A ROADMAP OF OSTEOGENIC DIFFERENTIATION IN HUMAN MESENCHYMAL STEM CELLS ASSESSED BY PROTEIN MULTIPLEX ANALYSIS

Human mesenchymal stem cells (hMSCs) are widely used to model osteogenic differentiation in vitro , yet few studies compare the numerous available protocols. The poor translation between in vitro and in vivo results in bone regeneration highlights the need for improved methodologies to assess osteogenic phenotype. Omics technologies generate vast biological data but are complex for comparing several osteogenic differentiation strategies. Conversely, conventional methods like Alizarin red S stains or alkaline phosphatase assays do not provide enough data. Here, we propose a targeted protein multiplex assay to characterize and compare several models of osteogenic differentiation in hMSCs, containing markers for osteogenesis, angiogenesis


Introduction
Bone marrow-derived human mesenchymal stem cells (hMSCs) are multipotent progenitors, able to undergo differentiation towards the osteogenic, chondrogenic and adipogenic lineages (Bhat et al., 2021; Xu et al., 2017).They are of interest for tissue engineering and regenerative medicine, often in combination with a biomaterial, due to their role in regeneration and healing (Herberg et al., 2021; Quade et al., 2020; Sha et al., 2019).For example, hM-SCs have been shown to secrete biologically active factors with immunoregulatory, tissue-supportive, and instructive functions in vitro (Kehl et al., 2019; Quade et al., 2020; Shin et al., 2021), and have successfully generated bone tissue, promoted host vascularization, and provided an immunomodulatory effect upon transplantation in several in vivo models (Clark et al., 2020; Quade et al., 2020; Taylor et al., 2019).Despite these successes, the clinical application of hMSCs, for example in bone regeneration, remains limited.A reason for this is the difficulty to translate findings from in vitro work to in vivo, suggesting the need to (a) Different strategies were used to induce osteogenic differentiation in human mesenchymal stem cells (hMSCs) over 28 days.After mineralization assessment by Alizarin red S staining at days 14 and 28, the strategies were classified according to their resulting mineralization status into inducers of non-, late-and earlymineralization, and protein patterns from the multiplex assay were correlated to the mineralization status.Assays that were executed on a specific time point are marked by X.(b) Cell morphology and mineral deposition at day 3. Initial mineral deposits were observed in hMSCs in phosphate ions (Pi) at day 3. (c) Mineral deposits at day 21 were observed in hMSCs cultured in OM, OM+BMP-2, calcium (Ca), and Pi at day 21.Scale bar: 100 µm.
European Cells and Materials Vol.47 2024 (pages 170-189) DOI: 10.22203/eCM.v047a12reconsider how differentiation of hMSCs is investigated in vitro (Hulsart-Billström et al., 2016).In particular, hMSC differentiation into the osteogenic lineage is important for bone regeneration and its application to heal critical-sized bone defects, which are currently treated by bone grafting, a common surgical procedure that is performed over 2.2 million times per year worldwide (Calori et al., 2011).
A lot of effort has been put into finding a method to stably and reproducibly differentiate hMSCs, resulting in several in vitro strategies to direct their osteogenic differentiation.While three-dimensional (3D) models have demonstrated the ability to more accurately mimic the physiological environment of bone, overcoming the limitations associated with two-dimensional (2D) models such as forced polarity and reduced cell-cell contacts, the use of 2D models in the field remains high (Braccini et al., 2005).This study will focus on commonly used 2D models within the field.Common medium components include dexamethasone to enhance alkaline phosphatase (ALP) expression, βglycerophosphate to serve as a phosphate source for mineralization, and ascorbic acid as a co-factor for collagen synthesis (Sordi et al., 2021).Dexamethasone is typically supplemented throughout the culture at a concentration range that varies (e.g., from 10 nM to 1 µM) in different studies.Growth factors such as bone morphogenetic protein-2 (BMP2), bone morphogenetic protein-7 (BMP7), bone morphogenetic protein-9 (BMP-9) or transforming growth factor beta-1 (TGF-beta-1) are also used to upregulate pathways related to osteogenesis (Han et al., 2020; Xu et al., 2020; Yang et al., 2018).Osteogenic differentiation can also be achieved through the enrichment of medium with calcium (Ca) or phosphate ions (Pi), or by culturing cells on calcium phosphate (CaP) scaffolds (Danoux et al., 2015; Lei et al., 2015).To date, there are few experimental studies comparing the considerable number of differentiation protocols available, and there is a lack of standardization in the methodologies used to assess the osteogenic phenotype.
When choosing the methodology to assess a phenotype, there are several important considerations.Conventional assays such as Alizarin red S (ARS) staining, quantitative RT-PCR and ALP assays are affordable and easy to perform, but provide limited biological information, which may contribute to the poor (pre-) clinical translation of in vitro findings (Hulsart-Billström et al., 2016).Alternatively, numerous studies have investigated the protein profiles of hMSCs using omics technologies (Kehl et al., 2019; Lei et al., 2015; Shin et al., 2021), which generate a large amount of biological data that can provide important insights.However, these methods are complex, expensive and require specialized infrastructure.A protein multiplex analysis could be considered a promising intermediate tool to characterize hMSCs in vitro as it is a simple assay that can simultaneously quantify a panel of selected proteins, making it a more targeted and cost-effective approach.The use of multiplex protein assays remains limited in tissue engi-neering, though a study by Fischer et al. (2022) evaluated the effect of cold physical plasma on osteogenic differentiation of hMSCs and used a 12-plex assay comprising cytokines, chemokines and growth factors at early time points (Fischer et al., 2022).
Here, we used a targeted proteomic approach with an 18-plex panel of markers for osteogenesis, angiogenesis, and inflammation in hMSCs.Osteogenic differentiation was induced using different strategies, which were then classified as non-, early-and late-mineralizing groups according to the resultant mineralization status (Fig. 1).Protein expression patterns were then associated with the different classifications.We identified osteogenic markers, such as matrix metalloproteinase-13 (MMP-13) and growth arrest-specific 6 (GAS-6), that could be correlated to the mineralization status, while others such as TGF-beta-1 were inversely correlated.We also showed higher angiogenic potential and increased inflammatory markers in hMSCs that did not mineralize.Based on our results, we consider a targeted protein multiplex assay as a practical tool to characterize differentiating hMSCs and to provide a better understanding of osteogenic differentiation in vitro, both informing the more effective use of hMSCs in regenerative medicine.

