Gingival epithelial cell-derived microvesicles activate mineralization in gingival fibroblasts

Soft tissue calcification occurs in many parts of the body, including the gingival tissue. Epithelial cell-derived MVs can control many functions in fibroblasts but their role in regulating mineralization has not been explored. We hypothesized that microvesicles (MVs) derived from gingival epithelial cells could regulate calcification of gingival fibroblast cultures in osteogenic environment. Human gingival fibroblasts (HGFs) were cultured in osteogenic differentiation medium with or without human gingival epithelial cell-derived MV stimulation. Mineralization of the cultures, localization of the MVs and mineral deposits in the HGF cultures were assessed. Gene expression changes associated with MV exposure were analyzed using gene expression profiling and real-time qPCR. Within a week of exposure, epithelial MVs stimulated robust mineralization of HGF cultures that was further enhanced by four weeks. The MVs taken up by the HGF's did not calcify themselves but induced intracellular accumulation of minerals. HGF gene expression profiling after short exposure to MVs demonstrated relative dominance of inflammation-related genes that showed increases in gene expression. In later cultures, OSX, BSP and MMPs were significantly upregulated by the MVs. These results suggest for the first time that epithelial cells maybe associated with the ectopic mineralization process often observed in the soft tissues.

www.nature.com/scientificreports/ We have reported previously that MVs released by keratinocytes regulate several genes in dermal fibroblasts via multiple intracellular signaling pathways, leading to increased cell migration and fibroblast-mediated angiogenesis in vitro 23 . In addition to MVs, exosomes released from oral mucosal epithelial cell sheets exhibit proregenerative effects on wound healing 24 . In general, EVs, especially from mesenchymal stem cells (MSCs), have been described to possess wound healing-supporting properties [25][26][27] . We have also reported that oral bacterial biofilms increase MV production in gingival epithelial cells (GECs) 28 . These epithelial MVs increased expression of genes associated with inflammation and matrix degradation in gingival fibroblasts 28 . Moreover, several studies have demonstrated MV-related vascular calcification. Mineralized vesicles were only abundant in the extracellular matrix (ECM) at sites of vascular calcification and not in healthy arteries. Another study reported that vascular smooth muscle cells (VSMCs) released EVs rich in osteogenic differentiation-related biomarkers under inflammatory condition 29 . EVs derived from stressed VSMCs aggregate and form microcalcification zones 30 . We have demonstrated previously that gingival fibroblasts are neural crest-derived cells that retain MSC-like properties and can differentiate to multiple lineages, including undergoing osteogenic differentiation, in vitro [31][32][33] . Since gingiva can develop various calcified lesions, we investigated whether MVs derived from gingival epithelial cells could regulate calcification of gingival fibroblast cultures in an osteogenic environment. Surprisingly, epithelial MVs promoted ectopic intracellular mineralization in gingival fibroblasts that was associated with early expression of inflammation related genes and late gene expression of some classical osteogenic differentiation markers and MMPs.

Results
Gingival epithelial cell-derived MVs induce calcification of human gingival fibroblast cultures. Calcification of HGF cultures after exposure to MVs was investigated by two methods. First, the von Kossa staining was used to identify mineralization nodules after seven and twenty-eight days of exposure. Second, the mineralization of cultures was quantified in real-time using the IncuCyte system with calcein green staining. Using von Kossa staining, first clear signs of appearance of mineralized nodules were observed in 7-days-old cultures exposed to the MVs in the OM, while no nodules were identified in the untreated controls (Fig. 1a). In 28-days-old cultures, epithelial MVs further stimulated the formation of mineralized nodules (Fig. 1a). HGFs in OM also showed some mineralization but significantly less than in cultures exposed to MVs (Fig. 1a). No deposition of mineralized nodules was detected in BM at any time point.
To quantify the mineralization in the MV-exposed cultures, calcein fluorescence was used to record real-time mineral deposition over time for up to 17 days. After 7 days, calcein fluorescence started to increase in MVtreated HGF cultures in the OM medium (Fig. 1b). Calcein fluorescence then linearly increased in the treated cultures to the end of experiment at 17 days (Fig. 1b, c). OM alone induced weak mineralization while cells in the BM did not mineralize at any time point of the experiment (Fig. 1b, c).

