In vivo cartilage-formation potency of somatic stem cells is associated with responsiveness to TGFβ stimulation in vitro

Background: Somatic stem cell transplantation has been performed for cartilage injury, but the reparative mechanisms are still conicting. The chondrogenic potential of stem cells are thought as promising features for cartilage therapy, however the correlation between their potential for chondrogenesis in vitro and in vivo remains undened. The purpose of this study was to investigate the intrinsic chondrogenic condition depends on cell types and explore an indicator to select useful stem cells for cartilage regeneration. Methods: The chondrogenic potential of two different stem cell types derived from adipose tissue (ASCs) and synovium (SSCs) of mice and humans was assessed using bone morphogenic protein-2 (BMP2) and transforming growth factor-β1 (TGFβ1). Their reparative potential was also validated through transplantation into a mouse osteochondral defect model. Results: All cell types showed apparent chondrogenesis under the combination of BMP2 and TGFβ1 in vitro, as assessed by the formation of proteoglycan- and type 2 collagen (COL2)-rich tissues. However, our results vastly differed with those observed following single stimulation among species and cell types; apparent chondrogenesis of mouse ASCs, mouse SSCs, and human SSCs was observed with supplementation of BMP2, either BMP2 and TGFβ1, and BMP2, respectively. Human ASCs showed no chondrogenesis following single stimulation. Among human SSCs, two of six donors formed partially COL2-positive tissues in response to TGFβ1. However, unlike mouse cells, human cells showed brous components (COL1/3) with all treatments. Mouse SSCs formed hyaline-like cartilage (proteoglycan+/COL2+/COL1-/COL3-) in the transplanted site in vivo, but other cell types mainly formed COL1/3-positive brous tissues. Remarkably, only donors that showed chondrogenic response to TGFβ1 in vitro formed partially COL2-positive tissues in vivo. However, the area was mainly composed of COL1/3, indicating brous cartilage-like tissues. Conclusion: The chondrogenic response to TGFβ1 may represent the fate of stem cells when locally transplanted into cartilage defects. those from tissue,


