Ultra‐high performance liquid chromatography coupled to tandem mass spectrometry‐based metabolomics study of diabetic distal symmetric polyneuropathy

Abstract

Aims/Introduction

Distal symmetric polyneuropathy (DSPN) is a common complication of type 2 diabetes mellitus, but the underlining mechanisms have not yet been elucidated. The current study was designed to screen the feature metabolites classified as potential biomarkers, and to provide deeper insights into the underlying distinctive metabolic changes during disease progression.

Materials and Methods

Plasma metabolite profiles were obtained by the ultra‐high liquid chromatography coupled to tandem mass spectrometry method from healthy control participants, patients with type 2 diabetes mellitus and patients with DSPN. Potential biomarkers were selected through comprehensive analysis of statistically significant differences between groups.

Results

Overall, 938 metabolites were identified. Among them, 12 metabolites (dimethylarginine, N6‐acetyllysine, N‐acetylhistidine, N,N,N‐trimethyl‐alanylproline betaine, cysteine, 7‐methylguanine, N6‐carbamoylthreonyladenosine, pseudouridine, 5‐methylthioadenosine, N2,N2‐dimethylguanosine, aconitate and C‐glycosyl tryptophan) were identified as the specific biomarkers. The content of 12 metabolites were significantly higher in the DSPN group compared with the other two groups. Additionally, they showed good performance to discriminate the DSPN state. Correlation analyses showed that the levels of 12 metabolites might be more closely related to the glucose metabolic changes, followed by the levels of lipid metabolism.

Conclusions

The finding of the 12 signature metabolites might provide a novel perspective for the pathogenesis of DSPN. Future studies are required to test this observation further.

Short abstract

We found taht 12 signature metabolites might provide novel perspective for the pathogenesis of DSPN. Future studies are needed to test this observation further.

INTRODUCTION

Distal symmetric polyneuropathy (DSPN) is one of the most common complications of type 2 diabetes mellitus. Due to the high incidence and disability rate, it is imperative to investigate further the pathological changes and underlying mechanisms of DSPN ¹ . DSPN is neuroendocrine disorders related to long‐term hyperglycemia, and is essentially a metabolic disease ² . The symptoms of neurological dysfunction posed by disease, such as persistent foot pain, foot ulcerations and even lower‐limb amputations, not only seriously affect the patient's quality of life, but also place considerable pressure on public medical systems ³ , ⁴ . Consequently, a deeper metabolic changes investigation might give a clue to the induction mechanism for the DSPN. Although some pathological metabolites, such as advanced glycation end‐products, methylglyoxal and neurone‐specific enolase, as well as pathological metabolic pathways, such as the polyol pathway, the hexosamine pathway and activation of protein kinase C, have been identified ⁵ , ⁶ , ⁷ , ⁸ , targeted therapies based on existing mechanisms have limited effectiveness. Meanwhile, the pathogenesis of DSPN is complex, and there might be discrepancies between animal models studies and clinical studies ⁹ . Therefore, it might be crucial from a clinical perspective to explore novel pathophysiological pathways and molecules in DSPN patients.

Distal symmetric polyneuropathy is generally a chronic and progressive process. Most patients with type 2 diabetes mellitus usually develop DSPN after a certain course of disease (≥5 years) ¹⁰ , which might suggest the abnormal accumulation of metabolic alterations resulting in gradual nerve damage. Thus, further understanding the global changes in endogenous metabolites at different disease stages might be an important strategy for an accurate understanding of DSPN. Metabolomics is emerging as a powerful tool to facilitate characterizing various pathological disease. Similary to other omics technologies, metabolomics is focused on exploring the overall changes of the organism. However, it is worth noting that metabolomics could tell us what is happening in the moment, as well as completely reflect a range of complex biological events under the physiological or pathological conditions. As such, metabolomics is considered the closest omics to the disease phenotype ¹¹ , ¹² . It aims to provide mechanistic insights by comparing all the metabolite changes in different groups or different time points, and is particularly appropriate for studies of metabolic diseases ¹³ .

Based on the aforementioned premises, the present study was designed to include three groups: healthy control participants (Con group), patients with type 2 diabetes mellitus (type 2 diabetes mellitus group) and type 2 diabetes patients with DSPN (DSPN group). Through untargeted metabolomics measurement of plasma samples obtained from aforementioned populations, we sought to explore and identify significantly changed metabolites, and key pathological change, as well as potential biomarkers in DSPN.

