Urine Tricarboxylic Acid Cycle Metabolites and Risk of End-stage Kidney Disease in Patients With Type 2 Diabetes

Abstract

Context

Metabolites in the tricarboxylic acid (TCA) pathway have pleiotropic functions.

Objective

To study the association between urine TCA cycle metabolites and the risk for chronic kidney disease progression in individuals with type 2 diabetes.

Design, setting and participants

A prospective study in a discovery (n = 1826) and a validation (n = 1235) cohort of people with type 2 diabetes in a regional hospital and a primary care facility

Exposure and Outcome

Urine lactate, pyruvate, citrate, alpha-ketoglutarate, succinate, fumarate, and malate were measured by mass spectrometry. Chronic kidney disease progression was defined as a composite of sustained estimated glomerular filtration rate below 15 mL/min/1.73 m², dialysis, renal death, or doubling of serum creatinine.

Results

During a median of 9.2 (interquartile range 8.1-9.7) and 4.0 (3.2-5.1) years of follow-up, 213 and 107 renal events were identified. Cox regression suggested that urine lactate, fumarate, and malate were associated with an increased risk (adjusted hazard ratio, [95% CI] 1.63 [1.16-2.28], 1.82 [1.17-2.82], and 1.49 [1.05-2.11], per SD), whereas citrate was associated with a low risk (aHR 0.83 [0.72-0.96] per SD) for the renal outcome after adjustment for cardiorenal risk factors. These findings were reproducible in the validation cohort. Noteworthy, fumarate and citrate were independently associated with the renal outcome after additional adjustment for other metabolites.

Conclusion

Urine fumarate and citrate predict the risk for progression to end-stage kidney disease independent of clinical risk factors and other urine metabolites. These 2 metabolites in TCA cycle pathway may play important roles in the pathophysiological network, underpinning progressive loss of kidney function in patients with type 2 diabetes.

tricarboxylic acid cycle, urine metabolomics, fumarate, citrate, end stage kidney disease, type 2 diabetes

Diabetic kidney disease (DKD) is not only a leading cause of end-stage kidney disease (ESKD) but also an established risk factor for cardiovascular disease (CVD) and premature death (1). The prevalence of DKD varies greatly among ethnic groups (1, 2). Compared with those of European descent, Asian people with diabetes are more susceptible to kidney injury and experience a higher risk of progression to ESKD (1, 3).

Beside filtration and reabsorption, the kidney performs various other physiologic functions including secretion, biosynthesis, and maintenance of fluid and electrolyte homeostasis. Hence, the kidney has a large number of mitochondria to fulfill the high demand of energy (4). Tricarboxylic acid (TCA) cycle is the central hub of energy production that integrates glucose, lipid, and amino acid metabolism in mitochondria. Intermediate metabolites in the TCA cycle pathway also regulate redox balance, provide substrate for macromolecule biosynthesis, function as messengers for interorgan crosstalk, and determine cell fate by modulating gene expression (5, 6). Robust preclinical studies have shown that TCA cycle metabolites may be causally involved in kidney pathology beyond bioenergetics (6-8). For example, Zecchini et al recently reported that accumulation of fumarate might cause leakage of mitochondrial DNA into cytosol and activate innate inflammatory response in the kidney (9).

Early clinical studies, including our own work, have associated urine TCA cycle metabolites with DKD pathogenesis and progression (10-12). However, these studies were limited by small sample size (13), cross-sectional design (10, 11), short follow-up duration (12, 13), or taking surrogate instead of “hard” clinical outcomes as endpoints (12). In this context, we sought to systematically examine the associations between urine TCA cycle-related metabolites (lactate, pyruvate, citrate, alpha-ketoglutarate, succinate, fumarate, and malate) and the risk of progression to ESKD in 2 cohorts of patients with type 2 diabetes.

Materials and Methods

We adopted a discovery and validation approach in this prospective study. Discovery substudy was nested in Singapore Study of Macro-angiopathy and Micro-Vascular Reactivity in Type 2 Diabetes (SMART2D) cohort. The validation substudy was nested in the Singapore Khoo Teck Puat Hospital-Diabetic Kidney Disease (KTPH-DKD) cohort.

