Characterization of sea urchin primary mesenchyme cells and spicules during biomineralization in vitro

During gastrulation, the sea urchin embryo is actively involved in biomineralization. At this stage of development, the primary mesenchyme cells initiate the assembly of triradiate spicules that grow to form the larval skeleton (Okazaki, 1960; Okazaki & Inoue, 1976). Prior to spicule formation, these cells migrate within the blastocoel, attach to the basal lamina, and fuse with each other via filopodia (Katow & Solursh, 1980, 1981; Galileo & Morrill, 1985) to form a syncytium (Gibbons, Tilney & Porter, 1969; Millonig, 1970) in which mineral (principally CaCO₃) is subsequently deposited (Okazaki & Inoue, 1976). This developmental program is retained in primary mesenchyme cells cultured in vitro (Okazaki, 1975; Harkey & Whiteley, 1980; McCarthy & Spiegel, 1983; Krishnamoorthy & Solursh, 1984; Karp & Solursh, 1985), although the geometry of the spicules found under these conditions is much simpler than that of those found in the embryo.

As an extension of earlier studies on the involvement of glycoconjugates and collagen-like molecules in spicule formation in vitro (Mintz & Lennarz, 1982; Mintz, DeFrancesco & Lennarz, 1981), we have devised a simple system to obtain cultures of isolated primary mesenchyme cells (Carson et al. 1985). This in vitro system is being used to elucidate the cellular processes controlling Ca²⁺ uptake and spicule formation (Carson et al. 1985; Farach et al. 1987). However, a complete understanding of these processes requires more extensive knowledge of the ultrastructural relationship between, primary mesenchyme cells and the growing spicule. Accordingly, in the current study we have used scanning and transmission electron microscopy (SEM, TEM) and electron microprobe techniques to study these cells and the spicules they produce in vitro.

Consistent with previous studies (Krishnamoorthy & Solursh, 1984; Karp & Solursh, 1985), light microscopy and SEM revealed that cultured primary mesenchyme cells from Strongylocentrotus purpuratus embryos extend filopodia and lamellipodia to form multicellular ‘networks’ wherein the spicules are deposited. Both SEM and TEM indicated that openings to the extracellular environment exist in the membrane sheath surrounding the spicule. Consistent with this observation, we found that low pH or the presence of either EDTA or EGTA dissolved the mineral in the spicule compartment, but had no obvious effect on granular deposits in the cell bodies. Subsequent TEM and electron-probe analysis revealed the presence of electron-dense deposits within these cells that, similar to the spicule, contained Ca, Mg and S.

In agreement with recent studies that localized certain glycoproteins to the organic matrix of the larval spicules (Benson, Jones, Crise-Benson & Wilt, 1983; Benson, Benson & Wilt, 1986), we found that, after removal of the electron-dense mineral phase of the spicules, a conjugate of concanavalin A (ConA)-gold penetrated the spicule cavity and specifically bound to membrane-associated glycoproteins in the cavity under conditions where no binding to intracellular glycoconjugates was observed.

Collectively, these results suggest that primary mesenchyme cells cultured in vitro elaborate an extracellular cavity into which they deposit both the mineral and matrix components of the spicule.

embryos were obtained and cultured at 14°C to mesenchyme blastula stage as previously reported (Heifetz & Lennarz, 1979). Primary mesenchyme cells were isolated and cultivated in vitro by a modification of our previously described procedure (Carson et al. 1985). 1 ml sediments of intact embryos in artificial sea water (ASW) were resuspended 1/40 (v/v) in Ca²⁺-, Mg²⁺-free sea water (CMFSW) containing 10 mM-Tris-HCl, pH 7· 8, and 2min-EDTA. These suspensions were pipetted through a large-bore plastic pipet to dissociate the embryos into single cells. As monitored by phase microscopy, dissociation was generally accomplished within 5 min after exposure of the embryos to CMFSW. To remove partially dissociated embryos, the cell suspensions were filtered through a 20μm Nitex filter, sedimented, resuspended in ASW and incubated at 14°C in 20 ml culture dishes in the presence of 4· 0% horse serum and 50μgml^-1 gentamycin sulphate at 14°C.

