Human mandibular growth is mainly through endochondral bone formation at the condyles and intramembranous bone formation at the periosteum (Enlow and Hans,1996). While condylar growth provides the overall elongation and rotation of the mandible (Bjork and Skieller,1983), periosteal activity causes extensive surface modeling/remodeling or reshaping of the mandible (Enlow and Hans,1996). One striking feature of mandibular periosteal activity is at the ramus, where bone resorption takes place at the anterior (rostral) border whereas bone apposition at the posterior (caudal) border (Robinson and Sarnat,1955; Seiton and Engel,1969; Hans et al.,1995). At the cell level, osteogenic proliferation is concentrated at the appositional posterior ramal region, whereas osteoclasts are mainly present at the resorptive anterior region (Ochareon and Herring,2007). However, the molecular mechanisms related to these differential periosteal activities are mostly unknown.
Recruitment and maturation of osteoclasts for bone resorption require two important molecules, colony-stimulating factor (Felix et al.,1990; Cecchini et al.,1997) and receptor activator of NF-κB ligand (RANKL) (Hsu et al.,1999; Wada et al.,2006), both of which are expressed by osteoblasts. The binding of RANKL to the RANK receptor on osteoclasts activates osteoclast differentiation and bone resorption. On the other hand, osteoblasts also produce osteoprotegerin (OPG), a decoy receptor for RANKL, to inhibit RANK activation by RANKL. The ratio of RANKL to OPG expression determines the extent of bone resorption activity. Accordingly, the differential periosteal activity between the anterior and posterior ramal borders may result from the variations in the relative RANKL/OPG expression.
Clarifying the molecular mechanisms related to the regional variation of the mandibular periosteal activity is important for a better understanding of mandibular growth, as well as for using periosteal cells in tissue engineering applications. Mandibular periosteal cells have been recently used to augment the maxillary sinus floor before dental implant placement (Schmelzeisen et al.,2003) and to enhance bone repair at calvarial defects (Miyamoto et al.,2004; Sakata et al.,2006). In these studies, periosteal cells were collected either from the lateral ramal surface (Schmelzeisen et al.,2003) or from the anterior ramal surface (Miyamoto et al.,2004; Sakata et al.,2006). Access to the lateral ramal surface requires a surgical procedure more invasive than access to the anterior ramal surface, which can be easily obtained during a routine wisdom tooth extraction. Despite easy access, given that the anterior periosteal surface is dominated by bone resorption (Enlow and Hans,1996), the anterior mandibular ramal surface may not provide an optimal periosteal cell source for bone engineering unless their molecular bone resorptive profile is reversed by in vitro treatments.
One treatment that can potentially change the phenotype of ramal periosteal cells in vitro is to apply mechanical stretching (tensile strain). During normal mandibular growth, both the fibrous (Grimm and Katele,1979) and the osteogenic (Ochareon and Herring,2007) layers of the periosteum are displaced toward the appositional posterior border. Such displacement is thought to be caused by mechanical stretching generated during bone elongation (Frankenhuis-van den Heuvel et al.,1992). In addition to physical displacement, mechanical stretching may also change the phenotype of periosteal cells, as abundant evidence has shown that mechanical stretching activates osteoblasts in a number of signaling pathways (Wang et al.,2001; Armstrong et al.,2007; Sunters et al.,2010) and promotes osteogenic differentiation of cultured mesenchymal stem cells (Ward et al.,2007; Huang et al.,2009). In this study, we sought to characterize the molecular basis of mandibular ramal periosteum in relation to the juxtaposing bone apposition/resorption activities and examine the effect of mechanical stretching on ramal periosteal cells during in vitro expansion.
MATERIALS AND METHODS
Young domestic pigs (3–4 months old; Sus scrofa) were used in this study. The pig has a masticatory apparatus comparable to the human (Herring,1976; Ström et al.,1986; Wang et al.,2007) and has been used before in mandibular growth and repair studies (Ochareon and Herring,2007; Sun et al.,2007). All procedures involving live animals were approved by the Ohio State University Institutional Animal Care and Use Committee.
