N‐lactoyl amino acids are potential biomarkers for insulin resistance and diabetic complications

Abstract

Aims

N‐lactoyl amino acids (Lac‐AA) are emerging as crucial players in metabolic research, with potential implications for disease mechanisms and therapeutic interventions. This study exploress the role of Lac‐AA in insulin resistance, type 2 diabetes (T2D), and its complications.

Materials and Methods

A cross‐sectional study was conducted using data from 2918 participants from Qatar Biobank. After quality control, 2907 individuals were retained and randomly divided into discovery (n = 1990) and validation (n = 917) cohorts. Untargeted metabolomics was employed to profile serum metabolites, and analysis was focused on three Lac‐AA species. Participants were stratified into insulin‐sensitive, insulin‐resistant, T2D without complications and T2D with complications. Associations with clinical traits were assessed using linear regression and Spearman correlation. Diagnostic performance was evaluated using Receiver Operating Characteristic (ROC) analysis in an independent cohort (n = 60). One‐sample Mendelian randomisation was performed to assess causality between genetic predisposition to T2D and Lac‐AA levels. Network analysis explored metabolic pathways linked to Lac‐AA.

Results

Lac‐AA levels were significantly higher in individuals with insulin resistance and diabetic complications. These findings were robustly replicated in the validation cohort. These metabolites showed strong positive correlations with markers of poor glycaemic control independent of metformin use. ROC analysis demonstrated that Lac‐AA could discriminate between insulin‐resistant and insulin‐sensitive individuals. Mendelian randomisation analysis indicated a potential causal association between genetic risk for T2D and increased Lac‐AA, particularly in patients with complications, supporting their role as downstream biomarkers of metabolic disease severity. Gaussian graphical model analysis revealed Lac‐AA as central nodes in metabolic networks, showing strong associations with mitochondrial dysfunction biomarkers.

Conclusions

Lac‐AA may serve as integrative biomarkers of metabolic dysfunction and diabetic complications. Further longitudinal and interventional studies are needed to clarify their mechanistic roles and clinical utility.

1. INTRODUCTION

N‐lactoyl‐amino acids (Lac‐AA) are a class of ubiquitous metabolites that have recently gained much research interest as bioactive metabolites and signalling molecules with implications in metabolism and disease. ¹ , ² , ³ , ⁴ Lac‐AA are rapidly formed through the reaction of specific amino acids with lactate, primarily catalysed by the enzyme cytosolic nonspecific dipeptidase 2 (CNDP2) via a process known as reverse proteolysis. ⁵ The controversy surrounding Lac‐AA centres on their conflicting research findings, with some studies suggesting potential metabolic benefits while others raise concerns about their complex interactions with human health conditions.

Alterations in levels of various members of Lac‐AA, especially N‐lactoyl phenylalanine (Lac‐Phe), were associated with physiological and pathological outcomes. Elevated levels of Lac‐AA were reported to be significantly increased in patients with mitochondrial encephalomyopathy lactic acidosis and stroke‐like episodes (MELAS), ² phenylketonuria, ⁵ rosacea, ⁶ diabetic retinopathy, ⁷ and maple syrup urine disease (MSUD). ⁵ Moreover, many members of Lac‐AA were classified as biomarkers of NADH reductive stress ² and have been shown to reflect mitochondrial dysfunction and overload, ⁸ , ⁹ predicting mortality in septic shock. ⁹

On the other hand, elevated levels of Lac‐Phe are associated with physical activity, ¹⁰ suppressing feeding and obesity and influencing systemic energy balance. ¹ , ¹¹ Moreover, metformin treatment was shown to have a profound effect on increasing Lac‐Phe levels in both mice and humans, which in turn mediates appetite‐suppressing and weight‐lowering properties. ³ , ⁴ Additionally, Lac‐Phe was negatively correlated with glycosylated haemoglobin (HbA1c) in T2D patients, suggesting that higher levels of this metabolite may be associated with improved glycaemic control. ⁸

The multifaceted roles of Lac‐AA extend beyond mere metabolic intermediates. The ongoing debate about their significance and relation to T2D risk underscores a complex metabolic network that remains incompletely understood. The objective of this study is to investigate the profile and role of Lac‐AA in the context of insulin resistance, diabetes and associated diabetic complications. This research aims to elucidate the metabolic pathways and identify the genetic predisposition involved in the fluctuation of these metabolites, particularly in metabolic diseases.

