Co-expression and Functional Interaction of Silicatein with Galectin

Sponges (phylum Porifera) of the class of Demospongiae are stabilized by a siliceous skeleton. It is composed of silica needles (spicules), which provide the morphogenetic scaffold of these metazoans. In the center of the spicules there is an axial filament that consists predominantly of silicatein, an enzyme that catalyzes the synthesis of biosilica. By differential display of transcripts we identified additional proteins involved in silica formation. Two genes were isolated from the marine demosponge Suberites domuncula; one codes for a galectin and the other for a fibrillar collagen. The galectin forms aggregates to which silicatein molecules bind. The extent of the silicatein-mediated silica formation strongly increased if associated with the galectin. By applying a new and mild extraction procedure that avoids hydrogen fluoride treatment, native axial filaments were extracted from spicules of S. domuncula. These filaments contained, in addition to silicatein, the galectin and a few other proteins. Immunogold electron microscopic studies underscored the role of these additional proteins, in particular that of galectin, in spiculogenesis. Galectin, in addition to silicatein, presumably forms in the axial canal as well as on the surface of the spicules an organized net-like matrix. In the extraspicular space most of these complexes are arranged concentrically around the spicules. Taken together, these additional proteins, working together with silicatein, may also be relevant for potential (nano)-biotechnological applications of silicatein in the formation of surface coatings. Finally, we propose a scheme that outlines the matrix (galectin/silicatein)-guided appositional growth of spicules through centripetal and centrifugal synthesis and deposition of biosilica.

etons are termed spicules; they are used as systematic characters for a given sponge species (2). Given the comprehensive studies of Bütschli (3) and Minchin (4), a descriptive view of the formation of the siliceous spicules has been well established. In demosponges, where most studies have been performed with Suberites domuncula, spicules are initially formed within specialized cells called sclerocytes (5). S. domuncula has the advantage of containing only macroscleres (tylostyles/oxeas), whereas most other sponges, e.g. Tethya aurantium, contain macroscleres (oxeas) as well as microscleres (spherasters) (6). Spicules have in their center a 1-2-m-wide axial canal (7), which contains the axial filament. In demosponges first siliceous deposits are arranged around this axial filament. When spicules reach lengths of about 10 m they are extruded from the cells. The spicules are completed extracellularly in the mesohyl (8), where they reach final sizes of 10 m (microscleres) and 200 m (macroscleres) as well as their final shapes by appositional growth (7,9).
Major progress in understanding the biochemical mechanism of spicule formation in Demospongiae came from the discovery that silica deposition in T. aurantium is catalyzed by the enzyme silicatein (10,11); silicatein exists in the axial filament around which the first layer(s) of biosilica are formed. The axial filament had been isolated by hydrogen fluoride (HF), 2 which dissolves the biosilica of the spicules. During this procedure O-glycosidic and phosphate ester bonds in proteins are split off. As reported for the silaffins in diatoms, this treatment results in a loss of phosphate groups of the structural molecules and also a modification of their functions (12,13). Therefore, the (apparently pure) preparation of the axial filaments (composed of silicatein) from T. aurantium studied hitherto (10,14) may not represent native silicatein. Furthermore, the reported presence of three different but related silicateins (silicatein-␣, -␤, and -␥) in the axial filaments of spicules had not been substantiated on molecular level. Until now, only genes encoding silicateins ␣ and ␤ could be isolated from T. aurantium (10,14) or other demosponges, e.g. S. domuncula (5,15); it may however be possible that the transcript level of the putative silicatein-␥ is too low. Very recently it could be proven that axial filaments of S. domuncula contain, in addition to silicatein, other proteins such as selenoprotein M or the spiculeassociated protein (9). A thorough cytological and electron microscopic study revealed that the axial canal, which harbors the axial filament, is filled with a non-homogenous material (8). It has been reported that the silicateins in S. domuncula are phosphorylated proteins, which are HFlabile. Only in this state are they able to oligomerize (form silicatein dimers and trimers) (8).
Even though the enzymatic activity of silicatein isolated from axial filaments by HF treatment, as well as the enzymatic activity of recombinant silicatein, can be demonstrated to a limited extent, which likely does not reflect the physiological in vivo activity seen in the animal (14,15). Under optimal conditions, spicules have an astonishing growth rate of 5 m/h (16), which certainly cannot be attributed to silicatein or recombinant silicatein alone (14). Because silicatein might be technically used to form silica nanoparticles and nanofibers under ambient (low temperature and low pressure) conditions (17)(18)(19), a knowledge on the protein composition of the axial filament and of the interactions of these proteins in their native states is also of interest in nanobiotechnology and nanomedicine. Silicatein, when chemically bound to metal surfaces, e.g. gold, retains its biocatalytic activity (18). Even more impressive are studies demonstrating that immobilized silicatein also allows the biosynthesis of titania and zirconia (19).
Here we present a comprehensive study on the role of silicatein from S. domuncula in silica/spicule formation. The question by which molecules silica deposition in sponge spicules is biologically controlled is addressed and partially answered. Evidence is presented that a galectin builds, in concert with collagen, a proteinaceous framework into which silica is deposited. These studies took advantage of (a) the primmorph culture system, a special form of three-dimensional cell aggregates that comprise proliferating and differentiating sponge cells (20), and (b) antibodies against silicatein.