Fluorescent Vital Staining and Assessment of Bone Formation
Fluorescent staining was used as described earlier (Sun et al.,2004). Briefly, the pigs were anesthetized with isofluorane, and the fluorescent dyes calcein and alizarin complexone (Sigma-Aldrich, St. Louis, MO) were injected intravenously (12.5 mg/kg) on 10 and 3 days prior to sacrifice, respectively. Following sacrifice, two sets of specimens containing the anterior and posterior ramal regions were harvested. One set of the specimens were fixed in Prefer fixative for 3 days (Anatech, Battle Creek, MI), dehydrated in an ethanol series, embedded with MicroBed resin solution (Electron Microscopy Sciences, Fort Washington, PA), and cut into 50-μm thick sections using a diamond circular saw (Leica SP 1600; Germany). The other set of specimens were decalcified with 15% formic acid after Prefer fixation, dehydrated, paraffin embedded, cut into 5-μm sections and stained with H&E. Undecalcified sections were used to evaluate mineralization activities using an Axioplane 400 Zeiss fluorescent microscope with Axioplan software and cameras (Carl Zeiss MicroImaging, NY). The interlabel width was measured as an indicator of new bone formation. Decalcified sections were used to evaluate histology of the periosteum and bone at the two regions.
In Vivo Tissue Harvesting and Processing
Immediately after sacrifice, pig ramal periosteum was harvested from the anterior and posterior regions (∼100 mg per region) and instantly frozen at −80°C. A total of four pairs (anterior and posterior from the same ramus bone) of samples were obtained. The tissues were then placed in cryogenic vials containing a stainless steel grinding ball, immersed in liquid nitrogen and pulverized using Mikro-Dismembrator S (Sartorius Stedim Biotech, Bohemia, NY) at 2500 rpm for 30 sec, for a total of four times. After each grinding, the shaking flask was immersed in liquid nitrogen for a few minutes to maintain tissue brittleness. Pulverized tissues were homogenized in 1 mL of Trizol Reagent (Invitrogen, Carlsbad, CA), followed by RNA and protein extraction according to manufacturer's recommended protocol.
Periosteal Cell Culture
Periosteal tissues harvested from the anterior and posterior ramal surfaces were immediately immersed in ice-cold Hank's Balanced Salt Solution (HBSS), 1× (Invitrogen, CA), minced, and transferred onto a 105-μm Spectra/Mesh Polypropylene filter (Spectrum Laboratories, CA) in the digestion chamber. HBSS was replaced by 15 mL of prewarmed TryplE Express (Invitrogen, CA) and gently stirred at 37°C for 10 minutes for enzymatic dissociation of superficial cells. Subsequently, TryplE Express was replaced by 15 mL of prewarmed 0.2% Collagenase I and incubated for 3 hr under constant stirring. Collagenase filtrate was transferred into 3 × 15 mL conical tubes. The polypropylene filter was washed with 25 mL of Ham's F12 (Mediatech, Manassas, VA), and the filtrate was added to the tubes containing collagenase filtrate. The mixtures of filtrates were spun at 1,100 rpm for 10 minutes, and the cell pellet was resuspended in DMEM/F-12 media containing Penicillin (100 U/mL), Streptomycin (100 μg/mL), 2 mM L-glutamine, and 10% fetal bovine serum (Invitrogen, CA). Cells were pooled, washed again, and grown in a 25 cm2 flask. Five days after the cells were seeded, the media were changed and supplemented with 50 μg/mL ascorbic acid (Sigma-Aldrich, MO) and 10 mM β-glycerolphosphate (MP Biomedicals, OH) for osteogenic induction. The media were changed every 3–4 days, and cells were passaged when reaching 80% confluence. Cells from passages three to five were used.
