Correction: Human bone marrow harbors cells with neural crest-associated characteristics like human adipose and dermis tissues

[This corrects the article DOI: 10.1371/journal.pone.0177962.].


Introduction
The neural crest (NC) is an embryonic ephemeral structure present in vertebrates and located at the margin of the neural tube [1,2]. Those neural crest-derived stem cells (NCSC) have the unique property to migrate from the dorsal neuroepithelium and invade multiple tissues. The initiation of cell migration involved an epithelial-to-mesenchymal transition (EMT) which triggers multiple molecular and cellular machineries [3]. After completing EMT, NCSC delaminate and start migrating to their final destination [4]. A multitude of neural and non-neural mature a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Human bone marrow stromal cell isolation Human bone marrow stromal cells were obtained using iliac crest puncture from 23 to 29 year-old healthy women. Cells were initially isolated and cultivated according to our previously described protocol [59,60]. Human bone marrow samples were obtained from healthy volunteer adults. Mononuclear cells were isolated by centrifugation over Ficoll Paque™ Plus (GE Healthcare, Diegem, Belgium) and washed in PBS (Lonza, Verviers, Belgium). Cells were seeded at a density of 5 x 10 5 cells/cm 2 in MSC growth medium (Lonza) and incubated into humidified incubator at 37˚C and 5% CO 2 .
Human adipose tissue and dermis-derived stromal stem cell isolation Human stromal stem cells from adipose tissue and dermis were isolated from residual tissue obtained after abdominoplasties or breast reductions on 17, 21, 23, 27 and 34 year-old healthy women. Cells were isolated and cultivated as previously described [61]. Adipose tissue stem cell preparation: immediately after surgery, adipose tissue was washed 3 times in PBS-glucose, then minced and incubated with 1% collagenase type I (C0130, Sigma-Aldrich, St-Louis, MO, USA) during 1 hour at 37˚C and 5% CO 2 ; 5 mg/ml of DNase (2 000 Kunit/ml) (AMPD1 Sigma-Aldrich) was added in order to decrease viscosity. Mature adipocytes and undigested fragments were separated from the pellet containing stromal vascular fraction (SVF) by centrifugation at 500 g, during 15 minutes. The SVF pellet was re-suspended in MesenCult culture medium (MesenCult, Stem Cells Technologies, Grenoble, France) supplemented with penicillin-streptomycin in 25 cm 2 T-flask (Corning, New York, NY, USA) into a humidified incubator at 37˚C and 5% CO 2 . Skin-derived stem cell preparation: after removal of epidermis and hypodermis, dermis was washed 3 times in PBS-glucose. 3 mm 2 -pieces were cut from the entire dermis explant and placed into 24-well plates. Dermis pieces were covered by culture medium (MesenCult culture medium with penicillin-streptomycin) and incubated into humidified incubator at 37˚C and 5% CO 2 . Explants were cultivated for one week, allowing NCSC migration from explants, then removed.

Stromal stem cell culture
Bone marrow, adipose tissue and skin-derived stem cells were all cultivated after isolation in MesenCult culture medium. After reaching confluence, cells were resuspended using 0.05% Trypsin-EDTA (Life Technologies, Carlsbald, CA, USA), and then sub-cultured (100 000 cells/ 25 cm 2 (Corning) at 37˚C, in a humid atmosphere containing 5% CO 2 . All experiments were performed on cells at passage 2 to 8.

IPS-derived neural stem cells
Those cells were used as negative control for chicken injection and migration experiments. Those cells were a gift from Anna Falks laboratory at the Karolinska institutet. Those NSC from human iPSC cells were cultured in DMEM/F12 medium supplied with EGF and FGF2.

CFU assay
Colony forming unit (CFU) potential was assessed for bone marrow, adipose tissue and skinderived stem cells. 50 cells/cm 2 were cultivated in 9 cm 2 dishes (Thermo Scientific, Waltham, MA, USA) during 14 days at 37˚C, in a humid atmosphere containing 5% CO 2 ; Mesencult culture medium was changed twice a week. Number of colonies is counted after 14 days.

