Tracing the destiny of mesenchymal stem cells from embryo to adult bone marrow and white adipose tissue via Pdgfrα expression

Mesenchymal stem cells (MSCs) display self-renewal capacity, proliferation sustainability in vitro and the potential to differentiate into various mesenchymal cell lineages, including adipocytes, chondrocytes, and osteocytes (Prockop, 1997; Pittenger et al., 1999). MSCs are also crucial for several tissue formation processes and injury repair, including bone fracture (Grcevic et al., 2012; Mizoguchi et al., 2014; Miwa and Era, 2015). MSCs have been proposed for use in stem cell therapy because they can be derived from adult bone marrow (BM) and white adipose tissue (WAT); however, the safe and successful application of MSCs for clinical use requires clarification of their origin and the development of reliable markers for fate mapping.

Previously, we have exploited two methods for inducing mesenchymal cell lineages from embryonic stem cells (ESCs) that have also proved useful for dissecting the differentiation process in vitro (Wobus et al., 2002; Kawaguchi et al., 2005; Sakurai et al., 2006). The first method, which was culturing ESCs on collagen IV-coated dishes under serum-containing medium, supported the generation of mesoderm cells that can give rise to chondrocytes, osteocytes and myocytes. The other method involved treating ESCs with retinoic acid (Dani et al., 1997), thus mimicking the natural MSC differentiation pathway through neuroepithelium. This second method was demonstrated in vivo using a Sox1-Cre knock-in mouse line and indicated that Sox1-positive neuroepithelial cells supply the earliest wave of MSC differentiation that occurs during embryogenesis and is subsequently replaced by MSCs from other origins during postnatal development (Takashima et al., 2007). Nevertheless, our observations in these previous studies did not rule out the possibility that mesoderm could give rise to MSCs under other appropriate conditions. The current study investigates this possibility further in vivo using our previously generated Pdgfrα-GFPCreER^T2 knock-in mouse line together with several other mutant mouse lines. The results define Pdgfrα as a key marker of MSC in vivo and show that almost all adult BM-MSC and WAT-MSC originate from mesoderm and embryonic Pdgfrα-positive cells.

To first investigate when MSCs emerge during mouse embryonic development, we sorted total cells from whole embryos at respective developmental stages by flow cytometry using Pdgfrα as a MSC marker, and then performed colony-forming unit-fibroblast (CFU-F) assays on the sorted cells (Fig. 1A, Fig. S1A). Colonies of CFU-F initially emerged at embryonic day 9.5 (E9.5) and were significantly increased in number between E11.5 and E14.5 (Fig. 1A). We observed no colonies in Pdgfrα-negative fractions (Fig. 1A), and E7.5 cells could not be cultured using our methods (data not shown). Correlating with the increased CFU-F colony numbers during embryo development, the percentages of Pdgfrα-positive fractions in total cells also rose from E11.5 to E13.5 (Fig. S1A). The CFU-F colony formation tendency in C57BL/6 mice was similar to that of another mouse strain, ICR (Fig. S1B,C). Furthermore, we divided whole embryos into four parts, head, trunk, limb and tail, and then sorted Pdgfrα-positive cells from each. Trunks showed a similar sorting pattern to whole embryos, whereas CFU-F colonies in heads and limbs only emerged at E13.5 (Fig. 1B). The percentages of Pdgfrα-positive fractions in all segmented parts of embryos peaked at E13.5, similar to the results for whole embryos (Fig. S1A,D). Although we examined expressions of several marker genes in Pdgfrα-positive and -negative fractions by RT-PCR, Sca1 (Atxn1) and Cd51 (Itgav), which have been reported as MSC markers (Morikawa et al., 2009; Pinho et al., 2013), were expressed in both fractions (Fig. S1E). These experiments suggested that embryonic MSCs emerge in waves and that Pdgfrα is a key marker of MSCs (Fig. 1A, Figs S1C and S2A); however, we could not confirm that all of CFU-F colonies were MSCs with the attendant features of self-renewal capacity and multipotency. We therefore picked up some of the CFU-F colonies derived from various stages and tissues, and established cell lines for induction into adipocytes, chondrocytes and osteocytes (Fig. S2B). CFU-F colonies from E9.5 samples showed no multipotency, whereas some was shown for those from E11.5 and E13.5 (Table 1). Because MSCs were shown to be located in perivascular niches of both adult BM and WAT (Traktuev et al., 2008; Morikawa et al., 2009), we derived CFU-F colonies from these tissues. Almost all of the BM-derived CFU-F colonies showed multipotential differentiation, but in adult mice about half of subcutaneous WAT (sWAT)-derived CFU-F colonies did not (Table 1). These results suggest that almost all of BM-derived CFU-F colonies and half of sWAT-derived CFU-F colonies were MSCs. Moreover, adult gonadal WAT (gWAT) did not give rise to CFU-F colonies with multipotency in our experiments (Table 1). These data show that CFU-F colonies do not necessarily have proliferation capacity and multipotency in the same way as MSCs do.

