Spatial Multi-Omics Maps Kidney Regeneration at Single-Cell Resolution
A landmark multi-omics atlas of human kidney tissue reveals how proximal tubule cells shift between repair and disease states after injury.
Summary
Researchers from Indiana University and the KPMP consortium combined spatial protein imaging (CODEX) with single-nucleus RNA sequencing to map the regenerative capacity of kidney proximal tubule cells in both healthy and diseased human tissue. By analyzing over 1.7 million cells across 58 kidney biopsy samples, they identified distinct cellular states — healthy, injured, adaptive, and failed-repair — and tracked how disease severity shifts cell populations. They also discovered that spatial neighborhoods around tubules predict clinical outcomes better than single-cell data alone, and validated key protein biomarkers of regenerative potential including VCAM1, CD10, and vimentin, offering a new framework for understanding why some kidneys recover from injury while others progress to chronic disease.
Detailed Summary
Chronic kidney disease (CKD) affects over 800 million people globally, yet the molecular mechanisms determining whether injured kidneys recover or progress to fibrosis remain poorly understood. The proximal tubule (PT), the kidney's primary filtering workhorse, is the most vulnerable segment to ischemic or toxic injury. A central unresolved question is why some PT cells regenerate after acute kidney injury (AKI) while others enter a maladaptive state that drives CKD progression. This study addresses that question with unprecedented spatial and molecular resolution using human kidney tissue.
The team integrated two cutting-edge technologies: CO-Detection by indEXing (CODEX), a highly multiplexed spatial protein imaging platform measuring 47 proteins simultaneously at subcellular resolution, and single-nucleus RNA sequencing (snRNA-seq) from the Kidney Precision Medicine Project (KPMP). They analyzed 58 human kidney biopsy samples spanning reference (healthy), AKI, and CKD cohorts, encompassing over 1.7 million segmented cells across CODEX datasets and over 100,000 single nuclei from matched snRNA-seq data. Samples were obtained from living donors, nephrectomies, and clinical biopsies, providing a clinically representative spectrum of disease severity.
Using unsupervised clustering of CODEX protein expression, the researchers identified seven distinct proximal tubule cell states. These ranged from PT-Healthy (expressing high LRP2/megalin, SLC3A1, AQP1) through PT-Injured (elevated VCAM1, vimentin, CD44), PT-Adaptive (intermediate phenotype with co-expression of injury and repair markers), to PT-Failed Repair (low canonical PT markers, high CD44 and VCAM1, loss of LRP2). The proportion of PT-Failed Repair cells increased significantly with CKD stage, from <5% in reference tissue to >25% in advanced CKD (eGFR <30), while PT-Healthy cells correspondingly declined. These shifts correlated strongly with eGFR (r = −0.71, p < 0.001 for failed-repair proportion vs. eGFR).
Crucially, the study leveraged the spatial context of CODEX to define cellular neighborhoods — microenvironments of 10–30 cells surrounding each PT cell. Neighborhood analysis revealed that failed-repair PT cells were consistently co-localized with activated myofibroblasts (α-SMA+), inflammatory macrophages (CD68+CD163−), and endothelial cells with reduced PECAM1 expression, forming a pro-fibrotic niche. Logistic regression models incorporating neighborhood features predicted CKD progression (doubling of serum creatinine or dialysis initiation within 2 years) with an AUC of 0.84, significantly outperforming models using single-cell protein expression alone (AUC 0.71). Integration with snRNA-seq using a label-transfer approach confirmed that CODEX-defined PT states mapped onto transcriptionally distinct populations, with PT-Failed Repair cells enriched for HAVCR1 (KIM-1), TGFB1, and senescence-associated gene signatures.
The findings carry significant implications for both biomarker development and therapeutic targeting. The protein VCAM1 emerged as a robust spatial marker of maladaptive PT cells detectable in tissue and potentially in urine, supporting its value as a non-invasive CKD progression biomarker. The identification of the pro-fibrotic cellular neighborhood as a discrete, spatially organized unit suggests that therapies disrupting PT-myofibroblast or PT-macrophage crosstalk might halt the transition from AKI to CKD. Limitations include the cross-sectional design, which limits causal inference about injury-to-repair trajectories, and the relatively small sample sizes within individual disease strata. The reliance on biopsy tissue also introduces selection bias toward more severely ill patients.
Key Findings
- Failed-repair proximal tubule (PT-Failed Repair) cells increased from <5% in healthy reference tissue to >25% in advanced CKD (eGFR <30), with proportion inversely correlated with eGFR (r = −0.71, p < 0.001)
- CODEX spatial neighborhood features predicted 2-year CKD progression with AUC of 0.84, significantly outperforming single-cell protein expression alone (AUC 0.71)
- Seven distinct PT cell states were identified across 1.7 million+ segmented cells from 58 biopsy samples spanning healthy, AKI, and CKD cohorts
- Failed-repair PT cells were spatially co-localized with α-SMA+ myofibroblasts, CD68+CD163− inflammatory macrophages, and PECAM1-low endothelial cells, defining a pro-fibrotic niche
- snRNA-seq integration via label transfer confirmed PT-Failed Repair cells are transcriptionally enriched for HAVCR1 (KIM-1), TGFB1, and senescence-associated gene signatures
- VCAM1 protein expression was identified as a robust spatial and potentially urinary biomarker of maladaptive PT state across disease stages
- 47-plex simultaneous protein imaging (CODEX) combined with >100,000 single-nucleus transcriptomes enabled direct protein-to-transcript validation in matched tissue
Methodology
Cross-sectional multi-omics study analyzing 58 human kidney biopsy samples from the Kidney Precision Medicine Project (KPMP), encompassing reference (healthy living donors), AKI, and CKD cohorts. CODEX (CO-Detection by indEXing) spatial proteomics measured 47 proteins at subcellular resolution across >1.7 million segmented cells; matched single-nucleus RNA-seq provided transcriptomic validation for >100,000 nuclei. Cell-state assignments used unsupervised clustering with downstream spatial neighborhood analysis; clinical outcome prediction used logistic regression with AUC comparison. Label transfer between modalities was performed computationally to link protein states to transcriptional identities.
Study Limitations
The cross-sectional design prevents causal inference about the temporal sequence from healthy to failed-repair PT states, and longitudinal validation cohorts are needed to confirm prognostic models. Biopsy-based sampling introduces selection bias toward more symptomatic or severe disease, potentially underrepresenting early-stage transitions. The authors note that while protein-transcript integration was strong for major cell types, rare transitional states may be underrepresented due to sampling depth constraints in individual cohorts.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.