Reconstituting the formation of hierarchically porous silica patterns using diatom biomolecules

Journal of Structural Biology xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Journal of Structural Biology journal homepage: www.elsevier.com/locate/yjsbi

Reconstituting the formation of hierarchically porous silica patterns using diatom biomolecules Damian Pawolskia, Christoph Heintzea, Ingo Meyb, Claudia Steinemb, Nils Krögera, a b

⁎

B CUBE, Center for Molecular and Cellular Bioengineering, TU Dresden, Arnoldstr. 19, 01307 Dresden, Germany Institut für Organische und Biomolekulare Chemie, Georg-August-University Göttingen, Tammannstr. 2, 37077 Göttingen, Germany

A R T I C LE I N FO

A B S T R A C T

Keywords: Biomineralization Biosilica Morphogenesis Long-chain polyamines Organic matrix Self-assembly

The genetically-controlled formation of complex-shaped inorganic materials by living organisms is an intriguing phenomenon. It illustrates our incomplete understanding of biological morphogenesis and demonstrates the feasibility of ecologically benign routes for materials technology. Amorphous SiO2 (silica) is taxonomically the most widespread biomineral, with diatoms, a large group of single-celled microalgae, being the most prolific producers. Silica is the main component of diatom cell walls, which exhibit species-specific patterns of pores that are hierarchically arranged and endow the material with advantageous properties. Despite recent advances in characterizing diatom biomolecules involved in biosilica morphogenesis, the mechanism of this process has remained controversial. Here we describe the in vitro synthesis of diatom-like, porous silica patterns using organic components that were isolated from biosilica of the diatom Cyclotella cryptica. The synthesis relies on the synergism of soluble biomolecules (long-chain polyamines and proteins) with an insoluble nanopatterned organic matrix. Biochemical dissection of the process revealed that the long-chain polyamines rather than the proteins are essential for efficient in vitro synthesis of the hierarchically porous silica patterns. Our results support the organic matrix hypothesis for morphogenesis of diatom biosilica and introduce organic matrices from diatoms as a new tool for the synthesis of meso- to microporous inorganic materials.

1. Introduction Diatoms are single-celled, eukaryotic algae that produce silica-based cell walls with species-specific morphologies. A hallmark of diatom biosilica is the presence of hierarchically arranged patterns of pores that have diameters in the range of ∼10 nm up to ∼1000 nm. The porous architecture endows the material with interesting optical properties including light confinement and selective optical transmission (De Tommasi et al., 2017). Despite the high porosity, diatom silica exhibits a remarkably high mechanical stability (Hamm, 2003; Aitken et al., 2016; Dimas and Buehler, 2012). Morphogenesis of diatom biosilica is therefore regarded as a paradigm for the synthesis of multifunctional low-density, high-strength materials. Previous research on the cell biology, molecular genetics, and biochemistry of diatom cell wall biogenesis has provided insight into the intracellular organization and molecular composition of the machinery

of biosilica formation. This process takes place inside the cell within lipid bilayer bound compartments called silica deposition vesicles (SDVs). The two types of biosilica building blocks (plate- or domeshaped valves and ring-shaped girdle bands) are produced in different SDVs during different stages of the cell cycle. Valves are produced during cell division and girdle bands during interphase when the cell expands. The deposition of solid silica and its shaping and patterning is accomplished entirely within the valve and girdle band SDVs. When morphogenesis is completed, the SDVs undergo exocytosis and the newly formed valve or girdle band is incorporated into the cell wall. Diatom biosilica is a hybrid material with amorphous SiO2 as the main component (∼90% by weight) and the remainder being organic components, which include unique proteins (e.g., silaffins, silacidins, cingulins, silicanins, SiMat proteins, SAPs), long-chain polyamines (LCPA), and polysaccharides (5). When the silica is dissolved using ammonium fluoride, silaffins, silacidins, and LCPA become solubilized, whereas

Abbreviations: AFSC, ammonium fluoride soluble components; DAP, 1,3-diaminopropane; DLA, diffusion-limited aggregation; LCPA, long-chain polyamines; NPI, Nmethylated propyleneimine; PAGE, polyacrylamide gel electrophoresis; PI, propyleneimine; SDS, sodium dodecyl sulfate; SDV, silica deposition vesicle; EDTA, ethylenediaminetetraacetic acid; TMOS, tetramethyl-orthosilicate ⁎ Corresponding author. E-mail addresses: [email protected] (D. Pawolski), [email protected] (C. Heintze), [email protected] (I. Mey), [email protected] (C. Steinem), [email protected] (N. Kröger). https://doi.org/10.1016/j.jsb.2018.07.005 Received 11 May 2018; Received in revised form 6 July 2018; Accepted 7 July 2018 1047-8477/ © 2018 Elsevier Inc. All rights reserved.

