Wnt signaling and osteoblastogenesis
- 1.3k Downloads
Wnts are a large family of growth factors that mediate fundamental biological processes like embryogenesis, organogenesis and tumorigenesis. These proteins bind to a membrane receptor complex comprised of a frizzled (FZD) G-protein-coupled receptor (GPCRs) and a low-density lipoprotein (LDL) receptor-related protein (LRP). The formation of this ligand-receptor complex initiates a number of intracellular signaling cascades that includes the canonical/β-catenin pathway, as well as several GPCR-mediated noncanonical pathways. In recent years, canonical Wnt signaling has been shown to play a substantial role in the control of bone formation. Clinical investigations have found that mutations in LRP-5 are associated with bone mineral density and fractures. For example, loss-of-function mutations in LRP-5 cause osteoporosis pseudoglioma syndrome, while gain-of-function mutations lead to high bone mass phenotypes. Studies of knockout and transgenic mouse models for Wnt pathway components like Wnt-10b, LRP-5/6, secreted frizzled-related protein-1, dickkopf-2, Axin-2 and β-catenin have demonstrated that canonical signaling modulates most aspects of osteoblast physiology including proliferation, differentiation, bone matrix formation/mineralization and apoptosis as well as coupling to osteoclastogenesis and bone resorption. Future studies in this rapidly growing area of research should focus on elucidating Wnt/FZD specificity in the control of bone cell function, the role of noncanonical pathways in skeletal remodeling, and direct effects of Wnts on cells of the osteoclast lineage.
KeywordsLow-density lipoprotein receptor-related protein High bone mass Dickkopf Secreted frizzled-related protein Axin-2 β-catenin Mesenchymal stem cell Osteoblast Proliferation Differentiation Mineralization Apoptosis
1 Introduction to Wnt signaling
Wnts are a family of 19 secreted proteins that mediate important biological processes like embryogenesis and organogenesis [1, 2, 3, 4]. Wnts bind to a membrane receptor complex composed of one-of-ten Frizzled (FZD) G-protein coupled receptors (GPCRs) and one-of-two low-density lipoprotein (LDL) receptor-related proteins (LRPs) [2, 3, 4]. This binding activates one of four known intracellular signaling pathways based upon the Wnt, FZD, LRP and cell-type involved [1, 2, 3, 4]. The best characterized of these pathways is the canonical or Wnt/β-catenin pathway that signals through LRP-5 or LRP-6 and leads to inhibition of glycogen synthase kinase (GSK)-3β and subsequent stabilization of β-catenin. β-catenin then translocates to the nucleus, where it binds to and activates lymphoid-enhancer binding factor (LEF)/T cell-specific transcription factors (TCFs) [1, 2, 3, 4, 5]. Wnts also activate additional noncanonical pathways that include the G-protein mediated Wnt/calcium  and Wnt/cAMP [7, 8] pathways, as well as the disheveled (Dsh)-mediated c-Jun NH2-terminal kinase (JNK) pathway . Although the biological consequences of canonical Wnt signaling are well established, the significance of the noncanonical pathways is less well understood. As a result of the physiological importance of Wnt signaling, there are many extracellular and intracellular proteins that modulate these pathways [1, 2, 3]. The extracellular regulators include secreted proteins like Wnt inhibitory factors (WIFs), secreted frizzled-related proteins (sFRPs) and dickkopfs (Dkks) [10, 11, 12] as well as SOST/sclerostin, Wise and connective tissue growth factor (CTGF) [13, 14]. These proteins bind Wnts (WIFs and sFRPs), bind to FZD receptors (sFRPs) or interact with LRPs (Dkks, SOST/sclerostin and CTGF). Proteins like sFRPs and WIFs have the ability to inhibit all Wnt-activated pathways , while Dkks only suppress canonical signaling .
2 Roles of LRP-5 and LRP-6 in osteoblast proliferation, function and apoptosis
Osteoblasts are bone-forming cells that synthesize and mineralize the skeleton [15, 16] (see article by Lian in this issue). These cells develop from bone marrow-derived multipotent mesenchymal stem cells (MSCs) of the colony forming unit-fibroblast (CFU-F) lineage that also give rise to fibroblasts, myoblasts, adipocytes, and chondrocytes [15, 16]. Many mammalian MSC and osteoblast models have been shown to express Wnts and other components of the pathways, indicating that these cells have the machinery to generate and respond to both canonical and noncanonical signals [13, 17, 18].
