Defective Proliferation and Osteogenic Potential with Altered Immunoregulatory phenotype of Native Bone marrow-Multipotential Stromal Cells in Atrophic Fracture Non-Union

Bone marrow-Multipotential stromal cells (BM-MSCs) are increasingly used to treat complicated fracture healing e.g., non-union. Though, the quality of these autologous cells is not well characterized. We aimed to evaluate bone healing-related capacities of non-union BM-MSCs. Iliac crest-BM was aspirated from long-bone fracture patients with normal healing (U) or non-united (NU). Uncultured (native) CD271highCD45low cells or passage-zero cultured BM-MSCs were analyzed for gene expression levels, and functional assays were conducted using culture-expanded BM-MSCs. Blood samples were analyzed for serum cytokine levels. Uncultured NU-CD271highCD45low cells significantly expressed fewer transcripts of growth factor receptors, EGFR, FGFR1, and FGRF2 than U cells. Significant fewer transcripts of alkaline phosphatase (ALPL), osteocalcin (BGLAP), osteonectin (SPARC) and osteopontin (SPP1) were detected in NU-CD271highCD45low cells. Additionally, immunoregulation-related markers were differentially expressed between NU- and U-CD271highCD45low cells. Interestingly, passage-zero NU BM-MSCs showed low expression of immunosuppressive mediators. However, culture-expanded NU and U BM-MSCs exhibited comparable proliferation, osteogenesis, and immunosuppression. Serum cytokine levels were found similar for NU and U groups. Collectively, native NU-BM-MSCs seemed to have low proliferative and osteogenic capacities; therefore, enhancing their quality should be considered for regenerative therapies. Further research on distorted immunoregulatory molecules expression in BM-MSCs could potentially benefit the prediction of complicated fracture healing.


A defective proliferation and osteogenic potential of uncultured and minimally-cultured BM-MScs.
Following the sorting of BM cells, the growth factor receptors and osteogenic markers were tested on isolated uncultured CD271 high CD45 low cells. Interestingly, fewer transcript levels of EGFR, FGFR1 and FGRF2 were detected in uncultured NU-CD271 high CD45 low than U-cells (p = 0.049, p = 0.021, and p < 0.001 respectively, Fig. 2a and Table 1) indicating potentially a low response to these growth factors. When measuring the growth factors related to angiogenesis, the VEGFA transcripts (but not VEGFC or ANGTP) were lower in NU-CD271 high CD45 low cells than U-cells (p = 0.032, Fig. 2a and Table 1). To test if serum has an impact (systemic) on the proliferation of BM-MSCs, BM cells were seeded in cultures with either NU-or U-serum then the levels of 18 s were assessed (without cell passage) as indicative of cell quantities. Noticeably, the donor-matched comparison showed that both NU-and U-passage-zero BM-MSCs had reduced 18 s levels when in NU serum cultures than those in U-serum cultures (p = 0.031 for both, Fig. 2b) indicating a negative effect of NU serum on MSC proliferation. www.nature.com/scientificreports www.nature.com/scientificreports/ The PCR data showed significantly lower levels of osteogenic markers, ALPL, BGLAP, SPARC and SPP1, but not LepR in uncultured NU-CD271 high CD45 low cells than U-cells (p = 0.040, p = 0.049, p = 0.043 & p = 0.047 respectively, Fig. 2c and Table 1) suggesting an impaired osteogenic potential for NU BM-MSCs. When cultured in NU-serum, passage-zero NU BM-MSCs had consistently less ALPL transcripts relative to U-serum cultures (p = 0.008) and U-MSCs in both serum cultures (both p = 0.029, Fig. 2d). However, BGLAP, SPP1 and SPARC transcripts were comparable in both U and NU serum-cultures (Fig. 2d). In summary, reduced the proliferative potential of NU BM-MSCs was noted, and apparently due to both cellular and serum-related causes. Also, osteogenic markers were less expressed on native BM-MSCs, but that was not as evident in serum-supplemented cultures.  