Single-cell characterization of monolayer cultured human dental pulp stem cells with enhanced differentiation capacity

Human dental pulp stem cells (hDPSCs) are easily obtained multipotent cells, however, their potential value in regenerative medicine is hindered by the phenotypic and functional changes after conventional monolayer expansion. Here, we employed single-cell RNA sequencing (scRNA-seq) to comprehensively study the transcriptional difference between the freshly isolated and monolayer cultured DPSCs. The cell cluster analysis based on our scRNA-seq data showed that monolayer culture resulted in a significant cellular composition switch compared to the freshly isolated DPSCs. However, one subpopulation, characterized as MCAM(+)JAG(+)PDGFRA(−), maintained the most transcriptional characteristics compared to their freshly isolated counterparts. Notably, immunofluorescent staining revealed that the MCAM(+)JAG(+)PDGFRA(−) hDPSCs uniquely located in the perivascular region of human dental pulp tissue. Flow-cytometry analysis confirmed that their proportion remained relatively stable (~2%) regardless of physiological senescence or dental caries. Consistent with the annotation of scRNA-seq data, MCAM(+)JAG(+)PDGFRA(−) hDPSCs showed higher proliferation capacity and enhanced in vitro multilineage differentiation potentials (osteogenic, chondrogenic and adipogenic) compared with their counterparts PDGFRA(+) subpopulation. Furthermore, the MCAM(+)JAG(+)PDGFRA(−) hDPSCs showed enhanced bone tissue formation and adipose tissue formation after 4-week subcutaneous implantation in nude mice. Taken together, our study for the first time revealed the cellular composition switch of monolayer cultured hDPSCs compared to the freshly isolated hDPSCs. After in vitro expansion, the MCAM(+)JAG(+)PDGFRA(−) subpopulation resembled the most transcriptional characteristics of fresh hDPSCs which may be beneficial for further tissue regeneration applications.


INTRODUCTION
Human mesenchymal stem cells (MSCs) are multipotent cells, which exhibit a specific cell surface marker spectrum, and have multilineage differentiation potentials. 1 MSC-like populations have been obtained and harvested from a variety of organs and tissues. 2 Amongst them, dental pulp-derived MSCs (DPSCs), have been considered as one of the most promising cell source for maxillofacial tissue repair and regeneration owing to its abundance and ease of cell acquisition. [3][4][5][6][7][8] Similar with other tissuespecific MSCs, the dental pulp-derived MSCs contain several subpopulations with different biological functions, which leads to a great heterogeneity. For instance, the heterogenous DPSC subpopulations have demonstrated differences in proliferation and differentiation potentials and proximally only two thirds DPSC subpopulations were able to form ectopic dentin in vivo, indicating their different regenerative capabilities. 3,9 In order to identify distinct subsets of DPSC populations, researchers have optimized DPSC isolation based on a series of cell surface markers. DPSC is known to express MSC surface markers, such as CD29, CD44, CD59, CD73, CD90, and CD146, but no hematopoietic stem cell markers (e.g., CD14, CD34, CD45, and CD11b). 10 Previous studies showed that DPSC subpopulations with different combinations of surface markers exhibited different tendency towards odontogenic, adipogenic and neurogenic commitment. 11,12 In addition, DPSCs reside in the perivascular niche and the perineural niche in the postnatal dental pulp tissues. 13,14 The intercellular communications between endothelial and stem cells within the perivascular niche influenced the self-renewing capacity of DPSCs, 15,16 Furthermore, the superior neural and vascular properties of DPSCs opens their promising potentials for cell-based therapy of neurodegenerative and oral diseases.
Conventionally, DPSCs were isolated from the extracted tooth and expanded in vitro in monolayer prior to further cell-based applications. 3 However, more and more evidences suggest that such monolayer expansion leads to an alteration of cell phenotype, evidenced 17 by the CD44 expression and differentially expressed of CD73, CD106, CD146, and CD271. Furthermore, a recent study revealed that the DPSCs after monolayer expansion tend to lose multipotential properties, 17,18 and continue with an unpredictable dynamic changes and functional variations of subpopulations. 19,20 Several articles have reported the comprehensive view of the heterogeneity within the human dental pulp environment using single-cell RNA sequencing (scRNA-seq). [21][22][23] However, up to date, the atlas of the monolayer expanded human dental pulp stem cells (hDPSCs) still remains unclear.