Approximately 24 h after plating, the unattached cells were decanted from the dish. The dish was then washed with 10 ml of ASW which was later decanted and replaced by 10 ml of ASW plus horse serum and gentamycin. The resultant cultures of attached primary mesenchyme cells were monitored microscopically for up to 5 days following this procedure.

Occasionally, prior to addition of the cells, plastic coverslips (Thermanox no. 1 1/2, 13 mm diameter, Miles Scientific, Naperville, IL) or coverglasses (no. 2,22 mm diameter circles, VWR Scientific, San Francisco, CA) were added to the culture dishes. Subsequently, aliquots of the cell suspension were added and cultured as described above. At time points corresponding to morphologically distinct stages (6, 24, 48 and 72 h; see Results) coverslips were removed from the dishes and examined as described below.

Light microscopy was performed in phase mode on a Nikon Diaphot inverted microscope equipped with ×4 and ×10 objectives and ×20 and ×40 objectives with adjustments for optimizing the working distance between the objective lens and the specimen. When coverslips were used, they were removed from the dishes at the appropriate time points and placed on glass slides (specimen side down). A second glass coverslip (Fisher no. 1 22 mm square) was mounted on top of the first to avoid formation of salt crystals and to slightly flatten the cell bodies to improve the optical resolution for visualization of the intracellular deposits. The specimens were photographically recorded on Kodak Panatomic X(ASA 32).

Because 25 mM-EGTA was often used to demineralize the spicules, it was important to determine if this chelator affected cell viability. Therefore, live primary mesenchyme cells (24 h in culture) were detached from the culture dishes in CMFSW containing 25 mM-EGTA (sea water C (Detering, Decker, Schmell & Lennarz, 1977)). After release from the culture dish, the cells were briefly sedimented for 5 min at 1000g and resuspended in ASW and then cultured as above. Such cultures were monitored in the phase microscope for an additional 48 h. Since cell-substratum attachments and formation of cellular networks occurred during this period (not shown), it was concluded that chelator-treated cells retained viability.

In addition, 72 h cultures were demineralized in CMFSW containing 2mM-EDTA or in the absence of chelator in 0·54M-NaCl. Dissolution of the spicules in CMFSW (approximately 15–30 min) was accelerated by acidification from an initial pH value of 8· 0 (not shown). In contrast, at approximately pH 6·0 in simple NaCl solutions, in the absence of chelator, demineralization proceeded relatively slowly (not shown). To prevent disruption of the thin sheath around the spicule, demineralization of the spicules was also performed in the presence of EGTA or EDTA and either 3· 0 % glutaraldehyde or 2· 0 % paraformaldehyde.

To study the surface features of mesenchyme cells cultured in vitro, cells attached to coverslips were fixed at 4°C for 2–12 h in a solution of ASW buffered with 0·05M-sodium cacodylate (pH 7·4) and containing 3· 0% glutaraldehyde and 0· 5 mg ml^-1 ruthenium red. The specimens were subsequently washed in ASW and postfixed for 1h in 1·0% OsO₄ in ASW. Dehydration, critical-point drying and sputter coating were performed as previously reported (Galileo & Morrill, 1985). SEM was performed on either an ISI SS-4O SEM operated at 10 kV or a Hitachi S520 SEM operated at 15 kV. After recording the spatial coordinates of selected 72 h syncytia that had been photographed, certain specimens were removed from the microscope and fractured with a no. 11 scalpel blade with the aid of a dissecting microscope and an X–Y–Z micromanipulator (Prior, England). To study the surface features of the spicules exposed by this process, the specimens were again sputter coated and photographed in the SEM.