Analysis of the mineralized nodules.
After we had established that epithelial MVs induce calcification of HGF cultures, we investigated whether this mineralization occurs in the ECM or intracellularly. In the SEM images, numerous nodules were detected, matching the shapes of the cells in the MV-treated cultures but not in the controls (Fig. 2a). This accumulation mimicked intracellular mineralization rather than extracellular accumulation. To confirm this observation, we used focused ion beam (FIB) to dissect the cultures in apico-coronal direction. The FIB confirmed that the nodules were located in the cytoplasm within vesicular structures (Fig. 2b, c). Using EDS, we determined that the Ca/P ratio of the mineralized nodules was 2.0 ( Fig. 2e). Outside of the nodules, only carbon was detected (Fig. 2d).

Localization of MVs in relation to mineralized nodules in the HGFs.
To find out whether the mineralization is directly initiated by the MV intake, we performed double staining of the MVs and nodules at the early stages of mineralization. After three days of exposure to the red fluorescent MVs, the MVs were clearly detected as dots and diffuse patches on the cell layer ( Fig. 3). At this time point, only a few calcein-fluorescent green dots were detected scattered at the periphery of the cells and not co-localized with MVs (Fig. 3). After a 7-day treatment with MVs, the number of calcium nodules significantly increased around nuclei (Fig. 3). However, no co-location of the MVs and the mineralized nodules were detected (Fig. 3). HGFs without MVs showed no green calcein positive nodules at this time point.

Gene expression profiling.
To investigate changes associated with the mineralization process induced by the epithelial MVs, we performed gene profiling of the early changes in 1-and 3-day-old cultures and analyzed the expression of osteogenic differentiation associated genes in cultures showing MV-induced mineralization on days 7 and 28. Total of 2125 and 1579 genes were differentially expressed in HGFs with fold-change ≥ 1.5 and at p < 0.05 in response to MV stimulation on days 1 and 3, respectively. The top 10 enriched gene ontology annotations of biological process (GOTERM_BP), cellular component (GOTERM_CC) and molecular function (GOTREM_MF) associated with differentially expressed genes (DEGs) are shown in Fig. 4a and b. After 1-day stimulation with the MVs, the categories related to cell proliferation, cytoplasmic components and protein binding were significantly regulated in HGFs (Fig. 4a). After 3-days of stimulation, the categories in treated cells had shifted to ECM components and organization (Fig. 4b). DEGs were further analyzed in groups related to ECM, Inflammatory response (INF) and Ossification (OSS) ( Fig. 5a and b). In the ECM group, 84 and 110 genes were differentially regulated by MVs after one-and three-days stimulation, respectively (50 genes down and 34 up for one day and 85 genes down and 25 up for three days). In the INF group, 65 and 52 genes were regulated by the MVs with upwards trend in the majority of the genes (N = 41 and 28). In the OSS group, 52 and 53 genes were significantly regulated by the MVs with majority showing downregulation (37 and 43 genes) ( Fig. 5a and