Calcium Phosphate Coating Preparation
For the first coating step, a five-times concentrated simulated body fluid solution was prepared by dissolving carbon dioxide (CO 2 ) gas into a solution comprising 733.5 mM Na 2+ , 7.5 mM Mg 2+ , 12.5 mM Ca 2+ , 720 mM Cl − , 5 mM HPO 4 2− , and 21 mM HCO 3 − in demineralized water at 37 °C, made from reagent grade NaCl, CaCl 2 •2H 2 O, MgCl 2 •6H 2 O, Na 2 HPO 4 •2H 2 O and NaHCO 3 salts, added one after the other, as previously described (Yang et al., 2010).The solution was sterile-filtered and 6-well tissue culture plates (Corning, Amsterdam, The Netherlands) were filled with 2 mL of the solution and incubated at 37 °C in 5 % CO 2 for 24 h, resulting in the formation of an amorphous CaP layer.The plates were washed twice with cell culture grade water.
In the second coating step, CO 2 was dissolved into a solution comprising 733.5 mM Na 2+ , 1.5 mM Mg 2+ , 12.5 mM Ca 2+ , 720 mM Cl − , 5 mM HPO  et al., 2010).The solution was sterilefiltered and the 6-well tissue culture plates (Corning) were filled with 2 mL of the solution and incubated at 37 °C in 5 % CO 2 for 3 days with daily refreshments, resulting in the formation of a crystalline CaP layer.Following the coating process, the plates were washed twice with cell culture grade water.Before cell seeding, the CaP coatings were sterilized in 100 % isopropanol for 2 min and 70 % isopropanol for 2 min, washed twice in cell culture grade water, and soaked in basic medium (BM) for 30 min.

Fourier-Transform Infrared Spectroscopy
CaP-coated well plates were washed twice with cell culture grade water and dried.Attenuated total reflectance Fourier-transform infrared spectroscopy (ATR-FTIR) (Nicolet is50, Thermo Fisher Scientific, Landsmeer, The Netherlands) was performed in the range of 500-4500 cm −1 in the transmission mode in order to identify structural groups.

X-Ray Diffraction (XRD) Analysis
Phase analysis and the degree of crystallinity of the CaP coatings were determined on an X-ray diffractometer (D2; Bruker) at room temperature in the range of 6 ≤ 2θ ≤ 60 with a scan rate of 1°/min and a step size of 0.03°.Pattern analysis was performed using Profex 4.2.(Solothurn, Switzerland) (Doebelin and Kleeberg, 2015).

Scanning Electron Microscopy (SEM) and Energy-Dispersive X-Ray Spectroscopy (EDS) Analyses
The coated wells were washed, cut, and glued onto aluminum stubs using carbon tape, after which they were gold sputter coated (SC7620, Quorum Technologies, Laughton, England).To visualize the CaP crystal morphology and their distribution on the surface of the coating, scanning electron microscopy (JSM-IT200, JEOL, Zaventem, Belgium) was performed at an acceleration voltage of 20 kV and magnifications of up to 10,000×.To determine the elemental composition of the coating and its homogene-ity, energy-dispersive X-ray spectroscopy analysis (JSM-IT200, JEOL) was performed.
hMSC Culture HMSCs originating from bone marrow were obtained from PromoCell on 27-11-2019 (Lot: 451Z021.2) (Promo-Cell, Huissen, The Netherlands, 65-year-old donor), with confirmed marker expression of CD73, CD90, and CD105 and absence of CD14, CD19, CD34, CD45, and HLA-DR markers, as recommended by the International Society for Cellular Therapy and free from any contaminants such as bacteria, fungi and mycoplasma.Cells at passage 2 were seeded at 2000 cells/cm 2 in BM composed of minimal essential medium (α-MEM GlutaMAX no nucleosides; Gibco, Landsmeer, The Netherlands) supplemented with 10 % fetal bovine serum, and were maintained in a humidified environment at 37 °C in 5 % CO 2 .Medium was changed every two days.Upon reaching 80 % confluence, cells were detached using 0.05 % trypsin-EDTA and reseeded at a density of 1000 cells/cm 2 for further expansion.Cells were used for experiments after passage 4, when they were seeded into 6-well plates (Corning) with or without CaP coatings at a density of 10,000 cells/cm 2 in 2 mL of BM.Prior to inducing osteogenic differentiation, cells were expanded to confluence.