Deletion of LRP-6 also causes decreased TBV. Kharode et al.  reported that micro-computed tomography (CT) analysis of femurs from 26-week-old LRP-6+/− females demonstrated that partial loss of LRP-6 decreased TBV by 55%. Reduced volumetric BMD (vBMD) as determined by peripheral quantitative computed tomography (pQCT) could be observed in the LRP-6+/− mice as early as 9 weeks of age. These preliminary results have been confirmed by Holmen et al.  using LRP-5 and LRP-6 double knockout mice. In this study, both distal femur TBV and mid-femur cortical thickness were reduced by deletion of LRP-5 and LRP-6 as determined by micro-CT analysis of 3-month-old females. In addition, the effects appeared to be dose-dependent, such that deletion of two or three alleles produced a larger decrease in bone accrual than loss of a single allele. Collectively, these observations indicate that all four alleles of LRP-5 and LRP-6 are required for normal trabecular and cortical bone formation.
In contrast to the osteopenic/osteoporotic phenotype of the loss-of-function LRP-5 knockout mice , our group showed that the LRP-5 gain-of-function transgenic mice have HBM . The HBM mice were developed by targeting expression of human LRP-5G171V to bone using the 3.6 kb rat type I collagen promoter. These mice have increased bone formation, but the mechanisms for this effect are different from those that lead to decreased bone formation resulting from loss of LRP-5. Heterozygous HBM mice (LRP-5G171V/+) have a 100% increase in distal femur trabecular vBMD as measured by pQCT that is detected as early as 5 weeks of age and persists until at least 52 weeks of age. In addition, cortical bone thickness is also increased by 30% in the LRP-5G171V/+ mice. Histological analysis of the femurs indicated that total bone area is increased four-fold in the HBM mice, while the mineralizing surface is increased 40%. However, MAR is not significantly elevated in the HBM mice, indicating that osteoblast activity is not affected by the mutation. Alkaline phosphatase (ALP) staining of calvaria is also elevated in the LRP-5G171V/+ mice, while TUNEL (terminal dNTP transferase-mediated dUTP nick end-labeled) staining demonstrated that osteoblast apoptosis is reduced by 70%. Like the LRP-5−/− mice, osteoclast numbers and bone resorption are not affected by the G171V mutation. In addition, the LRP-5G171V/+ mice have increased femoral and vertebral bone strength , but the bones are otherwise normal in size and shape. Thus, the primary mechanism for increased bone formation in the HBM mice appears to result from elevated osteoblast/osteocyte numbers due to decreased cell death (Fig. 1).
3 Role of β-catenin in osteoblastogenesis and osteoclastogenesis
Further evidence for the importance of the β-catenin pathway in osteogenesis was recently reported by Hu et al. , who evaluated bone development in β-catenin conditional knockout mice (β-catc/c) embryos. When examined at embryonic day 18.5, skeletons of β-catc/c embryos lacked bone, although cartilage was formed. In situ hybridization studies showed that osteoblast differentiation was arrested at the early progenitor stage, and that only type I collagen (TIC) and ALP were expressed. Therefore, β-catenin signaling is required for osteoblasts to complete the differentiation process and synthesize properly formed bone (Fig. 1). To study the effects of β-catenin in limb and head mesenchyme, Hill et al.  used β-catΔPrx1/− mice and showed that β-catenin activity is required for an early step of osteoblast differentiation. On the other hand, stabilization of β-catenin function in the mesenchyme using β-catΔex3Prx/+ mice did not result in increased osteoblastogenesis but instead suppressed chondrogenesis. Day et al.  used Catnbyc/c; Dermo1 (twist-2)-Cre mice to inactivate β-catenin in early mesenchymal progenitor cells as well as Catnbyc/−; Col2a1-Cre mice to remove functional β-catenin from late mesenchymal cells that have committed to the chondrocyte lineage and reported that β-catenin signaling is necessary to inhibit chondrocyte differentiation while allowing osteoblasts to form.