Discussion
Pathologic ectopic calcification of soft tissues remains a significant clinical problem. Calcifications occur in blood vessel walls, in many connective tissues and in benign and malignant tumors. The molecular mechanisms of soft tissue calcification are relatively poorly understood. We report in this paper that epithelia-driven mechanisms, including MVs, could be involved in a potential pathway leading to calcification of soft tissues. www.nature.com/scientificreports/ The key observations of our study show that epithelial MVs strongly induced mineralization of HGF cultures, and this was associated with remarkable early upregulation of genes linked to inflammation and MMP activity. Furthermore, some of the genes associated with osteogenic differentiation such as OSX 34 showed up-regulation with advanced mineralization that was largely contained in vesicles in the cytoplasm. Osteogenic differentiation is driven by key transcription factors RUNX-2 and OSX that shift the profile of gene expression to support matrix deposition (type I collagen expression) and mineralization mediated by ALP 35 . HGFs treated with GEC-MVs in osteogenic conditions showed consistent up-regulation of OSX. However, RUNX-2 was not altered at early time point and reduced in later time point which is curious given that OSX acts downstream of RUNX-2 36 . In addition, the expression of ALP and type I collagen were both strongly downregulated by the MVs. Therefore, intracellular mineralization does not seem to follow gene expression profile associated with typical osteogenic differentiation. ALP is present in the MVs (data in file for Bi et al., 2016 28 ) and could be provided for the intracellular mineralization process. EVs, including exosomes and MVs, are highly heterogenous group of vesicles produced by practically all cell types 37 . Before studies on EVs were common, vesicles associated with endochondral bone formation were identified and called matrix vesicles 38,39 . The size of these vesicles range between 50 and 400 nm and could, therefore, represent a mixture of exosomes and MVs 40,41 . MVs produced by osteogenic cells can accumulate Ca 2+ and P intracellularly, and these vesicles serve as sites of mineral nucleation 42 . Epithelial MVs were internalized by the HGFs but did not seem to directly mineralize. We also did not observe any mineralization in HGF cultures that were exposed to epithelial MVs in the basic medium supporting the view that these MVs stimulate mineralization through an indirect process that is likely associated with their cargo and a suitable environment. We have previously analyzed the protein content of MVs derived from the same gingival epithelial cell line used in the present study 28 . There are over 2000 proteins present in the epithelial MVs, and it is likely that, in osteogenic environment, some could be recycled to form new vesicles that have Ca 2+ -binding capacity. For example, epithelial MVs contain several annexins that can bind both Ca 2+ and phospholipids and can potentially act as nucleation sites for mineralization when released into the target cells. A previous study has demonstrated that EVs derived from mineralizing osteoblasts contain more annexins than from non-mineralizing osteoblasts 42 . The reasons why the newly formed vesicular structures that were mineralized with material having a Ca 2+/ P ratio close to hydroxyapatite were not released from the cells remains to be further investigated. Biogenesis and release of EVs is a complex process and involves often multiple steps and sorting machineries 37 that may be deficient in HGFs, leading to trapping of the vesicles in the intracellular milieu. Mitochondrial participation in mineralization has been proposed in several studies [43][44][45] . We have not confirmed mitochondrial role for epithelial-driven and MVassociated intracellular mineralization of HGFs, and it warrants further studies. However, intracellular accumulation of mineral-containing vesicles has been observed in several other studies using different cell lines [46][47][48] .
Epithelial MVs caused a strong and early response of the expression of genes related to inflammation and matrix metalloproteinase activity. These findings support our previous findings that showed MV-induced gene expression of IL-6, IL-8, MMP-1 and MMP-3 in HGFs 28 . MVs from epidermal keratinocytes also stimulate the  d), (e), The energy-dispersive X-ray spectroscopy (EDS) analysis. EDS analysis showed that both calcium and phosphorus were present in the highdensity nodules in the cell interior (e) while control areas contained carbon only (d). The Ca/P ratio of these nodules was determined to be 2:1. www.nature.com/scientificreports/ same genes in dermal fibroblasts, supporting the notion that this signaling machinery stands as universal communication mechanism between epithelial cells and fibroblasts 23 . Pro-inflammatory signals are likely to initiate normal bone healing and promote osteogenesis, but chronic inflammation is detrimental for bone formation [49][50][51][52] .
Inflammation is also linked to extraosseous mineralization in pathological conditions, such as atherosclerosis, bone metastasis and others 53,54 . Serum amyloid A proteins 1 and 2 (SAA1/2) are released by liver in response to inflammation but also produced locally to regulate inflammatory cytokine release through activation of tolllike receptors TLR2 and TLR4 55,56 . In mesenchymal stem cells, both autocrine and paracrine SAA1/2 stimulate inflammatory cytokine expression and enhanced mineralization 57 . In addition, the expression of inflammatory cytokines, such as IL-6 and IL-8, is highly elevated during osteogenic differentiation of bone marrow mesenchymal stem cells 58 . Interestingly, expression of both SAA1 and 2 were significantly increased in HGFs by the MVs and could, therefore, mediate strong up-regulation of cytokines such as Il-6, IL-8 and CXCL1. Another    www.nature.com/scientificreports/ inflammation-related gene that was upregulated by MVs was PTX3 (long pentraxin 3). Intriguingly, PTX3 appears to play a role in regulation of both normal and ectopic mineralization by promoting cell differentiation and mineral crystal formation 59 . Particularly, calcifications in normal and cancerous tissue in the breast, in prostate cancer and in atheromas are associated with high PTX3 expression [60][61][62] . As expected, TNF and NF-B signaling pathways were increased in MV-treated HGFs. Interestingly, a low dose of TNF stimulation has been reported to induce persistent Wnt expression through NF-κB and JNK signaling pathways, resulting in bone formation 63 . Our previous studies have shown that JNK signaling pathway is activated by epithelial MVs in HGFs and dermal fibroblasts 23,28 . Therefore, NF-κB and JNK signaling could participate in MV-induced mineralization of HGF cultures.
In the present study, MMP1 and MMP3 were strongly upregulated by epithelial MVs, corroborating our previous observations with dermal and gingival fibroblasts 23,28 . Epithelial MVs regulate MMPs mainly through ERK1/2 signaling. Increased expression of MMPs, including MMP-1 has been also found in bone marrow MSCs during osteogenic differentiation 64 . Other MMPs, such as MMP-2, MMP-9 and MT1-MMP, are also stimulated during early mineralization 64,65 . MMP-1 may directly promote osteogenic differentiation through ERK and JNK pathways 66 . MMP-1 stimulation of bone marrow MSCs upregulated osteogenic differentiation markers (RUNX2, OSX, OPN and OCN) and promoted mineralization 66 . Whether other MMPs could also directly stimulate osteogenic differentiation remains unknown. Overall, a number of MMPs play a crucial role in bone development and pathology but their role is multifaceted 67 .
Possible limitations of the present study include the use of MVs from spontaneously immortalized human gingival epithelial cells. It is well known that the cargo in the EVs can vary depending on cell types and pathological Table 1. Top 10 upregulated genes of differentially expressed genes (DEGs) on HGFs with epithelial MV stimulation related to (a) Extracellular matrix (ECM), (b) Inflammatory response (INF), and (c) Ossification (OSS) after 1-day culture. In ECM group, MMP1 and MMP3 were upregulated with MV stimulation. Genes related inflammation (IL-8, CXCL1 and C3) were highly upregulated. Also, SAA1 and SAA2 were upregulated. Among OSS, TNC, FOXC1 and MSX2 were upregulated. www.nature.com/scientificreports/ conditions 46 . However, epithelial MVs from spontaneously immortalized gingival and epidermal keratinocytes as well as from primary keratinocytes similarly induce MMP and cytokine expression on fibroblasts from different origins 23,28 , suggesting that at least some of the cargo proteins are shared. In addition, we were not able to determine the molecules involved in the induction of the mineralization and why the minerals were retained in the cytoplasm. These interesting questions will need to be answered in a separate future study.
In conclusion, the present study shows that gingival epithelial-derived MVs induce rapid mineralization of gingival fibroblast cultures in osteogenic conditions. This process is associated with early upregulation of inflammation and MMP activity-related genes that could positively regulate the osteogenic differentiation. Remarkably, that the mineralization remained largely intracellular, warrants further investigation.