Trilineage Differentiation Potential of hMSCs
Adipogenic differentiation was induced in hMSCs which were seeded at a density of 10,000 cells/cm 2 .After a 2-day expansion, adipogenic medium consisting of Dulbecco's modified Eagle medium (high glucose, no sodium pyruvate; Gibco, Landsmeer, The Netherlands) supplemented with 10 % FBS, 40 mM indomethacin, 83 mM 3www.ecmjournal.orgEuropean Cells and Materials Vol.47 2024 (pages 170-189) DOI: 10.22203/eCM.v047a12isobutyl-1-methylxanthine, 10 mg/mL insulin and 0.1 mM dexamethasone was added.Medium was changed every 2-3 days.After 21 days, cells were fixed in 4 % wt/vol paraformaldehyde for 15 min at room temperature.Cells were stained with a 0.2 % (wt/vol) Oil Red O solution in 60 % isopropanol for 15 min to detect lipid droplets.Chondrogenic differentiation was induced in hMSCs which were seeded at a density of 200,000 cells per well in a 96-well, U-bottom suspension culture plate.Aggregates formed within 24 h.Chondrogenic differentiation medium consisting of Dulbecco's modified Eagle medium (high glucose, no sodium pyruvate; Gibco) supplemented with 10 % FBS, 50 mg/mL ITS premix (Thermo Fisher Scientific, Landsmeer, The Netherlands), 40 µg/mL proline, 0.2 mM ascorbic acid phosphate, 100 µg/mL sodium pyruvate, 0.1 µM dexamethasone and 10 ng/mL TGF-b3 (PeproTech, Hamburg, Germany) was added.Medium was changed every two days.After 21 days in culture, aggregates were fixed in 4 % wt/vol paraformaldehyde for 25 min at room temperature.Aggregates were embedded in paraffin, sectioned, rehydrated and stained with 1 % Alcian blue solution in acetic acid for 30 min in order to detect proteoglycans.Osteogenic differentiation was induced in hMSCs in osteogenic medium (OM), with the formulation described in the next section.The capacity of hMSCs to mineralize, as an indicator for osteogenic differentiation, was assessed via ARS staining at day 28 of culture.
For lysis, the cells were washed twice with phosphate buffered saline (PBS) and frozen at -80 °C.After three freeze-thaw cycles at -80 °C, 500 µL of lysis buffer (Pro-cartaPlex, Invitrogen, Waltham, MA USA) supplemented with 1 mM phenylmethylsulfonyl fluoride was added to each well and incubated for 15 min on ice before collection.Samples were sonicated three times for 10 s.The immunoassay panel was performed according to the manufacturer's instructions.In brief, samples were diluted 2× in the universal assay buffer provided in the kit, and 50 µL of medium supernatant or cell lysate from each sample, along with standards and quality controls, were added to wells containing the antibody-coupled beads, shaken for 30 min at room temperature, incubated overnight at 4 °C, and shaken again for 30 min.After two washing steps, detection antibody-biotin reporters were added to each well and incubated for 30 min at room temperature with shaking, followed by incubation with the fluorescent conjugate streptavidin-phycoerythrin for 30 min while shaking.Protein concentrations were measured using a Luminex100 (Bio-Rad, Veenendaal, The Netherlands) with data acquisition in Bio-Plex Manager 6.0 software (Bio-Rad, Veenendaal, The Netherlands).Calibration and verification were performed before analysis.Standard curves were established for each analyte with known concentrations of proteins, and the concentrations of analyzed markers were expressed as picograms per milliliter (pg/mL).Blanks containing cell culture medium or lysis buffer and a known sample as quality control were analyzed in each assay in duplicate.For normalization of protein concentration, a single-plex assay for HSP60, a housekeeping protein, was done following the same steps.Each protein concentration normalized to HSP60 is represented as a ratio relative to expression at day 0.

Alizarin Red S Staining and Quantification
Alizarin red analysis was performed at days 14 and 21, for which hMSCs were washed twice with PBS and fixed European Cells and Materials Vol.47 2024 (pages 170-189) DOI: 10.22203/eCM.v047a12 in 4 % (wt/vol) paraformaldehyde for 15 min at room temperature.After three washes in distilled water, cells were stained with 2 % (wt/vol) Alizarin red S solution in distilled water (pH 4.2) for 15 min with gentle shaking.Calcium deposits, stained in red, were imaged by phase contrast microscopy.For colorimetric quantification, Alizarin red S staining was removed by incubation in 10 % hexadecylpyridinium chloride in 10 mM sodium phosphate buffer (pH 7.0) overnight at room temperature.The dissolved solution was transferred to a 96-well plate, diluted if necessary, and the absorbance was measured at 562 nm on a plate reader (CLARIOstar, BMG Labtech, Ortenberg, Germany).A standard curve with known amounts of Alizarin red S was used to estimate sample concentrations.Control wells without cells that had received the same treatments, and a CaP coating without cells, were used as a blank measurement for the different conditions and subtracted from experimental measurements.

Alkaline Phosphatase Activity Assay
ALP levels were measured to indicate osteoblast phenotype at days 3, 7, 10, 14 and 21 of culture.Cell lysis was done as previously described, and ALP activity was quantified using the CDP-Star solution (Sigma-Aldrich, Amsterdam, The Netherlands) (Alves et al., 2011; Zhang et al., 2014).The cell lysate was incubated 1:5 with the reagent for 20 min in the dark at room temperature in a white-bottom 96-well plate.Luminescence was read on a spectrophotometer (CLARIOstar, BMG Labtech, Ortenberg, Germany) with a 3600 gain and a measurement interval of 1 s.ALP values were normalized to the housekeeping protein HSP60 expression.

RT-qPCR
Samples were lysed in TRIzol at days 3, 7, 10, 14 and 21.RNA was extracted using the phenol chloroform method and purified using a Bioline kit.RNA purity and quantity were determined on a BioDrop.For each sample, 500 ng of RNA were reverse transcribed into cDNA using the iScript cDNA synthesis kit (Bio-Rad, Veenendaal, The Netherlands).The amplification of 20 ng of cDNA by qRT-PCR was performed on a CFX96 Real-Time PCR Detection System using the iQ SYBR Green Supermix (Bio-Rad, Veenendaal, The Netherlands).Transcript levels of osteogenic biomarkers including runt-related transcription factor 2 (RUNX2), osterix (OSX ), BMP2, ALP, SPP1, collagen type I alpha 1 chain (COL1A1), and osteocalcin (OCN), were determined.Fold expression values were determined using the ∆∆Ct method after normalizing each target gene with respect to the geometric mean of the expression of two stably-expressed housekeeping genes (RPL13A and YW-HAZ) and to the expression of the target gene in hMSCs at day 0 on culture plates in BM.Primer sequences can be found in Table 1.

Statistical Analysis
Two-way analysis of variance (ANOVA) was done on all data sets, except the protein data.Data are represented as mean ± standard deviation.Post-hoc analysis was performed using Dunnett's test.p-values < 0.05 were considered statistically significant.
Concerning the statistical analysis of the protein data, missing data points below the detection limit of the Luminex instrument were imputed using the lowest concentration of the standard curve of the corresponding protein or the lowest value in the data set if the latter was below the standard curve value.Missing data points above the detection limit were imputed using the highest value of the standard curve of the corresponding protein or the highest value in the data set if the latter was above the standard curve value.The proteins SPP-1 in the lysate and CALCA in the lysate had too many missing values, and were therefore removed before analysis.After data preprocessing, differential expression analysis on protein data was done using linear regression modelling on a log scale (DEP package 1.21.0 in R 4.2.0).p-values were computed using the Benjamini-Hochberg false discovery rate correction for multiple tests (p.adjust function in R).Principal component analysis (PCA) was performed per time point and per group of proteins (osteogenesis-, angiogenesis-or inflammation-related) in order to determine sample clustering and whether different experimental groups showed a different osteogenic, angiogenic or inflammatory marker expression.