Please cite this article as: Pawolski, D., Journal of Structural Biology (2018), https://doi.org/10.1016/j.jsb.2018.07.005

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

and Hildebrand, 2013). On the other hand, circumstantial experimental evidence has been presented that the insoluble organic matrices from the diatom Thalassiosira pseudonana may be largely covered by silica (Kotzsch et al., 2016; Kotzsch, 2017), which would only be possible if the organic matrices are present during silica formation inside the SDV. When exposed to a metastable solution of silicic acid, silica was rapidly deposited on the surfaces of T. pseudonana organic microrings (i.e., organic matrices derived from girdle band biosilica), yet no porous silica patterns were observed (Scheffel et al., 2011). It has previously been hypothesized that the soluble biosilica associated components (e.g., silaffins, LCPA, silacidins) need to bind to the insoluble organic matrices to template the formation of porous biosilica patterns (Scheffel et al., 2011). Here we have investigated this hypothesis using the valvederived organic matrix and the soluble biosilica-associated components from the diatom Cyclotella cryptica.

cingulins, silicanins and SiMat proteins are part of insoluble organic matrices which exhibit mineral-free nanopatterns that match structural features of the silica (Hildebrand et al., 2018). Previously, several hypotheses have been put forward regarding the mechanism for biosilica morphogenesis. These can be grouped into two main categories: (i) template-independent mechanisms, which rely on diffusion limited aggregation (DLA) (Gordon and Drum, 1994; Parkinson et al., 1999) or a reaction-diffusion system (Willis et al., 2010) and (ii) template-dependent mechanisms (Robinson and Sullivan, 1987; Schmid, 1994; Sumper, 2002; Lenoci and Camp, 2008; Kröger, 2007). In the following paragraph, these hypotheses are briefly summarized. (i) Template-independent morphogenesis: The DLA process assumes that 1–10 nm-sized silica nanoparticles are imported into the leading edge of the growing SDV lumen, subsequent diffusion throughout the SDV, and nucleation of silica deposition in the SDV center. Computer simulations of this process demonstrated the formation of radial, branched ribs of silica that resemble the spoke-like silica patterns of diatoms (Gordon and Drum, 1994; Parkinson et al. 1999). However, the process is unable to explain the formation of cross-connections between the ribs and the pore patterns in the silica plates. In contrast, computer simulations based on the Turing equation (reaction-diffusion system) have been shown to generate patterns of pores within pre-existing large pores (Willis et al., 2010). It is yet unknown whether reaction-diffusion systems would be able to explain morphogenesis of the basic diatom biosilica architecture that is made of cross-connected ribs. (ii) Templatedependent morphogenesis mechanisms rely on biomolecular clusters that self-assemble into patterned organic scaffolds with long-range order. The assembly of the scaffold may occur within the SDV lumen (i.e. internal patterning) (Sumper, 2002; Lenoci and Camp, 2008; Kröger, 2007) or on the cytoplasmic surface of the SDV (i.e. external patterning) (Robinson and Sullivan, 1987; Schmid, 1994). External patterning is thought to involve the cytoskeleton and other intracellular structures (mitochondria, spacer vesicles). Microtubules, actin, mitochondria and spacer vesicles are believed to mold the SDV membrane into the species-specific shapes including indentation patterns that template the pores (Hildebrand et al., 2018; Schmid, 1994). Cytoskeleton filaments have been hypothesized to mediate the positioning of transmembrane proteins that are able to nucleate silica formation in the SDV lumen (Robinson and Sullivan, 1987). In contrast, internal patterning mechanisms are believed to be governed by some of the previously identified biosilica associated biomolecules (see above). In vitro experiments have demonstrated that mixtures of silaffins and LCPA undergo phase separation generating biomolecule-rich liquid droplets within the bulk aqueous solution (Poulsen et al., 2003; Poulsen and Kröger, 2004; Sumper and Brunner, 2008). It has been proposed that the formation of hierarchical pores in the diatom genus Coscinodiscus is the result of an iterative process of liquid-liquid phase separation in hexagonal arrays of LCPA-rich droplets inside the SDV (Sumper, 2002). Computer simulations of the phase separation process were able to generate diatom-like porous patterns albeit no hierarchical pores were obtained (Lenoci and Camp, 2008). In an alternative model, it was proposed that the porous basal layer of biosilica is templated by a silaffin-LCPA matrix that spans the entire lumen of the SDV. The porous biosilica patterns are believed to reflect the distribution of activating and inhibiting silaffin-LCPA clusters (Poulsen and Kröger, 2004; Kröger and Sandhage, 2010) that define areas of silica deposition and silica free sites, respectively, within the matrix (Kröger, 2007). The presence of insoluble organic matrices with nanopatterns that match structural features of the silica in diatoms seems to support the model of template-directed silica morphogenesis within the SDV lumen (Reimann et al., 1966; Brunner, 2009; Scheffel et al., 2011; Tesson and Hildebrand, 2013; Kotzsch et al., 2016). However, it has been argued that the insoluble organic matrices may be identical to the diatotepum, which is an organic layer that is added onto the biosilica after SDV exocytosis and thus cannot be involved in silica morphogenesis (Tesson