To trace embryonic MSCs in vivo, we mated a Pdgfrα-GFPCreER^T2 knock-in mouse line generated previously (Miwa and Era, 2015) with Rosa26-Tomato mice, which can continuously express the Tomato reporter gene from various developmental stages under the control of Cre activity following treatment with 4-hydroxytamoxifen (4-OHT) (Feil et al., 1997; Madisen et al., 2010). Because pregnant mice injected with tamoxifen cannot bear pups normally, all pups were delivered by Cesarean section and nurtured by foster mothers (Samokhvalov et al., 2007; Zovein et al., 2008; Cebrian et al., 2014). We then isolated BM cells from tibiae and femora of Pdgfrα^GCE/wt;Rosa26^Tomato/wt mice once they reached 8 weeks of age, and counted the number of Tomato-positive and -negative CFU-F colonies (Fig. 2A). Although there were no Tomato-positive colonies derived from 4-OHT-treated Pdgfrα^GCE/wt;Rosa26^Tomato/wt BM at E6.5, almost all of CFU-F colonies from mice treated at and after E7.5 were positive for Tomato expression (Fig. 2B). Then, using Pdgfrα^GCE/wt;Rosa26^LacZ/wt mice that continuously express the lacZ gene under Cre recombinase activity following 4-OHT treatment, X-gal staining revealed that Pdgfrα-positive cells generated at embryonic stages remained in adult BM (Fig. S3B). Indeed, E7.5 Pdgfrα-positive cells contributed to a large proportion of BM, but positive cells at and after E9.5 were limited, supporting results reported previously (Fig. S3A) (Ding et al., 2013). Moreover, almost all of CFU-F colonies derived from adult BM were continuously expressing Pdgfrα (Figs S2A and S3C). Half the dose (25 μg/g) of 4-OHT injected into pregnant mice was insufficient to recombine loxP sites by CreER^T2 and 100 μg/g injection caused miscarriages or stillbirths in almost all cases (data not shown). Pdgfrα^GCE/wt;Rosa26^Tomato/wt CFU-F from mice treated with vehicle at E9.5 and E10.5 did not express Tomato (Fig. S3D). Pdgfrα is expressed in the mesoderm, somite and branchial arch of the embryo (Takakura et al., 1997; Miwa and Era, 2015), which have been assumed to be where MSCs originate. Together with previous data, these findings suggest that embryonic MSCs, adult BM-MSCs and their progenitors continuously express Pdgfrα from E7.5.

