Biosilicification of loricate choanoflagellate: organic composition of the nanotubular siliceous costal strips of Stephanoeca diplocostata

One of the major innovative steps in the evolution of uni- and multicellular animals was the acquisition of a hard, mineralized skeleton. The development of skeletal elements facilitated an increase in size of the organisms – a phyletic trend that is known in metazoans as Cope's rule (Nicol, 1966). As changes in body size affect almost every aspect of life (Schmidt-Nielsen, 1984), two strategies have been developed in animals to circumvent any constraints arising from body size increase. First, the acquisition of a hydrostatic skeleton and, second, the development of rigid solid skeletal elements (Biewener, 2005). Exclusively, the formation of inorganic structures in uni- and multicellular organisms is guided by organic templates (Lowenstam and Weiner, 1989). Those template-induced or controlled mineralization processes have been termed biomineralization. The ubiquitous occurrence of template-induced biomineralization processes in nature became obvious with the discovery that even the formation of the polymetallic nodules and crusts of the deep-sea is initiated and directed by biogenic templates (Wang and Müller, 2009). In 1924, Schmidt (Schmidt, 1924), who was the first scientist to compile template-caused/controlled biomineralization processes (Weiner and Dove, 2003), highlighted the importance of an inorganic skeleton in the establishment of a body plan. Two concepts of biomineralization were categorized by Weiner and Dove (Weiner and Dove, 2003), based on earlier systematic studies (Lowenstam and Weiner, 1989). They distinguished between biologically induced mineralization, whereby biological structures act as causative agents for nucleation and subsequent growth of biominerals, and biologically controlled mineralization, a process during which cells/organisms direct both the nucleation/growth and final location of the minerals within an organism. The composite biominerals can be deposited extracellularly, as in Foraminifera or in shells of mollusks, intercellularly as in some calcareous algae, or intracellularly, as in bacteria (magnetosome formation), plants or animals (reviewed in Weiner and Dove, 2003). Besides calcium-based skeletons, silica-based skeletal systems arose during the early evolution of uni- and multicellular eukaryotes in the Precambrian (Proterozoic), more than 542 million years ago (Müller et al., 2007).

Focusing on silica, the major taxa that use this monomeric inorganic molecule to form solid skeletons through controlled silica deposition processes are some protozoans, diatoms, choanoflagellates and silicoflagellates (Leadbeater and Jones, 1984), and metazoans, with the siliceous sponges (phylum Porifera) as the most prominent representative, as well as higher plants (see Müller et al., 2003; Perry, 2003). All of those organisms take up silica into their cells as monomeric silicic acid in order to polymerize/polycondensate amorphous and hydrated bio-silica. It is amazing that those organisms are able to deposit almost pure, amorphous quartz glass at ambient, physiological conditions from monomeric silicate (see Perry, 2003). There are two mechanisms by which bio-silica is formed, first by oversaturation and second by enzymatic synthesis. Oversaturation of mono-/oligo-silica results in polycondensation at concentrations above 100 mmol l^–1 at neutral/physiological pH and body temperature (Iler, 1979; Benning et al., 2005). The rate of aggregation/polycondensation from the monomeric silica increases with temperature at an activation energy of approximately 10 kcal/mol (Iler, 1979), a value that is about half of the average activation energy required for the breaking of an average covalent bond (Porter et al., 2009). In the second strategy, the silica-depositing organisms use an enzyme, silicatein, to lower the activation energy required for lower silica concentrations to deposit monomers and oligomers to polymerize amorphous silica (Cha et al., 1999; Krasko et al., 2000). This enzyme shows an affinity constant (Km value) to the mono-/oligomeric substrate of approximately 50 μmol l^–1 (Müller et al., 2008b), allowing the polymerization to proceed at environmental concentrations that in the sea amount to approximately 5 μmol l^–1 (Maldonado et al., 1999). It can be postulated that bio-silica deposition, irrespective of its way of formation, non-enzymatically or enzymatically, is facilitated if the guiding organic template remains surrounded by the inorganic polymer formed. This assumption stems from the observations that, at interfaces between two phases, sudden and considerable changes of the apparent activation energies occur (Wynn-Williams, 1976; Ben-Shooshan et al., 2002). Until now, only from the spicules of the siliceous sponges has a protein-bio-silica hybrid been described (Müller et al., 2008a; Müller et al., 2008c). In sponge spicules, this hybrid composition provides them with unusually high toughness combined with extreme flexibility, a feature that pushed sponge bio-silica to the forefront of material sciences (Mayer, 2005) (reviewed in Schröder et al., 2008).

