W. E. G. Müller, Institute for Physiological Chemistry, University Medical Center of the Johannes Gutenberg University Mainz, Duesbergweg 6, D-55128 Mainz, Germany Fax: +49 6131 39 25243 Tel: +49 6131 39 25910 E-mail: email@example.com
Biomineralization processes are characterized by controlled deposition of inorganic polymers/minerals mediated by functional groups linked to organic templates. One metazoan taxon, the siliceous sponges, has utilized these principles and even gained the ability to form these polymers/minerals by an enzymatic mechanism using silicateins. Silicateins are the dominant protein species present in the axial canal of the skeletal elements of the siliceous sponges, the spicules, where they form the axial filament. Silicateins also represent a major part of the organic components of the silica lamellae, which are cylindrically arranged around the axial canal. With the demosponge Suberites domuncula as a model, quantitative enzymatic studies revealed that both the native and the recombinant enzyme display in vitro the same biosilica-forming activity as the enzyme involved in spicule formation in vivo. Monomeric silicatein molecules assemble into filaments via fractal intermediates, which are stabilized by the silicatein-interacting protein silintaphin-1. Besides the silicateins, a silica-degrading enzyme silicase acting as a catabolic enzyme has been identified. Growth of spicules proceeds in vivo in two directions: first, by axial growth, a process that is controlled by evagination of cell protrusions and mediated by the axial filament-associated silicateins; and second, by appositional growth, which is driven by the extraspicular silicateins, a process that provides the spicules with their final size and morphology. This radial layer-by-layer accretion is directed by organic cylinders that are formed around the growing spicule and consist of galectin and silicatein. The cellular interplay that controls the morphogenetic processes during spiculogenesis is outlined.
Gray  reported for the first time that the spicules of some sponges, the bone-like skeletal elements (‘glassy needles, irregularly but closely put together lengthways’; Fig. 1D, insert) of the ‘zoophyte’Gorgonia briareus , are composed of silica. Gray had already stated that, in the animal kingdom, silica provides an inorganic support only in sponges (phylum Porifera), where it does so in an almost pure form. In 1885, the siliceous sponges were split into the two classes Hexactinellida and Demospongiae . The spicules (skeletal elements) of both classes of sponges comprise a central axial canal around which a silica sheath is deposited. The diameter of the axial filament residing in the axial canal is in the range of 1 μm, and the length varies with the spicule size. The larger megascleres have lengths of > 10 μm, whereas the smaller spicules, the microscleres, have lengths of only < 10 μm . The silica sheath around the axial canal has, in the phylogenetically oldest taxon , the Hexactinellida, a lamellar organization (Fig. 1F,G), whereas in demosponges it constitutes a solid rod (Fig. 1B,C; reviewed in ). The inorganic silica is made of glassy amorphous hydrated silica , similar to opal. In order to distinguish the organism-derived silica from the amorphous opal that is formed by precipitation of silica from silicate-containing solutions or colloidal suspensions without participation of any organic component, the biologically formed silica has been termed biosilica. It is established for over 100 years that the spicules are synthesized in specialized cells, the sclerocytes  (Fig. 1D). Spicule formation starts intracellularly, and is completed, after extrusion of the spicules into the extracellular space [8,9], in the mesohyl.
A breakthrough in the understanding of spicule formation came with the discovery that the axial filaments in the spicules of demosponges  and hexactinellids  contain an enzyme. This enzyme, which has been termed silicatein, catalyzes biosilica formation during axial and radial growth. In contrast to plant phytoliths and diatom frustules, where biosilica is deposited from a supersaturated solution onto organic templates, resulting in intricate 3D nanostructures and microstructures , the siliceous spicules in sponges are formed from a hyposaturated intraorganism environment in an enzymatic way. This property allows the sponges to form biosilica in the aquatic environment, which is poor in silicate. The present-day oceans, where most sponges live, contain only ∼ 5 μm silicate . After this discovery, it was determined that silicatein is the key enzyme/molecule in siliceous sponges that mediates silica deposition [reviewed in 14–16] in vitro [17,18] and also in vivo  under physiological conditions.