2. METHODS

2.1. Data source and study participants

This study utilised data from 2918 Qatari nationals and long‐term residents in Qatar, comprising males and females aged 18–80 years, recruited by Qatar Biobank (QBB). ¹² The data collection included a comprehensive socio‐demographic questionnaire and various clinical parameters. ¹³ Additionally, the dataset included information on medication usage, ¹⁴ past medical history of diabetes, and a metabolomics profile covering greater than 1000 metabolites using the Metabolon platform. ¹⁵ Participants were identified as having T2D if they met at least one of the following criteria: (i) HbA1c level >6.5%, (ii) fasting glucose ≥7 mmol/L, or (iii) self‐reported diagnosis of diabetes by a physician, or the use of diabetes‐related medications.

The control group was further categorised into insulin‐sensitive (IS) and insulin‐resistant (IR) groups based on BMI and Homeostasis Model Assessment of Insulin Resistance (HOMA‐IR) values. For participants with BMI ≥25 kg/m², insulin resistance was defined as HOMA‐IR >2.4, and for those with BMI <25 kg/m², insulin resistance was defined as HOMA‐IR >1.85, based on prior studies. ¹⁶ , ¹⁷ , ¹⁸

The T2D group was also stratified into two subgroups based on the presence or absence of diabetes‐related complications: T2D without complications (T2D‐NC) and T2D with complications (T2D‐C). Identification of complications relied on indirect indicators available in the Qatar Biobank dataset: (i) self‐reported medical history of diabetic complications, and (ii) medication usage records consistent with treatment for diabetes‐related complications. Participants classified as T2D‐NC were receiving anti‐diabetic medications but had no documented history or medication use indicative of diabetes‐related complications at the time of data collection. In contrast, T2D‐C participants met the diagnostic criteria for T2D and were also receiving treatment for at least one diabetes‐related complication, either microvascular (e.g., retinopathy, nephropathy, neuropathy) or macrovascular (e.g., cardiovascular or cerebrovascular disease). The presence of complications was determined by a combination of self‐reported diagnoses and concurrent use of relevant medications. Individuals in the T2D‐C group may present with multiple coexisting complications, and such overlap was anticipated. For the purpose of analysis, all individuals with one or more complications were grouped together, without further stratification by the number or type of complications.

The initial dataset consisted of 2918 participants. Prior to analysis, data quality assessments were conducted to ensure completeness and reliability. As a result, 11 participants were excluded due to incomplete or missing data. Following this exclusion, the final sample comprised 2907 participants. This cohort was then randomly allocated into two independent groups: Discovery cohort (n = 1900) and a validation cohort (n = 917). The overall study design and participant stratification are illustrated in Figure 1, and the baseline characteristics are detailed in Table 1.

FIGURE 1.

Study design depicting the characterisation of participants.

TABLE 1.

General characteristics of participants.

	Discovery cohort (n = 1990)	Validation cohort (n = 917)	p
IS/IR	893/708	393/345	0.27
T2D‐NC/T2D‐C	163/226	75/104	0.93
Sex M/F	999/991	471/446	0.58
Metformin users (%)	31.2	31.4	0.94
Age	38 (30–48)	39 (29–48)	0.876
BMI	28.33 (24.97–32.55)	28.32 (25.01–32.62)	0.785
Waist‐to‐hip ratio	0.84 (0.77–0.91)	0.84 (0.77–0.91)	0.367
Average systolic blood pressure (mmHg)	113 (103–123)	113 (105–124)	0.161
Average diastolic blood pressure (mmHg)	73 (66–79)	73 (66–80)	0.625
Average pulse rate (bpm)	69 (62–76)	68 (63–75)	0.873
Fasting blood glucose (mmol/L)	5.1 (4.74–5.7)	5.1 (4.7–5.7)	0.346
HbA1c (%)	5.4 (5.2–5.9)	5.5 (5.2–5.8)	0.42
Insulin (μU/mL)	10 (6.7–17.9)	10.5 (6.6–17)	0.426
C‐Peptide (ng/mL)	2.34 (1.65–3.42)	2.36 (1.63–3.36)	0.677
HOMA‐IR	2.35 (1.44–4.56)	2.44 (1.47–4.5)	0.605
Total cholesterol (mmol/L)	4.9 (4.3–5.55)	4.9 (4.36–5.5)	0.891
HDL‐cholesterol (mmol/L)	1.29 (1.08–1.54)	1.29 (1.08–1.56)	0.987
LDL‐cholesterol (mmol/L)	3 (2.32–3.53)	3 (2.4–3.5)	0.984
Triglycerides (mmol/L)	1.19 (0.82–1.75)	1.16 (0.8–1.7)	0.687
Haemoglobin (g/dL)	13.6 (12.4–14.9)	13.7 (12.4–14.8)	0.882
Creatinine (μmol/L)	66 (56–78)	66 (56–78)	0.577
Albumin (g/L)	45 (43–47)	45 (43–47)	0.615
Alkaline phosphatase (U/L)	67 (56–80)	68 (56–81)	0.512
ALT (U/L)	19 (13–29)	19 (13–28)	0.286
AST (U/L)	18 (15–22)	17 (15–21)	0.097
Bicarbonate (mmol/L)	26 (25–28)	26 (25–28)	0.68
Total bilirubin (μmol/L)	6 (4.47–8.6)	6 (4.23–8.38)	0.599
Calcium (mmol/L)	2.39 (2.33–2.45)	2.39 (2.33–2.45)	0.818
Total protein (g/L)	73 (70–76)	73 (70–76)	0.669
Urea (mmol/L)	4.3 (3.5–5.1)	4.3 (3.5–5.2)	0.391
Uric acid (μmol/L)	294 (239–350.75)	295 (242–357)	0.549

