A hormone-sensitive stem cell hides in endometrial organoids

What the work claims

Tang, Monsivais and colleagues claim that cells with high activity of aldehyde dehydrogenase 1 (ALDH HI cells) are a genuine adult stem-cell population of the endometrium, not just a marker of stress or differentiation. The claim rests on four pillars: lineage tracing showing ALDH1A1-positive cells are long-lived and expand to populate luminal and glandular epithelium; hormone manipulations showing their location is dynamically controlled by estrogen and progesterone; diphtheria-toxin ablation showing they are required for normal gland number and glandular identity in vivo; and organoid assays in both mouse and human showing ALDH HI cells have far greater clonogenic capacity.¹ The authors further propose, more speculatively, that ALDH1A1-positive cells shed in menstrual effluent could serve as a biomarker for endometriosis propensity.

This is peer-reviewed primary work (eLife reviewed preprint, version 2), which puts it a confidence tier above the preprints covered elsewhere in today's stream.

How it works

ALDH1A1 converts retinaldehyde to retinoic acid, so the marker ties a metabolic enzyme to a developmental signalling pathway. The authors sorted endometrial epithelial cells with the ALDEFLUOR assay, which measures the enzyme's activity directly rather than its abundance, and compared the two populations.

In mouse, the contrast was stark: ALDH HI cells formed organoids at 2.69% versus 0.01% for ALDH LO cells (n=5 and n=3, p below 0.05), with larger organoids (936±117 versus 664±119 µm perimeter; 65,068±17,014 versus 33,017±10,580 square micrometres in area) and a stemness-biased transcriptome, 346 transcripts up and 674 down at adjusted p below 0.005, with cell-cycle genes repressed and BMP and Wnt pathway genes induced. Single-cell RNA-seq of about 5,984 estrus and 3,995 diestrus cells located Aldh1a1 in quiescent epithelial stem-cell clusters co-expressing Lgr5 and Axin2, with a trajectory running from those quiescent cells through a proliferative compartment into glandular and then luminal fates.¹

Location was hormone-dependent: ALDH1A1 protein sat in glandular crypts during estrus, spread through luminal and glandular epithelium in diestrus, spread everywhere after ovariectomy, and was driven back into crypts by estradiol or progesterone treatment. Lineage tracing from postnatal day 7 or 14 labelled cells that persisted and expanded through day 56; in adult cycling mice the stromal labelled population expanded most clearly; after parturition, labelled cells concentrated in the regenerating stroma near placental detachment sites. In the ablation experiment, giving diphtheria toxin to mice whose ALDH1A1-positive cells expressed the toxin receptor cut gland counts nearly in half by postnatal day 56 (11.75±8.2 versus 22±3.4 glands, p below 0.05) and reduced FOXA2 intensity per gland (10,434±2,136 versus 15,079±3,006, p below 0.001), while matched organoids disintegrated with cleaved caspase-3 staining after toxin treatment in vitro.¹

In human eutopic endometrium from donors, ALDH HI cells again formed organoids more efficiently than ALDH LO cells, with two independent quantifications showing 1.98±0.38 versus 0.58±0.055 organoid-forming units (p=0.022) and 9.04±0.72 versus 1.7±0.38 (p=0.0008), and they propagated longer (23±3 versus 14±2 passages, n=3 donors, p=0.0576, not significant). Their transcriptomes ran toward gland development and proliferation genes and away from ciliated, luminal-like genes, and ALDH LO organoids carried about nine times more ciliated cells (7.2±1.56 versus 0.77±0.42 per field, p below 0.001).¹

Where a skeptic should push

The most load-bearing assumption is that high ALDH activity cleanly marks a stem-cell state rather than a transient metabolic state that any epithelial cell can enter. The evidence for flexibility is inside the paper: expression of ALDH1A1 is hormone-dependent and reversible, and complete ALDH1A1 knockout mice are viable and fertile, with gland development largely preserved, which the authors attribute to compensation by other ALDH isozymes. A marker whose loss is compatible with a near-normal organ is a softer stemness criterion than ablation implies. The partial gland phenotype after toxin ablation (an average loss of about ten glands, with high variance) is consistent with that softness.