In view of this, the current study was aimed to comprehensively investigate the atlas of hDPSCs after 10-day monolayer expansion. We performed single-cell RNA sequencing of both freshly isolated and monolayer expanded DPSCs. Our data confirmed that the monolayer expansion induced an significant cellular composition shift, while only one subpopulation characterized as MCAM(+)JAG (+)PDGFRA(−) most resembles their fresh isolated state, resulting in enhanced differentiation potentials and elevated ectopic bone forming capacity. Combined, we believe these findings can improve reliability, reproducibility, and robustness of the hDPSCs for future applications in cellular therapy and tissue engineering.

Cellular composition in human dental pulps
To explore the cellular composition of human dental pulp tissues, we first used scRNA-seq to profile 7000 isolated cells from the fresh erythrocyte-depleted dental pulp tissues (Fig. 1a). The major cell populations were defined using the SingleR computational method 24 to correlate single-cell gene expression with reference cell marker datasets of pure cell types recorded in the human cell atlas. 25 As shown in Fig. 1b, the unbiased clustering showed 14 clusters, in which the dental pulp mesenchymal lineage (cluster 0, 1, 2, 3, 7, and 8) were clearly separated from the immune cells (cluster 4, 9, and 12) and endothelial cells (cluster 5, 6, and 10). Gene ontology analysis revealed that cluster 0 highly expressed genes associated with cell differentiation and skeletal system development, while cluster 1 expressed genes associated with cell-cell signaling and extracellular organization. In contrast, the genes enriched in cluster 7 and 8 were strongly related to neuron projection morphogenesis and ossification ( Supplementary  Fig. 1a).
We further assessed the uniquely-enriched marker genes expressed in the dental pulp mesenchymal lineage (Fig. 1c). Specifically, VCAN, STAB2, SOX5, and GLI3 were highly expressed in cluster 2, which was strong related to blood vessel development and odontogenesis shown by gene ontology analysis. In addition, RGS5, NOTCH3, and PDGFRB, the perivascular cell markers, were exclusively expressed in cluster 3, while POSTN and IGFBP5, markers of the peri-odontoblastic layers, 21 were highly expressed in cluster 7 ( Fig. 1c and Supplementary Fig. 1b).
Interestingly, we observed that cells in cluster 4, 6, and 7 formed a successive cell-cycle loop based on RNA velocity analysis ( Supplementary Fig. 3b). Since the cell-cycle might influence gene expression, we further assigned each cell a cell-cycle phase according to a set of cell-cycle genes. Our data showed that Cluster 3, 4, 6, and 7 consisted of dividing cells assigned to the G2/ M and S cell-cycle phases (Fig. 1e). For instance, Cluster 4 was enriched in S phases marker genes CDCA7 and MCM3 (Supplementary Fig. 4a); while Cluster 6 exhibited characteristic genes in the G2/M phase, such as BIRC5 and CCNB2 ( Supplementary Fig. 4b).
GO enrichment assay further revealed that cells in different clusters showed different biological properties. Consistent with previously mentioned RNA velocity results, GO enrichment assay confirmed that cells in clusters 4, 6, and 7 highly expressed genes related to DNA replication, chromosome segregation, and cell-cycle process (Supplementary Fig. 4c). In comparison, cells in clusters 0, 1, and 2 highly expressed genes related to ossification, evidenced by high expressions of angiogenesis related genes such as CXCL12, ANGPT1, IGFBP6, and FGF7 in cluster 0 and 1 cells and mineralized matrix organization related genes such as COL1A2, COL5A3, ITGA11, and SPARC in cluster 1 and 2 ( Supplementary Fig.  5). Notably, cells in cluster 5 highly expressed perivascular cells or pericyte markers, 26 such as NG2 (CSPG4), PDGFRB, NOTCH3, and ACTA2 ( Supplementary Fig. 6a). The enriched GO terms were associated with blood vessel development and cell differentiation, suggesting that cluster 5 might be perivascular niche cells with multilineage potentials in dental pulps ( Supplementary Fig. 6b).