Participants and Follow-up

SMART2D is an ongoing cohort study on vascular complications in multiethnic Southeast Asian patients with type 2 diabetes. In brief, 2057 outpatients with type 2 diabetes were recruited from a regional hospital and a primary care facility between August 2011 and March 2014 (14, 15). Patients who were pregnant, on active treatments for autoimmune disease, cancer, infection, and those with kidney disease attributable to specific causes were excluded. We also excluded patients with high risk of hyperglycemic decompensation (point-of-care fasting glucose >15.0 mM or HbA1c >12% after phlebotomy) from cohort enrolment because their clinical course of disease might differ from the majority of outpatients with relative stable glycemic control. Participants were followed every 2 years by reviewing their electrical medical records in a centralized data repository. They were also invited for in-person research visits every 3 years, during which clinical data including major cardiorenal events were collected. We combined data from routine clinical care and research visits into 1 dataset to ascertain clinical outcomes. Follow-up was censored on December 31, 2021.

Details of Singapore KTPH-DKD cohort have been described previously (2, 16). Briefly, patients with type 2 diabetes who visited the hospital were recruited consecutively between March 2004 and October 2017. Patients who were pregnant, had kidney disease attributable to specific causes, and those on active treatments for autoimmune disease, infectious disease, and cancer were excluded. Participants were followed by reviewing electronic medical records and by data linkage with national disease registries. Follow-up for this study was censored on March 31, 2020. We considered participants recruited from January 2011 onward as candidates for the current study because the centralized electronic medical records were accessible from that time.

We included participants with baseline estimated glomerular filtration rate (eGFR) ≥ 30 mL/min/1.73 m² for both discovery and validation substudies. Participant selection is illustrated in Supplementary Fig. S1 (17).

The study was approved by Singapore National Healthcare Group Domain Specific Review Committee. Each participant gave informed consent.

Definition of Clinical Outcome

The primary outcome was a composite of progression to ESKD or doubling of serum creatinine. ESKD was defined as progression to eGFR < 15 mL/min/1.73 m² with at least 1 confirmatory measurement 3 months apart, sustained dialysis for more than 3 months, or death attributable to renal causes, whichever occurred first. Death events were identified from electronic medical records and by data linkage with the national death registry. Renal death was identified according to the primary cause of death on death certificate. Because serum creatinine was not measured at a regular interval in observational study, we identified doubling of serum creatinine (ie, 57% decline of eGFR) by extrapolation from eGFR slope derived from linear mixed model (random intercept, random slope, time coefficient) (18). We excluded eGFR readings measured during hospitalization in slope calculation.

We conducted a case-control substudy nested in KTPH-DKD cohort to examine the correlations between urine and plasma metabolites (citrate and fumarate) and their associations with chronic kidney disease (CKD) progression. We defined CKD progression in this substudy as eGFR decline 3 mL/min/1.73 m² or greater per year in more than 2 years of follow-up (19). We defined those with eGFR change ± 2 mL/min/1.73 m² per year in more than 5 years of follow-up as nonprogressors.

Clinical and Biochemical Variables

Sex, ethnicity, and smoking status were self-reported. CVD history that included myocardial infarction and stroke was self-reported and ascertained by review of medical records. Medication usage was obtained from medication dispensary database. Blood pressure was measured 3 times by a semi-auto sphygmomanometer in the discovery cohort and the mean value was used, whereas it was measured once in the validation cohort. Mean artery pressure was calculated as (systolic + 2×diastolic blood pressure)/3. HbA1c was measured by a point-of-care analyzer (DCA Vantage Analyzer, Siemens, Germany) in the discovery cohort, whereas it was measured by immunoturbidimetric method (Cobas Integra 800 Chemistry Analyzer, Roche, Basel, Switzerland) in the validation cohort. Creatinine was quantified by an enzymatic method that was traceable to an isotope dilution mass spectrometry reference in both cohorts. We calculated eGFR based on serum creatinine by the 2009 CKD-EPI equation. Urinary albumin was quantified by an immunoturbidimetric assay (Roche Cobas c, Roche Diagnostics, Mannheim, Germany) and presented as an albumin-to-creatinine ratio (ACR).

Quantification of Urine TCA Cycle Metabolites by gas Chromatography Mass Spectrometry

Urine specimens were collected after overnight fasting, aliquoted, and stored at −80°C. Samples used for the current study underwent only 1 freeze-thaw cycle. To account for run-order effects, samples in each cohort were coded and randomized, respectively, before sample preparation. For urine samples with creatinine concentration above 4 mM, the samples were diluted to 4 mM creatinine by ultrapure water before metabolite extraction. For samples with creatinine concentration below 4 mM (∼20% of all samples), neat urine samples were used for metabolite extraction but final concentrations of metabolites were normalized to 4 mM creatinine by ratio before data analysis.