A Carbonated Apatite Coating was Deposited on Well Plates
SEM analysis of the CaP-coated tissue culture plates showed that a homogenous mineral layer was deposited (Fig. 2a).ATR-FTIR spectroscopy showed the presence of phosphate and carbonate bands (Fig. 2b).The XRD pattern was typical of a crystalline carbonated apatite (Fig. 2c).

hMSCs were able to Undergo Trilineage Differentiation
The trilineage differentiation potential of the hMSCs was evaluated.Positive staining for Oil red O, Alizarin red S, and Alcian blue confirmed their capacity to differentiate into adipocytes, osteoblasts, and chondrocytes, respectively (Fig. 2d).served in hMSCs cultured in OM, OM+BMP2, as well as in Ca and Pi (Fig. 1c).Taking phase contrast images of hMSCs through the thick CaP coating was not feasible due to technical limitations.Increased osteogenic differentiation, as defined by mineral deposition, was confirmed by a positive Alizarin red S stain in hMSCs in Ca and Pi and on CaP after 14 days in culture and in OM and OM+BMP2 after 28 days in culture (Fig. 3a).The Alizarin red S stain was then dissolved and quantified (Fig. 3b).After 14 days in culture, the amount of Alizarin red S in hMSCs in Ca (2.7 mM ARS) and on CaP (4.7 mM ARS) was higher (all p < 0.001) than that in the BM control (0.026 mM ARS), while the amount in BM+BMP2, OM, OM+BMP2 and Pi was not significantly different from the control.After 28 days in culture, the amount from hMSCs in OM (5.6 mM ARS), OM+BMP2 (3.3 mM ARS), Ca (1.9 mM ARS), and Pi (3.2 mM ARS) and on CaP (6.5 mM ARS) was higher (all p < 0.001) relative to hMSCs in BM (0.035 mM ARS) (Fig. 3b).These findings led us to classify the osteogenic differentiation conditions into three groups, namely "nonmineralization" (in BM+BMP2), "late-mineralization" (in OM and OM+BMP2) and "early-mineralization" (in Ca and Pi, and on CaP).

Increased Alkaline Phosphatase Activity was Detected in HMSCs in OM, OM+BMP2, Ca and Pi
ALP activity, an enzyme known to promote mineralization that is upregulated in the early stages of osteoblast commitment, was measured at days 3, 7, 10, 14 and 21 (Fig. 3c-g).As expected, increased ALP activity was detected to be above the level of the BM control in both the earlyand late-mineralization groups, albeit at time points that did not necessarily correspond to their mineralization.After 3 days in culture, there were no significant changes (Fig. 3c) compared to hMSCs in BM, but ALP activity was increased after 7 days in OM (Fig. 3d), and after 10 days in OM, OM+BMP2 and Pi (Fig. 3e).After 14 days, ALP activity was increased in OM, OM+BMP2, and Ca, and decreased in hMSCs in BM+BMP2 compared to the BM control (Fig. 3f).After 21 days, ALP activity was increased in hMSCs in OM and OM+BMP2, and decreased in BM+BMP2 and Pi, and on CaP (Fig. 3g).