EXPERIMENTAL PROCEDURES
Chemicals, Materials, and Enzymes-Restriction enzymes, a Total RNA isolation kit, and reagents for the RACE procedure were purchased from Invitrogen; TriplEx2 vector from BD Biosciences; pET-41a(ϩ) vector and BugBuster reagent from Novagen (Darmstadt, Germany); TRIzol reagent from Invitrogen; Hybond-N ϩ nylon membrane from Amersham Biosciences; Cy3-conjugated F(abЈ) 2 goat antirabbit IgG from Jackson ImmunoResearch; PCR-DIG probe-synthesis kit, a BM chemiluminescence blotting substrate kit, proteinase inhibitor mixture, and CDP-Star from Roche Applied Science; sodium metasilicate from Aldrich; a silicon test colorimetric assay kit and silicon standards from Merck (Darmstadt, Germany); natural seawater (containing 5 M silicate) and tetraethoxysilane from Sigma; protein A/G-agarose beads from Santa Cruz Biotechnology (Santa Cruz, CA); 1.4-nm nanogold anti-rabbit IgG from Nanoprobes (Yapbank, NY); and rhodamineconjugated goat anti-rabbit immunoglobulin from Dako (Carpinteria, CA). The synthetic oligopeptide was custom-synthesized.
Sponges and Primmorphs-Specimens of the marine sponge S. domuncula (Porifera, Demospongiae, Hadromerida) were collected in the northern Adriatic Sea near Rovinj, Croatia, and then kept in aquaria in Mainz, Germany, at a temperature of 17°C for more than 5 months.
Primmorphs, the in vitro cell culture system, were prepared as described (15). These three-dimensional cell aggregates were used 5 days after their formation. Primmorphs were kept in natural seawater (containing 4 -6 M silicate) supplemented with 1% RPMI 1640 medium in the absence or presence of 60 M sodium silicate (as sodium metasilicate) for up to 5 days. Extracts from primmorphs or from sponge tissue were prepared by homogenization in lysis buffer (1ϫ Trisbuffered saline, pH 7.5, 1 mM EDTA, 1% Nonidet-P40, and protease inhibitor mixture (1 tablet/10 ml)) and centrifuged, and the supernatants were subjected to further analysis.
Isolation of the Proteins from Spicules-In earlier studies isolated spicules had been treated with HF to obtain the axial filament (10). In the present study this harsh procedure was replaced by treatment of the spicules with lysis buffer supplemented with 4 M urea. At first the spicules were isolated from tissue, collected (8), and then pulverized thoroughly in a mortar. The lysis-urea buffer was added, and the suspension was stirred for 1 h at 4°C and subsequently centrifuged at 10,000 ϫ g for 5 min at 4°C. The supernatant was used for further analysis by SDS-PAGE and Western blotting.
Differential Display and Cloning of Galectin and Collagen cDNAs-The technique of differential display of transcripts was performed as described previously (21). RNA (1 g) from untreated primmorphs (control, 0 days) and primmorphs treated for 2 days with 60 M sodium silicate, was reverse-transcribed using the T 11 CC oligonucleotide as 3Ј primer. The resulting cDNA was added to the PCR using the arbitrary primer GTGATCGCAG and the T 11 CC primer in the assay together with [␣-32 P]dATP. Among the dominantly expressed genes, collagen (SDCOLL3) and galectin (SDGALEC2) transcripts as well as transcripts of the silicatein-␣ gene (SDSILCAa) were identified. The ϳ300 -380nt-long sequences were completed by 5Ј-RACE with vector primers of the cDNA library from S. domuncula (22). The complete clone for collagen-3 (SDCOLL3) comprises 1302 nts, and that for galectin-2 (SDGA-LEC2) 955 nts.