Cell injections into chick embryos
Experiments performed on chicken embryos were carried out in accordance with Karolinska Institutet ethical committee rules. Chick embryos were obtained from Gallus gallus domesticus fertilized eggs (E S F-Produkter i Estuna AB) and stored at room temperature. Before injection, eggs were incubated during 72 hours at 38.5˚C in a humidified incubator (60% humidity), they were selected at Hamburger-Hamilton stage 18 (HHSt18), based on their morphology. We injected 0.1% (w/v in PBS) Fast Green (Amresco, VWR, Sweden) under the chorioallantoic membrane to aid visualization of the underlying embryo. Using an electrochemically sharpened tungsten wire we opened the vitelline membrane and introduced a small lesion in the thoracolumbar region directly dorsolateral to the NT, in-between two somites and thus directly in the neural crest migration stream [12,67]. About 300 adherent cells in suspension or one sphere (corresponding approximately 300 cells) are placed directly in the lesion. The egg was sealed and incubated 38.5˚C in 60% humidity. Three days later (HHSt30), embryos were removed from the shell, decapitated and fixed in PBS containing 2% PFA and 15% (w/v) sucrose in PBS for 2 hours at 4˚C with gentle agitation, before incubation overnight at 4˚C in 30% (w/v) sucrose in PBS. Embryos were frozen in Tissue O.C.T. Q-Path (Labonord) and 14 μm embryo slices were cut using cryostat microm HM 560 (LCAM-van Leeuwenhoek Centre for Advanced Microscopy, Amsterdam, Netherland).

Image acquisition and analysis
Image acquisition and analysis were either performed using a Zeiss AxioImager Z1 epifluorescence microscope (Zeiss, Zaventem, Belgium) coupled with FluoUp software; or using NIKON A1R hybrid resonant confocal microscope couple with NIS-Element software. The digitized images were adjusted for brightness and contrast using the NIH program ImageJ (Wayne Rasband, National Institute of Mental Health, Bethesda, MD, USA).

Chemotaxis assay
In order to investigate migration phenomenon observed in chick embryo we developed chemotaxis assay using human cells from adipose tissue, bone marrow and skin and chemokines like stromal cell-derived factor 1 (SDF-1), SCF and Neurotrophin-3 (NT3). 12h before starting the experiment, human cells were cultivated in serum-free DMEM. In a first time ATSC, BMSC and SKSC were labeled with Cell Tracker Green (CTG) (Life Technologies) in serumfree DMEM. 25 000 cells were then placed on 5.2 mm-diameter filters (each containing 25 000 10 μm-pores) (ChemoTx, NeuroProbe, Gaithersburg, MD, USA). Below the filter, in a bottom chamber, we placed 30 μl of DMEM containing chemokines. The plate was incubated for 48h at 37˚C. After incubation, non-migrating cells were removed from the top of the filter. Finally using fluorescence microscopy, we then used ImageJ software to quantify the percentage of filter area occupied by ATSC, BMSC or SKSC in response to different concentration of chemokine cocktail. Control condition contains on the bottom chamber DMEM alone and cells on the top compartment.

Reverse transcription and quantitative polymerase chain reaction (RT-qPCR)
RNA from adherent cells and spheres was extracted using TRIzol reagent (Life Techonologies, Carlsbad CA, USA). RNA amount was quantified using NanoDrop spectrophotometer (Thermo Scientific). After DNase treatment (Promega, Madinson, WI, USA), cDNA synthesis was carried out using Moloney-murine leukemia virus (MMLV) Reverse transcriptase (Promega) and random hexamer primers (Promega), following the manufacturer's instructions. cDNA product was then use for real-time PCR in a LightCycler 480 (Roche, Belgium) using SYBR Green Master Mix (Eurogentec, Liege, Belgium). Gene expression levels were normalized using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) housekeeping gene. Relative expression levels were obtained using ΔΔCt method. Amplifications were performed in triplicates, and on three biological replicates. Specific primers are listed in Table 1 (IDT, Leuven, Belgium).

Statistical analysis
Data were analyzed using GraphPad InStat software (GraphPad Software, Inc., La Jolla, CA, USA, www.graphpad.com). Data are reported as mean ± standard error of mean (SEM), with n corresponding to the number of donors for each cell type. Level of statistical significance was set at p < 0.05. Statistical analysis for RT-qPCR results was performed on square-root transformed data.