Holmen et al.  studied the role of osteoblastic β-catenin signaling during postnatal murine bone acquisition by conditionally deleting either β-catenin or adenomatous polyposis coli (APC) using the osteocalcin (OC) promoter to drive Cre expression. In the Δβ-catenin mice, micro-CT and histological analysis of long bones demonstrated that both trabecular and cortical bone volume were reduced. This osteopenia correlated with a decrease in osteoblast differentiation and matrix mineralization, as well as an increase in osteoclast differentiation and activity that resulted from down-regulation of osteoblastic osteoprotegerin (OPG) expression and up-regulation of receptor activated by nuclear factor-κB ligand (RANKL) expression. On the other hand, the Δ-APC mice, which had elevated osteoblastic β-catenin levels, exhibited an osteopetrotic phenotype that resulted primarily from reduced osteoclast differentiation and activity as a result of up-regulation of osteoblastic OPG expression and down-regulation of RANKL expression. Links between osteoblastic β-catenin signaling, OPG expression and osteoclastogenesis have also been obtained by Glass et al.  who utilized Cre-lox technology to delete portions of the β-catenin gene in murine osteoblasts. In addition, these authors showed that deletion of TCF-1 in mice produced a low bone mass phenotype as a result of diminished OPG expression. Surprisingly, the authors did not observe significant changes in osteoblastogenesis or bone formation in any of these transgenic or knockout mouse models. In a report evaluating the molecular events associated with Wnt-3a action on mouse C3H10T1/2 pluripotent mesenchymal stem cells, Jackson et al.  also found that OPG expression was up-regulated following activation of the β-catenin pathway. Thus, the canonical Wnt pathway seems to regulate both bone formation and bone resorption via cells of the osteoblast lineage. While alteration of LRP-5 and LRP-6 activity appears to result exclusively in bone formation changes, control of downstream pathway components like β-catenin seems to primarily yield alterations in bone resorption. The reason for these different observations is not entirely clear, but modulation of downstream components is likely to be refractory to feedback control and may therefore represent extreme phenotypes.
4 Role of axin-2 in osteoblast proliferation and differentiation
Axin is an intracellular inhibitor of canonical Wnt signaling [1, 2, 3]. To study the role of Axin-2 in craniofacial morphogenesis, Yu et al.  generated Axin-2−/− mice. These mice exhibited malformations of skull structures as a result of premature cranial suture fusion that resembled craniosynostosis in humans. Characterization of neonatal calvarial-derived osteoblast cultures in vitro demonstrated that deletion of Axin-2 led to an enhancement of cellular proliferation as measured by BrdU labeling. In addition, osteoblast differentiation as measured by ALP, osteopontin and OC expression was stimulated, and matrix mineralization as determined by von Kassa staining was increased, following loss of Axin-2. However, osteoblast apoptosis was not affected by ablation of Axin-2. Immunohistochemical analysis of calvarial suture sections and immunoblot analysis of whole calvaria from the Axin-2−/− mice showed increased levels of activated β-catenin when compared to wild-type controls. Therefore, loss of Axin-2 leads to increased osteoblastic canonical Wnt signaling that results in elevated cellular proliferation and differentiation (Fig. 1).
5 Role of DKK-2 in late-stage osteoblast differentiation and matrix mineralization
Li et al.  examined the role of DKK-2 in osteoblast physiology and bone formation by deleting the gene in mice. Since DKK-2 is an extracellular antagonist of LRP-5 and -6 [1, 2, 3], the authors anticipated that loss of this gene would lead to increased bone formation. However, characterization of the DKK-2−/− mice showed that these animals were osteopenic. Analysis of long bones from 4-month-old knockout mice by pQCT and static histomorphometry demonstrated a 14–16% decrease in trabecular and cortical bone mineral content (BMC) as well as a 31–33% reduction in trabecular bone volume and trabecular number. In addition, deletion of DKK-2 led to a 96% increase in osteoid surface in the absence of a corresponding elevation in osteoblast number or osteoblast surface. Moreover, the DKK-2−/− mice exhibited a 20% reduction in the MAR as measured by dynamic histomorphometry, a measurement of osteoblast activity. Thus, loss of DKK-2 appeared to cause a defect in terminal osteoblast differentiation and matrix mineralization. These in vivo analyses were confirmed by in vitro studies of bone marrow-derived and neonatal calvarial-derived osteoblast cultures, which demonstrated that deletion of DKK-2 resulted in delayed cellular differentiation and matrix mineralization even though canonical Wnt signaling was elevated in cells from the DKK-2−/− mice. When the authors analyzed the expression of DKK-2 mRNA as a function of osteogenic differentiation, they observed that the levels of this secreted Wnt antagonist increased ten-fold with advancing osteoblast development. Furthermore, when the authors over-expressed DKK-2 in bone marrow-derived and neonatal calvarial-derived osteoblast cultures obtained from wild-type mice, they observed an enhancement of matrix mineralization as determined by xylenol orange staining. Therefore, while elevation of canonical Wnt signaling may be required for pre-osteoblast proliferation as well as the initiation and/or progression of the osteoblast through cellular differentiation, suppression of this pathway by antagonists like DKK-2 appears to be important for terminal differentiation and matrix mineralization (Fig. 1).
