An investigation of cellular dynamics during the development of intramembranous bones: the scleral ossicles

Authors


Correspondence

Tamara A. Franz-Odendaal, Department of Biology, Mount Saint Vincent University, 166 Bedford Highway, Halifax, NS, B3M 4J6, Canada.

E: tamara.franz-odendaal@msvu.ca

Abstract

The development of intramembranous bone is a dynamic and complex process requiring highly coordinated cellular activities. Although the literature describes the detailed cellular dynamics of early mesoderm-derived endochondral bone, studies regarding neural crest-derived intramembranous bone have failed to keep pace. We analyzed the development of chick scleral ossicles from the onset of osteoid deposition to mineralization at morphological, histological, and ultrastructural levels. We find that the mesenchymal condensations from which ossicles develop change their shape from ellipsoidal to trapezoidal concurrent with an increase in size. Furthermore, the size of an ossicle is dependent upon its time of induction. Our histological analyses of condensation growth reveal cell migration and osteoid secretion as key cellular processes determining condensation size; these processes occur concomitantly to increase both the area and thickness of condensations. We also describe the formation of the zone of overlap between ossicles and conclude that the process is similar to that of cranial suture formation. Finally, transmission electron microscopy of early condensations demonstrates that early osteoblasts secrete collagen parallel to the long axis of the condensation. This study elucidates fundamental mechanisms of intramembranous bone development at the cellular level, furthering our knowledge of this important process among vertebrates.

Introduction

Morphogenesis of the skeleton is achieved through a series of highly coordinated cellular activities. These include migration, proliferation, differentiation, apoptosis and the secretion of factors involved in extracellular matrix production and mineralization (reviewed in Franz-Odendaal, 2011). This high level of coordination is achieved primarily through the formation of condensations – areas of high cell density which precede the formation of skeletal tissues (Hall & Miyake, 1992, 2000).

Condensations provide the raw material from which skeletal elements are built, and it is generally accepted that the shape of a condensation defines the basic shape of the element (Hall, 2005). Condensations give rise to bone in one of two ways: during endochondral ossification, mesenchymal cells differentiate into chondrocytes and form cartilage which is later replaced by bone, whereas during intramembranous ossification these cells differentiate into osteoblasts and lay down bone directly. Here we focus on intramembranous ossification, which gives rise to the flat bones of the skull (e.g. the calvariae).

During ossification, some osteoblasts become surrounded by organic matrix (osteoid) composed of ground substance and fibers (primarily collagen I) of their own production. These osteoblasts are trapped in the newly secreted bone matrix, transforming them into osteocytes (reviewed in Franz-Odendaal et al. 2006; Dallas & Bonewald, 2010). How this transformation process occurs remains somewhat unclear. In a recent review by Franz-Odendaal et al. (2006) they suggest three scenarios: osteoblasts may secrete collagen in all directions (i.e. cells are unpolarized), osteoblasts may secrete collagen from one cell surface but not in a specific direction (polarized but not arranged) or osteoblasts may secrete collagen from one cell surface and in a specific direction (polarized and arranged). That is, osteoblasts may bury themselves in matrix, may be buried by their neighbors, or both processes may occur.

In this study, we focus on the scleral ossicles of the chick (Gallus gallus), an excellent but underappreciated model of intramembranous ossification (Franz-Odendaal, 2008, 2011; Duench & Franz-Odendaal, 2012). Like the bones of the skull vault, scleral ossicles are flat bones derived from the cranial neural crest (Couly et al. 1993; Creuzet et al. 2005). Scleral ossicles are large in size, easily accessible, have an early and predictable sequence of development and their condensations do not split or fuse during development (Dunlop & Hall, 1995; Hall & Miyake, 2000; Franz-Odendaal & Vickaryous, 2006; Franz-Odendaal, 2008). In addition, scleral ossicles form as a series, thereby allowing for direct comparisons of development among bones of the same type. These features together make chick scleral ossicles an excellent model for studying cellular dynamics within condensations, and the process of intramembranous ossification.