Characteristics of NCSC and MSC isolated from mouse adult bone marrow
To isolate pure NCSC and MSC populations from bone marrow, we used a clonal selection approach [18] in which bone marrow stromal cells from Wnt1-Cre/R26R-LacZ mice were individually seeded into 96-well plates (mean density of 0.7 cell per well). Each clone was then classified according to its embryonic origin, based on the beta-galactosidase expression. MSC were beta-galactosidase-negative, whereas NCSC were beta-galactosidase-positive (Fig 1a and  1e). After further characterization, we observed that NCSC were positive for transcription factor Sox10, the intermediate filament protein Nestin and the neurotrophin receptor p75 NTR (Fig 1f-1h). MSC were negative for sox10, weakly positive for p75 NTR and nestin (Fig 1b-1d). More interestingly, only NCSC were able to grow in suspension as spheres (Fig 1i-1l). Altogether, both NCSC specific marker expression and sphere-forming ability could be used as a tool to specifically identify NCSC in bone marrow. As genetic tracking markers are obviously non available in human, we decided to use those specific characteristics to assess the presence of NCSC in the human bone marrow.

Characterization of human stem cells according to the International Society for Cell Therapy standard protocol, for mesenchymal stem cells
Since the embryonic origin of stromal stem cells is less evident to assess in human, we decided to use dermis (Skin derived stem cells, SKSC) and adipose tissue-derived stem cells (ATSC) as positive controls. Indeed, those two populations have already been described as heterogeneous populations containing neural crest derived cells [28][29][30][31]43]. SKSC and ATSC were isolated Table 1. List of primers used in quantitative RT-PCR.

Gene
Sequence https://doi.org/10.1371/journal.pone.0177962.t001 from healthy patients having breast reduction or abdominoplasty; bone marrow stromal cells (BMSC) were isolated from iliac crest aspiration, from healthy volunteers. BMSC donors were healthy men and women aged approximately 20 years old (Fig 2b). ATSC (Fig 2a) and SKSC (Fig 2c) were obtained from surgery residual material, from female donors aged from 18 to 52 years old. In order to check if donor gender and/or age could influence our results, we compared sphere forming ability and colony forming unit (CFU) potency between our different donors, in each cell type. No statistical significance was observed neither from sphere-forming nor CFU abilities in ATSC, BMSC and SKSC populations. We cultivated ATSC, BMSC and SKSC in adherent culture conditions, for a maximum of 8 passages. According to the recommendations issued by International Society for Cellular Therapy guidelines (ISCT) [44], using flow cytometry phenotyping, we observed that more than  Fig 3o to 3u). As we could expect, it appeared that those guidelines efficiently discriminate hematopoietic from non-hematopoietic cells, but would not be useful to discriminate neural crest-derived cells from the entire population.

Comparison of mouse and human cells with regard to NCSC-related characteristics
As previously described for mouse NCSC and MSC (Fig 1), we characterized the three human cell types (BMSC, ATSC and SKSC) using NESTIN, SOX10 and P75 NTR expression. Surprisingly, all three populations were homogeneously positive for NESTIN (Fig 4a to 4c), but negative for SOX10 (Fig 4e to 4g). The only difference we observed concerned P75 NTR expression, since SKSC were P75 NTR -positive (Fig 4k), while BMSC and ATSC were weakly P75 NTR positive (Fig 4i to 4k). We also performed these immunostainings using a positive control for NC identity (MeWo cell line) and as expected NESTIN (Fig 4d), SOX10 (Fig 4h) and P75 NTR (Fig  4l) were expressed. We then decided to analyze their ability to grow as spheres, and tested a sphere-forming protocol on those three human cell types. As a reminder, we previously described that only NCSC isolated from mouse adult bone marrow were able to grow as sphere (Fig 1) compared to adult bone marrow derived MSC. This property could therefore be used as a tool to discriminate mouse NCSC from mouse MSC. Using light and transmission microscope we were not able to differentiate adherent cells from adherent cells obtained after re-adherence from spheres (Fig 5a and 5b). As observed on Fig 5c, all cell types were able to grow as spheres. Those spheres reached a diameter from 80 to 350 μm. We also assessed the sphere-forming rate (Fig 5d) and it appeared that SKSC presented a significantly higher ability to grow as spheres compared to BMSC (p<0.0001) and ATSC (p<0.05). The proportion of cells growing as spheres was based on the total number of obtained spheres, after 12 days of culture, compare . ATSC (a) and NCSC (c) were obtained using residual material form abdominoplasties or breast reductions performed on healthy women from 18 to 52 years old. We compared the ability to grow as sphere and the CFU potential between donors from the same cell type and we did not observed statistically significant differences between them. Statistical analysis: one way ANOVA followed by HSD post hoc test. to the number of initially plated cells. We further characterized those spheres, which appeared to be NESTIN-positive (Fig 5e to 5h), p75 NTR -positive (Fig 5i to 5l) and SOX10-negative (Fig  5m to 5p). These observations were confirmed at the mRNA level using quantitative RT-PCR. Indeed, NESTIN (Fig 5h) and P75 NTR (Fig 5l) were highly expressed in spheres compare to adherent cells (a difference was observed in ATSC for both genes (p<0.0001), in SKSC for NESTIN (p<0.0001) and in BMSC for P75 NTR (p<0.0001). As observed on Fig 5p, we detected a weak expression level for SOX10 in the different culture conditions, whereas we never detected any signal by immunofluorescence. However, the expression level was really weak (p<0.0001) compared to the expression level observed in MeWo cells. This last was human malignant melanoma cells, and was the only cell type expressing SOX10 at the protein level (Fig 4h). Altogether, it appeared that mouse MSC and NCSC features were not fully Human bone marrow harbors cells with neural crest-associated characteristics transposable to their human counterparts, and further characterization was therefore required in order to attest the potential presence of NCSC in human adult bone marrow.

