The Skeletal site-specific role of connective tissue growth factor in prenatal osteogenesis



Background: Connective tissue growth factor (CTGF/CCN2) is a matricellular protein that is highly expressed during bone development. Mice with global CTGF ablation (knockout, KO) have multiple skeletal dysmorphisms and perinatal lethality. A quantitative analysis of the bone phenotype has not been conducted. Results: We demonstrated skeletal site-specific changes in growth plate organization, bone microarchitecture, and shape and gene expression levels in CTGF KO compared with wild-type mice. Growth plate malformations included reduced proliferation zone and increased hypertrophic zone lengths. Appendicular skeletal sites demonstrated decreased metaphyseal trabecular bone, while having increased mid-diaphyseal bone and osteogenic expression markers. Axial skeletal analysis showed decreased bone in caudal vertebral bodies, mandibles, and parietal bones in CTGF KO mice, with decreased expression of osteogenic markers. Analysis of skull phenotypes demonstrated global and regional differences in CTGF KO skull shape resulting from allometric (size-based) and nonallometric shape changes. Localized differences in skull morphology included increased skull width and decreased skull length. Dysregulation of the transforming growth factor-β-CTGF axis coupled with unique morphologic traits provides a potential mechanistic explanation for the skull phenotype. Conclusions: We present novel data on a skeletal phenotype in CTGF KO mice, in which ablation of CTGF causes site-specific aberrations in bone formation. Developmental Dynamics, 2012. © 2012 Wiley Periodicals, Inc.


Connective tissue growth factor (CTGF) is a matricellular protein with a multi-modular structure that affords it multiple roles in normal and pathologic development. Because of its unique modular structure, CTGF has been grouped with other proteins in the CCN family. Named for the first three members to be described in the literature—Cysteine-rich 61 (Cyr61/CCN1), CTGF (CCN2), and Nephroblastoma overexpressed protein (Nov/CCN3)—the CCN family has since been extended to include the Wnt-induced secreted proteins-1 (Wisp-1/CCN4), -2 (Wisp-2/CCN5), and -3 (Wisp-3/CCN6) (Brigstock et al., 2003). CCN proteins are 30–40 kDa proteins that are cysteine-rich, and are composed of the following four conserved modules: module 1 is an insulin-like growth factor (IGF)-binding domain; module 2 is a von Willebrand type C domain; module 3 is a thrombospondin-1 domain; and module 4 is a C-terminal domain containing a putative cysteine knot (Bork, 1993; Brigstock, 1999; Lau and Lam, 1999; Perbal, 2001). The mosaic structure of these proteins allows for their involvement in crucial cellular (mitosis, adhesion, extracellular matrix production) and physiologic (angiogenesis, chondrogenesis, osteogenesis) processes necessary for skeletal development (Brigstock, 2003; Dhar and Ray, 2010).

Matricellular proteins, such as CTGF, are a subset of extracellular matrix (ECM) proteins that are dynamically regulated and serve a complex role in their local environment. In addition to having direct ECM-related scaffolding functions, CTGF also modulates cellular responses to local environmental cues through functional interactions with cell surface integrins, growth factors, proteases, cytokines, or other ECM proteins (Rachfal and Brigstock, 2005; Cicha and Goppelt-Struebe, 2009; Arnott et al., 2011). Because the effects of CTGF are determined by the relative presence or absence of the aforementioned extracellular components in the local milieu, its functions are context dependent (Cicha and Goppelt-Struebe, 2009). As a result, the effects of CTGF on physiologic processes, such as osteogenesis, may differ depending on the skeletal site being investigated.

The importance of CTGF in skeletal development has been established from studies using the global CTGF ablation (knockout, KO) model in mice. First described in 2003, it was shown that CTGF KO mice had defects in growth plate chondrogenesis, angiogenesis, endochondral bone formation, and ECM production (Ivkovic et al., 2003). CTGF KO mice displayed chondrodysplasia in limb and rib cartilage, deformations in Meckel's cartilage, and aberrations in growth plate organization in endochondral bones (Ivkovic et al., 2003). Of interest, the gross skeletal defects seen in the CTGF KO mice were consistent and presented as kinking of the ribs, tibiae, fibulae, radii, and ulnae, as well as craniofacial abnormalities. Due to misshapen ribs and pulmonary hypoplasia, respiratory failure occurred shortly after birth resulting in neonatal lethality (Ivkovic et al., 2003; Baguma-Nibasheka and Kablar, 2008). This constitutes the chief difficulty in studying the role of CTGF in bone development, formation, and function: the inability to study CTGF KO mice at postnatal time points.

As bone formation occurs through both endochondral (replacing a preexisting cartilage anlage) and intramembranous (de novo) ossification processes, additional studies have been conducted to elucidate the effects of CTGF on prenatal skeletal development. In vitro studies using CTGF-null chondrocytes demonstrated decreased proliferation and deregulation of Indian hedgehog (Ihh) and parathyroid hormone-related peptide (PTH-rP), two critical players in proper growth plate organization and endochondral ossification (Kawaki et al., 2008a). Taking into account original evidence of defects in growth plate cartilage, it was concluded that endochondral ossification is negatively affected in the absence of CTGF. Suggested in this conclusion is that bone formed as a result of the endochondral process is also affected in CTGF KO mice. To date, the studies examining the bone phenotype in these mice have been largely qualitative, and demonstrate decreased ossification of intramembranous (cranial vault) and endochondral (metaphyseal trabeculae) bones (Ivkovic et al., 2003; Kawaki et al., 2008a, b). Furthermore, in vitro studies of CTGF-null osteoblasts have demonstrated that their differentiation and function were decreased compared with wild-type (WT) osteoblasts (Kawaki et al., 2008b).

In this study, we extended previous studies to provide an in-depth characterization of appendicular, axial, and craniofacial skeletal phenotypes of CTGF KO mice. We hypothesized that the effect of CTGF on skeletogenesis would be context dependent, and therefore, ablation of CTGF would result in skeletal site-specific aberrations in bone formation. Based on this hypothesis, we selected and analyzed several skeletal sites (bony elements) based on the ossification process(es) through which they form, as well as their location in either the appendicular or axial skeleton. The sites analyzed included femora and tibiae, vertebral bodies, and the craniofacial skeleton. Techniques used included micro-CT, histomorphometry, craniofacial landmark analyses, and quantitative real-time polymerase chain reaction (qPCR).