Note: Clinical parameters were checked for Gaussian distribution using Shapiro–Wilk's test. Data are then presented as median (IQR) or mean (SD). The median/mean between the study groups were compared using Mann–Whitney/Student t tests. A p‐value of <0.05 was considered significant.

Abbreviations: ALT, alanine transaminase; AST, aspartate aminotransferase; BMI, body mass index; HbA1c, glycated haemoglobin; HDL, high‐density lipoprotein; HOMA‐IR, Homeostasis Model Assessment of Insulin Resistance; IR, insulin‐resistant; IS, insulin‐sensitive; LDL, low‐density lipoprotein; T2D‐C, type 2 diabetes with complications; T2D‐NC, type 2 diabetes with no complications.

Analyses focusing on Lac‐AAs, the metabolites of interest in the study, were performed in the discovery cohort and subsequently confirmed in the validation cohort. Furthermore, Lac‐AA levels were evaluated for their ability to predict insulin resistance, as defined by HOMA‐IR, in an independent cohort of healthy Qatari individuals (n = 60) consisting of insulin‐sensitive (n = 21) and insulin‐resistant (n = 39) participants. Baseline characteristics of this independent cohort are detailed in Table S1.

2.2. Metabolomics

Metabolomic profiling was performed using Metabolon platform at Anti‐Doping Lab Qatar using established protocols that have been previously described. ¹⁵ Briefly, the methods involved the use of a Thermo Scientific Q‐Exactive high‐resolution/accurate mass spectrometer connected to a heated electrospray ionisation (HESI‐II) source and Orbitrap mass analyser with a mass resolution of 35 000 and Waters ACQUITY ultra‐performance liquid chromatography (UPLC). In short, methanol extraction was performed on serum samples to remove the protein component. The extract was separated into parts, each for a reverse‐phase UPLC tandem mass spectrometry (MS/MS) methods with positive ion mode electrospray ionisation (ESI), a reverse‐phase UPLC_MS/MS with negative ion mode ESI, a hydrophilic interaction chromatography/UPLC_MS/MS with negative ion mode ESI, and one was preserved as a backup. A detailed explanation of the procedure for LC–MS was previously described. ¹⁹ Briefly, raw peak data were identified, and quality procedures were performed using Metabolon hardware and software. ²⁰ The chemicals were identified by cross‐referencing them with either pure standard entries or frequently occurring unknown substances from an established library. This library included more than 3300 purified standard chemicals that are commercially available. The library matches were examined for each compound for each sample and modified as necessary.

2.3. One‐sample Mendelian randomisation

GWAS summary statistics for T2D and IR (HOMA2‐IR) were obtained from previously published studies conducted on the same Qatari participants used for this analysis. ²¹ , ²² The published GWAS for T2D involved 2765 T2D cases and 8671 controls, while the HOMA2‐IR analysis was conducted with 6217 non‐diabetic participants. We selected SNPs associated with T2D/HOMA2‐IR based on the GWAS summary statistics without BMI adjustment, since Spearman correlation showed no significant association between Lac‐Phe levels in serum and BMI.

Mendelian randomisation (MR) analysis was conducted for two outcomes: T2D and IR. For T2D, all SNPs reported in GWAS summary statistics were utilised as instrumental variables (IVs), including those that reached genome‐wide significance (p‐value <5 × 10⁻⁸) and replicated T2D‐susceptibility SNPs from the GWAS catalogue, to compute a single composite instrument. ²² In contrast, only replicated SNPs associated with HOMA2‐IR were included, as no genome‐wide significant SNPs were identified for this outcome. ²¹ We have implemented this flexible selection to include the replicated SNPs due to the limited number of reported genome‐wide significant SNPs, primarily because of the limited sample sizes affecting the power to identify GWAS‐level significant SNPs.

