Periosteal PTHrP regulates cortical bone modeling during linear growth in mice

Abstract

The modeling of long bone surfaces during linear growth is a key developmental process, but its regulation is poorly understood. We report here that parathyroid hormone-related peptide (PTHrP) expressed in the fibrous layer of the periosteum (PO) drives the osteoclastic (OC) resorption that models the metaphyseal–diaphyseal junction (MDJ) in the proximal tibia and fibula during linear growth. PTHrP was conditionally deleted (cKO) in the PO via Scleraxis gene targeting (Scx-Cre). In the lateral tibia, cKO of PTHrP led to a failure of modeling, such that the normal concave MDJ was replaced by a mound-like deformity. This was accompanied by a failure to induce receptor activator of NF-kB ligand (RANKL) and a 75% reduction in OC number (P ≤ 0.001) on the cortical surface. The MDJ also displayed a curious threefold increase in endocortical osteoblast mineral apposition rate (P ≤ 0.001) and a thickened cortex, suggesting some form of coupling of endocortical bone formation to events on the PO surface. Because it fuses distally, the fibula is modeled only proximally and does so at an extraordinary rate, with an anteromedial cortex in CD-1 mice that was so moth-eaten that a clear PO surface could not be identified. The cKO fibula displayed a remarkable phenotype, with a misshapen club-like metaphysis and an enlargement in the 3D size of the entire bone, manifest as a 40–45% increase in the PO circumference at the MDJ (P ≤ 0.001) as well as the mid-diaphysis (P ≤ 0.001). These tibial and fibular phenotypes were reproduced in a Scx-Cre-driven RANKL cKO mouse. We conclude that PTHrP in the fibrous PO mediates the modeling of the MDJ of long bones during linear growth, and that in a highly susceptible system such as the fibula this surface modeling defines the size and shape of the entire bone.

Introduction

The periosteum (PO) forms a continuous envelope that surrounds the skeleton. It comprises two layers, an outer fibrous layer made up of mesenchymal cells and an inner cambial layer that contains mesenchymal stem cells that form the populations of chondrocytes, osteoblasts and osteoclasts that sculpt and build the cortical surface during bone growth and development (Allen et al. 2004). The PO has a number of important functions associated with its capacity to add and subtract cortical bone (Turner et al. 1987; Pead et al. 1988; Lazenby, 1990; Burr, 1997; Martin et al. 1998; Duan et al. 2001; Ahlborg et al. 2003; Seeman, 2003; Allen et al. 2004). One such function concerns bone strength. The PO perimeter has an enormous impact on bone strength because of the fourth-power function of the moment of inertia (Seeman, 2003), which accounts for the stronger and less fracture-prone skeleton in the male (Martin et al. 1998; Duan et al. 2001; Ahlborg et al. 2003; Seeman, 2003; Allen et al. 2004). This sexual dimorphism is due to the fact that estrogens inhibit PO bone formation, whereas androgens and mechanical loading stimulate it, resulting in larger bones in the male (Burr, 1997; Duan et al. 2001; Seeman, 2003). A second such function concerns bone shape. This results from the capacity of PO bone cells to model the cortical surface of a growing long bone. A prime example here is the modeling or sculpting of the metaphysis to form the subjacent diaphysis, while the growing diaphysis simultaneously adds both endosteal and periosteal bone so that it strengthens as it lengthens (Martin et al. 1998; Allen et al. 2004). In spite of the biological and clinical importance of these aspects of bone size and shape, relatively little is known as to how these PO functions are regulated, particularly as regards local or paracrine regulatory factors.

Parathyroid hormone-related protein (PTHrP) is a member of a small gene family that includes PTH itself (Wysolmerski et al. 1998). The PTH and PTHrP genes have similar structures, and their products have highly homologous N-terminal sequences that act through a common receptor, known as the type 1 PTH/PTHrP receptor or PTH1R (Jüppner et al. 1991; Wysolmerski, 2013). The biological specificity of PTH and PTHrP results from the fact that they act in two separate domains, PTH as a classical systemic peptide hormone and PTHrP as predominantly a paracrine regulatory molecule. Known functions of PTHrP include regulation of the chondrocyte differentiation program during development, mediating tooth eruption, regulating tone in a variety of smooth muscle structures, and mobilizing bone mineral from the skeleton during lactation (Philbrick et al. 1998; Wysolmerski et al. 1998; Wysolmerski, 2013). In a number of these sites, PTHrP gene expression appears to be driven by mechanical loading (Wysolmerski, 2013).

