Stem Cells From Human Exfoliated Deciduous Teeth as a Potent Cell Therapy for Radiation‐Induced Osteonecrosis Model

ABSTRACT

Osteonecrosis is a serious disorder leading to subchondral bone collapse and joint dysfunction. Osteonecrosis has been increasingly diagnosed before bone collapse; however, effective treatments for regenerating necrotic bone remain unavailable. Mesenchymal stem cell (MSC)‐based therapies are promising; however, bone marrow‐derived MSCs (BMSCs) are limited by donor burden, low yield, and suboptimal regenerative efficacy. Stem cells from human exfoliated deciduous teeth (SHED) exhibit higher proliferative capacity, enhanced growth factor secretion, minimal senescence, and noninvasive harvesting, making them a strong alternative. We evaluated SHED and BMSCs in a refractory rat model of radiation‐induced tibial osteonecrosis. In vitro, SHED secreted similar or higher levels of proangiogenic and chemotactic factors—including vascular endothelial growth factor, angiopoietin‐2, stem cell factor, monocyte chemoattractant protein‐1, and C‐C motif chemokine ligand 5—than BMSCs and significantly enhanced macrophage migration. Hydroxyapatite/collagen scaffolds seeded with SHED or BMSCs were transplanted into irradiated tibial bone defects in vivo. Micro‐computed tomography revealed significantly accelerated bone regeneration with SHED, but not with BMSCs, compared with cell‐free scaffolds. Histology confirmed new bone and marrow formation at SHED‐treated sites, accompanied by increased periosteal proliferation and vascularity. Bulk RNA sequencing showed that SHED transplantation upregulated angiogenesis and immune cell recruitment–related pathways, whereas BMSCs upregulated genes associated with endochondral ossification without effective regeneration. Notably, SHED did not differentiate into osteoblasts or chondrocytes in vivo, indicating paracrine mediation of angiogenesis and macrophage recruitment. Given the limited regenerative capacity of necrotic bone, SHED's angiogenic and immunomodulatory activities position it as a promising candidate for cell‐based therapy in refractory osteonecrosis.

1. Introduction

Osteonecrosis, also known as ischemic or aseptic osteonecrosis, most commonly affects periarticular regions, particularly the epiphyseal and subchondral bone. Osteonecrosis can result from various causes, including drug‐related factors such as corticosteroids, trauma, infection, and radiation therapy‐related effects on tumors. Progressive necrosis lesions lead to articular surface collapse, resulting in secondary osteoarthritis and substantial loss of joint function. In advanced cases, joint replacement surgery may be necessary. However, given the finite longevity of prosthetic implants and the potential need for revision procedures, joint preservation strategies have received increasing attention, particularly among young adult patients. Recent advances, including the identification of osteonecrosis risk factors and early detection before bone collapse using magnetic resonance imaging (MRI) and bone scintigraphy, have further emphasized the necessity for joint‐preserving strategies [1, 2].

Bone regenerative therapy using mesenchymal stem cells (MSCs) has recently been suggested to promote bone repair in osteonecrotic lesions [3, 4, 5, 6, 7]. Bone marrow‐derived mesenchymal stem cells (BMSCs) have demonstrated substantial regenerative potential [8]; however, clinical translation is limited by donor morbidity, cell yield constraints, and variable therapeutic efficacy. Given these limitations, MSCs derived from dental tissues have garnered increasing interest as alternative candidates for bone regenerative, primarily due to their superior accessibility. Human dental pulp stem cells (DPSCs) were first identified in 2000 [9], followed 3 years later by the discovery of stem cells from human exfoliated deciduous teeth (SHED) [10]. SHED exhibit minimal cellular senescence and higher proliferative capacity than adult DPSCs, characteristics that are advantageous for stem cell banking [11]. SHED outperform both BMSCs and adult DPSCs in proliferation, osteogenic differentiation, and bone repair capacity, highlighting their potential as a robust cell source for bone tissue engineering [10, 12]. In addition, SHED secrete high‐level cytokines with pro‐angiogenic activity and have been shown to be a promising source of perivascular cells for in vivo angiogenesis [13, 14]. Consequently, SHED represent a promising therapeutic avenue for the regeneration of ischemic osteonecrotic lesions.

Based on these findings, we hypothesized that SHED‐based cell therapy would achieve superior bone regeneration compared with BMSCs in osteonecrosis treatment. This study aimed to compare the bone regeneration capacity of SHED and BMSCs in osteonecrotic lesions and to elucidate their underlying mechanisms of action. As this model does not reproduce subchondral bone collapse, this study was conducted as a proof‐of‐concept, pilot investigation preceding future evaluation in an osteonecrosis model with joint collapse.

2. Methods

This study comprised in vitro and in vivo phases. In the in vitro phase, SHED and BMSCs were expanded in culture, and their secretion of angiogenic and chemotactic factors, together with their capacity to induce macrophage migration, was compared. In the in vivo phase, radiation‐induced osteonecrosis was first established by local irradiation of the hind limb in SCID rats. Four weeks later, SHED or BMSCs were seeded onto a hydroxyapatite/collagen (HA/Col) scaffold and pre‐cultured for 2 days, after which the cell‐seeded scaffold was implanted into a surgically created bone defect in the irradiated proximal tibial metaphysis; a cell‐free scaffold was implanted into the contralateral tibia as an internal control. Bone regeneration was monitored longitudinally by micro‐CT over 8 weeks, after which the animals were euthanized and the implantation sites harvested for histological analysis to evaluate new bone formation, vascularity, and the fate of the transplanted cells. In a separate cohort, tissue was harvested 2 weeks after transplantation for bulk RNA sequencing to characterize the transcriptional programs induced by SHED and BMSCs in both the transplanted human cells and the host rat tissue. Specific procedural details for each step are described in the following subsections.

