What was tested
Mhmod and colleagues at Ain Shams University and Cairo University compared the GSK-3 inhibitor tideglusib against mineral trioxide aggregate (MTA), the clinical gold standard for direct pulp capping, in an in vivo dog model (Mhmod et al. 2025, BMC Oral Health, DOI 10.1186/s12903-025-06546-6). The work directly tests the small-molecule dentin-repair idea that Neves et al. (2017) established in a mouse tooth-injury model and that anchors the pharmacological thread of the pulp-dentin repair program (/programs/pulp-dentin-repair/).
Two adult male mongrel dogs, mean age two years and mean weight 20 kg, contributed 56 teeth. The team prepared Class V buccal cavities with standardized mechanical pulp exposures, achieved hemostasis with saline, and capped 14 teeth per subgroup: ProRoot white MTA in one subgroup and a resorbable Type I collagen sponge soaked in freshly prepared tideglusib solution in the other. Access cavities were sealed with glass ionomer. Dogs were euthanized for histology at three weeks (group A) or eight weeks (group B). The study was approved by the Ain Shams Faculty of Dentistry ethics committee (approval 105, 15 July 2020) and reports following ARRIVE guidelines.
One reporting problem deserves flagging before any numbers: the dose is stated inconsistently. The abstract says the collagen was soaked in a 50 nM tideglusib solution, the methods describe preparing a 30 nM/mL target solution, and the discussion states the team used 30 nM/mL, citing a rabbit study that found maximal pulp-cell viability at 34 nM/mL. The paper never reconciles these figures.
What was found
Histology was scored for inflammatory cell count (ImageJ-assisted, four fields per slide at 40x), pulp tissue disorganization (0 to 3 scale), and new hard tissue formation (0 to 2 scale) (Mhmod et al. 2025, Table 1).
On inflammation, tideglusib-capped teeth scored significantly worse at both timepoints. At three weeks the mean inflammatory cell count was 125.43 plus or minus 8.96 under tideglusib versus 86.23 plus or minus 1.05 under MTA; at eight weeks it was 224.00 plus or minus 10.78 versus 0.00 under MTA, with P less than 0.001 for both comparisons. On tissue disorganization, tideglusib scored 2.14 plus or minus 0.18 versus 1.05 plus or minus 0.23 at three weeks (P = 0.013) and 1.50 plus or minus 0.20 versus 0.71 plus or minus 0.16 at eight weeks (P = 0.018).
On the outcome that matters for dentin repair, there was no benefit. No tooth in either group showed any hard tissue at three weeks. At eight weeks, only two of fourteen MTA teeth showed a partial bridge (score 1, group mean 0.14 plus or minus 0.10), and zero of fourteen tideglusib teeth showed any (P = 0.157, not significant). The authors rejected their null hypothesis on inflammation and disorganization but found no hard-tissue advantage for the drug.
How to read the result
This is a negative result for tideglusib as a capping agent, delivered in one model with real limitations. Two dogs, a mechanical rather than carious exposure, a collagen carrier, and an eight-week maximum observation window bound the claim. The authors themselves offer a material-level explanation: the biodegradable collagen sponge may have lost the bacteria-tight seal that MTA maintains under the restoration, so the tideglusib group may partly reflect bacterial ingress rather than drug toxicity. They also note that tideglusib is dose-dependently cytotoxic to fibroblasts in vitro, citing their own 2024 data, and they flag the short follow-up as the main limitation.
The dose inconsistency noted above cuts both ways: if anything above roughly 34 nM/mL kills pulp cells in culture, a sponge soaked near that range is not obviously the right formulation, and the paper cannot say precisely what dose the pulps actually received.
Where it leaves the tideglusib story
The result does not overturn Neves et al. (2017), who reported denser reparative dentine with tideglusib than with MTA in mice, but it is the larger and more clinically shaped head-to-head, and it points in the opposite direction. It is also consistent with a human tooth culture study reporting no mineralization with 50 nM tideglusib (Sukajintanakarn et al. 2020, conference proceedings), a weaker source but the same sign. For the pulp-dentin repair program, the honest update is narrow: pharmacological dentin repair with GSK-3 antagonists remains preclinical, and the first dog-model comparison against MTA found inflammation without regeneration. The program’s standing note that true human pulp-dentin repair evidence still rests on small autologous cell trials is unchanged.
Where we differ from the coverage
Popular coverage of tideglusib has tended to run on the 2017 mouse result, framing it as an Alzheimer’s drug that regrows teeth. That framing was already ahead of the evidence, and this dog study is the clearest in vivo counterpoint yet: in a direct capping comparison, the drug group healed worse, not better. Nothing here shows that GSK-3 inhibition cannot promote dentin repair in some delivery format; it shows that a collagen-sponge format, at a poorly reported dose, failed against MTA in dogs.
Provenance: grounded in the full open-access text of Mhmod et al. 2025 (PMC12276705), read in full, with the Neves 2017 context taken from the program record at /programs/pulp-dentin-repair/. Method and sourcing standard at /method/.