Methods
Cell culture. Spontaneously immortalized human gingival epithelial cell (GEC) line was established in the authors' laboratory 68 . This cell line has partial triploid phenotype but no known mutations. Normal primary human gingival fibroblast strain (GFBL-HN; referred here as HGFs) was isolated from healthy attached gingiva of a healthy 18-year-old female 69 . This cell strain possesses an average phenotype of several primary gingival fibroblast strains tested in our previous study 69 . Both cell types were maintained in basic medium (BM) containing Dulbecco's modified Eagle's medium (Gibco, Life Technologies, Grand Island, NY, USA) supplemented with MV collection. MV collection was performed as previously described 28 . Briefly, confluent GECs were rinsed with phosphate-buffered saline (PBS) and cultured in FBS-free medium for 48 h. Conditioned medium from the cell cultures was collected and centrifuged (Sorvall RC 5B Plus, Mandel Scientific; manufactured by DuPont, Newtown, CT, USA) at 4℃, first at 3,000 g for 15 min to remove cellular debris and then at 25,000 g for 30 min to collect MVs. MV pellets were rinsed once with PBS, homogeneously re-suspended in FBS-free medium and kept at 4℃ until used. Total MV protein content was measured with Bio-Rad Protein Assay reagent (Bio-Rad Laboratories, Hercules, CA, USA) and spectrophotometry at 570 nm and used for standardizing the vesicle amounts in the experiments. Characterization of MVs (physical size, proteomics analysis) produced by the GEC line used has been published in a previous study 28 .
HGF stimulation by epithelial MVs in osteogenic conditions. HGFs were seeded in 24-well plates