Introduction
It is widely accepted that local injury of articular cartilage associated with joint trauma does not regenerate spontaneously. Transplantation of mesenchymal stem cells (MSCs) derived from bone marrow, synovium, and adipose tissue has shown to be a promising strategy for cartilage regeneration.
Several clinical trials using MSCs have been completed or are ongoing [1,2], although the actual outcome of transplantation of MSCs remain unclear.
There are con icting reports on the reparative mechanisms of local transplantation of MSCs. Koga et al. reported that transplantation of bone marrow MSCs (BMSCs) or synovial MSCs (SMSCs) repaired cartilage defects with differentiation of transplanted cells to chondrocytes in an in vivo environment in a rabbit osteochondral defect model [3]. After transplantation, the cells were engrafted and functioned as chondrocytes/cartilage over six months [4]. On the other hand, Nakamura et al. found in a porcine study that the transplanted SMSCs repaired the cartilage defect without differentiation to chondrocytes; moreover, the transplanted SMSCs disappeared within a month [5]. Recently, many studies have highlighted that the signaling effects of transplanted stem cells lead to endogenous repair by host stem cells through the secretion of growth factors, cytokines, microRNAs, extracellular vesicles, and/or cell-cell contact [6,7]. Thus, distinct mechanisms may be involved in transplantation therapy.
Several studies have suggested a relationship between in vitro chondrogenic potential and in vivo cartilage regeneration [3,8,9]. Among MSCs, SMSCs are reported to provide superior chondrogenic potential in humans [10,11], rodents [12,13], rabbit [3], and large animals [5,14]. Based on those ndings, clinical trials of transplantation of SMSCs for articular cartilage defects have been recently reported [15,16]. However, Dickhut et al. showed that transforming growth factor-β3 (TGFβ3) completely induced chondrogenesis from BMSCs, which was evident with the formation of tissue rich in type 2 collagen (COL2) and proteoglycan in vitro, whereas the stimulation with TGFβs alone was insu cient for SMSCs and adipose MSCs (AMSCs). The formed tissues contained a low amount of glycosaminoglycans and COL2 [17,18]. Previous studies have shown that supplementation with bone morphogenic proteins (BMPs) is necessary for induction of chondrogenesis from SMSCs and AMSCs [17,[19][20][21]. Moreover, other supplements such as serum and glucocorticoids also affect the chondrogenesis [19,[22][23][24][25][26]. Thus, conditions for adequate chondrogenesis are intrinsically different among MSCs, implying that chondrogenic potential is probably miscalculated depending on assay conditions. For further understanding the mechanisms underlying the action of stem cells transplantation for cartilage repair therapy, we aimed to evaluate the in vitro and in vivo chondrogenic potential of two different somatic stem cell types derived from synovium and adipose tissue in mice and humans, respectively.
The isolated cells were cultured with growth media (10% FBS-DMEM supplemented with 1 ng/mL bFGF (FUJIFILM Wako Pure Chemical Corporation (Wako), Osaka, Japan) at 37°C in a humidi ed atmosphere with 5% CO 2 and 3% O 2 . Mouse adipose stem cells (mASCs) were obtained from the stromal vascular fraction of the inguinal fat tissue by collagenase digestion for 30 min. Only dissociated cells were collected using a 70-μm strainer and cultured with growth media at 37°C in a humidi ed atmosphere with 5% CO 2 and 3% O 2 . Cultured cells were passaged with 0.25% trypsin/ethylenediaminetetraacetic acid (EDTA) at 80% con uency and replated at a density of 5,000 cells/cm 2 . The medium was changed three times per week. These cells have been already characterized in our previous report [27]. The cells were used for further experiments after 12-14 days of culture, as mouse somatic stem cells are sensitive to senescence [27,28]. We prepared over three batches obtained from several mice with mixed gender at different timings.
Human subcutaneous adipose tissue and synovium were obtained from osteoarthritis (OA) patients (N=6) during total knee arthroplasty (TKA) in accordance with a protocol approved by the institutional ethics committee. Written informed consent was obtained from all patients. The age, gender, and Kellgren-Lawrence grade of the patients are listed in Additional le 1. Adipose tissues were resected from the incision site at the knee, and synovium was resected from the suprapatellar pouch. Human ASCs (hASCs) were isolated as mentioned above and human SSCs (hSSCs) were isolated as previously established [18]. Human cells were cultured with growth media at 37°C in a humidi ed atmosphere with 5% CO 2 . Those cells were passaged by treatment with 0.25% trypsin/EDTA at 80% con uency (approximately once per week) and replated at a density of 5,000 cells/cm 2 . The media was changed twice per week. Human cells at passage 3 were used for further experiments.

Animal experiment
Twelve-week-old male mice were anesthetized by an intraperitoneal injection of a mixture of 0.3 mg/kg of medetomidine, 4.0 mg/kg of midazolam, and 5.0 mg/kg of butorphanol. For both knees, the femoral trochlear grooves were exposed via a medial parapatellar incision with laterally patellar dislocation. An osteochondral defect (diameter, 0.5 mm; depth, approximately 0.5 mm) was created in both knees by manual drilling. A pellet of 5 × 10 4 cells prepared by overnight culture with growth media were transplanted into the defects, and then patellar dislocation was reduced. The joint capsule and the skin were sutured as separate layers. After surgery, mice were allowed to be active without any xation device or immobilization. C57BL/6 and CAG-EGFP C57BL/6 mice were used for allograft study and C.B-17 SCID mice were used for xenograft study.

Micro-computed tomography
Whole knee joints were scanned using micro-computed tomography (μCT; inspeXio SMX-100CT system (Shimadzu, Kyoto, Japan)) at a resolution of 12 μm per voxel using the following consistent parameters: 75 kV and 140 mA. Three-dimensional images of bone were analyzed using the TRI/3D-BON software (RATOC System Engineering, Tokyo, Japan). Bone volume (BV) and bone mineral density (BMD) at the area of osteochondral defect trimmed into a cylinder (diameter, 0.5 mm; depth, 0.4 mm) were calculated as described previously [29].