In in vitro systems, tetraethyl orthosilicate (TEOS) and orthosilicic acid have been used as substrates for enzymatic biosilica formation; during this polycondensation reaction, ethanol is released [10,20]. Furthermore, the recombinant enzyme is also able to catalyze, in a phase-transfer reaction, the in vitro condensation of alkoxy silanes to silicones . The silicatein-mediated formation of silica proceeds at silica precursor concentrations (around 200 μm) far below those at which condensation in silicatein-free solutions occurs (1 mm and higher) . The organization of the spicule-based sponge skeleton is genetically fixed and architecturally complex. It is amazing that the up to 3-m-long glass-fiber-like giant basal spicules from Monorhaphis chuni (Fig. 1H) are able to transmit light over this distance without considerable loss of intensity (Fig. 1I) . Recent evidence indicates that the spicules also function in vivo as a light signaling system, perhaps substituting for a nervous system, which is lacking in sponges [24–26].
In the present minireview, we focus on the role of silicatein as the principal molecule/enzyme mediating biosilicification in sponges, and on the molecular biological, biochemical and functional aspects of this novel enzyme that causes the formation of the spicules. In order to obtain a comprehensive understanding of the structure-forming ability and the biocatalytic properties of this enzyme, the silicatein-interacting proteins, the silintaphins, as well as those of the silicase, the catabolically acting biosilica-degrading enzyme, are also highlighted. The purely nonenzymatic silica polycondensation reaction proceeds at higher orthosilicate substrate concentrations than the enzymatically driven biosilica polymerization. Both the chemical and physicochemical characterization of the reaction are given in the accompanying minireview article by the Perry group . The (potential) use and (bio)fabrication of both silica and biosilica are the subject of the third minireview in this series by the Tremel team .
Biosilica-forming and biosilica-metabolizing sponge enzymes and associated proteins
Silicatein – a unique biocatalyst
The silicateins were first identified in the axial filament of the demosponge Tethya aurantium  by partial protein analysis followed by molecular cloning. The silicateins comprise a family of related sequences, which consist of three isoforms, silicatein-α, silicatein-β, and silicatein-γ, in a molar ratio of 12 : 6 : 1 . In the axial filament, the molecular masses of the polypeptides have been determined (by SDS/PAGE) to 27–29 kDa for the demosponge T. aurantium  and 24–29 kDa for Suberites domuncula [9,29]. The silicateins belong to the papain-like cysteine protease superfamily, and are most closely related to the cathepsin family . The first cathepsin in sponges was identified and cloned from the demosponge Geodia cydonium .
Over 85 silicatein cDNA sequences have been deposited in the databases (reviewed in ). The silicateins undergo post-translational maturation/modification processes that are best understood for silicatein-α from S. domuncula [9,29] (Fig. 2A). The deduced size of the primary translation product of the silicatein-α transcript in this sponge has been calculated to have an Mr of 34 663. The predicted signal peptide ranges from amino acids 1–17. First by sequence comparison and then by the analytical data obtained from the silicatein of Petrosia ficiformis , the cleavage site between the propeptide (amino acids 18–112) and the mature protein chain (amino acids 113–330) has been identified as ProLeuGln↓AspTyrPro. As in cathepsins, the intramolecular disulfide bridges in the silicateins had first been predicted and then experimentally demonstrated [29,33]. However, more importantly, the catalytic triad in the silicateins differs in one position from that of the cathepsins . In silicateins, the catalytic triad is composed of the three amino acids Ser (in S. domuncula, Ser138), His227, and Asn350, whereas in the cathepsins the Ser is replaced by Cys. Evidence has been presented that the interaction between Ser and His is essential for the catalytic mechanism of the silicatein-mediated polycondensation reaction , as discussed below. Likewise, the presence of at least five phosphorylation sites in the silicateins was initially predicted  and then proven : two putative protein kinase C phosphorylation sites (amino acids 227 and 267), and three casein kinase phosphorylation sites (amino acids 160, 230, and 232) (Fig. 2A). In contrast to the cathepsins, the silicatein-α molecules possess cluster(s) of hydroxy amino acids. In demosponges (T. aurantium and S. domuncula), this stretch comprises eight amino acids (with six Ser residues), whereas in the hexactinellid silicatein from Crateromorpha meyeri, this Ser segment is longer (amino acids 145–158) (Fig. 2A). Intriguing in the hexactinellid sequence is the existence of a ‘hexactinellid-specific’ Ser stretch, between amino acids 169 and 173, in addition to the ‘conventional’ sponge Ser cluster . These two Ser clusters frame the His moiety of the active center of the enzyme in C. meyeri. It has been proposed that these Ser clusters are involved in the binding of the orthosilicate/oligosilicate substrate/product [35,36]. Finally, it should be mentioned that, in the silicateins, one hydrophobic region exists (between amino acids 135 and 150), which is located around the first amino acid of the catalytic triad close to the site where presilicatein is coverted to mature silicatein protein . Modeling studies have revealed that both the synthetic organosilicon substrate TEOS and the natural substrate orthosilicate fit very well into the active site pocket of the silicatein-α molecule  (Fig. 2C). Model prediction has revealed that, in the prosilicatein, this hydrophobic stretch, which is adjacent to the active site of the enzyme, is covered by the propeptide (Fig. 2Ca), whereas in the mature protein this hydrophobic region is exposed, together with the active site amino acids, to the surface of the molecule (Fig. 2Cb,c). In the axial filament, only the mature silicateins can be identified (24–29 kDa), whereas the 33-kDa large prosilicatein is found in the extraspicular ‘enriched sponge extract’ . In this compartment, the appositional, radial growth of the spicules occurs, which is likewise catalyzed by silicatein. Therefore, it is likely that, in the extraspicular space, prosilicatein is processed immediately after binding to the surface of the growing spicule .
The first mechanism was proposed by Cha et al. . These authors suggested a two-step mechanism for silicatein-mediated silica formation from silicon alkoxide (TEOS) substrates. This model does not explain the core function of silicatein, the silica polymerization/polycondensation reaction. A second model  was deduced from mutational studies with a ‘chimeric’ mutant of human cathepsin L.
With a new model  (Fig. 2B), we explain both the initial condensation/disilicic acid formation and the subsequent oligomerization process. We also accept that the initial step of the catalytic reaction involves a nucleophilic attack (SN2 type) of the (electronegative) oxygen atom of the hydroxy group of the active site Ser to the (electropositive) silicon atom of a silicic acid molecule, a process that occurs in the catalytic pocket of the silicatein molecule. This reaction is presumably facilitated by hydrogen bridge formation between the Ser OH and the His of the catalytic center, which increases the nucleophilicity of the Ser hydroxy group (Fig. 2B, Step 1). Next, a proton transfer from the His nitrogen (Ser·His hydrogen bond) to one of the silicon OH ligands of the pentavalent intermediate (transition stage) is formed, resulting in the release of a water molecule and ending in covalent binding of the enzyme Ser to the silicic acid. Subsequently, the Ser-bound silicic acid molecule undergoes a nucleophilic attack on the silicon atom of a second orthosilicic acid molecule, entering the substrate pocket of the enzyme (Fig. 2B, Step 2). In turn, the nucleophilicity of the attacking oxygen atom of the first silicic acid is thought to be increased by the formation of a hydrogen bond with the nitrogen imidazole of the His. Such a reaction had been proposed earlier . The loss of a water molecule generates a disilicic acid molecule (Fig. 2B, Step 2), which remains bound to silicatein through the ester-like linkage. Now, rotation of the ester bond allows interaction of a second OH ligand of the enzyme-bound silicic acid unit with the nitrogen imidazole of the His in the catalytic center, giving rise to further growth from the disilicic acid to a third orthosilicic acid molecule (Fig. 2B, Step 3) with release of a water molecule. A repetition of this reaction cycle (nucleophilic attack/proton transfer/loss of water/rotation) may result in the generation of higher-membered silicic acid oligomers (tetrasilicic acid, etc.). In analogy to previous data , this model describes a cyclization process (Fig. 2B, Step 3). The subsequent polycondensation reactions might be guided by the silicatein interactor silintaphin-1 , which is able to assemble silica nanospheres formed by silicatein, a process that could proceed even in the absence of the enzyme.
