Comparative proteomic analysis of osteogenic differentiated human adipose tissue and bone marrow‐derived stromal cells

Abstract Mesenchymal stromal cells are promising candidates for regenerative applications upon treatment of bone defects. Bone marrow‐derived stromal cells (BMSCs) are limited by yield and donor morbidity but show superior osteogenic capacity compared to adipose‐derived stromal cells (ASCs), which are highly abundant and easy to harvest. The underlying reasons for this difference on a proteomic level have not been studied yet. Human ASCs and BMSCs were characterized by FACS analysis and tri‐lineage differentiation, followed by an intraindividual comparative proteomic analysis upon osteogenic differentiation. Results of the proteomic analysis were followed by functional pathway analysis. 29 patients were included with a total of 58 specimen analysed. In these, out of 5148 identified proteins 2095 could be quantified in >80% of samples of both cell types, 427 in >80% of ASCs only and 102 in >80% of BMSCs only. 281 proteins were differentially regulated with a fold change of >1.5 of which 204 were higher abundant in BMSCs and 77 in ASCs. Integrin cell surface interactions were the most overrepresented pathway with 5 integrins being among the proteins with highest fold change. Integrin 11a, a known key protein for osteogenesis, could be identified as strongly up‐regulated in BMSC confirmed by Western blotting. The integrin expression profile is one of the key distinctive features of osteogenic differentiated BMSCs and ASCs. Thus, they represent a promising target for modifications of ASCs aiming to improve their osteogenic capacity and approximate them to that of BMSCs.


| INTRODUC TI ON
The treatment of bone defects caused by infection, trauma or neoplasms remains a clinical challenge. Autologous bone transplantation is limited by availability of donor sites, with iliac crest being the most common, apart from donor site morbidity that restricts the size of transplants, as well as the surgical risk factors. 1 This has given rise to stromal/stem cell-based therapy. 2 Adult mesenchymal stromal cells (MSCs) can be harvested from different tissues such as bone marrow, adipose tissue, dental pulp and other tissues. 3 They have stem-like properties and are able to undergo differentiation into different mature mesenchymal cell types, given certain conditions and stimuli. 4 In 2006, the International Society for Cellular Therapy (ISCT) proposed minimum criteria for classification as mesenchymal stromal cells. They must be plastic-adherent (eg to a tissue culture flask), express surface markers CD73, CD90 and CD105 (≥90%), lack haematopoietic lineage markers CD14, CD34, CD45, CD19 and HLA-DR (≤2%) and should be able to differentiate into mesodermal lineage (osteogenic, adipogenic and chondrogenic). 5 Lately, paracrine effects of MSCs have gained attention as an important mode of action, as exosomes represent a way of cell-free regenerative therapy. 6 Bone marrow-derived stromal cells (BMSCs) have been studied to a large extent and show a high regenerative potential, although their use is still limited by availability of donor sites for bone marrow aspiration, morbidity of the procedure-although lower than for bone grafting 7 -and the relatively low cell yield, as they represent <0.1% of cells harvested from bone marrow aspirate. 8,9 At the same time, they are the closest and most obvious mesenchymal stromal cells for bone tissue engineering, given their tissue origin, and unlike other mesenchymal stromal cells their ability to support formation of haematopoietic marrow. 10 Adipose tissue-derived stromal cells (ASCs) as part of the stromal vascular fraction of adipose tissue can likewise undergo osteogenic differentiation and may be isolated in sufficient quantities from lipoaspirates after liposuction. Here, it has been shown that there are no major differences in regard to proliferation or differentiation capacity of ASCs derived from subcutaneous fat of different anatomical regions. 11 It has been shown that BMSCs are more prone to senescence during expansion and passage and more affected by ageing in terms of proliferative capability than ASCs, while at the same time osteogenic differentiation capacity is reported to be the lineage least impacted by age. 12,13 Multiple studies have compared the characteristics of these two mesenchymal stromal cells in regard to bone tissue engineering in vitro. Most studies point to inferior extracellular matrix mineralization and lower expression of key osteogenic transcription markers like Runx2 in osteogenically differentiated ASCs compared to BMSCs. 14,15 An intraindividual comparison of human MSCs of three donors cultured on decellularized porcine bone confirmed superior osteogenic capacity of BMSCs compared to ASCs. 16 On the other hand, a study by Rath et al found better osteogenic differentiation of ASCS compared to BMSCs using 3D bioglass scaffolds as a particular culturing condition. 17 Brennan et al isolated human BMSC from bone marrow aspirates and ASCs from lipoaspirates of healthy donors and characterized the cells based on surface markers and tri-lineage differentiation as outlined above. In an ectopic nude mouse model, BMSCs but not ASCs were able to induce ectopic bone formation. 18 In a critical size defect model of sheep tibia, application of ovine BMSCs isolated from bone marrow aspirates resulted in a significantly higher amount of newly formed bone tissue than application of ovine ASCs isolated from excised subcutaneous fat tissue. 19 Importantly, osteogenically differentiated ASCs do not support the formation of a hematopoietic marrow. 10,20 Proteomics enables large-scale analysis of proteins present in a cell type in trying to gain mechanistic insight as to the underlying reasons for functional differences and can be used to identify differentially regulated key proteins in a comparative approach. To overcome this need, an intraindividual comparative data-independent acquisition (DIA)-based proteomic analysis of osteogenic differentiated human BMSCs and ASCs was performed in this study.