Scleral ossicle development has been previously described by a number of authors (Murray, 1943; Coulombre et al. 1962; Coulombre & Coulombre, 1973; Fyfe & Hall, 1983; Pinto & Hall, 1991 and others) and most recently by our lab (Franz-Odendaal, 2008; Duench & Franz-Odendaal, 2012) and by Zhang et al. (2012) and Palumbo et al. (2012). Scleral ossicles are induced to form in ectomesenchyme via an unknown signal emanating from the overlying conjunctival papillae in a 1 : 1 ratio. This epithelial–mesenchymal interaction results in the formation of an osteogenic condensation. Conjunctival papillae are known to form and degenerate in a specific pattern, such that each eye contains papillae and ossicles at various stages of development. Papilla #12, directly over the ciliary artery, forms first, followed by the rest of the temporal group (#10–14); the nasal (#2–6) and dorsal (#7–9) groups then form, and papilla #1 over the choroid fissure forms last (Franz-Odendaal, 2008). Ossicle number may vary both between and sometimes within individuals. In Gallus gallus ossicle number ranges from 13 to 16 ossicles per eye (Franz-Odendaal, 2008); the sequence of development is broadly conserved but with some individual variation in the number of ossicles in each group. Each adult eye possesses a ring of slightly overlapping plates tightly held together with dense connective tissue, known as the sclerotic ring (Fig. 1).

Figure 1.

Schematic showing the progression of scleral ossicle development. Hamburger and Hamilton (HH) stages of development are given. Gray circles represent papillae and gray rhomboids and ellipses represent condensations and mineralized ossicles, respectively. Ciliary artery is shown at HH 30 to demonstrate the position of condensation #12. Anterior is to the right. Condensations ultimately give rise to a ring of overlapping bones, the sclerotic ring. P19: post-hatching day 19.

The current study investigates how cellular dynamics give rise to scleral ossicles of specific size and shape. We first examine growth of ossicle condensations at a morphological level, and show that the relative size of an ossicle depends on its time of induction. We then conduct a detailed histological analysis which spans the range of ossicle development from the first sign of condensation (HH 36.5) to the onset of mineralization (HH 38). We show, for the first time, that continued cell migration into the presumptive ossicles and osteoid secretion occur concomitantly to increase ossicle size. We also describe the formation of the zone of overlap between adjacent ossicles and find that this area resembles cranial sutures. Finally, we examine the secretion of collagen by very early osteoblasts and find that, in contrast to previous results (e.g. Ferretti et al. 2002), these cells are both polarized and arranged parallel to the long axis of the condensation. These findings shed light on the complexities underlying the morphogenesis of intramembranous bones in vertebrates and demonstrate how a series of similar bones within the skull coordinate their development.

Materials and methods

Chick embryos

Fertilized White Leghorn (Gallus gallus) eggs were obtained from Cox Bros. Ltd. (Truro, NS, Canada) and incubated at 37 ± 1 °C in a forced-draft incubator. Chicken embryos were staged according to the Hamburger and Hamilton, HH, staging series (Hamilton, 1952) immediately prior to fixation. Isolated fixed eyes from hatchling chickens were provided by William Stell (University of Calgary, Canada).

Gross morphology

To determine accurately the size of ossicles at HH 38, we measured the area and spacing between each ossicle in both eyes of six embryos using NIS Elements (BR 3.00; Nikon Software). We focused on ossicles #3, 4 and 5 (of the nasal group) and ossicles #10, 11 and 12 (of the temporal group) for our analyses. Averages were calculated and statistical significance was determined using Student's t-test (Minitab version 14). Alizarin Red and alkaline phosphatase staining were performed as previously described (Franz-Odendaal, 2008; Edsall & Franz-Odendaal, 2010).