Histology and histochemistry
The samples were xed in 10% neutral buffered formalin and embedded in para n wax, which was followed by serial dehydration using ethanol and clearance using xylene. For bone tissues, samples were decalci ed with 10% EDTA (pH 7.4) after formalin xation and delipidation using ethanol. Sections cut

mRNA expression analysis
Cell pellets were initially homogenized with zirconia beads in TRI Reagent (Cosmo Bio, Tokyo, Japan). Total RNA was then extracted using Direct-zol RNA Kit (Zymo Research, CA, USA) according to the manufacturer's protocol. Total RNA was reverse transcribed to cDNA using ReverTra Ace qPCR RT Master Mix (TOYOBO, Osaka, Japan). Quantitative reverse transcription polymerase chain reaction was performed using THUNDERBIRD SYBR qPCR Mix (TOYOBO) and Thermal Cycler Dice Real Time System III (TaKaRa Bio, Shiga, Japan). The information about primers is listed in Additional le 3. The expression levels normalized to the levels of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) were measured using absolute standard curve method [30] or delta CT method [31].

Transcriptome Analysis
Total RNA extracted from human cells was used for transcriptome sequencing. The sequence libraries was prepared using a NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs, MA, USA) according to the manufacture's protocol. Sequencing was performed using an Illumina HiSeq 4000 System with 2 × 150 bp paired-end reads (Veritas Genetics, MA, USA). The raw sequence data were ltered to remove adaptor sequences, low-quality reads, sequences with a high content of N, and reads <50 bp length by using Trimmomatic (ver.0.39). The ltered data were aligned against the human reference genome (GRCh38.p13) using STAR (ver.