MATERIALS AND METHODS

Study design and participants

The current study was a cross‐sectional study. Ethical approval was obtained from the ethics committee of Shuguang Hospital Affiliated to Shanghai University of Traditional Chinese Medicine, Shanghai, China (No. 2018‐599‐28‐01). The study procedures were in accordance with the Helsinki Declaration, and all the participants provided informed consent before inclusion. A total of 30 patients with type 2 diabetes mellitus and 60 patients with DSPN were enrolled, in addition, 30 healthy control participants who were matched for age and sex were also included. Clinical type 2 diabetes mellitus diagnosis was based on the World Health Organization criteria. DSPN was diagnosed according to guidelines for prevention and treatment of type 2 diabetes in China (2017): patients showing typical clinical symptoms of peripheral neuropathy followed the diagnosis of type 2 diabetes mellitus; patients underwent neurosensory examination (pinprick, vibration perception, touch pressure, temperature sensation and ankle reflexes) by a trained clinician, and at least two of the items were abnormal; that is, the Toronto Clinical Scoring System scores ≥6 ¹⁴ . The Toronto Clinical Scoring System score included three parts: neurological symptoms, neurological reflex and sensory function examination score. The neurological symptoms included numbness, pain, pins and needles sensation, weakness, unsteadiness of walking in the lower extremity, and similar symptoms in the upper extremity, with 0 points for normal and 1 point for the presence of corresponding symptoms, totaling 6 points; the neurological reflexes included ankle reflex and knee reflex, with 0 points for normal, 1 point for reduced and 2 points for disappeared, totaling 8 points; the sensory function examination included pain, temperature, pressure, vibration and position sensation of the right bunion, with 5 points for normal and 1 point for abnormal. The aforementioned scores totaled 19 points. Those with scores of 0–5 had no diabetic peripheral neuropathy (DPN), those with scores of 6–8 had mild DPN, those with scores of 9–11 had moderate DPN and those with scores of 12–19 had severe DPN. Exclusion criteria included other causes of peripheral neuropathy; recent acute complications of diabetes; severe liver or renal dysfunction; women who were pregnant or taking hormonal contraceptives; and use of chemotherapy drugs or psychopharmacologic drugs that would significantly affect the metabolic environment in vivo.

Sample preparation and quality control

Fasting plasma (ethylenediaminetetraacetic acid anticoagulation) of all the participants was collected by the same clinician at 8:00 a.m. and stored at −80°C. While processing the samples, proteins were precipitated with methanol under vigorous shaking for 2 min (Glen Mills GenoGrinder 2000; Clifton, NJ, USA) followed by centrifugation. The resulting extract was divided into five fractions: two for ultra‐high liquid chromatography coupled to tandem mass spectrometry (UPLC‐MS/MS; negative ion mode), two for positive ion mode, and one sample was reserved for backup. A rigorous quality control protocol was carried out. Overall process variability was determined by calculating the relative standard deviation for all endogenous metabolites (i.e., non‐instrument standards) present in 100% of the pooled matrix samples. Experimental samples were randomized across the platform run with quality control samples spaced evenly among the injections, as outlined in Figure S1.

Methodology

The parameters of the untargeted metabolomics method were described in our previous study ¹⁵ . All methods utilized a Waters ACQUITY UPLC and a Thermo Scientific Q‐Exactive high‐resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI‐II) source and Orbitrap mass analyzer. The sample extract was divided into five fractions. One aliquot was analyzed using acidic positive ion conditions, chromatographically optimized for more hydrophilic compounds. Another aliquot was also analyzed using acidic positive ion condition; however, it was chromatographically optimized for more hydrophobic compounds. Another aliquot was analyzed using basic negative ion optimized conditions using a separate dedicated C18 column. The fourth aliquot was analyzed by negative ionization afyer elution from a HILIC column (Waters UPLC BEH Amide 2.1 × 150 mm, 1.7 μm) using a gradient consisting of water and acetonitrile with 10 mmol/L ammonium formate, pH 10.8. Then, raw data were extracted, peak‐identified and quality control processed ¹⁶ . Furthermore, biochemical identifications are based on three criteria: (i) retention index within a narrow RI window of the proposed identification; (ii) accurate mass match to the library ±10 ppm; and (iii) the MS/MS forward and reverse scores between the experimental data and authentic standards. In addition, a variety of curation procedures were carried out to ensure that a high‐quality dataset was made available for statistical analysis and data interpretation. For studies spanning multiple days, a data normalization step was carried out to correct variation resulting from instrument interday tuning differences (Figure S2).