Histology

Whole chick heads were fixed in 10% neutral buffered formalin (Fisher Scientific, 23245685) overnight at room temperature, rinsed in chick saline (0.85% NaCl) and stored in 70% ethanol. One eye per embryo was dissected such that tissue pieces contained condensation #12 and half of each condensation on either side (n = 20 eyes). Condensation #12 can be reliably identified by an anatomical landmark, namely, the ciliary artery, which lies below it (Franz-Odendaal, 2008). Dissected tissues were embedded in paraffin wax (Paraplast Xtra; Fisher Scientific) and sectioned longitudinally beginning at the inner corneal edge at 8 μm thickness with a Leitz 1512 microtome. Sections were stained with Masson's Trichrome as previously described (Franz-Odendaal et al. 2007) and viewed with a Nikon Eclipse 50i compound microscope.

Previous reports show unmineralized condensations at HH 37 (Franz-Odendaal, 2008). Therefore, in order to determine whether cell migration contributes to condensation size, we examined histological sections at HH 36.5. Although not an ideal methodology, the inherent difficulty of cell labeling in advanced stage chick embryos prevents this type of analysis at this time. We measured the angles of ectomesenchymal cells (i.e. the angle that the long-axis of the cell makes with the overlying conjunctival epithelium) adjacent to condensation #12 at HH 36.5 (i.e. the experimental areas). We defined adjacent cells as those cells which are found very near to, but are not a part of, the condensation (Fig. 2A, boxed area). Specifically, the angle of 10 cells above and 10 cells below each condensation for three non-consecutive sections were measured using NIS Elements (BR 3.00; Nikon Software). This process was repeated for three eyes from different embryos, resulting in 180 total measurements. For control areas, we measured the angle of cells (20 per section) in a region of non-condensed mesenchyme located immediately below the conjunctival epithelium at the same stage (HH 36.5).

Figure 2.

Analysis of cell angles. (A) An example showing the angles measured for three cells adjacent to a condensation. Inset shows an enlargement of the boxed area. The condensation is outlined by dashed lines. Scale bar: 20 μm. (B) Side-by-side boxplots comparing the cell angles in control (ctl) and experimental (exp) areas for each of three embryos examined.

The maximum osteoid thickness was also measured in a series of sections through condensation #12 at HH 37.5 and at HH 38 (n = 3 eyes from different individuals, 45 measurements in total and n = 2 eyes from different individuals, 30 measurements in total for HH 37.5 and HH 38, respectively).

Furthermore, to more closely examine the area of overlap between ossicles, we measured the angles of cells at the zone of overlap between ossicles #12 and #13 at HH 38 (n = 2 eyes from different individuals, 120 measurements in total). The space between ossicles #12 and #13 (i.e. the presumptive zone of overlap) was also measured; defined as the distance from the edge of one condensation to the edge of the adjacent one at HH 37.5 (n = 3 eyes from different individuals, 24 measurements in total).

Electron microscopy

Chick embryos staged HH 36.5–37 were decapitated and each eyeball was quickly injected with approximately 1 mL of primary fixative (a mixture of 2.5% glutaraldehyde, 2.5% paraformaldehyde, and 0.04% calcium chloride in 0.01 m sodium cacodylate buffer (pH 7.4)) at room temperature. Chicks were then stripped of eyelids and nictitating membranes. A piece of tissue containing papilla #12 at its center was excised and submerged in primary fixative overnight at 4 °C. Fixed tissues were washed in 0.1 m cacodylate buffer and postfixed with 0.5% osmium tetroxide in 0.1 m cacodylate buffer for 30 min at 4 °C. Tissues were washed in the same buffer, dehydrated in a graded ethanol series (10 min per concentration) and embedded in Epon. Semi-thin sections (2 μm) were stained with 1% toluidine blue O (T3260; Sigma), 1% sodium tetraborate decahydrate (B9876; Sigma) in distilled water and viewed with a Nikon Eclipse 50i compound microscope. Thin sections (60–90 nm) were then cut using a diamond knife, placed on pioliform-coated copper grids, contrasted with 1% uranyl acetate and lead citrate and viewed with a Tecnai G2 electron microscope housed at Dalhousie University (Halifax, NS, Canada).