Preparation of Recombinant S. domuncula Galectin-2 and Antibodies-
The complete open reading frame of SDGALEC2 was expressed in Escherichia coli using the glutathione S-transferase (GST) fusion system in the pET-41a(ϩ) vector as described (25,26). The GST fusion protein (65 kDa) was purified by GST-Bind Resign affinity chromatography. The fusion protein was then cleaved with enterokinase to separate the N-terminal GST tag and His tag as well as the S-tag from the recombinant galectin (rGALEC2_SUBDO). This resulting galectin-2 carried at its C terminus a small His tag. Gel electrophoresis of protein extracts was performed in polyacrylamide gels (12% gels) containing 0.1% SDS-PAGE according to Laemmli (27).
Polyclonal antibodies (pAb) were raised against the purified, recombinant galectin-2 (rGALEC2_SUBDO) in female rabbits (New Zealand White) as described (28). 10 g of recombinant protein per injection was dissolved in phosphate-buffered saline (PBS). After three boosts the serum was collected; the pAb against galectin-2 was termed pAb-aGA-LEC2. The titer of the antibodies was 1:3,000. In the controls, adsorbed pAb-aGALEC2 (100 l of antibodies were incubated with 20 g of rGALEC2_SUBDO) was used. Likewise, pAbs against silicatein (pAb-aSILIC) were raised against axial filaments from spicules, containing mainly silicatein, as described earlier (8).
Co-immunoprecipitation and Western Blot Analysis-For co-immunoprecipitation the described protocol (29, 30) was applied. Primmorph samples (200 mg each) were exposed in seawater to 60 M sodium silicate for 0 -5 days. Then they were shock-frozen in liquid nitrogen, subjected to twice the volume of lysis buffer (1ϫ Tris-buffered saline, pH 7.5, 50 mM NaCl, 1 mM EDTA, 0.1% Nonidet-P40, 10 mM NaF, protease inhibitor mixture (1 tablet/10 ml), and 1 mM sodium orthovanadate), homogenized, incubated for 60 min at 37°C, and then brought to 4°C. Finally the samples were subjected to co-immunoprecipitation or Western blot analyses. For each immunoprecipitation a 100-l aliquot of the lysate was incubated with 2 l of undiluted serum (preimmune or immune pAb (pAb-aGALEC2 or pAb-aSILIC)) for 1 h (4°C). Protein A/G-agarose beads (10 l) were added and after 3 h were washed with lysis buffer and resuspended in SDS-PAGE sample buffer. SDS-PAGE-fractionated proteins were analyzed on Western blots. The membranes were probed with the pAb (pAb-aGALEC2 or pAb-aSILIC) and detected by anti-rabbit IgG (peroxidase-coupled) using the BM chemiluminescence blotting substrate kit. In control experiments 100 l of the respective antibodies pAb-aGALEC2 and pAb-aSILIC were adsorbed with 20 g of the recombinant proteins (rGALEC2_SUBDO or rSILIC_SUBDO) during an incubation period of 30 min (4°C) prior to their use.
Competition experiments with the synthetic oligopeptide (corresponding to the hydrophobic C terminus of galectin-2) were performed as follows. The 100-l extract in lysis buffer was supplemented with 5 l of the oligopeptide (5 g), allowed to stand for 60 min at 37°C, and then brought to 4°C. Subsequently, the samples were processed with the antibodies as described above.
Silicatein Enzyme Assay-For these studies recombinant silicatein was used; preparation (17, 31) and reaction conditions (14, 31) were as described previously. In a total volume of 500 l, 100 g of recombinant silicatein was added to the 25 mM Tris-HCl buffer, pH 7.2 (10 mM DLdithiothreitol, 100 mM NaCl, 1 M ZnCl 2 , 5 M Fe 3ϩ citrate, 5 mM CaCl 2 , 5 mM MgCl 2 ). 1 mM tetraethoxysilane was used as substrate. Reaction proceeded at 22°C for 0 -120 min and was terminated by centrifugation (20,000 ϫ g; 4°C, 15 min). The sediment was hydrolyzed by NaOH, and the released silicic acid was determined as described (14). Where specified, the standard assays were performed in the absence of silicatein, or they contained 100 g of galectin-2 (rGALEC2_SUBDO) instead of silicatein or 100 g of galectin-2 together with 100 g of silicatein.
Aggregate Formation of Recombinant Galectin-2-To visualize aggregate formation of rGALEC2_SUBDO in vitro recombinant protein (0.1 mg/ml) was dissolved in PBS supplemented with 5 mM CaCl 2 . After standing on ice for 2-30 min, samples were dropped onto a carboncoated copper grid. After drying, the film was stained with platinum and examined by electron microscope (32).
Electron Microscopy-For transmission microscopic analysis, sponge primmorphs or tissue samples were cut into pieces (1 mm 3 ), incubated in phosphate buffer, and then washed in phosphate-buffered saline, pH 7.4, at room temperature as described (5,8). After the samples were treated with 1.25% NaHCO 3 , 2% OsO 4 , and 1% NaCl, they were dehydrated with ethanol. The dried samples were fixed in propylene oxide/araldite, covered with pure araldite, and hardened at 60°C for 2 days prior to cutting to 60-nm ultrathin slices (Ultracut S; Leica, Wetzlar, Germany). The samples were transferred onto coated copper grids and analyzed with a Tecnai 12 microscope (FEI Electron Optics, Eindhoven, Netherlands).
Electron immunogold labeling was performed with primmorph/tissue samples treated in glutaraldehyde/paraformaldehyde buffered in phosphate buffer, pH 7.4 (5,8). The material was dehydrated in ethanol FIGURE 1. The two S. domuncula galectins (GALEC1_SUBDO and GALEC2_SUBDO (in white letters on black)) were aligned with the G. cydonium galectin-1 (GALEC1_GEOCY, X93925) and the human galectin-8 isoform b (GALEC8b_HUMAN, O00214). In addition, the two segments within GALEC2_SUBDO (in white letters on gray) and GALEC8b_HUMAN, spanning the galactose-binding domains have been included. In S. domuncula galectin-2, the parts ranged from aa 1 to aa 117 and from aa 118 to aa 293 (designated GALEC2a_SUBDO and GALEC2b_SUBDO) and in the human galectin-8b from aa 1 to aa 150 and from aa 151 to aa 316 (GALEC8ba_HUMAN and GALEC8bb_HUMAN). These domains are marked Gal-binding-1 and Gal-binding-2. In addition, the N terminus of the mature galectin-2 (ϩ::::) and the hydrophobic terminus (hydrophobic) at the C terminus of this protein are indicated. Amino acids that are similar among all sequences are in reversed type, and those conserved in at least two sequences are shaded. and embedded in LR-White resin. 60-nm thick slices were cut and blocked with bovine serum albumin in PBS and then incubated with the primary antibody pAb-aSILIC (1:1,000) for 12 h at 4°C. In controls, preimmune serum were used. After three washes with PBS, 1% BSA, sections were incubated with a 1:100 dilution of the secondary antibody (1.4-nm nanogold anti-rabbit IgG; diluted 1:200) for 2 h. Sections were rinsed in PBS, treated with glutaraldehyde/PBS, washed, and dried. Subsequently, enhancing of the immunocomplexes was performed with silver as described (33). The samples were examined as described above with the Tecnai 12 microscope.
S. domuncula Collagen-The sponge collagen was isolated from fresh tissue by urea extraction followed by acid precipitation (32). The fibrils were visualized electron microscopically after negative staining with potassium phosphotungstate and shadowing with platinum.
Immunohistology-Tissue samples were embedded in paraformaldehyde. After dehydration the samples were embedded in Technovit 8100, and 8-m-thick sections were cut (8). The sponge tissue was not treated with HF/NaF solution to ensure that siliceous spicules were not removed. The sections were incubated with the anti-galectin-2 antibody (pAb-aGALEC2); rhodamine-conjugated goat anti-rabbit immunoglobulin was used as a secondary antibody. The preimmune rabbit serum was used as a control. Further details were given recently (9). The immunofluorescence analysis was done with an Olympus AHBT3 light microscope together with an AH3-RFC reflected light fluorescence attachment at an emission wavelength of 546 nm (filter G).
Further Analytical Methods-For the quantification of protein, the Bradford method (34) (Roti-Quant solution-Roth) was used. The N-ter-minal part of the mature galectin-2 protein was determined as described earlier (35). After size separation by SDS-gel electrophoresis the N terminus of galectin-2 was determined by direct protein sequencing.