2.1. In Vitro Study

2.1.1. Quantification of Growth Factors in SHED and BMSC Culture Supernatants

Three strains of SHED (S‐Quatre Corporation, Tokyo, Japan) and 2 strains of adult human BMSCs (PromoCell, Heidelberg, Germany) were used for the experiments. For both SHED and BMSCs, experiments were performed using cells between passages 6 and 10, starting from cryopreserved P6 stocks. A total of 5.76 × 10⁵ SHED and BMSCs were seeded in each respective proliferation medium (SHED: VERITAS, Tokyo, Japan, #ST‐05439; BMSC: Takara‐bio, Kyoto, Japan, #C‐28009) in a 6‐well plate. After 1 day of attachment, the medium was replaced with serum‐free MEM α (FUJIFILM, Tokyo, Japan, #135‐15175). After 2 days of culture, SHED and BMSC culture supernatants were simultaneously collected from three separate wells. Both cell types were cultured at 37°C with 5% CO₂ and 95% humidity. Protein levels in each culture supernatant were measured using the LEGENDplex Human Growth Factor Panel (#741061; BioLegend, Inc., San Diego, CA, USA) for all proteins except CCL5, and the Human Proinflammatory Chemokine Panel (13‐plex) (#740003; BioLegend, Inc., San Diego, CA, USA) for CCL5, according to the manufacturer's instructions, and the mean of three technical replicates was used for analysis.

2.1.2. Macrophage Chemotaxis Assay Using SHED and BMSC Conditioned Media

Since chemotactic factors secreted by MSCs play an important role in macrophage induction, which is essential for bone regeneration, we compared the macrophage‐inducing abilities of SHED and BMSCs.

SHED strain #1 and BMSC strain #1 were used in subsequent experiments. SHED and BMSCs were suspended in their respective growth media and seeded into 6‐well plates at 5.76 × 10⁵ cells per 2.5 mL per well; after 1 day of culture, the medium was replaced with serum‐free MEMα (FUJIFILM #135‐15175). 48 h later, the supernatants were collected from each well. The supernatant was centrifuged at 400 × g for 5 min and filtered through a 0.22‐μm filter. The supernatant was then diluted with MEMα to 30%, and fetal bovine serum (FBS; #175012, Nichirei Biosciences Inc., Tokyo, Japan) was added to achieve a final FBS concentration of 0.5%. MEMα alone, diluted SHED culture supernatant, and diluted BMSC culture supernatant were prepared in triplicate in a 24‐well plate. Corning® FluoroBlok™ cell culture inserts (#351152, 8.0‐µm pores) were placed in a 24‐well plate. Mouse RAW 264.7 cells were suspended in MEMα and seeded into the upper chamber (2.1 × 10⁴ cells in 350 μL per insert). After 24 h of migration, non‐migrated cells on the upper surface of the membrane were removed with a cotton swab. Migrated cells on the lower surface were fixed with 4% paraformaldehyde, permeabilized with 0.3% Triton X‐100, and stained with DAPI (1 μg/mL). Fluorescence images were acquired at 10× magnification, and migrated cells were quantified by counting DAPI‐positive nuclei using ImageJ [15].

2.2. In Vivo Study

2.2.1. Animals and Radiation‐Induced Osteonecrosis Model

All animal procedures were performed in accordance with the institutional guidelines for laboratory animal care. Ethical approval for the experiments was obtained from the institutional review board (approval numbers 24‐0075 and 25‐0005). In vivo experiments were conducted using 20‐week‐old female SCID rats (F344‐Il2rg^em1Iexas) obtained from the Institute of Medical Science, University of Tokyo, Japan. During radiation therapy, surgery, and micro‐computed tomography (micro‐CT) scans, anesthesia was induced using a mixture of 10 mL ketamine with 0.67 mL dexmedetomidine, administered at a dose of 1 mL per 200 g of body weight.

In this study, a radiation‐induced osteonecrosis animal model was adopted, as it is the most severe and reproducible ischemic osteonecrosis model in terms of impaired bone regeneration [16, 17, 18]. Unlike animal models of bone necrosis induced by the cauterization of bone‐nourishing blood vessels, this model does not exhibit angiogenesis from collateral circulation within a few weeks. Using an X‐ray irradiation device (Faxitron X‐ray CP160, AcroBio Inc., Tokyo, Japan), areas other than the hind limbs were shielded with 2‐mm‐thick lead plates, and a single 20‐Gy dose of radiation was administered to each hind limb.

2.2.2. In Vivo Experimental Design

An overview of the in vivo experiments is shown in Figure 1. Ten SCID rats were allocated to the SHED transplantation group, and another ten SCID rats were allocated to the BMSC transplantation group. The sample size was determined with reference to previously published studies employing similar experimental designs [12]. Radiation was administered to the hind limbs 4 weeks prior to cell transplantation. Pre‐culture of 3 × 3 × 3 mm bioabsorbable HA/Col product, ReFit (HOYA Technosurgical Co., Tokyo, Japan), seeded with 2.0 × 10⁵ SHED, BMSCs, or no cells, was performed 2 days prior to surgery. On the day of cell transplantation, the proximal tibia was surgically exposed, and a 3‐mm‐diameter bone hole was created in the proximal metaphysis of both tibias using an electric burr drill. The cell‐seeded ReFit was transplanted into the bone hole of one tibia, whereas the cell‐free ReFit was transplanted into the contralateral tibial bone hole. The treated rats were observed longitudinally for 8 weeks. In the bulk RNA sequencing (RNA‐seq) study, either cell‐seeded ReFit or cell‐free ReFit was transplanted into the bone hole, and gene expression profiles in the osteonecrotic region were analyzed 2 weeks after transplantation. Gentamicin (4 mg/kg) was administered subcutaneously immediately before surgery. All the surgical procedures were performed by the same surgeon (TM). After surgery, buprenorphine (0.05 mg/kg) was administered subcutaneously every 12 h for 2 days for pain relief. Animals were housed under standard laboratory conditions and observed throughout the experimental period.