Results

Gross morphology of developing scleral ossicles

We begin by describing the development of condensations which will give rise to scleral ossicles at a morphological level starting at the time of condensation formation (HH 36.5) and continuing to the onset of mineralization (HH 38).

Alkaline phosphatase (AP) staining is commonly used as an early marker for preosteoblast and osteoblast differentiation. We find that condensed cells express AP as early as HH 36.5, indicating that osteoblast differentiation has commenced at this stage (Fig. 3A). Interestingly, these early condensations exhibit an ellipsoidal shape and appear to span a range of sizes.

Figure 3.

Scleral ossicle development. (A) Alkaline phosphatase staining at HH 36.5 shows small, ellipsoidal condensations of variable size. Anterior is to the right. (B) Unstained embryo at HH 38 shows distinct trapezoidal condensations. (C) Ossicles continue to grow after the onset of mineralization and overlap one another extensively. The basic shape of the condensations is retained at post-hatching day 8 (P8). (D) Histology of condensation #12 at HH 36.5 shows a disorganized mass of cells among osteoid. (E) Ultrastructure of active osteoblasts at HH 36.5 shows a large nucleus, numerous mitochondria, cell–cell contacts, and distended rough endoplasmic reticulum, along with surrounding collagen fibers (asterisk). (F) Osteoblasts are polarized and arranged parallel to the long axis of the condensation at HH 36.5. Inset shows condensation under low magnification stained with toluidine blue. Asterisks indicate areas of active collagen secretion. (G) At HH 37, osteoblasts are separated by osteoid. (H) Ultrastructure of condensed cells, early HH 37. Intercellular spaces increase in size as more collagen fibers are laid down. (I) Whereas most cells move to the periphery of the condensation, some stellate-shaped cells (osteoid osteocytes, asterisk) remain surrounded by osteoid. (J) Osteoid osteocytes contain an extensive labyrinth of rough endoplasmic reticulum (asterisk) at HH 37. (K,L) The condensation expands in width via polarized secretion of osteoid. (M) The presumptive zone of overlap can be identified at HH 37.5. Condensation edges are marked by asterisks. Dashed line represents the conjunctival epithelium. (N) The zone of overlap expands at HH 38 as the condensations grow laterally towards one another via cell recruitment. Condensation edges are marked by asterisks. (O) Extensive overlap between two mineralized ossicles is present at post-hatching day 4 (P4). Ossicles outlined by black dashed lines. Scale bars: 1 μm (E,F,J), 2 μm (H,I), 20 μm (D,G,K,L), 50 μm (M,N), 100 μm (O); 1000 μm (A-C).

Ossicle condensations increase in size over development and are visible in vivo (eyelids removed) by HH 37. Condensations at this stage have undergone obvious changes in shape and are all now trapezoidal. Measurements of condensations in the temporal (early developing) and nasal (late developing) groups at HH 38 reveal that condensations differ significantly in size depending on their location within the sclerotic ring despite having the same general shape (Fig. 3B). Condensations of the temporal group are significantly larger in size than those of the nasal group (unpaired t-test, P < 0.005; = 22.2; df = 60) and are also more closely spaced than those of the nasal group (unpaired t-test, P < 0.005; = 3.0; df = 62). Specifically, condensations of the temporal group have a mean area of 0.59 ± 0.070 mm2 and an average spacing of 211 ± 28 μm, whereas condensations of the nasal group have a mean area of 0.23 ± 0.061 mm2 with an average spacing of 230 ± 30 μm (n = 72 measurements in 12 eyes). Thus, the timing of the onset of condensation induction dictates the relative size of the ossicle. Ossification of these mature condensations then occurs, giving rise to scleral ossicles with the same basic shape as the condensations (Fig. 3C).

Histology of developing scleral ossicles

To better understand how cellular activities contribute to the size and spacing of condensations, we conducted a thorough histological analysis beginning at the time of condensation formation (HH 36.5) and continuing to the onset of mineralization (HH 38).