Differentiation abilities of human bone marrow, adipose tissue and dermis-derived stem cells
In order to determine if human BMSC, ATSC and SKSC are able to differentiate into classical mesenchymal derivatives such as adipocytes, osteocytes and chondrocytes, but also into neural crest-derived cells like Schwann cells, melanocytes, and neurons, we placed those cells in different specific culture conditions in order to induce their differentiation (as described in experimental procedures section). As expected, the three cell types were able to differentiate into adipocytes (Oil-Red-O labeling, Fig 6a-6c), osteoblasts (Alizarin-Red labeling , Fig 6d-6f) and chondrocytes (Alcian Blue , Fig 6i-6k). Concerning osteoblasts and chondrocytes, we also performed quantitative RT-PCR for RUNX2, ALPL, AGGRECAN and BARX2 (Fig 6g, 6h, 6l and 6m) to validate their differentiated stage compare to undifferentiated cells.
Schwann cell differentiation was attested by S100-β (Fig 7a to 7c), MBP (red) and P0 (green) labeling (Fig 7d to 7f), as well as by quantitative RT-PCR for S100-β, PMP22, MBP and P0 genes (Fig 7g to 7j). Interestingly, S100-β was detected in a large amount of cells (observed at protein level, but also with the higher mRNA expression level compared to other genes) whereas only a few S100-β positive cells were also MBP and P0 positives. Indeed, these two markers were specific for Schwann cells at later differentiating stages. These results were confirmed at the mRNA level (Fig 7g to 7j). Melanocyte differentiation was attested by TYRP1 labeling (Fig 7k and 7l), as well as by quantitative RT-PCR for MITF and TYRP2 (Fig 7n and  7o). Quantitative RT-PCR for PMP22, P0 and MBP (Fig 7h to 7j) and MITF and TYRP2 (Fig  7n and 7o) confirmed observations previously performed at the protein level, concerning their differentiating potential. Neuronal differentiation was attested by NeuN (green) and Neurofilament (red) expression (Fig 7p to 7r). Altogether, it appeared that those three human cell populations (ATSC, BMSC and SKSC) were able to differentiate into mature mesenchymal cell types (adipocytes, osteoblasts and chondrocytes) and neural crest-derived cells (Schwann cells, melanocytes and neurons). Similarly to immunostainings performed in mouse (Fig 1), human ATSC, BMSC, SKSC expressed NESTIN (a-c), did not express SOX10 (e-g). ATSC and BMSC express weakly P75 NTR (i-j) in comparison to SKSC (k). We also used a cell line derived from human malignant melanoma cells (MeWo), which was a positive control for NC identity (d, h, l). (Scale bars = 20μm). Transcriptomic and proteomic characterization using NCSC-related markers The next step of this study was to characterize the expression profile of several markers that are classically expressed by NCSC. We first analyzed the expression of the following genes, using quantitative RT-PCR: FOXD3, PAX3, SLUG, SNAIL1, SOX9 and TWIST. Those genes are involved as neural crest specifiers and/or as neural plate border specifiers. Indeed, NGN1 is involved in neural differentiation; BRN3A is associated with PNS development; and finally MSI1 is linked with neural precursors. As we can observe in Fig 8, most of these transcripts The majority of ATSC, BMSC and SKSC were able to grow as sphere, in specific culture medium. Using light and transmission microscope, we were not able to distinguish adherent cells from re-adherent sphere-derived cells (a and b). Spheres presented a diameter between 80 and 350 microns (c). We quantified the ability to each cell type to grow as spheres, after 12 days of culture compare to the initial number of plated cells (75 000 cells in 25 cm 2 T-flask). SKSC presented a significantly higher ability to grow as spheres compared to BMSC and ATSC (respectively p<0.0001 and p<0.05 one way ANOVA followed by HSD post hoc test) (d). Similar immunostaining characterizations as performed in Fig 4 were performed on sphere-derived cells. In this case, human spherederived ATSC, BMSC and SKSC expressed NESTIN (e-h) and did not express SOX10 (m-p). However, unlike adherent cells, sphere-derived ATSC, BMSC and SKSC expressed P75 NTR (i-l). Quantitative RT-PCR were performed on adherent and sphere-derived cells and confirmed immunological observation made at mRNA level (h, l, p). Data were normalized to SKSC adherent cells expression level which was set as 1. Statistical analysis: one way ANOVA followed by HSD post hoc test. **** means p<0.0001. (Scale bars = 50μm). https://doi.org/10.1371/journal.pone.0177962.g005 Human bone marrow harbors cells with neural crest-associated characteristics were overexpressed in spheres compared to adherent cells in the three cell types (ATSC, BMSC and SKSC). We confirmed those results at the protein level for SOX9 (Fig 9a to 9g) and TWIST (Fig 9h to 9n). Interestingly, in adherent cells, we observed heterogeneous populations for several markers: 25.7% (±0.9%) of adherent ATSC, 20.5% (±2.7%) of adherent BMSC and 29.8% (±2.6%) of adherent SKSC expressed SOX9 (Fig 9d) and 42% (±2.7%) of adherent ATSC, 33,4% (±3.6%) of adherent BMSC and 44,2% (±1.7%) expressed TWIST (Fig 9k). Furthermore, more than 80% of the cells, in spheres, expressed SOX9 and TWIST. Altogether, those results reinforce the hypothesis that neural crest-derived cells are present in human dermis, bone marrow and adipose tissue. They also suggest that spheres are an enriched population of neural crest cells (NCC).