Figure 1.

A schematic overview of in vivo experiments. (a) Radiation exposure. Radiation was exposed only to hind limbs of rats to induce tibial osteonecrosis. (b) Cell transplantation to tibial osteonecrosis model in rats. Four weeks after radiation exposure, bone holes were created in the metaphysis of both tibias, and cell transplantation was performed. Prior to this, 2 days before transplantation, transplant cells were seeded on a scaffold material and pre‐cultured. (c) In the group undergoing bulk RNA seq, tissue surrounding the bone hole was excised, and RNA was collected 2 weeks after transplantation. GSEA was performed using species‐specific gene sets for human and rat to confirm the gene expression status of transplanted cells and host cells. In the group undergoing longitudinal micro‐CT observation, micro‐CT were taken every 2 weeks after transplantation, and histological evaluation was performed 8 weeks after transplantation. CT, computed tomography; GSEA, gene set enrichment analysis; RNA seq, RNA sequencing.

2.2.3. Micro‐CT Analysis

Bilateral tibias were scanned individually using micro‐CT (CT, R_mCT2; Rigaku, Tokyo, Japan) at 10‐μm isotropic resolution. Micro‐CT imaging was performed under anesthesia at 0, 2, 4, 6, and 8 weeks post‐transplantation to evaluate the bone repair rate over time. Two‐dimensional micro‐CT images (slice thickness: 60 μm) parallel to the medial cortical surface of the tibia were reconstructed from the three‐dimensional dataset using multi‐planar reconstruction. The bone repair rate was calculated using ImageJ software (NIH, Bethesda, MD, USA) as the percentage of the original 3‐mm‐diameter defect area occupied by newly formed bone. To calibrate the gray values, a hydroxyapatite bone density phantom (Ratoc, No. 01‐E5D1mmH) was scanned under identical conditions, allowing the conversion of CT values to bone mineral density (mg HA/cm³). A threshold of 300 mg HA/cm³ was applied to define bone tissue [19]. All measurements were performed in a blinded manner (HK). The boundaries of the original bone defect in follow‐up images (weeks 2–8) were identified by referencing the baseline (week 0) images.

2.2.4. Histological Analysis

8 weeks after cell transplantation, the animals were euthanized, and the treated tibiae were harvested. The harvested tibiae were subsequently fixed in 4% paraformaldehyde and demineralized using Morse's solution (FUJIFILM, #135‐17071). After decalcification, the tibiae were sectioned perpendicular to the bone hole and stained with hematoxylin and eosin (H&E), anti‐CD31, and anti‐lamin antibodies. The sections were incubated for 60 min at room temperature with an anti‐CD31 primary antibody (1:2000, Abcam, Cambridge, UK, ab182981) to assess neovascularization by anti‐CD31 immunostaining. After incubation with a secondary antibody (Histofine Simple Stain MAX‐PO(R); Nichirei Bioscience code: 414141 F), the sections were visualized and imaged using a light microscope (BZ‐X710, Keyence, Osaka, Japan). The number and total area of blood vessels were quantified within regions of the bone hole lacking newly formed bone. The analysis region was defined on the surface of the transplanted area, and the vascular area per unit area and the number of vessels per unit area were measured. Blood vessel analysis was performed on 5 SHED‐transplanted individuals and 5 BMSC‐transplanted individuals. Specifically, 5 SHED‐transplanted limbs, 5 BMSC‐transplanted limbs, and 10 limbs transplanted with the carrier alone were analyzed. Anti‐lamin immunostaining was performed to detect the nuclear membrane of SHED; sections were incubated for 60 min at room temperature with an anti‐human lamin A + lamin C primary antibody (1:2000, Abcam: ab108595), followed by incubation with a secondary antibody (Histofine Simple Stain MAX‐PO(R), Nichirei Biosciences code: 414141 F). The sections were examined and imaged using a light microscope to confirm SHED localization.

2.2.5. RNA Extraction, Sequencing, and Data Processing

Bone holes were created 4 weeks after irradiation using the same protocol as that used in the bone repair observation experiment, after which scaffold material alone, BMSCs plus scaffold material, or SHED + scaffold material was transplanted into both tibiae. Two weeks after transplantation, the surrounding soft tissue was removed from the tibia, and a 5 mm‐long section of the tibial diaphyseal end, including the bone defect, was excised using an electric cutter. This bone sample was homogenized, and RNA was extracted using Trizol and the RNeasy® Mini Kit (Qiagen, Hilden, Germany) with on‐column DNase digestion, and RNA integrity was assessed using a Bioanalyzer. Libraries were prepared by Novogene Japan Co. Ltd (Tokyo, Japan) using s poly(A)‐selected, strand‐specific, reverse‐stranded (ISR) protocol and sequenced as 150‐bp paired‐end reads. Adapter and low‐quality read trimming followed the provider's standard operating procedures; the resulting clean reads exhibited Q30 scores greater than 94% and an error rate below 0.03%.

Reads were independently aligned to human (GRCh38.p14 + GENCODE v49) [20] and rat (mRatBN7.2 + Ensembl 110) [21, 22] reference genomes using STAR v2.7.10b [23]. Species‐specific prefixes (“hs_” and “rn_”) were added to both FASTA and GTF files, and STAR indices were built with splice junction annotations using ‐‐sjdbOverhang 149. BAM files were sorted by read name and processed using Disambiguate [24] (STAR mode) to classify the reads. Ambiguous reads (0.12%–0.58%) were excluded. Species‐specific BAM files were converted to paired‐end FASTQ files using samtools fastq, retaining only properly paired reads and excluding singletons.