Primary mesenchyme cells that had been cultured for 72 h were gently scraped from the culture dish. The suspensions of isolated cells and spicules were sedimented for 5 min at 1000g. Aliquots of the pellets were mounted on copper supports and quickly plunged into Freon 22 previously supercooled in liquid nitrogen (Kowarski, Shuman, Somlyo & Somlyo, 1979). The specimens were subsequently transferred directly to liquid nitrogen and stored at –85°C. Anhydrous acetone at –85°C was used for freeze substitution, as described earlier (Kowarski et al. 1979). After 3 days, the specimens were transferred from –85°C to –20°C for 1 h and fixed at 4°C for 1 h in 4· 0 % OsO₄ in anhydrous acetone. The specimens were briefly washed in anhydrous acetone and impregnated overnight (12–18 h) with a 1:1 mixture (v/v) of Spurr resin and anhydrous acetone in a closed bell jar containing desiccant. After evaporation of the acetone, the specimens were bathed in Spurr resin in the bell jar for 1–3 days and then polymerized at 60°C for 18 h. Sections were cut with a diamond knife and collected on dry glycerol. Electron-probe analysis was performed using a Phillips 400 TEM equipped with a field emission gun, a Kevex Si (Li) energy dispersive X-ray detector and computerized storage accessories for processing the accumulated X-ray counts per unit assay time, as reported by Somlyo et al. (1981).

In some cases prior to demineralization, the cells and associated spicules were scraped from the dish in 3·0% glutaraldehyde in pH 8·0 ASW and fixed in suspension at 4°C for 30 min. Cell suspensions from five culture dishes were sedimented in 12 ml conical tubes at 1000g for 5 min to form individual pellets that were subsequently resuspended in CMFSW containing 25mM-EGTA for at least 30 min or until dissolution of the spicules was complete. The cells were again sedimented for 5 min and processed for TEM as described below. Cells and spicules attached to culture dishes were similarly fixed and demineralized. To stain the organic matrix of the spicule, the attached cultures were subsequently exposed to 3· 0% glutaraldehyde in ASW containing ruthenium red (0· 5 mg ml^-1) and 0· 05M-sodium cacodylate (pH 7· 4) for a period of 1–2 h at 4°C. After a brief wash in ASW, the substratum-attached cultures and the cell sediments were routinely postfixed at 4°C for 1–2 h in 1· 0% OsO₄ in ASW. The specimens were then washed briefly in ASW and dehydrated through a graded series of ethanol as reported earlier (Decker & Lennarz, 1979). The pellets were then infiltrated overnight in a 1:1 mixture (v/v) of acetone and Spurr resin and embedded in Spurr resin and polymerized at 60°C for 18 h. However, since Spurr resin adversely affected the surfaces of the culture dishes, Polybed 812 (Polysciences, Inc.) was used as embedding medium for the substratum-attached cultures. In this case, after removal of absolute ethanol (see above), the culture dishes were rinsed twice in Polybed 812 resin and allowed to partially polymerize overnight at 45–50°C. The specimens were shifted to 60°C and incubated for an additional 18 h. This procedure provided optically clear specimens for selecting primary mesenchyme syncytia for en face sectioning along the longitudinal axis of the spicule cavities or for transverse sectioning. For this purpose, the culture dishes were fractured by rapid cooling in liquid nitrogen and mechanically separated from the specimen. Pieces were cut from the embedded specimens and glued on to epoxy resin blocks with the basal aspect of the culture exposed for en face sectioning. Alternatively, pieces that had been stripped from the dishes were placed in flat embedding moulds with the specimen side up. Polybed resin was poured over the specimen and polymerized for 18 h at 60°C. The specimens were removed from the moulds, oriented and trimmed via a dissecting microscope for transverse sectioning of the syncytia.