Stable isotope-labeled internal standards were spiked into each study sample. Metabolites were then extracted by ethyl acetate after adding ethoxyamine to complete ethoxylation of ketoacids. The extractant was dried and derivatized with N,O-Bis(trimethylsilyl)trifluoroacetamide to form trimethylsilyl derivatives. The trimethylsilyl derivatives were separated on a VF-1 MS column on Agilent Technologies HP 7890A GC system and quantified by selected ion monitoring on a 5975C mass spectrometer. MS data were analyzed on a Mass Hunter Workstation vB.06.00 and quantification was based on stable isotope dilution. A total of 99.5% of metabolite readings were within the calibration curves. An aliquot from each study sample was pooled as quality control (QC) sample. The QC sample was injected after every eighth study sample. The SD of QC, which indicated stability of metabolite responses on MS, was < 5% for all 7 metabolites in both cohorts.

We measured the same panel of urine metabolites in 567 participants in discovery cohort in year 2016 (12) and remeasured them with other samples in year 2022 for the current study. The correlation coefficients between 2 measurements ranged between 0.82 and 0.93 (Supplementary Table S1) (17).

Blood sample was collected in EDTA-coated tube after overnight fasting. Plasma was separated by centrifugation, aliquoted, and stored at −80°C. Plasma citrate and fumarate concentrations were measured using gas chromatography MS as described previously.

Statistical Analysis

We presented clinical and biochemical variables as mean ± SD, median (interquartile range [IQR]), or percentage. Urine ACR and metabolites were log-transformed and scaled by SD. Binary correlation was assessed by Spearman correlation analysis. We handled missing values of clinical variables by listwise deletion because of very low missingness (Table 1).

Table 1.

Participant baseline clinical and biochemical characteristics

	Discovery cohort		Validation cohort
	N		N
Age (y)	1826	57.2 ± 10.7	1235	55.2 ± 12.4
Male sex (%)	1826	51.2	1235	60.1
Ethnicity	1826		1235
Chinese		51.9		46.9
Malay		20.9		31.9
Asian Indian		27.2		21.2
CVD history (%)	1826	7.9	1235	16.2
Active smokers (%)	1823	8.9	1234	16.0
Diabetes duration (y)	1818	10 (4-15)	1229	10 (5-16)
Body mass index (kg/m2)	1824	27.7 ± 5.2	1220	27.5 ± 5.6
HbA1c (%)	1826	7.8 ± 1.3	1229	8.7 ± 2.1
HbA1c (mmol/mol)	1826	61.7 ± 10.0	1229	71.6 ± 17.3
Blood pressure (mm Hg)
Systolic	1826	140 ± 18	1227	137 ± 19
Diastolic	1826	79 ± 9	1227	76 ± 12
Mean artery pressure	1826	99 ± 11	1227	97 ± 12
eGFR (mL/min/1.73 m2)	1826	89 ± 23	1235	87 ± 28
Urine ACR (mg/g)	1824	21 (6-81)	1235	37 (11-183)
Medication usage (%)
Insulin	1816	26.7	1235	43.5
RAS blocker	1812	59.4	1235	54.5

N refers to sample size with the variable available.

Abbreviations: ACR, albumin-to-creatinine ratio; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; RAS, renin-angiotensin system.

We employed Cox regression models to study the association between metabolite and the renal outcome. Covariates were a priori determined based on biological plausibility. We included age, sex, ethnicity (Chinese as reference), active smoking (yes or no), CVD history (yes or no), renin-angiotensin system blocker usage (yes or no), body mass index, diabetes duration, HbA1c, mean artery pressure, baseline eGFR, and urine ACR in the multivariable model. We tested proportional hazard assumption by modelling the product of metabolite × time as a covariate. Urine lactate, fumarate, and malate violated proportional hazard assumption in discovery substudy. Hence, they were modelled as variables with time-varying coefficients.

