefore, there were metabolic disorders of amino acid, nucleotide, and steroid hormones in IMN patients.