In this study, we employed a high-throughput NMR platform to analyse 51 metabolite-to-creatinine ratios from 24-h urine samples of 2670 individuals with type 1 diabetes and found that ten ratios were associated with progression of diabetic nephropathy after adjusting for baseline albuminuria category and CKD stage. Importantly, there were differences between the urinary profile for overall progression and progression from normo- and macroalbuminuria suggesting 2-hydroxyisobutyrate as a potential marker for progression from normoalbuminuria (i.e. new-onset diabetic nephropathy) and branched-chain amino acids as markers for overall progression and progression from macroalbuminuria to ESKD.
Several metabolites point towards a link between insulin resistance and progression of diabetic nephropathy. Urinary amino acid and in particular branched-chain amino acids were associated with overall progression as well as progression from macroalbuminuria to ESKD. These novel findings are noteworthy in the context of the many studies on plasma concentrations of branched-chain amino acids in the field of type 2 diabetes. The first study to show associations between plasma levels of branched-chain amino acids with obesity and insulin resistance dates back 50 years [17]. These amino acids also represent the most consistent biomarkers identified by blood metabolomics to be predictive of future type 2 diabetes risk [18]. In physiological studies and human genetics, there is even evidence that links higher levels of insulin resistance causally to higher concentrations of circulating branched-chain amino acids [19]. Furthermore, insulin resistance has been observed in individuals with type 1 diabetes and linked to the increasing prevalence of obesity in type 1 diabetes [20], and also in our cohort, progressors were more likely to be obese at baseline. Importantly, it has been shown in type 1 diabetes that insulin resistance precedes microalbuminuria [21]. A previous, small study of 200 individuals from FinnDiane [22] found that serum metabolites that have been associated with new-onset albuminuria have also been associated with insulin resistance. Urine concentrations of branch-chain amino acids may depend on protein intake. We do not have data on diet, but secretion for most amino acids is not strongly influenced by protein intake in healthy individuals [23]. However, high protein intake in prevalent CKD may further kidney function decline [24]. Reduced tubular reabsorption may play an important role. A recent review [25] reported that the expression of amino acid transporter Slc6a19, which has a higher affinity to neutral amino acids than branched-chain amino acids, was increased in animal models of diabetes. In line with reduced reabsorption, a study on type 2 diabetes found that individuals who progressed to ESKD had lower serum concentrations of leucine and valine [26].
Urinary 2-hydroxyisobutyrate is another metabolite that links progression of diabetic nephropathy with insulin resistance. It derives from the degeneration of proteins by gut microbiota [27]. It was shown that urinary 2-hydroxyisobutyrate was higher in obese, insulin-resistant men compared with age-matched, lean control participants [27]. In our cohort, 2-hydroxyisobutyrate was associated with overall progression, the progression from normoalbuminuria and nominally associated with the progression from macroalbuminuria to ESKD. The fact that 2-hydroxyisobutyrate is associated with both progressions is important as progression from normoalbuminuria is related to changes in albumin excretion, and progression from macroalbuminuria is more related to a drop in filtration capacity. A study found no correlation between eGFR and serum 2-hydroxyisobutyrate in non-diabetic individuals with CKD stages 3 and 4 [28] and reported higher serum concentrations of 2-hydroxyisobutyrate compared with healthy control participants, suggesting that renal clearance of the metabolite is mainly dependent on active tubular transport. Moreover, clinically relevant concentrations of 2-hydroxyisobutyrate given to a human renal proximal tubule cell line resulted in higher expression of mesenchymal markers and loss of epithelial features without affecting its mitochondrial activity [28]. Mitochondrial homeostasis is very important for the functioning of reabsorption [29]. Therefore, fitting with our observation of 2-hydroxyisobutyrate as an early marker of progression, there is some evidence that the higher concentrations in urine in individuals with normoalbuminuria might not result from reduced tubular reabsorption but from higher serum concentrations. In contrast to 2-hydroxyisobutyrate, the branched-chain amino acids did not show any association with progression from normoalbuminuria but a strong association with progression from macroalbuminuria, pointing more towards reduced tubular reabsorption.
Urinary pseudouridine has been discussed as a potential glomerular filtration marker, and in our study, it was associated with overall progression. Previous studies in serum have shown elevated concentrations of pseudouridine in renal failure and uraemia, and the concentrations were associated with CKD [30]. In type 1 diabetes, serum pseudouridine has been shown to be associated with eGFR decline in individuals with proteinuria in CKD stage 3 [31]. Pseudouridine-to-creatinine ratios measured in spot urine were slightly decreased in individuals with CKD, but it is unclear whether these cross-sectional differences were statistically significant [32]. Previously, pseudouridine was dismissed as a glomerular filtration marker due to tubular reabsorption. However, newer findings have shown that pseudouridine and eGFR are correlated, and while the calculation of eGFR has to take sex into account, pseudouridine excretion was shown to be less dependent on sex [30]. Therefore, our results suggest that pseudouridine is a glomerular filtration marker.
In addition to filtration, several urinary metabolites are linked to renal tubular damage or protection. Low urinary citrate was associated with overall progression. This observation is in line with a study showing that administration of citrate salts reduces tubulointerstitial injury and slows eGFR decline [33]. Additionally, metabolic acidosis in individuals with CKD was associated with lower urinary citrate [34]. In general, our results on citrate are in accordance with a previous small cross-sectional study on CKD [34] associating urinary citrate with tubular protection.