Self-assembly of silicateins
Based on hydrophobic segment(s) existing within the silicatein molecules, fractal oligomerization of the silicatein molecules takes place. This process was first described for demosponges (T. aurantium  and S. domuncula ), and then also for the silicateins from the hexactinellid M. chuni . Starting from monomeric silicateins, a stepwise and stoichiometric assembly of the two isoforms of silicateins occurs . Monomeric silicatein-α molecules form dimers and then tetramers, comprising a central cleft into which one silicatein-β molecule is inserted (Fig. 2D,E). The resulting pentamers are proposed to align to filaments. According to the model prediction, the Ser clusters of four silicatein-α molecules are exposed to the center of the tetrad (Fig. 2F). Within the planar tetramer formed, the active sites of the silicatein-α molecule are oriented con-axial (active centers on the outer surface of the assembly), whereas the Ser clusters are oriented towards the center of the tetrameric silicatein complex. Those clusters are likely to allow binding of the oligo(silicate) product.
The biosilica-degrading silicase
The biosilica-catabolic silicase was identified from S. domuncula . The protein had been deduced from the isolated cDNA, and has an Mr of 43 131 (accession number DD298191). Sequence similarity analysis revealed that the silicase is a member of the family of carbonic anhydrases. The expression of the silicase gene is upregulated in response to higher orthosilicate concentrations. Functional analysis of the enzyme was performed in situ with light and electron microscopy techniques . It was found that silicatein and the silicase are colocalized on the surfaces of growing spicules, as well as in the axial canal. It is assumed that the silicase is also involved in biosilica deposition, during the reorganization/remodeling of the silica sheath in the growing spicule, as part of a network of (diffusible) regulatory factor(s) controlling enzymatic silica deposition and, in turn, the shape and the size of the spicules.
The sponge-specific silicatein-interacting proteins – silintaphins
Two proteins that interact with silicatein have been isolated. They have no distinct sequence similarity with any of the proteins listed in the databases.
Silintaphin-1 was identified as the first protein of the group of silicatein interactors (the silintaphins) . Database analysis revealed significant homology with the pleckstrin homology (PH) domain. Silintaphin-1 is colocalized with silicatein both intraspicularly and extraspicularly.
Very recently, a second member of the silintaphins, termed silintaphin-2, has been identified . This protein contains four Ca2+-binding sites. In turn, it has been proposed that the calcium ions, bound to the secreted silintaphin-2, could induce conformational changes that allow binding of silicatein to the silicate substrate.
In a first approach to determine the kinetic parameters for silicatein, an optical test was developed. On the basis of the observation that silicatein is a reversible enzyme, and functions both as a silica polymerase and as a silica esterase, the reaction parameters for recombinant silicatein-α were determined with the substrate bis(p-aminophenoxy)-dimethylsilane at a wavelength of 300 nm . With this substrate, the Km was calculated to be 22.7 μm. Such a value is below the critical concentration required for silica polycondensation in protein-free solution  or for nonenzymatic (collagen-mediated) silica polycondensation, which is ∼ 4.5 mm sodium silicate . The turnover number (molecules of converted substrate per enzyme molecule per second) of the enzyme in the silica esterase assay is 5.2 . The optimum temperature range is 20–25 °C.
After the discovery that, in sponge tissue, silicatein coexists with silintaphin-1 and silintaphin-2, the enzymatic activity of silicatein in the presence of the interacting protein, primarily silintaphin-1, was studied . The recombinant proteins, silicatein-α and silintaphin-1, were mixed at stoichiometric ratios of 1 : 1, 4 : 1, or 10 : 1, and incubated with orthosilicate (prehydrolyzed TEOS) as a substrate. The experiments revealed that, if both proteins were allowed to interact prior to addition of orthosilicate, the extent of biosilica formation was substantially higher.