Characterization of bone marrow stem cell migrating abilities when injected into chicken embryos
To finalize our characterization, we decided to test the migrating properties of those three cell types, once injected into chicken embryos. Indeed, it has been well described [12,45,46] that ATSC, BMSC and SKSC were able to differentiate into Schwan cells as the majority of the cells were stained with anti-S100β antibody (a-c) and a small amount of S100β + cells was also stained with anti-MBP (red) and anti-P0 (green) antibodies (d-f). Quantitative RT-PCR for S100β, PMP22, MBP and P0 confirmed the results obtained by immunofluorescences, at mRNA level (g-j). The three cell types were also able to differentiate into melanocytes stained with anti-TYRP1 antibody (k-m), which was confirmed at mRNA level based on quantitative RT-PCR for MITF (n) and TYRP2 (o). Finally, neuronal differentiation was assessed using anti-NeuN (green) and anti-Neurofilament (green) antibodies (p-r). Data were normalized using SKSC undifferentiated cells expression level set as 1. Statistical analysis: one way ANOVA followed by HSD post hoc test. * means p<0.05, ** means p<0.005, *** means p<0.0005. (Scale bars a, b, c, d, e, f, p, q and r = 50μm; Scale bars k, l,and m 20μm).
https://doi.org/10.1371/journal.pone.0177962.g007 when injected into chicken embryos at a specific stage, NCSC are able to migrate and follow chicken NCSC migration pathways. More precisely, this experiment consisted in the injection of a small amount of cells into the dorso-lateral aspect of the thoraco-lumbar region of Hamburger and Hamilton stage 18 (HHst18) chicken embryos (corresponding to the early period of NC migration from the dorsal NT at the trunk level). We therefore injected around 300 adherent cells or one sphere into HHst18 chicken embryos. The embryos were sacrificed 72 hours post-injection. Injected cells and their derivatives were traced using a specific marker for human cells (i.e. anti-HUMAN NUCLEI marker). At this developmental stage, trunk NCSC could give rise to two main populations [47]: 1) sensory neurons and glial cells of the dorsal root ganglion (DRG) and associated nerves; 2) melanocytes of the skin. At this stage, the boundary cap cells are known to be NC-derived cells forming clusters (that will migrate later) at the surface of the NT [48]. Some injected cells were observed in the fiber tracts leaving DRG, which also include NCSC. In this context, the three cell types (adherent ATSC, BMSC and SKSC) were localized into the peripheral neural crest targets as described thereafter: most of the cells were migrating to the DRG (Fig 10a to 10c; S1a to S1c Fig), but some were also found in the area of the boundary cap (Fig 10d to 10f, S1d to S1f Fig), in the dermal layer of the skin (a few cells were observed at the surface of the embryo corresponding to melanocyte migration territory) (Fig 10j to 10l; S1j to S1l Fig) and finally in nerve fiber tracts leaving the DRG (Fig 10m to 10o, S1m to S1o Fig). Cells were also found in the injection site illustrating Here again, sphere-derived cells were detected in the DRG (Fig 11a to 11c, S2a to S2c Fig), in the area of the boundary cap (Fig 11d to 11f, S2d to S2f Fig), in the injection site (Fig 11g to  11i, S2g to S2i Fig), in migration territory of the melanocytes (Fig 11j to 11l, S2j to S2l Fig) and in the nerve fiber tracts leaving the DRG (Fig 11m to 11o, S2m to S2o Fig). Specificity of the migration was addressed by the injection of iPS-derived NSC (S3 Fig). In those conditions, we observed that iPS-derived NSC were also able to migrate to the DRG, but they also colonized the neural tube which was never observed with ATSC, BMSC or SKSC.
Altogether, we conclude that a subset of BMSC, ATSC and SKSC was able to migrate along neural crest migration pathways towards tissues classically colonized by neural crest derivatives. We also confirmed those migrating properties by performing an in vitro chemotaxis assay. Chemokine cocktail made of SDF-1, SCF and NT3 was used to attract adherent ATSC, BMSC and SKSC in order to mimic chicken embryo environment. We observed that human cells (from the three tissue origins) were responsive to this cocktail and migrate through the filter (Fig 12a to 12c). Human bone marrow harbors cells with neural crest-associated characteristics