Colloidal gold (8 nm average diameter) was prepared by the method of Muhlpfordt (1982). Salt-free Con A (Sigma) was electrostatically absorbed to the colloid as reported earlier (Armant et al. 1986). Primary mesenchyme cultures (72 h) attached to plastic coverslips were fixed and demineralized for 3h in CMFSW containing 3· 0% glutaraldehyde and 25mM-EGTA. The coverslips were rinsed in 1OSM saline buffered with sodium phosphate at pH 7·4 and incubated on ice for 1 h in 20 ml of fresh buffer. The lectin conjugate was mixed 1:1 (v/v) with the buffer or with 0· 5 M-α-methyl mannoside in 0·15M-sodium phosphate (pH 7·4). To establish the specificity of lectin binding to the demineralized syncytia, the coverslips were incubated for 1 h on ice in the presence or absence of the hapten sugar as described by Roth (1983). The specimens were subsequently washed for 30 min in 1· 0 OSM phosphate-buffered saline and fixed in 1·0 % OsO₄ in ASW and prepared for en face sectioning.

To study spiculogenesis, we used primary mesenchyme cells cultured in vitro on plastic dishes or coverslips. As reported in a related study (Farach et al. 1987), within 6 h after plating, a small percentage of the cells obtained from dissociated mesenchyme blastula stage embryos attached to the substratum. As shown in Fig. 1A–C, the cells that remained attached to the substratum at 24, 48 and 72 h after plating exhibited a number of characteristics that could be used for their identification. First, similar to primary mesenchyme cells within the embryo (see Introduction), the attached cells extended filopodia and gave rise to multicellular networks (Fig. 1A). Second, they stained positively with a monoclonal antibody known to bind specifically to syncytial primary mesenchyme cells in the blastocoel of the embryo or to cells cultured in vitro (Carson et al. 1985; Farach et al. 1987). A third characteristic was the appearance of incipient spicules associated with the cell bodies and with specialized regions of the filo-podial processes (Fig. 1B, arrow). The inset in Fig. 1B shows a minute spicule within a cell body. Such intracellular spicules appeared to emerge from the cell bodies into regions of the filopodial processes that interconnected two cell bodies. At close inspection, smaller, granular inclusions that resemble the spicules were discernible within the perinuclear regions of many of these cells (Fig. 1B, arrowheads), suggesting that constituents of the spicules might accumulate within the cell bodies and be mobilized to the sites of spicule elongation. Intracellular granules that resembled the spicules were also present in cells after 72 h in culture (Fig. 1C and inset). Such cultures exhibited numerous elongated spicules (Grant et al. 1985) that, as shown in Fig. 1C, sometimes grew to greater than 200μm in length.

To further define the structural relationship between primary mesenchyme cells and the spicules in the in vitro culture system, specimens were examined by SEM. As shown in Fig. 2A, as early as 6h after plating both filopodia and lamellipodia were expressed by some of the attached cells. After a longer time in culture (24 h), syncytial networks formed by the attached cells exhibited regions where multiple filopodia had aligned and fused to form a roughly cylindrical sheath (Fig. 2C, arrowheads). In Fig. 2B is shown an enlargement of the boxed region in Fig. 2C. That the spicule develops within a sheath is clearly demonstrated in Fig. 2D–F. Spicules were selected by SEM (Fig. 2D) and then fractured transversely, recoated and viewed again in the SEM (Fig. 2E). Higher magnification (Fig. 2F) clearly revealed that two spicules were partially dislodged from the two adjacent filopodial sheaths shown in Fig. 2D. At still higher magnification, granular substructures of somewhat irregular shape and size (Fig. 2G, arrowheads) could be seen on the spicule surface, particularly towards the tip of the spicule (Fig. 2F, arrow).

Although somewhat similar observations of larval spicule surfaces have been made after removal of associated cells with alkaline hypochlorite (Okazaki & Inoue, 1976), our results using mechanical fracturing clearly show that granular substructures exist on the surface of the mineral component of the spicule and are not the result of chemical etching.