U and NU BM-MSCs had similar functions when culture-expanded. The uncultured and minimally-cultured BM-MSCs showed reduced proliferative capacity, lower expression of osteogenic markers with altered expression immunoregulatory markers. To test if these detected changes have a functional impact on BM-MSCs, culture expansion of these progenitors was performed to get enough cells for the functional assays. The culture-expanded U-and NU BM-MSCs with/without cytokines had no significant differences in the XTT absorbance ( Fig. 4a) indicating similar in vitro proliferative capacity. To further confirm these data, BM-MSCs were loaded on scaffolds, and the numbers of NU-and U-MSCs (counted as CD45 − CD73 + CD90 + CD105 + cells using flow-cytometry, 48 ) were similarly increased after 3-week cultures relative to 1-week cultures (p = 0.007 and p = 0.003 respectively, Fig. 4b) confirming similar proliferation for culture-expanded NU-MSCs and U-MSCs. For testing the osteogenic differentiation, culture-expanded BM-MSCs were cultured for 3 weeks in osteogenic medium with/without cytokine treatment then the calcium deposition levels were measured. While IL-1 and IL-17 (but not IFN-γ, and TNF-α) seemed to induce calcium levels in differentiating NU and U BM-MSCs (Fig. 4c), measured calcium levels were comparable for NU and U BM-MSCs either with or without cytokine treatments (Fig. 4c). These data demonstrated the similar osteogenic capacity of culture-expanded NU and U BM-MSCs.
For immunosuppression, cytokine-treated culture-expanded MSCs were assessed for the intracellular IDO levels using flow-cytometry (Fig. 4d). As expected, the IDO levels were similarly induced when MSCs were treated by IFN-γ alone or combined with TNF-α, IL-1 or IL-17 (all p < 0.001, Fig. 4e). Importantly, these induced IDO levels were similar for NU-and U-MSCs (Fig. 4e). Also, the percentage of culture-expanded NU-and U-MSCs expressing LAP (surface TGF-β1) were similarly increased after cytokine treatments (p = 0.040 p = 0.006, respectively, Fig. 4f). These results indicated comparable immunosuppressive functions of culture-expanded NU-and U-MSCs. www.nature.com/scientificreports www.nature.com/scientificreports/ The expression of cytokine receptors on BM-MSCs and comparable serum cytokine levels during early fracture healing. As the minimal cultures data indicated that NU BM-MSCs were less immunosuppressive when cytokine-treated compared to U-MSCs, we next tested if the cytokine receptor levels could be a contributing factor. Interestingly, the uncultured NU-CD271 high CD45 low cells expressed fewer transcripts of IL-1R1 than U-CD271 high CD45 low cells (p = 0.005, Fig. 5a). However, no significant differences were detected for other cytokine receptor transcripts (Fig. 5a). Consistently, the surface protein expression of the cytokine receptors was similar for uncultured NU-and U-CD271 high CD45 low cells except for IL-1R1, which were less expressed on the surface of NU-CD271 high CD45 low cells (p = 0.049, Fig. 5b). The plots show the correlation between the transcript levels of BST2 versus those of osteogenic and growth factor receptors in CD271 high CD45 low cells. Spearman r test was used for the correlation (n = 17). (c) The plots show the correlation between the transcript levels of S100A8 versus those of osteogenic and growth factor receptors in CD271 high CD45 low cells. Spearman r test was used for the correlation (n = 17). (d) The figure presents the mean of relative gene expression of IDO, PTGES2 and TGF-β1 when passage-zero BM-MSCs were treated with a mixture of IFN-γ, TNF-α, and IL-1 in U-serum or NU-serum supplemented cultures. The Kruskal-Wallis test was used to compare the groups (n = 3).  www.nature.com/scientificreports www.nature.com/scientificreports/ To explore why contrarily culture-expanded BM-MSCs have similar immunosuppressive functions, the surface cytokine receptor expressions were compared between donor-matched culture-expanded MSCs and uncultured CD271 high CD45 low cells. The data showed various changes in all receptors surface expression (mainly a decrease after culture-expansion) in both U-MSCs and NU-MSCs (Fig. 5c). Furthermore, comparable cytokine receptor levels were noted between culture-expanded NU-and U-MSCs (Fig. 5d) verifying changes of MSC surface phenotype during in vitro culture-expansion.