To further explore the origin of MSCs in adult BM, we performed the same analyses using a Brachyury-GFPCreER^T2 knock-in mouse line (Imuta et al., 2013; Taguchi et al., 2014) to test for a recognized mesoderm and notochord marker at between E7.5 and E9.5 (Herrmann et al., 1990). Brachyury is a transcription factor that belongs to the T-box family of molecules and is required for mesoderm formation (Bennett et al., 1972; Yanagisawa et al., 1981). We treated Brachyury^GCE/wt;Rosa26^Tomato/wt mice with 4-OHT at E6.5, E7.5, E8.5, E9.5, E10.5 and postnatal day (P) 2, P3 and P4, and examined Tomato expression of CFU-F colonies derived from adult BM. Almost all colonies were Tomato-positive from E7.5-treated BM, but not from BM treated at any other stage during development regardless of the expression of brachyury (Fig. 3A, Figs. S1E and S4A). Additionally, adult CFU-F did not express brachyury (Fig. 3B). We further investigated whether neuroectoderm cells could contribute to adult BM-MSC using a Sox1-Cre knock-in mouse line as reported previously (Takashima et al., 2007; Yoshimura et al., 2013). Sox1 was shown to be the most specific marker for neuroepithelial cells (Fig. S1E) (Pevny et al., 1998; Aubert et al., 2003). Although we detected MSCs that originated from Sox1-positive cells in E14.5 heads (data not shown), no CFU-F colonies originating from these cells existed in adult BM (Fig. 3C). We therefore speculated that adult BM-MSCs develop from limb buds, and subsequently analyzed Pdgfrα-GFPCreER^T2, Brachyury-GFPCreER^T2, Sox1-Cre and Pmx1-Cre mutant mouse lines (Logan et al., 2002). Pmx1 (Prrx1) is expressed in early limb bud mesenchyme and plays an essential role in regulating skeletal development in limbs (Fig. S1E) (Cserjesi et al., 1992; Martin et al., 1995). We found that almost all cells in limb buds originated from brachyury-positive mesoderm at E7.5, when it also expresses Pdgfrα (Fig. 3D, Fig. S4A,B). Considering the results of tracing Pmx1-positive cells (Fig. 4A), MSCs in limb buds probably develop into adult BM-MSC because limb buds and adult BM-MSC have the same origin.

As for adult BM-MSC, we investigated the origins of adult sWAT-MSC using respective mutant mice. The results largely mirrored those for BM, i.e. CFU-F colonies derived from sWAT were Tomato-positive in adult Pdgfrα^GCE/wt;Rosa26^Tomato/wt mice treated with 4-OHT at and after E7.5 (Fig. 5A), and in adult Brachyury^GCE/wt;Rosa26^Tomato/wt mice at E7.5 and E8.5 (Fig. 5B). Negative CFU-F colonies in these assays possibly partly reflected a neuroectoderm origin (Sox1-positive cells) (Fig. 5C). Finally, Tomato-positive CFU-F colonies from Sox1 knock-in sWAT could not differentiate into three mesenchymal cell lineages (data not shown). These results suggest that almost all adult sWAT-MSC originate from mesoderm and embryonic Pdgfrα-positive cells as BM-MSCs.

Based on previous reports that Pmx1-positive cells show MSC differentiation potential (Krueger et al., 2014; Sanchez-Gurmaches et al., 2015), we examined adult BM and WAT in Pmx1-Cre^Tg/0;Rosa26^Tomato/wt mice using Pmx1 as a marker. The majority of adult BM-MSCs were Tomato-positive, but both positive and negative CFU-F colonies were derived from adult sWAT and gWAT (Fig. 4A). Lines established from adult sWAT CFU-F could differentiate into adipocytes, chondrocytes and osteocytes regardless of expression of Tomato (Fig. 4B, Table 1). On the other hand, lines derived from adult gWAT did not give rise to adipocytes (Fig. 4C, Table 1). Our results suggest that Pmx1 is a useful marker of limb bud and adult BM-MSC, but not WAT-MSC (Figs 3D, 4A-C, Fig. S4B). In addition, almost all BM-MSCs originated from Pmx1-positive cells, whereas some sWAT-MSCs originated from Pmx1-negative cells.

We next examined the role of adult Pdgfrα-positive cells under various abnormal conditions, primarily testing whether they could contribute to a bone fracture repair. Adult Pdgfrα^GCE/wt;Rosa26^Tomato/wt mice were treated with 4-OHT and their tibiae were fractured. Tomato-positive cells accumulated in the fractured sites during repair (Fig. 6A). Analyses of Pdgfrα^GCE/wt;Rosa26^Rainbow/wt mice (Rinkevich et al., 2011) revealed Cherry-, Orange- and Cerulean-positive cells in the repaired sites with a mosaic pattern (Fig. 6B), indicating that multiple clones of MSCs contributed to fracture repair through their proliferation and differentiation.

