Host, diet, microbiome, and xenobiotic plasma metabolites associate with kidney cell-type transcriptional programs in CKD

Background

Chronic kidney disease (CKD) is associated with widespread alterations in the circulating metabolome, including metabolites influenced by diet and the gut microbiome that may represent modifiable disease pathways. However, the extent to which metabolites from distinct origins are associated with kidney cell-type-specific injury programs remains incompletely characterized.

Methods

We analyzed plasma metabolomics data from 240 Kidney Precision Medicine Project (KPMP) participants (207 CKD, 33 healthy reference) and validated findings in an independent cohort (n=883). We applied unsupervised clustering to identify metabotypes and elastic net classification to generate a continuous metabotype score. We classified metabolites by origin (host-derived, diet-derived, microbiome-dependent, xenobiotic) and trained random forest models to predict filtration markers and structural injury. In 66 participants with matched kidney single-cell RNA sequencing, we regressed podocyte and proximal tubule pseudobulk expression against plasma metabolites and computed hub scores quantifying transcriptional connectivity.

Results

Two reproducible metabotypes separated CKD from healthy states (elastic net CV AUC=0.984 discovery, 0.997 validation), with the metabotype score strongly correlating with eGFR (ρ=−0.79). Although they represented a minor subset of the circulating metabolome, diet-and microbiome-dependent metabolites such as myo-inositol, guaiacol sulfate, and indole derivatives were among the strongest metabotype-defining features. Metabolites improved prediction of filtration markers beyond clinical covariates, most notably for BUN (ΔR²=+0.29), driven by urea, guanidinosuccinate, and the microbiome-derived arabitol/xylitol. Plasma metabolites were associated with interstitial fibrosis (415 metabolites) and tubular atrophy (505 metabolites) and distinguished opposing proximal tubule transcriptional programs. Across kidney cell types, environmental xenobiotics (PFHxS, halogenated benzoic acids) activated aryl hydrocarbon receptor, inflammatory, and fibrotic programs in podocytes, while endogenous metabolites engaged inflammatory and oxidative stress programs in proximal tubule.

Conclusion

Plasma metabotypes capture metabolic heterogeneity beyond eGFR, with metabolites of distinct origins associated with cell-type-specific kidney injury programs, supporting plasma metabolomics as a noninvasive approach for molecular patient stratification in CKD.

Key Points

• Unsupervised clustering defined reproducible plasma metabotypes spanning host-, diet-, microbiome-, and xenobiotic-derived metabolites that separated CKD from healthy reference states.

• Diet- and microbiome-dependent metabolites were among the strongest metabotype-defining features and showed the strongest associations with podocyte and proximal tubule gene expression.

• Metabolites of distinct origins associated with different cell-type transcriptional programs in podocytes and proximal tubule.