Transcript abundance was estimated using Salmon v1.10.2 [25] with decoy‐aware gentrome indices [26] and selective alignment (validateMappings; libType A resolved to ISR). This two‐stage approach (genome‐based species classification followed by transcriptome‐based quantification) was validated by mapping pure rat control samples, which showed minimal human misassignment (0.027%–0.091%). TPM and count matrices were imported using tximport v1.30.0 [27] with countsFromAbundance set to “no.” Differential expression analysis was performed using DESeq. 2 v1.42.0 [28] (design ~ group; Wald test), with significance defined as false discovery rate (FDR) < 0.05 and |log2FC| > 1. No fold change in shrinkage was observed. One‐versus‐rest contrasts were implemented as linear combinations with equal weighting of group means (e.g., BMSC vs. 0.5 × SHED + 0.5 × Control), regardless of sample size per group [29].

Pipeline robustness was assessed by comparing analyses conducted with and without decoy sequences (differentially expressed gene [DEG] overlap 96.8%, Spearman's r = 0.999).

2.2.6. Gene Set Enrichment Analysis (GSEA)

GSEA was performed on DESeq. 2 Wald statistics using named, descending‐sorted ranking vectors (names = official gene symbols; values = stat). Duplicate symbols were collapsed by retaining the row with the largest |stat| value. The rat contrasts analyzed were SHED versus Control, BMSC versus Control, and SHED versus BMSC. The human analysis used the SHED versus BMSC contrast. A positive normalized enrichment score (NES) indicates enrichment on the left side of each contrast name, according to convention. Expression‐based prefiltering was not applied. Gene sets were drawn from MSigDB Hallmark mapped via msigdbr v25.1.1 (Rattus norvegicus for rat analyses; Homo sapiens for human analyses), retrieved on October 21, 2025; gene sets were restricted to sizes of 15–500. Enrichment was computed with fgsea v1.32.4 (nperm = 10000, minSize = 15, maxSize = 500, nproc = 1, seed = 42), returning enrichment score (ES) and NES. Multiple testing correction was performed using the Benjamini–Hochberg FDR method, with significance set at FDR < 0.05. Redundancy among significant pathways was reduced by removing one of any pair with overlap ≥ 0.60 or Jaccard ≥ 0.35, retaining the pathway with the larger |NES|.

2.2.7. Heatmap Analysis of Upregulated DEGs Associated With Angiogenesis, Cell Induction, and Endochondral Ossification

DEGs were identified by RNA‐seq analysis to compare SHED‐transplanted samples with controls and BMSC‐transplanted samples with controls. Functional enrichment analysis was performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) Bioinformatics Resources (v6.8) to identify significantly enriched Gene Ontology (GO) Biological Process (BP) categories. Enrichment was evaluated using the Expression Analysis Systematic Explorer (EASE) score and a modified Fisher's exact test, with multiple testing corrections applied using the Benjamini–Hochberg method. GO BP categories with an adjusted p value (Benjamini) < 0.05 were considered statistically significant.

In the SHED versus control comparison, genes significantly upregulated in the SHED transplantation group were annotated to GO BP categories associated with angiogenesis and cell induction. Similarly, genes that were upregulated in the BMSC transplantation group and annotated to GO BP categories associated with endochondral ossification were extracted from the BMSC versus control comparison.

Heatmaps were generated using SRPLOT, based on the normalized expression levels of the selected genes in individual samples. Gene expression values were normalized on a per‐gene basis using Z‐score transformation, with the color scale set from −2 to +2 to visualize relative expression patterns across samples.

2.3. Statistical Analysis

Statistically significant differences among groups were determined by one‐way analysis of variance (ANOVA), followed by Tukey's test for multi‐group comparisons. Two‐tailed Student's t‐test was used for two‐group comparisons. p < 0.05 was considered statistically significant. All data are presented as mean ± standard deviation. Statistical analysis was performed using JMP® software (JMP® Pro17, SAS Institute, Cary, NC, USA).

2.4. AI‐Assisted Refinement of Author‐Generated Schematic Illustration

For the refinement of the author‐generated schematic illustration in Figure 7, ChatGPT (OpenAI, San Francisco, CA, USA) was used as an assistive tool for graphical polishing, while all scientific content and final decisions remained under the authors' responsibility.

Figure 7.

Schematic representation of the hypothetical mechanism of action of SHED, as suggested by the results of this study. When SHED collected from children's dental pulp is transplanted into necrotic bone, CCL2, which promotes macrophage induction, and pro‐angiogenic factors that promote angiogenesis are secreted. CCL2 induces macrophages, and pro‐ angiogenic factors cause revascularization of the necrotic area. Macrophages that migrate to the necrotic area secrete CCL5 and induce MX1‐positive periosteal stem cells. The periosteal stem cells differentiate into osteoblasts, promoting bone regeneration. CCL2, C‐C motif chemokine ligand 2; SHED, stem cells from human exfoliated deciduous teeth.

3. Results

3.1. In Vitro Study: SHED has Greater Potential than BMSCs for Angiogenesis and Macrophage Induction

Protein measurements of cell culture supernatants revealed that SHED secrete multiple factors associated with angiogenesis and cell migration (Figure 2a). Vascular endothelial growth factor (VEGF) secretion was similar between SHED and BMSCs. Protein levels of angiopoietin‐2, stem cell factor (SCF), monocyte chemoattractant protein‐1 (MCP‐1), and C‐C motif chemokine ligand 5 (CCL5) tended to be higher in SHED than in BMSCs, although this varied depending on the strain.

Figure 2.