TGFβ1 induced chondrogenesis from mSSCs but not mASCs in vitro
Previous reports indicate that optimal chondrogenic conditions differ among cell types [17,32,33]. To evaluate the actual chondrogenic capacity, in vitro pellet culture under stimulation with BMP2, TGFβ1, or their combination (B + T) was performed. As a reference, neither mASCs nor mSSCs differentiated into chondrocyte in the basal chondrogenic medium (see Additional le 4).
When mASC pellets were cultured under those three conditions, chondrogenic differentiation ascertained by semi-translucent tissue formation composed of proteoglycan and COL2 was recognized with supplementation of BMP2 (Fig. 1A, 1B). However, brotic matrices composed of type 1 and 3 collagen were observed in the chondrogenic tissues, indicating brocartilage development. Although there was no evidence for chondrogenesis in TGFβ1 treated mASCs even at the mRNA levels (see Additional le 5), TGFβ1 facilitated BMP2 induced chondrogenesis with reduction of COL1/3 component. In contrast, mSSCs underwent chondrogenesis under all conditions with the formation of a semi-translucent tissue with SOX9-positive chondrocytes embedded in the homogenous extracellular matrices enriched in proteoglycan, type 2 collagen, and low amount of COL1/3 (Fig. 1C, 1D). These observations show that mSSCs could form hyaline-like cartilaginous tissue. COL10, a hypertrophic marker, was observed in the area surrounding the hypertrophic cell lacuna only with BMP2 supplementation. The same trends were observed and validated at gene levels (see Additional le 5). Taken together, mASCs and mSSCs showed the potential for chondrogenesis but the suitable conditions were intrinsically different among the various cell types in vitro.
mSSCs, but not mASCs, formed cartilaginous tissue in the osteochondral defect in vivo The reparative capacities of mASCs and mSSCs in the osteochondral defect were evaluated by transplantation of the cell pellet into osteochondral defects. The osteochondral defect with no transplantation showed spontaneous repair of the subchondral bone, but the articular cartilage region did not regenerate, and was covered with brous tissue (N = 11/14) or exposed bone (N = 3/14) ( Fig. 2A).
When mASCs were transplanted, the defects were lled with brous tissue with no regeneration of articular cartilage (Fig. 2B). Although, COL1 is a major brous tissue component, its detection was complicated for the assessment of brous tissue because the surrounding bone was also positive for COL1. Therefore, COL3, which is another brous component, was adapted in this study and strikingly signi ed the brotic area. GFP detection demonstrated that transplanted cells remained within the transplanted site without differentiation toward chondrocytes or osteoblasts, resulting in the impeded spontaneous repair of subchondral bone. In some cases, Saf-O + /COL2 + area was partially observed (N = 4/25); however, they were brous (N = 3/25) or ectopic site that protruded from the articular cartilage region (N = 1/25) (see Additional le 6).
On the other hand, when mSSCs were transplanted into the chondral defect, cartilaginous tissue rich in proteoglycans and COL2 but low in COL1/3 (N = 15/20) was newly generated (Fig. 2C). Interestingly, in all unrepaired cases (N = 5/20), defect sites were located in the distal femur groove and the host synovium tissue was in ltrated from enthesis of crucial ligaments to the defect site (see Additional le 7). GFP mice-derived mSSCs were detected at neo cartilage tissue, indicating that mSSCs differentiated to chondrocytes and formed cartilage tissue in vivo (Fig. 2D). Notably, GFP-negative cells were also detected within the neo-cartilage tissue, particularly in the articular surface area and the boundary of bone marrow.
To investigate the contribution of host cells, wild-type (WT)-derived mSSCs were transplanted into GFP mice. As a result, approximately 20% host cells were also incorporated for cartilage regeneration with differentiation toward chondrocytes (GFP-positive cell rate: 83.1 ± 6.6% (N = 5) in GFP cell transplantation to WT mice and 20.8 ± 11.4% (N = 4) in WT cell transplantation to GFP mice). At 6 weeks, remodeling of the neo-cartilaginous tissue was histologically observed in the deep zone (Fig. 2E). COL10 was observed from the middle to the deeper area of the neo-cartilage, which expanded at 6 weeks. The µCT analysis revealed that ossi ed volume of neo-tissues increased at 6 weeks but the mineral density did not matured like bone, suggesting the proceeding cartilage ossi cation (Fig. 2F). These ndings indicate that mSSCs repaired the osteochondral defect via endochondral development, and that accompanied with the stimulation of host cells to differentiate toward chondrocytes.  [34]. In vivo specimens transplanted with mSSCs showed strong pSmad1 and weak pSmad2 activation during chondrogenesis. On the other hand, the brous area formed by mASCs contained cells weakly positive for pSmad1 and negative for pSmad2 (see Additional le 8). Therefore, the TGFβ-Smad2/3 pathway is important during in vivo chondrogenesis.
To elucidate the distinct outcome in response to TGFβ1 among cell types, the expression of their speci c receptors was con rmed. Among TGF receptor superfamilies, Alk1 and Alk7 were remarkably higher in mSSCs and mASCs, respectively. However, Alk5 which is a major TGFβ type 1 receptor, and Tgfbr2 were comparable in mSSCs and mASCs. There was no difference in Alk2, Alk3, Alk4, Alk6, and Bmpr2 (Fig. 3A).
Furthermore, the levels of most receptors did not dynamically alter in 2 weeks of chondrogenic culture (see Additional le 9A). Western blotting analysis showed that during short-term stimulation, Smad1/5 and Smad2 were activated in response to BMP2 and TGFβ1, respectively, in both cell types according to conventional theory (see Additional le 10). Thus, mASCs and mSSCs were comparable in terms of the expression of TGFβ receptors as well as of receptors mediating Smads phosphorylation.
Conversely, in chondrogenic pellet culture, TGFβ1 did not affect either Smad1/5 and Smad2 pathway in mASCs, while it temporally activated Smad2 and Smad1/5 in some batches of mSSCs with obscure trends (Figs. 3B, see Additional le 11). Notably, the activation of Smad2 was marked with the presence of BMP2 as well as Smad1/5 in both cell types. This was supported by the fact that SB-431542, an Alk5 inhibitor, strikingly abrogated chondrogenesis induced not only by TGFβ1 but also by BMP2 (Fig. 3C). Although TGFβ signal is associated with cell growth and survival, its inhibition led adipogenesis not cell death (see Additional le 12).
Considering the activation of Smad2 during BMP2 stimulation, it is possible that autocrine ligand expression contributed to chondrogenesis. Regarding endogenous chondrogenic growth factors, Bmp2 was expressed higher in mSSCs than mASCs whereas TGFβ1/3 were expressed at comparable levels (Fig. 3D). Bmp4 and Bmp7 were barely expressed in two cell types (data not shown). Tgfb1/3 were constant or marginally upregulated in chondrogenesis, and Bmp2 and Bmp6 were upregulated in BMP2 stimulation toward the end of induction, similar to Col10a1 (see Additional le 9B).
Based on the above in vitro ndings, it can be inferred that TGFβ receptor-mediated signal was indispensable for chondrogenesis. However, the output in response to TGFβ could not be determined in their pro les for receptors and ligands.