Co-expression of Silicatein with Galectin and Collagen
By differential display in addition to the silicatein-␣ gene (SDSILCAa; CAC03737.1), two major genes were identified that are expressed in primmorphs in response to sodium silicate treatment. Analysis revealed that one of these transcript codes was for a galectin and the other one for a collagen protein; both cDNAs were completed.
Galectin-The 955-nt-long cDNA for galectin (SDGALEC2) comprises one open reading frame between nt 70/72 and nt 949 -951 [stop] . The deduced protein (GALEC2a_SUBDO) of 293 aa (Fig. 1) has a predicted size of 33,091 Da. Galectin-2 shows two galactoside-binding lectin domains (Motif Scan), which span aa 1 to aa 117 (domain 1) and aa 118 to aa 293 (domain 2). A highly hydrophobic region in galectin-2, predicted according the described algorithm (36), is found between aa 275 and aa 293 . The mature protein, isolated after separation on an SDS-gel, begins at aa 10 of the predicted protein. Highest sequence similarity was found in the data base with S. domuncula galectin-1 with an "expect value" (E) (26)   2A), with a predicted size of 357 amino acids and a molecular mass of 36,665 Da. Three segments can be distinguished: (i) the N-terminal non-collagen region (aa 1 to aa 26 ); (ii) the region comprising the characteristic G-X-Y collagen triple helix repeats (in COLL3_SUBDO 37 triples; aa 27-136 ); and finally (iii) the non-collagen C-terminal part (aa 137-357 ). This arrangement has been described also for the other collagen proteins from S. domuncula (15,37). Data bank searches revealed that COLL3_SUBDO shares highest sequence similarity to the fibrillar collagen from Haliotis discus (BAA75668 (38)) with E ϭ 8e Ϫ39 and from Xenopus laevis (AAH49829 (39)) with E ϭ 2e Ϫ34 .

Expression Patterns of Silicatein-␣, Galectin, and Collagen
To verify that the three genes, silicatein-␣ (SDSILCAa, 1.4 kb transcript size), galectin-2 (SDGALEC2, 1.2 kb), and collagen-3 (SDCOLL3, 1.1 kb) were up-regulated, primmorphs were incubated with sodium silicate for at least 2 days. The steady-state expressions of the genes were determined semiquantitatively and showed only very low levels of these transcripts at day 0. A strong increase could be measured for the transcripts after 2 or 5 days (Fig. 3). Control experiments with the housekeeping gene ␤-tubulin (transcript size of 1.4 kb) showed that the same amounts of RNA were loaded onto the gels (Fig. 3).