Staining of Alkaline Phosphatase Activity
Osteogenic differentiation in primary osteoblasts from each passage was assessed by the presence of alkaline phosphatase (AP) activity using AP staining kit (Sigma-Aldrich, MO), according to the manufacturer's recommended protocol. Negative controls were fixed and stained with hematoxylin for 10 minutes.
Application of Cyclic Tensile Strain
Cells were cultured to 80% confluence on six-well Bioflex culture plates coated with collagen-type I (Flexcell International, Hillsborough, NC). Subsequently, cells were serum starved overnight in culturing media containing 1% FBS and subjected to dynamic tensile strain at a magnitude of 0 (control), 5 or 10% at 0.5 Hz for 2 hr using a FX-4000 Tension Plus system (Flexcell International, NC). One hour after the cessation of loading, total RNA was extracted using Qiagen RNeasy Mini Kit (Valencia, CA) and its concentration was determined using Nanodrop 1000 (Thermo Scientific, DE). In addition, 18 hr after loading, total cellular proteins were extracted in RIPA buffer containing protease inhibitors (Santa Cruz Biotech, CA), and the concentrations were measured using BCA Kit (Pierce, Rockford, IL).
Reverse Transcription and Quantitative Real-Time PCR Analysis
One microgram of total RNA was reverse transcribed into cDNA using a Superscript III Reverse Transcriptase kit according to manufacturer's protocol (Invitrogen, CA). Primer sets for each target genes were designed and validated by electrophoresis of PCR products. Primer sets for each of the target genes and a housekeeping gene β-actin are shown in Table 1. Real-time PCR was performed to quantify the expression of target genes with iQ SYBR Green Supermix using the iCyler iQ detection system (Bio-Rad, Hercules, CA). The expression levels of target genes were computed by using the comparative threshold (CT) cycle method (Giulietti et al.,2001; Livak and Schmittgen,2001). Briefly, all raw CT values were normalized to β-actin from the same sample to obtain ΔCT value. The average ΔCT values of the anterior periosteal tissue or unloaded anterior periosteal cells were calculated and used as the calibrators. The amount of target gene expression was determined by a formula 2−ΔΔCT, where −ΔΔCT = ΔCT (individual sample)−ΔCT(calibrator).
Table 1. Primer sequences for real-time RT-PCR
Western Blot Analysis
Fifty micrograms of protein samples were mixed with 4× SDS gel loading buffer, boiled for 5 minutes, and eletrophoretically separated on a 10% SDS polyacrylamide gel. Proteins were then transferred to a nitrocellulose membrane. The membrane was blocked with 5% nonfat milk for 1 hr and washed with PBS containing 0.5% Tween-20 (Bio-Rad, CA). Subsequently, the membrane was probed overnight at 4°C with 1 μg/mL goat polyclonal antibody against RANKL (sc-7628; Santa Cruz Biotechnology, CA) and rabbit polyclonal antibody against OPG (sv-11383; Santa Cruz) in PBS with 5% bovine serum albumin. To normalize total protein, the membranes were reprobed with monoclonal anti-β-actin (A1978; Sigma-Aldrich, MO). The membranes were then incubated with 1 μg/mL IRdye-conjugated secondary antibodies (LI-COR, Lincoln, NE) for 1 hr and the band intensity quantified using Odyssey Infrared Imaging System (LI-COR, NE).
PASW Statistics 18 software (SPSS, IL) was used for data analysis. For periosteal tissue, four paired samples from the anterior and posterior periosteum were assayed. The difference of target gene expression (fold of difference calibrated to the mean ΔCT of anterior tissue) was assessed by paired student t tests. All in vitro assays were performed in triplicates from different passages. The variation (fold of difference relative to the mean ΔCT of anterior cells without loading) caused by the two cofactors, region (anterior vs. posterior) and loading (0, 5, and 10%), was analyzed by two-way ANOVA.