Single-cell RNA sequencing reveals a cellular composition switch upon monolayer expansion Next, we integrated the scRNA-seq profiles from fresh dental pulp and monolayer cultured hDPSCs and our data confirmed that the monolayer culture leads to a significant cellular composition alteration compared to the freshly isolated DPSCs (Fig. 2a). Cell composition across both datasets was visualized and 13 clusters was yielded, in which the cluster 4 was the most closely overlapping zone, suggesting that cluster 4 cells after monolayer most resembled their freshly isolated counterparts ( Fig. 2b and Supplementary Fig. 7a).
We further assessed the differentially expressed genes enriched specifically in the cluster 4 cells from both the freshly isolated and cultured DPSCs. Our data showed that cluster 4 cells from both conditions expressed like NOTCH3, THY1, MCAM, JAG1, and ITGA1, the classic stem cell marker genes. However, TUBA1B, S100A10 and DAB2 were highly expressed in cultured samples whereas RGS6, RGS5, EBF2, and POUF2 were highly expressed in fresh samples (Fig. 2c, Supplementary Fig. 7b, c). Furthermore, gene expression dynamics along the pseudotime revealed that after in vitro culture, JAG1 was stably expressed and barely changed, while gene like RGS5 and S100A10 was down and upregulated, respectively, indicating the transcriptional changes still occurred during in vitro culture (Fig. 2c).
In addition, GO analysis on genes highly expressed in cluster 4 cells showed that biological process like cell morphogenesis involved in cell differentiation was enriched ( Supplementary Fig.  8a). In consistence with the GO analysis result, we also found that the cluster 4 cells express classic MSC markers, like ENG, NT5E, THY1, LEPR, and CXCL12, but no GLI1. Furthermore, the expression of LEPR significantly declined during in vitro culture, while the expression of ENG, NT5E, and CXCL12 showed significantly upregulated after monolayer expansion ( Supplementary Fig. 10).
The location of MCAM(+)JAG1(+)PDGFRA(−) cells in human dental pulp tissue and their resistance to aging and inflammation conditions Our next question was to identify the location of the MCAM(+) JAG1(+)PDGFRA(−) in the native healthy human dental pulp tissue. The immunofluorescent staining showed that cells positive in MCAM or JAG1 mainly located near blood vessels, as confirmed by co-staining with the vascular endothelial cell marker CD31 (Fig. 3a), and the MCAM and JAG1 simultaneously double-stained cells were mainly observed in the perivascular stem cell niches (Fig. 3b). On the other hand, PDGFRA was predominant within the odontoblastic layer and stroma, whereas the MCAM(+)JAG1(+) PDGFRA(−) cells were uniquely located in the perivascular region of human dental pulp tissue (Fig. 3c).