Recombinant Galectin, Expression, and Aggregate Formation
Galectin-2 was expressed as a fusion protein with a GST tag, His tag, S-tag, and again a His tag. This protein was analyzed by SDS-PAGE and found to display the expected size of 69 kDa (37 kDa for galectin-2 and 32 kDa for the tags ; Fig. 4A, lane a). Antibodies (pAb-aGALEC2) raised against recombinant galectin-2 recognized this protein on a Western blot (Fig. 4A, lane b). The fusion protein was subsequently cleaved with enterokinase to remove the N-terminal Tag region, and the final recombinant galectin (rGALEC2_SUBDO) with its C-terminal His tag had the expected size of 34 kDa (data not shown). The antibodies were also used to identify the native protein in the crude sponge extract (Fig. 4B, lane  a). In this preparation pAb-aGALEC2 recognized the 35-kDa native galectin-2 (Fig. 4B, lane b). To demonstrate the specificity of pAb-aGA-LEC2, the antibodies were adsorbed with rGALEC2_SUBDO prior to the application in the Western blot. The adsorbed antibodies did not show any visible immunocomplexes (Fig. 4B, lane c).
Based on earlier observations with Geodia cydonium galectin, we checked whether the S. domuncula galectin-2 forms aggregate in the presence of 5 mM CaCl 2 . Samples were reacted with the cation and analyzed electron microscopically. In the initial reaction only small particles (around 4 nm) were seen; during longer periods of incubation of up to 30 min, compact coil-like aggregates formed (Fig. 5). In the absence of CaCl 2 no aggregate formation was observed (not shown).

Interaction of Silicatein-␣ with Galectin-2 Proteins
The experiments were performed with primmorphs. They remained for 0 -5 days in seawater supplemented with 60 M sodium silicate. Then lysates were prepared, subjected to PAGE (Fig. 6), and analyzed for the presence of a complex between galectin-2 and silicatein, applying the co-immunoprecipitation technique. pAb-aGALEC2 and pAb-aSILIC were used as tools to identify the two proteins, as described under "Experimental Procedures." In one series of experiments, the primmorph lysate was reacted with pAb-aGALEC2 and in a second series with pAb-aSILIC. After immunoprecipitation, the proteins were size-separated and probed for Western blots with antibody combinations as follows. The proteins precipitated with pAb-aGALEC2 were reacted with pAb-aSILIC (Fig. 6a), or

FIGURE 6. Co-immunoprecipitation (ImmPrec) and Western blot analysis (WestBlot) of galectin-2 and silicatein.
The primmorphs were incubated for 0 -5 days with 60 M sodium silicate. Extracts were prepared, which remained either in the absence of the synthetic oligopeptide N-LLIFDGLKIVLYCVLYCI-C (minus peptide) or were pretreated with this peptide (plus peptide). Then the samples were subjected to immunoprecipitation with pAbs directed against either galectin-2 (aGALEC2) (a) or silicatein (aSILIC) (b). The immunocomplexes were isolated with protein A/G-agarose beads. The bound proteins were separated by SDS-PAGE. Then the proteins bound to pAb-aGALEC2 were processed by Western blotting with the polyclonal antibodies aSILIC, and those bound to pAb-aSILIC were identified on the blots with the polyclonal antibodies aGALEC2. Further details are given under "Experimental Procedures." the pAb-aSILIC precipitated material was probed with pAb-aGALEC2 (Fig. 6b). It can be seen that pAb-aGALEC2 also precipitated, in addition to galectin-2 (Fig. 4), the 24-kDa silicatein protein (Fig. 6a). Silicatein was identified by pAb-aSILIC. The immunoreaction was strong in lysates prepared from primmorphs that had been incubated with silicate for 2 or 5 days; at time zero almost no immunocomplexes could be seen. In a control experiment the specificity of the reaction was demonstrated by applying adsorbed pAb-aSILIC; this preparation did not react with any lysate protein (not shown). In the reverse series, lysates were treated with pAb-aSILIC (Fig. 6b). After precipitation and size separation the blots were incubated with pAb-aGALEC2, and the immunocomplexes were visualized with labeled anti-rabbit IgG. At time zero, only a faint band, corresponding to the 35-kDa galectin-2 was detected. After incubation for 2 or 5 days the amount of 35-kDa galectin-2 strongly increased. Again the specificity of the pAb-aGALEC2 antibodies was verified after adsorption; no strong reactions on the blot could be seen (not shown). Further controls were performed with preimmune serum, which revealed no signals on the Western blots (not shown). From these data we conclude that galectin-2 interacts with silicatein in lysates from primmorphs that have been treated for at least 2 days with silicate.
In a further series of experiments the site of interaction between galectin-2 and silicatein was investigated. As outlined above, galectin-2 carries at the C terminus a highly hydrophobic stretch (aa 275-293 ), which can be considered as one segment involved in such an interaction. To test this assumption competition experiments were performed with the synthetic oligopeptide. The peptide was added prior to the incubation step at 37°C for 60 min. These samples were processed with the two antibody combinations: immunoprecipitation with pAb-aGALEC2 and subsequent Western blotting with pAb-aSILIC (Fig. 6a); or immunoprecipitation with pAb-aSILIC followed by Western blotting with pAb-aGALEC2 (Fig. 6b). The signals visible in those Western blots were considerably lower than those seen with extracts not supplemented with the oligopeptide, e.g. the signal (extracts from primmorph after 5 days in silicate) of the co-precipitation experiment pAb-aGALEC2/ pAb-aSILIC was reduced to less than 30%, close to that seen in extracts from primmorphs not treated with silicate. These findings support the presumption that galectin-2 binds to silicatein, very likely through its C terminus.