By means of a PTHrP-lacZ reporter mouse, PTHrP expression was identified in the fibrous layer of the PO, as well as in ligament and tendon insertion sites (entheses) of the fibrous or periosteal type but not in entheses of the fibrocartilagenous type (Chen et al. 2006, 2007). In both the PO and fibrous entheses, PTHrP expression peaks during maximal linear growth, and in the insertion sites PTHrP expression has been found to be load-induced (Chen et al. 2006, 2007). Recent studies have shown that PTHrP regulates the modeling of the cortex that creates the root system by which ligaments and tendons are anchored into the bony cortex (Wang et al. 1987, 2014). We report here that PTHrP also appears to regulate the PO modeling of long bone surfaces during linear growth.

Materials and methods

Mice

Scleraxis (Scx) is a transcription factor that is highly expressed in all connective tissue structures that surround muscle and bone as well as the ligaments and tendons that join bone to bone and muscle to bone (Cserjesi et al. 1995; Brent et al. 2003; Murchison et al. 2007). In the developing axial skeleton, Scx is expressed in a region of the somite that lies between the myotome and sclerotome that has been referred to as the syndetome (Brent et al. 2003). The Scx-Cre mouse (Blitz et al. 2009) was outbred onto a CD-1 background and crossed sequentially to PTHrP-lacZ and PTHrP^lox/lox mice to create Scx-Cre/PTHrP^lacZ/lox conditional knockout (cKO) and littermate PTHrP^lacZ/lox mice (Ahlborg et al. 2003; Wang et al. 2013). The receptor activator of NF-kB ligand (RANKL)^lox/lox mouse (Xiong et al. 2011) was also outbred into a CD-1 background before crossing with the Scx-Cre mouse. Gender- and age-matched CD-1 mice served as wild-type controls for all experiments (Wang et al. 2013). The postnatal growth of CD-1 mice is described in detail in data from Charles River (http://www.criver.com/files/pdfs/rms/us-model-pricing/rm_rm_c_cd_1_mice.aspx), and is summarized in the legend of Fig. 1. We focused principally on specimens from 6 and 12–16 weeks old, corresponding to the periods of peak growth and skeletal maturity, respectively (Roach et al. 2003). The PTHrP-lacZ allele is a knock-in construct that is functionless and was used as the null allele in the cKO system (Wang et al. 2013). This enabled us to cross-compare PTHrP and Scx expression patterns in PTHrP-lacZ, Scx-Cre/R26R and the Scx-Cre/PTHrP^lacZ/lox cKO mouse in both anatomical and histological detail; these findings are summarized in the Supplemental Data of this manuscript. All mice were handled according to the US Department of Agriculture Guidelines, and with approval of the Yale Animal Care and Use Committee. The PTHrP-lacZ mouse has been deposited in the Mutant Mouse Regional Resource Center (MMRRC) National Program and is available to all investigators as MMRRC stock number 037092-UCD.

Fig. 1.

Schema of modeling of a long bone during linear growth. In the mouse, the growth spurt kicks in at 3 weeks old, peaks at 3–6 weeks, levels off between 6 and 9 weeks, and asymptotically ceases between 9 and 12 weeks. The modeling activities in the region of the metaphysis and MDJ mirror the growth rate and comprise four separate bone cell activities. One is the carving away of the metaphyseal surface by periosteal osteoclasts, the so-called metaphyseal cut-back phenomenon (Martin et al. 1998) that is the principal focus of this manuscript. The diaphyseal cortex is formed by a combination of endosteal osteoblastic new bone formation in the region of the MDJ and periosteal osteoblastic bone formation on the surface of the newly-formed diaphysis below. Endosteal osteoclasts in the diaphysis remove bone from the inner surface as a component in the overall circumferential growth of the diaphysis.

Anatomical and histological analyses

Specimens were analyzed by micro-CT X-ray tomography, and sagittal and axial cross-section images generated as described (Wang et al. 2013). The chemical and beetle larvae clearing methods were as described (Philbrick et al. 1998; Chen et al. 2007). For double-label experiments, calcein (30 μg g⁻¹) and alizarin (30 μg g⁻¹) were injected i.p. 6 days apart, and the mice killed 36 h after injecting the second fluor. Plastic sections were analyzed by florescent microscopy or stained with tartrate-resistant acid phosphatase (TRAP) or alkaline phosphatase (AP; Wang et al. 2013). Paraffin sections were stained by X-gal (5-bromo-4-chloro-2-indolyl-β-d-galacto-pyranoside), TRAP or AP as described (Chen et al. 2006, 2007; Wang et al. 2013), using a heat step to inactivate endogenous galactosidase activity. The mineral apposition rate (MAR) was calculated as described (Schilling et al. 1992). Immunohistochemistry for RANKL used the protocol from the supplier (Santa Cruz Biotechnology, Santa Cruz, CA, USA). All data were replicated in triplicate unless otherwise indicated.