Non-union patients
We further aimed to understand if serum cytokines levels are linked to the decreased proliferative potential and altered immunosuppressive phenotype of native BM-MSCs. Therefore, the levels of immunosuppression-related cytokines were measured in serum of fracture patients relative to control levels. The ELISA data showed that IFN-γ, TNF-α, and IL-1 levels were not different between U-sera, NU-sera and control sera (Fig. 6a-c). These findings indicated no systemic changes in these cytokine levels during fracture healing. Noteworthy, the measured IL-17 levels were alike for U and NU fractures during early healing. However, significant lower IL-17 levels were detected for late healing NU-serum than healed U-serum and control levels (p < 0.001 and p < 0.001, respectively, Fig. 6d). Collectively, apart from IL-17 measured in NU serum at late healing, comparable serum cytokine levels were noted for NU and U.

Discussion
The application of autologous BM to boost the healing of non-united fractures as a source of osteoprogenitors has gained lately a great popularity [49][50][51][52] . However, the quality of these BM-MSCs is poorly understood. In this study, we comprehensively compared various functional potentials of BM-MSCs from NU and U fracture patients utilising multiple assays and at different cell culture conditions. The functional potential of uncultured CD271 high CD45 low cells and minimal MSC cultures have shown significant differences between NU and U BM-MSCs, implying defective multifunction of NU BM-MSC. However, the functions of culture-expanded MSCs were similar. This masking of cell differences within the culture-expanded cells is most likely due to in vitro-related changes in cell phenotype such as those changes we detected for the expression of the surface cytokine receptors. Similarly, adipose-derived MSCs showed significant variations in cell cycle and functional transcripts during culture-expansion 53 . We think that foetal calf serum (FCS) could be a contributing factor to phenotype changes, particularly with previous research has shown that human serum, unlike FCS, can maintain the MSC genomic patterns 54 .
With regards to the proliferative capacity, we identified significant lower transcripts levels of EGF and FGF receptors for uncultured NU CD271 high CD45 low cells than U cells. These findings implied a weak response to EGF and FGF, which are known to be essential for the MSC proliferation and survival, particularly in fractures [55][56][57][58] . Also, a role for EGF in fracture healing has been suggested when its serum levels were found increased in patients with a brain injury combined with a limb fracture than those having brain injury only 59 . In agreement with our findings, a previous study has reported downregulation of FGF-R2 gene expression in human non-union osteoblast cultures 35 . In addition to the MSC response to growth factors, we noted an inadequate proliferation of passage-zero BM-MSCs in cultures supplemented with NU-serum compared to U-serum cultures inferring lower growth inducers in NU-serum. Our group previously showed that the proliferative capacity of BM-MSCs from fracture patients was positively correlated with serum growth factor levels, such as platelet-derived growth factors (PDGFs) 55 . In total, it seems that the proliferation of BM-MSCs in non-union patients could be defective due to both intrinsic cellular and extrinsic microenvironmental factors. Such complicated alterations of the NU www.nature.com/scientificreports www.nature.com/scientificreports/ BM-MSC proliferation could be potentially compensated for by applying adequate numbers of BM-MSCs as suggested for therapies of non-united fractures 51 .