Effect of Galectin-2 on the Silicatein Reaction
In earlier studies it was shown that silicatein mediates the formation of silica in solution (14,15). In the present study we determined the effect of galectin-2 on the extent of silica formation in assays containing 5 mM CaCl 2 in addition to silicatein and its substrate, tetraethoxysilane (1 mM). As summarized in Fig. 7A, in the absence of silicatein, the detectable amount of silica was low (19 Ϯ 4 ng/assay during the 120-min incubation period). The addition of silicatein (100 g/assay) resulted in silica production; after 120 min 47 Ϯ 12 ng/assay was formed. The addition of galectin-2 (100 g) alone also resulted in a significant but smaller increase in the amount of silica (23 Ϯ 5 ng/assay). However, if silicatein (100 g) was added together with galectin-2 (100 g), a strong increase in silica formation (115 Ϯ 20 ng/assay) was measured.
In a parallel series, the synthetic oligopeptide was added to the silicatein-containing assay in the presence and absence of galectin. Again it is seen that in the presence of silicatein (42 Ϯ 9 ng/assay; 120-min incubation period) or galectin (16 Ϯ 5 ng) a significant amount of silica was formed (Fig. 7B). However, if galectin-2 was added together with the oligopeptide and silicatein (53 Ϯ 16 ng), no significant stimulatory effect of galectin was measured.

S. domuncula Collagen
Collagen fibrils were isolated from fresh tissue and analyzed electron microscopically (Fig. 2B). After isolation and subsequent staining/shadowing, the banding pattern of the collagen fibrils became visible; the diameter of the fibrils is ϳ15 nm, and the simplified arrangement of bandings has a periodicity of about 20 nm.

Existence of Galectin-2 in Axial Filaments and on Surface of Spicules
Spicules were prepared, pulverized, and finally extracted with a lysis buffer as described under "Experimental Procedures." Those extracts contained as one dominant protein, the 24-kDa silicatein(s), as checked by SDS-PAGE (Fig. 8, lane b) followed by Western blotting with pAb-aSILIC (lane c). For comparison, the protein pattern of crude extracts from primmorphs obtained with this urea-supplemented lysis buffer is shown after SDS-PAGE (Fig. 8, lane a). The second major band in spicule extracts corresponds to 35 kDa, as seen on the Coomassie Brilliant Blue-stained gel (Fig. 8, lane b). This protein was identified by Western blotting with pAb-aGALEC2 as the 35-kDa galectin-2 (Fig. 8, lane d). In the high molecular range of the stained gels two further prominent protein bands exist that correspond to a size of Ͼ250 kDa.

Immunofluorescence Analysis; Identification of Galectin-2 in Spicules
Cryosections were prepared from sponge tissue and reacted with pAb-aGALEC2, and the immunocomplexes were visualized with Cy3labeled IgG. As shown in Fig. 9A, a bright staining on the sections is seen on the surface of the spicules and also in their axial canal/axial filaments; this localization becomes more obvious at a higher magnification (Fig.  9C). The corresponding Nomarsky interference image shows that spicules are embedded in the bulky mesohyl (Fig. 9D). A control series shows that the preimmune serum did not react with structures on the slides (Fig. 9B).  ). B, in a parallel series, 5 g of the synthetic oligopeptide N-LLIFDGLKIVLYCVLYCI-C was added to each assay. After incubation for 0 -120 min, the amount of insoluble silica formed was determined by measuring the silicic acid released following treatment with NaOH. The values, expressed in ng of silicic acid/assay, are the means Ϯ S.D. of five separate experiments.

Embedding of Spicules in the Mesohyl of the Sponge
The initial synthesis of the spicules occurs intracellularly in the sclerocytes (8). After reaching a critical length of 5-10 m, the spicules are extruded into the bulky extracellular matrix within the mesohyl. Transmission microscopic analysis shows that in this space the spicules are surrounded by cells that never come closer to the spicules than 10 nm. Some of these 10-nm spaces contain string-and net-like structures. In the mature state the axial filament (Fig. 10A), with a diameter of 300 -600 nm, is compact. In growing spicules the axial canal is filled with the axial filament, which is surrounded by membranous structures and a matrix consisting of bundles and string-and net-like structures of variable size between 0.5 and 4 nm (Fig. 10B). The extraspicular space likewise contains those small string-and net-like structures as well as solid rods. These rods are rarely curved and have diameters between 10 and 14 nm (Fig. 10C). At higher magnification they show a banding pattern, indicating that they are collagen fibers (Fig. 10D). In contrast to the rods, the string-and net-like structures are often, especially in the axial canal, closely associated with the silica shell of the spicules (Fig.  10E); collagen fibers are never found tightly associated with the silica shell of the spicules. However, frequently the collagen fibers existing in the extraspicular space are intimately associated with the 0.5-4-nmthick string-/net-like structures (Fig. 10F).

Association of Silicatein with String-and Net-like Structures
Silicatein molecules were visualized by the electron immunogold labeling technique using polyclonal anti-silicatein antibodies, pAb-aSILIC (Fig.  11). By this technique the labeled antibodies, recognizing silicatein, were never seen to be directly associated with collagen. pAb-aSILIC recognized aggregates of silicatein of different sizes, which were always associated with the string-and net-like structures (Fig. 11, A-E). In a control experiment, the preimmune serum did not react with the silicatein particles (Fig. 11F). When associated with collagen fibers, the silicatein/string-and net-like structures are organized in concentric rings (Fig. 11D). The following sequential association of the three components, silicatein (as immunogold signal), string-/net-like structures, and collagen, might be proposed. At first, silicatein is associated in an orderly manner with the string-and netlike structures (Fig. 11, A-C). In the initial phase the monomeric gold particles are arranged (in a "pearl string" manner) always in double rows (Fig.  11, A and B). Often the particles are aggregated together in areas that show also the string-and net-like structures (Fig. 11, B and C). When these immunogold particles together with the string-and net-like structures are seen in association with collagen fibers, a linear arrangement is induced (Fig. 11, D and E).

DISCUSSION
The synthesis of sponge spicules is a complex process which involves first the intracellular synthesis of the initial silica layer(s) and, after reaching a critical size, the extracellular completion of these skeletal elements in the mesohyl. Because the shapes of sponge spicules display an amazingly artistic diversity, which serves as the main taxonomic characteristic to determine the species, it must be assumed that several molecules are involved in the synthesis of the spicules and in the control of their size, form, and shape. In the present study we describe the  . Immunohistological analysis of sections through tissue with antibodies against galectin-2. The slices were stained with pAb-aGALEC2 (A and C). Parallel to the immunostaining in C, the corresponding Nomarsky interference images is given (D). As a control one section was reacted with preimmune serum (B). Spicules (sp) and canals (ca) are marked. Transmission microscopic analysis was performed as outlined under "Experimental Procedures." A, cross-section through a spicule within a primmorph showing that they are closely, but never intimately, associated by cells. Within the spicules the triangular axial canal with its compact axial filament (af) is seen. B, in addition to the axial filament, actively synthesizing spicules in primmorphs in the axial canal also contain fibrillar strings (st). C, in the extraspicular space (tissue) two types of fibrils are recognized: first, the solid rods, or collagen (col); and second the fibrillar, irregular strings/nets. D, at a higher magnification the collagen rods with their banding pattern are seen. E, in the axial canal (ac) the strings (or nets) are closely associated with the silica shell. F, more distantly located from the silica shell, the collagen fibers are often tightly associated with the strings/nets. With the exception of the image in C, all sections came from 12-day-old primmorphs.
differential expression of two genes that are involved directly in spicule formation: galectin-2 and collagen.
Galectins are the prevailing proteins in the extracellular space of demosponges, e.g. G. cydonium (35,41) and S. domuncula (42). In S. domuncula the differential display of transcripts revealed one dominant galectin, termed galectin-2. This lectin comprises, in contrast to galectin-1 from G. cydonium (35), not only two galactose-binding sites but also one hydrophobic region at its N terminus. It is important for the understanding of the silicatein function that S. domuncula galectin-2 is able to self-associate in the presence of Ca 2ϩ ; this property already has been described earlier for the G. cydonium galectin (41). Also in G. cydonium, galectin is a dominant protein in the mesohyl (41,43), which forms in the presence of Ca 2ϩ ions three-dimensional self-associates, with sizes of Ͼ1 m. Because the native S. domuncula galectin-2 is a glycoprotein, we assume that this self-association is due to the binding of the galactose-binding domain to the carbohydrate side chains within galectin-2, as shown for the G. cydonium galectin.
The second molecule expressed in S. domuncula after exposure to silicate is collagen. Until now, only genes encoding non-fibrillar collagens had been described in sponges, e.g. as for Ephydatia muelleri (40) or S. domuncula (37). The newly discovered S. domuncula collagen displays the typical structure for fibrillar collagens: the N-peptide, the triple helix segment, and the C-peptide. Fibril-forming collagens are characterized by the ability to assemble into highly oriented supramolecular aggregates with a characteristic suprastructure, the quarter-staggered fibril array (see Ref. 44). Electron microscopic analysis of the S. domuncula collagen molecules revealed a pronounced banding pattern. Those collagens in sponges had been identified earlier from electron microscopic analyses (45). It had been proposed that fibrillar collagens control spicule formation (46). In addition, collagen had been implicated in the directional movement of cells and transport of spicules (46,47). The collagen fibers have now been demonstrated in S. domuncula electron microscopically, and their role in the morphogenesis of the spicules has become likely, as outlined below.
As a further class of structural proteins occurring in Demospongiae, spongins have been described (reviewed in Ref. 48). This type of proteins was first reported by Croockewit (49); it is rich in glycine (50) but is distinguishable from collagen by the absence of the banding pattern (48). However, until now no genes coding for such (glyco)proteins have been discovered. This is surprising, because at present close to 20,000 ESTs (expressed sequence tags) have already been annotated from S. domuncula to our sponge data base; a domain search in this base gave no indication of the existence of an additional structural/fibrillar protein other than collagen.