Pdgfrα-positive cells have been associated with obesity (Lee et al., 2012); in this study, adult Pdgfrα–positive cells marked with Tomato matured into adipocytes in sWAT following a high-fat diet (Fig. 6C). Based on previous data that Pdgfrα-positive cells contribute to beige adipocytes (Lee et al., 2015), a beige marker was examined in adult Pdgfrα^GCE/wt;Rosa26^Tomato/wt mice treated with 4-OHT. Immunostaining of harvested tissues showed that Tomato-positive cells in adult Pdgfrα^GCE/wt;Rosa26^Tomato/wt sWAT expressed Ucp1 (Fig. 6D), which is a marker gene of beige cells (Wu et al., 2012; Kazak et al., 2015).

On the other hand, Pdgfrα-positive cells are also reported to proliferate in liver fibrosis (Campbell et al., 2005). Herein, Tomato-positive cells that also stained with Masson tricrome were detected in livers of adult Pdgfrα^GCE/wt;Rosa26^Tomato/wt mice treated with carbon tetrachloride (Fig. S5A). In addition, skin ulcerations were repaired by Pdgfrα-positive cells (Fig. S5B), and transplanted Tomato-positive BM-MSCs were engrafted at ulcerations (Fig. S5C) (Driskell et al., 2013). In summary, our knock-in mice directly demonstrate that adult Pdgfrα-positive cells are involved in some pathological conditions and could be useful in further studies of such disorders.

In the initial part of this study, we clarified when and where MSCs emerged during embryo development. CFU-F colonies initially appeared at E9.5, but did not show multipotency until E11.5 in whole embryos. On the other hand, MSCs were markedly abundant at E13.5 in heads and limbs. To our knowledge, this study provides the first detailed determination of MSC development and indicates that embryonic MSCs emerge in waves.

Almost all of CFU-F colonies derived from adult BM were Tomato positive in Pdgfrα^GCE/wt;Rosa26^Tomato/wt and Brachyury^GCE/wt;Rosa26^Tomato/wt mice treated at E7.5. Colony-forming cells have a multipotency for the differentiation into three mesenchymal lineages: adipocytes, chondrocytes and osteocytes. These findings suggest, in turn, that most of BM-MSC originate from both Pdgfrα- and brachyury-positive cells at E7.5, which are mesoderm cells (Bennett et al., 1972; Yanagisawa et al., 1981; Herrmann et al., 1990; Takakura et al., 1997; Imuta et al., 2013; Miwa and Era, 2015). In contrast, some CFU-F colonies derived from adult sWAT did not express the Tomato reporter gene. Interestingly, some Tomato-positive CFU-F were detected in sWAT from Sox1^Cre/wt;Rosa26^Tomato/wt mice, in which Sox1-positive neuroectoderm cells can be traced during development (Takashima et al., 2007; Yoshimura et al., 2013), but did not show multipotency, despite forming CFU-F-like colonies. Taken together, we can conclude that most adult BM- and sWAT-MSC originate from mesoderm, whereas the origin of some embryonic MSCs is neuroepithelium (Takashima et al., 2007). Recent studies showed that MSCs could differentiate into endoderm- and neuroectoderm-derived cells such as hepatocytes and neurons, respectively (Anjos-Afonso et al., 2004; Lee et al., 2004; Beltrami et al., 2007). Taking these findings together, MSCs should have the potential to cross germ layer boundaries.

Pdgfrα is an absolute marker of MSC in mice at least, and although MSC are currently being sorted using Pdgfrα and Sca1 from mouse BM, the latter is not always required for their isolation because MSCs exist in both Sca1-positive and -negative fractions (Morikawa et al., 2009). Furthermore, other species such as humans do not have the gene to encode Sca1 (Holmes and Stanford, 2007). Therefore, until better markers of MSCs emerge, Pdgfrα is the preferred option in mice. We confirmed that Pdgfrα-positive cells were involved in the repair of the bone fracture. Thus, Pdgfrα is one of the candidate markers for purifying functional MSC for clinical purposes, although its expression patterns in humans might be different from those in mice (Sakurai et al., 2012; Pinho et al., 2013; Li et al., 2014; Miwa and Era, 2016).

