What the study does
Mubeen, He, Gruenhagen, Satoskar, and Streelman at the Georgia Institute of Technology set out to measure how a vertebrate that replaces its teeth naturally responds when teeth are removed, and to map the cells and signals involved (Mubeen et al. 2026, bioRxiv, DOI 10.64898/2026.01.06.697137). Their model is Lake Malawi cichlids, which normally replace each tooth with shape fidelity every 30 to 40 days. Mammals are the wrong comparison group here: mice never replace a tooth, and humans do it once, so a regenerating system is the only place this question can be asked directly.
The team worked in three rock-dwelling species with divergent dentitions: Cynotilapia afra (n = 8), Metriaclima estherae MZ Red (n = 8), and Petrotilapia chitimba Thick Bar (n = 9), 25 fish in all. Teeth were plucked from the right jaw half only, the left half serving as each fish’s internal control. A vital-dye pulse-chase distinguished old from new mineralized tissue: Alizarin Red at 100 micrograms per milliliter for 24 hours before plucking, then, after a 15-day recovery, Calcein at 50 micrograms per milliliter for 24 hours, with euthanasia 24 hours later.
Replacement speeds up about threefold, only on the plucked side
Scored as new teeth divided by total teeth per jaw quadrant, tooth gain was about 3 to 4 times higher on the plucked side across all three species, with roughly 60 percent of plucked-side teeth new after 15 days (Figure 1C and 1D; p less than 0.0001, paired Student’s t test). The post-plucking rate did not differ significantly between species (one-way ANOVA with Tukey-Kramer post hoc), which is notable given that one species carries tens of large conical teeth and another hundreds of small tricuspid teeth. The acceleration is local and proportional: it is not an ablation response, it is the normal replacement cycle mobilized faster, confined to the manipulated side.
The cell map behind the cycle
For the molecular readout the team turned to MZ Red alone, chosen for its large jaws and dense replacement tooth germs, and collected plucked and control jaw quadrants at four time points: 60 minutes, 1 day, 3 days, and 7 days after plucking (4 biological replicates at day 0, 2 at each later point). About 3 billion reads yielded 27,114 nuclei passing quality control, which they resolved into 25 cell types in 4 broad classes, annotated with canonical markers: pitx2 for dental epithelium, pax9 for dental mesenchyme, aif1 for immune cells, spp1 for bone (Figure 2B).
Two trajectory results matter for the field. In epithelium, cells carrying successional-lamina and epithelial stem cell markers concentrated in the ventral enamel epithelium (VEE), and pseudotime analysis traced every epithelial lineage, including the full ameloblast maturation sequence, back to that compartment. In mesenchyme, there were two progenitor domains, one in the dental follicle marked by twist1 and one in the dental papilla marked by twist2, with dnmt1 expression overlapping both. Two small epithelial populations the authors call ES-1 and ES-2 sit beside these compartments; ES-1, enriched for slc7a5 and mmp15, has no clear counterpart among mouse incisor cell types, while ES-2 resembles mouse outer enamel epithelium and stratum intermedium.
The first week runs immune first, then construction
The time course is the paper’s most distinctive contribution. Differential expression is already widespread one hour after plucking, and the earliest up-regulated genes in epithelium lean toward inflammation and immune response. fosb is the standout: strongly up-regulated at day 0 in outer enamel epithelium and stratum intermedium, then below control by day 1. CellChat modeling of ligand-receptor traffic shows the same handoff in reverse: at day 0 the dominant senders are mesenchymal cell types and alveolar bone signaling to cycling ameloblasts; by day 1 the receivers switch to NK and T lymphocytes; day 3 is dominated by epithelial-to-epithelial crosstalk; by day 7 mesenchymal signaling to cycling ameloblasts returns, joined by ameloblast-to-pre-odontoblast signals. Across the whole week, plucked-side communication is driven by Collagen, MMP, Slit-Robo, SPP1, Semaphorin, and Notch signaling. One telling detail: the same collagen ligands pair with integrin receptors when aimed at ameloblasts at day 0, and with CD44 when aimed at immune cells at day 1, so one molecular family is doing remodeling and immune recruitment in sequence.
What it does and does not say about a third dentition
For the bioengineered tooth germ program (/programs/bioengineered-tooth-germ/), this is comparative biology, not a protocol step. Nothing here was shown in a mammal, and the engineering targets in that program, dissociation and reaggregation of dental mesenchyme and reconstituted germ transplant, are untouched by this work. What it supplies is a parts list and a schedule: which conserved cell types a replacing dentition runs on, and the time-ordered sequence of immune activation, vascular and nerve recruitment, odontogenesis, and bone formation that a replacement tooth follows. The authors’ own closing argument is that mammalian teeth retain homologous cell types and that regeneration was lost by changing what those cells do, not by losing them, so reinventing replacement may mean re-engaging ancestral molecular profiles in cells mammals still have. That is a hypothesis this dataset motivates, not one it proves.
The honest boundary sits close to the data. Every claim above is transcriptional inference from pooled samples: replicates were pooled by condition, individual identity was recovered computationally, and there is no spatial validation, which the authors name as the key missing experiment. The progenitor assignments come from trajectory algorithms on single time courses, not from lineage tracing. And the behavioral result, robust as the threefold figure looks across 25 fish, is one manipulation, whole-tooth plucking, in one lineage of fishes. Mouse, fish, organoid, and patient remain different statements, and this paper is a fish statement.
Provenance: grounded in the full open-access text of Mubeen et al. 2026 (bioRxiv 2026.01.06.697137, version 1), read in full, with program context from /programs/bioengineered-tooth-germ/. Method and sourcing standard at /method/.