Urinary amino acids patterns enlighten new predictive models for extensive renal proximal tubular injuries and kidney function

The kidney plays a crucial role in amino acids (AAs) retention and excretion through AA transporters located in the proximal tubules. The impairment of these transporters results in aminoaciduria, one of the features of extensive proximal tubular (PT) dysfunction, namely renal Fanconi syndrome (FS). This study aimed to explore the clinical application of urinary AAs excretion profile and investigate their associations with renal function and other PT dysfunction.

We enrolled 157 patients with PT dysfunction and detected urinary AAs levels by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Machine learning models were used to assess diagnostic values of urinary AAs in renal FS and renal function decline, with randomized search to optimize parameters and evaluate algorithm performance.

The patients enrolled had a median age of 47 (35-60) years and a median eGFR of 92.8 (50.0-115.7) mL/min/1.73 m², with primary Sjӧgren's syndrome as the most common identified primary cause. We mapped etiology-specific urinary AAs excretion profiles. The most susceptible urinary AAs were citrulline (Cit) and threonine (Thr) transported by SLC6A19 (B⁰AT1), while arginine (Arg), aspartate (Asp) and histidine (His) were the least affected. Diagnostic models of renal FS achieved area under curve (AUC) values of 0.81 (XGBoost) and 0.82 (RF) based on all 21 urinary AAs, outperforming traditional clinical indicators. An optimization model based on four urinary AAs (Cit, phenylalanine [Phe], asparagine [Asn] and Asp) showed similar diagnostic performance (AUC 0.88). The urinary proline (Pro) level was independently correlated with eGFR < 60 mL/min/1.73 m² (OR = 2.44, 95% CI: 1.53-3.91, P < 0.001). We found significant associations between abnormal urinary AAs excretion and hypouricemia or hypokalemia, particularly in the male patients.

We mapped the etiology-specific urinary AAs excretion atlases in patients with PT injuries, and established a fine and simplified machine-learning model based on only four urinary AAs to diagnose extensive PT dysfunction noninvasively and conveniently. Significant associations of aminoaciduria with the other PT dysfunction showed gender differences.