We detected significantly lower osteogenic marker levels in uncultured NU BM-CD271 high CD45 low cells despite the osteogenic ability shown for culture-expanded cells in this study and previous work 41 . In NU-serum supplemented cultures, all osteogenic genes, except ALPL showed comparable levels versus U-cultures. These observations could be explained by the positive osteoinductive role of human serum effect on progenitor osteogenesis as reported before 60 . Together, unlike culture-expanded MSCs, our data uniquely suggested a lower osteogenic potential of native NU BM-MSCs compared to U BM-MSCs. This apparent osteogenic defect could be probably improved by co-application of autologous serum growth factors, e.g., concentrated platelet lysate products. Consistent with our findings of low osteogenic potential of BM-MSCs in NU patients, the serum levels of several osteogenic markers have been shown to decrease in NU fractures. Kurdy et al. and Emami et al. detected lower levels of ALP in patients with delayed healed tibial fractures than normal healers 61,62 . Additionally, other groups reported reduced levels or delayed induction of osteocalcin levels in NU than U fractures [63][64][65] . Furthermore, the concentrations of C-terminal cross-linking telopeptide of type I collagen measured in the serum of NU patients were significantly less than those in U group 66 .
Although early inflammatory response is critical for the healing of fractures and bone loss, unbalanced inflammation marked by high levels of cytokines, TNF-α, IFN-γ and IL-1 TNF-α in fracture hematoma or callus could hinder a successful bone regeneration and lead to complicated healing (or non-union) 11,20,67-69 . Schmidt-Bleek et al. also have reported that higher relative expression of the pro-inflammatory cytokines together with a higher percentage of cytotoxic T cells in the hematoma of delayed healing than those of normal healed fracture 70 . Several Figure 6. The measurements of serum cytokine levels. The figures show the median of serum levels of IFNγ, TNF-α, IL-1 and IL-17 (a-d respectively) during a week of fracture (early healing), at or after 6 months of fracture (late healing) and healthy control levels (two-time points, 6-month apart). The Kruskal-Wallis test was used to compare levels of IFN-γ, TNF-α, IL-1 and IL-17 between U-serum (early healing: n = 9, 12, 9 and 13 respectively, late healing n = 9, 12, 9 and 11 respectively), NU-serum (early healing n = 5, late healing n = 9, 12, 9 and 6 respectively) and control serum (both time points: n = 8, 11, 7 and 5 respectively). (2019) 9:17340 | https://doi.org/10.1038/s41598-019-53927-3 www.nature.com/scientificreports www.nature.com/scientificreports/ studies reported that MSCs could drive immunosuppressive response as a part of their role in bone healing e.g., decreasing cytokines levels in association with enhancing the osteogenesis in fractures or osteoporosis models [24][25][26] . Uniquely, we showed here that NU BM-MSCs had fewer transcripts of IDO, PTGES2 and TGF-β1 when cytokine-treated than those of U BM-MSCs. Furthermore, low expression of IL-1 receptors in NU BM-MSCs might also suggest less response to such a priming cytokine. Altogether, our findings showed that NU BM-MSCs could have less immunosuppressive potential that could negatively affect normal bone healing.
In addition to immunosuppressive markers, other markers related to BM-MSCs immunoregulatory potential have been recently described. An example is the detection of high levels of BST2 expression in specific clones of BM-MSCs that secrete an immunoregulatory cytokine, IL-6 29 . We also have reported that S100A8 was highly expressed on hematopoietic cell-depleted cancellous bone populations that are capable of inhibition of T cell proliferation 31 . Our novel data here showed that NU BM-MSCs expressed lower BST2 transcripts, but higher S100A8 transcripts than U BM-MSCs. Furthermore, BTS2, but not S100A8 transcript levels were positively correlated with the osteogenic markers and growth receptor levels, confirming previous results of BST2 involvement in the MSC osteogenic differentiation 30 . The high S100A8 transcript levels in NU BM-MSCs could be linked to dysregulated bone remodelling as S100A8 has been shown as a contributor in the remodelling process 71 .
We detected significantly lower levels of IL-1 receptors consistently at protein and gene expression levels in NU BM-MSCs, also validating our PCR data. Growing evidence has indicated that IL-1 could have an individual or additive role to IFN-γ in enhancing MSC-mediated immunosuppression 72,73 . Additionally, IL-1 was shown to enhance osteogenesis, particularly for periodontal ligament stem cells 74 . Therefore, the functional impact of low expression of IL-1R on bone healing would be interesting for further investigations.