Another important observation made by SEM was the finding that many of the cell bodies were attached to the common spicule sheath via membrane-limited stalks (Fig. 3A, arrow). Numerous extracellular filaments interconnected the exterior membrane domains of this syncytial complex and were associated with the filopodia along the exterior aspect of the sheath surrounding the spicule (arrowheads). Consistent with the idea that the spicule might be deposited in an extracellular cavity, the spicules were found to be readily extractable from the syncytial sheaths in which they had formed (Fig. 3B). However, under a variety of demineralization conditions (see Materials and methods), the deposits in the cell bodies (Fig. 3B, arrows) were still present after dissolution of the spicules.

In an effort to elucidate the cellular route by which Ca²⁺ is accumulated and deposited by isolated primary mesenchyme cells, we prepared specimens by conventional fixation methods and by freeze substitution at –85°C in anhydrous acetone. In agreement with previous studies (Gibbons et al. 1969), electrondense deposits were only occasionally observed in primary mesenchyme cells prepared by conventional chemical fixation. However, as shown in Figs 4 and 5, quick freezing and low temperature substitution of the frozen water phase of the specimen preserved these cells and their associated mineral deposits. In particular, the presence of well-defined organelles, such as mitochondria, rough endoplasmic reticulum, nucleus (Fig. 4A) and the Golgi complex (Fig. 4B) indicated that freeze fixation had been achieved. Moreover, in some thin sections many of the cells displayed relatively massive accumulations of material of high density in perinuclear regions (Fig. 5A) and within mitochondria (Fig. 5B). Smaller dense deposits were occasionally seen in the cisternae of the rough endoplasmic reticulum showing the stacked morphology (not shown) and within dilated cisternae of this organelle (Fig. 5C). Electron-dense granules were also found in smooth-surfaced vesicles near the cell surface (not shown).

X-ray probe analysis of selected subcellular deposits enabled us to demonstrate the presence of sequestered Ca²⁺. As might be expected, spicules (Fig. 6A) were found to contain the highest level of Ca²⁺. Mitochondria (Fig. 6B) also contained relatively high levels of Ca, as well as phophorus, probably due to the presence of endogenous nucleotide phosphates or calcium phosphates. Spectra from other subcellular deposits (Fig. 6C) which could not be readily assigned to particular membrane-bound organelles were often found in perinuclear regions containing the Golgi complex. These deposits also contained high levels of Ca²⁺. In addition, relatively low levels of Mg²⁺ and S were found in some of the deposits that contained sequestered Ca²⁺. In contrast, Ca²⁺ could not be detected in regions of the cytoplasm that did not exhibit electron-dense deposits (Fig. 6D).

Visualization of the interior structural features of the cavity in which the spicule elongates required removal of the electron-dense mineral phase of the spicule. As shown in Fig. 7A, primary mesenchyme cells that had been isolated from the culture dish in the presence of glutaraldehyde could be demineralized in CMFSW containing EGTA. Examination of thin sections of these preparations revealed little structural material within the syncytial cavities, suggesting that, despite the use of prefixation, the organic matrix of the spicule (see Introduction) had been extracted. Similar to our observations using SEM (see Fig. 3B), in some thin sections (Fig. 7B) small openings were observed between the filopodial processes forming the spicule sheath. To gain further insight into the structural organization of the spicule cavity, 72h cultures were fixed, demineralized and embedded with the cells still attached to the culture dish. Similar to the results shown in Fig. 7A and B, transverse sections of the spicule cavity (Fig. 7C) generally appeared devoid of matrix material. This approach also showed that the lamellipodia contain cytoplasm and were continuous with the sheath forming the spicule cavity (Fig. 7C). Additionally, as shown in Fig. 7D, coated pits (arrowheads) were observed at both sides of the syncytial cavity.