| MATERIAL S AND ME THODS
The study was approved by the ethics committee of the Ruhr University Bochum (Approval number: 5045-14) and was conducted according to the Declaration of Helsinki. Written consent was obtained from all patients included in the study.
The study setup is illustrated in Figure 1A. In patients undergoing autologous bone transplantation from the iliac crest in the departments for trauma surgery and plastic surgery/hand surgery, cancellous bone that had been removed in excess and a small amount of subcutaneous fat from the surgical site at the iliac crest were harvested as paired samples. Patients aged 18-89 were eligible as study participants. No exclusion criteria were applied.

| Cell isolation
ASC and BMSC isolation was performed following modified standard protocols as described in our data article on the generation of the spectral library for this project. [25][26][27][28][29] In brief, adipose tissue was rinsed with pre-warmed (37°C) PBS (PAN Biotech, Germany). Blood vessels and connective tissue were carefully detached and discarded, before the tissue was minced. Samples were then weighed. 5 mL of collagenase IV (1 mg/mL, Cell Systems, Germany) per gram of tissue F I G U R E 1 A, Graphical study setup. Harvest of cancellous bone and subcutaneous fat samples from patients undergoing autologous bone graft surgery and subsequent isolation of ASCs/BMSCs, osteogenic differentiation and proteome analysis. B, Flow chart of identified and filtered proteins. C, Venn diagram of proteins quantified in 80% of ASCs, BMSCs and both cell types material was utilized for digestion. This was followed by incubation

| Osteogenic differentiation
Passage three hASCs and hBMSCs were utilized for osteogenic differentiation. 18 For this purpose, cells were washed and then detached using 0.05% Trypsin/EDTA (PAN Biotech, Germany). Cell suspension was centrifuged at 400 x g for 5 minutes at room temperature and the cell sediment resuspended in base medium. Cells were plated with a density of 2 × 10 5 cells per 10 cm culture dish.
Afterwards, osteogenic differentiation was induced by culturing the cells for 21 days in osteogenic differentiation medium consisting of the base medium, 10 mmol/L β-glycerophosphate, 100 nmol/L dexamethasone and 250 µmol/L ascorbic acid. Then, cells were washed and mechanically detached cautiously using a cell scraper.
Cells were collected and centrifuged at 160 x g for 5 minutes at 4°C. Supernatant was aspirated, and samples were stored at −80°C.
After 21 days culturing in osteogenic differentiation medium, extracellular matrix mineralization in hASCs and hBMSCs was examined. To this end, cells were carefully rinsed with DPBS and fixed with 4% paraformaldehyde (PFA) solution for 30 minutes at room temperature. Subsequently, cells were washed with distilled water and 2% Alizarin Red S solution (Sigma-Aldrich, Germany) was added. Cells were incubated for 45 minutes in darkness, before the staining solution was removed and cells washed four times with distilled water.

| Adipogenic differentiation
After preparation of cells in the same form as for osteogenic differentiation, adipogenic differentiation was induced by culturing cells for 21 days in base medium supplemented with 10 μg/mL Insulin, 1 μmol/L dexamethasone, 200 μmol/L indometacin and 0.5 mmol/L of 3-isobutyl1-methylxanthin (AppliChem, Germany). 30 Afterwards, Oil red O staining was performed to visualize lipids in the vacuoles of the cells. Briefly, cells were washed twice with PBS, fixed with 10% formaldehyde for 10 minutes and then stained with Oil Red O solution (Sigma-Aldrich) for 15 minutes. The cells were subsequently washed twice with distilled water.