Multiple Growth Plate Malformations Occur in CTGF KO Mice

We performed measurements in proximal tibial metaphyses of newborn (postnatal day 0, P0) CTGF KO mice and WT littermates (Fig. 1). Three distinct zones—proliferating, prehypertrophic, and hypertrophic—were determined based on previously established morphologic criteria (Farnum and Wilsman, 1987; Hunziker et al., 1987; Hunziker and Schenk, 1989; Kerkhofs et al., 2012). We demonstrated that the proliferating zone is significantly shorter in CTGF KO mice than their WT littermates. There was no significant difference in the length of the prehypertrophic zone in CTGF KO tibia. However, CTGF KO mice did demonstrate a significant expansion in the hypertrophic zone length (Fig. 1). An expansion in this zone has been noted in previous studies (Ivkovic et al., 2003; Kawaki et al., 2008a), however specific measurements were neither taken nor statistically analyzed.

Figure 1.

Growth plate malformations in connective tissue growth factor knockout (CTGF KO) mice. A: Sections through postnatal day (P) 0 wild-type (WT) and CTGF KO proximal tibiae with brackets outlining the proliferating zone (PZ), prehypertrophic zone (PHZ), and hypertrophic zone (HZ). B: Quantification of growth plate zone thicknesses are shown in the graph. **P < 0.01 and ****P < 0.0001. Scale bar = 100 μm.

CTGF KO Mice Have Skeletal Site-Specific Effects on Bone Formation

Using micro-CT, we performed a thorough analysis of the bone microarchitecture of multiple skeletal sites in CTGF KO mice; these included both appendicular (limbs) and axial skeletal divisions (vertebral bodies, craniofacial bones). Three-dimensional (3D) micro-CT reconstructions demonstrated that femora in CTGF KO mice appeared phenotypically normal when compared with WT littermates (Fig. 2A,B). However, tibiae from CTGF KO mice were misshapen, displaying a bend or kink (Fig. 2G,H); while this kink was present in all CTGF KO tibiae, the exact proximal-distal position varied between mice (data not shown). We analyzed the trabecular bone of distal femora and proximal tibiae, as well as whole bone from femoral and tibial mid-diaphyses, from CTGF KO mice and WT littermates (Fig. 2; Table 1). Our results revealed a significant decrease in trabecular percent bone volume (BV/TV) in distal femoral and proximal tibial metaphyses in CTGF KO mice compared with WT littermates (Fig. 2E,F,I,J). Trabeculae were both less numerous and had increased separation in both bones. These changes are indicative of functionally defective endochondral bone formation (Table 1). We also assessed bone volume in the mid-diaphysis (Fig. 2C,D,K,L). Analysis of femoral diaphyses revealed modest reductions in total bone volume (BV) CTGF KO mice compared with WT littermates; however, overall percent bone volume and periosteal perimeter (Ps.Pm) were not significantly affected. Analysis of tibial diaphyses (comprising the region of the kink) demonstrated that total tissue volume (TV), BV, and Ps.Pm were markedly increased in CTGF KO mice. However, the percent bone volume was not significantly different when compared with WT littermates (Table 1).

Figure 2.

Skeletal site-specific effects on formed bone in connective tissue growth factor knockout (CTGF KO) mice. A–L: Micro-CT reconstructions of postnatal day (P) 0 wild-type (WT) and CTGF KO femora (A,B) and tibiae (G,H) are shown, with isolation of mid-diaphyseal femoral bone (C,D), distal femoral trabecular bone (E,F), proximal tibial trabecular bone (I,J), and mid-diaphyseal tibial bone (K,L). Analysis of bone microarchitecture in these isolated regions is displayed in Table 1.

Table 1. Micro-CT analysis of long bone microarchitecture in WT and CTGF KO femora and tibia
Bone site and parametern = 10n = 9KO/WTbBone site and parametern = 9n = 9KO/WT
  • a

    Values are shown as mean ± SEM

  • b

    CTGF Knockout (KO) to Wild-type (WT) ratios (KO/WT; expressed as percentages) are displayed only for values with statistically significant differences.

  • c

    p < 0.05

  • d

    p < 0.01

  • e

    p < 0.0001

Femoral Distal Metaphysis (trabecular bone)  %Tibial Proximal Metaphysis (trabecular bone)  %
 Bone volume/tissue volume (BV/TV; %)29.90 ± 2.017a21.25 ± 3.163c71.1 Bone volume/tissue volume (BV/TV; %)29.03 ± 2.48617.60 ± 2.190d60.6
 Trabecular Thickness (Tb.Th; mm)0.025 ± 0.0010.023 ± 0.001  Trabecular Thickness (Tb.Th; mm)0.026 ± 0.0010.023 ± 0.0005c88.5
 Trabecular Separation (Tb.Sp; mm)0.037 ± 0.0010.045 ± 0.003c121.6 Trabecular Separation (Tb.Sp; mm)0.039 ± 0.0010.048 ± 0.003d123.1
 Trabecular Number (Tb.N; mm-1)11.94 ± 0.6708.96 ± 1.243c75.0 Trabecular Number (Tb.N; mm-1)11.28 ± 0.7817.625 ± 0.909d67.6
Femoral Mid-Diaphysis (whole bone)   Tibial Mid-Diaphysis (whole bone)   
 Bone volume (BV; mm3)0.072 ± 0.0030.060 ± 0.006c83.3 Bone volume (BV; mm3)0.058 ± 0.0030.073 ± 0.005c125.9
 Tissue volume (TV, mm3)0.174 ± 0.0050.160 ± 0.006  Tissue volume (TV, mm3)0.115 ± 0.0030.159 ± 0.003e138.3
 Bone volume/tissue volume (BV/TV; %)42.24 ± 1.35436.66 ± 2.661  Bone volume/tissue volume (BV/TV; %)50.55 ± 2.18145.42 ± 2.277 
 Periosteal perimeter (Ps.Pm; mm)2.25 ± 0.0322.19 ± 0.032  Periosteal perimeter (Ps.Pm; mm)1.783 ± 0.0242.235 ± 0.024e125.4

In the axial skeleton, we studied two locations within the vertebral column—8th thoracic (T8) and 1st lumbar (L1)—and also isolated bone from two sites in the craniofacial skeleton—parietal bone and mandible. Due to the small size of P0 mouse vertebral bodies, we assessed the percent bone volume of the entire vertebral body. Our micro-CT analysis demonstrated a significant decrease in the percent bone volume (BV/TV) of the more caudal (L1) vertebral body, where CTGF KO had 32.5% compared with 48.6% in WT mice (P < 0.03). However, the percent bone volume in the more cephalic (T8) vertebral body was similar in CTGF KO mice, with KO mice having 43.5% compared with 44.9% in WT mice (P = 0.76). We next assessed total bone volume of the parietal bones. Because at this age these flat bones have not yet developed a marrow cavity inside, total bone volume was measured. The bone volume in CTGF KO parietal bones was 0.165 mm3 compared with 0.192 mm3 in WT parietal bones (P < 0.02). Lastly, we analyzed the percent bone volume (BV/TV) of the mandible because the mandible develops around Meckel's cartilage, some of which persists in early postnatal development (Shimo et al., 2004). CTGF KO mice had 31.4% compared with 39.1% in WT mice (P < 0.03).