As an alternative approach for examining the spicule cavity, fixed, demineralized primary mesenchyme cell cultures were postfixed in glutaraldehyde in the presence of ruthenium red and prepared for en face sectioning. As shown in Fig. 8A, extracellular matrix remained associated with the basal aspect of the primary mesenchyme cell after removal of the plastic substratum. Numerous distinct filaments (arrows) were observed within this adhesive matrix.

Sectioning farther into the cellular complex revealed the interior of the spicule cavity (Fig. 8B) and a structurally distinct matrix associated with the inner face of the limiting membrane. As seen at higher magnification (Fig. 8C, arrowheads), the ruthenium-red-stained matrix exhibited regions of relatively intense staining, possibly indicative of regional variation in the distribution of anionic sites.

Further evidence that the spicule cavity is accessible to the external environment was obtained by the use of a membrane impermeant conjugate of Con–Agold. The results in Fig. 9A reveal binding of Con–Agold to sites associated with the interior membrane surface of the cavity (arrowheads) and to fragments of the residual spicule matrix. Examination of another cavity at higher magnification (Fig. 9B) clearly demonstrated that such Con–Agold binding sites (arrowheads) were often displaced from the membrane surface by distances of greater than 10 nm, which is similar to the position of the ruthenium-red-stained matrix components observed in Fig. 8C. Regions of the plasma membrane and extracellular filaments on the outside of the cavity also bound ConA-gold (Fig. 9B). However, in all cases ConA–gold binding was greatly reduced in the presence of a-methyl mannoside (Fig. 9C). The fact that Con–Agold was not seen within the cytoplasm under any of these conditions clearly indicates that the probe did not enter the spicule cavity through artifactually produced holes that might have formed during fixation and demineralization.

Spicule formation by primary mesenchyme cells of the sea urchin embryo cultured in vitro is a useful model system for studying biomineralization. This system has been utilized to study the possible involvement of glycoproteins and collagen-like molecules in spicule formation (Mintz & Lennarz, 1982; Mintz et al. 1981) and to study a cell surface antigen involved in the Ca²⁺ uptake process (Carson et al. 1985; Farach et al. 1987). Further studies of the molecular events in spicule formation require a better understanding of the structural relationships between the primary mesenchyme cells, the syncytium and the spicule. In the current study, we utilized a variety of microscopy techniques to elucidate several features of the system.

One important initial finding was that primary mesenchyme cells cultured in vitro exhibit a number of characteristics common to primary mesenchyme cells within the blastocoel of the embryo. When cultured in vitro these cells migrate and fuse (Krishnamoorthy & Solursh, 1984; Karp & Solursh, 1985) and establish cytoplasmic bridges that directly connect the individual cell bodies to the filopodia-derived sheath in which the spicule elongates. Extracellular filaments found in vitro appear to interconnect the exterior surfaces of the spicule sheath, filopodia and lamellipodia. Additionally, a filamentous matrix appears to be associated with sites of cell-substratum adhesion. Several previous studies have shown that similar filaments are associated with the exterior surfaces of the primary mesenchyme cells and the basal lamina in the intact embryo (Galileo & Morrill, 1985; Gibbons et al. 1969; Akasaka, Amemiya & Terayama, 1980). Based on the cross periodicities of filaments associated with these cells in the blastocoel, it has been suggested that some of these-filaments may be collagen (Crise-Benson & Benson, 1979). Other extracellular matrix components also appear to be associated with the primary mesenchyme cells within the embryo (Wessel, Marchase & McClay, 1980; Spiegel & Burger, 1982).