Following this, we selected proxy SNPs for each phenotype based on window sizes of 250 kb and a threshold of r ² <0.001 based on the genomics data of our study cohort. Subsequently, we investigated whether any of the proxy SNPs were significantly associated with BMI, age, and sex, as well as metformin use for SNPs associated with T2D or fasting‐time for SNPs associated with HOMA2‐IR, utilising three different databases including PheLiGe, ²³ the GWAS catalogue, and the Common Metabolic Disease Knowledge Portal. ²⁴

Total number of SNPs used to conduct MR analysis was 66 for T2D and 8 for IR (Table S2). Following this, the three core assumptions of MR were examined: (i) Relevance assumption: the association between the IVs and the exposure of interest was evaluated by assessing instrument strength using the F‐statistic, with a cut‐off value greater than 10 indicating no evidence of weak instruments; (ii) Independence assumption: the direct association between GRS and the outcome of interest was cross‐checked by linear regression analysis (p‐value <0.05), and (iii) Exclusion restriction criteria (no pleiotropy): we analysed potential pathways between the genetic variant(s) and the outcome other than via the exposure of interest using multivariate regression. This was conducted to investigate whether the selected SNPs were independent of the covariates, including age, sex, and BMI. Multivariate regression analysis identified 17 SNPs associated with T2D and 2 SNPs associated with IR that exhibited strong associations (p‐value <0.05) with the confounding variables. Consequently, these SNPs were excluded from the MR analyses to mitigate the effects of pleiotropy. Additionally, we performed a leave‐one‐out analysis to investigate the influence of individual SNPs on the overall results. Further validations were conducted by performing a t test between the constructed GRS and sex (p‐value 0.77), as well as regression models for BMI and age with the constructed GRS, which yielded p‐value s of 0.19 and 0.69, respectively.

In total, 49 SNPs associated with T2D and 6 SNPs associated with IR were utilised to construct a genetic risk score (GRS) by adding the number of risk alleles in each individual, weighted by the effect size from the genome‐wide association studies using PLINK version 1.9 software. ²⁵ The constructed GRS was used as an IV for performing one‐sample MR analysis; hence, both GWAS summary statistics and Lac‐Phe levels were obtained from the same population. The MR analysis for IR was not conducted as it violated the first assumption of MR, which requires a strong association with the phenotype. On the other hand, MR for T2D subgroups was taken in separate models to elucidate the causal relationship between genetic predisposition to T2D and Lac‐Phe levels using the ‘ivreg’ Package (https://zeileis.github.io/ivreg/) using R version 4.1.3, as previously shown in Lawn et al. ²⁶ Our MR model included the following explanatory variables: age, sex, BMI and metformin use for T2D.

2.4. Statistical analysis

Metabolomics data were log‐transformed for normality. QBB metabolomics data include 3 measurements of three Lac‐AA family members, namely N‐lactoyl phenylalanine, N‐lactoyl tyrosine (Lac‐Tyr), and N‐lactoyl valine (Lac‐Val). Linear regression models were performed for each metabolite (as the response variable) versus the phenotype of interest. Metabolites were compared between all control versus all T2D, IS versus IR, controls versus T2D‐NC, controls versus T2D‐C, and T2D‐NC versus T2D‐C. The model also included the following confounders: age, sex, BMI, fasting time and metformin use for each comparison. The model was first applied in the discovery cohort and then tested again in the validation cohort.

Partial correlations were calculated using the R package GeneNet based on the Gaussian graphical models (GGMs) using metabolite data measured from all participants. GGMs partial correlations with Lac‐AA metabolites scoring an FDR corrected p‐value of 0.05 were collectively displayed in a network using the software Cytoskape (version 3.10.2). The edges are based on the strength of correlation (correlation coefficient) and nodes reflect all metabolites significantly different between controls and T2D (FDR 0.05). To further validate a subset of our findings, a Receiver Operating Characteristic (ROC) curve analysis was performed for the three studied Lac‐AA in an independent cohort (n = 60). For ROC, the Wilson–Brown method was used to compute confidence intervals for sensitivity and specificity, and the area under the curve (AUC) was determined to assess predictive performance.

Shapiro–Wilk normality tests were performed to determine normality. Students' two‐tailed unpaired tests were performed for data with a normal distribution, and Mann–Whitney tests were performed for data with non‐normal distributions. Metabolomics data were analysed using built‐in linear model functions and MatchIt packages in R. MR was performed using the ivreg R package. GraphPad Prism was used for visualisation. Results were expressed as means ± standard error of the mean (SEM). Statistical analysis was performed using Excel (Microsoft Corp.), RStudio (version 4.0.3), BCFtools, Plink (version 1.9) and GraphPad Prism (GraphPad Software Inc., version 10.1.0).

