What was tested
Sol Jeong, So Young Park, Hyemin Ku, Byeong Soo Eom, Dohyun Kim and Jin Man Kim, at Yonsei University and collaborators in Seoul, Republic of Korea, published in Experimental & Molecular Medicine on 6 October 2026 (open access; data deposited as GSE299319). The paper links two questions: how macrophages and dental pulp stromal cells (DPSCs) physically communicate in carious human pulp, and whether adding the extracellular matrix protein fibronectin to regenerative endodontic revascularization improves outcomes in immature teeth. The human tissue arm compared pulp from healthy teeth and teeth with moderate or severe caries by immunofluorescence. The in vitro arm co-cultured U937-differentiated macrophages with primary human DPSCs, sorted the two cell types by Thy-1 and CD14, and ran bulk RNA-seq on each; it also reanalyzed two published single-cell RNA-seq datasets of human pulp (GSE185222 and GSE146123, 19,768 cells in total). The in vivo arm used three dogs (one mongrel, two beagles), immature premolars, and a split-mouth design with 22 roots per group: the fibronectin group got a final 10-minute irrigation with 100 ug/ml fibronectin and a fibronectin-soaked collagen sponge placed on the induced blood clot, the control group a PBS-soaked sponge of the same size, with MTA coronal seals in both. Outcomes were read at eight weeks by periapical radiography, micro-CT and histology.
What they found
In tissue, CD68+ macrophages made increasing contact with Thy-1+ DPSCs as caries advanced: 11.2% of macrophages contacted stromal cells in healthy pulp, 23.7% in moderate caries and 48.7% in severe caries. Adhesion profiling pointed to fibronectin from DPSCs engaging integrin alpha5/beta1 on macrophages rather than homophilic adhesion molecules: a blocking RGD peptide reduced macrophage-DPSC contact in a concentration-dependent way, and CRISPR knockout of ITGA5 in RAW264.7 macrophages shortened the duration of contacts. Under co-culture, bulk RNA-seq found 586 differentially expressed genes in macrophages (459 up, 127 down, at P < 0.05 and |log2FC| at least 2). The chemokine pattern shifted away from pro-inflammatory CCL2 (log2FC -2.22) and CCL3 (-1.28) toward ELR-positive CXC chemokines (CXCL2, CXCL3, CXCL5, CXCL6, CXCL8) and VEGFA, and the CD206+ M2-like fraction rose while CD86 did not change. The single-cell reanalysis matched this: immune cells expanded from 1.5% of pulp cells in health to 6.7% in moderate and 18.1% in severe caries, while endothelial cells fell from 16.8% to 8.6% and 9.5%, and inferred cell-cell interaction counts rose from 278 (healthy) to 1,646 (moderate) and 2,470 (severe). In HUVEC tube-formation assays, conditioned medium from co-cultures built more vessel-like networks and less cleaved caspase-3 staining than either monoculture; this was partly blocked by an RGD peptide, by the CXCR2 inhibitor SB225002 and by the VEGFR inhibitor AAL-993. Fibronectin-coated surfaces activated an NFkB reporter (about 90.7% more GFP signal than the LPS-treated control over 12 hours; collagen-coated surfaces did not), and co-culture raised NFkB p65 intensity in macrophages to 146.2% plus or minus 17.7% in cytoplasm and 123% plus or minus 20.5% in nucleus.
The load-bearing result is the dog experiment. At eight weeks, mean cross-sectional root dentin area was 3.40 plus or minus 1.58 mm2 in fibronectin-treated roots versus 2.39 plus or minus 1.34 mm2 in controls (paired t-test, P = 0.014), and apical closure was observed in 13 of 22 treated roots (59.1%) versus 5 of 22 controls (22.7%) (two-sided Fisher’s exact test, P = 0.031). Histology in the treated roots showed an organized layer of polarized odontoblast-like cells, dentinal tubule-like structures, dentin sialoprotein-positive staining, more vascular structures, and more CD14+CD163+ M2 macrophages in the stroma; control roots held loosely organized fibrous tissue with thin dentin walls.
What it does not show
The in vivo arm is three animals with 22 roots per group. Roots from the same jaw share animal-level biology, so this is not a 44-unit trial despite the paired statistics; the split-mouth design pairs treated and control roots within animals, which helps, but the number of independent animals remains three. Follow-up stops at eight weeks, so nothing is known about long-term stability, mechanical strength of the new dentin walls, or pulp function. The regenerated tissue is described by the authors as pulp-dentin-like: organized and dentin sialoprotein-positive, but not shown to be innervated or physiologically equivalent to normal pulp. The mechanistic chain from fibronectin through integrin, NFkB and CXCL/VEGFA to angiogenesis is assembled across separate in vitro assays; in the animals, group differences could equally reflect direct effects of fibronectin on stromal cells, a possibility the authors themselves raise by citing fibronectin-driven DSPP upregulation in DPSCs, so the macrophage pathway is a proposed route, not a demonstrated one in vivo. The single-cell compositional numbers rest on two healthy donors and one donor per caries group, so they are donor snapshots rather than distributions. And the canine immature premolar, while the standard large-animal model for revascularization, is not a human jaw: delivery parameters are unoptimized and no toxicology or dose work is reported.
Where it sits in the field
For the pulp and dentin repair program, this is a positive preclinical efficacy signal in the exact clinical niche regenerative endodontics targets: necrotic immature permanent teeth, where standard root canal treatment stops root development and leaves thin, fracture-prone walls. Its specific addition to the record is not the scaffold material, which is an ordinary collagen sponge, but the claim that a single endogenous matrix protein, applied by irrigation and sponge at a modest 100 ug/ml, measurably improved root dentin deposition and apical maturation in a designed split-mouth comparison. If that holds up in larger animal series, the route to a clinical adjunct is short, since fibronectin is already manufactured to pharmaceutical grade. This is rescue of the existing tooth, 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 bound two author framings. First, “minimal risk of adverse reactions” is a biocompatibility expectation for an endogenous protein, not a finding of this study: no safety arm, toxicology or dose escalation appears in the paper. Second, “reprograms the damaged pulp environment” is mechanism language resting on the co-culture assays; in the dogs, what was measured is dentin area, apical closure and histology, and the macrophage story is one candidate explanation. We also note a reporting wrinkle: the text quotes mean dentin area as 3.40 plus or minus 1.58 versus 2.39 plus or minus 1.34 mm2 while the figure displays medians and interquartile ranges; we quote the text values and flag that the spread around each mean is large relative to the difference between them, which the paired design addresses but a casual read of the means alone would miss.
Provenance: grounded in the full open-access text of Jeong et al., Experimental & Molecular Medicine, published 6 October 2026, DOI 10.1038/s12276-026-01863-4, read in full from the publisher PDF; every number above was checked against that text. Exact contact-proportion sample sizes for the human tissue quantification appear only in figure panels and are not quoted here. Method and sourcing standard at /method/.