A Multiplex Assay to Determine hMSC Phenotype when Differentiating towards the Osteogenic Lineage
An 18-plex protein multiplex assay was performed at days 3, 7, 10, 14 and 21 on the cell lysates (LY) and/or the medium supernatants (SN), depending on the protein, in order to establish protein profiles for the different conditions (Fig. 4a).At each time point, differential expression analysis determined significant changes relative to control hMSCs in BM.To identify potential protein expression patterns between groups, PCA analysis was performed using the abundances of the proteins, grouped as osteogenic, angiogenic and inflammatory markers, for each time point.
The PCA was plotted based on the first and second principal.The contribution of proteins to principal components can be found in Fig. 5a-c.An overview of the significant markers can be found in Fig. 6d.
As expected, hMSCs that had been classified as nonmineralizing (BM+BMP2) formed a well-defined cluster at most time points.Compared to hMSCs in BM, this cluster at day 3 was defined by an increase in SPP-1, BMP-9, FGF-23, RANKL and OPG, all in the supernatant.And at day 7, this cluster expressed an increase of TGF-beta-1, GAS-6, MMP-13, RANKL, SPP-1 and BMP-9, all in the supernatant.By day 10, the non-mineralization group was showing more similarities to hMSCs in OM+BMP2.This correlation was determined by an increase of BMP-9, SPP-1 and FGF-23, and a decrease in BMP-2, all in the supernatant.As time continued to days 14 and 21, the nonmineralization group was the least correlated to the other conditions.This was determined by an increase in TGFbeta-1, FGF-23, BMP-9, and OPG at day 14, all in the supernatant and a decrease of TN-C in the lysate, and by an increase of OPG and TGF-beta-1 in the supernatant, and a decrease of TN-C, MMP-13 and OPG in the lysate at day 21.
The group of late-mineralizing hMSCs clustered together with the control in BM at day 3.However, both conditions belonging to the late-mineralizing condition did not cluster closely together until day 14 and we noted significant differences between hMSCs in OM compared to OM+BMP2, despite their similar timelines of mineralization and ALP activity.For example, at day 7, hMSCs in OM+BMP2 were defined by the increase of MMP-13 in the lysate, and FGF-23 and BMP-9 in the supernatant, while OPG in the lysate decreased.The two conditions diverged further at day 10, where FGF-23 and BMP-9 in the supernatant were decreased in OM and increased in OM+BMP2.The OM+BMP2 condition was further defined by the increase in MMP-13 in the lysate, which was not seen in hMSCs in OM.By day 14, the two conditions in the latemineralization group converged.HMSCs in OM expressed diminished osteogenic markers, namely TGF-beta-1, FGF-23 and BMP-9, all in the supernatant, and TN-C in lysate.Conversely, hMSCs in OM+BMP2 were defined by an increase of MMP-13 (in both the supernatant and lysate), and a decrease in BMP-2 in the supernatant and OPG in the lysate.TN-C in lysate was also decreased.Despite these differences, the overall profile of the two conditions was similar, and this trend continued until day 21, at which point  European Cells and Materials Vol.47 2024 (pages 170-189) DOI: 10.22203/eCM.v047a12 both groups were defined by the decrease of TN-C and OPG in the lysate, and GAS-6, RANKL, BMP-9 and FGF-23 in the supernatant.Only one difference was found, which was that MMP-13 in the lysate was decreased in OM and increased in OM+BMP2.A decrease in CD73 was only shown in late-mineralizing hMSCs from day 10 and later time points which may indicate a loss of stemness.
The conditions classified as early-mineralizing demonstrated diversity in their accompanying protein profiles even though they clustered closely beginning on day 7.At day 3, they were similar in their increase of CD73, MMP-13 and GAS-6 in the lysate and their decrease of GAS-6 and TGF-beta-1 in the supernatant.An increase in TN-C in the lysate was found in both hMSCs in Ca and Pi and in the supernatant in Pi and on CaP.At day 7, the early-mineralizing hMSCs were distinguished from non-mineralization hMSCs by their increase in CD73, GAS-6 and MMP-13 in the lysate and their decrease of TGF-beta-1 and GAS-6 in the supernatant.TN-C was increased in the supernatant of hMSCs in Pi and on CaP until day 10.At day 10, the early-mineralizing group was most dissimilar to the other groups, and was marked by an increase in TN-C and GAS-6 in the lysate and a decrease of GAS-6, RANKL, TGF-beta-1, and SPP-1 in the supernatant.It was surprising to see the similarities between the early-mineralizing and the non-mineralizing groups at day 14, where significant differences were only detected in CD73, MMP-13 and GAS-6 in the lysate, all of which were increased in the early-mineralizing group.Ultimately, by day 21, the early-mineralizing group formed a well-defined cluster based on the increase of GAS-6 and TN-C in the lysate and of FGF-23 in the supernatant, and a decrease of OPG in the lysate and of MMP-13, OPG, and TGF-beta-1 in the supernatant, suggesting that we found markers that can distinguish this group, even after mineralization had occurred.

Increased Angiogenesis-Related Marker Expression in Early Mineralizing hMSCs at Multiple Time Points
In addition to osteogenic markers, we determined the protein expression of two angiogenic markers: VEGF-A and ANG-1 (Fig. 4a,c, Fig. 6b).Given we had classified the conditions based on their osteogenic phenotype, it was not surprising that the conditions clustered less closely based on angiogenic markers, but we were nonetheless able to determine informative trends.For example, at day 7, the early-mineralization and late-mineralization groups were clearly different.The late-mineralizing hM-SCs expressed higher ANG-1 in the supernatant and lower VEGF-A in both the supernatant and the lysate compared to both the control and the non-mineralization group.Conversely, VEGF-A in the lysate was increased in the earlymineralizing hMSCs and ANG-1 in the supernatant was decreased.This trend continued at day 10, and by day 14, the early-mineralization group had a marked increase in ANG-1 in the lysate compared to the late-mineralization group, suggesting a different differentiation trajectory.After 21 days, the non-mineralization, late-mineralization and early-mineralization hMSCs all had increased ANG-1 in the lysate and decreased ANG-1 in the supernatant.

Inflammation-Related Markers Increased in the Late-Mineralizing hMSCs and Decreased in Late-Mineralizing hMSCs
The protein multiplex assay also contained markers related to the immunomodulatory effect of hMSCs, which is an important process related to bone healing, namely IL-6, TNF-A and GM-CSF (Fig. 4a,d, Fig. 6c).Overall, we found that clustering based on inflammation-related markers was similar to the classification based on mineralization.
Looking first at the non-mineralization hMSCs, which overall clustered closely based on inflammation markers, we found a marked increase in TNF-A and GM-CSF in the supernatant at day 7.By day 14, IL-6 in the supernatant was significantly increased, a trend that continued until day 21.In contrast, the hMSCs classified as late-mineralizing had decreased IL-6 levels in the supernatant at day 3.The different conditions did not cluster closely at days 7 and 10, and by day 14, two different profiles had emerged, both of which had decreased TNF-A in the supernatant, but were distinguished by IL-6 and GM-CSF.Namely, IL-6 was lower in OM than in OM+BMP2, whereas GM-CSF in the supernatant was lower in OM+BMP2 than in OM.The most pronounced difference within the late-mineralization hM-SCs cluster was observed at day 21.
The early-mineralization hMSCs had a different profile of inflammatory markers and clustered more closely than the other groups.Until day 7, the inflammatory markers were decreased.Beginning at day 10, IL-6 in the lysate was increased, which continued until day 21.TNF-A in the supernatant was also increased at day 14.At day 21, early-mineralization hMSCs showed a more similar inflammatory profile to control cells in BM than other conditions and were marked by a decrease in IL-6 in the supernatant and an increase in IL-6 in the lysate.

Osteogenic Transcript Expression was Correlated to Mineralization Status
Having evaluated the protein multiplex assay data, we questioned how qRT-PCR of transcripts related to osteogenic differentiation, namely RUNX2, OSX, BMP2, ALPL, SPP1, COL1A1 and SPP1, would correlate to our findings (Fig. 7).Overall, while there was some correlation to the mineralization classification, the results at the transcript level highlighted the importance of evaluating protein expression.
Non-mineralizing hMSCs did not show an upregulation of any osteogenic genes indicating that no osteogenic differentiation is taking place in these cells.The gene expression profiles of late-mineralizing hMSCs differed from www.ecmjournal.orgEuropean Cells and Materials Vol.47 2024 (pages 170-189) DOI: 10.22203/eCM.v047a12early-mineralizing hMSCs.For example, an upregulation in RUNX-2 and OSX, both transcription factors in osteogenesis, was detected at early time points in late-mineralizing hMSCs (days 7 and 10), as well as ALPL and BMP2 expression.In comparison, early-mineralizing hMSCs showed an upregulation of OCN early on and maintained this until later time points (days 3 until 21).SPP1 expression was upregulated in both mineralizing groups at a later time point.However, when comparing gene expression results with the protein multiplex analysis, differences between conditions and mineralization status were detected more clearly on the protein level, confirming that a qPCR is a less powerful assay to characterize differentiation.