We measured the serum levels of cytokines that are known to prime BM-MSCs for immunosuppression. Our data showed no differences in serum levels of IFN-γ, TNF-α, IL-1, and IL-17 between patients or healthy non-fracture controls. Unlike serum, high levels of TNF-α were noted in fracture hematoma initially, then these levels were decreased with repair, but were highly expressed again during remodelling [75][76][77] . Also, IFN-γ expression levels were reported to be increased at hematoma 78 . Exceptionally, the IL-17 levels in NU-sera were significantly lower than U-and control serum in late healing. IL-17 was detected in soft callus during bone healing 79 and shown to have a complicated role during remodelling being supportive of osteoblastic bone formation but also involved in the stimulation of osteoclasts formation [80][81][82] . Our data showing lower IL-17 levels in NU sera than controls might suggest a relation to the disturbance in the bone remodelling process. The expression levels of several serum cytokines, growth factors, and other inflammation-related molecules in NU fracture patients were thoroughly assessed in many previous studies. However, and up to our knowledge, we are the first to compare serum cytokines, IFN-γ, TNF-α, IL-1, and IL-17 between NU, U patients, and healthy controls and found comparable levels. Likewise, another study showed that serum levels of IL-8, which is linked to human MSC migration 83 were not different between NU fractures and controls 84 . Furthermore, serum levels of other cytokines linked to BM-MSC migration, such as Macrophage colony-stimulating factor (M-CSF) and SDF-1 85 , were found to be similar between patients with physiological fractures and those with impaired fracture healing 84,86 . In contrast, Mathieu et al. found higher serum levels of IL-6 and lower levels of soluble IL-6 receptors in NU patients compared to healthy individuals 84 . Interestingly, an in vitro work showed that high concentrations of IL-6 with low expression of soluble IL-6 receptors could inhibit MSC differentiation 87 .
In terms of growth factors, TGF-beta, PDGF, FGF, and IGF are critically important for differentiation and growth of MSCs 58,88 . In fractures, TGF-β1 serum levels were found peaked between 1 and 6 weeks after trauma in normal healers, but it lasted for a shorter time and lower levels in NU patients 89,90 . Additionally, reduced PDGF serum levels in NU fractures were noted compared to U fractures 84,90,91 . Similarly, lower serum levels of FGF-2 in paediatric NU patients than normal healers were reported 92 . Previous data also showed low and stable levels of serum insulin-like growth factor-1 (IGF-1) and its specific binding proteins in NU fractures 84,90,93 . While serum VEGF concentrations in NU patients were found comparable to those in patients with normal fracture healing in some studies 84,94 , others found that serum VEGF is increased when NU fracture is treated 95 . Compared to systemic changes in these growth factors, other growth factors seemed to have local changes. Serum BMP-2 and BMP-4 were below the detection level in both U and NU patients 90 . Another work demonstrated no significant difference in plasma levels of BMP-2, -4, -6, and -7 between patients with NU and those with normal fracture healing 96 . However, the expression levels of these BMPs were significantly lower when assessed within fibrous NU tissue relative to a standard healing callus 33,97 .
Additional serum factors that are related to skeletal tissue inflammation or BM-MSC immunomodulatory functions were also quantified by other groups. In one study, NU patients had significantly higher serum concentrations of metalloproteinase protein-1 (MMP-1) and MMP-8, which are linked to skeletal tissue inflammation 98 compared with those with normal fracture healing 99 . S100A9 is also secreted by blood cells, and its high serum levels are linked to inflammation, particularly of musculoskeletal tissues 100 . Interestingly, serum S100A9 has been found elevated than normal in NU patients 101 , reflecting a local inflammation of NU tissues in which MSCs could be involved. Another work by Wang et al. demonstrated that Nitric oxide (NO), which is linked to MSC differentiation and bone remodelling 102 , had significantly lower levels in persistent NU patients relative to those having good healing progress 95 . Furthermore, proteomic analysis of serum showed that complement C6, C3 and C4, which their strongly activating phenotype is less favourable for MSC-related immunosuppression 85 , were up-regulated in the serum of NU patients compared to healthy controls 101 . Altogether, the systemic changes of various serum mediators could explain the lower proliferative and osteogenic potential as well as the altered immunomodulatory potential of the MSC pool in BM that we noted in our study and by others for NU fracture patients 103 .