We performed the following sensitivity analyses: (1) we redefined the renal outcome as incident ESKD (sustained progression to eGFR < 15 mL/min/1.73 m², dialysis, or renal death); (2) we combined renal events and nonrenal death as a composite outcome to study whether nonrenal death might affect the association between urine metabolites and renal outcome as a competing risk; and (3) we tested whether urine metabolites interacted with CKD status (eGFR ≥ 60 vs < 60 mL/min/1.73 m²), albuminuria levels (ACR < 30, 30-299, and ≥300 mg/g), sex, and ethnicity in their associations with the renal outcome by adding the multiplicative interaction term as a covariate in the Cox regression models.

We employed binary logistic regression models to study the association between urine citrate, plasma citrate, and CKD progression. CKD progression (yes or no) was the binary outcome. Urine and plasma citrate were included in the same model as covariates. The same set of clinical and biochemical risk factors were adjusted as that in the Cox regression models discussed previously. We used the same approach to study the relationship between urine and plasma fumarate and their associations with CKD progression.

Data analyses were performed using SPSS (version 27) and R software (version 3.4.2). A 2-sided P value <.05 was considered statistically significant.

Results

Participant Characteristics

A total of 1826 participants were included in the discovery substudy. Participant average age was 57 (SD 11) years, 51% were male, and median of diabetes duration was 10 (IQR 4-15) years (Table 1). During a median of 9.2 (IQR 8.1-9.7) years of follow-up (15, 635 patient-years), we identified 213 renal events. Among them, 119 progressed to sustained eGFR <15 mL/min/1.73 m², 1 was identified as sustained dialysis, 5 death events were attributable to renal causes, and 88 experienced doubling of serum creatinine. The crude incidence rate was 1.36 per 100 patient-years.

A total of 1235 participants were included in the validation substudy. Participant baseline characteristics were similar to the discovery cohort except that they had a higher HbA1c and were more likely on insulin treatment (Table 1). During a median of 4.0 (IQR 3.2-5.1) years of follow-up (5, 263 patient-years), we identified 107 renal events. Among them, 57 progressed to sustained eGFR <15 mL/min/1.73 m², 3 death events were attributable to renal causes, and 47 experienced doubling of serum creatinine. The crude incidence rate was 2.03 per 100 patient-years.

Compared with participants without events, those with renal event occurrence had higher levels of HbA1c, blood pressure, urine ACR, a lower eGFR, and were more likely to be of Malay ethnicity (Supplementary Table S2) (17).

Urine metabolites were only modestly correlated with urine ACR (all Spearman rho < 0.30). Urine citrate was positively correlated with eGFR (Spearman rho 0.55 and 0.46), whereas the other metabolites were only modestly or moderately correlated with eGFR (Spearman rho between 0.06 and 0.38, Supplementary Table S3) (17).

Association of Urine TCA Cycle Metabolites With the Composite Renal Outcome in Discovery substudy

In univariable analysis, an increasing concentration of urine lactate, fumarate, and malate was associated with a higher risk of the composite renal outcome, whereas a higher concentration of citrate and succinate was associated with a lower risk. In multivariable model, 1 SD increment in lactate, fumarate, and malate was associated with 1.63- (95% CI, 1.16-2.28), 1.82- (1.17-2.82), and 1.49- (1.05-2.11) fold increased risk, whereas 1 SD increment in urine citrate was associated with a 17% (adjusted hazard radio 0.83 [0.72-0.96]) lower risk of the renal outcome after adjustment for cardiorenal risk factors. Urine pyruvate and alpha-ketoglutarate were not significantly associated with renal outcome in either univariable or multivariable analysis (Fig. 1). We obtained consistent outcomes when we analyzed urine metabolites as categorical variables in quartiles (Supplementary Table S4) (17).

Figure 1.

Association of 7 urine TCA cycle metabolites with the composite renal outcome in the discovery (SMART2D) cohort.

Association Between Urine TCA Cycle Metabolites and Renal Outcome in Validation substudy

All 4 urine metabolites in association with the composite renal outcome identified in the discovery substudy were also significantly associated with the renal outcome in the validation cohort in univariable analysis. The associations between urine lactate, fumarate, malate, and the composite renal outcome remained statistically significant after adjustment for cardiorenal risk factors. The association between urine citrate and the renal outcome was statistically significant after adjustment for cardiometabolic risk factors and urine ACR (adjusted hazard ratio 0.77 [0.67-0.89]). However, the strength of association was markedly attenuated after further adjustment for eGFR (adjusted hazard ratio 0.93 [0.78-1.10], Fig. 2). Consistent outcomes were obtained when the metabolites were analyzed as categorical variables (Supplementary Table S5) (17).

Figure 2.