3. RESULTS

3.1. Elevated Lac‐AA in insulin resistance, T2D, and diabetic complications

To investigate the levels of Lac‐AA in our cohort, linear regression analysis was performed, adjusting for age, sex, BMI, fasting time, and metformin use. The results revealed that serum levels of Lac‐Phe, Lac‐Tyr and Lac‐Val were significantly elevated in individuals with T2D compared to controls in both the discovery and validation cohorts (Figure 2A–C,G–I).

FIGURE 2.

Lac‐AA levels are elevated in the insulin resistant individuals and T2D patients with complications. Violin plots showing the linear regression results comparing the levels of Lac‐Phe, Lac‐Tyr, and Lac‐Val between phenotypes of interest. (A–C) Comparing levels of Lac‐AA between all non‐diabetic general population (Control, n = 1601) and all type 2 diabetic individuals (T2D, n = 389) from the discovery cohort. (D–F) Comparing the levels of metabolites after stratifying controls into insulin sensitive (IS, n = 893) and insulin resistant (IR, n = 708), and T2D into T2D without complications (T2D‐NC, n = 163) and T2D with complications (T2D‐C, n = 226). Similarly, (G–I) comparing levels between non‐diabetic population (Control, n = 738), and T2D (n = 179) from the validation cohort. (J–L) Comparing levels of Lac‐AAs after stratifying controls into insulin sensitive (IS, n = 393) and insulin resistant (IR, n = 345), and T2D into T2D without complications (T2D‐NC, n = 75) and T2D with complications (T2D‐C, n = 104). *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Moreover, when the control group was stratified into insulin‐resistant and insulin‐sensitive individuals, IR participants exhibited significantly higher levels of the three studied Lac‐AA (Figure 2D–F,J–L). Interestingly, when individuals with T2D were further stratified into those without complications (T2D‐NC) and those with complications (T2D‐C), elevated Lac‐AA levels were observed primarily in the T2D‐C subgroup, while levels in the T2D‐NC group were comparable to those in the control group (Figure 2D–F,J–L). These findings suggest that increased circulating Lac‐AAs may be associated not only with T2D but more strongly with its complications. (See Table S3 for additional details.)

Lac‐AA levels were also evaluated for their ability to predict insulin resistance, as defined by HOMA‐IR. Therefore, Receiver Operating Characteristic (ROC) curve analysis was performed for the three studied Lac‐AA in an independent cohort (n = 60) of insulin‐sensitive (n = 21) and insulin‐resistant (n = 39) individuals. ROC analysis (Figure 3) yielded an AUC of 0.823 (p < 0.0001) for Lac‐Phe, 0.852 (p < 0.0001) for Lac‐Val and 0.727 (p = 0.004) for Lac‐Tyr. This indicated a strong discriminatory power of Lac‐AA in distinguishing insulin‐sensitive from insulin‐resistant individuals.

FIGURE 3.

ROC curves showing the diagnostic performance of the three studied metabolites in differentiating insulin‐resistant and insulin‐sensitive individuals in an independent cohort (n = 60). (A) Lac‐Phe, (B) Lac‐Tyr, and (C) Lac‐Val. The area under the curve (AUC) values/effect size and p‐values indicate the ability of each Lac‐AA to discriminate between insulin‐resistant and insulin‐sensitive individuals.

3.2. Association between Lac‐AA and clinical traits

Spearman correlation demonstrated a strong positive association between all measured Lac‐AA levels and key metabolic indicators of IR and T2D, including the Homeostatic Model Assessment for Insulin Resistance (HOMA‐IR), glucose, and HbA1c (Figure 4A). Interestingly, the significant correlation persisted after excluding all metformin users from the analysis (Figure 4B). This suggests that the association between Lac‐AA levels and clinical traits of IR and T2D is independent of metformin use. These observations highlight the significant role of these metabolites in the pathophysiology of T2D and its associated complications. Furthermore, they suggest that Lac‐AA could serve as biomarkers for evaluating IR and the severity of T2D.

FIGURE 4.

Association of elevated Lac‐AA levels with clinical traits in IR and T2D is independent of metformin use, Spearman correlation analysis examining the relationships between Lac‐Phe, Lac‐Tyr, Lac‐Val, and various clinical traits, divided into two panels: Panel (A) includes all participants from the general population, while panel (B) excludes metformin users. Positive correlations are represented in blue, while negative correlations are shown in red. The size of the circles reflects correlation coefficient. Significance was defined as * (p‐value ≤0.05), ** (p‐value ≤0.01), *** (p‐value ≤0.001).

3.3. Mendelian randomisation (MR) analysis reveals causal association between T2D genetic predisposition and serum Lac‐Phe levels independently of metformin treatment

The causal relationship between IR/T2D and Lac‐Phe levels was investigated using a two‐stage least‐squares (2SLS) method for one‐sample MR analysis, after adjustment for metformin use, since the exposure (T2D subgroups, and IR) and outcome (Lac‐Phe levels) data were obtained from a single population.