Ramal Periosteal Growth
Periosteal growth at the anterior and posterior ramal surfaces labeled by fluorescent dyes is shown in Fig. 1A,B. The posterior surface was characterized by fast primary bone formation in a sinusoidal pattern. Approximately 1.2 mm of new bone was formed during a 1-week interval of dye injections, and approximately another 0.5 mm of new bone formation between the alizarin injection and sacrifice (3 days) as indicated by the distance between alizarin and bone front. Contrarily, the anterior region consisted of dense cortical bone with minimal mineralization activity at the periosteal surface and islands of bone that may have been separated from the main cortex by bone resorption.
Histology of the periosteum and bone at the two regions is shown in Fig. 1C–F. The posterior periosteum had a thick osteogenic layer with abundant osteoblast-like cells present, while the anterior periosteum was characterized by areas of bone resorption around bone islands with multinucleated osteoclast-like cells.
AP Activity in Periosteal Cells
Osteoblasts contain high levels of AP activity required for the mineralization. Therefore, we examined the presence of AP activity in periosteal cells as a marker for their osteogenic phenotype. Cultured periosteal cells from the anterior and posterior regions were stained for the presence of AP activity. Cells from both anterior and posterior regions exhibited the presence of AP activity. A comparison of cells grown from the two regions showed that cells from the posterior ramus tended to have stronger AP activity than cells from the anterior ramus (Fig. 1G and H).
Regulation of Gene Expression (mRNA)
The expression of runt-related transcription factor 2 (Runx2), osteocalcin (OCN), bone morphogenetic protein 2 (BMP2), RANKL, and OPG is related to the osteoblastic/osteoclastic activity during bone apposition and resorption. Therefore, we examined the mRNA expression levels of these genes by real-time RT-PCR. These primers were tested for their specificity by electrophoresis of PCR products with the exhibition of single bands of the expected base-pair size (Fig. 2A).
Periosteal tissue gene expression at the mRNA level is shown in Fig. 2B. Anterior periosteal tissue had significantly stronger expression of BMP2 (by ∼2-fold) but significantly less expression of OPG (by ∼3-fold) than posterior periosteal tissue (paired student t tests, P < 0.05). The RANKL/OPG ratio also tended to be greater at the anterior periosteal tissue (P = 0.058), whereas the expression of RANKL, Runx2, and OCN was statistically comparable between the regions. A tendency of stronger OCN expression was shown in anterior periosteal tissue. This tendency was mainly caused by one outlier sample, which also increased the variability. When that outlier sample was excluded from the analysis, the two regions were identical in OCN expression (not shown).
The mRNA expression in cultured periosteal cells with and without loading exhibited relatively high variability among some samples (from different passages; Fig. 3). However, despite the variability, a statistically stronger expression of BMP2 and higher RANKL/OPG ratio were present in anterior cells than in posterior cells (two-way ANOVA, P < 0.05) with or without mechanical loading. No statistical significances were found for the regional variations of Runx2, OCN, OPG, and RANKL, although the posterior cells tended to have stronger expression in OPG. Upon tensile strain loading, anterior cells showed a tendency (two-way ANOVA, P > 0.05) of magnitude-dependent upregulation in Runx2, OCN, and RANKL, whereas posterior cells exhibited a tendency of downregulation in Runx2, OCN, and OPG expression. RANKL/OPG ratio also tended to increase in anterior cells but remained constant in posterior cells, resulting in an even greater difference between these two regions than control cells (without loading). Compared with these molecules, BMP2 was minimally changed by the strain.
OPG and RANKL protein expression was examined by Western blot analysis. Most in vivo periosteal tissues produced relatively small amounts of protein samples, so these experiments were mainly performed on cultured cell samples. The relative migration of proteins on the gels and their staining by specific antibodies showed that positive OPG and RANKL bands were approximately at 60 kD and 50 kD, respectively (Fig. 4A). After β-actin normalization, both OPG and RANKL expression tended to be stronger in anterior cells than posterior cells, but the differences were not statistically significant. With the application of tensile strain, only minimal changes in OPG and RANKL protein expression was observed (Fig. 4A and B). RANKL/OPG ratio was also similar between the two regions with or without loading (not shown).