Validation of the associations between urine lactate, citrate, fumarate, malate, and the composite renal outcome in the KTPH-DKD cohort.

Sensitivity Analyses

We identified 125 and 60 incident ESKD events in 2 cohorts, respectively. We obtained results similar to the primary analysis when taking incident ESKD as the study endpoint (Table 2).

Table 2.

Association of 4 urine TCA cycle metabolites with the risk of incident end-stage kidney disease (ESKD)

	Discovery cohort (125 ESKD events)		Validation cohort (60 ESKD events)
	HR (95% CI)	P value	HR (95% CI)	P value
Lactate (per SD)
Unadjusted	1.26 (1.08-1.47)	.003	1.64 (1.36-1.96)	<.001
Multivariable	1.17 (0.97-1.40)	.095	1.33 (1.02-1.72)	.033
Citrate (per SD)
Unadjusted	0.60 (0.55-0.67)	<.001	0.61 (0.52-0.72)	<.001
Multivariable	0.81 (0.67-0.97)	.025	0.90 (0.71-1.13)	.36
Fumarate (per SD)
Unadjusted	1.39 (1.17-1.66)	<.001	1.92 (1.46-2.54)	<.001
Multivariable	1.35 (1.06-1.72)	.014	1.69 (1.15-2.49)	.008
Malate (per SD)
Unadjusted	1.27 (1.09-1.47)	.002	1.67 (1.34-2.09)	<.001
Multivariable	1.25 (1.04-1.49)	.016	1.40 (1.02-1.92)	.036

Cox proportional hazard regression: time to ESKD as outcome. Multivariable model adjusted age, sex, ethnicity (Chinese as reference), active smoking (yes or no), cardiovascular disease history (yes or no), renin-angiotensin system blocker usage (yes or no), body mass index, diabetes duration, HbA1c, mean artery pressure, estimated glomerular filtration rate, and urine albumin-to-creatinine ratio.

We identified 149 and 92 nonrenal deaths in 2 cohorts, respectively. All 4 urine metabolites (lactate, citrate, fumarate, and malate) were significantly associated with the composite of renal events and nonrenal deaths after adjustment of clinical risk factors (Supplementary Table S6) (17).

In the subgroup analysis, CKD status modified the association between urine citrate and the composite renal outcome in both cohorts (P < .05 for interaction term, Supplementary Table S7) (17). It appeared that the association between urine citrate and the renal outcome was more pronounced in participants with preserved eGFR (eGFR > 60 mL/min/1.73 m², Supplementary Table S8) (17). Sex, ethnicity, and albuminuria category did not modify the associations between urine metabolites and the composite renal outcome (all P values > .05 for interaction terms) (17).

Urine Fumarate and Citrate Were Independently Associated With the Composite Renal Outcome After Additional Adjustments for Other Urine Metabolites

Given the high correlation among metabolites in the TCA cycle pathway, we further studied which of the 4 metabolite(s) was associated with the composite renal outcome after additional mutual adjustment for other metabolites. Cox regression model with backward selection suggested that urine citrate and fumarate were retained as the determinants for the composite renal outcome independent of clinical risk factors (Table 3). The associations between urine lactate, malate, and the renal outcome were markedly attenuated after adjustment for clinical risk factors and additional mutual adjustment for the other urine metabolites (Supplementary Table S9) (17).

Table 3.

Urine metabolites and clinical risk factors associated with the composite renal outcome in Cox regression models with backward selection^a

	Discovery cohort		Validation cohort
	HR (95% CI)	P value	HR (95% CI)	P value
Fumarate (per SD)	1.51 (1.22-1.86)	<.001	1.64 (1.29-2.09)	<.001
Citrate (per SD)	0.68 (0.57-0.81)	<.001	0.86 (0.73-1.01)	.058
Age (y)	0.98 (0.96-0.99)	.006	0.97 (0.95-0.99)	.03
Male sex	1.34 (0.97-1.86)	.07	b
Malay ethnicity (yes/no)	1.63 (1.20-2.22)	.002	b
RAS blocker usage (yes/no)	b		1.84 (1.14-2.96)	.01
BMI (kg/m2)	b		0.96 (0.93-1.00)	.033
HbA1c (%)	1.12 (1.01-1.24)	.034	1.10 (1.01-1.20)	.036
Mean BP (mm Hg)	1.02 (1.00-1.03)	.018	b
eGFR (mL/min/1.73 m2)	.98 (0.97-0.99)	<.001	.98 (0.97-0.99)	<.001
Ln ACR	1.57 (1.45-1.70)	<.001	1.99 (1.74-2.28)	<.001

Cox proportional hazard regression with backward selection: time to the composite of renal events as outcome.