In the present study, the inorganic, siliceous skeletal framework of choanoflagellates was studied with the aim of clarifying whether their siliceous structures are composed of an inorganic [siliceous]: organic [proteinaceous] composite, as well. Choanoflagellates [phylum Choanozoa (Shalchian-Tabrizi et al., 2008)] are globally distributed free-living unicellular or colonial flagellate eukaryotes living in marine and freshwater environments (Thomsen and Larsen, 1992; Buck and Garrison, 1988). The choanoflagellates are subdivided into three families (based on the composition and existence and/or structure of the extracellular matrix, the periplast) the: Codonosigidae (lacking any periplast), Salpingoecidae (encased/coated into a firm theca composed of organic material), and Acanthoecidae (captured and protected by a basket-like lorica). The lorica comprises siliceous ribs or ‘costae’, termed costal strips. After being taken up by the unicellular eukaryotes, the ortho-silicate is deposited in the fibril-shaped siliceous strips, with a diameter of approximately 100 nm (Leadbeater, 1989; Leadbeater et al., 2008). Initially, the siliceous strips are formed intracellularly in membrane-sealed vesicles that are located within the peripheral cytoplasm (Leadbeater, 1989). Those strip-containing vesicles are always associated with the membranes of the Golgi apparatus (Arndt et al., 2000). Once the strips are developed, they are released from the cells, stored at first in the top of the collar, from where they are taken for the assembly of the lorica; this organelle comprises a two-layer arrangement and is pieced together within a few minutes while the cells undergo a rotational movement (Buck and Garrison, 1988; Leadbeater, 1979a; Leadbeater, 1979b; Leadbeater et al., 2008; Leadbeater et al., 2009). Initial observations suggest that a connection between the strips exists that might contain some kind of organic connective material (Mann and Williams, 1983).

Until now, no biochemical evidence has been presented that could indicate that organic substances are located within the different zones of the siliceous cylinder of costal strips. To test this issue, we used the choanoflagellate Stephanoeca diplocostata as a model organism. The unicellular S. diplocostata are free-living flagellates that are common in coastal water. They take up free silica from the environment and use it as a building block for the synthesis of the lorica; their culture conditions are well defined, especially with respect to the formation of the lorica (Leadbeater, 1979a; Leadbeater, 1979b; Leadbeater and Davies, 1984; Leadbeater and Jones, 1984; Leadbeater, 1985; Leadbeater, 1989; Leadbeater et al., 2009). In the present study, we outline the ultrastructure of the siliceous strips and demonstrate the existence of organic components within the strips. The respective proteins were isolated from costal strips and their distribution within those strips was mapped using the immunogold labeling technique. Moreover, in vitro silica precipitation experiments were performed that led to the conclusion that the organic components exert a silica-inductive effect, resulting in bio-silica polycondensation.

The choanoflagellate Stephanoeca diplocostata Ellis was obtained from the ATCC (ATCC50456; Manassas; VA, USA). The following materials were purchased: Percoll, polyvinylpyrrolidone (M_r 360,000), biotinylated wheat germ agglutinin, alkaline phosphatase conjugated avidin, goat-anti-mouse serum coupled to 5 nm gold particles, goat-anti-mouse IgG-conjugated alkaline phosphatase and tetraethylorthosilicate from Sigma-Aldrich (Taufkirchen or Steinheim; Germany); blocking reagent and BCIP/NBT (5-bromo-4-chloro-3-indolyl-phosphate/4-nitro-blue tetrazolium chloride) were from Roche (Mannheim, Germany); peptone and yeast extract were from Roth (Karlsruhe; Germany). Culture flasks were obtained from Greiner Bio-one (Frickenhausen, Germany). Artificial seawater was prepared according to the recipe described by Harrison and colleagues (Harrison et al., 1980) and autoclaved.

The choanoflagellate S. diplocostata cells were grown in an artificial-seawater-based medium, supplemented with bacteria (Klebsiella pneumoniae subsp. pneumoniae), at 16°C as described (Harrison et al., 1980; Leadbeater and Davies, 1984; Leadbeater and Jones, 1984; Leadbeater, 1985). The medium for this heterotrophic culture comprised 800 mg/l peptone, 400 mg/l yeast extract, 1% wheatgrass extract and enrichment solution (55.3 mg Na₂EDTA.2H₂O, 466.7 mg NaNO₃, 300 mg Na₂SiO₃.9H₂O, 66.7 mg β-glycerophosphate Na₂ salt, 38 mg H₃BO₃, 0.14 mg CoCl₂.6H₂O, 1.5 mg ferric citrate, 5.4 mg MnSO₄.4H₂O, 0.73 mg ZnSO₄.7H₂O and 100 ml H₂O; pH 7.4). The wheatgrass extract was prepared as proposed by the ATCC, with some modifications. Ten grams of wheatgrass (Bienenschwarmmm; Emmerich, Germany) were boiled in 1 l artificial seawater for 5 min; after cooling, the extracts were passed through a 0.2 μm filter (Nalgene; Rochester, NY, USA).

Cultures were grown for 14 days in T-500 flasks (Greiner Bio-one; Frickenhausen, Germany) and then centrifuged at 2000 g (15 min) to collect the cell pellets. Percoll discontinuous density centrifugation was used (Gong et al., 2008a) to separate the S. diplocostata cells from the bacteria. In detail, decreasing concentrations of Percoll, 60%–40%–15%, were layered on top of each other in a 15 ml tube. After centrifugation (2000 g; 16°C; 40 min), the interface between the 40% and 60% layers was collected by aspiration. The cells were washed twice in artificial seawater by centrifugation (2000 g; 10 min) and then stored as a pellet at –20°C until the isolation of the siliceous costal strips.