At HH 36.5, condensation #12 (positioned above the ciliary artery and the first to be induced) is visible as a cluster of rounded osteoblasts interspersed between a sparse collagenous extracellular matrix (Fig. 3D). The overall arrangement of cells appears haphazard; however, a few cells are completely surrounded by osteoid.

Since little evidence of proliferation was previously identified by Fyfe & Hall (1983) and by Franz-Odendaal (2008), we wanted to determine how the condensation enlarges. We wanted to explore previous hypotheses (Fyfe & Hall, 1983; Franz-Odendaal, 2008) that this may occur by continued migration of mesenchymal cells into presumptive ossicles. Given that migrating cells align their long axes parallel to the direction of movement (Wyngaarden et al. 2010), and since cell labeling in opaque embryos is very difficult, we measured cell angles following the method of Sepich et al. (2005). Cells migrating dorso- or ventro-medially into the condensation should have different orientations relative to the surface epithelium than cells in nearby uncondensed mesenchyme. Indeed, we found that cells immediately adjacent to the condensation are orientated differently to those in the control area. Specifically, they have a significantly larger angle relative to the overlying conjunctival epithelium than those in control mesenchyme (25 ± 9° and 17° ± 10°, respectively; unpaired t-test, P < 0.005; = 7.1; df = 354; Fig. 2B). This suggests that these cells migrate dorso- and ventro-laterally into the condensation.

As development proceeds, the condensation becomes reorganized such that the majority of osteoblasts completely surround the osteoid layer, with a few cells embedded in the matrix (Fig. 3G,K). At HH 37.5 the ossicle anlage is composed of two distinct rows of osteoblasts separated by an osteoid layer of 5.5 ± 1.1 μm thickness (n = 45 measurements). From this point onwards, the osteoid layer substantially thickens to 8.2 ± 1.2 μm at HH 38 (n = 35 measurements). We have never observed osteoid between adjacent osteoblasts within the same row (i.e. lateral to cells) or on the contralateral side of the osteoid layer at HH 37–38. This suggests that osteoblasts are polarized and arranged such that collagen secretion occurs towards the existing osteoid at these stages.

These results provide further evidence that (i) condensation size increases via migration, as previously hypothesized (Fyfe & Hall, 1983; Franz-Odendaal, 2008); (ii) the inclusion of migratory cells concomitantly with osteoid deposition acts to increase the size of the ossicles over the course of development; and (iii) osteoblasts in mature condensations secrete collagen in a polarized manner towards the existing osteoid between HH 37 and HH 38.

Development of the zone of overlap between scleral ossicles

Our histological analyses revealed that ossicle #12 begins to approach ossicle #13 at HH 37.5. These ossicles form a highly disorganized zone of overlap composed of the two condensation edges separated by a thin mass of mesenchymal tissue (Fig. 3M). At HH 38, 8 days before hatching, the overlap region becomes more cellular and the edges of the ossicles appear to deflect away from one another (Fig. 3N). Four days after hatching the mineralized bones overlap one another extensively and are separated by a very thin layer of connective tissue (Fig. 3C,O).

We hypothesize based on the above evidence that expanding ossicles recruit cells at the zone of overlap, similar to the process that occurs within sutures (Opperman, 2000). To test this hypothesis we measured cell angles relative to the overlying conjunctival epithelium for cells in the zone of overlap compared with cells in control areas. At HH 38, cells in the zone of overlap form an average angle of 26 ± 9° with the overlying conjunctival epithelium, whereas those in control mesenchyme form an angle of only 12 ± 5°; this difference is statistically significant (unpaired t-test, P < 0.005; = 10.92; df = 90). This finding is consistent with our hypothesis and suggests that recruitment of mesenchymal cells at the lateral edges of condensations actively contributes to ossicle enlargement.