Discussion
Inducing osteogenic differentiation in hMSCs has been achieved through multiple approaches, especially by the addition of pro-osteogenic factors to achieve an osteoblast-like phenotype and the secretion of a mineralized extracellular matrix.Unfortunately, there has been a poor correlation between in vitro and in vivo work, suggesting a need to improve how osteogenic differentiation is achieved and investigated in vitro.In vitro differentiation models oftentimes are evaluated using limited readouts to assess osteogenic differentiation, whereas a broader range of markers could offer better insight.Therefore, in this study, we propose a protein multiplex assay to evaluate the phenotype of hMSCs undergoing osteogenic differentiation.We set up a targeted 18-plex protein panel of markers for osteogenesis, angiogenesis, or inflammation in order to characterize different strategies of inducing osteogenic differentiation in hMSCs, to identify relevant patterns, and to gain a better understanding of osteogenic differentiation in vitro.HM-SCs were stimulated with a combination of dexamethasone, www.ecmjournal.orgEuropean Cells and Materials Vol.47 2024 (pages 170-189) DOI: 10.22203/eCM.v047a12β-glycerophosphate and ascorbic acid (OM), with growth factors such as BMP-2 (BM+BMP2 and OM+BMP2) or by supplementing calcium (Ca) or phosphate ions (Pi) or by seeding on a CaP-based biomaterial (CaP), all conditions that have previously been shown to induce osteogenic differentiation.
Mineral deposition, as measured by ARS was used for the classification of these conditions.It is important to note that, in the context of our measurements, we cannot rule out the possibility that some calcium deposition observed may be attributed to the release of calcium upon cell death, which could have contributed to the overall mineralization measurement.Protein patterns were associated with different experimental groups classified as inducers of non-(BM and BM+BMP2), late-(OM and OM+BMP2) and early-(Ca, Pi and CaP) mineralization.It was interesting to note that differences were also found within the mineralization groups, which speaks to the complexity of assessing osteogenic differentiation.For example, many differences in the osteogenic protein profiles between hMSCs in OM and OM+BMP2 were detected until day 14, even though their mineralization status was similar.Their protein profiles related to inflammation and angiogenesis were, however, much more similar.Overall, while osteogenic markers varied substantially within mineralization classifications, the angiogenic and inflammatory markers were more similar according to mineralization status.
No significant difference in LDH-B protein release as compared to the control in BM was detected (Fig. 3a), suggesting the absence of cytotoxicity.This is further supported by the phase contrast images obtained during the culture period (Fig. 1b,c).However, to provide a definitive confirmation of cell viability, it is advisable to conduct additional assays such as an MTT assay.
The protein panel contained nine markers selected to characterize osteogenic differentiation in hMSCs: BMP-2, BMP-9, GAS-6, FGF-23, TGF-beta-1, RANKL, OPG, SPP-1 and MMP-13.Two of the markers tested, BMP-9 and FGF-23, failed to distinguish the differences we observed in the mineralization status between groups.However, our results indicate that the increased expression of these two markers may be linked to the supplementation of BMP-2 to the medium, which was sufficient to stimulate BMP-9 and FGF-23 at early time points (day 3 until 10) in hMSCs in both BM+BMP2 and OM+BMP2, but was insufficient to promote an osteoblastic phenotype.
In contrast, TGF-beta-1 in the supernatant emerged as an interesting factor to correlate to mineral deposition: Its decrease was detected in late-(OM) and early-(Ca, Pi and CaP) mineralizing hMSCs.TGF-beta-1 plays a crucial role in chondrogenesis and bone formation and resorption (Zhao and Hantash, 2011).It promotes osteoprogenitor proliferation and early differentiation, but also inhibits mineralization in the late stage of osteoblast differentiation have shown that TGF-beta-1 supplementation to hM-SCs can decrease their mineralization (Cals et al., 2012; Nam et al., 2020), which is in line with our results.While there are also studies that state the opposite, it is widely accepted that TGF-beta-1 has a broad range of effects on bone metabolism and its level in the supernatant could prove to be a useful predictor of osteogenic differentiation (Jilka et al., 1998; Kasagi and Chen, 2013; Kassem et al., 2000; Pfeilschifter et al., 1990).
We also found two markers, GAS-6 and MMP-13, indicative of osteogenic differentiation, but with different expression patterns in the supernatant and the lysate, underlining the power of this particular assay.First, we found GAS-6 to be an indicative marker for early-mineralizing hMSCs.GAS-6 has previously been reported in osteoblasts to have a stimulatory effect on bone resorption by osteoclasts (Nakamura et al., 1998; Shiozawa et al., 2010).Its expression in the supernatant in the non-mineralizing hM-SCs (BM+BMP2) was increased, whereas its expression in the lysate was decreased.The opposite was true for the early-mineralizing hMSCs (Ca, Pi and CaP).This indicates that GAS-6 may not be released or may remain in the extracellular matrix.Second, MMP-13 in the supernatant was found in conditions lacking mineralization, whereas MMP-13 in the lysate was found in mineralizing hMSCs.MMP-13 in the native bone ECM plays a role in bone remodeling and is a key enzyme for collagen type 1 degradation (Wu et al., 2002).Our results indicate that MMP-13 could be successfully correlated to the differentiated phenotype marked by mineral deposition.
Because we were introducing a novel approach to characterize differentiation in hMSCs, we also wanted to use conventional methodology.ALP activity was therefore quantified as an important enzyme involved in mineral deposition in bone remodeling (Hessle et al., 2002).ALP activity directly correlated to mineralization for latemineralizing hMSCs, but was inversely correlated to mineralization status in early-mineralizing hMSCs.While the upregulation of ALP activity in late-mineralizing hMSCs has been previously explained by the supplementation of dexamethasone in the medium (Wong et al., 1990), it could be that the early upregulation of ALP activity reduced inorganic pyrophosphate, a hydroxyapatite-forming blocker, and supplied inorganic phosphate for hydroxyapatite production (Whyte et al., 1995).These findings may be explained by the known downregulation of ALP caused by high concentrations of phosphate (Liu et al., 2009), or the upregulation known to be caused by BMP-2 supplementation (Vanhatupa et al., 2015).Together, this shows that investigating ALP activity alone is insufficient to determine whether hMSCs are undergoing osteogenic differentiation.
A correlation between the transcript expression of osteogenic markers and the mineralization state of hMSCs was only found for SPP1, the gene coding for Osteopontin that is commonly expressed in osteoblasts, at a late time point (day 21) as it was upregulated in all conditions that European Cells and Materials Vol.47 2024 (pages 170-189) DOI: 10.22203/eCM.v047a12 were able to mineralize.The differences in relative gene expression levels were not as pronounced as the variations observed at the protein level.OCN, which is involved in matrix mineralization by binding hydroxyapatite, as well as SPP1 were upregulated in early-mineralizing hMSCs at several time points (Danoux et al., 2015; Tsao et al., 2017).No differences in any gene expression profiles were found between hMSCs in BM and BM+BMP2.Even though the supplementation of BMP-2 activated certain pathways, the concentration used here, 50 ng/mL, was insufficient to induce osteogenic differentiation in hMSCs on its own.In hMSCs in OM+BMP2, however, the supplementation seemed to enhance osteogenesis in comparison with hM-SCs in OM, which has been previously reported (Vanhatupa et al., 2015).Angiogenic protein marker expression did not correlate to mineralization status of hMSCs but it was similar within mineralization groups, possibly due to the limitations of the in vitro model used in this study.The use of hMSCs in 2D does not recapitulate the complexities of the angiogenic processes in vivo (Liang et al., 2017).However, it can indicate whether the differentiating hMSCs release angiogenic factors which potentially will stimulate angiogenesis in surrounding cells.Here, we measured the expression of VEGF-A and ANG-1, which promote angiogenesis and immunomodulatory functions, and have been shown to improve bone formation (Rundle et al., 2006; Shin et al., 2021).Non-mineralization hMSCs had increased VEGF-A in the supernatant and showed an increase in VEGF-A over time, indicating a pro-angiogenic effect in agreement with previous reports (Krawiec et al., 2015).Furthermore, the increase in TGF-beta-1, which besides being an osteogenesis-related marker is also considered an angiogenic factor that directly targets VEGF-A, supports this claim (Sánchez-Elsner et al., 2001).Late-mineralization hMSCs showed an opposing trend in angiogenic factors than early-mineralization hMSCs at early time points, with an increase in ANG-1 and a decrease of VEGF-A.Hoch et al. (2012) attributed this to the dexamethasone-containing differentiation medium and concluded that differentiated MSCs have diminished angiogenic potential (Hoch et al., 2012).However, we found that hMSCs classified as earlymineralizing maintained their angiogenic potential with an initial increase of VEGF-A.
Inflammation plays an important role in bone healing (Kon et al., 2001; Rundle et al., 2006), and hMSCs are known to secrete immunomodulatory factors such as IL-6, TNF-A or GM-CSF (Shin et al., 2021).The nonmineralizing hMSCs (BM+BMP2) expressed higher IL-6 in the supernatant, suggesting a positive effect on bone healing.IL-6 is a pro-inflammatory cytokine shown to influence the MAPK signaling cascade, which is an essential process for bone formation.It is, however, also correlated with the severity of inflammation (De Benedetti et al., 2006).High IL-6 in the supernatant was also seen in the early-mineralizing group, but not in the late-mineralizing group, which had diminished IL-6 in the lysate in agreement with previous studies (Pricola et al., 2009).Earlymineralizing hMSCs had diminished TNF-A, an inflammatory factor shown to stimulate osteoclast-induced bone resorption with a described inhibitory effect on osteogenic differentiation at high concentrations (Zhao et al., 2011), as well as GM-CSF, a cytokine related to bone resorption that regulates the fusion of mononuclear osteoclasts into boneresorbing osteoclasts (Lee et al., 2009).Combined with the finding on IL-6, which may result in suppression of neutrophil apoptosis, favoring M2 polarization in macrophages in vivo, it could be that these hMSCs could participate in a healing process (Philipp et al., 2018; Raffaghello et al., 2008).Bastidas-Coral et al. (2016) have shown an increased induction of mineralization in hMSCs stimulated with IL-6 (Bastidas-Coral et al., 2016).Therefore, increased IL-6 expression in early-mineralizing hMSCs may also be connected to the increased mineral deposition.