For the isolation of the costal strips, a total of approximately 4 g (wet mass) of choanoflagellates was broken up by osmotic shock in distilled water. The cell particles were collected (2000 g; 15 min) and washed twice in distilled water. The resulting pellets containing most siliceous costal strips and some possible cell debris were cleaned by 4% NaDodSO₄ water solution for 24 h, or washed by concentrated sulfuric acid and nitric acid (v/v, 4:1) for 1 h to remove possible organic substances mixed in the isolated costal strip (Schmid, 1980). The pellets thus obtained were thoroughly washed by water and kept at 4°C for further experiments.

Two protocols were used for demineralization of the costal strips (‘costal strip’ fraction). First, the aliquot (ca. 200 mg) was treated with buffered fluoric acid [hydrofluoric acid (HF)/0.5 M ammonium fluoride (NH₄F); pH 5], as described by Shimizu and colleagues (Shimizu et al., 1998). By this procedure, a complete demineralization of the bio-silica structure could be expected. The exposure time to buffered fluoric acid was 20 min at 24°C. Subsequently, the material was transferred into dialysis tubes (Spectrum Laboratories; CA, USA) with a molecular mass cut-off of 1000 and then dialyzed against distilled water to remove fluoric acid. Finally, the sample was dialyzed twice against phosphate-buffered saline. The solid HF-insoluble material (ca. 20 mg) was collected by centrifugation (2400 g; 10 min). Separately, the ‘costal strip’ fraction was resuspended in 1% SDS (200 μl) and incubated for 30 min at 95°C. After centrifugation, the resulting supernatant including the total proteins from costal strips was collected and stored at –20°C. The pellets were thoroughly rinsed with water. The samples were used for EDX-analysis.

A second protocol was applied to partially demineralize the costal strips. The ‘costal strip-fraction’ was transferred to the 0.2 mol l^–1 NaOH/Na₂PO₄ buffer (pH 10.0 or 12.0), and the costal strips were treated for 2 days at 24°C; these conditions are known to partially dissolve amorphous silica (van Dokkum et al., 2004). After treatment, those partially demineralized strips were collected by centrifugation (2400 g; 10 min), suspended in 400 μl distilled water and then used in the silica precipitation assays.

For transmission electron microscopy (TEM) observation, samples including the intact siliceous costal strips as well as the partially to completely demineralized strips were dehydrated in an ascending alcohol series (incubated for 5-min periods in ethanol: 30%, 50%, 70%, 80%, 96% and, finally, twice in 100% ethanol). Then, one drop of the suspension was placed onto a formvar carbon-coated copper grid and left to dry on filter paper in air at room temperature. Morphological studies were performed by Philips EM420. Line-scan and spot EDX were performed in scanning transmission electron microscopy (STEM) mode by a FEI Tecnai F30 equipped with an energy-dispersive X-ray spectrometer (Mugnaioli et al., 2009).

Analysis by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE), using a 4–15% discontinuous polyacrylamide gel, was performed as described previously (Laemmli, 1970). After running, the gels were stained with Coomassie brilliant blue.

The proteins were electrophoretically transferred to polyvinyldifluoride membranes (PVDF; Millipore, MA; USA) using a Trans-Blot SD Semi-dry Transfer Cell (Bio-Rad; Hercules CA, USA). The membranes were blocked with 5% skimmed milk and incubated with primary antibody [diluted 1:2000 fold in 3% BSA (bovine serum albumin), PBS (phosphate buffered saline)] for 2 h. The membranes were washed with PBST (PBS, supplemented with 0.05% Tween 20) and incubated with goat-anti-mouse IgG-conjugated alkaline phosphatase (Sigma-Aldrich, Steinheim; Germany); detection was carried out using BCIP/NBT (5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium chloride) solution (Roche) (Gong et al., 2008b).

After size separation by SDS-PAGE, the proteins were transferred onto PVDF and blocked in 2% (w/v) polyvinylpyrrolidone-360 in 50 mmol l^–1 Tris-HCl buffer (pH 7.5; 0.5 mol l^–1 NaCl) overnight at 4°C. The membranes were transferred into a 2 μg ml^–1 biotinylated wheat germ agglutinin (WGA) solution (Tritium vulgaris) and then incubated for 2 h at 24°C. Finally, the membranes were rinsed three times (10 min each) in 50 mmol l^–1 Tris-HCl (pH 7.5; 0.5 mol l^–1 NaCl, 0.1% Triton X-100 [v/v]), followed by an incubation with alkaline-phosphatase-conjugated avidin at a dilution of 5×10^–3 for 1 h at 24°C. The lectin-biotin complexes were visualized with the phosphatase substrate BCIP/NBT (Bédouet et al., 2001).