Results

The working hypothesis examined here concerned whether PTHrP expressed in the fibrous layer of the PO might regulate modeling of the periosteal surface during long bone growth. We studied this question in the proximal tibia and fibula of CD-1 wild-type mice, PTHrP-lacZ gene reporter mice, and mice with a Scx gene-driven cKO of PTHrP in the PO. Cross-comparison of β-gal activity in the PTHrP-lacZ and Scx-Cre/R26R mice allowed us to identify concurrent sites of PTHrP and Scx expression with both anatomical and histological precision in these experiments.

Linear growth begins in the mouse at 3 weeks old, and proceeds in long bones at a rate some six times that of appositional growth (Shaw & Benjamin, 2007). During linear growth the metaphysis is modeled or cut back to form the subjacent diaphysis. This modeling involves a coordinated pattern of bone cell activities, two osteoclastic (OC) and two osteoblastic (Fig. 1). Osteoclasts sculpt the metaphyseal cortex into the forming diaphyseal cortex (often referred to as the metaphyseal cut-back). Osteoblasts drive bone formation on both the endosteal and periosteal surfaces of the forming diaphysis so that its cortex grows larger and stronger as it lengthens. Osteoclasts also remove bone from the endosteal surface as the growing diaphysis enlarges circumferentially. The net result is an adult long bone structure with the same topography as its younger self, but longer and stronger in its diaphysis.

We focused here on OC modeling of the periosteal surface at the metaphyseal–diaphyseal junction (MDJ) of the lateral tibia and proximal fibula. The lateral tibia is a classical site for studying modeling during postnatal bone growth, and the fibula proved to be a particularly informative system that has previously been but little-studied. We describe the findings in these two locations in separate sections below.

Tibia

Data from wild-type and PTHrP cKO tibia are illustrated anatomically in Fig. 2, and histologically in Figs 3 and 4. Modeling of the lateral tibial metaphysis and MDJ was profoundly abnormal in the PTHrP cKO mouse. This was manifest as a near-complete failure to cut back the lateral metaphyseal surface, resulting in a vertical rather than concave metaphysis in the cKO mouse (Fig. 2B,F). In cross-sections, this failure to model took the form of a bulge-like deformity that involved both the lateral as well as posterior surfaces of the proximal tibia (Fig. 2D,C). In addition to the failure to model the periosteal metaphyseal surface, there was a thickening of the cortex in the cKO MDJ region (Figs 2B,D and S4–6). Altogether we examined a total of 15 pairs of gender-matched CD-1 and cKO mice at 12–16 weeks old, and noted the deformed posterolateral MDJ and thickened cortex in 100% of the cKO and none of the CD-1 mice (see Supplemental Data in Fig. S4–6). This cKO phenotype was well established in growing mice at 8 weeks old (Fig. S6A–D), and was particularly dramatic in the few aged cKO mice we examined at 8 months old (Fig. S6E,F).

Fig. 2.

Anatomical views of the adult tibia from control and parathyroid hormone-related protein (PTHrP) conditional knockout (cKO) mice. (A, B) Sagittal micro-CT sections of CD-1 (A) and cKO (B) tibia at 16 weeks old. The arrow in (B) identifies the lateral MDJ, and the arrowhead the insertion site of the MCL. The findings of the MCL site have been reported previously (Wang et al. 2014, ), and will not be further discussed. (C, D) Micro-CT cross-sections of the CD-1 (C) and cKO (D) tibia from (A) and (B) taken at the level of the arrow/arrowhead in (B). The sections are oriented so that the anterior tibial crest is at the top and the lateral MDJ to the left, and the lateral (L) and posterior (P) surfaces as indicated. The arrows in (C) and (D) identify the lateral MDJ; note that in the cKO specimen the MDJ is enlarged and thickened, and that the involved region corresponds to the posterior aspect of the lateral surface as well as the lateral aspect of the posterior surface. (E, F). Anterior view of 16-week-old control (E) and PTHrP cKO (F) tibia cleared by dermastid beetle larve. The tuberosity at the MCL site is indicated (*), as is the deformity reflecting the failed modeling at the lateral MDJ (arrow). Note that the lateral metaphyseal region in the cKO specimen corresponds to a straight vertical surface as compared with the concave surface in the control specimen, and that the entire proximal tibia is enlarged and deformed by these two abnormalities. The space bars in (A) and (B) correspond to 1 mm, and in (C) and (D) to 0.5 mm.