To further explore the osteogenic and adipogenic potentials in vivo, cells from both groups were seeded on calcium phosphate scaffolds or collagen membrane, respectively, and subcutaneously implanted in nude mice for 4 weeks. MicroCT analysis revealed that cell from both MCAM(+)JAG1(+)PDGFRA(−) and PDGFRA(+) cells induced significant higher new bone formation compared to the bare scaffolds, although no statistical difference was detected between each other (Supplementary Fig. 9). Interestingly, histological analysis showed abundant bone marrow formation within the PDGFRA(+) explants, compared to the newly-formed osteoid-like structure observed from the MCAM(+)JAG1(+) PDGFRA(−) explants (Fig. 8a). Furthermore, the immunohistochemical staining human specific mitochondria (hMitochondria) and osteocalcin (hOCN) staining revealed that the implanted MCAM(+)JAG1(+)PDGFRA(−) cells actively contributed to the newly-formed bone tissue. In contrast, the majority of the newlyformed bone from the PDGFRA(-) cells was negative for hMitochondria and hOCN staining, suggesting that these cells did not directly contributed to the new bone formation (Fig. 8b, c). Similar trend was observed from histological samples for in vivo adipose tissue formation. After 4 weeks, HE staining illustrated that the MCAM(+)JAG1(+)PDGFRA(−) cells induced more adipose tissue formation compared to the PDGFRA(+) group (Fig. 7a). Furthermore, the human mitochondria-positive cells were both observed in the newly-formed adipose tissues (Fig. 7b)   DPSCs are traditionally expanded in vitro prior to their basic and clinical application. However, long-term culture results in continuous changes to these cells and it is difficult to predict the dynamic changes and functional variations of subpopulations after prolonged culture duration or exposure to external factors. 19,20 After extended in vitro culture, cells in cluster 0, 1, and 2 all displayed elevated expression of genes associated with ossification. In addition, aging was one of the enriched biological processes in cluster 0. These results are in line with the previous study, that is, MSCs undergo replicative senescence, tend to lose multipotential properties and differentiate to osteogenic cell lineage. 17 Studies have also reported that the differentiation capability of hDPSCs was gradually reduced following cell cultivation. 18 According to our data, though monolayer culture decreased most cell types of hDPSCs and induced an apparent phenotype switch, MCAM(+)JAG1(+)PDGFRA(−) subpopulation displayed minimal transcriptional changes and functional alterations. The attenuated potency of monolayer cultured hDPSCs may partly due to the limited fraction of cells that maintain their cellular identity and multilineage potentials. However, although the MCAM(+)JAG1(+)PDGFRA(−) hDPSCs exhibited higher multilineage differentiation abilities, the proportion of this subpopulation decreased in the long-term expansion in vitro. The gradual decline of this subpopulation may be responsible for the reduced biological function of in vitro cultured hDPSCs.
The maintenance of the cellular identity and its specific stem cell pool requires collaboration and molecular crosstalk among the heterogeneous subpopulations. Stem cells typically reside in their respective microenvironments or niches that provide signaling cues to regulate their fates and maintain tissue homeostasis. 27 Cell interactions form networks of cytokines and growth factors. Various cell types and their secreted key factors, especially endothelial cell-secreted factors in the vascular niche, are necessary to promote self-renewal of stem cells. 28 Our cultured data revealed that cycling cells in cluster 4, 6, and 7 support the stem cell reservoir of cluster 0-2 and 5, which might play a role in maintain the self-renewal and heterogenous of hDPSCs. In addition, cells in cluster 0 express several cytokines, 29 like CXCL12, KITLG, and ANGPT1. These secreted factors may also contribute to the functional maintenance of the cultured hDPSC pool.
Compared to the recently published studies about scRNA-seq of human dental pulps, we found several markers they identified were also expressed in the MCAM(+)JAG1(+)PDGFRA(−) subpopulation. Pagella et al. 23 demonstrated that MSCs in dental pulps were characterized by the higher expression of NOTCH3, MYH11, THY1, and ID4. Consistent with this study, our data also showed that the PDGFRA(−)MCAM(+)JAG1(+) subpopulation of hDPSC (cluster 4) expressed these markers (Supplementary Fig.  11a). Furthermore, PDGFRA(−)MCAM(+)JAG1(+) subpopulation of DPSC (cluster 4) also expressed the other characteristic markers of DPSCs such as BGN, SOX4 and JAG1, which was consistent with the study of Yin et al. 22 (Supplementary Fig. 11b). Taken together, our data was consistent with the previous studies using DPSCs obtained from different patients, which illustrated the reproducibility of our sequencing data.
