The first human taste bud organoids that respond like taste buds

What the work claims

This is a primary methods-and-characterization study: the first reported long-term expandable human taste bud organoid system. The claims are threefold. First, epithelial cells from adult human circumvallate papillae can be maintained for more than 20 passages and over 50 weeks in a defined medium, and a two-phase protocol that separates a mitogen-rich expansion phase from a low-mitogen differentiation phase drives efficient maturation of taste receptor cells. Second, single-cell RNA sequencing shows the organoids traverse the same lineage states as native tissue, from basal progenitors through transitional basal and LGR5-positive cells to Type I to IV taste receptor cells, with differentiated taste cells reaching nearly half of all cells under optimized conditions. Third, the organoids are functional: a subset of cells responds to bitter denatonium with dose-dependent, PLC inhibitor-sensitive calcium transients, distinct populations respond to sweet and umami stimuli, and the canonical TAS2R38 PAV-versus-AVI genotype-phenotype relationship is reproduced donor by donor.¹

How it works

The protocol was built by systematic subtraction. Starting from a kitchen-sink medium used for other epithelial organoids, the authors removed one factor at a time and found that Wnt3a, R-spondin 1, Noggin, the TGF-beta inhibitor A83-01, FGF10, and forskolin are each required for long-term growth, while EGF, nicotinamide, and a p38 inhibitor are dispensable for maintenance. EGF was kept anyway for biomass: with it, organoids reach roughly 700 micrometers in diameter by day 14, versus about 100 micrometers without it.

The decisive step is temporal. Sustained mitogenic signaling blocks terminal differentiation, because proliferating progenitors and maturing taste cells are forced to share one signaling environment. The authors expand for 7 days, then switch to a differentiation medium lacking EGF and Wnt3a-conditioned medium, with the GSK3 inhibitor CHIR99021 at 3 micromolar providing defined Wnt activation. Raising R-spondin 1 conditioned medium from 10 to 80 percent during differentiation, the high-R-spondin condition, further amplified canonical Wnt signaling and pushed the differentiated taste cell fraction from 10.1 percent of cells under low R-spondin to 35 percent under high R-spondin at the same time point, and to 48.7 percent after 21 days of differentiation. Across 37,897 single-cell transcriptomes from three donors, Wnt pathway activity peaked in the LGR5-positive compartment, placing these cells at the center of the differentiation response.

Functionally, 10 to 30 percent of dye-loaded regions of interest in differentiated organoids responded to 20 millimolar denatonium, with dose dependence and reproducibility across repeated stimulation and across donors. The PLC inhibitor U-73122 at 5 micromolar markedly attenuated the responses, tying them to the canonical taste transduction pathway. Salt responses were not detected with an amiloride-withdrawal protocol, and citric acid gave broad, non-discrete signals, so sour and salty modalities remain unresolved. The genotype result is the cleanest in the paper: sequencing identified PAV and AVI haplotypes at TAS2R38, PAV organoids contained cells responding to phenylthiocarbamide, AVI organoids contained none, and denatonium responses were comparable in both.¹

Where a skeptic should push

Three cautions deserve weight. First, donor and tissue scope: the organoids derive from a single papilla region, the circumvallate papilla, with three donors for the maintenance experiments and three for single-cell analysis. Whether the protocol generalizes across the broader human tongue, across ages, and across many donors is untested, and the cross-species divergence the paper itself documents makes it plausible that culture requirements differ by donor as well as by species.

Second, architecture is explicitly not recapitulated. Proliferating cells in the organoids lack the basal spatial organization of native epithelium, and the calcium-imaging readout is restricted to cells at the organoid periphery where dye loading and optics work. This is a dissociated epithelium with correct cell types and correct signaling, not a reconstructed taste bud; paracrine geometry, innervation, and the roughly 10 to 14 day in vivo turnover rhythm are all absent.

Third, the headline human-versus-mouse divergence could partly be an artifact of the culture itself. Human LGR5-positive cells carry a distinct transcriptional program, including the endothelial-associated factor ERG where mice use Foxe1, and human LGR5 cells express Wnt-associated genes more broadly. But the differentiation conditions deliberately deliver intense Wnt stimulation, and the authors concede they cannot separate true species biology from culture-induced state. The new Type I markers (MSLN, FGF3, NOTUM) are validated by immunostaining in native human and mouse tissue, which is real evidence, but the functional repertoire is still demonstrated for only three of five taste qualities.

Taste organoids and the drug hunt

The obvious use is a human-mimetic platform where none existed: taste dysfunction from chemotherapy, viral infection, and aging has no good model system, and drug candidates that modulate taste (bitter blockers for pediatric formulations, sweetener optimization, appetite-modulating compounds) have been screened almost entirely in rodents or heterologous cells. A renewable human tissue with a functional calcium readout and a demonstrated genotype-phenotype link turns taste into a tractable organoid pharmacology problem for the first time. The TAS2R38 result has a subtler implication: bitter receptors are expressed outside the tongue, including in airway and gut epithelium, and a platform that preserves receptor genotype in donor-derived tissue could let drug programs test whether bitter receptor polymorphisms drive inter-individual variability in off-target drug responses rather than assume it.

The threat cuts the other way, and it is the one this paper documents from inside: the mouse taste stem cell program is not the human one. LGR5 was imported from mouse literature as the master taste stem cell marker; here it marks a transcriptionally distinct state in human cells, regulated differently and wired to Wnt differently. Any chemosensory program, toxicity program, or publication that treats mouse taste bud biology as a proxy for human has a concrete counterexample in this dataset. The two-phase design lesson generalizes further: organoid maturation is gated by mitogen withdrawal, so expansion-optimized protocols systematically under-report the differentiated phenotypes that toxicology and efficacy screens actually need to measure. A screen run in the expansion medium would have seen mostly progenitors and concluded the model does not work.

For organoid-based drug discovery broadly, this is a template for claiming a new tissue: demonstrate lineage trajectories at single-cell resolution, demonstrate a functional readout tied to a named signaling pathway, and demonstrate that a human genetic polymorphism predicts the response. Few organoid papers meet all three bars; this one does, with the unresolved modalities and single-region sourcing stated honestly enough to bound the claim.

The bottom line

Established: human circumvallate papilla organoids expand long-term, differentiate into Type I to IV taste receptor cells under a two-phase Wnt-tuned protocol reaching nearly half taste cells, respond to bitter, sweet, and umami stimuli through the canonical PLC pathway, and reproduce the TAS2R38 genotype-phenotype relationship. Hypothesis: the platform will model human taste dysfunction and support chemosensory drug screening at scale. What would confirm it: replication across more donors and papilla types, demonstration of drug-induced taste injury and rescue, and extension to sour and salty responses. What would break it: failure of the differentiation efficiency to hold beyond the three donors shown, or evidence that the high-Wnt culture state, rather than human biology, drives the divergent LGR5 program the paper leans on.