Fig. 3.

Normal and abnormal modeling of the tibial MDJ. (A, B) Modeling surface at the MDJ of parathyroid hormone-related protein (PTHrP)-lacZ (A) and PTHrP conditional knockout (cKO) (B) mice at 6 weeks (X-gal and TRAP stains). (C) By areal histomorphometry, the OC surface was reduced by 72% in the cKO section (shown as mean ± SD for triplicates, P ≤ 0.05, paired test). (D) Modeling surface at the tibial MDJ at 4 weeks in a PTHrP-lacZ mouse (X-gal and TRAP stains). This section was chosen to highlight the mesenchymal cell layers of the cambial PO (indicated by *) that lie between the PTHrP-expressing fibrous layer of the PO and the periosteal osteoclasts on the bone surface; these intermediate layers correspond to the region in which RANKL was identified by immunohistochemistry (see E). The cambial layer of the PO in the metaphyseal region of growing bone is considerably thicker and more complex than it is in the subjacent diaphysis or the adult metaphysis (Fan et al. 2008) (E, F). RANKL immunohistochemistry in MDJ sections from CD-1 (E) and PTHrP cKO (F) mice. (E) The immunostaining is in the intermediate mesenchymal zone (*) highlighted in (D); this layer (*) is immunonegative in the cKO section (F). The counterstain in these sections (Weigert's iron hematoxylin) conveniently stains osteoclasts blue-green, these being abundant in the CD-1 (E) and virtually absent in the cKO (F) section. The space bars correspond to 50 μm.

Fig. 4.

Endosteal osteoblastic bone formation at the MDJ in control and parathyroid hormone-related protein (PTHrP) conditional knockout (cKO) mice. (A, B) Alkaline phosphatase (AP) activity on the endosteal surface of PTHrP-lacZ (A) and PTHrP cKO (B) mice at 6 weeks old (X-gal and AP stains). (C) Quantitation of the AP activity in (A) and (B) by areal histomorphometry (A was assigned a value of unity and B was 3.4-fold increased above this value; *P < 0.005). Fig. S7 supplements these data via images of the endosteal lining osteoblast population in control and cKO mice at 6 weeks. (D, E) Alizarin (red) and calcein (green) double-labels on the endosteal surface at the MDJ site in CD-1 (D) and cKO (E) mice; the double-headed arrows correspond to the distances between the labels. (F) Mineral apposition rates in the sections shown in (D) and (E). This experiment was performed in triplicate twice at 6 weeks. *P < 0.005. The space bars correspond to 300 μm in (A) and (B), and 20 μm in (E) and (F).

As reported previously, there was also abnormal modeling at the insertion site of the medial collateral ligament (MCL) on the medial surface of the proximal tibia in the cKO mouse, which was associated with a traction deformity of this surface (Fig. 2B,D,F; Wang et al. 2014, ). In combination, these two modeling abnormalities deformed and enlarged the entire proximal tibia in the adult cKO mouse (Fig. 2B,D,F).

Tartrate-resistant acid phosphatase staining revealed that the periosteal osteoclast population on the MDJ surface was markedly reduced in the cKO as compared with the PTHrP-lacZ sections at 6 weeks (Fig. 3A–C). The microanatomy of the periosteal layers is illustrated in Fig. 3D. The PO is considerably thicker in the region of the MDJ than on the diaphyseal surfaces (Fan et al. 2008). In Fig. 3D, the outer fibrous layer is easily identifiable by the β-gal staining, and the osteoclasts on the bone surface by the TRAP staining; the cambial layer here is made up of a thickened intermediate zone of mesenchymal cells that lie between the fibrous layer and the surface osteoclasts that have formed from cambial precursor cells. Immunohistochemistry for RANKL localized its production to this intermediate cambial layer (Fig. 3E). Both RANKL staining and osteoclasts were essentially absent in the cKO section (Fig. 3F).

A thickening of the lateral tibial cortex was uniformly present in the cKO mice (Figs 2, 4, S4 and S6). This thickening was centered on the region of failed periosteal modeling (Fig. 4B). This is a region that normally experiences very active endocortical osteoblastic bone formation during linear growth, as was apparent by the strong AP signal and the abundant osteoblast population lining the endosteal surface in the control mice at 6 weeks (Figs 4A and S7A). This endosteal bone formation was magnified several-fold in the cKO mouse, as quantitated by both areal histomorphometry of the AP staining (Fig. 4C) and determination of the MAR using alizarin/calcein double-labels (Fig. 4D–F, see also Fig. S7A,B). The bone so formed was osteon-containing mature cortical bone and not woven bone (Fig. S7C,D). These data suggested that the cortical thickening in the cKO MDJ region might well be the result of an enhancement of the endosteal bone formation that occurs normally in this region in growing mice.