Polyclonal antibodies were raised in one BALB/C mouse against the protein band with the apparent molecular mass of 14 kDa obtained by loading SDS-PAGE gels with costal strip proteins. The antiserum, termed Poly-anti-S14, was achieved after three rounds of immunization during a 1.5 month period and the titer (1:10,000) was measured by ELISA assay (Mayer and Walker, 1987). Poly-anti-S14 was diluted 2000 fold and then applied for immunogold labeling. Three different samples of costal strips were analyzed. First, freshly isolated costal strips; second, costal strips incubated in water for 40 days; third, costal strips treated with alkaline solution (0.2 M NaOH/Na₂PO₄; pH 12.0) for 2 days.

Before labeling, the costal strips were blocked in 1% western blocking reagent (Roche, Mannheim, Germany) in 3% BSA/PBS solution overnight at 4°C and subsequently incubated with either Poly-anti-S14 or with preimmune serum, obtained from the same animal used for immunization, for 2 h at room temperature. The samples were then incubated with a 20-fold diluted goat-anti-mouse serum coupled to 5 nm gold particles. After each step of incubation, the samples were washed three times with PBST for 5 min. Finally, the samples were fixed in 2.5% glutaraldehyde for 15 min, washed extensively with distilled water and dehydrated in the ascending alcohol series and then inspected by Philips EM420 (Gong et al., 2008b).

A solution of orthosilicic acid was freshly prepared by dissolving tetraethylorthosilicate (TEOS) in 1 mmol l^–1 HCl to a final concentration of 0.9 mol l^–1. The silica precipitation assay was performed in a final volume of 190 μl and composed of one of the two buffers, either of a 100 mmol l^–1 sodium acetate buffer (HAc-NaAc) [acetic acid (HAc), buffered with 1 mol l^–1 NaOH to pH 5.0, 5.5, 6.0 or 6.5] or 100 mmol l^–1 Hepes buffer (buffered with NaOH to pH 7.0 and 7.5). Routinely, the assays were performed at pH 6.0, a value that had been found to promote optimally the reaction. The samples (10 μl), containing either untreated strips or partially demineralized ones, were suspended in water. 10 μl of the suspended samples were transferred into the 190 μl buffer and then supplemented with 10 μl of orthosilicic acid/TEOS solution. The final concentration of silicic acid was adjusted to 40 mmol l^–1 or 9 mmol l^–1, respectively. The periods of incubation (24°C) were 10 min and 30 min, respectively. Next, the samples were centrifuged (12,000 g; 5 min) and the pellets were washed twice with distilled water to remove free silicic acid. The pellets thus obtained were resuspended in 200 μl of 1 mol l^–1 NaOH (95°C; 30 min) to form monomeric silica. Silicic acid concentrations were quantitatively determined in these solutions with β-silicomolybdate using the silicate colorimetric kit from Merck (Darmstadt, Germany), according to the manufacturer's instructions (Iler, 1979; Krasko et al., 2002). The OD_650nm values were determined, and the concentration of silicic acid was calculated on the basis of a linear calibration curve obtained with orthosilicic acid.

For protein quantifications, the Bradford method (Compton and Jones, 1985) (Roti-Quant solution–Roth) was used.

The cells of S. diplocostata are characterized by their basket-like lorica (Fig. 1), which surrounds one single flagellum (Fig. 1A,B). The choanoflagellates contain longitudinal as well as vertical costal strips that are woven into a basket, the lorica (Fig. 1B). To obtain sufficient starting material for the isolation of the costal strips (4 g of wet mass), cells from 5 l cultures were collected. Subsequently, the choanoflagellates were separated from the bacteria by discontinuous Percoll density gradient centrifugation. The S. diplocostata cells were collected within the 40 and 60% Percoll layers and then used for the analysis.

The lorica of S. diplocostata was separated from the protoplast by osmotic shock treatment in distilled water. An intact lorica (Fig. 2A) is built of longitudinal and transverse costal strips, with diameters of >75 nm (Fig. 2B,C); the length of the strips is approximately 3 μm. TEM observation revealed that large rod-like strips forming the base and also the longitudinal basket-like lorica function as a bracket-like support. The smaller strips (diameter <100 nm) are additionally used in the other parts of this basket, especially at the fringe and the intervening space. It appears that the links of the costal strips are reinforced by a connective-like substance (Fig. 2B,C), which can be removed by SDS.

TEM inspection showed that the costal strips have a marked texture reflecting distinct structures. Depending on their size, the costal strips can be grouped into large costal strips, with a diameter of >100 nm (Fig. 2C,D), and smaller strips, with a diameter of <100 nm (Fig. 2C,E). The larger strips appear as compact rods with a uniform texture, whereas the smaller strips show a less compact texture and display, especially in the central region, a fluffy appearance. Common to both types of costal strips is the presence of one tubular-like structure in the central, electron-lucent region of each strip. In contrast to the center region, the periphery of the strips is more electron dense (Fig. 2D,E). It is striking that especially the smaller costal strips have a wider electron-lucent central region (diameter of approximately 10 nm) than seen in larger rods (diameter of approximately 2 nm) (Fig. 2E,F). Moreover, another distinctive structure of smaller costal strips has been observed under the high-resolution of TEM. The outer edge is the most electron-lucent region, which makes it look like a capsule surrounding the surface of the strips (Fig. 3B1).