6 Role of Wnt-10b in osteoblastogenesis
Wnt-10b seems to be one of the ligands that are important in controlling bone formation. Transgenic expression of Wnt-10b in mice using the FABP4 promoter that targets the gene to marrow decreases both white and brown fat formation, provides resistance to diet-induced obesity and increases glucose intolerance [34, 35]. In addition, as described by Bennett et al. , these mice have increased bone formation suggesting that osteogenesis is enhanced while adipogenesis is suppressed (Fig. 1). FABP4-Wnt10b mice exhibit up to a four-fold increase in TBV of the distal femur as determined by micro-CT analysis. Increased trabecular bone mass in the transgenic mice was seen as early as 8 weeks of age and persisted until 23 months, but cortical bone properties remained unaffected by the transgene. Female FABP4-Wnt10b mice are also resistant to ovariectomy-induced trabecular bone loss at 3 months of age. Consistent with these results with the transgenic animals, Wnt-10b−/− mice show a 30% reduction in TBV of the distal femur at 8 weeks of age as measured by micro-CT . Serum OC levels are also reduced by about 30% in the Wnt-10b−/− mice, while serum concentrations of TRAP-5b do not change relative to the wild-type controls. Thus, similar to the loss-of-function and gain-of-function mutations of LRP-5, bone formation and not resorption are altered by deletion of Wnt-10b.
7 Role of sFRP-1 in osteoblast proliferation, differentiation, function and apoptosis
We discovered that sFRP-1 played a role in osteoblast physiology during a series of transcription profiling experiments that sought to elucidate the molecular events associated with human osteoblast differentiation and bone formation . Basal sFRP-1 mRNA levels were observed to increase over 20-fold during human osteoblast (HOB) differentiation from pre-osteoblasts to pre-osteocytes, and then decline in mature osteocytes. This expression pattern correlated with levels of cellular viability such that the pre-osteocytes, which had the highest levels of sFRP-1 mRNA, also had the highest rate of cell death. In addition, expression of sFRP-1 mRNA was induced over 30-fold following prostaglandin E2 (PGE2) treatment of pre-osteoblasts and mature osteoblasts that have low basal message levels. In contrast, sFRP-1 expression was observed to be down-regulated over 75% following transforming growth factor (TGF)-β1 treatment of pre-osteocytes that have high basal mRNA levels. Consistent with this observation, treatment of pre-osteoblasts and mature osteoblasts with PGE2 increased apoptosis three-fold, while treatment of pre-osteocytes with TGF-β1 decreased cell death by 50%. Likewise, over-expression of sFRP-1 in HOB cells that express low levels of the gene accelerated the rate of cell death three-fold. Therefore, these results implied that sFRP-1 is key modulator of human osteoblast and osteocyte survival.
In order to confirm these in vitro observations, we characterized a knockout mouse model . These mice expressed the LacZ gene in place of exon 1 of sFRP-1 so that promoter activity could be measured by β-galactosidase staining. Loss of sFRP-1 in mice increased distal femur TBV by 80% in 35-week-old females as determined by micro-CT. In addition, other trabecular bone parameters like connectivity density, trabecular number, trabecular thickness and trabecular spacing were improved by loss of the gene. But in contrast to the LRP-5G171V/+ mice, loss of sFRP-1 had no effect on cortical bone parameters. An interesting observation about the sFRP-1−/− mice was that prior to 13 weeks of age, there was no difference in trabecular vBMD of the distal femur as determined by pQCT between wild-types and knockouts. However, as the mice aged, the sFRP-1+/+ animals lost trabecular bone, while the sFRP-1−/− mice gained trabecular bone such that by 38 weeks of age there was a 100% increase in vBMD. This difference was then maintained until at least 52 weeks of age. Thus, deletion of sFRP-1 delays and enhances the onset of peak bone mass and suppresses senile bone loss.