In brief summary, the lateral MDJ of the proximal tibia is a prototypical modeling site during postnatal bone growth, and deleting PTHrP in the fibrous PO of this site aborted RANKL-driven osteoclast surface modeling in this location. This failed surface modeling was associated with a curious thickening of the subjacent endosteal surface, which was presumably somehow related to the dysfunctional modeling at the surface.

The fibula

The fibula is a little-studied bone, and it drew our attention here because of its striking phenotype in the PTHrP cKO mouse. The fibula has several unusual features that might predispose it to being particularly sensitive to dysfunctional modeling. First, the distal fibula fuses with the tibia during the second postnatal week in the mouse so that subsequent growth and modeling of the fibula occurs exclusively at the proximal growth plate and MDJ. Second, the fibular metaphysis and diaphysis have remarkably different 3D structures (Fig. 5A), so that extensive sculpting of the metaphysis is required to convert it into the subjacent diaphysis.

Fig. 5.

Micro-CT analysis of the fibula in CD-1 and parathyroid hormone-related protein (PTHrP) conditional knockout (cKO) mice. (A, B) Lateral 3D images of the tibiofibular complex in CD-1 (A) and cKO (B) mice at 12 weeks; the arrowheads identify the MDJ above, and the arrows the mid-diaphysis below at the two sites at which the cross-sections in (C–F) were taken. Five additional paired examples are shown in Fig. S6 and S7. (C, D) Cross-sections through the MDJ taken 2 mm below the growth plate of the bones shown in (A) and (B), respectively; the medial surface (M) and anterior margin (A) are indicated. (E, F) Cross-sections through the mid-diaphysis (at half the distance from the epiphysis to the fusion site) of bones shown in (A) and (B), respectively, with the medial (M) and anterior (A) surfaces identified. (G, H) Lateral 3D images of the tibia and fibula of CD-1 (G) and cKO (H) mice at 6 weeks; note that the phenotype in the cKO mouse at 6 weeks is as fully developed as it is in the adult at 12 weeks. Identical findings were seen in three additional pairs of CD-1/cKO mice at 6 weeks. (I) Periosteal circumference at the MDJ site shown in (C) and (D) in six pairs of CD-1 and cKO mice at 12 weeks (1.45 ± 0.07 vs. 2.03 ± 0.09 mm in CD-1 vs. cKO, mean ± SEM, *P < 0.001 by paired t-test). (J) Periosteal circumference at the mid-diaphysis site shown in (E) and (F) in the same six pairs (1.40 ± 0.07 vs. 2.03 ± 0.09 in CD-1 vs. cKO, mean ± SEM, *P < 0.002 by paired t-test). The space bars correspond to 1 mm in A, B, G and H; 0.25 mm in C and D; and 0.2 mm in E and F.

The 3D structure of the fibula in adult CD-1 mice could be likened in appearance to an archer's bow, whereas in the cKO mice it has the appearance of a primitive club (Fig. 5A,B). The cKO fibula was enlarged throughout but particularly in the region of the MDJ/metaphysis, giving it its club-like appearance (Fig. 5B). The cross-sections at the MDJ as well as through the mid-diaphysis provided clear visualization of the extent of the enlargement of the fibula (Fig. 5C–F). The periosteal circumference was measured in these two regions in six pairs of CD-1 cKO mice at 12 weeks, and there were highly significant increases in this circumference in both the MDJ and diaphyseal cross-sections (Fig. 5I,J). The cortical thickness was not increased in the cKO mice at either site (Fig. 5C–F). These abnormalities in the cKO mice were well established by 6 weeks old (Fig. 5G,H).

We examined fibular modeling histologically at 6 weeks, when growth and modeling are at their peak (Fig. 6). The modeling of the MDJ in the CD-1 mouse was remarkable in its exuberance, particularly at the anterior, medial and posterior surfaces (Fig. 6A,C,E,G,I). These surfaces correspond precisely to the sites of PTHrP expression as well as OC modeling in the MDJ (Fig. 6I,J), and the von Kossa-stained sections from the CD-1 specimens revealed that the surfaces were riddled by resorption pits in both longitudinal sections (Fig. 6A) and cross-sections through the MDJ (Fig. 6E,G).

Fig. 6.