Statistical analysis

To obtain more accurate and robust results, an online analysis platform, “MetaboAnalyst 5.0” (https://www.metaboanalyst.ca) ¹⁷ , SPSS software (version 26; IBM Corp., Armonk, NY, USA), and R software (version 4.1.1; The R Foundation for Statistical Computing, Vienna, Austria) were used. Welch's two‐sample t‐test was used to test two independent populations. For statistical significance testing, P‐values are given, and P < 0.05 was considered statistically significant. However, for a large number of tests, we needed to account for false positives, so the false discovery rate was used to correct for multiple testing. The metabolites were considered to be significantly different if P < 0.05 and the false discovery rate (adjusted P‐value) was <0.05 ¹⁸ . The supervised multivariate method, orthogonal partial least squares discrimination analysis, the diagrams of validation models and the receiver operating characteristics (ROC) curve for each differential metabolite were completed by MetaboAnalyst 5.0. A heat map of all the selected differential metabolites and the box plots for changes of individual metabolites were drawn using R software. Spearman's correlation analyses between the clinical parameters (blood glucose, blood pressure, blood lipids and body mass index) and the screened differential metabolites were completed using SPSS 25.

RESULTS

Characteristics of participants

We included three diverse populations representing disease progression (30 healthy control participants, 30 type 2 diabetes mellitus patients and 60 DSPN patients). Participants in the Con and type 2 diabetes mellitus group did not have clinical evidence of typical neuropathy. The patients in the DSPN group all had a Toronto Clinical Scoring System score of >6, with a median of 8. Among DSPN patients, mild DSPN patients accounted for the largest proportion, specifically 63.33%, followed by moderate DSPN, and the least proportion was severe patients. Although the patients in the DSPN group were older, there was no significant difference between the three groups, nor was the sex ratio. There were no significant underweight or overweight participants in any of the three groups, with a body mass index of 22–27. As DSPN is a metabolic disease, clinical indicators associated with metabolism are well documented. There were no significant differences between DSPN and type 2 diabetes mellitus patients in blood pressure, glucose metabolism and liver function, but there were some differences in partial indexes, such as high‐density lipoprotein cholesterol, total cholesterol and estimated glomerular filtration rate. Specific parameters are outlined in Table 1.

Table 1.

General characteristics of participants in each group

Variables	Con (n = 30)	Type 2 diabetes mellitus (n = 30)	DSPN (n = 60)	P‐value (type 2 diabetes mellitus vs DSPN)
Median age, years (range)	61.50 (54.50–67.25)	57.50 (47.25–66.25)	63.00 (58.00–71.00)	0.001
Sex (female/male)	16/14	12/18	31/29	0.296
Median diabetes duration, years (range)	NA	5.21 (3.23–8.58)	7.17 (5.58–10.06)	0.033
Diabetes retinopathy, n (%)	NA	3 (10.00)	9 (15.00)	0.511
Diabetes nephropathy, n (%)	NA	0 (0.00)	3 (5.00)	0.213
Median BMI, kg/m2 (range)	23.72 (22.74–26.46)	24.39 (22.12–27.38)	24.82 (22.67–27.25)	0.851
Median TCSS score (range)	NA	1.50 (0.00–3.00)	8.00 (7.00–9.75)	<0.001
6–8, n (%)	NA	NA	38 (63.33)
9–11, n (%)	NA	NA	15 (25.00)
12–19, n (%)	NA	NA	7 (11.67)
Median symptom score (range)	NA	1.00 (0.00–2.00)	2.00 (1.00–2.00)	<0.001
Median reflex score (range)	NA	0.00 (0.00–1.00)	4.50 (4.00–6.00)	<0.001
Median sensory test score (range)	NA	0.00 (0.00–0.00)	2.00 (1.00–2.00)	<0.001
Median SBP, mmHg (range)	122.50 (113.25–130.00)	128.00 (119.75–137.25)	129.00 (123.25–134.00)	0.610
Median DBP, mmHg (range)	75.00 (69.00–80.00)	76.5 (72.7–84.00)	79.00 (75.00–82.00)	0.484
Median FPG, mmol/L (range)	4.86 (4.61–5.19)	6.60 (6.08–8.73)	7.00 (6.23–8.03)	0.681
Median HbA1c, % (range)	5.75 (5.50–5.90)	7.30 (6.48–9.63)	7.50 (6.63–8.78)	0.962
Median TG, mmol/L (range)	1.51 (0.93–2.24)	1.16 (0.79–2.04)	1.30 (0.90–1.68)	0.749
Median TC, mmol/L (range)	4.92 (4.39–5.63)	4.67 (4.20–5.12)	4.87 (3.87–5.66)	0.480
Median HDL‐C, mmol/L (range)	1.34 (1.05–1.60)	1.05 (0.93–1.24)	1.11 (0.93–1.35)	0.646
Median LDL‐C, mmol/L (range)	2.86 (2.47–3.50)	2.65 (2.41–3.35)	2.77 (1.98–3.42)	0.531
Median ALT/AST (range)	0.94 (0.60–1.16)	1.35 (0.98–1.58)	1.12 (0.82–1.33)	0.016
Median SCr μmol/L) (range)	71.50 (59.75–85.25)	64.00 (50.75–69.25)	60.50 (49.13–74.00)	0.834
Median eGFR, mL/min/1.73 m2 (range)	95.50 (87.00–108.50)	108.00 (101.00–127.25)	100.00 (84.00–112.00)	0.008