In our study, the identified markers positively labeled the perivascular area, which further confirmed that the perivascular area was the niche for most types of MSCs. 26 PDGFRA is highly expressed in many adult fibroblasts and has been assigned a mesenchymal progenitor identity, such as bone marrow resident skeletal progenitor cell population, 30 myofibroblast progenitors, 31,32 and fibro/adipogenic progenitors. 33,34 PDGFRA functions as a gatekeeper of adipogenesis and distinct PDGFRA-marked perivascular cell subpopulations possess different osteogenic potentials within adipose tissue. 35 We observed that PDGFRA(+) expressing cells were also present in perivascular locations and displayed weaker multiple differentiation potentials. In addition, the number of PDGFRA(+) cells increased dramatically after in vitro expansion but PDGFRA was rarely detected in freshly isolated dental pulp cells. It indicated that the cell surface marker expression profiles differ from their in vivo environment and the upregulation expression of PDGFRA might infer the cellular phenotypic switch. Some cells in the cell-cycle phase also expressed PDGFRA, consistent with the previous observation that PDGFRA(+) stromal cells proliferate faster than their PDGFRA(−) counterparts. 34 MCAM/CD146 is primarily used as a biomarker of endothelial lineages and plays important roles in cell growth, migration, and angiogenesis. 36 MCAM is closely associated with the multipotency, proliferation, and stemness of MSCs. 37-39 MCAM may regulate cellular senescence of human umbilical cord bloodderived MSCs. 40 Cells with high levels of MCAM were identified as human pericytes that ensheathed the vasculature of multiple organs. Presently, MCAM and PDGFRA were expressed differently in two separate partitions. Although they are both thought to be markers for multipotent progenitor cells, the trilineage differentiation capabilities of MCAM(+) JAG1(+)PDGFRA(−) cells were stronger than that of PDGFRA(+) cells in vitro. A similar result was reported previously, in which MSCs expressing distinct levels of MCAM and PDGFRA differed in their establishment in the supporting niche and in the functional maintenance of HSCs. 41 Other marker genes, like NOTCH3, THY1 and ITGA1 also expressed in MCAM(+) JAG1(+)PDGFRA(−) hDPSCs. Our in vivo osteogenic findings also showed similar results with Yasui et al. 42 who found that THY1 (high+) hDPSCs promote new bone formation using lineage tracing.
Like other tissues, dental pulp undergoes age-associated changes, such as decreases in volume and vascularization, and modifications in the biological characteristics of DPSCs. Differentiation potentials and proliferative capabilities of DPSCs were reportedly decreased in aged donors. 43 We observed a slight reduced MCAM(+)JAG(+)PDGFRA(−) cells in the aged group but this change was not obvious. This result did not contradict previous conclusion of an age-dependent decline of the stem cell niche in dental pulp. 43 Because the total number of stem cells may decreased but the percentage of MCAM(+)JAG(+)PDGFRA (−) cells stayed stable. Inflammation is a common cause of the destruction of the dentin-dental pulp complex. Inflammation can activate and promote DPSC proliferation and differentiation to odontoblasts or fibroblasts in response to injury. 44,45 The present observations revealed insignificantly increased numbers of MCAM(+)JAG(+)PDGFRA(−) cells, suggesting that the dental caries may evoke the proliferation and differentiation of different subpopulations of hDPSCs. Due to the heterogeneity of hDPSCs, other subpopulations may take a part in this process. Other reason for the lack of significant difference between normal and dental caries groups may result from the asymmetric division of stem cells. Inflammation stimulates stem cells to produce two daughter cells, one is similar to its mother cell and maintains the original stem cell pool, the other differentiate into downstream cell types. However, detailed mechanisms and whether aging and inflammation affect the proliferation and differentiation potentials of MCAM(+)JAG(+)PDGFRA(−) cells need further explored in the future.
In summary, our findings revealed the heterogenous cell subpopulations in fresh and cultured hDPSCs and a cellular composition switch during in vitro expansion. Interestingly, one subpopulation characterized as MCAM(+)JAG(+)PDGFRA(−) maintained the transcriptional characteristics as the fresh isolated DPSCs. Specifically, the MCAM(+)JAG(+)PDGFRA(−) subpopulation located in the perivascular niche and the expression level remained stable in dental pulps regardless of aging or inflammation conditions. Furthermore, the MCAM(+)JAG(+) PDGFRA(−) subpopulation appeared higher proliferation and trilineage differentiation capabilities. Taken together, our data revealed that in vitro monolayer culture significantly altered the hDPSCs transcriptional characteristics and biological performance, which may affect their effective use in regenerative medicine.