We also demonstrated that Pmx1 is a viable marker of MSC in adult BM, but not in WAT. There were no differences in morphology or proliferation between Tomato-positive and -negative MSCs derived from Pmx1-Cre^Tg/0;Rosa26^Tomato/wt sWAT studied herein (data not shown). These data suggest that the specificity of MSC is independent of Pmx1 expression. Alternatively, it is possible that these mutant mice non-specifically express Cre recombinase. Indeed, reporter genes were occasionally expressed wholly after mating with Pmx1-Cre transgenic mice, as reported by the mouse originators (data not shown) (Logan et al., 2002).

In studies such as this one, knock-in mice rather than transgenics should be generated when reproducing precisely the expression pattern of a target gene and activating CreER specifically (Feli et al., 1997). For example, Nestin-Cre and Wnt1-Cre transgenic mice do not always recapitulate the expression and activation of targets (Itoh et al., 2012; Lewis et al., 2013), and a Pmx1-GFPCreER^T2 transgenic mouse line has leaky CreER activity (data not shown) (Kawanami et al., 2009). On the other hand, a knock-in Pdgfrα-GFPCreER^T2 mouse line shows a correlation between expression of the transgene and Pdgfrα, and high-efficiency CreER function (Miwa and Era, 2015). However, depending on the design of construction, transgenes cannot occasionally be expressed substantially even in knock-in mice. It might cause some of the discrepancies between our and previous results (Ding et al., 2013). Moreover, tracing experiments from embryo to adult are rare using the CreER system due to the time and effort required: specifically, because pregnant mice administrated with tamoxifen are not able to bear pups naturally, Cesarean section and foster mothers are necessary to grow mutant mice (Samokhvalov et al., 2007; Zovein et al., 2008; Cebrian et al., 2014). The amount of tamoxifen administered to mice must be adjusted to avoid stillbirth of the pups, but still activate CreER recombination. In addition, tamoxifen does not work sufficiently on adult skin and WAT in many instances (data not shown) (Konishi et al., 2008). Therefore, although the CreER system is very powerful, a simpler and safer system is required for future studies.

We focused on mesoderm, BM and WAT in this study, although several tracings of MSC in other tissues/organs have been reported previously (Nagoshi et al., 2008; Wislet-Gendebien et al., 2012; Zhao et al., 2014; Kaukua et al., 2014; Isern et al., 2014). These studies did not investigate the derivation of MSCs in BM and WAT during the mouse development. As the in vitro features, including CFU-F and multipotency, initially identified and defined MSCs, the proper markers are required for detecting and tracing MSCs in vivo. Our results demonstrated that MSCs, which had multipotency and self-renewal capacity in vitro, expressed Pdgfrα at all times and could be traced using our mutant mice in vivo. We hope that the previous evidence is confirmed by our materials or by a similar strategy to this study in the future.

Finally, as CFU-F colonies derived from adult gWAT did not exhibit MSC multipotency in the present study, it might be necessary to investigate the most appropriate tissues to use when MSCs are derived for clinical therapy. Unfortunately, such determination was beyond the scope of this study. Several studies have discussed the application of WAT-derived MSC for clinical therapy (Lo Furno et al., 2016). Based on our results, we recommend the careful investigation of the differences in characteristics among various adult WATs before the clinical trials (Tchkonia et al., 2005; Konishi et al., 2008).

In this study, we have demonstrated that most adult BM- and WAT-MSCs originate from mesoderm, and that this is an important finding for the safety of clinical therapy. Our results thus highlight important concepts in both developmental biology and regenerative medicine.

The Ethics Committee of Kumamoto University approved all study protocols. The genetic background of all mice was C57BL/6. Mutant mice were genotyped using genomic DNA extracted from the tail tip and amnion in lysis buffer containing proteinase K (Wako). PCR was carried out as previously described (Soriano, 1999; Logan et al., 2002; Takashima et al., 2007; Kawanami et al., 2009; Madisen et al., 2010; Rinkevich et al., 2011; Imuta et al., 2013; Miwa and Era, 2015). All wild-type mice were purchased from CLEA Japan.

In order to examine differentiation potentials, we picked up single colonies of CFU-F by micropipette under a microscope and re-cultured with the MSC culture medium, which was αMEM (Gibco) containing 10% FBS and 0.1 mM 2-mercaptoethanol (Sigma) (Takashima et al., 2007; Miwa and Era, 2016).