Abbreviations: ACR, albumin-to-creatinine ratio; BMI, body mass index; BP, blood pressure; RAS, renin-angiotensin system.

^aUrine lactate, citrate, fumarate, malate, age, sex, ethnicity, active smoking (yes or no), CVD history (yes or no), RAS blocker usage (yes or no), BMI, diabetes duration, HbA1c, mean artery pressure, estimated glomerular filtration rate, and natural log-transformed urine ACR (Ln ACR) were included in the model. Variable with P value < .1 was excluded in each iteration.

^bVariable that was not selected as an independent predictor for the composite renal outcome in each cohort.

Association of Urine and Plasma Citrate, Fumarate With CKD Progression

We included 147 participants with CKD progression (eGFR decline −7.1 [IQR −5.0 to −9.8] mL/min/1.73 m² per year, follow-up duration 5.0 [3.7-6.0] years) and 271 nonprogressors (eGFR change 0.18 [−0.74− 1.00] mL/min/1.73 m² per year, follow-up duration 6.0 [IQR 5.3-7.0] years) in the substudy to examine the relationships between urine and plasma citrate, urine and plasma fumarate, as well as their associations with CKD progression. Plasma citrate was modestly correlated with urine citrate (rho = −0.007) and moderately correlated eGFR (rho = −0.21). Only urine citrate was significantly associated with CKD progression after adjustment for clinical cardiorenal risk factors and mutual adjustment for each other (Table 4). Similarly, plasma fumarate was moderately correlated with urine fumarate (rho = 0.24) and eGFR (rho = −0.24, Supplementary Table S10) (17). Urine fumarate, but not plasma fumarate, was significantly associated with the risk of CKD progression in the multivariable model (Table 4).

Table 4.

Association of urine and plasma citrate and fumarate with rapid kidney function decline in logistic regression models (N = 418)

	Unadjusted model		Multivariable model
Per SD increment	OR (95% CI)	P value	OR (95% CI)	P value
Plasma citratea	0.76 (0.62-0.93)	.007	0.83 (0.64-1.08)	.16
Urine citratea	0.73 (0.60-0.90)	.003	0.69 (0.53-0.91)	.009
Plasma fumarateb	1.20 (0.97-1.48)	.09	1.16 (0.90-1.50)	.26
Urine fumarateb	1.52 (1.22-1.90)	<.001	1.32 (1.00-1.74)	.05

Binary logistic regression: rapid kidney function decline (eGFR decline 3 mL/min/1.73 m² per year or greater) was outcome. Multivariable model adjusted age, sex, ethnicity (Chinese as reference), active smoking (yes or no), renin-angiotensin system blocker usage (yes or no), body mass index, diabetes duration, HbA1c, mean artery pressure, estimated glomerular filtration rate, and log-transformed urine albumin-to-creatinine ratio.

Abbreviation: OR, odds ratio.

^aPlasma and urine citrate were included the same model.

^bPlasma and urine fumarate were included in the same model.

Discussion

In this prospective study in 2 cohorts of patients with type 2 diabetes, we found that a high level of urine fumarate predicted an increased risk of progression to ESKD, whereas a high level of urine citrate was associated with a low risk of CKD progression independent of clinical risk factors and other key TCA cycle metabolites. These data provide robust evidence supporting that dysregulation of metabolites in TCA cycle pathway may be an integral component of the pathophysiologic network underpinning progressive loss of kidney function in patients with type 2 diabetes.

Several clinical studies have assessed the relationship between urine TCA cycle metabolites and kidney disease progression in diabetic and nondiabetic populations (10-13, 20). In our early study, we reported that both fumarate and malate were significantly associated with rapid kidney function decline in patients with type 2 diabetes (12); however, that was a case-control study with limited generalizability. It had a relatively small sample size, short follow-up duration, and took eGFR decline as a surrogate endpoint. In the current prospective cohort study with the “hard” renal outcome as endpoint, we confirmed our findings early in the study by showing that both fumarate and malate predicted the risk for progression to ESKD in patients with type 2 diabetes (Figs. 1 and 2). Of note, these data were also consistent with findings in nondiabetic population in recent years. For example, Jo et al found that a high level of urinary fumarate predicted the composite renal outcome in patients with membranous nephropathy (21). Kim et al recently also reported that urine fumarate predicted kidney disease progression in patients with glomerular disease (13).