Skull Phenotypes of CTGF KO Mice

Phenotypic differences of 10 WT and 11 CTGF KO skulls at P0 were compared using 3D morphometric methods to identify both overall variation in skulls and localized shape differences in the two samples of mice. Three-dimensional coordinates of 32 cranial landmarks were recorded on 3D micro-CT isosurfaces (Fig. 3, Supp. Table S1, which is available online). The total landmark set was used to estimate a “global skull” shape for each sample and used to statistically compare overall skull shape between KO and WT mice. We further defined subsets of landmarks to represent the shapes of the three skull regions: cranial vault (green), cranial base (yellow), and facial skeleton (blue) (Fig. 3); these are the three major developmentally and phylogenetically distinct divisions of the vertebrate skull. We performed a Generalized Procustes analysis (GPA) of the global skull and of the three regions to determine the axes along which variation in these data sets were most profound and to explore the differences in shape variation between WT and KO mice. Localized differences in shape for the three landmark subsets were statistically evaluated using Euclidean Distance Matrix Analysis (EDMA) (Lele and Richtsmeier, 2001).

Figure 3.

Landmarks collected from isosurfaces of micro-CT reconstructions of postnatal day (P) 0 murine skulls. A–C: All landmarks used in the analysis of the global wild-type (WT) and connective tissue growth factor knockout (CTGF KO) skulls are shown from inferior (A), superoinferior (B), and lateral (C) views; depicted is a WT skull. Landmarks used in the analysis of the three landmarks subsets are shown on the facial skeleton (blue), cranial base (yellow), and cranial vault (green). Landmarks that define the subsets of regions depicting the facial skeleton, cranial base, and cranial vault are listed in Supp. Table S1 and defined on the laboratory website:

Differences in Cranial Shape in CTGF KO Mice

A GPA was used to superimpose coordinate data sets and to extract shape information from the data set for each skull region (Rohlf and Slice, 1990; Dryden and Mardia, 1998). A Principal Components Analysis (PCA) based on the Procrustes coordinates of the global skull configuration (Fig. 4A) shows a clear separation between CTGF KO mice and their WT littermates along Principal Components axis 1 (PC1), which accounts for 69% of the total shape variation. While the skulls of CTGF KO mice show a wide range of variation along both PC1 and PC2 axes, the WT littermates are more tightly clustered, indicating less variation in skull morphology, especially along PC1. We also saw a separation between CTGF KO mice and their WT littermates along PC1 in the cranial base and face, with little to no separation between these groups in the cranial vault (Fig. 4B–D).

Figure 4.

Differences in skull shape in connective tissue growth factor knockout (CTGF KO) mice. A–D: Results of Principal Components Analysis (PCA) analyses based on Procrustes coordinates from landmarks of the global skull (A), cranial vault (B), cranial base (C), and facial skeleton (D). In each figure, the shape of the cranial region of each mouse is represented as a single dot in the scatterplots of principal component 1 and principal component 2 scores before adjusting for the effects of allometry. The percentage of variation explained by each component is entered in parentheses after the axis label.

Procrustes superimposition reduces the effects of scale, but does not eliminate allometric shape variation that is related to size. We mathematically corrected for correlations among shape variables due to allometry (size related differences in shape) by computing a regression of shape on our chosen measure of size (centroid size) following the methods of Drake and Klingenberg (Drake and Klingenberg, 2008) (see the Experimental Procedures section for more details). The residuals for each specimen estimated from the regression analysis were then used to compute another PCA which did not include allometric effects. Though the distance between KO and WT mice along PC1 is reduced after this correction, the separation between the groups is maintained (data not shown). Additionally, the extent of variation of the WT mice is extended along PC1 after allometric correction indicating a greater variation in the nonallometric effects of shape variation in the WT mice. In summary, both allometric (those closely related to size) and nonallometric shape differences contribute to global differences in shape between the two samples of mice.

Localized Differences in Craniofacial Shape in CTGF KO Mice

We used the nonparametric bootstrap algorithm of EDMA to statistically test for differences in shape between the skulls of CTGF KO mice and WT littermates using landmark subsets that represented cranial vault, cranial base, and facial skeleton (Table 2). A test of the hypothesis of overall difference in shape for the cranial vault and facial skeleton revealed a statistical difference between the two samples (P ≤ 0.05) but a test of difference in overall shape of the cranial base between the CTGF KO and WT mice did not reach statistical significance.

Table 2. Results of EDMA testing for significant localized differences in CTGF KO skulls
Landmark SubsetLandmarks includedaP-valuebStatistically Different Linear Differencesc
Relatively larger in CTGF KORelatively smaller in CTGF KO
  • a

    See Figure 3 for landmark location

  • b

    p-value for test of overall shape difference between CTGF KO and WT skulls for regions of study

  • c

    Linear distances (identified by landmark endpoints) that show statistical difference between groups by confidence interval (α ≤ 0.10)

Cranial baseamsphlsyn0.173  bas-rsyn
Basrimo  bas- lsyn
Facelflacrmaxi0.006lpmx - rflaclzya - rpmxrmaxi - rpmx
lmaxirnslalflac - rpmxlflac - lnslalmaxi - lpmx
lnslarpmxlpmx - rzyalflac - rzyalnsla - rnsla
lpmxrzyalzya - rzyalzya - rflac 
lzya rflac - rnslalflac - rflac 
rflac lnsla - rflaclflac - rnsla 
Cranial Vaultlflacrpfl0.045  lpfl - lpsq
lpflrpsqlflac - rflacrpfl - rpto
lpsqrptolpto - rptolflac - lpsq
lpto lpfl - rflacrflac - rpsq
rflac   lpfl - lpto

To identify localized differences in morphology, nonparametric confidence interval testing was used to identify those linear distances significantly different in the two groups. The results (Table 2; Fig. 5) indicate that many measures of the facial skeleton and anterior cranial vault are significantly increased along the mediolateral axis in the CTGF KO mice, while the rostro-caudal length of the cranial vault of CTGF KO mice is reduced relative to WT littermates. Although the overall shape of the cranial base subset was not shown to be statistically different between the two groups (Table 2), two distances demonstrate significant reduction in the rostro-caudal dimension of the basi occipital bone in the CTGF KO mice (Fig. 5).