The genetic risk scores (GRS) for T2D patients were significantly associated with T2D incidence (β = 0.92; 95% CI, 0.65–1.89; p‐value <0.0001). The constructed GRS of T2D was regressed against the exposure and each phenotype (control, n = 524, T2D, n = 533, T2D‐C, n = 311 and T2D‐NC = 222). The analysis revealed a significant association between genetic predisposition to T2D and Lac‐Phe levels in T2D patients (OR = 0.496; p‐value = 0.013). Upon further categorising, a strong association was detected in T2D patients with complications compared to non‐diabetic control (OR = 0.856; p‐value = 0.0002). However, no association was observed between T2D patients with no complications when compared with non‐diabetic control (OR = 0.056; p‐value = 0.82) or T2D with complications (OR = 0.096; p‐value = 0.93). A leave‐one‐out analysis was performed to validate the robustness of our results. The results indicate that the main estimate was not driven by any individual SNP, as shown in Figure S1. The strong causal association between genetic predisposition to T2D and Lac‐Phe after adjusting for metformin use suggests that T2D and its associated complications may play a potential causal role in the elevation of Lac‐AA levels independent of metformin treatment.

The results of the HOMA‐IR GRS showed no association with IR (β = 0.55; 95% CI, −0.48 to 1.59; p‐value = 0.29), possibly due to a poor statistical significance of the instrumental variables and insufficient SNPs‐associations; therefore, further analysis was not carried out. It is also noteworthy that a metabolite genome‐wide association study (mGWAS) was performed for Lac‐Phe to investigate whether Lac‐Phe levels drive insulin resistance or diabetes (forward causation). However, the mGWAS results showed no statistical significance; thus, the analysis was not pursued further (data not shown).

3.4. Network analysis reveals distinct and shared metabolic pathways associated with Lac‐AA in T2D

In order to shed light on the underlying metabolic pathways linking Lac‐AA and increased risk of T2D, the association between Lac‐AA and other metabolites was explored using Gaussian Graphical Model (GGM) network analysis. ²⁷ , ²⁸ Figure 5 illustrates partial correlations between Lac‐AA, specifically Lac‐Tyr, Lac‐Val, and Lac‐Phe, and metabolites in the general cohort. The model exhibited a strong positive correlation among the three Lac‐AA as expected. Lac‐Val showed significant correlations with various amino acids, fatty acids, and intermediates involved in energy metabolism. Lac‐Tyr showed significant correlations with a distinct set of metabolites involved in amino acid and lipid metabolism. Lac‐Phe was correlated with both shared and unique metabolites, suggesting its potential role in connecting various metabolic pathways in T2D.

FIGURE 5.

Metabolomics network analysis reveals metabolic pathways associated with Lac‐AA in T2D. A Gaussian graphical model (GGM) representing the partial correlations between Lac‐AA and T2D‐associated metabolites in the general cohort is shown. Red colour indicates positive correlation. Blue colour indicates negative correlation. Each node corresponds to a metabolite, and the edges indicate partial correlations, with the thickness of the edges reflecting the strength of these correlations.

4. DISCUSSION

N‐lactoyl‐amino acids are recently identified metabolites that have emerged as promising biomarkers for metabolic diseases. Our study integrates metabolic, genetic, and clinical data to deepen the understanding of Lac‐AA contribution to disease pathogenesis and progression. Our results provide strong evidence that N‐lactoyl‐amino acids are associated with insulin resistance and diabetic complications independent of metformin usage. Our findings were supported by multiple statistical approaches, including linear regression, receiver operating characteristic analysis, correlation studies, Mendelian randomisation, and network analysis.

Linear regression analysis in both the discovery and validation cohort demonstrated that serum levels of three studied Lac‐AA are significantly higher in individuals with T2D compared to controls. Notably, among T2D patients, only those with complications showed elevated Lac‐AA levels. Within the control group, insulin‐resistant individuals also exhibited higher Lac‐AA levels than their insulin‐sensitive counterparts. ROC analysis in a small independent cohort confirmed that these Lac‐AA metabolites have a good discriminatory power for distinguishing between IR and IS individuals. Our results are in line with previous studies which have established a significant link between metabolic diseases and Lac‐AA, revealing important insights into their complex relationship. Scott et al. ³ reported that Lac‐Phe levels were 5.7‐fold higher in obese T2D individuals compared to obese non‐T2D controls, and significantly higher in obese T2D when compared to pre‐diabetic individuals. Additionally, Fernandes et al. ⁷ showed that Lac‐AA were associated with an increased risk of diabetic retinopathy. Furthermore, our previous investigations identified Lac‐Phe as the most discriminating metabolite between insulin sensitive and insulin resistant individuals. ¹⁷ , ²⁹ Recently, Sharma et al. ³⁰ demonstrated that the progression from prediabetes to diabetes could be mediated by Lac‐AA.