Fibrosis and inflammation are common pathways driving kidney disease progression (22). Fumarate promotes fibrosis by at least 2 mechanisms: (1) it inhibits ten eleven translocation demethylase and induces epithelial-mesenchymal transition, an important driver of kidney fibrosis (23, 24) and (2) it inhibits prolyl hydroxylases, leading to stabilization of hypoxia inducible factor (HIF)-1 and activation of the TGF-β pathway, the master regulator of fibrosis (25, 26). On the other hand, fumarate also plays a role in the regulation of inflammation. It activates adaptive immunity by inducing an epigenetic program (27). Moreover, accumulation of fumarate might cause leakage of mitochondrial DNA into cytosol and activate innate inflammation in the kidney (9). It is therefore biologically plausible that fumarate may be causally involved in the progressive loss of kidney function.

We reported that urine lactate was not independently associated with a fast eGFR decline in our early work (12). In the current study, we found that a high level of urine lactate was strongly associated with the composite renal outcome in both discovery and validation cohorts (Figs. 1 and 2). The discrepancy may be attributable to the vast differences in design (case-control vs longitudinal), study endpoints (surrogate vs clinical outcome), sample size, and follow-up duration. A high level of urine lactate may suggest the presence of an increased anaerobic glycolysis in patients with a high risk of CKD progression, which was partly supported by the finding that metabolic flux into glycolysis was significantly increased in db/db kidney and in patients with type 2 diabetes (10). The diabetic kidney is characterized by hypoxia (28). The low oxygen milieu activates HIFs, which bind to hypoxia-response elements of target genes including lactate dehydrogenase-A and pyruvate dehydrogenase kinase, leading to suppression of pyruvate dehydrogenase complex activity with a concomitant increase in aerobic glycolysis (29). Indeed, the pyruvate dehydrogenase complex, the gateway for pyruvate to enter the TCA cycle, was found to be hyperphosphorylated and inactivated in diabetic kidneys (30).

Of note, urine lactate, fumarate, and malate were highly correlated with one another (Supplementary Table S3) (17). Hence, it is reasonable to postulate that they are biologically interrelated. As an extension of the early studies, we attempted to explore which TCA metabolite(s) might be upstream of the pathophysiologic pathway underpinning DKD progression. Our analytical outcome suggests that fumarate may be at the upstream of lactate and malate because the strength of associations between the latter 2 metabolites and the renal outcome was markedly attenuated after adjustment for fumarate (Table 3 and Supplementary Table S9) (17). Fumarate accumulation may inhibit prolyl hydroxylases and activate HIF/hypoxia signalling pathway, which lead to increased glycolysis and lactate production (25, 31). This hypothesis is partly supported by (1) fumarase-mutant tumors with cytosolic fumarate accumulation exhibit enhanced glycolytic metabolism (32) and (2) the association of lactate with renal outcome was markedly attenuated after mutual adjustment for fumarate (Supplementary Table S9) (17). Similarly, only fumarate was significantly associated with the renal outcome, whereas the association between malate and renal events was diminished after mutual adjusting for other metabolites. Future studies are warranted to examine whether fumarate may be the upstream regulator of lactate and malate production in diabetic kidney.

The relationship between urine citrate and progressive CKD should be discussed. An early cross-sectional study found that patients with DKD had significantly lower level of urine citrate compared with patients with diabetes but without kidney disease (11). The association of urine citrate with the composite renal outcome could not be validated in the multivariable model (Fig. 2), which was similar to findings in our early case-control study (12). Further analysis showed that urine citrate interacted with CKD status in association with renal outcome. The inverse association between urine citrate and risk of ESKD was more pronounced in patients with preserved eGFR (Supplementary Tables S7 and S8) (17). The mechanistic linkage between a high level of citrate and a low risk of CKD progression may be multifactorial. (1) Citrate synthase is 1 of the rate-limiting enzymes in the TCA cycle (6). Hence, a lower urine citrate may indicate impaired oxidative phosphorylation in TCA cycle because of mitochondrial dysfunction. (2) Urinary citrate normalized to creatinine is a good marker of acid-base status, reflecting acid retention even in patients without overt metabolic acidosis (33, 34). Therefore, a low level of urine citrate may suggest an increased tubular uptake and metabolism of organic anions to normalize acid-base balance in individuals with a high risk of CKD progression (35). (3) Citrate may also play a role in immunomodulation. In adenine-induced CKD in rats, administration of citrate reduced the production of pro-inflammatory cytokines IL-6 and IL-17, whereas it increased the anti-inflammatory cytokines IL-10 (36). It is plausible that a high level of citrate may be associated with low risk of progressive CKD partly by reducing the inflammation tone.