Fibular histology in CD-1 and parathyroid hormone-related protein (PTHrP) conditional knockout (cKO) mice. (A, B) Undecalcified longitudinal sections of 6-week-old fibulae from CD-1 (A) and cKO (B) mice (von Kossa stains; the arrows identify the anterior margins in the MDJ region). (C, D) Tartrate-resistant acid phosphatase (TRAP)-stained sections contiguous to those in (A) and (B); the arrow in (C) highlights the osteoclast population driving the extensive periosteal resorption evident in (A), (E) and (G). (E, F) Cross-sections through the MDJ of CD-1 (E) and cKO (F) mice at 6 weeks (von Kossa stains of undecalcified sections); the double-headed arrows emphasize the larger width of the cKO fibula. The orientation of the cross-sections is designated in (E) (A is the anterior and M the medial surface). Note that the lateral surface shows minimal modeling (see also Fig. S3D,E). (G, H) Higher magnifications of the von Kossa images shown in (E) and (F). The arrow in (G) identifies the extensively resorbed anterior region of the CD-1 fibula. (I, J) Cross-sections of the fibular MDJ region from X-gal- and TRAP-stained PTHrP-lacZ (I) and PTHrP cKO (J) mice at 6 weeks old. Note that the β-gal pattern (best seen in J) reflects abundant PTHrP expression on the anterior, medial and posterior but not lateral surface of the fibula, that this activity corresponds to the very active TRAP-positive resorbing surfaces in (I) (entirely absent in the cKO specimen in J), and that the cross-sectional footprint in MDJ region is enlarged in the cKO as compared with the PTHrP-lacZ specimen (J vs. I). These β-gal and osteoclast patterns correspond precisely to the sites of the phenotypic abnormalities in the cKO fibular MDJ. Space bars correspond to 200 μm in A–D; 100 μm in E, F, I and J; and 50 μm in G and H.

Deleting PTHrP in the PO of the MDJ was associated with a profound failure of modeling, as assessed by both von Kossa and TRAP staining (Fig. 6B,D,F,H,J). This failure resulted in the club-like appearance of the anterior fibular neck in the MDJ region (Fig. 6B,D), and in the enlarged cross-sectional footprint of the cKO MDJ (Fig. 6E–J). Although the footprint of the cKO specimens was enlarged, the cortical thickness in the cKO was not (Fig. 6E–J).

In brief summary, the fibula is an unusual long bone in that its growth and modeling during the period of linear growth are essentially confined to the proximal growth plate and MDJ. The degree of modeling seen normally at the proximal MDJ is extraordinary, resulting in a cortex in the neck region that is so moth-eaten in appearance as to not resemble a periosteal surface at all. Conditionally deleting PTHrP in the fibrous PO largely aborted surface modeling at this site, resulting in a metaphysis and subjacent diaphysis that were markedly enlarged in overall size. Thus, the periosteal modeling of the proximal fibula appears to be largely responsible for defining the size and shape of the entire bone. While PTHrP appears to have a similar function elsewhere in the skeleton (e.g. the lateral tibial MDJ), the unusual postnatal development of the fibula amplifies the phenotypic effects of deficient PTHrP-induced periosteal modeling in this particular location.

RANKL conditional deletion

We attributed the phenotype in the PTHrP cKO mouse to a failure of OC modeling on the periosteal surface, which in turn was seen as a consequence of failure of PTHrP induction of RANKL and osteoclasts on this surface (Wang et al. 2013). In order to further corroborate this interpretation, we carried out a second conditional deletion experiment in which Scx-Cre was used to delete RANKL on the periosteal surface (Xiong et al. 2011). The working hypothesis here was that the phenotype in the Scx-Cre/RANKL cKO mouse would phenocopy that of the Scx-Cre/PTHrP cKO mouse. This proved to be the case. The lateral tibial MDJ in the RANKL cKO mouse lacked any evidence of concave sculpting in micro-CT sagittal and cross-sections (Fig. 7A–D). Similarly, the fibula was misshapen in the MDJ region, and its cross-sectional footprint enlarged in a fashion that fully phenocopied the PTHrP cKO fibula. There was a moderate degree of trabecular osteopetrosis in the RANKL cKO mouse (Fig. 7B,D), so that this system was not regarded as an informative or reliable means of determining whether there was a response on the endosteal surface similar to that seen in the PTHrP cKO mouse.

Fig. 7.

