Endometrial organoids rebuild a uterus, then become something else

What the work claims

This is a primary in vivo study in mice, built as a functional test of what transplanted endometrial epithelial organoids (EEOs) can actually do. The authors crossed epithelium-specific Ltf-iCre mice with the Rosa26-iDTR line, so diphtheria toxin (DT) selectively kills uterine epithelial cells. One DT administration removed about 94 percent of the luminal and glandular epithelium while sparing stroma. The denuded surface re-epithelialized within 10 days from retained cells, but these mice stayed infertile across a three-month breeding trial (n = 4 mice per group): resurfacing is not regeneration¹.

The rescue is striking. EEOs were grown from donor mice carrying a Cxcl15-iCre driver and a Rosa26 dual-reporter that switches from red to green fluorescence when the glandular Cxcl15 program switches on. At passage 3 the organoids were roughly 80 percent red (non-glandular, luminal-type) and 20 percent green (glandular). Transplanted into the uterine lumen 24 hours after ablation, they engrafted, and by day 30 donor cells had built both luminal and glandular epithelium. Mated 30 days after transplant, recipients showed implantation sites by ultrasound on gestational day 12 only in the transplanted horn, carried pregnancies to term and delivered live pups, with smaller litters consistent with pregnancy confined to one horn. Donor epithelium persisted even after pregnancy and postpartum regeneration¹.

The load-bearing experiment comes next: the authors ablated the green glandular cells out of donor organoids with DT in culture, leaving a purely red, luminal-type population, and transplanted that. In the dish these cells never produced glands. In the uterus, by day 30, glandular structures appeared that were green, meaning they had activated Cxcl15, and expressed the gland marker FOXA2. Luminal-derived organoids had converted to glandular identity in vivo, and the regenerated glands supported pregnancy¹.

How it works

The conversion is not a transplantation artifact but an endogenous repair program. In a second mouse model, Cxcl15-iCre;Rosa26-iDTR/nTnG, DT administered into the uterus ablated only the glandular epithelium in situ. Glands regenerated progressively: at day 3, red buds extending from the luminal surface into the stroma expressed the luminal marker CALB1 with variegated FOXA2; by day 6 transitional green cells appeared; by day 10 mature glands were green, FOXA2-positive and CALB1-negative. Gland-ablated mice recovered full fertility, with litter sizes statistically indistinguishable from controls (n = 5 mice per group, P = 0.7854), normal implantation-site numbers, correct embryo positioning and normal stromal decidualization marked by PTGS2. During gestation the regenerated glands re-expressed Cxcl15¹.

The interpretation the authors favor is that adult uterine epithelium harbors facultative progenitors: differentiated luminal cells retain a latent competence, established during early postnatal development when glands first bud from the luminal surface, to re-enter a gland-forming program when injury and stromal cues instruct it. The instructive signal itself is unidentified; the stroma is preserved in both ablation models and is the obvious candidate source¹.

Where a skeptic should push

Start with the strongest caveat, which the authors state themselves: this is established in the mouse, and relevance to human endometrial regeneration and to organoid therapeutic strategies is untested. Human endometrial organoids differ from mouse EEOs in clonogenic behavior and marker logic; nothing here guarantees the conversion program exists in humans¹.

Second, the ablation is about 94 percent efficient, not total. Re-epithelialization after full-surface ablation came from retained cells, and the authors cannot formally exclude rare mistargeted labeling with the Cxcl15 driver; the lineage conclusions rest on the reporter behaving exactly as designed. Third, the design is epithelial-centric by construction: stromal, immune and vascular contributions to regeneration are unexplored, and the niche signals driving the luminal-to-glandular switch are a black box. Fourth, the functional rescue is real but partial: litter sizes in the transplant experiment were reduced, and only the transplanted horn carried pregnancy, so engraftment coverage or quality limits function even in this cleanest-case model¹.

Finally, resist the seductive one-line summary. What failed after ablation was not structure but function: the lumen resurfaced yet pregnancy did not establish. That means whatever the retained cells rebuilt, it lacked gland-derived signals needed for implantation, and the organoid replaced those signals. The paper does not identify which glandular outputs matter, so the mechanism of rescue remains descriptive.

Plasticity upends organoid identity and purity specs

For the organoid field, the uncomfortable implication is that in-vitro identity is an incomplete specification of an organoid product. A luminal-type endometrial organoid is not committed to being luminal; place it in an ablated uterus and the host stroma rewrites its program within a month, activating a glandular lineage switch that culture never triggers. Every cell-therapy program that writes potency, purity and identity specifications from dish-based markers is making the same assumption this paper breaks: that the state you ship is the state that engrafts. The assay consequence is direct. Transplantation into an ablation model with a lineage reporter, scored by function (implantation, gestation, live birth) rather than histology, is the only kind of potency assay that would have caught the conversion. Dish assays alone would have certified the red organoids as purely luminal and missed the biology entirely¹.

The opportunity mirrors the threat. This ablation-plus-transplant scaffold is a reusable functional testbed: any epithelial organoid program, intestinal, airway, endometrial, could in principle be challenged to rebuild a native compartment and prove function in vivo, with lineage switches read out by an inducible reporter rather than inferred from markers. For endometrial disease specifically, where Asherman-type scarring destroys the epithelial compartment and no regenerative option exists, a human-validated version of this platform would be genuinely new therapeutic territory. But the same plasticity that makes engraftment powerful makes it unpredictable: transdifferentiation in a patient is not automatically benign, and fate after transplant must be measured in vivo over time, not assumed from the certificate of analysis. One mouse model, one injury paradigm, one fluorescent readout: generalization to other organs and to human tissue is exactly the experiment that has not been done¹.

The bottom line

Established in mice: near-total uterine epithelial ablation causes infertility despite luminal resurfacing; lineage-labeled endometrial organoids engraft, rebuild luminal and glandular epithelium, and restore fertility; and luminal-type organoids convert to glandular identity in vivo through a stepwise program that also operates during endogenous repair. Open: the instructive niche signals, the human relevance, and the molecular identity of the gland-derived factors required for implantation. What would confirm translational value: the same conversion demonstrated with human endometrial organoids in a humanized injury model, and identification of the stromal cues so the switch becomes controllable rather than observed.