Data from our current study have biological and clinical implications. Given that TCA cycle metabolites are involved in several aspects of kidney pathophysiology, measuring these metabolites in urine could gain insights into mechanisms of kidney diseases (37). Interestingly, a preclinical study showed that disturbance of TCA cycle metabolites might be alleviated by HIF-PH inhibitor enarodustat (38). SGLT2 inhibitors may also reverse the accumulation of TCA cycle metabolites in renal tissues (39), suggesting that the renal protective effects of enarodustat and SGLT2 inhibitors may be partly mediated by restoration of TCA cycle pathway. Whether reducing the accumulation of TCA cycle metabolites such as fumarate may serve as a potential strategy to target the dysregulated renal energy metabolism and slow down CKD progression deserves further study.

The current study has several strengths. We adopted the discovery and validation design in 2 cohorts of patients with type 2 diabetes. The high risk of DKD in Asian patients with diabetes, the relatively large sample size, and the long follow-up allowed us to choose the “hard” clinical outcome as study endpoint. We performed several sensitivity analyses, which might enhance the robustness of our analytical outcomes. Nevertheless, several important weaknesses must be highlighted. First, this is an observational study. Residual confounding is inevitable, although we have considered the main cardiorenal risk factors in data analysis. We cannot infer causality for the same reason. Conceivably, Mendelian randomization approach may be applied to elucidate the potential causal relationship between urine metabolites and risk of DKD progression in future studies. However, the genetic instruments (ie, variants robustly associated with urine TCA cycle metabolites) are still unavailable in Asian populations for now. Second, the TCA cycle pathway is complex and involves many intermediate metabolites. We only measured the key metabolites in this pathway. In addition, we measured the urine metabolites only once at baseline. Repeated measurements will further improve the robustness of analytical outcomes. Third, our participants were Southeast Asian. Further studies are needed to evaluate whether our findings are generalizable to other ethnic groups.

In conclusion, urine lactate, fumarate, malate, and citrate were associated with the risk of progression to ESKD independent of known clinical risk factors. Of them, citrate and fumarate may be the 2 key TCA cycle metabolites at the nexus of the pathophysiological network underpinning the progressive loss of kidney function in patients with type 2 diabetes.

Acknowledgments

We sincerely thank participants in the SMART2D and KTPH-DKD cohorts for their generous contributions to the study. We thank staff from clinical research unit in Singapore Khoo Teck Puat hospital for participant recruitment, biosample collection, and data curation.

Funding

This work was funded by Alexandra Health Systems Grants (STAR20201 and STAR23201) and National Medical Research Council Grants (MOH-000066, MOH-000714-01, and MOH-001327-02). The funders have no role in the study design, data analysis, manuscript writing, or decision for publication.

Disclosures

K.S. received consulting fee from Otsuka, honoraria from Pfizer, served on the CARA data monitoring board, received support from the American Society of Transplantation for attending meetings, and is the scientific founder of SygnaMap. The other authors declare that they have no conflicts of interests to disclose.

Data Availability

Data used in this study are not publicly available. However, anonymized data may be shared on reasonable request after approval from Singapore National Healthcare Group Ethics Review Committee.

Abbreviations

 
• ACR

  albumin-to-creatinine ratio

 
• CKD

  chronic kidney disease

 
• CVD

  cardiovascular disease

 
• DKD

  diabetic kidney disease

 
• eGFR

  estimated glomerular filtration rate

 
• ESKD

  end-stage kidney disease

 
• HIF

  hypoxia inducible factor

 
• IQR

  interquartile range

 
• KTPH-DKD

  Khoo Teck Puat Hospital-Diabetic Kidney Disease

 
• MS

  mass spectrometry

 
• QC

  quality control

 
• SMART2D

  Study of Macro-angiopathy and Micro-Vascular Reactivity in Type 2 Diabetes

 
• TCA

  tricarboxylic acid