Figure 5.

Localized differences in skull morphology in connective tissue growth factor knockout (CTGF KO) mice. Distances between landmarks found to be significantly different when comparing wild-type (WT) and CTGF KO groups by EDMA are shown on a micro-CT reconstructed isosurface of a postnatal day (P) 0 WT mouse. The skull is seen from above with the rostral end to the left and caudal end to the right. Yellow lines indicate linear distances that are significantly larger in CTGF KO mice; magenta lines represent linear distances that are significantly smaller in CTGF KO mice.

Morphologic Traits Unique to CTGF KO Skulls

In addition to the aforementioned quantitative changes in the skull morphology of CTGF KO mice, we also found several unique, morphologic traits that are present in all CTGF KO mice (Fig. 6). Not contained to a specific region of the skull, these dysmorphisms occur in the facial skeleton, mandibles, palate, and cranial base. The nasal bones in WT mice have a relatively straight superior surface (Fig. 6A), while in CTGF KO mice they are curved (Fig. 6B). CTGF KO mice also presented with defects in formation of the mandible resulting in a characteristic S-shaped bend of the mandibular body compared with WT littermates (Fig. 6C,D).

Figure 6.

Morphologic traits seen in connective tissue growth factor knockout (CTGF KO) skulls. A–D: Micro-CT reconstructions of postnatal day (P) 0 wild-type (WT) and CTGF KO global skulls from lateral (A,B) and inferior views (C,D, left) are shown, with isolation of the palate (C,D right) and sphenoid bones (E,F). Increased curvature of the nasal bones is seen in CTGF KO mice (B, yellow arrow) as well as an S-shaped bend in the mandibular body (D, red arrow). A close-up of the palate shows a characteristic kink in the CTGF KO vomer (D, blue arrow), as well as failed midline convergence of the maxillary (purple arrow) and palatine processes (green arrow). Isolation of the sphenoid bones demonstrates a difference in pterygoid plate morphology (white arrow) in CTGF KO mice. Sphenoid bones (E,F) are shown from a posterior view (top row) and inferior view (bottom row) with anterior directed upward.

It was previously reported that CTGF KO mice have a cleft palate at birth secondary to improper development of endochondrally forming cranial bones. We found that in CTGF KO mice the maxillary and palatine processes failed to converge at the midline. Additionally, CTGF KO mice have a characteristic kink of the vomer (Fig. 6C,D), although the side to which this bone kinks alters in fairly equal numbers. Lastly, the pterygoid plates in CTGF KO mice differ quite notably from their WT littermates. WT mice have nearly vertical pterygoid plates, while CTGF KO mice have much more horizontally oriented plates; these appear to be laterally displaced and flattened closer to the body of the sphenoid (Fig. 6E,F). However, the actual body of the sphenoid is not severely affected in CTGF KO mice, although it appears more convex perhaps due to the lateral displacement of the pterygoid plates.

In Vivo Gene Expression Patterns in CTGF KO Skeletal Sites

To examine the effect of CTGF ablation on key osteogenic markers in CTGF KO mice, we performed qPCR analyses using mRNA from two skeletal sites in CTGF KO and WT: (1) the tibial midshaft, which includes the kinked region in CTGF KO mice; and (2) the parietal bone. The patterns of expression of these osteogenic genes were opposite between the two sites. In the tibial midshaft of CTGF KO mice, Runx-2, alkaline phosphatase (ALP), and osteocalcin (OC) were significantly up-regulated compared with WT tibiae (Fig. 7A). In the parietal bone, expression of Runx-2, ALP, and OC was significantly decreased compared with WT parietal bone (Fig. 7B). To determine whether decreased expression in the parietal osteogenic markers was due to altered cellular expression of these genes and not the result of fewer osteoblasts in CTGF KO bone, we performed histomorphometric analyses of the parietal bones using von Kossa stained sections counterstained with toluidine blue to measure the number of osteoblasts per bone perimeter (N.Ob/B.Pm). We demonstrated that CTGF KO parietal bones had 34.2 osteoblasts per 1 mm of bone surface compared with 30.5 in WT mice (P = 0.24).

Figure 7.

Gene expression patterns in connective tissue growth factor knockout (CTGF KO) skeletal sites. A,B: In vivo mRNA gene expression from CTGF KO tibia (A) and parietal bone (B) is graphed relative to WT samples. Specific genes studied include those critical in osteoblast differentiation and function and CCN family members. C: In vivo expression of TGF-β ligand/receptor components from the craniofacial skeleton of CTGF KO mice is graphed relative to WT samples. Abbreviations include: runt-related transcription factor 2 (Runx-2), alkaline phosphatase (ALP), osteocalcin (Oc), TGF-β1 ligand and TGF-β receptors 1 and 2 (TGF-β RI and RII). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

We also investigated the expression levels of CTGF along with two closely related CCN family members—Cyr61/CCN1 and Nov/CCN3—at these two skeletal sites. Deletion of CTGF resulted in a significant reduction in CCN3 levels in the tibial midshaft in CTGF KO mice, while CCN1 levels were not significantly different (P = 0.25) (Fig. 7A). CCN1 expression levels were significantly decreased in the parietal bone CTGF KO mice. While CCN3 levels were decreased relative to WT levels, they were not significantly different (P = 0.06) (Fig. 7B). To rule out the possibility that normal variation in CTGF expression between skeletal sites could account for site-specific changes seen in the CTGF KO mice, we compared CTGF expression from the WT tibia, femur, parietal bone, and mandible. We found that there was no statistical difference between CTGF expression between these sites (data not shown).

