Changes in proteins and gene activity in stem cell secretions from ligament cells in a lab model of gum disease

Human PDLSCs were purchased (CELPROGEN, CA, USA) and cultured in α-Minimum Essential Medium (α-MEM, Gibco-BRL, Gaithersburg, MD, USA) containing 10% fetal bovine serum (FBS, Gibco-BRL) as previously described [22]. PDLSCs at passages 4–6 were used for all experiments. To perform each experiment, cells were cultured in an osteogenic medium (α-MEM containing 5% FBS, 50 μg/ml ascorbic acid, 1 μM dexamethasone, and 3 mM β-glycerophosphate). For the experiments involving PDLSC-CM and LPS (from Porphyromonas gingivalis) treatment, cells were divided into four groups: osteogenic medium (Control), osteogenic medium + LPS (5 μg/ml) (LPS), PDLSC-CM, and LPS + PDLSC-CM. For the PDLSC-CM and LPS + PDLSC-CM groups, PDLSC-CM was supplemented with the osteoinductive factors mentioned above, corresponding to the composition of osteogenic medium. Culture media were changed every 2 days. All reagents and laboratory consumables were obtained from the Sigma Chemical Company (St. Louis, MO, USA) and SPL Lifescience (Pocheon, Korea), respectively.

PDLSC-CM was prepared from cells grown under healthy culture conditions. PDLSCs were cultured in 100-mm cell culture dishes until they reached 70% confluency and were then washed with phosphate-buffered saline (PBS), after which they were cultured in serum-free α-MEM without FBS and antibiotics for 48 h. Culture supernatants were then collected and centrifuged at 3,000 rpm for 5 min to remove cell debris. The resulting medium was defined as PDLSC-CM for all experiments. In LC–MS/MS analysis, control medium (as Control) was obtained by collecting the culture medium (serum-free α-MEM) from culture dishes without cells after 48 h of incubation.

Prior to digestion, proteins were processed using filter-aided sample preparation (FASP) on a Microcon 30 K centrifugal filter device (Millipore, Billerica, MA, USA) and reduced with Tris(2-carboxyethyl)phosphine (TCEP) at 37℃ for 30 min [23]. Each sample was then alkylated with iodoacetic acid (IAA) at 25℃ for 1 h in the dark and washed with lysis buffer and 50 mM ammonium bicarbonate (ABC). The proteins were then digested with trypsin at 37℃ for 18 h. The digested peptides were then desalted using C18 spin columns (Harvard Apparatus, Holliston, MA, USA) and eluted with 80% acetonitrile in 0.1% formic acid.

LC–MS/MS analysis was performed according to a previous report [23]. Briefly, the digested peptides were resuspended in 0.1% formic acid and analyzed using a Q-Exactive Orbitrap hybrid mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) with an Ultimate 3000 system (Thermo Fisher Scientific, Waltham, MA, USA). We used a 2 cm × 75 μm ID trap column packed with 3 μm C18 resin and a 50 cm × 75 μm ID analytical column packed with 2 μm C18 resin to separate peptides according to their hydrophobicity. Data-dependent acquisition was performed, and the top 10 precursor peaks were selected for fragmentation. Ions were scanned in high resolution (70,000 in MS1 and 17,500 in MS2 at m/z 400), and the MS scan range was 400–2,000 m/z at both the MS1 and MS2 levels. Precursor ions were fragmented with 27% normalized collisional energy. Dynamic exclusion was adjusted to 30 s. Each MS/MS raw files were assessed using Proteome Discoverer™ software (ver. 2.5), and the Homo sapiens database was downloaded from Uniprot. The workflow of proteome data analysis included a peptide-spectrum match (PSM) validation step and SEQUEST HT as a database search algorithm. Cut-offs below a false discovery rate (FDR) of 1% were adopted, and sequences were filtered to peptides containing at least 6 residues. The relative amount of the proteins among samples was calculated by label-free quantitation.

To evaluate the osteogenic differentiation of PDLSCs, alkaline phosphatase (ALP) activity was evaluated as described in a previous study [22]. In brief, the total protein was extracted, and its concentration was determined. Then, 200 μl of p-nitrophenylphosphate (pNPP) was added to each sample and incubated for 30 min at 37 °C. The mixture was then stopped with the addition of 3 M NaOH, and the optical density was measured using a spectrophotometer at 405 nm. The ALP activity was calculated as mM/100 μg of protein.

The intracellular calcium level was assessed according to our previous report [22]. Briefly, cells were cultured in every experimental condition for 7 days, and the intracellular calcium concentration was measured using a calcium assay kit (BioAssay Systems, Hayward, CA, USA) according to the manufacturer’s instructions. The absorbance was then evaluated at 612 nm. The calcium content was calculated as mg/100 mg of protein.

To evaluate calcium deposits in cell cultures, Alizarin Red S Staining was performed according to our previous study [22]. The cells were washed with PBS and fixed in 4% paraformaldehyde for 15 min after a 14-day incubation. The cells were then rinsed three times with PBS and stained with 2% Alizarin Red S solution (pH 4.2) for 5 min at room temperature. The stained cells were washed and visualized using a light microscope. The stained area was measured with Image J software (National Institutes of Health, Bethesda, MD, USA).