The functional interaction between galectin-2 and silicatein has been demonstrated by co-immunoprecipitation studies using antibodies raised against these proteins. These experiments revealed that silicatein (24 kDa) co-precipitates with galectin-2 (35 kDa) in extracts obtained from primmorphs, which had been exposed to 60 M silicate for more than 2 days. This interaction was abolished in the presence of a peptide corresponding to the hydrophobic C terminus of galectin-2. The assumption that this interaction could alter the biocatalytic activity of silicatein was strongly supported by enzymatic studies. In the presence of galectin-2 the amount of biosilica formed increased by more than 2-fold as judged by analytical determination, suggesting at least an additive effect. This substantial stimulatory effect of galectin-2 on silicateinmediated biosilica formation could be suppressed by the presence of the synthetic peptide.
For a further proof that in the native state galectin-2 and silicatein co-localize, a new extraction procedure was introduced. SDS-PAGE analysis of the axial filaments obtained by this method revealed that they comprise, in addition to the 23-25-kDa silicateins, further distinct proteins with sizes of 35 and Ͼ250 kDa. The 35-kDa protein was identified as galectin-2 in Western blots; the Ͼ250-kDa polypeptide remains unidentified. The calculated size of the processed form of galectin-2, without the short N terminus, is 32 kDa. This difference to the 35-kDa form is very likely due to the glycosylation of galectin-2; the evidence that galectin-2 is a glycoprotein came from staining studies of polyacrylamide gels with periodic acid-Schiff base reagent (not shown). The existence of galectin in the axial canal and also on the surface of the spicules has been additionally and unequivocally demonstrated here by immunofluorescence. The fact that in S. domuncula the axial filament FIGURE 11. The electron immunogold labeling technique, using pAb-aSILIC, was applied to identify silicatein in the extraspicular space. Applying this method, silicatein can be identified to be attached to the string-and net-like structures (A-E). A, the immunogold particles are arranged around the string-and net-like structures (st). B and C, at a higher magnification it is seen that the monomeric particles (ϩ) are arranged in an orderly and linear manner. Frequently, these pearl strings are self-associated (ϩϩ). C, in the vicinity of spicules the associates reaches sizes of ϳ400 nm. D and E, association of the string-and net-like structures carrying the silicatein molecule (as indicated by the gold particles) with collagen rods (col). E, a higher magnification of the view shown in D. F, as a control, these sections were treated with a preimmune serum. FIGURE 12. Schematic outline of the appositional growth of spicules from demosponges. In the mesohyl, galectin molecules associate in the presence of Ca 2ϩ to strings/nets that allow binding of silicatein molecules. Collagen fibers orient the silicatein-galectin strings concentrically round the growing spicules. In the last step, biosilica deposition is mediated in two directions, originating both from the silicatein-galectin strings and from the surface of the spicules (centripetal and centrifugal). Finally a third silica lamella (3) is formed, which is layered onto the previous two lamellae (1 and 2).
does not contain only one type of protein (silicatein), but also further proteins, had been demonstrated previously (8,9).
The finding that the filament present in the axial canal is not homogenous (does not consist only of silicatein) was supported by electron microscopic analysis. Sections through primmorphs revealed that the axial canal contains not only electron-dense material, very likely silicatein, but also membraneous structures, 0.5-4-nm-thick string-like structures and less electron-dense material. The thin strings exist in close association with the silicate shell of the spicules, both inside and outside of the spicules. They are markedly distinguished from the 10and 14-nm large solid rods, which are striated. Although the rods represent collagen, the thin string-and net-like structures very likely represent self-associated galectin. This assumption was supported experimentally by immunogold electron microscopy investigations. The images show that the silicatein molecules are closely linked to the stringand net-like structures. Only rarely is labeled silicatein, bound to the string-/net-like structures, associated with collagen fibers.
A further protein in the axial filaments is galectin, which undergoes self-association in the presence of Ca 2ϩ . In vitro assembly experiments followed by electron microscopic analysis showed that the self-associated proteins are positioned along an axis and not irregularly distributed. Also the next step, which is the formation of the galectin-2-containing string-and net-like structures, could be documented electron microscopically. Immunogold electron microscopic images led to the assumption that collagen fibers mediate the functional orientation of silicatein-galectin-2 complexes by arranging them concentrically around the longer axis of a growing spicule. Then, silicatein mediates silica deposition onto the surface of the existing silica layer. Because the surface of a new siliceous spicule is also covered with silicatein, the appositional growth/thickening of a spicule proceeds from two directions (centrifugal and centripetal) (8,9); a scheme that outlines the probable growth of the spicules is given in Fig. 12. The lamellar composition of siliceous spicules has not only been demonstrated for Hexactinellida (51) but also for Demospongiae (5).
These data will contribute to the potential application of silicatein in bio(nano)technology. In view of the finding that galectin is used by silicatein as a matrix for an extensive synthesis and deposition of biosilica, embedding of the enzyme in suitable organic polymers/aggregates that support the silicatein-mediated silica formation at ambient temperature will be the prime application for the enzyme.