Discussion
The NC is a transient ephemeral structure containing cells that share the unique property, in vertebrates, to migrate, invade multiple tissues and give rise to diverse mature cell lineages. Most NC development knowledge derived from studies performed on chicken and zebrafish. It is obvious that human NC information is more difficult to obtain because of its transient nature and the limited availability of human fetal cells. In mouse, adult post-migratory NCSC were identified in several locations like gut, cornea, heart, bone marrow, adipose tissue and skin [49]. In human, the confirmation of the presence of neural crest cells in adult tissues is less obvious and mainly based on their expression profile, associated with their differentiating capacity and their ability to follow migration pathways when injected in chicken embryos. In those conditions, adult human NCSC have been identified in skin [28][29][30][31][32], hair follicle [30,31], in dental pulp [33][34][35][36], periodontal ligament [37,38], in cornea [39], inferior turbinate [40], olfactory or oral mucosa [41,42] and in adipose tissue [43].
In this study, we analyzed the putative presence of neural crest-derived cells in adult human bone marrow (BMSC). We compared human bone marrow cells with human adipose tissue (ATSC) and dermis (SKSC), two tissues already described as potential sources of neural crestderived cells. Those three cell types proliferated in the same culture conditions. All cell types were CD73 + , CD90 + , CD105 + and NESTIN + . Moreover, all three populations were heterogeneous and showed some positivity for BRN3A, FOXD3, MSI1, NGN1, PAX3, SLUG, SNAIL1, SOX2, SOX9, and TWIST. Finally, sphere-derived cells showed a higher expression level of those genes, suggesting a potential enrichment in neural crest derived cells, similarly to our observations for mouse bone marrow derived NCSC. The Table 2 represents sphere forming ability and relative expression level of the three human populations studied here. Our results are in accordance with Abe and collaborators studies focusing on human putative NCSC isolated from human apical dental pulp [34] or even oral mucosa stroma [42]. Indeed, in these two studies it has been shown that: 1) 0.023% of human oral mucosa stromal cells are able to grow as spheres. 2) Neurospheres, obtained from dental pulp or oral mucosa, were expressing NESTIN, MSI1, P75 NTR , SNAIL1, SLUG and SOX9.
Even if it is well accepted that sox10 transcription factor is a classical NCSC marker for mouse embryonic and adult cells, in human adult cells, SOX10 expression is not systematically observed. Indeed, in Fournier et al., 2016, these authors described human gingival NCSC. Those cells are NESTIN, SNAIL1, TWIST, PAX3, SOX9 and FOXD3 positive and able to generate neurospheres [50]. Abe et al., 2016 described sphere-derived multipotent progenitor cells obtained from human oral mucosa, that were enriched in NCSC [42]. In that study, the authors characterized those cells as NESTIN, CD44, SLUG, SNAI1, MSX1, HES1, and SOX9 positive, but not SOX10. Hara et al., 2014 described potential NCSC isolated from human corneal endothelial progenitor cells that were P75 NTR , SOX9, and AP-2β positive [39]. Bueno et al., 2013 described human adult periodontal ligament-derived cells that expressed stem cell Chemotaxis assay on human adherent ATSC, BMSC and SKSC. Chemotaxis assay were build to confirmed migration abilities observed in chick embryos. We placed ATSC, BMSC or SKSC on top of a 10 μm-filter, with cocktail of SDF-1, SCF and NT3 in the bottom chamber. After 48 hours, migration rate was evaluated by the quantification of the filter area occupied by human cells. Passive migration was observed and quantified in control condition (only serum-free DMEM in the bottom chamber) and this value was used to normalize the results obtained with different cocktail concentrations (Table). Statistical analysis: (n = 3), one way ANOVA followed by HSD post hoc test. * means p<0.05, ** means p<0.005, *** means p<0.0005, **** means p<0.0001.