Similar to humans, the pig mandibular ramus grows in an anterior-resorption and posterior-apposition pattern (Fig. 1A–F), confirming that the pig is an acceptable model to study the unique ramal periosteal growth in the human (Robinson and Sarnat,1955; Ochareon and Herring,2007). Therefore, the findings obtained from this study about pig molecular changes at the ramal periosteum provide a new understanding of periosteal growth in the human mandible, an area important for clinical orthodontics and craniofacial orthopedics.
In this study, the expression of several molecules involved in bone resorption/apposition regulation was quantified from harvested periosteal tissues and cultured periosteal cells. When compared with direct tissue processing, in vitro expansion and passaging, especially with the use of serum and supplements for osteogenic induction, can potentially impact gene expression (Tamama et al.,2010). Relatively, large intersample variability was observed in the expression of some molecules (Fig. 3). The samples used in this study were cells from different passages (passage three to five), rather than triplicates from the same passage. Very likely, this variability is at least in part contributed by changes caused by passaging. On the other hand, despite such intersample variability, we found that periosteal tissues and cultured cells consistently showed significantly greater expression of BMP2 and RANKL/OPG ratio at the anterior region than at the posterior region (Figs. 2, 3), suggesting that these genes may be involved in regulating differential periosteal activities at these regions.
Increased RANKL/OPG ratio accompanies bone resorption involved in a variety of processes such as pathological osteolysis (Grimaud et al.,2003; Tay et al.,2004), periodontitis (Crotti et al.,2003), tooth eruption (Heinrich et al.,2005), and orthodontic tooth movement (Yamaguchi,2009). The current data adds physiologic mandibular periosteal growth to that list. During normal growth, infection or inflammation is absent at the anterior ramal surface. Therefore, an increased RANKL/OPG ratio is not caused by infection or inflammation. Periosteal tissue from the posterior region had significantly greater expression of OPG than that from the anterior region, while their expression of RANKL was similar (Fig. 2). Cultured posterior periosteal cells without mechanical stretching also demonstrated a tendency of greater OPG expression than anterior cells, whereas their expression of RANKL was comparable (Fig. 3). These data suggest a likelihood that the switch for periosteal resorption/apposition control at the mandibular periosteum is mainly through OPG rather than through RANKL. More specifically, a default amount of RANKL may be present at both resorptive and appositional regions, while the level of OPG expression induction determines which activity dominates. The current data is inadequate to depict a detailed molecular control of mandibular periosteal resorption/apposition, but following this line, future studies can focus on factors that regulate OPG expression in the mandibular periosteum.
As an established regulators for skeletal development and bone formation (Cao and Chen,2005), BMP2 was expected to express more strongly at an appositional region than a resorptive region. We found the opposite in this study, which suggests that BMP2 may also be involved in regulating periosteal bone resorption. In recent years, studies have in fact found that BMP2 can enhance osteoclastogenesis indirectly by stimulating osteoblasts or stromal cells to produce osteoclast-promoting factors, such as RANKL (Abe et al.,2000; Otsuka et al.,2003), or directly by activating osteoclast differentiation in the presence of RANKL (Jensen et al.,2009). It is hence likely that stronger expression of BMP2 at the anterior periosteal region plays an assisting role in promoting osteoclastogenesis by synergizing with RANKL.
Western blots in this study detected bands for OPG and RANKL at ∼60 kD and 50 kD, respectively, from positive controls (Santa Cruz Biotech, CA) and our samples. A molecular weight of ∼50 kD was unexpectedly higher than the predicted molecular weight for transmembrane RANKL (Wong et al.,1997), but similar findings have been reported before. Using the same antibody (Santa Cruz Biotech, CA), Dossing and Stern (2005) reported a band of ∼52 kD for RANKL, and the increased molecular weight has been attributed to RANKL glycosylation (Wong et al.,1997; Lum et al.,1999). These reports confirm that our identification of a ∼50-kD RANKL band by Western blot was specific and reliable.