Element determination by EDX spot analysis showed that costal strips (Fig. 3A1,B1) comprise only silicon and oxygen (Fig. 3A2,B2). The copper peaks came from the grids. Furthermore, line EDX scans were used to measure silicon and oxygen concentrations across a large and a small costal strip. EDX analysis indicated that the outer capsule-like structure contains the least silicon (Fig. 3B3, indicated by arrow). As the probing continued, the silicon and the oxygen concentrations increased from the peripheral to the central region until the electron-lucent region was reached. The silicon concentration in the electron-lucent region was lowest; the oxygen levels were more uniform throughout a strip (indicated by arrowheads; Fig. 3A3, Fig. 3B3). These findings revealed that costal strips have no homogeneous composition; the two distinctive structures, an outer capsule-like structure and an electron-lucent central region, contain lower levels of silicon than the remaining regions. Furthermore, some observations led to the assumption that the central region might be filled by filament-like substances. The broken costal strips displayed a filament-like structure in the central canal (Fig. 2B). Even more supportive was the observation that a filament protruded from the central canal (Fig. 2E). Frequently, those filaments link individual costal strips together, and they can be removed by SDS. Taken together, we presume that the central silica-poorer region might not be hollow but contain organic substances, besides silica.

In order to obtain pure siliceous costal strips not associated with surface-bound organic material, 4% SDS was used for 24 h to free isolated lorica from organic materials. The obtained crescent-shaped rods with a length of approximately 3 μm were used for demineralization experiments (Fig. 4A,B).

For the analysis of the possible organic scaffold in the costal strips, the isolated strips were further cleaned by concentrated sulfuric acid and nitric acid (4:1 v/v) for 1 h. After thoroughly washing with water, the obtained costal strips were treated with buffered HF to remove the silica phase. The mass of costal strips was determined before and after HF treatment. Approximately 10% of the initial mass remained in the insoluble form after HF treatment; no protein was detected in the HF-soluble phase after dissolution of the costal strips, checked by SDS-PAGE.

As a first attempt to analyze the chemical nature of the insoluble residues after HF treatment, the ‘hydrofluoric acid-insoluble’ (HF-IS) fraction was analyzed by EDX. The spectrum obtained from the HF-IS sample showed a prominent peak for sulfur, as well as for phosphorus (Fig. 5A), supporting the conclusion that the scaffold comprises organic material. The HF-IS sample was subsequently treated with 1% SDS (95°C; 30 min), and the remaining insoluble material was subjected to EDX analysis. The spectrum obtained showed distinct peaks for sulfur and copper, whereas the peak for phosphorus was insignificant (Fig. 5B). The copper peaks came from the support grid, and calcium is believed to originate from distilled water.

Besides the method to demineralize silica skeletal elements with fluoric acid, a second protocol that was based on alkaline treatment was applied to achieve (partial) dissolution of silica (van Dokkum et al., 2004). Therefore, the SDS-cleaned ‘costal strip’ fraction was treated with pH 10.0 or 12.0 NaOH-Na₂HPO₄ buffer as described in the Materials and methods. Compared with the intact costal strips (Fig. 4B), the costal strips after 2 days of alkaline treatment lost their smooth surfaces, converted to blurry structures and released, to some extent, the tattered (organic) matrix (Fig. 4C). In addition, those flocculated membranous materials remained attached to the partially demineralized costal strips.

The costal strips after demineralized in pH 12.0 NaOH-Na₂HPO₄ were extracted with either 8 mol l^–1 urea (pH 7.0) or HAc-NaAc (pH 5.0); likewise, an equal amount of untreated costal strips was extracted in the same way. The supernatant solutions were analyzed by SDS-PAGE and the gel was stained with Coomassie brilliant blue (Fig. 6A). The gel loaded with the 8 mol l^–1 urea extract of pH 12.0 solution-treated strips showed a band with an apparent molecular mass of 14 kDa (Fig. 6A, lane c), whereas the band was absent in the extracts treated by HAc-NaAc (pH 5.0) or 1% HAc (Fig. 6A, lanes d and e). Likewise this band was also absent from 8 mol l^–1 urea extract of the untreated costal strips (Fig. 6A, lanes a and b). These results showed that the proteins only came from alkali-treated costal strips, but not from intact costal strips. 8 mol l^–1 urea can solubilize these proteins, but neutral or acid solution can not. Meanwhile, the supernatant was also examined after costal strips were treated with alkaline solution, and no protein was detected even in the concentrated sample (concentrated by 10-fold; Fig. 6A, lane f).