Electron microscopy of early scleral ossicle condensations

While it is well accepted that mature osteoblasts are polarized towards the bone surface with respect to collagen secretion, it remains unclear whether the first osteoblasts to deposit collagen do so in an organized fashion (Franz-Odendaal et al. 2006). To determine whether osteoblasts in very early condensations are polarized and arranged with respect to collagen secretion, we examined the cell ultrastructure within condensation #12 at HH 36.5–37; these are the earliest stages of development in which osteoid is observed under the light microscope (Fig. 3D). Condensed cells at HH 36.5 have many of the features of typical secretory osteoblasts as described by Holtrop (1990): a large, eccentrically located nucleus, one to three prominent nucleoli, an extensive network of rough endoplasmic reticulum and numerous cell–cell contacts that extend over large areas of cell contact or via thin filopodia (Fig. 3E), as described by Palumbo et al. (1990a,b). The nucleus tends to be localized to one pole of the cell, whereas the endoplasmic reticulum is often seen in well-organized sheets, indicating cell polarization. The lumen of the endoplasmic reticulum tends to be highly distended, suggesting a role in material (e.g. procollagen) storage. In addition, mature collagen fibers, with the characteristic 67 nm banding pattern, are haphazardly arranged in the extracellular space immediately adjacent to these cells (not shown).

Unlike the mature osteoblasts described by Holtrop (1990), osteoblasts at this early phase of development have yet to line up along the osteogenic surface and rarely contain a prominent Golgi apparatus. This makes it difficult to determine whether these cells are polarized with respect to collagen secretion at the electron microscopic level. Some authors consider the location of the Golgi apparatus an indicator of cell polarity (Trelstad, 1970; Yadav et al. 2009) whereas others consider the active site of collagen secretion as an indicator of cell polarity (Garant & Cho, 1979). After careful examination of multiple thin sections, noting both cell polarity and the location of collagen bundles, we conclude that active secretion by a single osteoblast occurs in one direction only. The direction of secretion varies from cell to cell but is always oriented parallel to the long axis of the condensation (Fig. 3F). Furthermore, considerably more active secretion events occur by cells within the center of the condensation than by those near its edges.

There is a noticeable increase in collagen density at HH 37 compared with HH 36.5 (Fig. 3F,H-J). In addition, cell–cell contacts at this stage take place mainly via thin filopodia with fewer broad contacts, and the amount of extracellular space within the condensation appears to increase as a result. This increase in extracellular space is further mediated by a shift in cell position as cells come to surround the osteoid layer. This shift is apparent in both light micrographs (Fig. 3D,G,K) and electron micrographs (Fig. 3F,I). However, not all cells make this shift, and those that remain embedded in osteoid likely represent osteoid-osteocytes, precursors to mature osteocytes (Franz-Odendaal et al. 2006). These cells retain their connections to osteoblasts on the edge of the osteoid and appear to span the width of the osteoid at this stage (Fig. 3I). Interestingly, this cell type typically retains an extensive labyrinth of rough endoplasmic reticulum (Fig. 3J), indicative of active collagen production (Palumbo, 1986).

Discussion

In this study, we describe the features of osteogenic condensations which give rise to chick scleral ossicles (flat bones) at a morphological, histological and ultrastructural level with a focus on how cellular dynamics shape the developing bone. In particular, we address (i) the importance of the timing of condensation induction in determining ossicle size, (ii) the role of cell migration in condensation formation and growth, (iii) the importance of the zone of overlap between ossicles and (iv) osteoblast polarization and arrangement with respect to collagen secretion during early development.

