TMT-based quantitative proteomic analysis revealed that FBLN2 and NPR3 are involved in the early osteogenic differentiation of mesenchymal stem cells (MSCs)

MSCs culture and osteogenic/adipogenic induction

This work received ethical approval from Jiujiang University, Jiangxi, China. As reported previously, MSC extraction, identification, maintenance, and storage were conducted [18]. 6th generation MSCs were cultured for 14 days in MEM α with 10% FBS, 100 mM dexamethasone, 10 mM sodium glycerophosphate, and 50 ng/ml vitamin C to induce osteoblast differentiation, and osteogenesis was detected using Alizarin Red staining (ARS). Next, upon a 15 min stain elution using 10% cetylpyridinium chloride in 10 mM sodium phosphate solution, absorbance was read at 590 nm via a spectrophotometer. To induce adipogenesis, 6^th generation MSCs were cultured in minimum essential medium (MEM) α with 10% fetal bovine serum (FBS), 1.0 μM dexamethasone, 0.5 mM 3-isobutyl-1-methylxanthine, and 0.01 mg/ml insulin for 14 days. The adipogenic potential following adipogenic induction was confirmed through Oil Red O staining.

Protein extraction and digestion

Cells were divided into five groups, namely, undifferentiated MSCs (MSC), osteogenic day 3 (OS3), osteogenic day 7 (OS7), adipogenic day 3 (AD3), and adipogenic day 7 (AD7). Following harvest, cells underwent sonication thrice in lysis buffer on ice via a high-intensity ultrasonic processor (Scientz) (8 M Urea, 1% Protease Inhibitor Cocktail), and debris removal was performed via a 10 min 12,000 g centrifugation at 4° C. A BCA kit was used to quantify the produced supernatant following kit instructions. To digest proteins, they were first reduced for 30 min at 56° C with 5 mM dithiothreitol and then alkylated for 15 min at RT in the dark with 11 mM iodoacetamide. The sample underwent dilution using 100 mM TEAB to a urea concentration <2M. Ultimately, trypsin was introduced at a 1:50 trypsin-to-protein mass ratio for an overnight (ON) initial digestion and a 4 h 1:100 trypsin-to-protein mass ratio secondary digestion.

LC-MS/MS analyses and database screening

LC-MS/MS Analyses and Database Screening were carried out by the PTM Biolab (Hangzhou, China). In short, tryptic peptides were re-suspended in 0.1% formic acid (solvent A) before loading onto a laboratory-made reversed-phase analytical column. Our gradient consisted of an elevation from 6-23% solvent B (0.1% formic acid in 98% acetonitrile) for 26 min, 23-35% for 8 min, and 80% for 3 min before holding at 80% for the remaining 3 min. Subsequently, the peptides were exposed to an NSI source, with subsequent tandem mass spectrometry (MS/MS) in Q Exactive TM Plus (Thermo) attached to the UPLC. The parameters were adjusted: electrospray voltage to 2.0 kV, m/z scan range 350-1800 for a full scan, and intact peptides was identified in Orbitrap at 70,000 resolutions. Peptide selection for MS/MS used an NCE setting at 28, and fragment detection employed a 17,500 resolution in Orbitrap. Then, we used a data-dependent technique that rotated between 20 MS/MS scans with 15.0 s dynamic exclusions and a single MS scan. The parameter was set as follows: Fixed first mass: 100 m/z. The MS/MS data were processed via the MaxQuant search engine (v1.6.5.0). Tandem mass spectra were validated against the combined reverse decoy and human Uniprot databases.

Trypsin/P served as the cleavage enzyme, and a maximum of 4 missed cleavages were permitted. The following parameters were used: precursor ion mass tolerance: 20 ppm (preliminary screening) and 5 ppm (main screening), fragment ion mass tolerance: 0.02 Da, FDR: < 1%, minimum modified peptide score: >40. All experiments were conducted twice.

Bioinformatics analysis

We used different databases or software to perform Gene Ontology (GO), protein domain, and the Kyoto Encyclopedia of Genes and Genomes database (KEGG). GO analyses were conducted via three major classifications: biological process (BP), cellular compartment (CC), and molecular function (MF). KEGG analyses involved the identification of critical networks related to the pre-identified DE proteins. To perform protein domain enrichment analysis, we used InterPro, a software application that categorizes protein sequences into families to determine crucial domains and locations. A p-value of <0.05 was adopted as the significant level in all analyses described above, and a two-tailed Fisher's exact test was employed to identify DE protein enrichment. Lastly, protein-protein associations were assessed using the STRING database (https://string-db.org/).