Staining of mineralized nodules by von Kossa.
Von Kossa staining was performed as previously reported 70 . In brief, the cells were fixed in 4% formaldehyde (Fisher Scientific, Fair Lawn, NJ, USA) after 7-and 28-days culture periods and incubated with 2% silver nitrate (Fisher Scientific) in dark for 10 min and then exposed to bright light for 15 min. The plates were then washed with distilled water and dehydrated in 100% ethanol. The samples were then examined by light microscopy 71 .

Calcification of cultures measured by IncuCyte real-time imaging.
For real-time imaging of calcification, 10 mM calcein green (Sigma-Aldrich) solution, which fluoresces when bound to calcium crystals, was prepared in 0.1 M NaOH, and the solution was further diluted to 1 mM and sterile-filtered. Calcein solution was Table 3. Top 10 of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of differentially expressed genes (DEGs) in MV stimulated HGFs compared to control after 1 day (a) and 3 days (b) using The Database for Annotation, Visualization, and Integrated Discovery (DAVID). At 1 day, pathways of cell cycle were highly upgraded. TNF signaling pathway was up regulated both after 1-day and 3-days stimulation. NF-κB signaling pathway was upregulated after 3 days of culture.   Quantitative reverse transcription PCR (RT-qPCR). RT-qPCR was performed as previously described 75 . Briefly, total RNA was obtained as above. Total RNA (1 μg) was reverse-transcribed with high-capacity cDNA reverse transcription kit (Applied Biosystems, Life Technologies, Grand Island, NY, USA), according to the manufacturer's instructions, and Mastercycler Gradient 5331 Reverse-Transcriptase PCR Instrument (Eppendorf AG, Hamburg, Germany). The cDNA was diluted to a concentration at which the threshold-cycle value was well within the range of its standard curve. cDNA (5 μl) was mixed with 10 μl of 2 × iQ SYBR Green I Supermix (Bio-Rad) and 5 pmol of primers in 96-well plate wells. Asparagine-linked glycosylation 9 (ALG9) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were used as reference genes. Real-time qPCR amplification was performed using the CFX96 system (Bio-Rad Laboratories). The data were analyzed and are presented according to the comparative Ct method (CFX Manager Software, version 2.1; Bio-Rad Laboratories). PCR primers used are listed in the supplementary Table S1.
Statistical analysis. All data are expressed as mean ± standard error of the mean (SEM) from at least three independent experiments. Statistical analysis was performed using SPSS 24.0 software. Statistical differences were evaluated by Student's t-test for paired comparisons or by one-way ANOVA followed by Tukey's post hoc test for multiple comparisons. Statistical analysis for RT-qPCR data was done using log2-transformed data. P values < 0.05 were considered as statistically significant.