| Sample preparation for protein analysis
Spectral library-For creation of a spectral library, subcutaneous fat and cancellous iliac bone specimen of a healthy, non-smoker, 24-year-old male patient was retrieved during an autologous bone transplantation procedure. Cell isolation, expansion and osteogenic differentiation were performed as described above.
The methods for the creation of the spectral library are described elsewhere. 29 Patient specific samples-The patient samples were prepared in the same way as the samples for the spectral library, followed by protein concentration determination by Bradford assay. These samples were used for data-independent acquisition (DIA)-based mass spectrometry as well as for Western blot analysis.

| Data analysis of patient samples using the generated spectral library
Preparation of the patient specific samples for DIA-based measurements was performed analogously to the samples for preparing the spectral library, with minor changes. 20 µg protein was loaded on the SDS gel, and electrophoresis was stopped after 15 minutes obtaining shorter gels compared to the approach described earlier for the spectral library. Afterwards, in-gel trypsin digestion and peptide extraction were performed as described earlier. From the resulting peptide extract, 2 µL was used for determination of the peptide concentration by amino acid analysis, as described by Steinbach et al 31 Samples were prepared for mass spectrometry, and 1 µL of iRT-peptide (Biognosys AG, Switzerland) was added to each sample.
For mass spectrometric analysis, again a Q Exactive HF™ (Thermo Fisher Scientific Inc, USA) mass spectrometer was used and operated in DIA mode. The full MS1 scan ranged from 350 to 1200 m/z at a resolution of 120 000. Fragment ions were generated by HCD at a resolution of 30 000 and a stepped NCE of 25.5%, 27% and 30%, respectively. Default charge state was set to ≥+4, and first fixed mass was set to 200 m/z (ACG 1e6, maximum injection time 20 ms). The dataset has been uploaded to ProteomeXchange with the identifier PXD015223.
Data evaluation was carried out with the interface of Spectronaut™ Pulsar under standard settings. In short, the spectral library generated here was taken as a reference database and false discovery rate (called Qvalue) was set to a threshold of 1%. Proteins that could be quantified in at least 80% of samples from each cell type were used for further statistical evaluation. As additional filter criteria, fold changes (>50%) and adjusted P-values (<5%) based on the Benjamini-Hochberg method were calculated manually.

| Western blot analysis of patient specific samples
Immunoblotting was performed, as described earlier for GFAP detection by Kurz et al, 32 with minor changes. 50 µg of protein lysate was separated using 10% Bis-Tris gels according to manufacturer's recommendations (Life Technologies, Germany). Proteins were transferred to nitrocellulose membranes using the iBlot transfer system (Thermo Fisher Scientific) followed by incubation in StartingBlock™ (Pierce, Woburn, USA) for 30 minutes and subsequent probing with primary antibody for 2 hours. For this, the primary antibodies were diluted in 50% TBS buffer/50% StartingBlock™. To remove unbound primary antibodies, the nitrocellulose membrane was washed three times for 10 minutes before incubation, followed by incubation with the fluorescent secondary antibody in 50% TBS buffer/50% StartingBlock™ for 1 hour. Finally, the membrane was washed in TBS buffer three times for 10 minutes. The Odyssey™ system (LI-COR Biosciences GmbH, Germany) was used for fluorescence read out.

| Pathway analysis
Pathway analysis was performed for all proteins up-regulated ≥1.5 fold in BMSCs, using the Reactome database. 33 In order to reference the number of up-regulated proteins per pathway, the number of quantified proteins for each pathway was also determined.

| RE SULTS
The study involved 30 donor patients, after one patient being excluded due to insufficient yield in the process of protein isolation.
Samples of one patient were used to generate a spectral library, while analyses were performed on samples of 29 patients. Median age was 52 (range 22-85), and 21 of the patients were male. Mean BMI was 28.7 ± 5, and 15 of the patients were smokers. Patient characteristics are presented in Table 1.

| Identification of ASCs/BMSCs
To characterize the hASCs and hBMSCs (passage 3), flow cytometry was used to identify the expression of different cell surface markers.
Here, expression of typical mesenchymal stem cell markers such as CD90 and CD105 and lack of expression of haematopoietic cell surface markers such as CD14, CD11b, CD34 and CD45 were analysed.