Distinct chondrogenic conditions and outcomes between hASCs and hSSCs with variations among donors
To expand the scope of application of murine experiments in human somatic stem cells, the responses of hASCs and hSSCs to chondrogenic growth factors were evaluated in vitro.
When hASCs were cultured with BMP2 or TGFβ1 alone, weak staining for Saf-O and aggrecan was recognized in the formed tissues in some cases but COL2 was completely absent (Fig. 4A). Instead, COL1/3 were abundant in their tissues. Their chondrogenic potential could be demonstrated with use of the combination of BMP2 and TGFβ1, as evident from the development of proteoglycan-and type 2 collagen-enriched translucent tissue. With increased cartilaginous matrices, COL3 was restricted to the peripheral zone; however, COL1 was observed, similar to that in other single treatments. These results were consistently observed in all six donors (Fig. 4B).
In hSSCs, BMP2 + TGFβ1 stably induced chondrogenesis in all donors similar to hASCs, whereas a distinct pattern for chondrogenic response to single stimulation was uniquely exhibited among the donors (Fig. 4C, 4D). When hSSCs were cultured with BMP2 alone, different grades of chondrogenesis were observed among donors: partially intense staining for chondrogenic markers was observed in 3/6 donors (D1, D3, and D4) and entire intense staining in 2/6 donors (D2 and D5) within developed tissues. Conversely, TGFβ1 alone induced proteoglycan synthesis, as observed in Saf-O and immunohistochemistry for aggrecan. Nevertheless, COL2 was recognized only in 2/6 donors (D1 and D3). Correlation analysis supported the inconsistency of the relationship between proteoglycans and COL2 (see Additional le 14A). Of note, all tissues developed from hASCs and hSSCs clearly contained COL1/3 under all given chondrogenic conditions. COL3 tended to be weakened in the specimen with intense staining for COL2; however, there was no negative correlation between COL1 and COL2 (see Additional le 14B and 14C).
Interestingly, the chondrogenic potential based on the response to BMP2 was not representative of the chondrogenic response to TGFβ1 (see Additional le 15). Thus, unique chondrogenic responses were observed between cell types and donors, although principal component analysis plot and hierarchical correlation analysis of mRNA-seq data showed low consistency in gene pro les between hASCs and hSSCs (Figs. 5A, 5B, see Additional le 16 and 17). The expression of receptors and ligands of the TGFβ superfamily was similar among cell types and donors (Fig. 5C). Moreover, no speci c enrichment was observed in multiple gene sets related to TGFβ signaling, proteoglycan, and chondrogenesis/cartilage development (see Additional le 18).
Thus, hASCs and hSSCs required distinct chondrogenic condition beyond those required by mouse stem cells. Moreover, the brous components were more abundant in developed tissues in human stem cell cultures, which characteristically resembled brocartilage in human stem cells. However, similar to that in mouse stem cells, the chondrogenic potential could not be predicted based on receptor/ligand expression.

Only hSSCs reactive to TGFβ1 in vitro formed brous cartilage in vivo
To investigate the in vivo chondrogenic potential of human stem cells, hASCs and hSSCs were transplanted into the osteochondral defect model using SCID mice without pre-chondrogenic induction and evaluated at 4 weeks.
In case of hASC transplantation, the subchondral bone was partially regenerated, but the surface zone Collectively, TGFβ1-responsive hSSCs could form brous cartilage in vivo. Moreover, the engraftment rate was intrinsically different among cell types regardless of the chondrogenic potential. Furthermore, the local transplantation of cells with lower potential for in vivo chondrogenesis may lead to other side effects such as erosion of the bone and cartilage.