Comparison of the angiogenic and cell induction capacity of SHED and BMSCs in vitro. (a) Growth factor secretion levels of SHED and BMSCs using a multiplex assay. (b) The number of mouse RAW cells that migrated using a migration assay with culture medium containing SHED, culture medium containing BMSCs, and culture medium alone. Values shown are mean ± SD (*p < 0.05) (c) Representative example of migration assay microscope image. Scale bar = 500 μm. BMSC, bone marrow‐derived mesenchymal stem cell; CCL2, C‐C motif chemokine ligand 2; MCP‐1, monocyte Chemoattractant Protein‐1; SCF, stem cell factor; SD, standard deviation; SHED, stem cells from human exfoliated deciduous teeth; VEGF, vascular endothelial growth factor.

MCP‐1, also known as CCL2, has been shown to induce macrophage migration, a process essential for bone healing. Cytokine profiling revealed that one of the most significant difference between SHED and BMSCs was its high expression of MCP‐1, leading us to hypothesize that this characteristic might be advantageous for necrotic bone regeneration. To facilitate testing this hypothesis, in subsequent in vitro and in vivo experiments, we selected SHED #1, which highly expresses MCP‐1, and BMSC #1, which expresses it in a low manner. A migration assay using RAW cells (a macrophage cell line) was performed to investigate the cell induction capacity of SHED #1 and BMSC #1. The migration assay results showed that SHED significantly promoted macrophage migration compared with the negative control and BMSCs (p < 0.05) (Figure 2b).

3.2. In Vivo Study: Bone Regeneration‐Promoting Effect of SHED in Radiation Induced Osteonecrosis Model

In preliminary experiments, ReFit transplanted into 3‐mm‐diameter bone hole in the proximal tibia of rats was spontaneously repaired within 4 weeks (Supplementary Figure S1). However, in a radiation‐induced osteonecrosis model, bone hole repair was markedly delayed, and the defect remained unrepaired even 8 weeks after transplantation of Refit.

Prior to conducting cell‐seeded ReFit transplantation experiments, the attachment and viability of the seeded cells within the ReFit scaffolds were evaluated. The number of viable cells after 48 h of pre‐culture in ReFit was confirmed to be equivalent to the initial transplantation cell number for both SHED and BMSCs (Supplementary Figure S2).

We longitudinally analyzed bone hole repair after transplantation with SHED‐ or BMSCs‐seeded ReFit in a radiation‐induced osteonecrosis model using micro‐CT images, to compare the bone‐regeneration‐promoting effects of SHED and BMSCs on necrotic bone. Humane endpoints were predefined, and animals showing severe complications related to irradiation or surgery were to be euthanized. No animals met these criteria; however, one of ten animals in the SHED group and four of ten animals in the BMSCs group died due to an unintended anesthetic overdose attributable to a technical error during anesthesia for micro‐CT imaging. Reconstructed micro‐CT images showed that bone hole repair on the SHED‐transplanted side was markedly accelerated compared with the cell‐free scaffold‐transplanted side, whereas repairs on the BMSCs‐transplanted side showed no significant difference compared to that on the cell‐free scaffold‐transplanted side (Figure 3a,b). In the quantitative evaluation, when bone was defined as having a density of 300 mg HA/cm³ or higher, the bone repair rate in the SHED group was significantly higher than that in the cell‐free scaffold group at 4, 6, and 8 weeks (Figure 3c). In contrast, the repair rate in the BMSCs group was not significantly different from that in the cell‐free scaffold group (Figure 3d).

Figure 3.

The longitudinal micro‐CT observation after transplantation of scaffold material seeded with cells into one tibial bone hole and transplantation of scaffold material alone into the opposite tibial bone hole. (a) Representative micro‐CT 3D reconstruction images of tibial bone hole transplanted with SHED and tibial bone hole transplanted with only scaffold on the opposite side at 0, 2, 4, 6, and 8 weeks postoperatively. (b) Representative micro‐CT 3D reconstruction images of tibial bone hole transplanted with BMSCs and tibial bone hole transplanted with only scaffold on the opposite side at 0, 2, 4, 6, and 8 weeks postoperatively. (c, d) Graphs showing the bone repair rate of the tibial bone holes receiving cell‐seeded scaffold and scaffold‐only controls, respectively. (e) Graph showing the mean bone repair rate in tibial bone holes receiving SHED‐ or BMSCs‐seeded scaffolds, and (f) the mean bone repair rate in contralateral tibial bone holes receiving scaffold alone. Values shown are mean ± SD (*p < 0.05). BMSC, bone marrow‐derived mesenchymal stem cell; CT, computed tomography; SD, standard deviation.

Direct comparison between the SHED‐seeded scaffold group and the BMSCs‐seeded scaffold group demonstrated significantly greater bone repair rate in the SHED group from 4 weeks onward (Figure 3e). In contrast, comparison of the contralateral scaffold‐only defects between the SHED and BMSCs groups showed no significant difference in repair rate (Figure 3f).

3.3. In Vivo Study: Mechanism of Bone Regeneration, Promoting Effect of SHED in a Radiation‐Induced Osteonecrosis Model

Histological observations 8 weeks after cell transplantation revealed that the bone hole in the tibia transplanted with the cell‐free scaffold mainly consisted of fibrous tissue, with minimal bone formation observed on and around the scaffold. The cortical bone at the base of the bone hole exhibited necrosis, characterized by empty lacunae and by fibrosis in the nearby bone marrow. At the edge of the bone hole, the periosteum was thickened with observable cell proliferation; however, new bone formation was minimal (Figure 4A(a–d)). These changes were attributed to radiation exposure; in the absence of irradiation, no necrotic bone was observed in the surrounding tissue, and new bone formation within the bone marrow, vigorous periosteal proliferation, and bone repair were observed (Supplementary Figure S1).

Figure 4.