Transforming growth factor beta (TGF-β) signaling is required for proper bone formation, and CTGF has been shown to functionally interact with components of this pathway (Abreu et al., 2002). We also noted unique morphologic traits in the CTGF KO skulls that closely resembled the craniofacial phenotype seen in Tgfbr2fl/fl;Wnt1-Cre mice (see Discussion). Therefore, we examined the expression of TGF-β1, TGF-β receptor 2 (RII) (the ligand binding receptor), and TGF-β receptor 1 (RI) (the signaling receptor) in two sites of the craniofacial skeleton: the parietal bone and the mandible. Expression of TGF-β RI in both the parietal bone and mandible significantly decreased in CTGF KO mice. Expression of the ligand, TGF-β1, was also decreased at both sites, with the mandible showing statistical significance and the parietal bone approaching significance (P = 0.052). Although levels of TGF-β RII were also lower in CTGF KO parietal bones and mandibles, the differences were not significantly different (P = 0.078 and P = 0.068, respectively) (Fig. 7C). These findings are consistent with a dysregulation of the TGF-β1 ligand/receptor interaction in the craniofacial skeleton in CTGF KO mice.


Our morphometric results demonstrated a significant decrease in the length of the proliferating zone in CTGF KO mice. This zone is normally maintained by both adequate PTH-rP production in the resting zone as well as Ihh production in the prehypertrophic and hypertrophic zones (Chung et al., 2001). Our findings are consistent with the previous report of decreased PTH-rP and Ihh gene expression in the CTGF KO growth plate (Kawaki et al., 2008a). Furthermore, we found that while the hypertrophic zone length was increased, the prehypertrophic zone length was unaffected. This expansion is likely due to the previously reported decrease in Ihh production by these cells (Kawaki et al., 2008a). It has been shown that abrogation of PTH-rP signaling can still result in proper progression from proliferating through prehypertrophic to hypertrophic chondrocytes (Provot and Schipani, 2005). This could explain why the length of the prehypertrophic zone is unaffected in CTGF KO mice while hypertrophic zone expansion occurs.

Although bone forms through both endochondral (replacing a preexisting cartilage anlage) and intramembranous (de novo) ossification processes, individual bones may form from a combination of the two processes, as seen in the occipital bone of the skull (Rice, 2008). Furthermore, both chondrocytes and osteoblasts originate from precursor mesenchymal cells, the location of their condensations during prenatal skeletal development determines the embryonic cell lineage of the future bone—craniofacial bones from neural crest and paraxial mesoderm, the remaining axial skeleton from somatic mesoderm of the sclerotomes, and the appendicular skeleton from lateral and intermediate mesoderm (Karsenty, 1998). Taking this into account, we examined various skeletal sites to determine if CTGF ablation produces global or site-specific changes in osteogenesis.

Our micro-CT and gene expression analyses demonstrated site-specific differences in bone formation within the appendicular skeleton. Using micro-CT, we found significant decreases in metaphyseal trabecular bone in both the phenotypically abnormal (kinked) tibiae and phenotypically normal (straight) femora in CTGF KO mice compared with WT littermates. These differences indicate defective endochondral ossification, as the trabeculae in this region form through this process. Nonetheless, we cannot discount that altered chondrogenesis could be contributing to these changes in bone mass as we and others have demonstrated dysregulation of the growth plate zones as well as production of crucial cartilage ECM components, such as aggrecan and collagen type X (Ivkovic et al., 2003; Kawaki et al., 2008a). However, it is less likely that aberrant osteoclast function is solely responsible for the bone phenotype presented herein. As it was previously demonstrated that CTGF KO mice have fewer osteoclasts in the metaphyseal region (Ivkovic et al., 2003), this decrease in resorption would be expected to cause a resultant increase in trabecular bone; this is not the case in the CTGF KO mice.

Our analyses of the midshaft of these same bones demonstrated different trends in bone formation, such that it was normal in the femur and increased in the tibia. In the phenotypically normal femur, the percent of bone at the midshaft was not significantly different in CTGF KO mice compared with WT littermates. On the contrary, in the phenotypically abnormal tibia, the midshaft showed a dramatic increase in total bone volume in CTGF KO mice compared with WT littermates. These results demonstrate specific differences in the bone volume at these various sites. To confirm if this was due to an increase in osteoblast function, we analyzed the gene expression of Runx-2 (marker of early osteoblast commitment), ALP (marker of osteoblast maturation), and OC (marker for terminal osteoblast differentiation). Our qPCR analyses demonstrated increased expression of all three osteoblast markers, consistent with increased bone formation at the cellular level. We also found a significant increase in Nov/CCN3 expression, which has previously been shown in the growth plate of CTGF KO mice (Kawaki et al., 2008a). These results demonstrate that not only is bone formation variably affected depending on the skeletal site, and in this case the site within an individual bone, but also that CTGF ablation does not result in a global decrease in bone formation.

An interesting point worth noting is the specificity and reproducibility of the abnormal bones in the CTGF KO skeleton. The CTGF KO phenotype involves distinct kinks in appendicular long bones, specifically bones of the embryonic zeugopod region, which comprises the radius and ulna in the forelimb (FL) and the tibia and fibula of the hindlimb (HL). However, the bones of the stylopod (humerus [FL] and femur [HL]), and autopod (carpus [FL], tarsus [HL], and digits [FL&HL]) remain unchanged. Why CTGF ablation only affects the zeugopod region could be due to changes in skeletal patterning, mechanobiological alterations, or a combination of the two. One of the key classes of genes involved in limb development includes the Hox genes, and it has been shown that specific mutations in Hoxa11, Hoxc11, and Hoxd11 deleteriously affect normal development of the zeugopod (Wellik and Capecchi, 2003; Zakany and Duboule, 2007; Koyama et al., 2010). Mechanobiological cues are also important in proper limb development, where muscle-induced mechanical load is necessary for proper bone formation in utero (Nowlan et al., 2007, 2008, 2010, 2012; Sharir et al., 2011).

The craniofacial skeleton is unique in terms of skeletal development in that it derives from multiple cell types and involves both forms of ossification (Noden and Trainor, 2005; Sperber et al., 2010). Crania are historically analyzed by subdividing them into three regions—the cranial vault or calvaria, the cranial base, and the facial skeleton. During development, a dual-layered capsular membrane known as the ectomeninx surrounds the developing brain. This ectomeninx, of both paraxial mesoderm and neural crest origin, forms the dura mater, which has an outer layer with chondro-/osteogenic properties. In the region of the future calvarium, this membrane will undergo intramembranous ossification, while in the area of the future cranial base this membrane will undergo endochondral ossification. The facial skeleton is of solely neural crest origin and forms through only intramembranous ossification (Sperber et al., 2010). Therefore, generally speaking, global aberrations in endochondral ossification are seen only in the cranial base, while effects on intramembranous ossification present as defects in cranial vault or facial bones, the differences between which can result from their different ossification processes.