Further experiments were used to test the effects of alkaline solutions with different pH on the costal strips. The gel loaded with the 8 mol l^–1 urea extract from pH 12.0 alkali-treated costal strips (Fig. 6B, lane c and d) showed a denser band at 14 kDa than that the one from pH 10.0 alkali-treated costal strips (Fig. 6B, lanes a and b). This result confirmed that an alkaline solution with higher pH can release more proteins from costal strips, which correlates with the increasingly pH-dependent demineralization level of amorphous silica (van Dokkum et al., 2004). The isolated 14 kDa protein was named S14.

Wheat germ agglutinin labeling revealed highly positive staining of S14 in the probing of SDS-PAGE gels, indicating that this protein is highly glycosylated (Fig. 7, lanes a and b). Furthermore, this protein was sized-separated in a SDS-PAGE gel, and the band at 14 kDa was cut out, ground and suspended in PBS for immunization of mice to raise polyclonal antibodies. The specificity of the achieved antibodies was proved by immunoblotting assay, and the band at 14 kDa was prominently recognized (Fig. 7, lane c). A faint band at approximately 28 kDa was also recognized. It might be suggested that the band at 28 kDa represents a dimer of the 14 kDa monomer. Therefore, the antibodies, named with Poly-anti-S14, were used to map the distribution of S14 in the costal strips.

For this experiment, three kinds of costal strip preparations were used for immunogold labeling studies using the polyclonal antibodies Poly-anti-S14 (Fig. 8). First, newly isolated costal strips (Fig. 8A); second, costal strips that had been incubated in water for 40 days (Fig. 8B); and, third, alkali-treated costal strips (Fig. 8C).

Colloidal gold particles (attached to the secondary antibodies) were used for visualization of the protein S14 and detecting its distribution within the costal strips. When newly isolated costal strips were incubated with Poly-anti-S14, sparse immunocomplexes could be identified with gold labeled secondary antibodies on the surface of costal strips (Fig. 8A). The images from the negative controls (treated with pre-immune serum; Fig. 8A-pre) indicated the reliability of the labeling. By contrast, for costal strips that had been stored for 40 days in distilled water, immunogold particles could be detected in the samples treated with Poly-anti-S14 (Fig. 8B-ab), whereas those immunocomplexes were absent in controls (Fig. 8B-pre). The complexes accumulated especially on the surfaces of the strips. The extent of immunogold complexes was drastically increased when the costal strips were treated with alkali (pH 12; 2 days, 24°C). Those strips exposed on their surfaces large numbers of proteins that could be recognized by Poly-anti-S14 and visualized by colloidal gold particles (Fig. 8C-ab-1, 2 and 3). The images revealed that the colloidal gold particles were clustered on the etched surfaces of costal strips, suggesting a possible organic scaffold embedded within a silica matrix. Once the silica structure was partially dissolved, the embedded organic scaffold could be exposed, approached by the antibodies and labeled by the colloid gold particles. In the parallel control, no immunoreactions could be seen (Fig. 8C-pre), suggesting the good reliability of the positive labeling.

For the above outlined data, it can be summarized that the costal strips are traversed by a proteinaceous scaffold that is exposed after alkali treatment. In order to test whether an organic scaffold can accelerate the silica deposition, partially demineralized strips were incubated with orthosilicate. The strips were treated for 4 days in a buffer of pH 12 (see Materials and methods), a period during which the majority of the strips lost their silica (Fig. 9). Those samples were incubated with orthosilicate (9 mmol l^–1 and 40 mmol l^–1). After incubation, the newly polycondensated silica, as well as the residual siliceous matrix from the strips, was collected by centrifugation. Subsequently, the poly(silicate) was quantified by using the β-silicomolybdate assay (after solubilization with NaOH). The experiments revealed that, in assays without additional orthosilicate, the level of poly(silicate) in the partially demineralized strips was low and amounts to 0.025 OD_650nm units, whereas the concentration of poly(silicate) was – as expected – in the intact strips much higher (0.22 OD_650nm units) (Fig. 9). The concentration of poly(silicate) in the assays with partially demineralized strips strongly increased concentration dependently from 0.12 OD_650nm units (9 mmol l^–1 orthosilicate) to 0.4 OD_650nm units in the presence of 40 mmol l^–1 orthosilicate. In parallel, the concentration of poly(silicate) did not change significantly in assays containing non-treated strips, and the values remained at OD_650nm units of 0.24 to 0.26. Moreover, the reaction of poly(silicate) deposition at partially demineralized costal strips was pH dependent. The maximal newly formed poly(silicate) was seen in reactions at a pH value of 6.0–6.5 (Fig. 10).

The newly formed poly(silicate) in assays comprising partially demineralized strips could be visualized by TEM inspection. In the assays supplemented with 9 mmol l^–1 orthosilicate, the organic lobate scaffold became densely associated with strings of beads, which we interpreted as silica particles (Fig. 11B). In control assays lacking orthosilicate, none of those structures is seen (Fig. 11A).