| Comparative proteomic analysis
Of the patients, a 24-year-old healthy non-smoker with no medical history who had undergone bone grafting for treatment of a scaphoid non-union was chosen as reference patient to create a spectral library for DIA-based proteomic analysis of cell samples.
Here, 96 546 peptides were identified, which could be assigned to 7162 proteins. This patient was not included in further comparative analysis. A flowchart of the filtering process is presented in  Table 2. Of these, the 10 proteins with the highest mean abundance per cell type are presented in Table 2.
Comparing the abundance of 2095 proteins quantified in ≥80% of both cell types revealed 281 proteins with a fold change of at least 1.5 and statistical significance after application of a Benjamini-Hochberg correction with a 5% false detection rate. Of these, 204 were more abundant in BMSCs while 77 were more abundant in ASCs.

| Functional and pathway analysis
Results of Reactome overrepresentation pathway analysis of the 204 proteins with higher abundance (fold change ≥1.5, P < 0.05 after Benjamini-Hochberg correction with a 5% false detection rate) in BMSCs are presented in Figure 3A. The abundance of the 20 proteins with the highest fold change is presented as boxplots in Figure 4B, with the corresponding data in Table 3.
Western blot analysis of integrins alpha 3, alpha 5, alpha 7 and alpha-11 as functionally relevant proteins with a highly significant difference in regulation was performed to validate the results of proteomic analysis. Here, significantly higher protein levels of all 4 Integrins in BMSCs compared to ASCs could be confirmed; results are shown in Figure 4C.

| D ISCUSS I ON
The aim of this study was to utilize intraindividual comparative proteomics to identify key proteins that are differentially expressed between ASCs and BMSCs and are potential candidates for improvement of the osteogenic potential of ASCs. We were able to successfully isolate cells from subcutaneous fat tissue and cancel- Integrin β1 itself has also been shown to be essential for osteoblast mineralization in mice. 53,54 Integrin α9β1 has been shown to mediate osteogenic effects of fibrinogen by Runx2 activation. 55 While there were a total of 5 different alpha integrins that were up-regulated in BMSCs compared to ASCs, integrin alpha-11 (ITGA11) showed a 2.76-fold change and was highly significant even after Bonferroni correction and was thus noteworthy. ITGA11 is one of the main mediators of cell adhesion of MSCs to collagen I and is up-regulated upon osteogenic differentiation. Its silencing leads to a marked decrease in MSC survival and in focal adhesion kinase activity. 56 Also, deficiency of integrin α2 and α11 leads to dwarfism with functional impairment of bone and systemic decrease in insulin-like growth factor concentration. 57 ITGA11 has only recently been identified as a receptor of osteolectin, an osteogenic growth factor that also has been discovered recently, and is required for maintenance of adult skeleton and osteogenic potential of BMSCs. 58,59 It has been shown that ITGA11 signalling activates the canonical Wnt pathway, and blockage of the latter also annihilates the osteogenic effect of osteolectin. 58 While up-regulation of multiple integrins has been found in transcriptome analysis of human ASCs upon osteogenic differentiation, ITGA11 was not up-regulated. 43 How altering ITGA11-expression in ASCs affects their osteogenic potential is part of an ongoing study.
While multiple approaches to improve osteogenic potential of ASCs have been undertaken, one such is hypoxic preconditioning, which has been shown to improve proliferation and osteogenesis. 60 Remarkably, a study on BMSCs similarly demonstrated positive  This study has high statistical power, given the inclusion of 30 patients and identification of more than 7000 proteins. We were able to identify integrin expression profile as one of the key differentiators between osteogenically differentiated ASCs and BMSCs, and indicate its functional relevance, in the context of previous studies and the present literature.
In this study of intraindividual proteomic analysis of osteogenic differentiated human ASCs and BMSCs, we were able to identify integrin expression profile as one of the key differentiators. Further research is needed to investigate the role of integrins in general, and particularly integrin α 11, in osteogenesis of ASCs, and their potential as therapeutic targets to approximate osteogenic capacity of ASCs to that of BMSCs.

CO N FLI C T O F I NTE R E S T
The authors have no potential conflicts of interest to report.

DATA AVA I L A B I L I T Y S TAT E M E N T
The proteomic dataset has been uploaded to ProteomeXchange with the identifier PXD015223. Tables S1 And S2 contain all proteins quantified in ≥80% of both cell types and all proteins quantified in ≥80% of only one cell type, respectively. Any additional data are available upon reasonable request.