Data are expressed as a median (range). ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; Con, control group; DBP, diastolic blood pressure; DSPN, distal symmetric polyneuropathy group; eGFR, estimated glomerular filtration rate; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; SBP, systolic blood pressure; SCr, serum creatinine; TC, total cholesterol; TCSS, Toronto Clinical Scoring System; TG, triglyceride.

Global metabolite profiling

A total of 938 compounds of known identity were detected. Glucose, 1,5‐anhydroglucitol and metformin were all present at levels that track with diabetes status, confirming disease status. After the exclusion of obvious exogenous substances intervention factors, the remaining endogenous metabolites can be divided into seven various subclassifications, including amino acid, carbohydrate, lipid, peptide and so on (Figure S3). After log transformation and imputation of missing values, if any, the minimum value for each compound was observed. Metabolites differed significantly between experimental groups.

Compared with the Con group, A total of 342 significantly differential metabolites were identified (P < 0.05 and false discovery rate <0.05) in the type 2 diabetes mellitus group (222 up and 120 down). Similarily, in the comparison between the type 2 diabetes mellitus group and DSPN group, 53 differential metabolites were obtained, and all of them were significantly increased in the DSPN group (Table 2). Meanwhile, orthogonal partial least squares discrimination analysis was carried out in high‐dimensional metabolomic data analysis to show the differences between the groups. The results showed that the difference of metabolite profiles between the Con group and type 2 diabetes mellitus group was more significant, and the established model was more stable (Figure 1).

Table 2.

A summary of the significantly altered metabolite

Statistically significant biochemicals	Type 2 diabetes mellitus vs Con	DSPN vs type 2 diabetes mellitus
Total biochemicals (P < 0.05 & adjust P < 0.05)	342	53
Biochemicals (↑|↓)	222|120	53|0

Con, control group; DPSN, distal symmetric polyneuropathy group.

Figure 1.

Orthogonal partial least squares discrimination analysis between Con/type 2 diabetes mellitus and type 2 diabetes mellitus/distal symmetric polyneuropathy groups. (a) Orthogonal partial least squares discrimination analysis and model validation between Con and type 2 diabetes mellitus group; (b) Orthogonal partial least squares discrimination analysis and model validation between type 2 diabetes mellitus and distal symmetric polyneuropathy.