Cell migration and osteoid secretion act concomitantly to determine ossicle size

Scleral ossicle condensations are first visible in histological section at HH 36.5, and continue to grow in size until at least HH 38, when mineralization begins (Franz-Odendaal, 2008; Zhang et al. 2012). Previous studies attempting to explain this increase in condensation size over development have found a limited role for cell proliferation in the scleral ossicle system. Studies by Fyfe & Hall (1983) and Van de Kamp & Hilfer (1985), using tritiated thymidine, revealed high levels of proliferation within ectomesenchyme underlying conjunctival papillae at HH 35, prior to condensation formation, but not at HH 36.5, when condensation induction is complete (Fyfe & Hall, 1983; Van de Kamp & Hilfer, 1985; Franz-Odendaal, 2008). Low levels of proliferation were observed at the periphery of condensations but condensed cells themselves were never labeled. Franz-Odendaal (2008) confirmed these early results using both bromodeoxyuridine and proliferating cell nuclear antibody staining methods. Since cell proliferation does not occur within condensations, it cannot be responsible for the observed increase in condensation size over development (i.e. from HH 36.5 to HH 38). The alternative mechanism proposed by these authors is cell migration. Here, we provide further (albeit indirect) evidence that mesenchymal cell migration into presumptive ossicles increases osteoblast numbers within condensations. This approach is feasible due to the elongated morphology of mesenchymal cells and their migratory nature (reviewed in Hay, 2005). Our results show that mesenchymal cells adjacent to condensation #12 align their long axes toward the condensation, whereas control cells align themselves almost parallel to the conjunctival epithelium. It should be noted that, based on our results, it cannot be determined whether these cells are migrating towards or away from the condensation. However, given the increase in size observed in condensations at this time, the latter is unlikely. Furthermore, and importantly, this migration appears to occur concomitantly with osteoid secretion to increase both the thickness and surface area of ossicles (summarized in Fig. 4).

Figure 4.

Schematic summarizing the cellular dynamics within neural crest-derived scleral ossicle condensations. (A) At HH 36.5, polarized osteoblasts secrete collagen fibers parallel to the long axis of the condensation, which thickens via addition of mesenchymal cells dorso- and ventro-laterally. (B) By HH 37, osteoblasts have surrounded the osteoid. (C) Throughout HH 37–38, the condensation expands via polarized secretion of osteoid and addition of mesenchymal cells at the condensation edges (zone of overlap). The long axis of each condensation is in the horizontal plane.

Cells which migrate into the condensation are unlikely to do so randomly and are more likely actively recruited by molecular signals emanating from condensed cells. Bone morphogenetic proteins (BMPs) and activin have been suggested to promote recruitment of mesenchymal cells into chondrogenic condensations (Hall & Miyake, 1995; Duprez et al. 1996). BMPs are expressed in cells surrounding both chondrogenic and osteogenic condensations (Kolodziejczyk & Hall, 1993; Duprez et al. 1996) and therefore likely play a role in increasing condensation size. Recruited cells must also become specified to the osteogenic lineage, another important function of BMP (Chen et al. 2012). Introduction of ectopic Noggin, a BMP inhibitor, results in small or absent ossicles (Duench and Franz-Odendaal, 2008), indicating a role for this protein in the scleral ossicle system. However, this result may be due to a feedback loop involving Sonic Hedgehog (Duench & Franz-Odendaal, 2012), which also appears to have a proliferative role in the overlying epithelium (Franz-Odendaal, 2008). A detailed knowledge of ossicle development at morphological, histological and ultrastructural levels, as described here, will provide the groundwork needed to understand intramembranous ossification at the molecular level.

Additionally, our data show that the formation of the zone of overlap between ossicles is histologically very similar to cranial suture formation (see Opperman, 2000 for a review). Previous research in our laboratory suggests that the overlap region is important in determining ossicle size. Specifically, when formation of one ossicle is inhibited, such that a normal overlap does not form, the two ossicles on either side expand to fill the space (Franz-Odendaal, 2008; Duench & Franz-Odendaal, 2012). It is only when these ossicles approach one another that further growth is arrested. This suggests that neighboring ossicles likely inhibit the growth of one another. This inhibition may occur via molecular signals, physical interaction, or both (Duench & Franz-Odendaal, 2012). Further experiments in which the sclerotic ring is forced to compensate for the loss of multiple ossicles will help to determine the limits of the system and will provide an intriguing model for studying skeletal plasticity.