Previous studies investigating tubular damage showed links with urinary glycine, alanine and pyroglutamate. In our cohort, we observed differing associations for progression for urinary glycine in individuals with normo- and macroalbuminuria. Lower glycine in urine was nominally associated with progression from normoalbuminuria and higher glycine with progression from macroalbuminuria. Urinary threonine followed the same directions of association, and there are several degradation processes for threonine, one leading to the synthesis of glycine [35]. Low urinary glycine has previously been associated with incident eGFR reduction below 60 ml min−1 [1.73 m]−2 in the case–control matched Framingham Offspring cohort of 386 individuals [36]. This setting is more comparable to our subanalysis in individuals with normoalbuminuria with a mean eGFR of 105 ml min−1 [1.73 m]−2 at baseline. As in the Framingham Offspring cohort, low urinary glycine was nominally associated with progression from normoalbuminuria. In animal models, dietary glycine has been shown to protect against cyclosporin-mediated proximal tubular damage [36]; such damage may occur early in kidney disease and may even precede glomerular changes [37]. We have previously shown that genetic variants in the GLRA3 gene encoding a glycine receptor are associated with albuminuria [38]. Interestingly, in our cohort, the association changed when looking at the progression from macroalbuminuria. In this scenario, higher and not lower urinary glycine was associated with progression. The reabsorption of glycine through the kidney plays an important role for the bioavailability of glycine [39] and is facilitated to a similar extent by energy-dependent transport and non-energy-dependent transport [40]. Lower urinary glycine in those individuals who progressed from normoalbuminuria might therefore be related to lower serum concentrations. On the other hand, the higher concentrations in progressors to ESKD point towards a reduced rate of energy-dependent reabsorption. Furthermore, increased concentrations of glycine and alanine were found in rat urine after inducing damage to the proximal tubule [41] reflecting that alanine and glycine energy-dependent reabsorption is mediated by neutral amino acid transporters such as SLC6A19 [40] and both were associated with incident ESKD in this study. Urinary pyroglutamate, which was associated with progression from macroalbuminuria to ESKD, has previously been linked to incident macroalbuminuria in individuals with type 2 diabetes and microalbuminuria [42]. A previous study in type 1 diabetes with 25 individuals linked pyroglutamate to the progression of early kidney disease from normoalbuminuria to microalbuminuria [43]. However, in our study, we did not find any link between pyroglutamate and such early progression. We suggest that, in a healthier kidney, pyroglutamate is efficiently reabsorbed [44], and with worsening kidney disease, the reabsorption capabilities diminish.
Finally, we also found urinary markers that have been previously linked to responses to oxidative stress and hypoxia, such as tyrosine. Oxidative stress is a hallmark of CKD [45], and the interplay between oxidative stress and hypoxia critically contributes to kidney injury [46]. In our study, higher urinary tyrosine was associated with progression to ESKD in individuals with macroalbuminuria. An inverse association was shown in a previous study in type 2 diabetes that only investigated the earlier stages of kidney disease in individuals who had either normo- or microalbuminuria at baseline [47]. These observations may be influenced by oxidative stress [45]. Molnár et al. suggested that both the healthy and the damaged kidney retain para-tyrosine that is converted from phenylalanine. In the presence of free radicals, phenylalanine and consequently tyrosine are hydroxylated in para, meta and ortho positions, and the authors observed that the ortho-tyrosine excretion was enhanced in type 2 diabetes through an increased tubular secretion and production [45]. Therefore, we would indeed expect increased concentrations of urinary tyrosine in the presence of increased oxidative stress. A previous study on urine metabolites in diabetes also found links to mitochondrial activity [7]. In further concordance with that study, we found the same direction of associations for 3-hydroxyisovalerate and glycolic acid in the overall and macroalbuminuria analysis.
There are some limitations of our study. The associations between urine metabolites and progression do not allow for any causal conclusions. Furthermore, the results have not as yet been replicated in an independent cohort as we did not have access to other large cohorts with type 1 diabetes, 24-h urine collections and long-term follow-up. Large-scale urinary NMR analysis, such as this study, will shortly be taken up by many cohorts. In the future, it will probably be possible to validate these results with other large cohorts, and therefore, these biomarkers will potentially go beyond providing disease aetiology information to disease prediction. These univariate analyses did not consider the correlation between metabolites (ESM Fig. 7), and future studies will need to investigate which signals are independent. In addition, we do not have dietary data to assess the influence of diet. We cannot exclude an effect of storage time; however, sensitivity analyses for creatinine measured by clinical chemistry at sampling time did not find a change in the direction of the associations. The study also has several strengths starting with the size of the cohort. In addition, the urine was measured over 24 h, which is the gold standard of urine collections. However, the applicability of the results to morning or spot urine collections remains to be addressed. The FinnDiane cohort is thoroughly characterised and has a long follow-up period.
In summary, this study found that ten out of 51 urinary metabolites were associated with progression of diabetic nephropathy even after adjusting for baseline albuminuria category and CKD stage. We found differences between overall progression and progression from normo- and macroalbuminuria such as 2-hydroxyisobutyrate as a potential marker for progression from normoalbuminuria. Amino acids and, in particular, the branched-chain amino acids were strongly associated with progression and especially with the progression from macroalbuminuria to ESKD. These results provide new potential urinary biomarkers that were associated with progression beyond the traditional markers of albuminuria and estimated kidney filtration rates. This study highlights the potential of routinely analysing urinary metabolites on a larger scale as urinary NMR metabolomic profiling is a reliable high-throughput method.