The assumption that silicateins are primarily responsible for biosilica formation in sponges (at least in demosponges) is supported by a comparison of the rates of biosilica synthesis in the intact cell network and in vitro . In 3D cell cultures (primmorphs) from the freshwater demosponge Ephydatia fluviatilis, the spicule growth rate is high (1–10 μm·h−1 ). This species has spicules with an average length of 200–350 μm and a thickness of 15 μm. Consequently, the total weight of one spicule amounts to ∼ 88 ng, a value that is equivalent to approximately 4.4 × 1013 silica units being formed per hour in one spicule. Taking into account the 5% of protein in a spicule, as measured for M. chuni , it follows that 0.22 ng of protein (mainly silicatein-α) catalyzes the synthesis of 4.4 ng of biosilica per hour. As the molecular mass of mature silicatein is 25 kDa, it follows that 8.8 fmol of silicatein (5.3 × 109 molecules) in the spicules in vivo synthesize 4.4 ng of biosilica [73.3 pmol (4.4 × 1013 silica units)] per hour. This means that 8.3 × 103 silica units are formed by one silicatein molecule. This value is in the same range as the turnover rate for silicatein-α calculated in vitro, with recombinant silicatein. If the enzyme was allowed to reassociate with silintaphin-1 in a 4 : 1 molar stoichiometric ratio, the silica-forming activity was even higher, and similar to that of the native protein, isolated from spicules with the glycerol/Tris-based extraction buffer system.
Silintaphin-1 – the structure-guiding building block during biosilica formation
As outlined above, S. domuncula silintaphin-1 contains a PH domain, which most likely promotes the interaction between silintaphin-1 and silicatein. It is known that PH domain-containing proteins do not abolish the enzymatic function of the target proteins, but promote their ability to form fibers .
In initial studies with native silicateins from axial filaments of T. aurantium, it was shown that fibrous structure formation, via fractal intermediates, is a diffusion-limited process that is entropic and mediated by the interaction of hydrophobic patches exposed on the surfaces of the silicatein molecules . It was concluded that the fractal network formed subsequently condenses and organizes into a filamentous structure, the axial filament. More recently, similar self-assembly studies have been performed with recombinant silicatein-α and silintaphin-1, as well as with silicateins, extracted from spicules in a Tris-buffer system . With these samples, controlled reassembly to form axial filaments could be visualized (Fig. 3A,E,F). In contrast, when silicatein samples that were prepared with buffered hydrofluoric acid from the spicules were used, only nonordered aggregates/filaments without a distinct fractal pattern could be visualized (Fig. 3B).
To clarify the role of silintaphin-1 during the aggregation process, reassembly experiments with recombinant silicatein-α and recombinant silintaphin-1 were performed . Self-assembly of recombinant silicatein-α revealed no distinct fractals, and only ruffled aggregates could be visualized (Fig. 3C). However, when the recombinant protein was supplemented with recombinant silintaphin-1 (Fig. 3D), organized aggregates became visible after only 15 min. The aggregates increased in size and compactness after 30 min, and even more so after 60 min or 120 min. The lengths of the filaments exceeded 5 μm, but the diameters remained almost constant, at 20 nm. In order to determine whether, during silicatein-mediated biosilica formation, the structural pattern is changed, both the silicatein-α filaments and the silicatein-α/silintaphin-1 filaments were incubated with orthosilicate (prehydrolyzed TEOS) . During the 24-h incubation period, the silicatein-α fibers organized with the 30-nm-thick silica particles (nanospheres) to form densely packed rods (Fig. 3E,F). However, in contrast to the fibers formed with silicatein-α alone, the silicatein-α/silintaphin-1 fibers showed a filamentous backbone structure that was surrounded by the nanoparticles, demonstrating a cooperative effect of silicatein-α and silintaphin-1 during biosilica formation. These experiments show that silintaphin-1 not only enhances the enzymatic activity of silicatein-α, but is also crucial for the assembly of silicatein into organized filaments.