Early osteoid deposition is not random

Although it is largely accepted that osteoblasts synthesize and secrete individual collagen polypeptides, known as procollagens, which then assemble into large collagen fibers in the extracellular space (Kadler et al. 1996; Manolagas, 2000; Canty & Kadler, 2005), how the very first osteoblasts that deposit collagen fibers are polarized and arranged with respect to procollagen secretion is unclear (see Franz-Odendaal et al. 2006 for a review). This is despite efforts to clarify the process. Ferretti et al. (2002) showed that the early phases of intramembranous ossification occur via a process known as static osteogenesis, in which bone is produced by pluristratified cords of irregularly arranged, variously polarized osteoblasts which transform into osteocytes in the same place that they differentiate (i.e. stationary osteoblasts). In contrast, we propose that early osteoblasts which give rise to scleral ossicles are both polarized and arranged, and that only some of these cells remain stationary and are embedded by the secretions of their neighbors and themselves. Others move and line up adjacent to the secreted osteoid, resembling the monostratified laminae of movable, polarized osteoblasts which carry out dynamic osteogenesis, as described by Ferretti et al. (2002) (Fig. 3). This movement of cells is accompanied by a change in the nature of cell–cell contacts from thin and numerous to broad and few, thereby increasing the amount of extracellular space within the condensation. A similar mechanism has been described during chondrogenesis of the chick limb bud (Searls et al. 1972). Thus, osteogenic condensations which give rise to scleral ossicles are composed of a mixture of static and dynamic osteoblasts. Further research examining movement of cells using cytoskeletal markers could shed further light on this dynamic process when methods are improved for opaque, advanced stage chicken embryos.

Previous studies based on gene expression analyses suggest that skeletogenic condensations are not composed of a homogeneous population of cells in which all cells are at the same stage of maturity (Abzhanov et al. 2007; Li et al. 2009). Specifically, the medial cells within a skeletogenic condensation express late-stage osteoblast markers (such as osteonectin and osteopontin), whereas the lateral cells express early-stage osteoblast markers (such as Runx2). Here, we observed more active collagen secretion events along the midline of the condensations compared to the edges, providing further support for the differential maturity of cells within condensations.

With respect to the phases of collagen synthesis, Gerstenfeld et al. (1988) showed in rat calvarial cell cultures that phases of rapid collagen synthesis are followed by phases of rapid collagen secretion. Our finding that large distensions of the rough endoplasmic reticulum are commonly observed at HH 36.5 suggests that a similar mechanism occurs during the development of scleral ossicles. Electron micrographic analysis of later stages of scleral ossicle development, along with experiments investigating the kinetics of collagen synthesis in these condensations, will help to further our knowledge of this process.

Conclusions

Clearly, the scleral ossicle system is incredibly complex. It does, however, provide the opportunity to explore intramembranous ossification in an accessible system. We add the following knowledge to the growing data regarding morphogenesis of intramembranous bones; however, similar detailed analyses of other flat bones are required to determine if these results are widespread.

  1. Ossicle condensations undergo a change in shape from ellipsoidal at HH 36.5 to trapezoidal at HH 38. Mineralization results in ossicles with the same basic shape as the condensations. Those ossicles induced earlier are larger in size at the onset of mineralization compared with those which are induced later, indicating that the time of induction is important in determining the relative size of ossicles.
  2. Once osteoid deposition has begun, condensations continue to increase in size by the addition of migratory ectomesenchymal cells from surrounding mesenchyme. Cell migration and osteoid secretion occur concomitantly to increase both the area and thickness of ossicles.
  3. The zone of overlap between ossicles is histologically similar to that of cranial sutures; this region is thought to be important in ossicle morphogenesis.
  4. Osteoblasts are polarized and arranged parallel to the long axis of the condensation with respect to collagen secretion.

Acknowledgements

We thank H. Seyan, Z. Lu, G. Roberts and C. Leggiadro for helpful advice and technical assistance. We thank P. Li for use of Dalhousie University's Microscopy Imaging Suite. This work was funded by a Discovery Grant to T.F.O. and an Undergraduate Student Research Award to J.J. from the Natural Sciences and Engineering Research Council of Canada.

Author contributions

Study designed by T.F.O.; experimentation and data analysis performed by J.J. and S.H.; J.J. and T.F.O. wrote the paper.

Ancillary