Identification of signature metabolites

Taking the intersection of the results from between‐group analysis, 30 metabolites were obtained. Heatmap visualization of >30 metabolites highlights their differential patterns of distribution among groups (Figure 2). Among them, only 12 metabolites showed the same change trend with disease phenotype. They were dimethylarginine (ADMA), N6‐acetyllysine, N‐acetylhistidine, N,N,N‐trimethyl‐alanylproline betaine (TMAP), cysteine, 7‐methylguanine (m7Gua), N6‐carbamoylthreonyladenosine, pseudouridine, 5‐methylthioadenosine (MTA), N2,N2‐dimethylguanosine (m2(2)G), aconitate and C‐glycosyl tryptophan. Boxed plots showed that the relative abundance of 12 signature metabolites exhibited an increased tendency with the progression of DSPN (Figure 3). In addition, all 12 metabolites showed an increasing trend in the stratification of the severity of DSPN patients (mild DSPN vs moderate/severe DSPN). Among them, three metabolites (TMAP, ADMA and C‐glycosyltryptophan) showed significant differences between groups (Figure S4). To avoid the influence of other microvascular complications of diabetes on the results, we excluded patients with retinopathy and nephropathy in the type 2 diabetes mellitus and DSPN groups for further analysis. The results showed that >12 metabolites remained significantly different between the type 2 diabetes mellitus group and the DSPN group. Furthermore, correlation analysis of age with individual metabolites also showed that age might not be a confounding factor (Table S1). All the aforementioned results suggested that the 12 metabolites we obtained are closely associated with the phenotype of patients with simple neuropathy (Figure S5).

Figure 2.

Heat map showing the intersection of differential metabolites among the three groups.

Figure 3.

Box plots of 12 screened signature metabolites. The abscissa represents the grouping and the ordinate represents the relative content of each metabolite, after log transformation. *P < 0.05; **P < 0.01; ***P < 0.001. MTA, 5‐methylthioadenosine; (m2(2)G), N2,N2‐dimethylguanosine; m7Gua, 7‐methylguanine; T2DM, type 2 diabetes; TMAP, N,N,N‐trimethyl‐alanylproline betaine.

The ROC curve and the correlation analyses

In the ROC curve analysis, the 12 signature metabolites showed good ability to differentiate DSPN patients, and all had an area under the curve (AUC) >0.75 (Figure 4). The three metabolites with the strongest discriminatory power were MTA (AUC 0.86, 95% confidence interval 0.80–0.92), N6‐acetyllysine (AUC 0.85, 95% confidence interval 0.77–0.94), and aconitate (AUC 0.84, 95% confidence interval 0.76–0.90). Correlation analysis between >12 metabolites and glycemia‐related measures showed a positive correlation, which is closely linked with disease phenotype. In addition, some of the metabolites were significantly associated with lipid metabolism indicators. No significant correlations were found between body mass index or blood pressure and most metabolites. Among them, aconitate, pseudouridine and TMAP were associated with more clinical metabolic indexes, suggesting the three metabolites might be more sensitive to changes in glucose and lipid metabolism (Figure 5).

Figure 4.

Receiver operating characteristic curves of 12 signature metabolites for distal symmetric polyneuropathy versus none distal symmetric polyneuropathy comparison. AUC, area under the curve.

Figure 5.

Correlation analysis between 12 signature metabolites and metabolism‐related indicators. *P < 0.05; **P < 0.01. BMI, body mass index; DBP, diastolic blood pressure; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin; LDL‐C, low‐density lipoprotein cholesterol; MTA, 5‐methylthioadenosine; (m2(2)G), N2,N2‐dimethylguanosine; m7Gua, 7‐methylguanine; TC, total cholesterol; TG, triglycerides; TMAP, N,N,N‐trimethyl‐alanylproline betaine.

DISCUSSION

Currently, there have been a small number of studies that have explored abnormal metabolites and metabolic pathways in diabetic neuropathy status. However, there were certain differences between the present study and previous studies. Some studies focusing on obese populations identified specific metabolites and metabolic pathways to differentiate obese individuals with and without peripheral neuropathy through the omics ¹⁹ , ²⁰ . Another study targeted the major risk factor for DSPN, lipids, and used lipidomics to assess the correlation between changes in the abundance of various lipid classes and neuropathy status ²¹ . In addition, some studies also found that the development of peripheral neuropathy might be associated with impaired tricarboxylic acid circulation and other glycolytic pathways by metabolomic methods. Although the results are similar to the present study, it belonged to basic research ²² , ²³ . In the present study, focusing on a non‐obese Chinese DSPN population, we explored the plasma metabolomics signatures of DSPN by applying a nontargeted metabolomics based on UPLC‐MS/MS. We found that 12 metabolites were significantly increased in the DSPN group as compared with other groups, and might become potential biomarkers of DSPN.