Real-time reverse transcription polymerase chain reaction (RT-PCR) was performed after RNA extraction and cDNA synthesis to validate RNA sequencing results as previously described [22]. Real-time RT-PCR was performed according to the instructions provided with the QuantiTect SYBR Green PCR kit (Qiagen) with an iCycler iQ Multi-Color Real-Time Detection System (Bio-Rad). The thermal cycling conditions were 95 °C for 30 s, 95 °C for 5 s, 55 °C for 30 s, and 72 °C for 30 s for 30 cycles. The primers were as follows: 5′-CGCTTCGATGACTGGTACCT-3′ (sense) and 5′-TAGGCCTCCAAGGACTGGAA-3′ (antisense) for endoplasmic reticulum (ER) degradation-enhancing alpha mannosidase-like protein 2 (EDEM2); 5′-CAGCTCTTTCCTCCAACCCT-3′ (sense) and 5′-GAAATTCCCGGAGCTCCAGA-3′ (antisense) for C-X-C motif ligand 3 (CXCL3); 5′-CTGCACAGATGAGAGACAAATTCC-3′ (sense) and 5′-GAAGCTGCAAAGATCCCAATG-3′ (antisense) for interleukin 11 (IL11); 5′-CTGGGACAGCGCCACATTCGCCGGAGGCGG-3′ (sense) and 5′-TCCGCAGAAAGCAGCCATAGGGGGTAGGCT-3′ (antisense) for a disintegrin and metalloproteinase 15 (ADAM15); 5′-TGTTAAGGCCATAGCTGCGT-3′ (sense) and 5′-TCGCACAGACACCTGGAAAA-3′ (antisense) for mohawk homeobox (MKX); 5′-CGCCTCTTCTTATCAAGCTCGTG-3′ and (sense) and 5′-GAAGCTGTCGTAATTCTGCCAGG-3′ (antisense) for phosphoinositide-3-kinase regulatory subunit 1 (PIK3R1); and 5′-GCTCTCCAGAACATCATCC-3′ (sense) and 5′-TGCTTCACCACCTTCTTG-3′ (antisense) for glyceraldehyde-3-phosphate dehydrogenase (GAPDH).

The extracted protein samples were separated by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred electrically to polyvinylidene difluoride (PVDF) membranes. To identify the osteogenic differentiation of PDLSCs, osteocalcin (OCN), osterix (OSX), and Runt-related transcription factor 2 (RUNX2) protein levels were analyzed using primary and secondary antibodies (goat anti-rabbit immunoglobulin G (IgG) and goat anti-mouse IgG conjugated to horseradish peroxidase). The antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA) and Cell Signaling Technology (MA, USA).

RNA sequencing was performed as described previously [24]. In brief, total RNA was extracted from PDLSCs cultured under different conditions, and the 500 ng of total RNA was prepared. The sequencing library was constructed using a QuantSeq 3′ mRNA-Seq Library Prep Kit (Lexogen, Inc., Austria) according to the manufacturer’s specifications. The library was amplified to add the complete adapter sequences required for cluster production. To sequence the libraries, they were mixed, denatured with NaOH to achieve single-stranded DNA, and then sequenced using an Illumina NextSeq 500 (Illumina, Inc., USA).

The reads generated in the RNA-seq data were aligned to a Bowtie2 index in which the reference genome and gene model annotation files (obtained from the Gene Expression Omnibus [GEO] website) were used for annotation [25]. Differentially expressed genes (DEGs) were assessed based on counts from unique and multiple alignments using the coverage command in Bedtools [26]. The RC (read count) data were obtained from the quantile normalization using the Bioconductor software package EdgeR [27]. The analyses of gene classification, ontology, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were processed by DAVID (http://david.abcc.ncifcrf.gov/↗). Data extractions and image visualizations were accomplished using ExDEGA (Ebiogen Inc., Korea). The protein–protein interaction (PPI) networks were constructed using STRING v3.8.2. A PPI score of > 0.9 (P < 0.05) was considered significant. The PPI networks were imaged using Cytoscape software (http://www.cytoscape.org↗).

The data are presented as means ± SD (n ≥ 3) using the Student’s t-test. One-way analysis of variance was applied for multiple comparisons (Duncan’s multiple range test). A P-value < 0.05 was considered statistically significant. The figures shown are representative of the data.

An LC–MS/MS proteomics approach was first performed to identify differentially abundant proteins in PDLSC-CM compared to control-CM (serum-free α-MEM). Among the 203 total proteins detected (Supplementary data 1), we distinguished 187 significant protein changes according to a fold change cut-off of ≥ 2 and a P-value < 0.05 (Supplementary data 2). Of these secreted proteins, 121 were present in increased levels in PDLSC-CM compared to control-CM, and 66 were present in decreased levels (Fig. 1A). We grouped these abundantly secreted proteins into broad categories based on various cellular activities, including angiogenesis (11 proteins), cell cycle (4 proteins), cell differentiation (52 proteins), cell migration (16 proteins), extracellular matrix (ECM) (72 proteins), ossification (11 proteins), and osteogenesis (8 proteins) (Fig. 1B). The top 10 PDLSC-secreted proteins were further classified and are listed in Table 1. Among these paracrine proteins, key components in the ECM including fibronectin and type I collagen were the abundantly secreted proteins from PDLSCs; it is known to participate in the bone formation and periodontal tissue regeneration [28–30]. In addition, keratin, one of major cytoskeleton related to stem cell proliferation and osteogenic differentiation of dental tissue-stem cells [31, 32], found at large number of peptides. Based on these proteomics profiling results, PDLSC-CM was determined to contain an extensive array of proteins that are potential candidates for regulating PDLSC activity and stimulating periodontal tissue regeneration.