Histological analysis 8 weeks after cell transplantation to tibial bone hole. (A) Representative histological micrographs of the proximal tibia including the bone hole, from control, SHED‐transplanted, and BMSC‐transplanted groups. The first row shows the histological images from the group in which a cell‐free scaffold was implanted into the tibial bone hole. The second row shows the histological images from the group in which SHED was transplanted with scaffold, and the third row shows the histological images from the group in which BMSCs were transplanted with scaffold. The four columns from the left show the results of the H&E staining, and the rightmost column shows the results of CD31 staining. (a, f, k) Low magnification H&E images. Scale bar = 1000 μm. (b, g, l) H&E staining of the bottom of the bone hole to confirm the repair reaction of the medullary side. Scale bar = 100 μm. (c, h, m) H&E staining of the rim of the bone hole to confirm the repair reaction from the periosteum. Scale bar = 100 μm. (d, i, n) H&E staining of the scaffold to confirm the repair reaction of the scaffold itself. arrow: Scaffold. Scale bar = 50 μm. (e, j, o) CD31 staining of the bone hole surface to confirm neovascularization. Scale bar = 50 μm. (B) Graph showing analysis of blood vessel area and number on the bone hole, based on observation of CD31‐stained histological images. Values shown are mean ± SD (*p < 0.05). BMSC, bone marrow‐derived mesenchymal stem cell; H&E, hematoxylin and eosin; SHED, stem cells from human exfoliated deciduous teeth.

In the tibia transplanted with SHED, new bone and bone marrow formed within the bone hole, and new bone formation was observed on and around the scaffold. Necrotic bone, characterized by empty lacunae, persisted in the cortical bone at the base of the bone hole; however, new bone formed on the surface of the cortical endosteum. Significant periosteal tissue proliferation and new bone formation were observed at the edges of the bone‐hole. (Figure 4A(f–i)). In the BMSC‐transplanted tibia, the bone hole was primarily composed of fibrous tissue; however, bone formation was occasionally observed around the scaffold. The cortical bone at the base of the bone hole was primarily necrotic tissue characterized by empty lacunae with minimal new bone formation. Periosteal tissue proliferation and limited new bone formation were observed around the margins of the bone holes (Figure 4A(k–n)).

Angiogenesis, a key factor in regenerating necrotic bone, was evaluated by immunohistochemical staining using a CD31 antibody. The SHED‐transplanted specimens showed significantly greater vascularization than the scaffold‐only and BMSC‐transplanted specimens (Figure 4A(e, j, o)). Quantitative assessments of blood vessel number and area were performed specifically within the fibrous bone marrow region, excluding areas of newly formed bone within the defect. The SHED transplantation group exhibited a significantly larger blood vessel area compared with the other groups. While no significant difference in blood vessel number was observed between the SHED and BMSCs transplantation groups, the SHED group showed a significantly higher vessel count than the control group (Figure 4B).

Proximal tibia tissue containing transplanted and host cells was subjected to RNA‐Seq analysis 2 weeks after cell transplantation to clarify the mechanism by which SHED promote necrotic bone regeneration. A transcriptomic overview is shown in Figure 5. Figure 5a confirms robust species disambiguation across groups (mean ± SE; n = 3). In transplanted human cells (Figure 5b), SHED, compared with BMSCs, activate TNF‐α/NF‐κB and IL6–STAT3 programs, increases angiogenesis, and suppresses E2F/G2M‐related programs. In host rat tissue, type‐I interferon signaling, antigen processing/MHC‐I, and complement pathways were upregulated, accompanied by cell‐cycle downregulation; these changes were most pronounced in SHED compared with BMSCs and were similarly observed in SHED versus control (Figure 5c–k). Hallmark angiogenesis increased in SHED versus control (Figure 5d), and C8 endothelial signatures increased in SHED versus control (Figure 5g), consistent with SHED‐associated neovascularization. C8 chondrocyte signatures were increased in BMSC‐transplanted tissue compared with controls (Figure 5h), supporting a chondrogenic/endochondral differentiation tendency of BMSCs at a later stage. Effects were attenuated in BMSC versus controls for the IFN/antigen‐processing/complement pathways (Figure 5e,h,k), excluding a generic injury response and supporting SHED‐specific immune–vascular–matrix remodeling preceding bone regeneration.

Figure 5.

Integrated QC and gene set enrichment linking implanted human programs to host responses. (a) Composition of RNA‐seq reads by species per group (SHED, BMSC, Control): stacked bars showing the proportions of human, rat, ambiguous, and unmapped reads (mean ± SE; n = 3 per group). Ambiguous reads were 0.12%–0.58%; human misassignment in pure rat controls was 0.027%–0.091%. (b) Human (transplanted cells)—Hallmark GSEA for SHED vs, BMSC indicating activation of TNF‐α/NF‐κB and IL6–JAK–STAT3, induction of angiogenesis, and suppression of E2F/G2M in SHED. (c–e) Host (rat), Hallmark GSEA for SHED versus BMSC, SHED versus Control, and BMSC versus Control, respectively, showing type‐I interferon, antigen processing/MHC‐I, and complement upregulation with cell‐cycle downregulation; effects are strongest in SHED versus BMSC and reproduced in SHED versus Control. Panel (d) shows increased angiogenesis in SHED versus Control. (f–h) Host (rat), C8 (cell‐type signatures) for the same contrasts, highlighting enrichment of monocyte/macrophage/microglia/Kupffer programs under SHED; panel (g) shows increased endothelial signatures in SHED versus Control. Panel (h) shows increased chondrocyte signatures in BMSC versus Control, consistent with a chondrogenic/endochondral bias in BMSCs. (i–k) Host (rat), Reactome for the same contrasts confirming type‐I interferon, antigen processing/MHC‐I, and complement activation with ECM/vascular remodeling (e.g., integrin/collagen modules), and suppression of DNA replication/pre‐RC. Bubble plots encode NES (color; centered at 0), leading‐edge gene counts (size), and FDR tiers (outline: solid < 0.05; dashed 0.05–0.25). For each contrast, NES > 0 indicates higher expression in the first‐named group (e.g., SHED in “SHED vs. BMSC”, BMSC in “BMSC vs. Control”). BMSC, bone marrow‐derived mesenchymal stem cell; FDR, false discovery rate; GSEA, gene set enrichment analysis; NES, normalized enrichment score; SHED, stem cells from human exfoliated deciduous teeth.