During the process of the metabolomics data screening, we used relatively broad criteria. However, we ensured that changes in concentrations of key metabolites were consistent with the disease phenotype. There, to a certain extent, the 12 signature metabolites we eventually screened might be closely related to the stepwise progression of DSPN. The ROC curves further proved that the 12 metabolites could reliably differentiate the DSPN status and non‐DSPN status, suggesting that the changes of their relative contents in patients with DSPN were specific. We did not carry out Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis because of the limited number of differential metabolites. However, the correlation analysis might provide insight into the metabolic process that the 12 signature metabolites might participate in. In combination with the biological properties of the metabolites and previous reports on the above‐mentioned differential metabolites in the literature, we finally divided them into seven categories, and summarized the metabolic processes and pathogenic mechanisms they might be involved in (Figure 6).

Figure 6.

The diagram for the possible mechanism of 12 signature metabolites in the development of distal symmetric polyneuropathy. eNOS, endothelial nitric oxide synthase; ROS, reactive oxygen species; TCA, tricarboxylic acid.

ADMA is a compound formed by the methylation of arginine residues in proteins and is released by proteolysis ²⁴ . It mainly exists in endothelial cells and smooth muscle cells of the cardiovascular system. Several studies showed that ADMA is an inhibitor of nitric oxide synthase, and that plasma levels of ADMA are closely associated with endothelial dysfunction ²⁵ . Therefore, ADMA is considered a cardiovascular risk factor ²⁶ and can be detected at high levels in various diseases, such as hypertension, coronary atherosclerotic heart disease, hypercholesterolemia and diabetes ²⁴ . The present study focused on DSPN patients and found significantly elevated ADMA in their plasma. As DSPN is essentially a microvascular disease, the increased ADMA is likely to be closely related to endothelial damage in DSPN patients. It was also reported that enhanced oxidative stress in cells could significantly inhibit the activity of dimethylarginine dimethylaminohydrolase (an enzyme that hydrolyzes ADMA) in hyperglycemic conditions. We speculated that the increased level of ADMA could affect endothelial function and blood supply by inhibiting nitric oxide synthase activities competitively, and finally affect neurological function ²⁷ , ²⁸ . Possible mechanisms are shown in Figure 6.

Another familiar metabolite is aconitate, an intermediate in the tricarboxylic acid cycle. It is well known that the tricarboxylic acid cycle is the center of the energy metabolic system ²⁹ . The results of correlation analysis in the present study also showed that aconitate was significantly associated with more clinical metabolic indices, which was consistent with previous studies and showed that aconitate is sensitive to energy metabolism changes in the body. Although the present study only found that aconitate was specifically elevated in DSPN patients, a comprehensive analysis of other intermediates changes throughout the tricarboxylic acid cycle in the present metabolomic results showed that the relative content of partial detectable intermediates, such as isocitrate, α‐ketoglutarate, succinate, fumarate and malate, exhibited an obvious decreasing trend in the DSPN group (Figure S6), especially the decrease in the content of isocitrate, which is directly located downstream of aconitate. The increase of aconitate content and decrease of isocitrate suggested that the activity of the intermediate metabolism enzyme, aconitase, might be inhibited (Figure 6).

Previous studies have also shown that the activity of aconitase decreases significantly in dorsal root ganglion neurons cultured in vitro after exposure to a hyperglycemic environment ³⁰ . In another study, the authors found that aconitase enzyme activity in neural tissue was gradually decreased from proximal to distal extremities in type 2 diabetic db/db mice through metabolomic methods ³¹ , which is consistent with the present findings. The aconitase activity can be inhibited by reactive oxygen species, such as O₂– and H₂O₂, which might strongly associate with a marked imbalance of energy metabolism and enhanced oxidative stress in the neural tissue of DSPN patients.

The metabolite, TMAP, was first identified in renal insufficiency. Subsequently, several studies observed that TMAP might be produced by the degradation of myosin light chain protein, and be associated with muscle and neurological diseases ³² , ³³ . In the present study, TMAP was found to be elevated in blood samples from DSPN patients for the first time, and correlation analysis results also suggested that TMAP was closely related to glucose and lipid metabolism. From this, we speculated that this metabolite might be pertinent to impaired muscle function influenced by lengthy metabolic disorder and neurovascular injury ³⁴ , ³⁵ .