Having NU BM-MSCs collected retrospectively was a limitation in our study. We could not prove that the altered immunoregulatory phenotype of these NU BM-MSCs (i.e. being less immunosuppressive, low BST-2 and IL-1R and high S100A8 expression) was consistently the same phenotype at the early phase of healing. Further (2019) 9:17340 | https://doi.org/10.1038/s41598-019-53927-3 www.nature.com/scientificreports www.nature.com/scientificreports/ research using a large panel of NU BM-MSCs prospectively collected within the early phase of fracture healing would be of great value to confirm our findings of the altered immunoregulatory phenotype. Such research could potentially help to introduce new predictive molecules for non-union fracture at the early phase and perhaps allowing earlier intervention with therapies to enhance bone repair.
In conclusion, our data of uncultured BM-MSCs indicated some interesting differences between union and non-union MSCs. These native NU BM-MSCs seemed to have a low proliferative capacity and osteogenic potential relative to U BM-MSCs. Additionally, our study is the first report of an altered immunoregulatory phenotype of NU BM-MSCs with these cells appeared to have reduced immunosuppressive potential and lower expression of BST2. Clinically, the quality of native uncultured autologous BM-MSCs used for enhancing bone regeneration should be considered. One may argue that the therapeutic value of autologous native BM-cells could be improved with growth adjuvants as well as placing adequate numbers of these BM-MSCs in situ to compensate for potentially low performance.
The functions of culture-expanded NU and U BM-MSCs were comparable most probably due to changes in cell phenotype. Therefore, to assess the MSC functions to a close picture as possible to in vivo cell performance, functional experiments would need to be optimised on minimally-cultured MSCs and in the presence of human serum. In such an experimental approach, future analysis of the functional impact of low IL-1 receptor levels in NU BM-MSCs could potentially help to improve cell-based bone regenerative therapies. Additionally, further work would be required to confirm a specific immunoregulatory phenotype of NU BM-MSCs at the early healing phase, which could potentially help to predict the risk of non-union fractures.
Methods ethics statement. Informed written consent was obtained from all the study participants before the samples were collected, and research was carried out in accordance with the Helsinki Declaration of ethics. The consents and sample collection (blood and bone marrow samples used to extract MSCs) for this study were under ethical approval with NREC number, 06/Q1206/127, National Research Ethics Committee Yorkshire and Humber-Leeds East. All study experiments were conducted according to the appropriate guidelines and regulations. participants and samples. Overall seventy-one participants were included in the study ( Table 2). All patients had long bone fractures (femur, tibia, humerus). NU was defined by the absence of radiological features of fracture healing (lack of callus formation in at least 3 cortices) either on plane radiographs or computed tomography scans after 9 months from fracture fixation and with ongoing pain at the NU site during ambulation. Exclusion criteria were children, cancer, diabetes, bone metabolic diseases, inflammatory/immune disorders and intake of drugs with a negative impact on fracture healing (i.e. NSAIDs). BM samples (15 ml) were aspirated from the anterior iliac crest as previously described 104 ; within one week following injury for U patients, and after diagnosis (>9 months) for NU patients. Peripheral venous blood samples (12 ml) were collected (from the patients) within one week from the date of fracture (early healing phase) or after fracture had united (4-6 months from injury) for U fractures and at the time of NU diagnosis. Serum from healthy individuals who did not have any bone fractures was used as control (twice, 6-month apart). Comparing the age of three groups showed no significant difference (Kruskal-Wallis test, p = 0.497).
The quantitative real-time PCR using TaqMan probes (ThermoFisher, Table 1) was conducted for measuring the relative gene expression levels of chosen markers. RNA extraction was performed as per manufacturer recommendation (Norgen Biotek). Extracted RNA samples were processed for reverse transcription and then pre-amplification using kits (Fluidigm) as recommended. Finally, the gene expression assays were performed using 48.48 chip/Integrated Fluid Circuit (IFC) and using the Biomark ™ HD system (Fluidigm). The marker transcript levels were calculated relative to that of a housekeeping gene, Hypoxanthine Phosphoribosyl transferase 1 (HPRT1).