Our correlation analysis showed that, in both metformin users and non‐users, the three studied Lac‐AA demonstrate strong and statistically significant positive correlations with key markers of insulin resistance and T2D, including fasting insulin, C‐peptide, HOMA‐IR, glucose, and HbA1c. Further reinforcing their association with metabolic dysfunction, the Lac‐AA correlate positively with triglycerides and negatively with HDL cholesterol in both groups. This lipid profile is characteristic of dyslipidaemia commonly observed in insulin‐resistant states and contributes to cardiovascular risk. Additionally, Lac‐AA are positively associated with markers of compromised renal function, namely creatinine, urea, and uric acid. These findings suggest a potential link between elevated Lac‐AAs and early kidney dysfunction, a common complication of T2D. The presence of these associations in both groups implies that Lac‐AA may be involved in renal pathophysiology independently of metformin treatment. Similarly, positive correlations with hepatic enzymes such as ALT and AST hint at a possible connection between Lac‐AA and hepatic stress or steatosis, further broadening the scope of their clinical relevance in diabetes‐related organ complications. Interestingly, markers of inflammation such as C‐reactive protein show weak or no consistent associations with Lac‐AA, which might suggest that these metabolites are more reflective of metabolic, rather than inflammatory, dysregulation. Nonetheless, the consistent associations of Lac‐AA with glycaemic control, lipid abnormalities, and renal and hepatic markers strongly support their potential role as integrative biomarkers of metabolic health and disease progression.

Mendelian randomisation analysis using a two‐stage least squares approach further supported the association between genetic predisposition to T2D and elevated Lac‐AA levels. This association was significant even after adjustment for metformin, reinforcing the argument that Lac‐AA elevation may result from the pathobiological processes of T2D, especially in individuals with complications. Interestingly, this link was not present in T2D patients without complications or when examining the HOMA‐IR genetic risk score, suggesting that Lac‐AA elevation may be more closely related to diabetes progression and complications than to early insulin resistance. Additionally, the lack of significant findings in forward causation analysis via metabolite GWAS (mGWAS) indicates their elevation is likely a response to, rather than a cause of, advanced metabolic disease. These results suggest that Lac‐AA elevation is not merely a marker of diabetes per se, but is closely linked to the presence of severe insulin resistance and the development of diabetic complications. The relationship between insulin resistance and diabetic complications is well established in the scientific literature. ³¹ , ³² , ³³ Our stratification underscores the specificity of Lac‐AA as a potential biomarker for metabolic dysfunction severity, rather than just T2D diagnosis.

Our GGM analysis revealed a strong positive correlation among the three studied Lac‐AA. These findings are consistent with previous studies confirming the function of Lac‐AA as a cohesive family. ² , ³ Expectedly, the GGM showed a significant correlation of Lac‐AA with amino acids and derivatives resulting from CNDP2 action, such as histidine and alanine. Indeed, the enzyme hydrolyses carnosine into histidine and alanine and cleaves threonyl dipeptides. ³⁴ Intriguingly, Lac‐Phe was found to exhibit a positive correlation with fructose levels. Fructose has been associated with an increased risk of incident T2D. ³⁵ Furthermore, elevated fructose concentrations have also been implicated in the development and progression of diabetic retinopathy among individuals with T2D. ³⁶ Lac‐Phe also showed a notable positive association with N6‐carbamoylthreonyladenosine, a promising biomarker for kidney function, showing a strong association with declining GFR in CKD patients. ³⁷ , ³⁸ Interestingly, the GGM identified a significant correlation between Lac‐AA and metabolites that are indicative of mitochondrial dysfunction. One key finding was a negative correlation with uridine, a glycosylated pyrimidine that plays a crucial role in various metabolic pathways, particularly those related to mitochondrial function. ³⁹ , ⁴⁰ In contrast, a positive correlation was revealed with aconitate and succinyl carnitine. The accumulation of these metabolites is often linked to mitochondrial impairment. ⁴¹ , ⁴² These findings enhance our existing understanding that Lac‐Phe is a potential indicator of mitochondrial dysfunction. ² , ⁸ , ⁹

Given the biosynthetic context of Lac‐AA, their accumulation is most likely a reflection of increased glycolytic flux secondary to impaired mitochondrial oxidative metabolism. Thus, Lac‐AA may serve as sensitive markers of mitochondrial stress in insulin resistance and diabetes, particularly in the setting of diabetic complications, rather than simply indicating systemic metabolic imbalance. Our observations may stem from the metabolic and biochemical factors driving Lac‐AA accumulation. Indeed, the formation of Lac‐AA is thermodynamically favoured when lactate and amino acids are present in high concentrations. ⁵