Discussion
Our ndings showed that the chondrogenic potential of somatic stem cells in vitro, which has been de ned in BMP2 supplemented conditions, did not correlate with the reparative capacity for local transplantation therapy of the osteochondral defect in vivo. Moreover, this study is the rst to demonstrate that the response to TGFβ1 in vitro represents the outcome after local transplantation, which may be a potential indicator to predict in vivo chondrogenesis. Although transplantation therapy of human SSCs have been conducted for the purpose of repair articular cartilage [15,16], it may be more suitable for brous cartilage repair such as meniscus [26,35,36]. Moreover, the test for chondrogenic response to TGFβ in vitro will provide a new criteria to select cell source from allogenic cell banks.
As the major limitation of a small animal study, the subchondral bone defect must be employed for the purpose of studying only the cartilage region. However, our data showed lower reparability of the cartilage even in an osteochondral injury, whereas bone defect region was spontaneously regenerated with membranous ossi cation. For articular cartilage repair, cartilaginous callus formation is required to remodel the osteochondral region for endochondral development. As SSCs are known to induce host synovial tissues [36], location of the defect site is critical for mSSC transplantation because in ltration of the host synovial tissue was observed when the defect was close to the enthesis of crucial ligaments. Furthermore, transplanted mSSCs detected below the marrow area did not exhibit signs of chondrogenesis, and in case of synovial in ltration, vessel-like shapes were found (see Additional le 7).
Thus, location and site-speci c signals, including mechanical stress, may be necessary for in vivo chondrogenesis as chondrogenesis was not observed in the knee with patellar dislocation (data not shown) [37,38].
Although we concluded the importance of chondrogenic response to TGFβ1, the distinct chondrogenic result of each cell type could not be elucidated in the analysis of TGFβ superfamily receptors, endogenous ligands, and downstream pSmads. In fact, all cell types used here were considered to react to TGFβ with unique tissue formation, unlike that found in basal or BMP2 supplemented medium. Other studies have shown the inconsistencies in cellular receptor expression in in vitro chondrogenic activity [39,40]. In general, TGFβs and BMPs bind their speci c receptors; ALK5 is a type I receptor for TGFβs; ALK2, ALK3, and ALK6 are type I receptors for BMPs, which activate downstream Smads, Smad2/3 and Smad1/5/8, respectively. However, TGFβs are known to activate Smad1/5/8 pathway via binding to ALK1, a predominant BMP type I receptor [34], and BMPs can activate Smad2 pathway via binding to ALK5 [41]. The pathway that will be activated is determined by the abundance of receptors on the cell membrane and their a nity to ligands [42]. For example, TGFβ signaling in OA chondrocytes switch from the anabolic ALK5-Smad2/3 pathway to catabolic ALK1-Smad1/5/8 pathway due to disproportionate ALK1/ALK5 ratio, which accelerates cartilage destruction [43][44][45][46]. Using western blotting, we demonstrated that Smad activation in the order of hours and days were distinctly regulated. It is known that transcriptional targets of Smad2/3 dynamically change in the order of minutes and hours [47]. These results suggest that there are inadequate downstream gene markers to assess the activation of TGFβ signaling. Moreover, the non-Smad pathways also regulate cellular function and differentiation, including chondrogenesis [34,48]. Thus, the entirety of TGFβ signaling from cue to output in each cell type is di cult to interpret; therefore, the chondrogenic response to TGFβ cannot be predicted without chondrogenic culture.
