Two Serum Proteins Could Reveal Hidden Kidney Damage in Diabetes

For decades, doctors have relied on a single number to decide whether diabetes is quietly destroying a patient’s kidneys: the urinary albumin-to-creatinine ratio, or UACR. If albumin spills into the urine at 30 milligrams per gram or above, the diagnosis of diabetic kidney disease is made. Below that threshold, the kidneys are presumed safe. A new study from Mianyang Central Hospital in China, published in Frontiers in Endocrinology, argues that this presumption is wrong in a specific and clinically important way. By integrating serum proteomics and metabolomics with machine learning, the research team identified two proteins, glutathione S-transferase A1 (GSTA1) and aldehyde dehydrogenase 1 alpha 1 (ALDH1A1), as the strongest candidate biomarkers for low-level albuminuric diabetic kidney disease, a stage in which albuminuria sits between 10 and 29 milligrams per gram and the estimated glomerular filtration rate has already fallen below 60 milliliters per minute per 1.73 square meters.

The problem the study addresses is one of the most underappreciated diagnostic blind spots in modern endocrinology. Non-albuminuric diabetic kidney disease, in which patients show significant renal impairment without crossing the albuminuria threshold, has been reported to account for between 20 and 70 percent of all diabetic kidney disease cases across different regions, and a recent meta-analysis suggests nearly half of patients with diabetes-related renal impairment have kidney disease without albuminuria. Yet diagnosing these patients is extraordinarily difficult. Measured glomerular filtration rate, the gold standard, is not routinely available, and guideline-recommended estimated GFR equations can deviate from actual GFR by up to 25 percent, leading to both misdiagnosis and missed diagnosis. The absence of specific biomarkers has left clinicians with almost no tools to identify kidney damage before it announces itself in the urine.

What makes the new study particularly valuable is that it does not treat non-albuminuric diabetic kidney disease as a single entity. In 2009, a workshop sponsored by the National Kidney Foundation and the US Food and Drug Administration proposed subdividing the original normal albuminuria category into low albuminuria (UACR 10 to 29 milligrams per gram) and strict normal albuminuria (UACR below 10 milligrams per gram). Previous research rarely distinguished between these two subgroups, leaving open the question of whether they represent unique disease subtypes or different clinical stages of the same trajectory. The Chinese team designed a three-phase study to answer this question directly: a retrospective cross-sectional survey to establish prevalence, a prospective omics profiling phase to identify differential proteins, and a preliminary validation phase to test diagnostic performance.

The retrospective phase examined 10,179 patients managed by the Department of Endocrinology between January 2023 and December 2024, all of whom had at least three serial measurements of UACR and eGFR obtained at monthly or bimonthly intervals. The results were striking. Strictly normoalbuminuric diabetic kidney disease, defined as UACR below 10 milligrams per gram with eGFR at or below 60, accounted for only 0.7 percent of all diabetes patients and 1.6 percent of all diabetic kidney disease patients. In contrast, low-level albuminuric diabetic kidney disease was far more common, representing 6.7 percent of all diabetes patients and 14.0 percent of all diabetic kidney disease patients. Both subgroups were characterized by advanced age and significantly lower eGFR than the other groups, with median ages of 68 and 69 years respectively, compared with 54 years in the simple diabetes group.

These prevalence data suggest something important about disease biology. The overwhelming rarity of strictly normoalbuminuric diabetic kidney disease, combined with the advanced age profile of both non-albuminuric subgroups, points toward a possible disease trajectory in which diabetes progresses to low-level albuminuric diabetic kidney disease and then to overtly albuminuric disease. The older age profile, however, also supports the possibility of multiple parallel disease pathways, in which some patients develop renal impairment through routes that never involve significant albuminuria. During the three-month prospective omics sample collection from newly diagnosed patients, the team encountered only one strictly normoalbuminuric case and 28 low-level albuminuric cases, indirectly confirming the prevalence distribution they had observed retrospectively.

The prospective omics phase enrolled patients with at least five years of type 2 diabetes duration and newly diagnosed diabetic kidney disease between January and March 2025. After a three-month follow-up with monthly UACR and eGFR measurements, participants were classified into five groups. To address the methodological uncertainty inherent in estimated GFR, the team performed dynamic 99mTc-DTPA clearance on a Symbia Intevo 6 SPECT/CT system for cases with UACR between 5 and 15 milligrams per gram and eGFR between 45 and 75, using the Gates formula to calculate bilateral measured GFR. The final omics cohort included 34 simple diabetes patients, 28 low-level albuminuric patients, 21 high-albuminuric patients, and 17 very-high-albuminuric patients. The single strictly normoalbuminuric case was excluded from omics profiling for insufficient statistical power.

