Abstract
Polyomaviruses establish long-term infection in the kidney and are intermittently shed in urine. However, the relationship between kidney-resident viral genomes and urinary shedding during persistent infection remains poorly defined. Using a genetically barcoded murine polyomavirus library, we tracked thousands of viral lineages in vivo by pairing longitudinal urine sampling with endpoint barcode sequencing of kidney tissue in four mice. Across all animals, kidney infection consistently resolved into two stable viral populations, with near-silent persistence as the dominant fate. Most kidney-resident barcodes were never detected in late urine at late stages of infection, even though many reached substantial abundance within the kidney, demonstrating that kidney viral genome levels alone do not predict urinary shedding. In contrast, only a small minority of kidney barcodes contributed disproportionately to urine virus output at late timepoints, and these barcodes exhibited stable longitudinal behavior, with repeated detection in urine over time and markedly higher peak urine abundance than late non-shed or random barcode controls. Shedding behavior was not explained by input virus stock abundance, barcode sequence features, predicted miRNA targeting, or ongoing reseeding from blood or other tissues. Instead, barcodes that ultimately dominated late urine already showed elevated urine detection early after infection, indicating that shedding fate is established early and maintained throughout persistent infection. Together, these findings reveal that persistent kidney infection is a structured reservoir composed of a large population of deeply restricted viral genomes and a smaller, stable subset that repeatedly produces urine-detectable viruses, with concurrent smoldering infections and latency-like restriction representing one possible model to explain the sharply different probabilities of shedding among kidney-resident genomes.