Growth of the siliceous spicules
With the primmorph cell culture system, the first solid evidence was provided that the synthesis of the spicules starts intracellularly [8,9]. The process of spicule formation can be divided into an initial intracellular phase and a subsequent extracellular shaping phase (Fig. 1D). Silicic acid, the substrate for silicatein, is actively taken up by cells (sclerocytes) via the Na+/ [Si(OH)4] cotransporter . In parallel, silicatein is synthesized, processed, and stored with silicic acid in special organelles of the sclerocytes, the silicasomes. Within these organelles, the axial filaments are formed, around which silica is deposited enzymatically. After formation of a first biosilica layer, immature spicules are released into the extracellular space, where they grow in length (axial direction) and in diameter (radial direction), by appositional layering of silica lamellae . During growth in the axial direction, biosilica is formed through the enzymatic function of the axial filament silicateins. The growth of the spicule is driven by the elongating biosilica core cylinder, which is synthesized by the 23-kDa processed form of silicatein. The radial thickening of the spicules, their appositional growth (radial direction), occurs by deposition of silica on the surface of the growing spicule, and is mediated by the extraspicular 34.7-kDa immature silicatein (prosilicatein) . Very likely, this 34.7-kDa silicatein undergoes processing to the 23-kDa protein by cleaving off (autocatalytically) of the N-terminal propeptide immediately prior to the onset of biosilica synthesis . There is no evidence at all that, in demosponges, either in the axial filament or on the surface of the spicules, there is collagen that is causatively involved in biosilica formation, as has been speculated .
The radial, appositional thickening of the spicules [9,51] and then the axial elongation process  were first understood. In the extracellular space, the silicatein molecules are organized into larger entities, by concentric rings/cylinders (Fig. 1A) around the spicule surface  (Fig. 4A). These structures become stabilized by the protein galectin and Ca2+; within these cylinders, the silicatein-mediated biosilica formation occurs. In hexactinellid spicules, these appositionally layered silica lamellae remain separated, and can reach 1000 in number  (Fig. 1F,G). In contrast, in demosponges, the individual lamellae fuse/biosinter  together, and form a ‘solid’ siliceous shell that surrounds the centrally located axial filament (Fig. 1B,C).
We provided, for the first time, experimental evidence for the involvement of cellular processes in the control of the axial growth of spicules. Using the primmorph system, we demonstrated that these cell processes originate from evaginations of the spicule-forming cells (sclerocytes) into the growing and elongating axial canal . The experiments showed that, around a cell extension that protrudes from a sclerocyte into the axial canal of a given spicule, silicatein molecules are released into the extracellular space of the axial canal (the space between the cell membrane and the inner surface of the siliceous mantel) to catalyze biosilica deposition from the inner surface. This causes the axial canal to narrow from ∼ 1 to ∼ 0.5 μm. Intracellularly, both the enzyme and its substrate (silica precursor) are stored in vesicles that have been termed silicasomes . These vesicles are released into the axial canal to allow the enzymatic polycondensation reaction. Therefore, spicule formation requires two mechanistically independent biosilica condensation/deposition processes: first, biosilica formation, which gives rise to the inner core of the growing spicule, mediated by silicatein present in the axial filament; and second, lamellar, appositional growth (thickening), which proceeds on the elongating siliceous core via the formation of organic cylinders, enabling the layer-by-layer deposition of individual silica lamellae, which biosinter together in demosponges or remain separated, to a large extent, in hexactinellids.
The formation of cell extensions during axial growth of the spicules is surely an energy-requiring mechanism, very probably dependent on ATP cleavage by the sponge arginine kinase. In S. domuncula, arginine kinase gene expression is induced by silicate , and is especially active around the spicules. Interestingly, in addition to silicatein and arginine kinase, a third protein, aquaporin-8, has been identified [56,57]. It is reasonable to assume that the removal of water, formed during the polycondensation reaction, from the site of biosilica synthesis will accelerate the condensation reaction and facilitate the maturation/aging process of the biosilica matrix .
The prevalent spicule type found both in tissue and in primmorphs from S. domuncula, the tylostyles (Fig. 5B,C), are monaxonal rods (length, 150–320 μm) that possess one blunt end, appearing as a knob (diameter, 6.5–9.2 μm), and one pointed tip. Each monaxonal rod contains a central axial canal (Fig. 1C), in which an organic axial filament is located (Fig. 1C) . In transmission electron microscopy (TEM) images, the axial filament has a dense appearance and is embedded in a granulated structure, filling the axial canal. As the synthesis of the spicules is a rapid process, and developing spicules can therefore only be seen very rarely in tissue from adult animals, a phase/time-specific analysis of the initial extracellular growth of the spicules has been performed with the primmorph system .