https://doi.org/10.1371/journal.pone.0177962.g012  [40]. The lack of SOX10 expression in our putative adult human bone marrow neural crest derived cells was surprising, as SOX10 transcription factor is commonly used as a NC marker. However, several facts could justify the absence of SOX10 expression. 1) SOX10 expression is mainly described in NC cells at the time of their emigration and is essential for their selfrenewal and survival [49]. After reaching their final destination, SOX10 expression may disappear, depending on their environmental condition [53]. 2) Adult rodent bone marrow NCSC were systematically sox10 + , while adult human bone marrow NCSC appeared to be SOX10 negative. This non-transposability of mouse results to human has already been described for NESTIN. Indeed, in mouse, it has been demonstrated that most NESTIN-positive cells from adult bone marrow are neural crest derived cells [18,54]. However, in human, all bone marrow stromal cells (including MSC and NCSC) are NESTIN + [55,56]. 3) As in many cell lineages and tissues, SOX10 and SOX9 are functionally redundant [57]. It is therefore possible that in human adult bone marrow NCSC, SOX9 expression could substitute for SOX10 expression  Similarly, it has been demonstrated that, in some conditions, SOX10 expression could be down-regulated by SOX9 [58]. Indeed, those authors demonstrated that SOX9 was able to interact with SOX10 promoter and regulated its level of expression.
The phenotypic characterization of ATSC, BMSC and SKSC was followed by the assessment of their ability to differentiate in neural crest-derived mature cell types and to follow the neural crest migration track when injected into chicken embryos. In this last experiment, all cell types were able to migrate and follow endogenous neural crest migration pathways, as all cell types were localized in dorsal root ganglia, in skin, and in boundary cap, but none of them were observed in the neural tube or any specific location of neural cell types attesting of the specificity of the migration.
Altogether, and as summarized in Fig 13, the phenotypic characterization, the differentiating and the migrating abilities of ATSC, BMSC and SKSC strongly suggest the presence of neural crest-derived cells in human bone marrow, as it has already been described for skin and adipose tissue. Belgium, and by a grant of Leon Fredericq Foundation. François Lallemend is supported by the Swedish Medical Research Council, Karolinska Institute, Ragnar Söderberg foundation, Knut and Alice Wallenberg Foundation, Åke Wiberg foundation and the Swedish Brain Foundation. We thank Professor Nizet from CHU de Liège for providing us residual material after abdominoplasties and breast reductions. We would like also to thank Dr Ormenese and the GIGA platform of flow cytometry and cell imaging for their valuable. We also would like to thank Dr Mansoureh Shahsavani, in Pr Anna Falk's laboratory for giving us iPS-derived NSC.

Author Contributions
Conceptualization: SW AG.