Micro-CT, gene expression, and landmark analyses of the axial skeleton of CTGF KO mice demonstrated multiple aberrations in the bone formation these mice. It has been previously shown that the ribcage of CTGF KO mice presents with characteristic kinks in the bone (Ivkovic et al., 2003). When assessing bony elements in the axial skeleton of CTGF KO and WT mice, we analyzed the amount of bone at two sites within the craniofacial skeleton and two locations within the vertebral column. We saw a dramatic difference within the vertebral column, such that the more caudal vertebral bodies (L1) had decreased bone while the more rostral (T8) were not significantly affected. Ossification of the vertebral column at birth is most prominent in the thoracic region of normal newborn mice while other more rostral and caudal sites fully ossify perinatally (Theiler, 1989). Therefore, the decreased ossification found at L1 and not T8 could have resulted from a delay in ossification in CTGF KO mice.

We also analyzed the parietal bones, which form solely from intramembranous ossification. These bones demonstrated decreased ossification in CTGF KO compared with WT mice. This coincides with our findings of decreased expression osteogenic markers in CTGF KO parietal bones (Fig. 7B). These results are in agreement with a previous study that showed a reduction in the expression of some of these markers during osteogenic differentiation of primary osteoblast cultures derived from CTGF KO mice compared with WT littermates (Kawaki et al., 2008b). Furthermore, we demonstrated through histomorphometric analyses that the number of osteoblasts in CTGF KO parietal bones is similar to that in WT parietal bones. Therefore, the decreased expression in osteoblast markers was not due to a decrease in osteoblast numbers but rather a decrease in gene expression on a per cell basis. We also demonstrated a decrease in expression of closely related CCN family member, Cyr61/CCN1 in CTGF KO parietal bones. Since it has been demonstrated that CCN1 can stimulate osteogenesis in vitro (Su et al., 2010), the absence of CTGF coupled with decreased CCN1 expression is consistent with the aforementioned reduction in the expression of osteogenic markers.

We determined phenotypic differences between CTGF KO and WT skulls using 3D coordinates of cranial landmarks for the entire skull, and then analyzed landmarks representing regions of the skull separately. Our PCA analysis of the global skull landmark set demonstrated a clear separation between CTGF KO and WT littermates with both allometric and nonallometric skull shape differences contributing to the changes seen in the skulls of CTGF KO mice. PCA analyses of the landmark subsets representing the three cranial regions also revealed separation between groups.

We further computed localized differences in craniofacial shape using the nonparametric bootstrap algorithm of EDMA (Lele and Richtsmeier, 2001). Overall differences in craniofacial shape were significant in CTGF KO cranial vault and facial skeleton regions, but not so for the cranial base. Confidence interval tests for each linear distance demonstrated relative increases along the medio-lateral axis for the facial skeleton and anterior cranial vault and relative decreases in distances along the rostro-caudal axes of the cranial vault.

In addition to our landmark analyses, we also noted several obvious morphologic differences in CTGF KO skulls; these included changes in the nasal bones, mandibles, palate and vomer, and pteryogoid plates of the sphenoid bone. These abnormal phenotypic traits were conserved in all CTGF KO mice studied. To relate the site specificity of these changes to any potential underlying mechanisms, we need to take into account (1) the embryonic cell origins of these bones, and (2) the signaling pathways known to involve CTGF in craniofacial development. The nasal bones, mandibles, palate, and vomer derive entirely from the cranial neural crest population, while the sphenoid has a dichotomous embryonic cell origin such that the sphenoid body is derived from paraxial mesoderm, while the pterygoid plates are derived from neural crest cells (Noden and Trainor, 2005).

The mandibles of CTGF KO mice, while aberrant in phenotype (Fig. 6D), displayed decreased percent bone volume compared with WT mice. The mandible has a complex process of bone formation involving both intramembranous and endochondral ossification, as it forms around Meckel's cartilage (Lee et al., 2001; Shimo et al., 2004). Additionally, it has been shown that CTGF is required for Meckel's cartilage development (Shimo et al., 2004). Therefore, this decrease in bone could be due to a combination of alterations in the preexisting Meckel's cartilage, decreased ossification at this site, or both.

The TGF-β signaling family has been shown to be involved in CTGF expression and signaling in bone development (Arnott et al., 2011). A role for CTGF in TGF-β signaling-mediated craniofacial development was shown by the Chai laboratory using a cranial neural crest-specific knockout of the TGF-β receptor II (Tgfbr2fl/fl;Wnt1-Cre) (Ito et al., 2003; Oka et al., 2007; Iwata et al., 2010). These mice demonstrated decreased CTGF expression in developing Meckel's cartilage, concomitant with a mandibular phenotype similar to that of the CTGF KO mice (Oka et al., 2007). Analysis of the craniofacial phenotype in the Tgfbr2fl/fl;Wnt1-Cre mice also demonstrated similarities with the CTGF KO mice in the development of their nasal bones, vomer, and palate (Ito et al., 2003; Iwata et al., 2010). Bones formed by endochondral ossification were targeted in mice in which the TGF-β receptor II was conditionally inactivated in chondrocytes under the type II collagen, alpha 1 (Col2a1) promoter (Col2acre+/−;TgfbrloxPloxP). These mice demonstrated changes in the sphenoid body (Baffi et al., 2004, 2006), which were not observed in our analysis of CTGF KO skulls. Therefore, the effects of CTGF ablation on TGF-β RII signaling in neural crest cells may provide an explanation for the craniofacial phenotype seen in our analyses.

We found aberrations in the TGF-β signaling pathway at both the receptor and ligand levels in CTGF KO parietal bones and mandibles. We demonstrated reductions in the expression of both TGF-β RI and RII; TGF-β RI is responsible for propagating the TGF-β signal through phosphorylation of Smads 2 and 3 (Hendy et al., 2005). It is interesting to note that in Tgfbr2fl/fl;Wnt1-Cre mice (discussed above), the expression of CTGF was decreased in developing Meckel's cartilage, and that the abnormal craniofacial phenotype was partially rescued by adding back exogenous CTGF (Oka et al., 2007). These results demonstrate the importance of CTGF as a necessary downstream player in TGF-β-mediated craniofacial development. A comparison of phenotypic changes seen in CTGF KO skulls with those in the skulls of Tgfbr2fl/fl;Wnt1-Cre mice highlights a potentially crucial role of the TGF-β-CTGF signaling loop in craniofacial development. We postulate that the absence of CTGF in mice results in a dysregulation of TGF-β signaling, thus providing a possible mechanistic explanation for the abnormal skull phenotype.