Analysis of the process of evagination of cell protrusions into the axial canal, using a longitudinal section through a complete growing monaxonal spicule of length 50 μm, revealed that the axial canal is closed at one end with a siliceous mantel, whereas the other end is open and associated with a sclerocyte (Fig. 4G) . Consequently, this site represents the growth region of the spicule. In the middle of the growing spicule, densely packed vesicles (silicasomes) can be identified, surrounded by a membrane that is probably the cell membrane (Fig. 4H). At the more terminal end, at the apex, the axial canal no longer shows any cellular fragments, and harbors only the axial filament (Fig. 4I). Cross-sections through growing spicules showed the existence of cell protrusions in those regions that are close to the connection with the sclerocytes (Fig. 4J,K), whereas in the more distantly located regions, which are closer to the tips of the spicules, only the axial filament is seen in the axial canal (Fig. 4L).
These data suggest that axial growth of the spicules is driven by the extension of cellular processes into the axial canal of the spicules through hydrodynamic forces that enable the evagination process.
Hardening of biosilica – an essential step during spicule formation
The biosilica matrix of the spicules is a solid hard rod . Hence, the biosilica product formed after the silicatein reaction must undergo a drastic process of hardening. A prerequisite for this reaction is the extrusion of water (syneresis) . This process has been studied recently in sponge 3D cell cultures (primmorphs). If primmorphs are cultivated in the presence of manganese sulfate, a change in morphology occurs . The primmorphs formed in the presence of manganese sulfate are never spherical, as in controls (Fig. 5A), but instead have a flat organization (Fig. 5E). They show similar viabilities to those in control primmorphs but, in contrast to them, they form spicules with a different shape and morphology. The surfaces of the newly formed spicules in nontreated primmorphs are smooth (Fig. 5B,C), but the surfaces seen in manganese sulfate-treated cultures have a different texture. Furthermore, on the surfaces of the spicules from manganese-sulfate treated cultures, silica deposits are formed (Fig. 5F,G) that are rough and porous; often, the spicules fuse/biosinter together (Fig. 5F) . Such a fusion/biosintering process of individual spicules is unknown in nontreated demosponges, but is frequently observed in hexactinellid taxa .
On the basis of the observations and spectroscopic analyses in primmorphs, it is reasonable to propose that a functional interaction of the aquaporin water channel with the enzymatic biosilica polycondensation mechanism exists that can be described as follows: (i) impaired synthesis of spicules; (ii) increased phosphorylation of silicatein; and (iii) reduced expression of the sponge aquaporin gene that is paralleled by a reduction in the formation of aquaporin protein . A schematic representation, given in Fig. 5D, highlights these sequential effects of manganese sulfate on the biosilica-synthesizing machinery in primmorphs. The observed increase in phosphorylation of silicatein presumably results in the formation of a less compact axial filament by aggregation/assembly of silicatein molecules. The removal of the water generated during the polycondensation reaction, which is essential for the maintenance and the velocity of spicule formation, from the area of biosilica synthesis to the surrounding extraspicular environment is surely not a rapid process, as compared with the fast growth of the spicules. The extraspicular environment is relatively poor in cells but rich in bulky extracellular matrix proteins, consisting of galectins as the dominant adhesion molecules. Galectin molecules form a gel-like scaffold in the presence of Ca2+ , and this in turn provides a fluid/stable dynamic matrix for the highly mobile ‘wandering cells’. Therefore, the environment of the spicules is highly suitable as a platform for cells transporting metabolites and also water. In view of the in vitro data, obtained with primmorphs, we have reasons to assume that, also in vivo, in intact tissue, water molecules formed during biosilica synthesis are trapped in the galectin matrix, from where they are imported into cells. Finally, these cells migrate, and thereby transport the water further away, resulting in an adjustment of the water equilibrium.
It should be noted that, during spicule formation, the diameter of the biosilica mantel decreases. Whereas the initial diameter of the developing spicule is ∼ 10 μm, it subsequently diminishes to approximately 2–3 μm [51,63]. Such a shrinking process is also observed during nonenzymatic sol–gel polycondensation processes at higher temperatures, and can be quantified with theoretical calculations, as follows. The diameter of a silicate monomer is 4.48 Å, whereas the dimension of one silica unit embedded in an almost completely polycondensated silica mesh is only 3.56 Å . As an example, and assuming that 30 molecules of silicate monomer are arranged in a space-filling manner (Fig. 5H), the same number of silica molecules organized in a completely polycondensated mesh would occupy a 30% smaller space (Fig. 5I).