Primary cell cultures
The entire protocols were conducted in accordance with the Ethics Committee of the School of Stomatology, Wuhan University. The isolation of human dental pulp cells (hDPSCs) was performed as our previous study. 46 Briefly, immediately after patients' healthy third molars (aged [15][16][17][18][19][20][21][22][23][24][25][26] were extracted, the samples were transported to ice-cold phosphate-buffered saline (PBS, Hyclone, Logan, UT) with 100 U·mL -1 penicillin/ streptomycin (PS, Hyclone) and rinsed repeatedly. The dental pulp tissues from five individual patients were then softly removed, pooled together, cut into tiny pieces and digested by 3 mg·mL -1 type I collagenase and 4 mg·mL -1 Dispase II (Roche) for 1 h at 37°C in water batch, aided by shaking every 10 min. After enzymatic dissociation, cell suspension was passed through a 40 μm Falcon Cell Strainers. The freshly obtained cell suspension was then centrifuged for 5 min at 1000 r·min -1 . The cell pellet was then resuspended into alpha Dulbecco's modified Eagle's medium (α-MEM, Hyclone, Logan, UT) with 20% fetal bovine serum (FBS, Gibco) and 100 U·mL -1 PS, at 37°C and 5% CO 2 in 25 cm 2 flasks. Cells started to expanded; and after 10 days, the cultured cells were trypsinized, rinsed and prepared for the single-cell suspension used for scRNA-seq or FACS cell sorting.
Single-cell suspension preparation for scRNA-seq The freshly isolated or the cultured single-cell suspension was rinsed twice with Dulbecco's phosphate-buffered saline (DPBS, Hyclone, Logan, UT) free of magnesium and calcium. The pellets were resuspended in DPBS + 0.04% bovine serum albumin (BSA, Sigma-Aldrich) and finally filtered by 40-μm cell strainers to avoid cell aggregation. The total number of cells needed to be supplied is at least 10 times than target cell numbers. Cell concentration and viability were confirmed using a LUNA automated cell counter (Logos Biosystems). The sample consisted of >90% viable cells and the cell concentration is required to be controlled in the range of 700-1200 cells per μL. The cell debris rate and large cell clumping rate were less than 5%.
10× Genomics scRNA-seq library generation and sequencing Reverse transcription and library construction were carried out following the Chromium TM Single-Cell 3ʹ Reagent Kits v2 protocol (10× Genomics) according to the manufacturer's guides. Four steps are introduced to perform: Gel Bead-In-Emulsions (GEMs) generation and barcoding, cleanup and cDNA amplification, library construction by enzymatic fragmenting and size selection, sequencing prepared libraries. Qubit 3.0 was used preliminarily for quantification after the library was constructed and the library was diluted to 1 ng·μL -1 . Agilent 2100 was then used to detect the insert size of the library. If the insert size met expectation, the StepOnePlus real-time PCR System fluorescence quantitative PCR instrument was performed to accurately quantify the effective concentration (>10 nmol·L -1 ) of the library. The qualified libraries were sequenced on Illumina platform.
The original offline sequences (raw reads) generated from Illumina platform is processed to get high-quality sequences (clean reads) through removing low-quality sequences and connector contamination. All subsequent analysis was based on the clean reads. We have uploaded our original sequencing data to GSA-human data base (no. HRA001320) Single-cell RNA-sequencing analysis Initial read alignment and quality control. scRNA-seq samples were processed and aligned to Ensembl genome GRCh38 using the Cell-Ranger 3.0.0 pipeline. The Cell-Ranger output a genebarcode matrix which was loaded into Seurat v3.0 for scRNA-seq data analysis. Cells with high mitochondrial features (>10% of total mapped reads) were excluded from the analysis to remove the influence of dead cells. A small number of cells with unusually high or low numbers of mapped reads were also removed from the dataset, as these outliers could be duplets (expression of at most 7500 genes for in vitro cultured data and 5000 genes for uncultured data) or poorly-captured cells (expression of at least 1000 genes for in vitro cultured data and 500 genes for uncultured data). After these filters 8963 in vitro cultured cells/6970 uncultured cells remained (out of 10 116/7000 initially provided by the Cell-Ranger pipeline).