Data-independent acquisition proteomics on a Vanquish Neo ultra-high performance liquid chromatography system coupled to an Orbitrap Astral mass spectrometer identified 1,597 proteins from the UniProt database, of which 1,466 were quantifiable. Partial least squares discriminant analysis revealed clear separation between most groups, with partial overlap between the high and very-high albuminuric groups. Volcano plot analysis identified 61 differentially expressed proteins between low-level albuminuric disease and simple diabetes, and 56 between low-level albuminuric disease and the albuminuric groups, yielding 97 differential proteins after accounting for overlap. Global targeted metabolomics on an Orbitrap Exploris 120 system identified 1,926 secondary metabolites, of which 506 were annotated at Metabolomics Standards Initiative Level 1. After excluding five overlapping metabolites, 162 differential metabolites were finalized for integrated analysis.

The integrated proteomic-metabolomic analysis, drawing on metabolic pathway data from 11 databases including KEGG, Reactome, and the Human Metabolome Database, found that differential proteins and metabolites in the low-level albuminuric group were significantly enriched in 33 metabolic pathways. The most significantly perturbed metabolic maps included cysteine biosynthesis, retinol biosynthesis, phase II conjugation, and selenoamino acid metabolism. Gene Ontology enrichment revealed that the 29 identified differential proteins, 8 upregulated and 21 downregulated, are over-represented in extracellular exosomes, cytosol, nucleus, membrane, extracellular space, and cytoplasm. KEGG enrichment indicated their primary engagement in metabolic pathways, amino acid biosynthesis, ribosome function, the one carbon pool by folate, and carbon metabolism. After 16 iterative collinearity diagnostic procedures eliminated proteins with variance inflation factors above 10, backward elimination yielded a final panel of 10 candidate proteins with significant clinical relevance.

The preliminary validation phase randomly enrolled 810 patients between April and December 2025, divided into 270 simple diabetes, 270 low-level albuminuric, and 270 high or very-high albuminuric patients, split into training and testing sets at a 7:3 ratio. Serum concentrations of the candidate proteins were quantified using enzyme-linked immunosorbent assay. LASSO regression and best subset selection consistently identified eight feature variables, including age, VEGFD, ALDH1A1, GSTA1, AHCY, RPL37A, B3GALT6, and OMD. Nine common machine learning classifiers were tested, and only the XGBoost classifier achieved all performance metrics above 0.8 on the cross-validation set, with a mean AUC of 0.843 and testing set AUC of 0.982. Shapley additive explanations analysis confirmed that GSTA1 and ALDH1A1 consistently ranked first and second in feature importance across both training and testing sets, with SHAP values of 0.219 and 0.203 on the training set and 0.229 and 0.174 on the testing set respectively.

The biological significance of these two proteins is compelling. GSTA1 is primarily expressed in proximal tubular epithelial cells and hepatocytes, and is released into blood and urine upon kidney injury. It catalyzes the conjugation of glutathione with electrophilic compounds and plays a central role in antioxidant defense. Multiple studies have demonstrated its diagnostic value for early tubular injury, particularly in acute kidney injury and drug-induced kidney damage. ALDH1A1, a core member of the aldehyde dehydrogenase superfamily, mediates retinoic acid synthesis, aldehyde detoxification, and antioxidant stress response. Upregulation in proximal tubule cells promotes retinoic acid synthesis, activates the RAR/RXR pathway, and drives tubular epithelial cell regeneration. In diabetic patients, high-glucose-induced oxidative stress reduces ALDH1A1 activity, aggravating aldehyde accumulation and promoting advanced glycation end product formation. Together, the two proteins collectively implicate tubular injury as a potential trajectory in low-level albuminuric diabetic kidney disease, suggesting that the earliest detectable damage may occur in the renal tubules rather than the glomeruli, and offering a promising detection method for managing chronic kidney injury progression in diabetic patients before it becomes visible in the urine.

Subject of Research: Integrated omics analysis identifies GSTA1 and ALDH1A1 as candidate biomarkers for low-level albuminuric diabetic kidney disease: a mixed prospective-retrospective cross-sectional study

Article Title: Integrated omics analysis identifies GSTA1 and ALDH1A1 as candidate biomarkers for low-level albuminuric diabetic kidney disease: a mixed prospective-retrospective cross-sectional study

Article References: Chen, X., Su, W., Yang, T., Huang, J., Wu, B., Yang, Q., Jin, Y., Tang, J., Yang, P., & Yang, Y. (2026). Integrated omics analysis identifies GSTA1 and ALDH1A1 as candidate biomarkers for low-level albuminuric diabetic kidney disease: a mixed prospective-retrospective cross-sectional study. Frontiers in Endocrinology, 17, Article 1978947. https://doi.org/10.3389/fendo.2026.1978947

Image Credits: AI Generated

DOI: 10.3389/fendo.2026.1978947

Keywords: Integrated, omics, analysis, identifies, GSTA1, ALDH1A1, candidate, biomarkers, low-level, albuminuric, diabetic, kidney