Micro-CT images of the Scx-Cre/receptor activators of NF-kB ligand (RANKL) conditional knockout (cKO) mouse system. (A, B) Sagittal sections of the proximal tibia from CD-1 (A) and RANKL cKO (B) mice at 12 weeks. Note the absence of modeling in the proximal lateral tibia in (B) (arrow) as well as the trabecular osteopetrosis (see also D). (C, D) Cross-sections at the level of the tibial MDJ in CD-1 (C) and cKO (D) mice. (E, F) 3D images of the fibula in CD-1 (E) and RANKL cKO (F) mice at 12 weeks. (G, H) Cross-sections through the MDJ regions of the fibula in CD-1 (G) and cKO (H) mice at 12 weeks; the anterior (A) and medial (M) surfaces are indicated in (G). This experiment was carried out in quadruplicate littermates and was highly reproducible. Space bars correspond to 0.5 mm in A–D; 1 mm in E and F; and 0.1 mm in G and H.

Discussion

Our working hypothesis focused on PTHrP as a candidate regulatory factor in periosteal modeling. This hypothesis was fashioned from a number of observations. First, PTHrP gene expression was identified in the fibrous layer of the PO, with the PTH1R being deployed in the subjacent cambial layer on the cortical bone surface (Chen et al. 2006, 2007). Second, PTHrP was noted to have a temporospatial expression pattern in classical modeling sites that mirrored the pattern of linear growth (see Fig. S2A–D). Third, PTHrP was also identified in fibrous ligament and tendon insertion sites on the cortical surface, and found to be load-induced and capable of regulating bone cell activities in these sites (Wang et al. 2014, ). Finally, the PO is anchored in the bone bark that flanks the growth plate, and in modeling sites such as the lateral tibial metaphysis/MDJ is thought to be heavily loaded during peak linear growth, much as is the case for structures such as the MCL (Crilly, 1972; Shapiro et al. 1977; Harkness & Trotter, 1978; Houghton & Dekel, 1979). The most direct means of testing this working hypothesis was via the cKO approach reported here, which revealed that conditionally deleting PTHrP in the PO led to profoundly dysfunctional modeling on the cortical surface.

The term ‘modeling’ is used in connection with osteoclast and osteoblast populations that operate separately and independently to determine the size and shape of a developing bone (Martin et al. 1998). In contrast, ‘remodeling’ refers to coupled or coordinated OC and osteoblastic activities in a localized region in which the osteoclasts and osteoblasts work as a team (Martin et al. 1998; Pederson et al. 2008). What we describe in the tibial MDJ and proximal fibula are classical examples of modeling in which PTHrP induces RANKL and osteoclasts that sculpt the proximal metaphysis/MDJ regions into the form of the subjacent diaphyses. This sculpting is entirely an OC process. We suspect that PTHrP is being induced in these modeling sites by growth-induced loading of the fibrous PO, but do not have direct evidence for this. The fibrous and cambial layers of the PO are thicker and more cellular in these metaphyseal regions than in the diaphyseal PO (Fan et al. 2008), and contain the mesenchymal cells that bear PTHrP receptors and produce RANKL.

It is of interest to compare the findings in these modeling sites with what is seen in the PTHrP-expressing fibrous insertion sites on the cortical surfaces of long bones. There are two principal categories of insertion sites, classified as fibrocartilagenous and fibrous entheses (Doschak & Zernicke, 2005; Benjamin et al. 2006; Chen et al. 2007; Shaw & Benjamin, 2007; Blitz et al. 2009; Sugimoto et al. 2013; Wang et al. 2014, ). Fibrocartilagenous and fibrous entheses differ in their location, development, structure and molecular regulation.

Fibrocartilagenous entheses are associated with major muscle insertions (e.g. the Achilles and quadriceps insertions), and are the most complex, being composed of four distinct layers (tendon/ligament connective tissue, unmineralized fibrochondrocytes, mineralized fibrochondrocytes, and bone) that transmit mechanical force to bone in a buffered or graded fashion. Fibrocartilagenous sites develop by a forme fruste of the endochondral pathway, lack a periosteal component, and do not express PTHrP, form osteoclasts or excavate the cortical surface. Several recent studies have characterized the development of fibrocartilagenous sites, which involve BMP-4 in the case of the endochondral program associated with the formation of bony ridges (Blitz et al. 2009), and an interplay of Sox 9 and Scx transcription factors in the development of the cartilaginous, junctional and connective tissue layers of fibrocartilagenous entheses (Sugimoto et al. 2013). Scx-Cre has been used successfully to target conditional deletions in fibrocartilagenous sites, but the phenotypes seen in these experiments do not resemble those seen in the Scx-Cre-driven PTHrP cKO mouse, as the fibrocartilagenous sites in question do not express the PTHrP gene or the RANKL-osteoclast-cortical excavation cascade (Wang et al. 2014, ).