2. Materials and methods 2.1. Chemicals Tetramethyl-orthosilicate (TMOS), Tris base, cyanoborohydride, formaldehyde, and ammonium formate, chitinase (from Streptomyces griseus) and proteinase K (from Tritirachium album) were purchased from Sigma-Aldrich. NH4F, 37% HCl, ethylenediamine tetraacetic acid (EDTA), 85% glycerol, acrylamide, ammonium acetate, β-mercaptoethanol, formic acid (98–100%), acetic acid (p.a.), acetonitrile (Prepsolv®) and sodium dodecyl sulfate (SDS) were purchased from Merck. Tricin base and chloroform were purchased from Roth. Tetramethylethylendiamin and ammoniumpersulfate were purchased from Bio-Rad. Pronase (from Streptomyces griseus) was purchased from Roche and methanol from VWR. MilliQ-purified H2O (resistivity: 18.2 MΩ cm) was used throughout this study except for mass spectrometry analysis for which LC-MS grade H2O (LiChrosolv®) from Merck was used. 2.2. Culture conditions C. cryptica strain CCMP332 was grown in an enriched artificial seawater medium according to the Canadian Center for the Culture of Microorganisms at 18 °C under constant light at 5000–10,000 lx for 20–26 days. 2.3. Isolation of ammonium fluoride soluble components (AFSC) C. cryptica biosilica was isolated according to the same method as described previously for T. pseudonana (Poulsen and Kröger, 2004) and incubated with 10 M NH4F (adjusted to pH 4.5 with 6 M HCl) for 1 h at room temperature to dissolve the silica. After centrifugation (30 min, 3200g) the supernatant was dialyzed against 100 mM ammonium acetate (MWCO 6–8 kDa, Spectra/Pore RC) and concentrated using an Amicon Ultra-15 centrifugal filter unit (MWCO 3 kDa; Merck). The concentrated solution contained the AFSC. The proteins in the AFSC were quantified via their phosphate content according to a previously described method (Buss and Stull, 1983). 2.4. Preparation of LCPA and LCPA-free AFSC proteins To isolate LCPA and prepare LCPA-free AFSC proteins, the isolated AFSC was separated into LCPA and proteins using the following protocol. The lyophilized AFSC was resuspended in 2 M NaCl, centrifuged (4 °C, 30 min, 3200g), filtered through a polyethersulfone syringe filter (pore size 0.2 μm; Carl Roth) and subjected to gel filtration chromatography on a Superose 12 10/300 GL column (GE Healthcare) equilibrated with 200 mM ammonium formate (flow rate: 0.4 ml/min). Mixtures of AFSC proteins and LCPA eluted between 17.5 and 43.5 min (fraction F1), and pure LCPA eluted between 45 and 48.5 min (fraction 2