CTGF heterozygous mice (CTGF+/LacZ) were used as breeders to obtain CTGF KO (CTGFLacZ/LacZ) mice as previously described (Crawford et al., 2009). Newborn animals used for this study were killed at birth (P0). Genotype was determined as previously described (Crawford et al., 2009), and animals were split into three groups: wild-type (CTGF+/+), heterozygous (CTGF+/LacZ), and knockout (CTGFLacZ/LacZ). For our studies, CTGF+/+ were used as wild-type (WT) controls.

All animals were maintained and used according to the principles in the NIH Guide for the Care and Use of Laboratory Animals (U.S. Department of Health and Human Services, Publ. No. 86-23, 1985) and guidelines established by the IACUC of Temple University.

Tissue Preparation and Histology

Animals used for this study were euthanized at birth (P0). Subsequently, tails were removed and used for DNA extraction and PCR amplification (Sigma, St. Louis, MO), or X-gal staining (Quiagen, Valencia, CA). Skulls, forelimbs, hindlimbs, and vertebral columns were dissected in phosphate buffered saline (PBS) and fixed immediately in 4% paraformaldehyde (PFA) for 24 hr at 4°C, and replaced with fresh PFA at 48 hr and 72 hr. Specimens for micro-CT scanning were placed in PBS 3 hr before scanning.

Specimens for plastic embedding were fixed in 4% PFA, dehydrated and cleared, followed by infiltration with and embedding in methylmethacrylate resin (Osteo-Bed Bone Embedding Kit, Polysciences, Warrington, PA). The 5-μm sections were obtained from the polymerized resin blocks, then de-plasticized and stained with von Kossa staining technique, for mineralization, and counterstained with toluidine blue (Ivkovic et al., 2003; Abdelmagid et al., 2010; Zhou et al., 2010). High-resolution images were captured with a digital camera attached to a Nikon Eclipse E-600 microscope.


Growth plate zones were quantified using a microscope (Nikon E800) interfaced with a digital camera (QImaging cooled Retiga camera, Surrey, BC, Canada) and bioquantification software system (BioquantOsteo II, Bioquant, Nashville, TN). The zones were identified based on previously established morphologic criteria (Farnum and Wilsman, 1987; Hunziker et al., 1987; Hunziker and Schenk, 1989; Kerkhofs et al., 2012). Briefly, the PZ was determined using the broad, flattened morphology of proliferating chondrocytes arranged in columns. The upper margin of the PHZ is determined as chondrocytes cease proliferating and undergo cytoplasmic expansion. This PHZ exhibits great variation in the morphology of the cells therein as it is a transition zone between the PZ and HZ. The separation of the lower PHZ from the upper HZ is determined by the increased size of hypertrophic chondrocytes and a lack of longitudinal space between neighboring chondrocytes in the HZ. Lengths of specific zones were measured using the auto-width tool of the Bioquant Image Analysis program in which the inner and outer boundaries of structures were traced, and then mean layer thicknesses were automatically generated, using similar methods as previously described (Bove et al., 2009). The assessment and analysis of the data were carried out in a blinded fashion. Three different sections were measured per bone for WT (n = 9) and CTGF KO (n = 7) mice, and the mean values ± SEM are reported.

Micro-computed Tomography (Micro-CT) Analysis

Bones were scanned in air with a Skyscan 1172, 11 MPix camera model, high-resolution cone-beam micro-CT scanner. Long bone images were scanned at a pixel size of 5.2 μm with an X-ray tube potential of 40 kV and X-ray intensity 250 μA. Entire hind limbs were scanned, with each slice equal to 5 μm. A 0.5-mm aluminum filter was used to remove image noise, with a ring artifact correction of 10 and a beam hardening correction of 40%. Skulls were scanned at a pixel size of 9.4 μm with an X-ray tube potential of 80 kV and X-ray intensity 125 μA, with each slice equal to 9 μm. A 0.5-mm aluminum filter was used to remove image noise, with a ring artifact correction of 15 and a beam hardening correction of 40%. After scanning, 3D microstructural image data was reconstructed using the SkyscanNRecon software.

Volumes of interest were isolated and structural indices calculated using the Skyscan CT Analyzer (CTAn) software. For long bones, cortical and trabecular bone were separated manually at 15 μm away from the endocortical surface with an irregular region of interest (ROI) tool. For vertebral bodies, mandibles, and parietal bones, individual bones were isolated using an irregular ROI tool. Left parietal bones and mandibles were used for analysis. Morphometric traits were computed from binarized images using direct 3D techniques, which do not rely on prior assumptions from the underlying structures. The volume of interest for trabecular microarchitectural variables, started 10 μm below the transition into the zone of ossification from the femoral distal epiphysis or tibial proximal epiphysis, and then extended 250 μm toward the diaphysis. An upper threshold of 255 and a lower threshold of 121 were used to delineate each pixel as “bone” or “nonbone.” Trabecular bone volume (BV), trabecular BV per total volume (BV/TV), mean trabecular thickness (Tb.Th), mean trabecular number (Tb.N), and mean trabecular separation (Tb.Sp) indices were computed using a marching-cubes algorithm in 3D. Morphological traits of the mid-diaphyseal region were identified at a site equidistant from the aforementioned starting points, and then extending from this position 50 slices in the proximal and distal directions, totaling 500 μm. Threshold values were identical to trabecular analysis, and BV, TV, BV/TV, and periosteal perimeter (Ps.Pm) indices were computed using a marching-cubes algorithm.