Conclusion
In summary, it was shown that early-and latemineralizing hMSCs undergo osteogenic differentiation, albeit with a difference in the way this is taking place.Nonmineralizing hMSCs in BM+BMP2 did not undergo osteogenic differentiation in our experiments, however, a beneficial effect on angiogenesis and inflammation markers was detected, both relevant for the bone healing process.While late-mineralizing hMSCs in OM and OM+BMP2 had minimal potential to secrete angiogeneic factors, earlymineralizing hMSCs did show an increased effect.Both non-and early-mineralizing hMSCs showed expression of markers related to inflammation.TGF-beta-1 was identified as an interesting marker inversely correlating to mineral deposition.GAS-6 as well as MMP-13 expression were highlighted as markers indicative of osteogenic differentiation.A comprehensive roadmap of osteogenic differentiation was established, and potential markers of interest were highlighted.A targeted proteomic screening approach such as a protein multiplex assay could be an improvement of how osteogenic differentiation is analyzed in vitro and be a first step to improve the poor correlation between in vitro and in vivo results (Hulsart-Billström et al., 2016).This approach may potentially improve the in vitro evaluation of biomaterials for bone regeneration, as has been shown here on the example of hMSCs on a CaP coating.In future research, building on these results, it would be interesting to transition to a 3D model which could further enhance the translational capacity of our findings.Overall, we were able to compare different commonly used in vitro models of osteogenic differentiation in hMSCs, giving a side-by-side comparison that has been lacking.