Dimensional reduction. Highly variable genes were identified and subjected to dimensional reduction by principal component analysis. The Elbow method implemented in Seurat was used to select a ranking of principal components based on the percentage of variance explained by each one. We can observe an 'elbow' around PC 10-12, suggesting that majority of the true signal is captured in the first 12 PCs. The Elbow plot of in vivo data indicated that PC 8 would be fine.
Identification of clusters. The identification of biologically relevant communities was performed by means of a graph-based clustering in Seurat through the construction of a KNN graph and posterior cell clusterization using the Louvain algorithm (resolution at 0.5). Cell clusters were visualized using uniform manifold approximation and projection plots with previously selected significant components as an input.
Immunofluorescence staining Human third molar samples were fixed in 4% paraformaldehyde (PFA) at 4°C overnight and then washed twice with distilled water. After decalcified in 10% ethylenediaminetetraacetic acid (EDTA) for more than 3 month, the demineralized tooth samples were dehydrated in an alcohol gradient and embedded in paraffin. Sections of 5 μm thickness were de-waxed and rehydrated followed by antigens retrieval with gastric enzyme. The slides were blocked by 2% BSA and incubated with primary antibodies at 4°C overnight. The following primary antibodies were diluted in PBS and used: mouse anti-MCAM (1:100) (Abcam ab85763), rabbit anti-JAG1 (1:100) (Abcam ab78479), and rabbit anti-PDGFRa (1:500) (CST AF1062). After washing in PBS, sections were incubated with secondary antibodies for 1 h at room temperature. The following secondary antibodies were diluted in PBS and used: anti-rabbit Alexa Fluor 488 (antgene ANT024), anti-mouse Alexa Fluor 594 (antgene ANT029) and anti-rabbit Alexa Fluor 647 (antgene ANT032). Afterward, sections were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) to label nuclei. Stained samples were imaged with a ZEISS LSM 710 confocal laserscanning microscope and analyzed by ZEISS ZEN microscope software.
Flow-cytometry analysis and FACS-based cell sorting Adequate cultured primary hDPSCs were collected, washed, filtered and finally resuspended in 2% FBS/PBS for further cell sorting.
The method of obtaining fresh human dental pulp cell suspensions were described above. After filtration and centrifugation, the cell pellets were resuspended in red blood cell lysis solution (Milteny) for 15 min at RT and then centrifugated at 300 × g for 5 min. Cells were washed twice with 2% FBS/PBS and filtered to obtain fresh cell suspensions for further flow-cytometry analysis.
Cell suspensions (10 5 for analysis and 10 7 for sorting) were then performed to multi-color immunofluorescence staining. Each of the following conjugated anti-human monoclonal antibodies were incubated with the cell suspensions at 37°C for 30 min in the dark: MCAM/PE (BD 550315), JAG1/APC (R&D FAB1277A) and PDGFRa/ BB700 (BD 746023). The data were analyzed and sorted by the FACS Calibur flow cytometer (Becton Dickinson, Franklin Lakes, NJ). PBS tube was used as blank control and single stained tubes as compensation controls. Data were analyzed using FlowJo TM software (Ashland, OR).
In vivo subcutaneous implantation Cells were trypsinized, counted and resuspended in PBS. Afterwards, 20 μL of cell suspension were drop seeded onto the tricalcium phosphate scaffolds (RB-a-SK-005G, Shanghai, China). The loading dosage of cells (50,000 cells/mm 3 ) were normalized to the open volume (28 mm 3 ) and the total volume space (42 mm 3 ) of the scaffolds, which were determined from the microCT-based image analysis in our preliminary experiment. After 30 min of incubation at 37°C, 5% CO 2 , and 95% relative humidity, 2 mL of α-MEM was added. The seeded scaffolds were incubated overnight to allow cell attachment before implanted in BALB/c nude mice. Cell-laden scaffolds were randomly implanted subcutaneously in the shoulder and the back area of the cervical region of animals. After 4 weeks of implantation, the mice were sacrificed and the implants were collected for further analyses.