Western blot (WB)

Following total protein isolation, equal quantities of proteins underwent separation on 10% SDS-PAGE before transfer to PVDF membrane (Bio-Rad, CA, USA), with subsequent ON blocking in 5% non-fat milk at 4° C and 2 h exposure to targeted antibodies at RT. The employed antibodies were as follows: anti-COL11A1 (ab166606, Abcam), anti-COMMD5 (67043-1, Proteintech), anti-SRPX2 (66266-1, Proteintech), anti-LRRC1 (ab127568, Abcam), anti-NPR3 (A8138, ABclonal), anti-APOL2 (25925-1, Proteintech), anti-GGT5 (A14374, ABclonal), anti-FBLN2 (sc-271843, Santa Cruz) and anti-β-actin (ab8227, Abcam). All antibodies mentioned above were used in dilution 1:1000, except for anti-β-actin, which was used in 1:2000 dilution. Corresponding secondary HRP-conjugated antibodies were employed in 1:10000 or 5000 dilutions, and protein visualization was done via an enhanced chemiluminescence system and quantification via Image Lab software.

RNA interference

Per kit directions, MSCs were incorporated with FBLN2- or NPR3-targeted shRNA lentiviral particles (Genechem, Shanghai, China). The target sequences are summarized in Supplementary Table 1. To induce transduction, 5 × 10⁵ cells were plated in 25 cm² in complete medium for an ON incubation, before a 12 h infection with 5 μL of shRNA in 5 g/mL polybrene media mixture, with subsequent refreshing of the old transduction medium to complete medium. Cell harvest was performed 3 days post-transduction for WB-based protein analysis or osteogenic/ adipogenic induction.

Availability of data and materials

Data are available via ProteomeXchange with identifier PXD041404.

Quantitative analysis of protein expression during early differentiation of MSC towards osteogenesis/adipogenesis

We initially assessed the inducible effects of osteogenic/adipogenic induction media to investigate the in vitro model of osteogenesis/adipogenesis. The 6^th generation of MSCs underwent ARS and ORO staining after a 14-day stimulation with the appropriate medium, confirming effective MSC differentiation (Supplementary Figure 1).

Using TMT-based quantification, we identified 609196 secondary spectral proteins. Upon comparison of this data with the theoretical protein data, the number of available spectra was 146661, and the spectral utilization rate was about 24%. In all, 96006 peptides were screened via spectral analysis, among which 92486 specific peptide segments were identified. Moreover, out of the 6943 proteins screened in this investigation, 6543 provided quantitative information (Figure 1A). The samples demonstrated satisfactory reproducibility based on our principal component (PCA) and Pearson correlation coefficient analyses (Figure 1B, 1C). Significance was considered when the inter-group fold change was <0.67 or >1.5, and the t-test p-value was < 0.05. Relative to the MSCs, 86 proteins were highly expressed, and 32 were scarcely expressed in the OS3 group; 104 proteins were highly expressed, and 92 were scarcely expressed in the OS7 group; 209 proteins were highly expressed, and 179 were scarcely expressed in the AD3 group; and 78 proteins were highly expressed, and 57 were scarcely expressed in the AD7 group. In addition, we performed a cross-range comparison between the OS3 and AD3 groups or the OS7 and AD7 groups. In the OS3 group, compared to the AD3 group, we found 187 highly expressed proteins and 222 proteins with low expression. In addition, 269 proteins were expressed at high levels and 321 at low levels in the OS7 group compared to AD7 (Figure 1D).

On the 3^rd and 7^th days of osteogenesis, 38 proteins were significantly increased (over 1.5 folds), whereas, on the 3^rd and 7^th days of adipogenesis, 86 proteins were significantly increased (over 1.5 folds). These highly expressed proteins have potential research value and are detailed in Supplementary Table 2A, 2B.

Bioinformatics analysis the differentially expressed proteins

We mainly focused on DE proteins that showed substantial overexpression according to bioinformatics analysis. Considering BP, the GO stratification of highly expressed proteins between each comparison group was largely similar and comprised physiological regulation, stimulus-response, cellular organization, and cellular, single-organism, and metabolic activities (Figure 2A). However, the subcellular location of highly expressed proteins exhibited considerably more variation between these comparison groups. For instance, (1) between the OS3 and MSC groups, there were 36 elevated proteins localized within the extracellular space, and between the OS7 and MSC groups, the number increased to 60. (2) between the AD3 and MSC groups, there were 50 enhanced proteins localized within the mitochondria, and between the OS7 and MSC groups, the number increased to 72. Lastly, (3) between the OS3 and AD3 or between the OS7 and AD7 groups, the maximum enhanced protein proportions were localized within the nucleus (Figure 2B).