The human data carry specific fragilities. The headline organoid-formation numbers come from one patient sample run as technical replicates, with additional patients relegated to supplements; the RNA-seq compared pooled ALDH HI and ALDH LO organoids from three patients, all with endometriosis, so the eutopic tissue analysed here comes from a disease context and was never compared against healthy endometrium. That matters doubly because the biomarker proposal is endometriosis-specific: the study shows ALDH HI cells behave differently within patients who already have the disease, not that ALDH HI abundance predicts who will get it. Whether ALDH activity in menstrual effluent stratifies risk is untested in this paper.

There is also a mouse-to-human translation wrinkle: mice resorb the endometrium each cycle rather than menstruate, and the postpartum repair process, where ALDH1A1-positive stromal cells were most prominent, has no exact human counterpart. The human organoid work captured epithelial behaviour only; the stromal expansion story, arguably the more novel, rests on mouse data. Finally, the mechanistic chain from ALDH1A1 through retinoic acid to estrogen-antagonised epithelial fate is a plausible synthesis of the group's prior work, not a demonstrated pathway in this dataset.

What this means for disease-model organoid banking

For organoid-based drug discovery, the practical lesson is about sorting. If a two-hour ALDEFLUOR sort enriches clonogenic endometrial epithelium roughly 3-fold to 9-fold in establishment efficiency, then every endometrial organoid biobank that banks unsorted tissue is carrying a variable mixture of stem-cell-enriched and differentiated fractions, and that variable will show up as unexplained line-to-line behaviour in drug-response experiments. The same logic generalises: enzyme-activity sorts are cheap, orthogonal to marker-antibody panels, and capture metabolic state, which is exactly what static transcriptomic reference atlases miss. Any organoid bank built for reproductive-toxicity screening, endometriosis modelling or regenerative-medicine work should be asking whether its lines are stem-cell-enriched by design or by accident.

The opportunity is a stratified endometriosis model. The study's own framing, that basalis-like, stem-marker-high cells in menstrual effluent may seed peritoneal lesions, suggests an experiment the organoid field is positioned to run: derive matched ALDH HI and ALDH LO lines from effluent and eutopic tissue of women with and without endometriosis, and test whether the ALDH HI fraction is selectively enriched for adhesion, invasion or lesion-initiating behaviour in assembloid co-culture with peritoneal mesothelium. That would convert a biomarker hypothesis into a druggable target: if lesion-initiating capacity tracks ALDH activity, inhibiting retinoic-acid production in those cells is a mechanism with a readout already built.

The threat is subtler. Endometrial organoids are being promoted as tools for personalised testing of progesterone-pathway drugs and for fertility-preservation toxicity screens. This paper shows those same organoids are exquisitely sensitive to the hormonal history of the donor and to which epithelial fraction was captured, while lacking the stromal compartment that hormones act through. A hormone-response assay run in epithelium-only endometrial organoids may systematically misestimate drug effects whose mechanism is stromal-epithelial crosstalk, which is most of reproductive endocrinology. The model is not wrong; it is answering a narrower question than its users may think.¹

The bottom line

Established, with peer review: ALDH1A1-high endometrial cells are long-lived, hormone-repositioned, clonogenically superior in organoid assays, and necessary for full gland complement in mice, with concordant behaviour in human eutopic endometrium from three donors. Hypothesis, not result: that ALDH1A1-positive cells in menstrual fluid predict endometriosis, and that the retinoic-acid pathway is the causal link between ALDH activity and stemness. Confirmation would require prospective, healthy-control cohorts and functional perturbation of retinoic-acid signalling in sorted human cells; the claim would break if ALDH activity turns out to mark a reversible metabolic state that correlates with, but does not cause, the stem-cell programme. Either way, the sorting result alone is worth acting on in how endometrial organoid banks are built.