For ectopic adipose tissue formation, cells were cultured in the adipogenic media (Cyagen Biosciences Inc, Sunnyvale, CA) for 1 week. 2 × 10 6 cells from each groups were typsinized, resuspended and mixed with the collagen membrane scaffold material. The seeded scaffolds were placed in a shaker at 37°C for 2 h and then centrifuged at 150 × g for 5 min to allow adherence. After implanted into the subcutaneous pockets of nude mice for 4 weeks, the tissues were harvested and analyzed by staining with H&E.
Microfocus computed tomography (microCT) and 3D image anasis Samples are scanned using the SkyScan 1276 system (Bruker) microCT. The operation condition was 60 kV, 200 μA and a 0.25 mm filter of aluminum. 3D parametric analysis was performed within the region of interest (ROI) using CTAn software (Bruker). We calculated the bone volume in the total explant space as defined by the ROI as well as the inside scaffold volume only.
For adipogenic differentiation, cells were cultured in the adipogenic medium (Cyagen Biosciences Inc, Sunnyvale, CA). Lipid droplets were then stained with Oil Red O stain kit (Solarbio) according to the recommended protocol. Briefly, cells were fixed in ORO Fixative solution for 30 min and washed twice with water. Followed by immersion in 60% isopropyl alcohol, cells were covered by fresh prepared ORO Stain solution for 30 min. Washing 5 times to remove excessive staining solution and then photographed.
For chondrogenic differentiation, cell pellet culture system was used to access the chondrogenic differentiation potentials. Chondrogenic media (Cyagen Biosciences Inc, Sunnyvale, CA) consisted of 1% (v/v) ITS supplement, 0.1% (v/v) sodium pyruvate, 0.3% (v/v) absorbate, 0.1% (v/v) proline, 0.01% (v/v) dexamethasone and 1% (v/v) TGF-β3. Pellets were made by centrifugation of 5 × 10 5 cells at 250 g for 5 min in 15 mL polypropylene tubes. Media was changed every 3 days. After 28 days, 3 pellets of each group were fixed, embedded, sectioned and then stained for 30 min with alcian blue solution 8GX. Blue staining indicated the synthesis of proteoglycans by chondrocytes. For quantification, the other three cultured pellets were fixed and stained. Then alcian blue dye was extracted with 4 mol·L -1 guanidine HCL (pH5.8) overnight, and its absorbance was measured at 590 nm.
Quantitative real-time PCR (qPCR) analysis Total RNAs from primary hDPSCs were isolated using the HP Total RNA Kit (Omega bio-tech, Norcross, GA), followed by cDNA synthesis using the Revert Aid First Strand cDNA Synthesis Kit (Thermo). qPCR assay for mRNA detection was performed by the SYBR Green Kit (Roche) using the QuantStudioTM 6 Flex Real-Time PCR System (Applied Biosystems, Foster City, CA). GAPDH was used as internal reference. The primers used in qRT-PCR were showed in Supplementary Table. Western blot analysis Cells were lysed using lysis buffer added with protease inhibitor at various time points after induction. The total protein concentrations were measured using a BCA protein assay kit (Thermo). Equal amounts f cell lysates were loaded onto a 10% SDS-polyacrylamide gel and transferred to 0.22 μm polyvinylidene fluoride membranes (Roche). After blocking with fat-free milk, the following antibodies were used: anti-RUNX2 (1:1000, Cell Signaling Technology), anti-SOX9 (1:1000, Abcam), and anti-PPARG (1:1000, Abcam). β-ACTIN was used as internal reference.
Statistical analysis qPCR and flow-cytometry data are presented as means ± SD. Differences were analyzed by unpaired two-tailed Student's t test between two groups using Graphpad Prism 7. All measurements were from distinct biological samples. Statistical parameters are found in the figure legends, including exact n and number of biological repeats. Exact P values are shown, P < 0.05 was considered as significant.

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