In fact, TGFβ1 is constantly supplied to the synovial uid from cartilage and synovium [53,54]. Therefore, tests for response to TGFβ can reasonably predict cell fate after transplantation into joints. BMSCs undergo chondrogenesis with TGFβ treatment alone [17,55], and are responsible for cartilage regeneration in focal cartilage defect using the bone marrow stimulation method [56]. We previously showed the in vivo chondrogenic potential of hBMSCs in a rat osteochondral model [22]. To further verify our proposal, we should have included hBMSCs in our analysis; however, as one of the limitations, it was di cult to prepare a su cient number of hBMSCs from bone marrow effusion after osteotomy in TKA operation. Besides, we used subcutaneous fat from the knee area, which may have different characteristics from abdominal fat [57,58]. Recently, medical waste-derived stem cells such as adipose tissue, synovium, umbilical cord, and dental pulp have been highlighted for stem cell therapy because of low additional invasion for tissue collection and their superior proliferative capacity compared with BMSCs [11,59,60]. Our data are limited and may not be applicable to other cell types; therefore, further study is required using other sources and cells puri ed from speci c populations.  [39]. The study also pointed a discrepancy in the expression of aggrecan and COL2 in chondrogenic culture of hSSCs, as shown herein as well as in our previous study [18]. Thus, conventional cell isolation from synovium without any selection yields low reproducibility, whereas our results provide a new criterion for selecting a useful cell source, even from OA synovium.
It is known that synovial broblasts (SFs) contribute to cartilage and bone erosion in rheumatoid arthritis (RA) [65]. Given that cultured SFs have traditionally been studied in arthritis using the same isolation procedure used in therapeutic study, their erosive potential of hSSCs should be considered. Recent singlecell transcriptome analysis showed that in ammatory broblasts expressing TNFα, IL6, and IL1B as well as immune cells were found in OA synovium. Moreover, they also expressed several proteinases such as ADAMTSs and MMPs [66]. Tsuchiya et al. showed that the responses of OA-SFs to the stimulation of in ammatory cytokines were shared in RA-SFs, indicating the common potential for proin ammation in expanded cells derived from OA, RA [67], and probably healthy synovium. Furthermore, other groups demonstrated that hSSCs either from healthy subjects or OA patients could induce osteoclastogenesis from PBMCs in vitro [68]. These results support our data regarding the TRAP activity at the border of brous neo-tissues. Altogether, cultured synovial cells have two aspects as stem cells and diseasemodifying broblasts. Regarding the purpose of regenerative therapy, further study will be needed to avoid the negative effects of hSSCs. These ndings have not been reported in hASCs; however, the erosion of host tissue was also observed in cell transplantation. Our data indicate that local transplantation for chondral injuries using cells with low chondrogenic potential in vivo should be carefully considered.
There are several limitations in this study. Firstly, we did not verify the surface antigens and potential of cells for osteogenesis and adipogenesis, which are often used in MSC studies; however these have been con rmed in our previous studies [27,69] or elsewhere with comparable culture methods [17,60]. In fact, the criteria and terminology for MSCs have recently been reconsidered [70]. Therefore, we referred to somatic stem cells using the name of the source tissue in this study. With respect to repair in bone diseases, evaluation of the osteogenic capacity in vitro may be helpful; however, in vivo osteogenesis was not recognized in the cells studied herein. Secondly, we suggested that endogenous expression of TGFβ superfamilies may contribute during chondrogenesis, but it was only at the gene level. Further study will be needed to clarify at the protein levels, including dominant ligands for chondrogenesis. Finally, the endpoint of in vivo study was 3-6 weeks, which is insu cient to evaluate the actual fate of neo-tissue. After written informed consent was obtained from all patients, surgical wastes were used in this study.
The protocol was approved by the ethics committee of the University of Tokyo (No.0622-12). Animal experiments were also approved by the University of Tokyo (No.P17-091).

Data Availability Statement
Nucleotide sequence data reported are available in the DDBJ Sequenced Read Archive under the accession number DRA011462. Other data that support the ndings of this study are available from the corresponding author upon reasonable request.

Competing interests
The authors declare that they have no competing interests.

Consent for publication
Not applicable.

Fundings
This work was supported by JSPS KAKENHI (Grand Numbers: 20K18053, 18J01264, and 19H05565), The Nakatomi Foundation, and the Takeda Science Foundation.