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

Fig. 1. Electron microscopy analysis of valve biosilica from C. cryptica. (A) TEM image of a single valve. The boxed region is shown in (B) at higher magnification. (C, D) Details of SEM images of the (C) distal surface of a valve and the (D) proximal surface of another valve. The symbols in the images indicate the positions of characteristic biosilica structures: white triangle = triangular sector of the valve; yellow circle = fultoportula at valve rim; red circle = fultoportula near valve center; yellow dotted line = wide costa; green dotted line = narrow costa; orange arrowhead = cribrum pore bordered by wide costa; red arrow = cribrum pore bordered by narrow costa, blue ring = areola pore.

microscopy (TEM). The majority of the H2O was removed by blotting with a piece of filter paper, and the samples were air-dried before imaging. To prepare the insoluble organic matrices, the wet, surface-adsorbed biosilica was overlaid with 10 M NH4F (adjusted to pH 4.5 with HCl) and incubated for 1 h at room temperature, followed by extensive washing with H2O. The surface-adsorbed organic matrices were kept hydrated for re-mineralization experiments (see below) and were air dried for imaging. Secondary electron microscopy images were taken using a JSM 7500F field emission scanning electron microscope (Jeol) at an acceleration voltage of 5 kV for biosilica and 1 kV gentle beam for insoluble organic matrices and re-mineralized insoluble organic matrices. For transmission electron microscopy objects were immobilized on goldcoated finder grids (G200F2-Au from EMS), and images were taken using a Morgagni 268D (FEI) instrument at an acceleration voltage of 80 kV.

F2). Fraction F1 was freeze-dried, redissolved in 100 mM ammonium acetate supplemented with 2 M NaCl, and incubated for 2 h at 0 °C to disassemble protein-LCPA aggregates. The solution was then filtered through a polyethersulfone membrane (pore size 0.2 μm; Carl Roth) and subjected to gel filtration chromatography on a Superdex peptide 10/ 300 GL column (GE Healthcare) equilibrated with 2 M NaCl in 100 mM ammonium acetate (flow rate: 0.4 ml/min). LCPA-free AFSC proteins eluted between 17.5 and 28.5 min (fraction F3), and pure LCPA eluted between 28.5 and 48.5 min (fraction F4). Fraction F3 and fraction F4 were desalted and concentrated using Amicon Ultra-0.5 ml membranes (MWCO 3 kDa, Merck) and a buffer consisting of 100 mM ammonium acetate. The desalted fraction F3 constituted the LCPA-free C. cryptica AFSC protein. The desalted fraction F4 was combined with the LCPA from fraction F2 constituting the C. cryptica LCPA. Analysis by SDSPAGE with Coomassie and “Stains All” staining confirmed the purity of the LCPA and AFSC proteins. 2.5. Preparation of biosilica and insoluble organic matrices for electron microscopy

2.6. SDS extraction of insoluble organic matrices

Isolated biosilica was critical point dried using the Leica CPD 300 instrument (Leica Microsystems). The dried biosilica was mounted on carbon pads on aluminum stubs and sputter coated with platinum using a Baltec SCD 050 instrument and argon as the process gas (40 mA, 40 s). To detach girdle bands from valves, C. cryptica biosilica (from 2 × 108 cells) was suspended in 1 ml H2O in a 1.5 ml tube and sonicated with an MS72 sonotrode tip (Bandelin) applying a total of 0.24 kJ over 20 s. A drop of sonicated biosilica was spotted on a platinum-coated polycarbonate membrane (0.2 µm pore size; Whatman) for scanning electron microscopy, and on Formvar-coated gold grids (EMS) strengthened with evaporated carbon for transmission electron

Isolated biosilica was incubated with 10 M NH4F (adjusted to pH 4.5 with 6 M HCl) for 1 h, and subsequently, the insoluble organic matrices were pelleted by centrifugation (7 min at 3200g) and the supernatant (AFSC) was collected. The pellet was washed with H2O by three resuspension-centrifugation cycles and then extracted for 1 h with 2% SDS at 95 °C. The SDS extract was collected by centrifugation (7 min at 3200g). The AFSC and the SDS extract were dialyzed (MWCO of 500 Da) twice against 100 mM ammonium acetate overnight and subsequently lyophilized. The lyophilized material was dissolved in 100 mM ammonium acetate and analyzed by SDS-PAGE. To detect LCPA, the extracts were subjected to 16% Tris/Tricine SDS-PAGE 3