Morphometric Analysis of Skull Phenotypes

Isosurfaces were reconstructed for the CTGF KO (N = 10) and WT (N = 11) mice from the microCT data to characterize all cranial bone using the software package Avizo 6.0 (Visualization Sciences Group, VSG, Burlington, MA). To statistically determine differences in shape of the skulls of CTGF KO mice, 3D coordinate locations of 32 cranial landmarks were recorded using Aviso 6.0. Landmarks were identified on endo- and ectocranial surfaces of cranial bones and used in morphometric analysis. Each specimen was digitized twice by the same observer (TP) and measurement error was minimized by averaging the coordinates of the two trials (Richtsmeier et al., 1995; Aldridge et al., 2005). To ascertain the accuracy and reproducibility of landmark placement, intraobserver error (i.e., absolute difference between the two trials) was estimated for every landmark. If landmark placement differed by more than 0.05 mm, the landmark was remeasured. Landmark definitions are provided in Table S1 and detailed on the website

Variations in skull phenotype were evaluated by analyzing 3D landmark data using two methods of analysis. Generalized Procrustes analysis was used to compare craniofacial shape as defined by landmark coordinate data (Rohlf and Slice, 1990; Dryden and Mardia, 1998). This method superimposes the coordinate data and adopts a single orientation for all specimens by shifting the landmark configurations to a common position, scaling them to a standard size and rotating them until a best fit of corresponding landmarks is achieved. We estimated landmark configurations of global skull using all landmarks, and separately estimated cranial vault, facial, and cranial base configurations by superimposition of specific landmark subsets (Fig. 3) using MorphoJ (Klingenberg, 2008). This procedure reduces the effects of scale (Rohlf and Slice, 1990), but does not eliminate the allometric shape variation that is related to size. To estimate the effect of allometry on shape information, we computed a regression of shape (represented by Procrustes coordinates) on centroid size (Drake and Klingenberg, 2008), estimated as the square root of the summed distances between each landmark location and the centroid of the landmark configuration (Dryden and Mardia, 1998). We explored the variation in the various data sets (global skull, vault, cranial base, facial skeleton) using a Principal Component Analysis (PCA), which performs an orthogonal decomposition of the data and transforms the resulting Procrustes coordinates into a smaller number of uncorrelated variables called principal components (PCs). For each landmark configuration, PCA performs a coordinate rotation that aligns the transformed axes (PCs) with the directions of maximum variation. The first PC (PC1) accounts for the largest amount of variation in the data set, the second PC (PC2) accounts for the second largest amount of variation, and so on (Reyment et al., 1984).

To statistically determine localized shape differences between KO and WT groups we used EDMA, Euclidean Distance Matrix Analysis (Lele and Richtsmeier, 1995; Lele and Richtsmeier, 2001). EDMA converts 3D landmark data into a form matrix consisting of all possible linear distances between unique landmark pairs, computes a relative comparison of form matrices for the samples of interest, and statistically tests for differences in shape between samples using nonparametric confidence intervals (Lele and Richtsmeier, 1995, 2001). We tested for morphological differences in CTGF mice relative to their WT littermates for groups of landmarks that define the whole skull (global skull) and specific regions (facial skeleton, cranial base, cranial vault). Within each group (WT, nonmutant littermates), for each landmark set, an average form matrix, consisting of the linear distance for all unique linear distance pairs, is estimated using the 3D landmark data. Differences in 3D size and shape are statistically compared as a matrix of ratios of all like linear distances in the two samples. The null hypothesis for each comparison is that there is no difference in shape between groups. For each linear distance, a ratio of the average values of that distance for the WT and the KO is computed. A ratio of 1 indicates that the two groups are similar for that measure; whereas a ratio significantly greater or less than 1 shows that they are different. The null hypothesis of similarity in shape for groups of landmarks representing the various landmark subsets is initially evaluated by a bootstrap approach providing an overall indication of difference in shapes between the samples (Lele and Richtsmeier, 2001). Confidence intervals for the null hypothesis of similarity in shape were also evaluated for each linear distance using 100,000 pseudo-samples generated from the data using a nonparametric bootstrapping algorithm. For each linear distance the null hypothesis is rejected if the 90% confidence interval does not include 1.0. Rejection of the null hypothesis enables localization of differences to specific landmarks and linear distances (Lele and Richtsmeier, 1995, 2001). EDMA analyses were performed using WinEDMA (Cole, 2002).

RNA Isolation and qPCR

Total RNA was isolated from P0 calvaria (parietal bones) from WT and CTGF KO mice as described previously (Abdelmagid et al., 2010). Briefly, bones from neonatal mice pups were dissected free from soft tissues and placed immediately in TRizol (Invitrogen, Austin, TX). Samples were homogenized in TRizol until bones were completely dissolved. Total RNA was extracted using acid-phenol-chloroform using centrifugation (10,000 g). The aqueous phase was isolated and transferred to a fresh tube, an equal volume of isopropanol was added, and the sample was incubated at room temperature for 10 min to precipitate the RNA. The RNA was pelleted by centrifugation (10,000 g), washed with 70% ethanol, air-dried, and dissolved in RNase-free water. Spectrophotometer readings were used to determine the concentration of each RNA sample, and the integrity of all samples was confirmed using 1% formaldehyde-agarose gels.

Gene expression for CTGF, runt-related transcription factor 2 (Runx-2), alkaline phosphatase (ALP), collagen, type I, osteocalcin (OC), transforming growth factor beta 1 (TGF-β1), and transforming growth factor beta receptors I and II (TGFβ RI, TGFβ RII), was determined by qPCR using Sybr Green Master Mix (Applied Biosystems, Foster City, CA) using 1 μl of cDNA, as described previously (Song et al., 2007). Reactions were run on a 7,500 Real-Time PCR system (Applied Biosystems). All samples were normalized to control glyceraldehyde-3-phosphate dehydrogenase (GAPDH). At least three independent experiments were performed for each gene, and each experiment was conducted in triplicate. Primers are listed in Table 3. The cycling program was as follows 50°C, 2 min; 95°C, 10 min; 40 cycles of 95°C, 15 sec, and 60°C, 1 min.

Table 3. Primer sequences used for qpcr
Gene NamePrimerSequence (5′-3′)PCR product (bp)GenBank accession no.

Statistical Analysis

Statistical analyses for craniofacial landmark analyses were performed as described above. All other statistical analyses were performed using GraphPad Prism 5 ( Briefly, a two-tailed Student's t-test was used to determine statistically significant differences between group means for histomorphometric, micro-CT, and qPCR results. A P-value of less than 0.05 was considered statistically significant.


We thank Roshanak Razmpour for her unwavering commitment to provide high quality processing, sectioning, and staining of undecalcified, plastic-embedded bones. We also thank Dr. Robin A. Pixley for his expertise in and assistance with micro-CT scanning. S.N.P was funded by NIH/NIAMS, A.G.L. was funded by the PA Department of Health, and J.T.R. was funded by the NIH/NIDCR.