What was tested
Wanmin Zhao, Wenzhe Wang, Jia Guo (co-first authors) and colleagues, working at the Fourth Military Medical University in Xi’an, China, with collaborators, published this study in Exploration on 8 October 2026 (open access; DOI 10.1002/exp2.70229). The question is how transplanted donor stem cells integrate with host tissue during dental pulp regeneration, a known weak point of cell therapies. The group previously regenerated pulp in human patients with autologous deciduous-tooth stem cells (SHED), restoring tooth sensation (Xuan et al., Science Translational Medicine, 2018); here they ask what the implanted cells become and how host nerves talk back.
Two in vivo models anchor the work. Human SHED were implanted into emptied minipig incisor root canals after pulpectomy and read out at 3 and 6 months. Human SHED were also implanted into rat maxillary first-molar root canals for mechanistic tracing, with trigeminal ganglion (TG) neurons cultured beside them in transwell and direct co-culture, plus subcutaneous ectopic implants of cell-loaded empty root canals in immunocompromised mice.
What they found
Regenerated minipig pulp at 6 months had an odontoblast layer, blood vessels, and full-length beta-III-tubulin axons wrapped by S100b- and MBP-positive Schwann cells (SCs), with normal-pulp-like root canal morphology on micro-CT (n = 3 pulps per condition for quantification). Single-cell RNA sequencing (three normal and three regenerated tissues at 6 months) resolved nine cell clusters present in both tissues, but the proportions differed sharply: about 49.3% of cells in regenerated pulp expressed SC-related markers, against 1.0% in normal pulp.
Tracking implanted versus host cells across time showed the switch. At 3 months, implanted-cell progeny were 19.9% MSCs and 3.7% SCs and myelinating SCs, while recruited host cells were mainly MSCs (18.7%) and endothelium (15.3%). At 6 months, implanted cells had shifted to 0.4% MSCs and 19.0% SC-marker-positive cells, and host cells were dominated by SCs and myelinating SCs (29.7%) plus dental pulp cells (15.5%). Human-mitochondria labeling confirmed the split: the SCs were donor-derived, the axons they enwrapped were host trigeminal.
The communication runs on extracellular vesicles (EVs) in both directions. SHED-EVs (average diameter 143.5 nm; CD9, CD63, TSG101 and Alix positive, Calnexin negative) migrated from root canals to the ipsilateral TG in rats, appeared in trigeminal neuron soma, and, at 30 micrograms per milliliter, boosted axonal growth of cultured trigeminal neurons (RNA-seq: 391 genes upregulated, 36 downregulated, with axon-regeneration terms enriched; n = 30 neurons). Blocking EV secretion with GW4869 before implantation cut axon regrowth and impaired SC differentiation in minipigs. The return signal: neurons that had internalized SHED-EVs secreted their own EVs (ST-EVs, 122.6 nm), which carried elevated TUFM and GFM2 (two of twelve proteins upregulated more than 1.5-fold, p less than 0.05, in ST-EVs versus unstimulated neuron EVs) and drove SHED toward SCs. Knockdown of Tufm, but not Gfm2, in this EV fraction abolished the effect; TUFM knockdown in SHED suppressed oxidative phosphorylation and SC differentiation, both rescued by ST-EVs; overexpressing TUFM raised SC markers. Consistent with a strict order of events, axon density peaked at week 6 in co-culture while SC differentiation kept rising to week 8, and injecting GW4869 into the rat TG left axons growing but significantly blocked SC differentiation (n = 4 pulps per condition), with RAB27A knockdown giving the same picture.
What it does not show
No sensory function was measured after the xenogeneic transplants; the authors state this plainly as a limitation, so “functional” here means histological architecture, not proven sensation. The temporal-sequence evidence leans on ectopic subcutaneous implants, which the authors say need orthotopic validation. Immune responses to human cells in pig and rat were not assessed, which matters for the chimeric framing. GW4869 is a blunt, off-target-prone inhibitor; the RAB27A knockdown arms support the EV story, but EV work always carries cargo-purity caveats, and the ST-EVs definition is operational (neuron EVs released after SHED-EV uptake). The single-cell proportions come from small sample counts (three tissues per group), and human relevance of the TUFM cargo rests on sequence conservation (93.81% rat-to-human identity) rather than human data.
Where it sits in the field
For the pulp and dentin repair program, this is mechanism work one level below the program’s clinical question: it explains how a transplanted pulp stem cell population could rebuild not just tissue but wiring, by farming out the neuron half of the job to the host trigeminal ganglion through a reciprocal vesicle exchange. It extends, rather than replaces, the human autologous-SHED result from 2018 by mapping donor cell fates in xenogeneic minipig and rat models. It is repair of the existing dentition, not a third-dentition route, and it moves no program tier; the current field assessment stands at /field/.
Where we differ from the coverage
We found no press or popular coverage of this paper to differ from. Against the paper itself, we hold two framings back. “Functional tissue regeneration” overstates what was measured: no nerve-conduction or sensory readout was performed after xenogeneic implantation, so the honest claim is anatomical reinnervation. And the roughly 49% Schwann-cell share of regenerated pulp is a single-model scRNA-seq proportion (three pigs per group), a strong signal but not yet a field-level constant.
Provenance: grounded in the complete open-access full text (PMC13650028) of Zhao et al., Exploration, published online 8 October 2026, DOI 10.1002/exp2.70229, read in full via Europe PMC, with bibliographic metadata cross-checked against Crossref and the Europe PMC record; the prior human SHED study is cited from the paper’s reference list as Xuan et al., Science Translational Medicine 2018, DOI 10.1126/scitranslmed.aaf3227. Method and sourcing standard at /method/.