Cells were dissociated with 0.25% Trypsin-EDTA and washed with 1% bovine serum albumin-HBSS. Following staining with biotinylated anti-Pdgfrα (eBioscience) and Streptavidin APC (eBioscience), the cells were analyzed by FACS (BD Biosciences). The results were reanalyzed using FlowJo software (Tree Star).

CFU-F assays were performed as described previously (Takashima et al., 2007; Miwa and Era, 2016). Cells were set up on six-well plates at 1000-10,000 cells/well and cultured with the MSC culture medium. The culture medium was changed every 3 days. After 14 days, colonies were stained with Leishman (Merck) and counted.

Differentiation into adipocytes, chondrocytes and osteocytes was induced as described previously (Sakurai et al., 2006; Takashima et al., 2007; Miwa and Era, 2016). Briefly, cells were cultured on 24-well plates with various induction cocktails. After a certain period of time, adipocytes were stained with Oil Red O, chondrocytes with Alcian Blue and osteocytes with Alizarin Red S.

Bone fracture was studied using a standard closed fracture model (Bonnarens and Einhorn, 1984). Before making bone fractures, 27-G needles were inserted into the tibia canals of anesthetized mice. Sections of bones were prepared using the Kawamoto method (Kawamoto and Shimizu, 2000).

Sections were fixed with 4% paraformaldehyde in PBS and then stained with anti-Ucp1 (Abcam, ab10983, 1:500) and anti-RFP antibodies (Chromotek, 5f8, 1:1000). Alexa Fluor488- and 594-conjugated secondary antibodies (Invitrogen, A-11008, A-11007, 1:50) were used for primary antibody detection.

The day of vaginal plug formation was recorded as E0.5. 4-OHT (Sigma) was dissolved in ethanol at a concentration of 100 mg/ml and then diluted in corn oil (Sigma) to a concentration of 10 mg/ml. Pregnant mice were injected intraperitoneally with 4-OHT at 50 μg/g body weight. Adult mice were injected intraperitoneally with 4-OHT once a day or fed with a powdered diet containing 500 μg/g of tamoxifen (Sigma) for 5 days.

Total RNA was purified with Sepasol Super G reagent (Nacalai Tesque) and transcribed to DNA using PrimeScript (Takara). Quantitative PCR was performed using THUNDERBIRD SYBR qPCR Mix (Toyobo) as described previously (Kitagawa et al., 2012). The data were analyzed with PikoReal 96 Real-Time PCR System (Thermo Scientific) and normalized to β-actin and Gapdh. The sequences of primers are listed in Table S1.

To detect expression of LacZ, sections were fixed and stained as described previously (Houzelstein et al., 1997). The fixative solution used was 2% formaldehyde and 0.2% glutaraldehyde in PBS. The staining solution used was 1 mg/ml X-Gal (Wako), 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM MgCI₂, 0.01% sodium deoxycholate and 0.02% Nonidet P-40 (NP-40) in PBS.

Liver fibrosis was induced by repeated intraperitoneal carbon tetrachloride (Wako) injections into mice (eight times per day every 3 days). Carbon tetrachloride was dissolved in an equal volume of corn oil (Sigma) and injected at a dose of 1 μl/g body weight. Sections stained with Masson trichrome were examined with light microscopy (Hotta et al., 2008).

Mice were anesthetized with an intraperitoneal injection of ketamine and xylazine. The dorsal surface was shaved and a disposable 0.6 cm diameter skin-punch biopsy tool was used to create a full-thickness excision wound that extended to the fascia, as described elsewhere (Goto et al., 2006). For MSC transplantation, we suspended 100,000 MSCs in 160 μl PBS and subcutaneously injected them around ulcerations.

Data are expressed as the mean±s.e.m. All experiments were independently repeated at least three times. The statistical significance of differences in mean values was assessed using Student's t-test.

We thank Dr Hiroshi Sasaki for kindly providing Brachyury-GFPCreER^T2 mice and Dr Hiroo Ueno for kindly providing Rosa26-Rainbow mice. We also thank the Center for Animal Resources and Development (CARD) of Kumamoto University for generating and housing the mutant and control mice.