Fig. 1 .
Fig. 1.Schematic overview of experimental set-up and cell morphology.(a) Different strategies were used to induce osteogenic differentiation in human mesenchymal stem cells (hMSCs) over 28 days.After mineralization assessment by Alizarin red S staining at days 14 and 28, the strategies were classified according to their resulting mineralization status into inducers of non-, late-and earlymineralization, and protein patterns from the multiplex assay were correlated to the mineralization status.Assays that were executed on a specific time point are marked by X.(b) Cell morphology and mineral deposition at day 3. Initial mineral deposits were observed in hMSCs in phosphate ions (Pi) at day 3. (c) Mineral deposits at day 21 were observed in hMSCs cultured in OM, OM+BMP-2, calcium (Ca), and Pi at day 21.Scale bar: 100 µm.
from reagent grade NaCl, CaCl 2 •2H 2 O, MgCl2•6H 2 O, NaHCO 3 and Na 2 HPO 4 •2H 2 O salts, added one after the other, in demineralized water at 50°C to prepare a second five-times concentrated simulated body fluid solution with a decreased concentration of Mg 2+ and HCO 3 − (Yang Detected in hMSCs in OM, OM+BMP-2, Ca, Pi and on CaP HMSCs were subjected to seven different conditions: BM, which served as a non-differentiating control, BM+BMP2, OM, OM+BMP2, Ca, Pi and on CaP.Morphological changes and mineral deposition patterns in hM-SCs were examined over the course of culture time.Initially, mineral deposits were observed in hMSCs in Pi early as day 3 (Fig. 1b).At day 21, mineral deposits were obwww.ecmjournal.orgEuropean Cells and Materials Vol.47 2024 (pages 170-189) DOI: 10.22203/eCM.v047a12

Fig. 2 .
Fig. 2. Characterization of calcium phosphate (CaP) coating and of hMSCs.(a) A homogenous CaP coating was deposited as observed by scanning electron microscope.Scale bar: 10 µm.(b,c) Attenuated total reflectance Fourier-transform infrared spectroscopy (ATR-FTIR) spectrum and X-Ray Diffraction (XRD) pattern show a carbonated apatite.(d) hMSCs are able to undergo trilineage differentiation as show by Oil red O stain at day 21 of culture, Alizarin red S stain at day 28 of culture and Alcian blue stain at day 21 of culture.Scale bars: 100 µm, 100 µm and 200 µm, respectively.

Fig. 3 .
Fig. 3. Increased mineralized matrix was formed in hMSCs in OM, OM+BMP2, Ca, Pi and on CaP, while ALP activity was increased only in hMSCs in OM and OM+BMP2.Phase contrast images of hMSCs stained with Alizarin red S after 14 and 28 days in culture to visualize mineral deposits (a).Data are representative of at least three independent experiments (N = 3) with similar results.Scale bars represent 100 µm.The stain was quantified (b) to classify the conditions as non-mineralizing (basic medium (BM) and BM+BMP2), early-mineralizing (Ca, Pi and CaP) or late-mineralizing (OM and OM+BMP2).ALP activity was quantified after 3, 7, 10, 14 and 21 days in culture (c-g) relative to a housekeeping protein (HSP60) set to 1.0.ALP activity relative to the BM control was decreased in hMSCs in BM+BMP2 after 14 and 21 days.It was increased in hMSCs in OM beginning on day 7, in OM+BMP2 beginning on day 10, in Ca beginning on day 14, and in Pi at day 10.At day 21, ALP activity in hMSCs in Pi and on CaP was decreased relative to the control.Bars represent the mean of three independent experiments (N = 3), error bars represent standard deviation and * indicates p < 0.05 relative to BM.

Fig. 4 .
Fig. 4. Multiplex analysis identified differentially expressed proteins as markers for osteogenic differentiation.Protein multiplex analysis of differentiating hMSCs at days 3, 7, 10, 14 and 21 measured in the cell lysate (LY) and supernatant (SN) (a) Protein data are represented relative to day 0 expression and normalized to a housekeeping protein (HSP60).Differentially expressed proteins (p < 0.05 relative to hMSCs in BM) are indicated with an open white circle.The data represent the mean of three independent experiments (N = 3)."undet": undetected.Principal component analysis was done for all osteogenesis-related protein data (b), angiogenesis-related proteins (c) and inflammation-related proteins (d) at days 7, 14 and 21 (for days 3 and 10, see Fig. 6).

Fig. 5 .
Fig. 5. Principal component analysis (PCA) loading plots showing the contribution of proteins to the principal components.PCA loading plots were generated for osteogenesis-related proteins (a), angiogenesis-related proteins (b) and inflammation-related proteins (c) at days 3, 7, 10, 14 and 21.

Fig. 6 .
Fig. 6.Principal component analysis and overview of protein expression patterns.PCA was done for osteogenesis-related proteins (a), angiogenesis-related proteins (b) and inflammation-related proteins (c) at days 3 and 10 to show the reproducibility of results.(d) Summary of the primary protein expression patterns discussed in this study.Red circles denote decreased protein expression, while green circles signify increased protein expression.

Fig. 7 .
Fig. 7. Osteogenic gene expression over time showed the upregulation of osteogenic markers in late-and early-mineralizing hMSCs.RT-qPCR was done at days 3 (a), 7 (b), 10 (c), 14 (d) and 21 (e).Ct values were normalized to the geometric mean of two housekeeping proteins (RPLP13A and YWHAZ) and represented relative to a control in BM at day 0. The data represents the mean of three independent experiments (N = 3).Differentially expressed proteins (p < 0.05 relative to hMSCs in BM) are indicated with an open white circle.