Heatmap analysis was performed by classifying genes that showed significant increases in expression into functional categories related to angiogenesis, cell induction, and endochondral ossification. Visualization using SRPLOT revealed distinct gene expression patterns between the two transplantation groups. The SHED‐transplantation group exhibited a higher expression of genes associated with angiogenesis and cell induction, whereas the BMSC transplantation group showed preferential upregulation of genes involved in endochondral ossification. Notably, MX1, CCL5, and CCR5—genes suggested to be involved in periosteal stem cell regulation—were markedly upregulated in the SHED‐transplantation group (Figure 6).

Figure 6.

Heatmap analysis of genes associated with specific biological processes. Heatmaps showing relative expression levels of genes associated with significantly enriched GO biological process categories identified by DAVID. (a) Angiogenesis, (b) Cell induction, and (c) Endochondral ossification. Expression levels of each gene are shown for individual samples. Gene expression values were normalized per gene and displayed as Z‐scores, with the color scale ranging from −2 to +2. GO, Gene Ontology.

4. Discussion

The results of this study demonstrated that cell therapy using SHED exerted a superior bone regeneration–promoting effect compared with BMSCs in a highly refractory rat model of radiation‐induced osteonecrosis. Because these results are based on a comparison between a single SHED cell line and a single BMSC cell line, it cannot be concluded that SHED cell preparations from all donors are superior to BMSC cell preparations from all donors. However, these findings are consistent with previous reports showing that SHED possess greater bone repair capacity than BMSCs or DPSCs in mouse calvarial defect models [12]. Furthermore, the fact that SHED demonstrated bone repair‐promoting effects in a refractory radiation‐induced osteonecrosis model where BMSCs failed to do so highlights the potential usefulness of SHED.

The superior ability of SHED to promote necrotic bone regeneration is likely attributable to their high angiogenic potential (Figure 7). Marx et al. identified three key pathophysiological mechanisms of radiation‐induced osteonecrosis—hypoxia, cytopenia, and vasopenia—with vasopenia playing a central role in disease development [30]. Several in vitro and in vivo studies have demonstrated the superior angiogenic potential of SHED: Kim et al. showed that SHED function as perivascular supporting cells to promote vessel formation in vivo [14], Kato et al. demonstrated that SHED‐conditioned medium directly activates endothelial cells to promote all stages of angiogenesis through the secretion of soluble factors, without requiring direct cell contact [31], and Guo et al. confirmed a paracrine angiogenic role of SHED mediated by growth factors such as VEGF [13]. Taken together with our present findings, we propose that SHED potently induce angiogenesis predominantly through the paracrine secretion of growth factors and, potentially, exosomes, rather than through direct differentiation into vascular endothelial cells. Although BMSCs promote endochondral ossification, as demonstrated by RNA‐seq analysis, their inability to facilitate bone formation in this model may be attributable to insufficient angiogenesis [32, 33, 34, 35].

Macrophage induction by SHED may contribute to necrotic bone regeneration (Figure 7). Results from multiplex and migration assays indicated that SHED promoted macrophage recruitment through the secretion of CCL2, a chemokine critical for macrophage induction. Macrophages play a pivotal role in the early stages of bone repair during both endochondral and membranous ossification [36, 37, 38]. Recent studies have shown that macrophages accumulating at fracture sites recruit MX1⁺/αSMA⁺ periosteal stem cells, whose migration is dependent on CCL5, to sites of bone injury, thereby facilitating bone repair [39]. In this study, we also observed upregulation of CCL5, CCR5, and MX1 expression in SHED‐transplanted tissues, suggesting that macrophage/periosteal stem cell recruitment is one of the mechanisms through which SHED promote bone regeneration. This interpretation is further supported by our finding that periosteal thickening and cell proliferation were significantly increased at bone defect sites where SHED were transplanted.

The enrichment of type I interferon‐related biological processes was observed only in lesions transplanted with SHED, suggesting that this response is not simply a result of radiation exposure but may reflect a context‐dependent interaction between SHED and the irradiated bone microenvironment. One plausible explanation is that SHED may amplify radiation‐induced DNA damage and subsequent host innate immune signaling through mechanisms that are currently unknown [40]. Therefore, the observed interferon signature may represent a transient, antiviral‐like innate immune response associated with early tissue remodeling rather than a sustained inflammatory response [41]. However, the functional significance of this response in bone regeneration remains unclear, and further mechanistic studies are required to clarify the causal role of nucleic acid‐sensing pathways in SHED‐mediated tissue repair.

Regarding the bulk RNA‐seq results, one sample in the BMSCs transplantation group exhibited marked variability in gene expression compared with the remaining samples. Since all transplanted BMSCs were derived from a single donor (BMSC #1; bone marrow obtained from the femoral head of a 60‐year‐old Caucasian male), donor‐to‐donor variability is unlikely to account for this observation. Rather, the variability is more likely attributable to host‐derived factors. One plausible explanation is heterogeneity in radiation exposure: although irradiation was performed under general anesthesia, subtle differences in hindlimb positioning or inadvertent displacement of lead shielding plates may have resulted in uneven radiation doses across animals. Consistent with this possibility, the outlier animal exhibited markedly attenuated hair loss — a hallmark of local radiation exposure — compared with the other irradiated animals. Nevertheless, since other biological and technical confounders cannot be excluded, these findings should be interpreted with caution.