The KEGG pathway enrichment-based cluster analysis revealed that, in the OS3 group, the highly expressed proteins were enriched in the ECM receiver interaction and focal adhesion pathway. In the OS7 group, they were enriched in the ECM receiver interaction and cell adhesion motors. In the AD3 and AD7 groups, the PPAR axis and peroxide were the primary clustering pathways. In addition, the KEGG clusters of DE proteins were substantially different between the osteogenic and adipogenic comparison (OS/AD). The highly expressed proteins in osteogenesis were clustered in the tight junction, focal adhesion, and cell adhesion molecules. In contrast, the highly expressed proteins in adipogenesis were clustered primarily in the fatty acid and glycolide metabolism networks, thus demonstrating distinct protein functions during osteogenic and adipogenic differentiation processes (Figure 3A).

The highly expressed proteins in the OS3 group were clustered in the insulin-like growth factor-binding protein and the EGF domains. In contrast, according to our protein domain clustering analysis, the highly expressed proteins in the OS7 group were clustered in the immunoglobulin-related domains. Highly expressed proteins in the AD3 or AD7 group were clustered in the NAD(P)-binding domain, amine oxidase domain, and so on. Between the adipogenesis and osteogenesis processes, the highly expressed proteins were clustered in numerous domains, such as the mitochondrial carrier domain, acyl-CoA-related domain, Thiolase domain, etc. (Figure 3B).

We determined the top 50 proteins in each comparison group with the highest number of direct associations to plot the protein association axis and better understand the relationships between DE proteins. The corresponding results are presented in Supplementary Figure 2.

Expression validation

We selected random 4 proteins (NPR3, APOL2, GGT5, and FBLN2) that showed substantially high expression during osteogenic differentiation and 4 proteins (COL11A1, COMMD5, SRPX2, and LRRC1) that showed significantly high expression during adipogenic differentiation for WB-based verification. Based on our analysis, the expression profile of the proteins, as mentioned above, during MSCs differentiation was consistent with data from proteomics-seq, thereby suggesting the accuracy of the sequencing data (Figure 4A, 4B).

FBLN2 and NPR3 participate in the osteogenic differentiation of MSC

According to the proteomics-seq, FBLN2 and NPR3 were significant because they had previously received little attention when studying osteogenic differentiation. Using lentivirus-mediated shRNA delivery, we successfully and independently knocked down FBLN2 and NPR3 in these experiments (Figure 4C). FBLN2 or NPR3 deficiency did not influence MSCs proliferation within 7 days post-infection (Data not shown). However, following 48 h of incorporation, MSCs were exposed to osteogenic inducer for 14 days. ARS revealed that FBLN2 or NPR3 knockdown partially suppressed extracellular calcification during osteogenesis (Figure 4D). Using quantitative analysis of ARS, we demonstrated that FBLN2 deficiency reduced ARS by approximately 60% and NPR3 deficiency by 65% (Figure 4E).

Functions of FBLN2 or NPR3, as evidenced by quantitative proteomics analysis

We used quantitative proteomics to evaluate the DE proteins on day 7 of MSC osteogenic differentiation after FBLN2 or NPR3 deficiency to investigate the underlying processes behind the FBLN2- and NPR3-mediated impact on MSC osteogenic differentiation. Overall, 381313 peptides were screened by spectral analysis, among which specific peptide segments were observed in 53353 proteins. In all, 6447 proteins were screened in this investigation, among which 6330 proteins provided quantitative information (Figure 5A). Compared to the parental control cells, FBLN2 knockdown resulted in 590 highly expressed proteins, 775 scarcely expressed proteins (by 1.2-fold), 102 highly expressed proteins, and 115 scarcely expressed proteins (by 1.5-fold). In the NPR3 knockdown cells, 1174 and 1235 proteins were markedly elevated and diminished (by 1.2-fold), respectively, or 240 and 178 proteins were strongly elevated and decreased (by 1.5-fold), respectively, relative to the parental control cells (Figure 5B).

The DE proteins were then subjected to bioinformatics analysis. Concerning the BP category, DE proteins were primarily enriched in physiological modulation and stimulus-response between the FBLN2 deficient and control groups at OS7. In the NPR3 deficient versus control group, DE proteins were increased in the cellular and metabolic processes (Figure 6). Based on the KEGG network enrichment analysis, DE proteins between the FBLN2 deficient and control groups were relatively scattered, including cancer-associated pathways, PI3K-AKT axis, focal adhesion, etc. By comparison, over 10% of DE proteins between the NPR3 deficient and control groups were clustered in the metabolic pathway (Figure 7).

The PPI networks of DE proteins were assessed using the STRING network. Among DE proteins between the FBLN2 deficient and control groups, we selected 35 scarcely expressed proteins associated with the focal adhesion pathway for PPI analysis. The ITG and collagen protein families, in particular, were employed to construct the major PPI complex network (Figure 8A). Among DE proteins between the NPR3 deficient and control groups, we selected 146 scarcely expressed proteins associated with the metabolic pathway for PPI analysis. The associations between DE proteins were observed to be more dispersed and complex (Figure 8B).