MTA is a naturally occurring sulfur‐containing nucleoside. Recently, some studies have explained its role in neuroprotection, it was reported that MTA could exert some neuroprotective effects across the blood–brain barrier through multiple ways, including anti‐oxidant capacity ³⁶ . The presence of peripheral nerve damage in DSPN patients is well established, and whether the elevated MTA reflects the protective response of the body in the face of damage warrant future research.

Cysteine, a class of non‐essential amino acids we are familiar with, has been reported more than once in neuropathy‐related studies. Notably, a study, also from metabolomics, found significantly higher plasma levels of cysteine in patients with diabetic neuropathy in type 1 diabetes ³⁷ . The result was similar to the present findings, but the underlying mechanism of cysteine also deserves further investigation.

m7Gua and m2(2) are critical components of the transfer ribonucleic acid molecule, and are essential to maintain a stable cloverleaf structure. Previous studies have reported that aforementioned two metabolites are closely associated with the risk of developing type 2 diabetes mellitus ³⁸ . This is consistent with the present study. Our study not only identified the aforementioned differential metabolites simultaneously in the type 2 diabetes mellitus group, but also their specific elevation in patients who developed DSPN. In addition, it has been shown that circulating transfer ribonucleic acid fragments were correlated to oxidative stress ³⁹ , suggesting that both substances might also still be associated with significantly enhanced oxidative stress responses in DSPN patients. The last category of identified signature metabolites were N6‐acetyllysine, N‐acetylhistidine, pseudouridine, C‐glycosyl tryptophan and N6‐carbamoylthreonyladenosine. They all belong to modified metabolites and are subject to various post‐translational modifications, including acetylation, C‐glycosylation, formylation and chemical modification of messenger ribonucleic acid. There is almost no research on the aforementioned modified metabolites, except for a 2017 paper from Diabetes Care that identified the same metabolites that could be the biomarkers of type 1 diabetes patients with impaired renal function progressing to end‐stage renal disease by untargeted metabolomic approaches ⁴⁰ . Our studied population showed good baseline kidney function and no significant differences among three groups, which suggested that the modified metabolites identified in our study could be more likely related to the disease phenotype. It is worth noting that advanced glycation end‐products are also modifying metabolites that are now clearly involved in the development of DSPN, and whether these modifying metabolites identified in our study for the first time also contribute to disease progression deserve in‐depth research in the future.

In conclusion, we identified 12 signature differential metabolites in DSPN patients by a non‐targeted metabolomics approach. These differential metabolites include the reported metabolites, such as ADMA, cysteine, aconitate, m7Gua and m2(2)G, in previous studies, as well as those newly identified metabolites, such as N6‐acetyllysine, N‐acetylhistidine, pseudouridine, C‐glycosyltryptophan, N6‐carbamoylthreonyladenosine, TMAP and MTA, in the present study for the first time. They showed higher concentrations in DSPN patients when compared with the healthy control group and type 2 diabetes mellitus group. Elevated levels of signature metabolites are not only highly correlated with disease progression, but also with obesity and disorders of glucose and lipid metabolism. Future studies with larger clinical datasets are required to further validate the present findings. However, our speculation on the biological functions of the above signature metabolites does provide us with better ideas and entry points for subsequent targeted in‐depth studies of DSPN.

DISCLOSURE

The authors declare no conflict of interest.

Approval of the research protocol: The study protocol was reviewed and approved by the ethics committee of Shuguang Hospital Affiliated to Shanghai University of Traditional Chinese Medicine (No. 2018‐599‐28‐01).

Informed consent: The patients/participants provided their written informed consent to participate in this study.

Registry and the registration no. of the study/trial: N/A.

Animal studies: N/A.

Supporting information

Figure S1 | Preparation of specific technical replicates.

Figure S2 | Visualization of data normalization steps for a multiday platform run.

Figure S3 | Hierarchical clustering analysis of endogenous metabolites among the three groups.

Figure S4 | Box plots diagram of 12 signature metabolites in the mild distal symmetric polyneuropathy patients (n = 38) and moderate/severe distal symmetric polyneuropathy patients (n = 22).

Figure S5 | Box plots diagram of 12 signature metabolites in the type 2 diabetes mellitus (n = 27) and distal symmetric polyneuropathy groups (n = 49; patients without retinopathy and nephropathy were included for analysis).

Figure S6 | Box plots diagram of other five intermediates metabolites in the tricarboxylic acid cycle.

Table S1 | The correlation between 12 metabolites and age in all participants.