Some workers have speculated that spicules deposited by primary mesenchyme cells in vivo are exposed to the external environment of the blastocoel (Kingsley, 1984). However, based on the observation of electron-dense material found in perinuclear coated vesicles and in large cytoplasmic vacuoles, Gibbons et al. (1969) suggested that spiculogenesis is an intracellular process. To clarify this issue, we have used primary mesenchyme cells cultured in vitro. The results of this study indicate that the mineral phase of the CaCO₃-rich spicule can be extracted from the living cells with EGTA without loss of the intracellular inclusions that resemble the spicule. To determine whether intracellular inclusions contained Ca²⁺ in primary mesenchyme cells, X-ray microprobe analysis was used after first preserving these cells by rapid freezing, freeze substitution and anhydrous fixation in OsO₄ (Kowarski et al. 1979; Somlyo et al. 1981). These experiments revealed that Ca²⁺ and other elemental ions found in the spicule were present in electron-dense, perinuclear deposits. The presence of small dense deposits in the rough endoplasmic reticulum and massive accumulations in the vicinity of the Golgi complex suggest that these organelles sequester Ca²⁺ during spiculogenesis.

Since calcification processes, including bone formation in higher animals (Veis, 1984), involve secretion of glycoproteins, proteoglycans and collagen, it may be that the organelles comprising the secretory pathway are involved in delivery of both the macromolecules and the Ca²⁺ required for mineralization. In fact, it has been established that sea urchin spicules contain glycoproteins that may function in the binding of Ca²⁺ (Benson et al. 1986). This idea is also consistent with earlier observations that an inhibitor of glycoprotein biosynthesis, tunicamycin, interferes with the migration of primary mesenchyme cells (Akasaka et al. 1980) and blocks spiculogenesis in embryos (Schneider, Nguyen & Lennarz, 1978) and in cell cultures (Mintz et al. 1981). Our finding of coated pits within the limiting membrane of the spicule cavity also supports the idea that the secretory pathway is involved in spiculogenesis. This observation agrees with the earlier proposal (Gibbons et al. 1969) that coated vesicles may carry Ca²⁺ to the growing spicule.

The observations that the spicule was readily dissolved by addition of Ca²⁺ chelator or by lowering the pH of the sea water suggested the possibility that the spicule was assembled in an extracellular compartment. TEM analysis of demineralized spicules provided evidence that is consistent with this idea. This evidence included visualization of small openings in the membrane sheath that forms the spicule cavity. TEM also revealed occasional structures in the cavity that had the appearance of organic matrix. It seems likely that this matrix is more abundant in the intact spicule and that demineralization leads to losses. Further support for the presence of an organic matrix containing glycoconjugates was obtained by staining demineralized preparations with ruthenium red and by showing that a conjugate of ConA-gold binds to material in the spicule cavity. The presence of matrix reacting positively with ruthenium red indicates the presence of acidic glycoconjugates. Deposition of ConA-gold in the cavity indicates the presence of polymannose-type glycoproteins, which is consistent with the finding that such glycoproteins can be recovered from isolated spicules (Benson et al. 1986). The ability to detect ConA-gold in the spicule cavity of demineralized preparations also provides support for the presence of openings to the external environment, because in the absence of such fenestrae the entry of this membrane-impermeant probe could not occur.

The results of this study have revealed many structural features in the cellular system for assembly of spicules in vitro that exist inside the embryo. Additionally, our results suggest that the overall process of deposition of CaCO₃ and glycoconjugates in the spicule occurs via a secretory pathway that delivers these components to the spicule cavity. Further kinetic studies will test the validity of this postulated pathway for the biomineralization process.

We are indebted to Dr Maria Valdizan, Mr Dana Earles and Mrs Helen Park for their invaluable assistance. Dr Andrew P. Somlyo generously performed the electronprobe analysis. Critical comments on the preliminary version of this manuscript, provided by Drs Daniel Carson, Mary C. Farach, Norka Ruiz-Bravo, Barry Shur, Stephen Grant and Mss Judy Roe and Leanne Brooks-Scott, are greatly appreciated. Ms Diana Welch is acknowledged for her expert editorial assistance. This work was supported by a National Institutes of Health grant (HD 21483) to W.J.L. Dr William J. Lennarz, who is a Robert A. Welch Professor of Chemistry, gratefully acknowledges the Robert A. Welch Foundation.