Finding. Bulk kidney proteomes average a diabetic argument. Zheng, Fenton and colleagues cut the argument into pieces. Sixteen weeks after streptozotocin, they use deep-learning visual proteomics to laser out proximal tubule and glomeruli (Dice 0.878 and 0.914). The tubule drops proteostasis, cell-cycle, and structure and turns mitochondrial and lipid metabolism up. The glomerulus drops oxidative metabolic capacity and does not replace it with another program. Fourteen of their top twenty tubule proteins, LARS2 and ANXA2 among them, move the same way in human CKD tubulointerstitium from KPMP. The mouse PT proteome correlates with that human set. The glomerular one does not.
Why this paper matters
Diabetic kidney talk is glomerular. The mitochondrial compensation is tubular. If you treat “the kidney” as one OXPHOS organ, you will prescribe the wrong move to one of those compartments.
LARS2 is a mitochondrial translation factor showing up on a human-conserved list. That is a concrete follow-up, not a pathway word.
How to read the score
Eighty. Compartment-resolved mitochondria, human overlap, a usable method. Confidence is high for the mouse split, medium for STZ-as-human-DKD.
Caveats
Toxin diabetes, males only. Protein is not flux. Do not start a LARS2 drug from this brief.
What to do with it
Split DKD mitochondrial claims by nephron segment. Carry the 14-gene/protein PT overlap. Pull LARS2 before you write a biomarker pitch.