As with the LRP-5−/− mice, deletion of sFRP-1 also affected osteoblast activity. Dynamic histomorphometric analysis of proximal femurs from 35-week-old sFRP-1+/+ and sFRP-1−/− female mice showed that deletion of sFRP-1 increased MAR by 30%, indicating that osteoblast activity was increased by loss of the gene. In addition, like the LRP-5G171V/+ mice, deletion of sFRP-1 also suppressed apoptosis. TUNEL staining of calvaria from 33-week-old female mice demonstrated that loss of sFRP-1 led to a 15–20% increase in calvarial thickness and a 50% decrease in osteoblast and osteocyte programmed cell death (PCD). In contrast to the LRP-5−/− mice, deletion of sFRP-1 also affects osteoblast differentiation. When bone marrow from 27-week-old sFRP-1+/+ and sFRP-1−/− female mice was differentiated to osteoblasts in culture by incubation with ascorbic acid, β-glycerolphosphate and dexamethasone, the number of ALP+ cells was increased three- to four-fold by deletion of sFRP-1. Analysis of the differentiating cultures for LacZ expression showed that osteoblast development and matrix mineralization increased as sFRP-1 promoter activity became elevated, suggesting that control of Wnt signaling by sFRP-1 modulates osteoblast differentiation and function. In addition, evaluation of the bone marrow cultures from knockout mice by TUNEL staining showed that cellular apoptosis was suppressed by 70% when compared to cultures from wild-type controls. Finally, like the LRP-5−/− mice, deletion of sFRP-1 also affects osteoblast proliferation. Measurement of DNA synthesis in cultures derived from newborn sFRP-1+/+ and sFRP-1−/− mice calvaria showed that osteoblast proliferation increased two-fold in the knockout cells during the proliferative-phase. However, when the sFRP-1−/− cultures reached confluence and proliferation ceased, the rate of DNA synthesis returned to normal, indicating that the proliferation–differentiation transition was not altered by loss of sFRP-1. Therefore, deletion of sFRP-1 enhances osteoblast proliferation, differentiation and function, while it suppresses osteoblast and osteocyte apoptosis.
Confirmation that Wnts prevent osteoblast apoptosis in vitro was recently reported by Ameida et al. . Using murine C2C12, OB-6 and MC-3T3-E1 cells, the authors showed that treatment with both canonical (Wnt-3a) and noncanonical (Wnt-5a) Wnts suppressed programmed cell death. In addition, while canonical Wnt signaling appeared to play a role in controlling cell survival, additional signaling pathways like Src/Erk (extracellular signal-regulated kinase) and PI3K (phosphatidylinositol 3-kinase)/AKT were also involved in this process.
At least one mechanism for the increased osteoblast differentiation seen in the sFRP-1−/− mice bone marrow cultures is elevated Runx2 expression. Although deletion of LRP-5 does not alter osteoblast differentiation and Runx2 expression , when RNA was isolated from long bones of the sFRP-1−/− mice, Runx2 mRNA levels were increased four- to eight-fold when compared to the sFRP-1+/+ controls . Analysis of the Runx2 promoter identified a putative LEF/TCF response element about 100 bp upstream from the transcription start site, which is adjacent to a Runx2 binding site. Co-transfection of MC-3T3-E1 mouse osteoblasts with a 0.6 kb Runx2 promoter-luciferase construct, TCF-1 and various Wnts showed that canonical Wnts up-regulated Runx2 promoter activity. Moreover, this effect was suppressed by co-transfection with sFRP-1. Thus, canonical Wnts increase Runx2 expression and this is blocked by sFRP-1. The explanation for the lack of effect of LRP-5 deletion on Runx2 expression is not clear, but this may relate to the ability of LRP-6 to compensate for some of the effects of LRP-5 .