An unexpected finding was that although SHED possess mesenchymal stem cell properties and can differentiate into osteoblasts, chondrocytes, vascular endothelial cells, and adipocytes in vitro, they rarely differentiate into osteoblasts or vascular endothelial cells after transplantion into osteonecrotic lesions. Histological examination of SHED nuclear membranes stained with an anti‐lamin antibody revealed that the cells forming new blood vessels and newly formed bone did not correspond to anti‐lamin‐positive cells; instead, anti‐lamin‐positive SHED were localized only on or around the scaffold (Supplementary Figure S3). These observations suggest that necrotic bone regeneration by SHED is mediated not by direct differentiation but rather through angiogenesis and stem cell homing induced by growth factors and exosomes.

This study has some limitations. First, although radiation‐induced osteonecrosis and drug‐ or alcohol‐induced osteonecrosis share ischemic features, radiation‐induced osteonecrosis also involves additional nonischemic mechanisms, including apoptosis, necrosis, and DNA damage. Therefore, the pathogenesis and molecular mechanisms identified in this study may not be generalizable to all forms of osteonecrosis. This model was selected because in animal models of osteonecrosis caused by cauterization of the bone nutritional vessels, angiogenesis from the collateral circulation occurs after several weeks, whereas steroid‐induced models show poor reproducibility and high variability. If angiogenesis and tissue regeneration can be achieved in a more severe radiation‐induced osteonecrosis model, similar therapeutic approaches may also be effective in treating drug‐ or alcohol‐induced osteonecrosis. Second, this study employed a xenotransplantation model. Although the clinical application of SHED is intended for allogeneic transplantation, this study used immunodeficient rats for xenotransplantation. Similar to BMSCs, SHED exhibit low immunogenicity and may suppress immune responses, making allogeneic transplantation feasible. However, the in vivo dynamics of SHED may differ between allografts and xenografts, and further studies are necessary to confirm whether comparable effects can be achieved with allografts. In this study, SHED were present at the transplantation site 8 weeks after implantation; therefore, future studies should include longer observation periods to determine the duration of SHED persistence in vivo. Third, this study used a metaphyseal tibial bone‐defect model to reproduce the vasculopenia and regenerative failure characteristic of osteonecrotic bone; however, this model does not reproduce the subchondral/epiphyseal location of clinical osteonecrosis or the mechanical loading conditions associated with articular surface collapse. This study should therefore be regarded as a pilot, proof‐of‐concept investigation of the angiogenic and regenerative potential of SHED rather than a direct model of joint‐preserving therapy. Future studies should evaluate SHED transplantation in models that reproduce subchondral osteonecrosis and incipient joint collapse (e.g., epiphyseal or femoral head osteonecrosis models) to confirm that the regenerative effects observed here translate into preservation of the articular surface and prevention of joint collapse.

5. Conclusion

This study demonstrated that SHED promote necrotic bone regeneration predominantly through paracrine‐mediated angiogenesis and stem cell homing, rather than through direct differentiation. Given their accessibility and potency, SHED are a promising candidate for cell‐based therapy of refractory osteonecrosis. Further studies are warranted to validate the efficacy of SHED in human clinical settings.

Author Contributions

Toshifumi Murakami contributed to research design, data acquisition, analysis and interpretation of data, and drafting of the paper; Noritaka Fukuda contributed to data acquisition, analysis and interpretation of data; Yasuyuki Mitani contributed to research design and interpretation of data; Ryo Fujita contributed to research design, data acquisition, analysis and interpretation of data, and drafting of the paper; Hideyuki Kobayashi, contributed to data acquisition, analysis and interpretation of data; M. Alaa Terkawi contributed to data acquisition, analysis and interpretation of data; Kazuo Tomizawa contributed to analysis and interpretation of data; Norimasa Iwasaki contributed to interpretation of data; Masahiko Takahata contributed to research design, data acquisition, analysis and interpretation of data, and drafting and critically revising the paper. All authors have read and approved the final submitted manuscript.

Conflicts of Interest

Noritaka Fukuda and Yasuyuki Mitani are an employee and a board director at S‐Quatre Corporation, respectively. S‐Quatre provided the SHED used in this study. This research was conducted as a joint research project and received partial funding from the S‐Quatre.

Supporting information

Figure S1: In a non‐irradiated rat tibia, the bone hole is repaired 3–4 weeks after creation. (a–d) Representative micro‐CT 3D reconstruction images of the tibial bone hole from a non‐irradiated rat implanted with only a scaffold. (a) 10 days after surgery, (b) 3 weeks after surgery, (c) 4 weeks after surgery, (d) 6 weeks after surgery. (e, f) Histological micrographs of the tibial bone hole from a non‐irradiated rat implanted with only a scaffold 10 days after surgery. Scale bar for (e) = 1000 μm, scale bar for (f) = 200 μm. Unlike the irradiated rat tibia, the tibial bone hole is repaired by vigorous bone formation from the bone marrow side. CT, computed tomography.

Figure S2: Number of viable cells after 48 hours of pre‐culture following cell seeding in ReFit. To quantify viable cells within the ReFit scaffolds, cells were isolated via collagenase treatment, stained with calcein‐AM, and counted using a hemocytometer under a fluorescence microscope.

Figure S3: Histological micrographs 8 weeks after cell transplantation. (a, c) H&E. Scale bar = 100 μm. (b, d) Anti‐lamin staining. Scale bar = 100 μm. Arrows indicate cells stained positive for anti‐lamin antibody. (a, b) In the newly formed bone, osteoblasts, osteocytes, and vascular endothelial cells were negative for anti‐lamin antibody, indicating that new bone was formed by host rat cells. (c, d) Numerous cells stained positively with anti‐lamin antibody were observed around the scaffold, indicating the presence of SHED. H&E, hematoxylin & eosin; SHED, stem cells from human exfoliated deciduous teeth.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.