A human kidney is built exactly once. Everyone is born with somewhere between 200,000 and 2 million nephrons, the microscopic filtering units that clean the blood, and the body never makes another one. Nearly everything known about how that construction works has been inferred from mice.
Researchers at the Penn-CHOP Kidney Innovation Center have now read instructions directly from human tissue, and one of their findings contradicts a long-standing assumption. Cells that had already started down the path toward becoming one kind of kidney cell were caught turning into a different one.
The study was published on July 30, 2026, in Nature Genetics.
Cells Reversed Decisions They Had Already Made
The prevailing model held that once a developing kidney cell commits to a fate, that is the end of it. In the spatial map, cells that had begun differentiating toward proximal tubule identity were observed converting into other cell types of the renal corpuscle, the structure that performs actual filtration.
That flexibility may solve an old puzzle. The chemical gradients thought to pattern the kidney are extremely subtle, yet somehow they produce a million correctly proportioned nephrons. The findings suggest the early gradients do not need to be precise. Patterning appears to happen in two passes, with early signals laying down a rough blueprint and later signals correcting individual cells to match their final position.
"We have long assumed that once a kidney cell commits to a fate, it stays there," said Katalin Susztak, MD, PhD, co-director of the center and co-lead of the study, in a Penn Medicine statement. "Our data show that development is more forgiving than that. The tissue lays down a rough plan and then corrects itself."
Reading the Signals a Cell Actually Sits In
The technical obstacle was that standard single-cell sequencing requires dissolving tissue into a slurry of separate cells, which destroys exactly the information development depends on: where each cell sits and which signals reach it.
Earlier single-cell work on human fetal kidney established that mouse and human kidney development differ in ways that matter, but those studies could catalog which cells were present without showing what was passing between them.
The team combined single-cell RNA sequencing with two spatial transcriptomics platforms and analyzed more than 700,000 cells from human fetal kidneys between 12.5 and 20.5 weeks of gestation. The tissue came from donors undergoing elective abortion at the University of Pennsylvania and at Sheba Medical Center in Israel, all with written informed consent and institutional review board approval.
The work came out of the Penn-CHOP Kidney Innovation Center, a collaboration between the Perelman School of Medicine and Children's Hospital of Philadelphia, working with bioengineers from Penn's School of Engineering and Applied Science. That structure mattered to the result: the genomics produced the map, and the engineering lab supplied the cultured kidneys and organoids needed to test whether anything on it was real.
Keeping the spatial context, let them run a genome-wide analysis of cell-to-cell communication and pinpoint the anatomical locations where cells commit to one path or another. They also introduced a different way of defining a cellular neighborhood: not by which cell types happen to sit nearby, but by the full mixture of soluble signals a cell is bathed in. Several of those signal-based neighborhoods cut straight across boundaries anatomists have drawn for a century.
One Growth Factor Kept the Stem-Like Cells Going
The communication analysis pointed to insulin-like growth factor 2 (IGF2) as strongly associated with the renewal of nephron progenitor cells, the stem-like population that generates most nephrons and is depleted by late gestation. Two other signals, GDNF and RSPO3, were also tracked with progenitor renewal.
A correlation in a map is not a mechanism, so the team tested it. In cultured mouse embryonic kidneys, adding IGF2 increased branching of the ureteric tree and enlarged the caps of progenitor cells at each branch tip. Blocking the receptor did the reverse, shrinking and disorganizing those caps, and at higher doses, wiping out the progenitor population. Human kidney organoids grown from stem cells behaved similarly.
Human population genetics points in the same direction. Variants near IGF2 and its receptor were recently associated with differences in adult kidney size.
What This Does Not Change for Patients
Nothing here is a treatment, and the nearest practical application is far from clinical. Laboratory-grown kidney tissue is a rapidly advancing field that still produces organoids that fall well short of real kidneys, and identifying which signals sustain progenitor cells is a concrete lever for improving them.
The stakes for getting kidney construction right are long-term rather than immediate. A low nephron count at birth carries a higher lifetime risk of high blood pressure and chronic kidney disease, and severe disruptions of kidney formation are the leading cause of chronic kidney disease in children.
The limits are worth stating plainly. The map is human, but the functional validation ran in cultured mouse embryonic kidneys and stem-cell-derived organoids, neither of which is a developing human. The link between the fetal signaling environment and kidney health decades later remains an inference, not a demonstrated causal chain. And the plasticity finding describes what cells were observed doing in fixed tissue, which is a snapshot rather than a movie.
The framework itself is not specific to kidneys. Any organ whose cells have to coordinate across space could be mapped the same way.
Key Questions Answered
What did the researchers actually build? A spatial map of the developing human kidney, assembled from more than 700,000 cells taken from fetal kidneys between 12.5 and 20.5 weeks of gestation.
Why is it surprising that cells change fate? Because the standard model treated commitment as final. Seeing cells already headed toward one identity become another suggests the kidney patterns itself in two passes and corrects its own errors.
What is IGF2 doing? It appears to sustain nephron progenitor cells, the stem-like population that gives rise to most nephron components. Adding it expanded those cells in laboratory models, and blocking its receptor depleted them.
Can this regrow kidneys in adults? No. There is no therapy here. The most immediate use is improving laboratory-grown kidney organoids, which remain research tools.
Why does the number of nephrons at birth matter? Because no new ones form afterward. A lower count is associated with a higher lifetime risk of high blood pressure and chronic kidney disease.
Where did the tissue come from? Human fetal kidneys donated by patients undergoing elective abortion at the University of Pennsylvania and Sheba Medical Center, with written informed consent and ethics board approval at both sites.