It is a challenge for the future to disclose the organic scaffolds in biominerals at the chemical/biochemical level and to understand their roles in the establishment of the individual morphologies of the inorganic matrixes, which are usually highly elaborated and complex. Hence, a first task is to identify the organic scaffold in any kind of inorganic skeletons in organisms. For a few organisms, the structure and the biochemical properties of the organic scaffold have been thoroughly studied. From the two taxa the diatoms and the siliceous sponges, the importance of the organic templates for the formation of their biomineralic skeletons has been well elucidated. The ornate and elaborated diatom frustules are formed onto organic templates (Kröger et al., 1994; Kröger et al., 1997; Kröger et al., 1999; Sumper and Brunner, 2008; Hildebrand et al., 2008), the biosilica-associated peptides (silaffins) and the long-chain polyamines (see Poulsen et al., 2003). Furthermore, it was found that those organic molecules accelerate silica formation from a silicic acid solution in vitro. In addition, it had been speculated that the silica deposition vesicles contain a matrix of organic macromolecules that not only regulate silica formation but also act as templates to mediate biosilica nanopatterning (Robinson and Sullivan, 1987; Kröger et al., 1999). The understanding of the formation of spicules from siliceous sponges has been considerably pushed ahead by the finding that the enzyme silicatein mediates the formation of the poly(silicate) (Cha et al., 1999; Krasko et al., 2000; Müller et al., 2008b) and also by the experimentally based finding that this protein remains entrapped in the poly(silicate) product (Müller et al., 2008a; Müller et al., 2009). Hence, the sponge siliceous spicules represent intriguing and fascinating model systems allowing investigations and understanding of biomineral formation from the genomic level to morphological realization.

A third taxon of organisms forming a silica skeleton, termed silica lorica, are the choanoflagellates. While the genome structure and organization of those protozoans are well established (Carr et al., 2008; Abedin and King, 2008; King et al., 2008), the nature of the organic scaffold for the formation of their siliceous structural elements is not well understood. Based on the pioneering studies of Mann and Williams (Mann and Williams, 1983) and the detailed structural and biomechanical investigations of Leadbeater (Leadbeater, 1979a; Leadbeater, 1979b), it could be postulated that the siliceous loricae with their costal strips are formed along an interior or exterior organic scaffold. Otherwise the morphology and functional interactions of their silica structures cannot be understood. Silica is taken up by the choanoflagellates from the aqueous environment very likely by an energy-consuming pumping system and then stored in special organelles, membrane-enveloped silica-deposition vesicles (SDVs) (Leadbeater, 1979a; Mann and Williams, 1983; Leadbeater 1989; Preisig, 1994). Within the cells, the SDVs are closely associated with the Golgi apparatus and the endoplasmic reticulum, suggesting that there silica formation around the costal strips proceeds in an intimate neighborhood of a protein-processing machinery. However, before the present study, only first suggestions (Mann and Williams, 1983) but no clear evidence has been documented that the process of strip formation and, in turn, their silicification require the presence of a proteinaceous scaffold.

In a first series of experiments, the morphology of the costal strips is described. We have classified these skeletal elements, building the lorica, into the group of large costal strips (diameter >100 nm) and small strips (<100 nm), according to their sizes. Both classes of strips disclose a central, electron-lucent canal. It is striking that the diameters of these central canals are wider in the smaller strips (diameter approximately 10 nm) than in the larger strips (diameter approximately 2 nm). At present, we postulate that the canals in the primordial strips are wider and become smaller during the ageing and maturation process. This process is difficult to study experimentally as the maturation of the strips proceeds rapidly in SDVs in the short period of approximately 12 min (Leadbeater, 1979a; Leadbeater, 1979b). Cross-sections through the costal strips, as well as breaking studies, supported electron microscopically that the central cylinder of the strips is made by a material, different from silica, that is organic in nature. The TEM images showed that, from the center of the strips, filament-like structures protrude, which might suggest that those organic structures might be involved in the longitudinal organization of the strips. The subsequent EDX analyses, line scanning, supported this assumption by the finding that the center of the strips is poorer in silica. Again, it should be highlighted that the difference in the chemical composition, a decrease/lowering of the silica concentration in the center of the strips, is much more pronounced in the smaller strips, compared with the larger ones.

Another distinctive structure of the smaller costal strips is the capsule-like outer edge, which contains least silica and therefore is distinguished from the inner region – although this structure is not significant in the large costal strips. It can be proposed that the capsule-like edge is the growing front of costal strips, followed by a scaffold that has been prebuilt on the surface of costal strips. This scaffold is supposed to support the precipitation of silica particles. With time, additional silica is deposited onto those surfaces, allowing a growth of the costal strips and also a reinforcement of the silica materials due to an increased density of this inorganic polymer. Gentle treatment with alkali (pH 10.0 and 12.0) had been shown to dissolve the outer edge of the strips and resulted in the release of a fluffy matrix. This observation can be taken as further evidence that a possible organic scaffold exists within the costal strips. Moreover, our EDX analysis revealed that also other elements, such as sulfur and phosphorus, exist within the residual materials of siliceous strips after complete demineralization. We also show that, after treatment of the strips with 1% SDS and by heat, these elements, including also phosphorous, are absent. By analogy to the silaffins from diatoms (Kröger et al., 2002), we assume that the organic matrix within the costal strips might be phosphorylated.