We next verified whether the secretome/PDLSC-CM can regulate the differentiation potential of PDLSCs in the context of an in vitro inflammatory microenvironment. Under LPS-induced inflammatory conditions, the osteogenic differentiation of PDLSCs was significantly inhibited, as shown by the ALP activity (Fig. 2A); furthermore, the intracellular calcium levels ([Ca²⁺]_i) (Fig. 2B), the protein expression levels of osteogenic factors such as OCN, OSX, and RUNX2 (Fig. 2C), and extracellular calcium deposits (Fig. 2D) were downregulated in the LPS-treated group. However, when individual cell cultures were treated with PDLSC-CM, the osteogenic activity of the PDLSCs was markedly upregulated compared to the control group. Moreover, the application of PDLSC-CM relieved the decrease in the osteogenic differentiation of PDLSCs under the LPS-stimulated inflammatory conditions. These findings suggest that PDLSC-CM represents a functional formulation of the PDLSC secretome that can enhance the osteogenic differentiation potential of PDLSCs and could be considered a cell-free therapeutic resource for periodontal tissue repair and regeneration.

We then distinguished the DEGs in the control vs. LPS, control vs. PDLSC-CM, and LPS vs. LPS + PDLSC-CM groups by RNA sequencing analysis (with fold changes ≥ 2 and P-values < 0.05 representing significant differences), resulting in the identification of 252 (79 upregulated and 173 downregulated), 1,326 (571 upregulated and 755 downregulated), and 776 (568 upregulated and 208 downregulated) DEGs, respectively (Fig. 3A and B, Supplementary data 3). In addition, 290 of common genes (5 upregulated, 2 downregulated, and 283 contra-regulated) among the different conditions were identified (Supplementary data 4). The transcript profiling was then validated by real-time RT-PCR, which verified that the mRNA expression levels of EDEM2, CXCL3, IL11, ADAM15, MKX, and PIK3R1 were consistent with the RNA-seq data in the different comparisons (Fig. 3C).

To assess the gene functions, we performed GO enrichment analysis of the DEGs in the different comparison groups. GO terms were arranged according to biological process and molecular function classifications, and the top 10 enriched GO terms were identified. The predominant biological processes included protein phosphorylation and apoptotic process in the control vs. LPS group and transcription regulation in the control vs. PDLSC-CM and LPS vs. LPS + PDLSC-CM groups (Fig. 4A). In the molecular function category, the cell organelle and protein binding terms were enriched in each comparison (Fig. 4B).

We also analyzed the top 20 KEGG enrichment pathways of the DEGs in the three comparisons, the results of which are presented in Fig. 5. The DEGs in the control vs. LPS group included the ubiquitin-mediated proteolysis, epidermal growth factor receptor family (ErbB) signaling, steroid biosynthesis, tumor necrosis factor (TNF) signaling, and Fc epsilon RI signaling pathways (Fig. 5A). The DEGs in the control vs. PDLSC-CM group were enriched in the endocytosis, base excision repair, regulation of actin cytoskeleton, focal adhesion, and mRNA surveillance pathways (Fig. 5B). The DEGs of the LPS vs. LPS + PDLSC-CM group included the protein processing in the ER, cell cycle, p53 signaling, mammalian target of rapamycin (mTOR) signaling, and Wnt signaling pathways (Fig. 5C).

A functional PPI network was constructed with the DEGs identified using the results of STRING analysis combined with confidence scores. The predicted PPI network in each comparison was mapped with nodes (representing proteins) and edges (representing interactions) (Fig. 6). There were 242 nodes and 58 edges in the control vs. LPS group (Fig. 6A), 1,273 nodes and 1,265 edges in the control vs. PDLSC-CM group (Fig. 6B), and 739 nodes and 354 edges in the LPS vs. LPS + PDLSC-CM group (Fig. 6C). The top 5 hub proteins in each comparison were also selected (Table 2). The proteins with the highest numbers of interactions were ubiquitin-conjugating enzyme E2C (UBE2C) (control vs. LPS group), heat shock protein 90-kDa alpha family class A member 1 (HSP90AA1) (control vs. PDLSC-CM group), and heat shock protein family A (Hsp70) member 5 (HSPA5) (LPS vs. LPS + PDLSC-CM group).

Not applicable.

Not applicable.

The authors declare no competing interests.

The sequence data supporting the findings of this study are available within the paper and its supplementary information files. Scanned images of X-film for western blots are also provided within the supplementary information files.