Unlike mRNA expression, protein expression of OPG and RANKL examined by Western blot was not distinctly different between the two regions. While this apparent discrepancy may indicate differential transcriptional vs. translational regulations of OPG and RANKL expression at the ramal periosteum, it is more likely caused by an incomplete measurement of protein expression by Western blot used in this study. Because only proteins extracted from cell lysates were assayed, the expression of total OPG, a secreted glycoprotein, was only partially represented by the current Western blot data. The distribution of secreted/unsecreted OPG protein in periosteal cells is currently unknown, but it has been reported that normal OPG can be quickly secreted in a number of cell lines (Middleton-Hardie et al.,2006). Similarly, the current Western blotting data probably only reflected the expression level of transmembrane RANKL but not soluble RANKL. Because of these reasons, the current protein expression data was not conclusive. A future analysis of protein expression in the culture media and in in vivo tissues will clarify whether the regional variation of RANKL/OPG mRNA expression at the ramal periosteum also exists at the protein level.
The periosteum from long bones contains mesenchymal progenitor cells in the osteogenic layer (De Bari et al.,2006). Upon osteogenic induction, these cells can differentiate into osteoblasts through multiple stages. The mandibular ramal periosteal cells cultured in this study also contained such cells as exhibited by positive AP activity and strong expression of osteoblast differentiation markers (Figs. 1–3). Because of this feature, periosteal cells have been used for tissue engineering in the craniofacial region (Schmelzeisen et al.,2003; Miyamoto et al.,2004; Sakata et al.,2006). The anterior ramal surface is easy to access and has been used as a location for periosteal cell harvesting (Miyamoto et al.,2004; Sakata et al.,2006). Whether this resorptive surface is a good location to harvest periosteal cells for tissue engineer needs to be clarified. In this study, we found that during in vitro expansion, cells from the anterior surface did not differ from cells harvested from the appositional posterior surface in expressing osteogenic markers (Runx2 and OCN). However, the expression profile for RANKL/OPG ratio and BMP2 were significantly different between the two surfaces in the same way as in vivo tissues. These data suggest that cells harvested from a resorptive periosteal surface can remain osteoclastogenesis-promoting even after in vitro expansion with osteogenic media. This feature may predispose more bone resorption after transplanting them to a receiving site than using cells harvested from an appositional surface. If this speculation was further confirmed by in vivo studies, it might be a disadvantage to use periosteal cells from a resorptive surface for tissue engineering unless in vitro expansion can reverse their resorptive molecular traits.
In this study, we applied a single-dose cyclic tensile strain to cultured periosteal cells to examine whether such mechanical stimulation can change the molecular profile of periosteal cells, especially those from the anterior resorptive region. Our data showed that this dynamic tensile strain did have a tendency to stimulate osteogenic differentiation (Runx2 and OCN) of anterior cells. However, the RANKL/OPG expression ratio in cultured anterior cells also had a tendency to be increased with the application of cyclic strain, suggesting that this mechanical treatment is not effective to reverse the osteoclastogenesis-promoting feature carried by the anterior cells. While many reasons can be speculated for this result, the difference between physiological strain that the periosteal cells receive and the strain used in this study has to be considered. Physiologically, the tensile strain that displaces the mandibular periosteum, which probably also stimulates periosteal cells to be more osteogenesis-promoting, was thought to derive mainly from bone elongation (Frankenhuis-van den Heuvel et al.,1992). The exact nature of this bone-elongation strain has not been clarified, but it is probably less dynamic and more persistent than the strain used in this study. A future study using a more continuous strain with longer durations could clarify this point.
In conclusion, this study found that the unique anterior-resorption and posterior-apposition periosteal growth pattern of the mandibular ramus is likely controlled by a variation in RANKL/OPG ratio (mainly through OPG) and BMP2 expression between these two regions.
We are very grateful to Dr. Sudha Agarwal for her great advice on this project and her thoughtful discussion of the manuscript. We also thank Sarah Hueni, Priyangi Perera, and Dr. Jin Nam for providing help for the experiments.