Fibrous insertion sites anchor tendons and ligaments to the metaphyseal and diaphyseal cortical surfaces by means of implantations that have been likened to the root systems of trees (Chen et al. 2006, 2007; Shaw & Benjamin, 2007; Wang et al. 2014, ). Fibrous entheses are continuous with the PO, contain periosteal elements and are thought to have evolved from the primitive PO (Shaw & Benjamin, 2007; Wang et al. 2014, ). The cortical bone at these sites is modeled by periosteal osteoclasts and osteoblasts derived from mesenchymal cells that lie on the cortical surface. These periosteal bone cells form via the intramembranous (periosteal) rather than the endochondral bone program (Chen et al. 2006, 2007; Shaw & Benjamin, 2007; Wang et al. 2014, ). The PTHrP gene is expressed in the mesenchymal cells that overlie the bone cell-forming layer in fibrous entheses, and PTHrP in these sites serves essentially as a load-induced modeling tool that drives the RANKL-dependent osteoclast formation that excavates the root systems by which these sites become anchored (Wang et al. 2014, ). This OC cortical resorption is coupled to periosteal osteoblast bone formation, which cements the insertion into the cortex in what eventually becomes mature osteon-containing cortical bone (Wang et al. 2014, ). There is a spectrum of fibrous entheses, some with a cord-like tendon/ligament and a deep root system, and others associated with a shallow root system spread over a relatively large cortical surface, the simplest of these being the so-called periosteal-muscle insertions that correspond to little more than the two-layered PO (Chen et al. 2006, 2007; Shaw & Benjamin, 2007; Wang et al. 2014). One very curious example of a fibrous entheses is the insertion of the MCL, which migrates along the medial cortical surface of the tibia during linear growth via a migratory canal. This canal is created by PTHrP-driven osteoclast resorption that is uncoupled to osteoblastic bone formation (Wang et al. 2013), a process that is very like the pure OC resorption that is simultaneously modeling the MDJ on the lateral surface. In all of these sites, Scx-driven conditional deletion of PTHrP abrogates excavation of the root system at fibrous entheses, leading to a variety of phenotypic abnormalities and deformities that are distinctly different from those seen in the Scx-driven conditional deletion experiments in fibrocartilagenous sites (Blitz et al. 2009; Sugimoto et al. 2013; Wang et al. 2014, ).

There are several additional examples of PTHrP function that merit mention in this context. One is that PTHrP-driven osteoclast induction mediates the eruption of teeth, and this too represents an entirely uncoupled OC process (Philbrick et al. 1998). Another is that PTHrP is expressed in the perichondrium that surrounds the costal cartilage and serves to prevent the subjacent chondrocytes from mineralizing (Karaplis et al. 1994; Chen et al. 2006). In the global PTHrP KO mouse, the costal cartilage mineralizes, leading to the death of these mice at birth (Karaplis et al. 1994). PTHrP regulates the chondrocyte differentiation program in a number of different chondrocyte populations (Kronenberg, 2003; Chen et al. 2008), but the regulation of the costal chondrocyte is noted here as it reflects PTHrP expression from a surface lining cell population, much as is the case for the PO reported here.

Both the fibula and tibia are foreshortened in the cKO mouse. This might reflect a spillover of Scx-Cre expression into peripheral PTHrP-expressing growth chondrocytes and some degree of impairment of the chondrocyte differentiation program (see Supplemental Data), a limitation in linear growth imposed by the disruption of surface modeling, and/or abnormal loading resulting from the disturbed modeling. The degree of foreshortening was modest and is presumed to have had little impact on the principal features of the PTHrP cKO phenotype. The osteopetrosis in the RANKL cKO system was also unanticipated and may well be a reflection of the aforementioned spillover of Scx expression into peripheral growth chondrocytes and consequent deletion of RANKL from a chondrocyte source (Xiong et al. 2011).

The increase in endosteal bone formation in the region of the tibial MDJ in the PTHrP cKO mouse was highly reproducible. As noted, there was ample endosteal bone formation ongoing in this region in the control CD-1 mouse, so that it would appear that what was occurring in the cKO mouse was an enhancement of a normal or ongoing developmental process rather than the induction of a new process. One would further infer that this enhancement might be a consequence of the conditional deletion of PTHrP, RANKL or an osteoclast product on the cortical surface. The most interesting possibility here would involve a signal from the cortical surface acting via the osteocyte to regulate endocortical osteoblastic activity, a particularly apt candidate in this regard being sclerostin (Keller & Kneissel, 2005; Zuo et al. 2012). Such a signal might serve to couple the surface resorption of the MDJ to the endocortical bone formation in the overall modeling process that converts it into the subjacent diaphysis.

Supporting Information

Additional Supporting Information may be found in the online version of this article: