Efficient silica synthesis from tetra(glycerol)orthosilicate with cathepsin- and silicatein-like proteins

Silicateins play a key role in biosynthesis of spicules in marine sponges; they are also capable to catalyze formation of amorphous silica in vitro. Silicateins are highly homologous to cathepsins L – a family of cysteine proteases. Molecular mechanisms of silicatein activity remain controversial. Here site-directed mutagenesis was used to clarify significance of selected residues in silica polymerization. A number of mutations were introduced into two sponge proteins – silicatein A1 and cathepsin L from Latrunculia oparinae, as well as into human cathepsin L. First direction was alanine scanning of the proposed catalytic residues. Also, reciprocal mutations were introduced at selected positions that differ between cathepsins L and silicateins. Surprisingly, all the wild type and mutant proteins were capable to catalyze amorphous silica formation with a water-soluble silica precursor tetra(glycerol)orthosilicate. Some mutants possessed several-fold enhanced silica-forming activity and can potentially be useful for nanomaterial synthesis applications. Our findings contradict to the previously suggested mechanisms of silicatein action via a catalytic triad analogous to that in cathepsins L. Instead, a surface-templated biosilification by silicateins and related proteins can be proposed.

to be non-efficient for silica polymerization catalysis and is commonly described as a feature of cathepsins L 3 . Site-directed mutagenesis confirmed the crucial role of Ser26 and His165 in silicateins 7,10 .
However, two recent works demonstrated that in spite of the presence of Cys in the active site, cathepsin L can possess silica-polymerizing activity in some conditions 11,12 . We hypothesized that previously described effects of substitutions of the catalytic residues were overestimated due to the suboptimal silica acid precursor used. Recently, a highly water soluble silica acid precursor tetra(glycerol)orthosilicate (TGS) was introduced that provides a convenient and efficient way to assess protein activity 13 . Here, to clarify the role of certain amino acid residues, activity of silicatein and cathepsin L mutants with TGS was systematically evaluated.

Results
Protein isolation and purification. Usually bacterially expressed silicateins are poorly soluble and require some extra tags to be produced in a soluble form 7 . Genes of two wild-type proteins from marine sponge Latrunculia oparinae were cloned, namely silicatein A1 (LoSilA1) and cathepsin L (LoCath), in several vectors, and different expression systems to express proteins were tested: pQE30 in XLBlue and BL21 strains, pBAD in BW and XJB strains, pET-40b(+) in XJB, BL and BL21codon+ strains. Also, different temperature (room temperature and 37 °C) and different concentration of inductors were tested. The conditions described in experimental section was the only one, which allowed to express proteins in soluble fraction. LoSilA1 and LoCath were still difficult to purify from the cell lysate by metal-affinity chromatography until His-tag was relocated to the C-terminus. The yield of the protein was about 1 mg per 1 liter of the cell culture. The purity of the proteins (more than 85%) was confirmed by the SDS-PAGE (representative example is shown in Fig. 2). Due to high protein similarity the same conditions were used for CTSL, but it was found that this protein should be expressed on 25 °C after induction.
To verify protein functionality scanning electron microscopy (SEM) was used. The data showed that both LoSilA1 and LoCath form similar amorphous silica particles with TGS (Fig. 3).  LoCath -were made. Notably, LoSilA1 and LoCath have very different residues upstream the catalytic Ser26 (or Cys26) -silicatein has GAS sequence at positions 24-26, while L. oparinae cathepsin L has KSC. So, triple mutants were made, where these moieties was reciprocally exchanged between the two proteins. In addition, Y27W substitution was introduced because this flanking residue is conservative in all the silica-condensing proteins, but differs in CTSL (Fig. 4).

Mutagenesis of sponge
In the previous work, pH optimum of LoSilA1 was measured to be 5.5 13 . LoCath has pI similar to that of LoSilA1, so both proteins were analyzed in the same conditions (PBS pH 5.5 at 0 °C). Kinetics of the silica formation was studied for LoSilA1 (0-120 min, with 20 min intervals). The reaction yield reached the maximum at 40 min (not shown). So, in the further experiments, 1-h reaction time was used to ensure completeness of the reaction together with low level of TGS spontaneous polymerization in this period of time 13 .
Surprisingly, all the mutants demonstrated silica polymerizing activity, even proteins with alanine in the active site (Fig. 5). LoSilA1 Ala-mutants (Q20A, S26A, H165A, N185A) were slightly less active, while LoCath-H165A mutant was a bit more effective than the corresponding wild-type proteins. Substitutions Q20A, C26A, and N185A did not alter LoCath activity. LoSilA1-G24K/A25S/S26C mutant was almost 3-fold more efficient than the wild type LoSilA1. Cathepsin mutant LoCath-K24G/S25A/C26S demonstrated a similar value of activity enhancement compared to the parental protein.
None of the mutations abolished silica condensing activity of the proteins. These unexpected results turned our attention to CTSL. We hypothesized that CTSL could also demonstrate silica-polymerizing activity with TGS as a silica acid precursor.    The data are represented by mean ± SD. The values were compared using two sample t-test (mutants vs the corresponding wild-type protein, wild-type LoCath was compared to wild-type LoSilA1), *corresponds to the p < 0.05. catalytic Cys and both flanking residues by the corresponding silicatein residues (SCW to ASY) 8 . Thus, three single-point mutants (G24A, C25S, W26Y), two double mutants (G24A/C25S and C25S/W26Y), and one triple mutant (G24A/C25S/W26Y) were obtained. Analysis of silica polymerizing activity both in PBS pH 5.5 at 0 °C and Tris-HCl pH 6.8 at RT was performed, because neutral Tris-HCl was a common buffer for the silicatein analysis in other works 1,7 . BSA and fluorescent protein mKate with 6His-tag were used as negative controls.
CTSL and its mutants showed silica polymerizing activity comparable to that of sponge silicatein A1 (Fig. 6). In the acidic conditions (PBS pH 5.5, 0 °C), mutants C25S and N187A demonstrated about 3-fold enhanced activity compared to the parental CTSL, whereas other mutants showed no significant difference from CTSL. In neutral conditions (Tris-HCl pH 6.8, 22 °C) only C25A and G24A/C25S showed the activity similar to the wild-type CTSL, while other mutants were more active than the wild-type protein; again, C25S and N187A variants were the most active. BSA showed some low-level activity, 10-and 30-fold lower than LoSilA1 in Tris-HCl and PBS, respectively. Silica polymerizing activity of mKate-6His was below the detection limit.
Secondary structure analysis. To analyze the effect of mutations on protein structure, the circular dichroism (CD) spectra were obtained for the wild-type proteins and some mutants at positions 23-26 (Fig. 7). CD spectra deconvolution showed that the wild-type proteins in the sodium-phosphate solution at pH 5.5 are well-folded and have a high content (>60%) of α-helix (Table 1). At the same time, three out of four studied mutants, namely LoCath-K24G/S25A/C26S, CTSL-G24A/C25S/W26Y, and CTSL-C25S showed strongly affected protein folding resulting in partial loss of α-helix elements (Table 1).

Figure 6.
Silica polymerizing activity of wild-type human cathepsin L (CTSL) and its mutants. All values were normalized to the wild-type LoSilA1. Reactions were performed in PBS pH 5.5 at 0 °C (A) and Tris-HCl pH 6.8 at room temperature (B). The absolute amount of the silica polymerized by LoSilA1 was 0.2 mM and 0.1 mM in PBS and Tris-HCl, respectively. The data are represented by mean ± SD. The values were compared using two sample t-test (mutants vs the corresponding wild-type protein, wild-type CTSL was compared to wild-type LoSilA1), *corresponds to the p < 0.05.

Discussion
In the present work we found that CTSL possesses silicatein-like activity and catalyzes silica polymerization. Analysis of the mutants of CTSL and L. oparinae LoSilA1 and LoCath showed that even mutants with the replacement of the one of the catalytic triad residues by alanine still have rather high silica-polymerizing activity. These observations contradict the generally accepted molecular mechanisms of silicatein enzymatic action 7-9,14 and call for further investigation and interpretation. Two reactions occur during silica condensation -hydrolysis of the silica acid precursor (Si-O bond hydrolysis) and condensation of the silica acid or partially hydrolyzed precursors (Si-O-Si bond formation) 15 . A protein can potentially catalyze: (i) hydrolysis, (ii) condensation, and (iii) both reactions.
It is difficult to discriminate between these possibilities as both hydrolysis and condensation can happen spontaneously, and thus one will observe protein-induced acceleration of formation of silica particles under any scenario. Nevertheless, there were several studies aiming to evaluate hydrolysis and condensation separately. Dakhili and coworkers studied silicatein α of S. domuncula with 4-nitrophenyl pivaloate or pivalamide as a substrate 10,16 . These compounds do not undergo further condensation, so, the experiment proved that silicatein is able to catalyze the hydrolysis reactions. Müller et al. demonstrated that the same silicatein could interact with bulky bis(p-aminophenoxy)-dimethylsilane which contains two Si-O and two S-C bonds, that confirms that silicatein could interact not only with orthosilicic acid itself 10,16 . Besides that, silicateins possess directly confirmed proteolytic activity, which is hydrolytic too [17][18][19] .
On the other hand, Fairhead et al. proved the silica condensing activity of the silicatein-cathepsin L chimera with sodium silicate. The hydrolysis step is absent is this system, the protein can catalyze only condensation reaction 8 .
Altogether these works demonstrated that the silicateins catalyze both the hydrolysis and condensation reactions. Mutations can affect both of them or one reaction only. In our experimental system, where TGS can be involved in every reaction, there is no possibility to distinguish effects. Due to spontaneous hydrolysis of TGS even mutants with no hydrolyzing activity could be defined as silica polymerizing.
Previous works on silicatein mutagenesis demonstrated a key role of Ser26 and His165 from the cathepsin-like catalytic triad, since S26A and H165A mutants failed to polymerize silica 7,10 . Both groups used tetraethyl orthosilicate (TEOS) as a silica acid precursor, which is poorly hydrolyzed in water at neutral pH. In contrast, here we used highly water soluble and hydrolysable TGS 13 that can potentially unmask the remaining silica condensing activity of the mutants.
Fairhead et al. performed mutagenesis of CTSL to make a chimerae with the silica-condensing activity 8 . A variant similar to our CTSL-G24A/C25S/W26Y mutant (called AS2 in their work) was the first mutant with notable silica-condensing activity. In the present work, silica-forming activity somewhat increased in the row CTSL, CTSL-G24A/C25S/W26Y, CTSL-C25S, but all these proteins demonstrated comparable activities (see Fig. 6). Again, differences between results obtained in the present work and in the paper by Fairhead et al. can probably be attributed to different substrates used -TGS versus sodium silicate, respectively.
Another line of research suggested catalytic triad-independent mechanism of silicatein activity. For example, synthetic block copolypeptides with silica condensation activity were obtained 20 . All the lysine-containing polypeptides showed some activity; the most effective were poly-L-lysine-poly-L-cysteine block copolypeptides. Later, silica condensing activity was shown for cysteamine and even ethanolamine 21 . It should be noted that catalysis by the copolypeptides and small bifunctional molecules required much higher concentrations of silica precursor and catalyst, but the resulting particles of amorphous silica looked similar to that formed by the silicateins. Also, a surface-templated biosilification was proposed to occur in glass sponges with highly hydroxylated collagen 22 , histidine-rich protein glassin 23 or even chitin [24][25][26] . The diversity of the involved polymers 27 suggests some common principle of the biosilification instead of the unique enzyme mechanism for every compound. Indeed, collagen and chitin as regular polymers probably cannot form specific active sites for silica condensation. Moreover, silicateins are known to condense a variety of compounds -titanium dioxide, zirconia dioxide, poly(L-lactide), barium oxofluorotitanate, calcium carbonate, silver, and ceria oxide or ceria-zirconia oxide nanocrystals [28][29][30][31][32][33][34] . Silicatein is usually entrapped in the center of the formed particle. It can be hypothesized that silicatein can orient these substances onto its surface and ensure a templated growth of the particles 28 .
Our findings support the mechanism of surface-templated silica condensation by silicateins and related proteins. Silica acid precursor TGS has relatively high level of the spontaneous hydrolysis. Thus, even in the absence of hydrolytic activity the enzyme still has enough silica acid to work with, and alanine substitutions of the catalytic triad could be not crucial. At the same time, mutations affect the charge of the protein surface, protein conformation and possibly oligomerization that can result in the increased silica polymerization. Importantly, it was demonstrated that silicatein-cathepsin L chimera loses its α-helix regions during the silica condensation 35 . In agreement with this work, there was a strong decrease in α-helix content in CD spectra of LoCath and CTSL mutants. Notably, the most crucial changes in the silica-condensing activity in our work were associated with substitutions at positions 24-26 in α-helix (Supplementary Fig. S2). Our data suggest that conformation of the protein affects their silica condensing or hydrolytic activity more than presence or absence of residues supposed to be catalytic.
A quite unexpected practical outcome of the present work was some mutants with enhanced silica-forming activity. In particular, sponge cathepsin L mutant LoCath-K24G/S25A/C26S possessed about 7-fold higher activity compared to the natural silicatein A1. Silicatein-directed formation of silica and other substances is considered to be a perspective way to nanomaterial synthesis with desired properties 32,[36][37][38] . Cathepsin L mutants generated here or developed in further works might be a useful addition to the available silicateins. Finally, our data raise a fundamental question of whether cathepsins L and silicateins belong to clearly different functional classes as it is currently thought.

Experimental procedures
Cloning and protein Purification. Mutations were introduced by the AQUA cloning technique 39 and confirmed by sequencing. Protein purification procedure was similar to described previously 13 . Genes were cloned into pET-40b(+) vector with C-terminal His-tag and without N-terminal tags (N-terminal DsbC-and His-tags were separated from the target proteins by two stop codons and a frame shift between them). Constructions were expressed in E. coli BL21-Codon+ strain. Expression was induced by 0.1 mM IPTG at 37 °C for 16 h for sponge proteins and at 25 °C for human cathepsin L. Cells were centrifuged, sonicated in 25 mM Tris-HCl pH 6.8, 150 mM NaCl with PMSF protease inhibitor (Thermo Fisher Scientific) and 5 mM DTT. Proteins were purified from the soluble fraction using Excel metal affinity resin (GE Healthcare Life Sciences), eluted with 25 mM Tris-HCl pH 6.8, 5 mM DTT, 200 mM imidazole, and stored at 4 °C for up to 3 days. Prior to the analysis, the proteins were transferred to assay buffer (25 mM Tris-HCl pH 6.8, 150 mM NaCl or PBS pH 5.5) using centrifugal filter units (Amicon ultra-4). Protein concentration was determined using optical density at 280 nm. Wild-type proteins, their mutants and reference silicatein A1 were isolated in parallel for each experiment. Purified protein samples were analysed using standard SDS-PAGE in 12% polyacrylamide gel stained with Coomassie Brilliant Blue G-250.
CD spectra measurements. The circular dichroism (СD) spectra of the proteins (0.08 mg/ml in 0.06 M sodium phosphate buffer pH 5.5) were recorded using a Chirascan spectrophotometer (Applied Photophysics, UK), equipped with a thermostatic cuvette holder, in a 0.05-cm path length quartz cuvettes at 20 °C. The bandwidth used was 1 nm with a scan time per point of 1 s. For each sample, two repeat scans were averaged and baseline-corrected by subtraction of the blank buffer. CD spectra deconvolution was performed using a CDNN (version 2.1) software tool (Applied Photophysics, UK) in the 195-260 nm spectral region because of its accuracy over the entire spectral region examined (i.e. with a total percentage sum closest to 100%). Substrate synthesis. Tetra(glycerol)orthosilicate (TGS) was synthesised as described previously 13 . It should be noted that TGS can not be accurately characterized in a standard way. We studied the 1H NMR spectra of the obtained substance immediately after the synthesis and after several variants of purification and found that it is an identical mixture in all cases. The NMR spectra of this mixtures characterized by two broad groups of multiplets in area of 4.3-4.8 ppm (corresponds to OH groups) and in area of 3.2-4.2 ppm (corresponds to CH and CH 2 groups) whose shape were more or less identical from one sample to another ( Supplementary Fig. S3). Probably, it represents a certain equilibrium mixture of different silicates of glycerol substituted by 1st and/or 2nd alcohol groups, containing different linear and cyclic, as well as monomeric and oligomeric derivatives. Formation of this equilibrium mixture occurs quickly enough, that does not allow us to purify the individual tetraglyceride, but the consistency of its composition was confirmed by the reproducibility of the results of the repeatedly performed synthesis and the identity of the obtained spectra. Thus, the name "TGS" should be treated not as the name of an individual compound, but as the ratio of the residues of glycerol to the silicon atom in a complex mixture.
As TGS was used without purification and the presence of free glycerol in obtained mixtures was revealed by NMR ( Supplementary Fig. S3), possible influence of glycerol on silica-condensing activity was tested. These control experiments showed that silica polymerizing activity of LoSilA1 was unaffected by as high as 10-fold excess of glycerol (0.1% TGS, 1% glycerol) (not shown).
Silica-condensing activity measurement. To evaluate silica-condensing activity, 1 ml of the 0.1% TGS solution was added to the 600 µl of the protein solution (0.06 mg/ml). After 1 h of incubation in the desired conditions samples were centrifuged, washed 3 times with 96% ethanol, air dried and dissolved in 200 µl 2M NaOH. The amount of the polymerized silica was determined by the colorimetric molybdate assay as described previously 13 . In spite of efforts to standardize experimental conditions, we observed rather strong difference in the absolute amount of the polymerized silica in different experiments. The probable reason for this is a poor stability of the purified proteins. To solve this problem, the wild-type protein, its mutants, and reference silicatein LoSilA1 were expressed, purified and measured in parallel in every experiment (4-5 independent experiments for all sets of mutants). Normalized data obtained in such a way were highly reproducible. Also, there was a negative control of spontaneous silica precipitation without any proteins in every experiment. Spontaneous TGS polymerization