To summarize this work from the sFRP-1−/− mice, it appears that sFRP-1 affects many aspects of osteoblast physiology (Fig. 1). Even though expression of sFRP-1 peaks in pre-osteocytes  or osteoid-osteocytes [41, 42], loss of sFRP-1 enhances osteoprogenitor proliferation and differentiation, pre-osteoblast proliferation and maturation, and mature osteoblast activity. Deletion of sFRP-1 also suppresses osteoblast and osteocyte apoptosis. These observations suggest that sFRP-1, which is a secreted Wnt antagonist, is able to regulate osteoblast lineage cells in both autocrine and paracrine manners.
Deletion of sFRP-1 also affects chondrogenesis and endochondral bone formation. As recently reported by Gaur et al. , sFRP-1 is highly expressed in cartilaginous tissues of the developing mouse. Histological analysis of long bones from 4-week-old mice demonstrated that loss of sFRP-1 led to shorted columnar zones as well as increased calcification of the growth plates and primary spongiosa. Using micromass cultures of mouse embryo fibroblasts (MEF), we showed that deletion of sFRP-1 allows the cells to undergo chondrogenesis in the absence of bone morphogenetic protein (BMP)-2 treatment, and that canonical Wnt signaling as well as chondrocyte differentiation were enhanced in the cultures. Transcription profiling of the cells indicated that there was a global down-regulation of Wnt antagonist gene expression in MEFs from sFRP-1−/− mice when compared to wild-type controls, and that this led to suppression of Indian hedgehog mRNA levels and an acceleration of chondrocyte maturation.
8 Conclusions and key unanswered questions
It is clear that canonical Wnt signaling is an important regulator of bone formation through actions on cells of the osteoblast lineage, and essentially each step of the osteogenic process can be affected by this pathway. But this regulation is also complex, and the site of intervention into this pathway clearly determines the mechanisms by which osteoblast physiology is altered (Fig. 1).
Although we have learned much in recent years regarding the role of canonical Wnt signaling in bone formation, some important questions remain to be addressed. For example, we know that most Wnts and FZDs are expressed in bone, but is there a role for Wnt and FZD specificity in the control of osteoblast physiology? The canonical pathway is clearly important for the regulation of bone formation, but do non-canonical pathways also play a role in bone metabolism? Finally, canonical Wnt signaling appears to control osteoclastogenesis through actions on osteoblasts, but do Wnts also have direct effects on bone resorbing cells? These and other questions are likely to be answered in the coming years.
- 16.Lian JB, Stein GS, Aubin JE. Bone formation: maturation and functional activities of osteoblast lineage cells. In: Favus MJ, editor. Primer on the metabolic bone diseases and disorders of mineral metabolism. Washington, District of Columbia: American Society for Bone and Mineral Research; 2003. pp. 13–28.Google Scholar
- 18.Bodine PVN, Robinson JA, Bhat RA, Billiard J, Bex FJ, Komm BS. The role of Wnt signaling in bone and mineral metabolism. Clinical Reviews in Bone and Mineral Metabolism 2006;4:73–96.Google Scholar
- 22.Kharode YP, Green PD, Marzolf JT, Zhao W, Askew R, Yaworsky P, et al. Alteration in bone density of mice due to heterozygous inactivation of LRP6. J Bone Miner Res 2003;18:S60.Google Scholar
- 39.Almeida M, Han L, Bellido T, Manolagas SC, Kousteni S. Wnt proteins prevent apoptosis of both uncommitted osteoblast progenitors and differentiated osteoblasts by beta-catenin-dependent and -independent signaling cascades involving Src/ERK and phosphatidylinositol 3-kinase/AKT. J Biol Chem 2005;280:41342–51.PubMedCrossRefGoogle Scholar
- 41.Lian JB, Trevant B, Symons J, Komm BS, Bodine PVN, Stein GS. The SFRP-1 gene, a Wnt antagonist, is expressed during embryonic development in bone and cartilage. J Bone Miner Res 2002;17:S405.Google Scholar
- 42.Trevant B, Symons J, Gaur T, Hussain S, Komm BS, Bodine PVN, et al. Normal embryonic mouse development does not require sFRP-1, a Wnt antagonist expressed in soft tissues and the skeleton. Submitted for publication, 2005.Google Scholar
- 43.Gaur T, Rich L, Lengner CJ, Hussain S, Trevant B, Ayers A, et al. Secreted frizzled-related protein-1 regulates Wnt Signaling for BMP2 induced chondrocyte differentiation. J Cell Physiol 2006;208:87–96.Google Scholar