The mechanism underlying the removal of water from biosilica via the process of syneresis might also help in understanding the formation of the various species-specific morphologies of the spicules formed in different sponge taxa. In both classes of siliceous sponge, spicule types exist that are richly ornamented and armed with hooks or thorns, or are bent in their longitudinal axes , a process that is very likely caused by a localized removal of water from the newly formed biosilica matrix of the growing spicule (Fig. 5J).
Cellular interplay during spicule formation
Recently, molecular markers have become available that allow a follow-up of some of the steps in the differentiation of stem cells to spicule-forming sclerocytes . With the primmorph system, it was shown that the following differentiated cell types, involved in skeletal formation, are derived from the totipotent stem cells: sclerocytes, archaeocytes, lophocytes, and chromocytes (Fig. 6). The chromocytes  store carotenoids that are produced by bacterial symbionts in the sponge [67,68]. These cells synthesize and release bone morphogenetic protein-1, a proteinase that is involved in the processing of a series of extracellular structural proteins, e.g. collagen and laminin , and also, in S. domuncula, silintaphin-2 . In the latter experimental model, experimental evidence has been presented showing that the Ca2+-binding protein silintaphin-2, which is formed in sclerocytes, delivers Ca2+ to the site of galectin filament/cylinder formation around the growing spicules . The galectin sheets provide the platform for binding of silicatein molecules, which in turn form an organic cylinder into which orthosilicate is channeled from the sclerocytes. Collagen might be involved in the setup of the organization framework for the organic cylinder, as previously proposed , but is surely not causatively involved in silicatein-mediated biosilica formation. The galectin molecules are synthesized in archaeocytes, whereas the silicateins come from the sclerocytes [29,70].
The synthesis of a functional inorganic polymer in an organism is dependent on the presence of an organic template that allows the deposition of the inorganic mineral. These biomineralization processes have been categorized , and grouped as either biologically induced mineralization, e.g. in the deposition of deep-sea crusts and nodules , or as biologically controlled mineralization, e.g. in the formation of bone . With the discovery of silicatein as the first enzyme that governs the polymerization of an inorganic material [17,29], a third highly specific biomineralization process can be added, biosilica formation in siliceous sponges. As described here, this enzyme allows the deposition of silica at undersaturated orthosilicate concentrations, and, even more interesting, this enzyme remains within the biosilica product, where it acts as an organic scaffold . In this way, the hybrid material provides the spicule with an unusual combination of mechanical properties in the sense of strength, stiffness, and toughness .
In contrast to other metazoan inorganic skeletal elements, the genetic blueprint for the synthesis of sponge spicules has been identified. The key genes and their expressed proteins required for their synthesis are now known. In turn, these molecules can now be used as tools for the design of bio-inspired materials in a rational way. It can be predicted that, by modification of these molecules, new materials can be designed that are provided with new properties. Hence, the elucidation of the silicatein-mediated synthesis of sponge spicules has introduced for the first time the concept of molecular bionics and changed a paradigm. Hitherto, it was assumed that enzymes are only able to form organic molecules/products from organic substrates, but, since the discovery of silicatein, it is accepted that, in living systems, inorganic products/materials can also be synthesized from inorganic precursors via a biocatalytic (enzyme-mediated) mechanism.
W. E. G. Müller is a holder of an ERC Individual Advanced Grant (No. 268476 BIOSILICA). This work was supported by grants from the Bundesministerium für Bildung und Forschung Germany (project ‘Center of Excellence BIOTECmarin’), the International Human Frontier Science Program, the Deutsche Forschungsgemeinschaft (Schr. 277/10-1), the European Commission (244967 ‘Mem-S’), the Public Welfare Project of Ministry of Land and Resources of the People’s Republic of China (Grant No. 201011005-06), and the International S & T Cooperation Program of China (Grant No. 2008DFA00980).