BM-MSc cultures.
BM-cells were cultured till plastic-adherent MSCs become confluent then lysed and analysed without passage (passage-zero cells). These cultures were maintained in DMEM medium (Sigma) supplemented with 10% pooled U-or NU-serum (from five donors and collected within one week of fracture). Some cultures were treated with cytokines (20 ng/ml IFN-γ, 20 ng/ml TNF-α, 10 ng/ml IL-1 and 100 ng/ml IL-17) as per previous studies [106][107][108] . For the culture-expansion, BM-cells were maintained in the StemMACS MSC Expansion medium (Miltenyi Biotec) and plastic-adherent MSCs were passaged then analysed at passage 3-4. proliferation. The proliferation of culture-expanded BM-MSCs was assessed using the XTT colorimetric assay kit (Merck). For testing the growth of culture-expanded BM-MSCs on Orthoss collagen ® scaffold (Geistlich), 10 × 10 4 cells were loaded on 60 mm 2 pieces of the scaffolds for 3 hours before maintaining in cultures. The loaded scaffolds were digested after 1 and 3 weeks of culture using collagenase (Stem cell technologies). The released cells were then characterised and counted by flow-cytometry as shown previously 105,109 . For passage-zero BM-MSCs in serum-supplemented cultures, the housekeeping 18 S expression was evaluated, indicating cell quantities as reported before 110 . (2019) 9:17340 | https://doi.org/10.1038/s41598-019-53927-3 www.nature.com/scientificreports www.nature.com/scientificreports/ osteogenesis and immunosuppression assays. The culture-expanded BM-MSCs were assessed for the osteogenesis by measuring the calcium levels after 3 weeks in osteogenic cultures 105 . The calcium levels were measured using a colorimetric kit (Sentinel) 105,99,78,78 .
The immunosuppressive capacity of culture-expanded BM-MSCs was tested following treatment for 5 days with cytokines. The cells were then processed for measuring the intracellular IDO and surface TGF-β1 latency-associated peptide (LAP) levels using specific conjugated monoclonal antibodies (ThermoFisher) for flow-cytometer as demonstrated in previous studies [111][112][113] . For passage-zero cultured BM-MSCs that were cytokine-treated, the expression levels of IDO, PGE2 and TGF-ß transcripts was evaluated indicating the MSC immunosuppressive potential.
Surface cytokine receptor measurement. Flow-cytometry was used to measure the cytokine receptor levels on the surface of uncultured NU and U CD271 high CD45 low cells (and to validate PCR data) or on culture-expanded NU and U MSCs. The flow-cytometry conjugated monoclonal antibodies against IFN-γ receptor 1 (CD119), TNF-α receptor 1 (CD120a), (both from Miltenyi Biotec), IL-17 receptor A (CD217), (R&D systems) and IL-1β receptor 1 (BD Biosciences) were used for these assays.
Serum cytokine measurements. The blood samples were processed by centrifugation (2000g for 15 minutes) to extract the serum, as previously described 114 . The separated sera were analysed for the IFN-γ, TNF-α, IL-1 and IL-17 levels using the hypersensitive ELISA kits (R&D systems) as recommended by the manufacturer. The optical density values were acquired using MULTISCAN EX reader and analysed with Ascent software (Thermo electron corporation).
Statistical analysis. The statistical analysis and figures' preparation were performed using GraphPad Prism 7 software. The Shapiro-Wilk normality test was applied to choose the comparative tests between groups. These comparative tests were also applied according to the group numbers and being paired or non-paired; all indicated in figure legends. Correlations were tested using the Spearman rho test. Any difference between the groups was considered as statistically significant when the p value < 0.05.

Data availability
The datasets supporting the conclusions of this study are included within the article and its additional supporting file.