SDS-PAGE analyses revealed that the organic scaffold is composed of proteins (positive for Coomassie brilliant blue); and the amount of released proteins relates to the extent of demineralization. It is known that amorphous silica is stable in water but increasingly dissolves as pH increases (van Dokkum et al., 2004). Our SDS-PAGE gels showed that few proteins were present after water treatment. More proteins were released after treatment with pH 12.0 alkali than following a pH 10.0 alkali treatment. The released proteins are naturally in an insoluble form. Urea or SDS can effectively dissolve them, but neutral or acidic solutions cannot. As these proteins were specifically detected after silica had been dissolved, we are convinced that these proteins originated from costal strips but not from the surrounding impurities. Further blotting studies and subsequent use of WGA agglutinin as a probe revealed that a glycoprotein S14 with an apparent molecular mass of 14 kDa was dominant among the released proteins. The specificity of this plant lectin had been attributed to the sugar moieties NeuNAc as well as GlcNAc (Kronis and Carver, 1982). Glycoproteins comprising those sugar residues occur frequently, from plants to metazoa (Karpati et al., 1999). However, this finding is noteworthy, especially in view of earlier findings that revealed that choanoflagellates (King et al., 2003) are rich in genes coding for sugar-binding adhesion molecules – lectins and adhesion receptor tyrosine kinases, of unknown function. These two pieces of data might suggest that those sugar-recognizing proteins in choanoflagellates are involved in the formation of their skeletal element. This assumption is based on the fact that, in both classes of siliceous sponges, the Demospongiae and the Hexactinellida, the formation of the extracellular spicule is mediated by silicatein, which is spread in a galectin cylinder laterally around the growing spicules (Schröder et al., 2006; Wang et al., 2009). Future studies must show whether in the costal strips of S. diplocostata a galectin exists that can recognize the 14 kDa dominant glycoprotein in the strips.

This glycoprotein S14 was separated in SDS-PAGE gels and prepared for immunization. The achieved polyclonal antibodies were specific to recognize S14 as proven by immunoblotting. Using them as a tool, it could be demonstrated that the organic scaffold of the strips became decorated after the rods had been (partially) demineralized, whereas the intact costal strips bound with few immuno-gold particles. These results confirmed that the capsule-like structure of costal strips is a protein-silica hybrid, and the protein S14 is embedded in this structure. Once silica was dissolved, protein S14 could be detected by the antibodies. First sequencing studies by MALDI technique (N.G. and W.E.G.M., unpublished observations) revealed peptides that, however, did not match with any protein sequence deduced from the choanoflagellate Monosiga brevicollis (King et al., 2008). Our electron-microscopic studies at the 5 to 20 nm scale did not result in any indication that the organic scaffold, released from the strips, originated exclusively from the central canal. It appears more likely that the organic fibers and aggregating proteins were released from the entire siliceous strip.

In the last series of experiments it was tested whether the scaffold within the costal strips accelerates/facilitates silica deposition. To test this issue, the costal strips were partially de-mineralized and subsequently exposed to orthosilicate that had been generated by hydrolysis from TEOS (Matsuoka et al., 2000). The assays to determine the extent of silica deposition were performed at pH 6.0/6.5, which had been found to be optimal for this reaction. The data revealed that the deposition of silica on the proteinaceous scaffold increased concentration-dependently; and the first measurable silica formation could be detected with the 9 mmol l^–1 orthosilicate in the β-silicomolybdate-based optical test, under the incubation conditions used. Electron-microscopic inspections revealed that the newly formed poly(silicate) in vitro in the presence of the scaffold from the strips becomes associated with the organic template in a string-like arrangement of the silica nanoparticles.

Taken together, the data presented in this report demonstrate that the costal strips are formed of hybrid silica, comprising a silica matrix and a proteinaceous scaffold. The structural details of the costal strips revealed that this biomineral is not a single silica rod, but has a more complex construction. The formation and the development of the costal strips are schematically outlined in Fig. 12. It is sketched that, in the center of the strips, a canal exists that is filled with an organic filament. This axis organizes the longitudinal growth of the strips and also facilitates the appositional growth of the strips with silica. During the lateral growth of the strips, the protein molecules become associated with the central organic filament and form a scaffold that facilitates the deposition of silica. The data gathered might indicate that, also during the lateral growth of the strips, the 14 kDa (glyco)protein plays a crucial role. The latter assumption arises from the fact that this protein is prevalent in the protein scaffold from costal strips.

It is clear that the cloning of the genes encoding the 14 kDa protein will be of prime importance for us. After having succeeded in preparing the recombinant protein, a wide array of approaches will open up. It can be anticipated that this 14 kDa can be used as template for the formation of silica-based, and surely also subsequently titania- and zirconia-based (Tahir et al., 2005), homogenous nanotubes and nanowires.