Mitochondrial dysfunction is a hallmark of insulin resistance and type 2 diabetes, and is further exacerbated in diabetic complications. ⁴³ , ⁴⁴ Impaired mitochondrial oxidative phosphorylation leads to decreased ATP production and a compensatory increase in anaerobic glycolysis. This metabolic shift results in the accumulation of glycolytic intermediates, most notably pyruvate and its reduction product, lactate. ⁴⁵ Insulin resistance and mitochondrial impairment are also associated with altered amino acid metabolism, particularly elevated plasma levels of branched‐chain amino acids (BCAAs) and aromatic amino acids like phenylalanine. ⁴⁶ , ⁴⁷ Impaired catabolism of these amino acids further increases their availability for conjugation with lactate.

Diabetic complications are further characterised by tissue hypoxia, oxidative stress, and chronic inflammation. ⁴⁸ These conditions further compromise mitochondrial function, intensifying the shift towards anaerobic glycolysis and lactate production. The resulting metabolic milieu is thus highly conducive to Lac‐AA accumulation.

Our study has several limitations that should be acknowledged. First, the MR analysis, while leveraging genetic variants that have reached genome‐wide significance in larger association studies, is constrained by the limited statistical power of our own cohort. Although we performed multivariable analyses to assess pleiotropy and confirmed that our constructed GRS was significantly associated with the studied phenotype, the MR assumption for insulin resistance versus insulin sensitivity was not fully met. Specifically, the number of proxy SNPs available for the HOMA‐IR GRS was limited, resulting in a weak genetic association and potential instrument bias for this phenotype. Additionally, the lack of detailed data on treatment duration and dosage for all participants may have introduced residual confounding, as treatment intensity is a known factor that can influence metabolite levels in biomarker studies. However, our analyses are robust due to the large sample size and substantial heterogeneity in both treatment regimens and medication combinations among participants, which likely minimises the influence of any single treatment pattern on the observed metabolite associations. Furthermore, the cross‐sectional design of our primary analyses restricts our ability to draw definitive conclusions about temporal or causal relationships between Lac‐AA levels and disease progression. While our findings are supported by multiple statistical approaches, the independent validation cohort used for ROC analysis was relatively small, which may limit the robustness and generalisability of the diagnostic performance results. Therefore, these findings should be interpreted with caution and validated in larger and independent cohorts. Although we adjusted for key covariates such as metformin use, our analysis was limited by the absence of detailed socioeconomic, lifestyle, and dietary data, which prevented us from fully accounting for these important confounders. Despite these limitations, our study provides important new insights into the role of Lac‐AA as potential biomarkers of metabolic dysfunction and diabetic complications.

5. CONCLUSION

This study establishes N‐lactoyl amino acids as promising biomarkers for insulin resistance and diabetic complications. Their robust associations with clinical and metabolic indicators, independent of metformin use, suggest strong pathophysiological relevance. Mendelian randomisation supports a potentially causal relationship between T2D genetic risk and elevated Lac‐AA, reinforcing their role as downstream markers of disease severity. Future studies are warranted to elucidate the precise mechanistic pathways linking N‐lactoyl amino acids to metabolic dysfunction and diabetic complications through longitudinal and interventional designs. Future research should focus on translating these findings into clinical tools for risk stratification, early intervention, and personalised treatment strategies for T2D and its complications.

AUTHOR CONTRIBUTIONS

Khaled Naja contributed to data analysis and manuscript writing. Asma A. Elashi, Aleem Razzaq, and Shamma Almuraikhy performed GWAS and Mendelian randomisation analyses and also contributed to manuscript writing. Najeha Anwardeen, Laila Hedaya, and Ilhame Diboun contributed to statistical, metabolomics, and GGM analyses and assisted in manuscript writing. Karsten Suhre and Omar AlBagha provided critical review. Mohamed A. Elrayess conceived the idea, supervised the project, and contributed to manuscript writing. All authors have reviewed and approved the final manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

PEER REVIEW

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/dom.16633.

ETHICS STATEMENT

The study was approved by the Institutional Review Boards of the Qatar Biobank (E‐2024‐QF‐QBB‐RES‐ACC‐00158‐0269) and by the Institutional Review Board of Qatar University (QU‐IRB 157/2024‐EM). Informed consent was obtained from all participants involved in the study.

Supporting information

Data S1.

Naja K, Elashi AA, Anwardeen N, et al. N‐lactoyl amino acids are potential biomarkers for insulin resistance and diabetic complications. Diabetes Obes Metab. 2025;27(10):5793‐5804. doi: 10.1111/dom.16633

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.