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

Fig. 2. TEM analysis of organic microplates from C. cryptica. (A) Image of a single microplate that exhibits valve biosilica-like morphology. The boxed region of (A) is shown in (B) at higher magnification. (C) Image of a single microplate that lacks the areola-like electron dense walls in large areas of the surfaces. The boxed region in (C) is shown in (D) at higher magnification. The symbols in the images indicate the positions of characteristic structural elements that resemble structural elements in the biosilica: white triangle = triangular sector; yellow circle = fultoportula-like element at the rim of the microplate; yellow dotted line = structure resembling a wide costa; red arrow = electron translucent circle resembling a cribrum pore; blue ring = electron dense wall resembling an areola pore.

2.9. Atomic force microscopy

followed by Coomassie staining. To prevent LCPA from diffusing out of the gel, the gel was fixed with 5% glutaraldehyde for 1 h prior to staining. To detect proteins, the extracts were subjected to 6% Tris/ Tricine SDS-PAGE followed by staining with the dye “Stains-All” (Goldberg and Warner, 1997).

Surface-adsorbed biosilica, organic matrices, and re-mineralized organic matrices on TEM grids were immobilized in a petri dish using a two-component biocompatible glue (JPK Instruments). The AFM measurements were carried out with a NanoWizard 4 (JPK Instruments) utilizing the QI™ mode with a trigger force of 250 pN. Biolever mini (BL-AC40-TS, Olympus Micro) cantilevers were used. Calibration of the cantilevers was done by the contact-free method as provided by the manufacturer (JPK Instruments). Localization of the samples on the TEM grid prior to AFM imaging was achieved using a BioMat Station (JPK Instruments) in combination with an Olympus BX51 microscope equipped with a 40 × water immersion objective). Image correction and flattening was performed within the JPK data processing software (JPK Instruments).

2.7. Re-mineralization of insoluble organic matrices For re-mineralization experiments, silicic acid was prepared freshly by hydrolysis of TMOS (1M TMOS in 1 mM HCl, 15 min constant shaking at room temperature). Immediately prior to use, the silicic acid solution was diluted to a final concentration of 100 mM by combining 100 µl of the solution with 1.9 ml 210 mM sodium acetate buffer pH 5.5. Where necessary, AFSC (final phosphate concentration: 64 µM), AFSC proteins (final phosphate concentration: 56 µM), or LCPA (final concentration: 1 µM) or a mixture of both were added to the silicic acid solution immediately prior to use. Hydrated, surface-adsorbed insoluble organic matrices were then incubated with the silicic acid-containing solutions for 5 min at room temperature. After extensive H2O washing, the samples were air dried at room temperature.

2.10. Localisation of chitin Insoluble organic matrices immobilized on TEM grids (see above “Preparation of biosilica and insoluble organic matrices for electron microscopy”) were incubated for 1 h with 1 mg/ml chitinase in PBS (50 mM phosphate buffer, pH 7.4, 150 mM NaCl) or in PBS as a negative control. After extensive washing with PBS, the samples were incubated for 1 h with 1% BSA in PBS, followed by 1-hour incubation with 80 ng/µl of a GFP-tagged chitin binding protein (Weiss and Schönitzer, 2006) in a PBS solution containing 1% BSA. After extensive washing with PBS, the samples were imaged using epi-illumination with a 488 nm laser at 10 mW and respective filter sets (laser bandpass: 475/35, dichroic longpass: H 488 LPXR and emission bandpass: 525/45). The fluorescence intensities were adjusted to prevent saturation of the detector. GFP and bright field z-stage images-series were subsequently acquired

2.8. Electron dispersive x-ray analysis The sample loaded TEM grids were immobilized on carbon adhesive tabs (Science Services) immobilized on specimen stubs and imaged with a Supra 40VP scanning electron microscope (Zeiss) equipped with a Quantax XFlash 6/100 detector (Bruker), at an acceleration voltage of 6 kV. The data obtained were analyzed with the Esprit 2.0 software (Bruker). 4

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

Fig. 3. AFM images of organic microplates from C. cryptica. Images of single organic microplates that exhibit triangular sectors (white triangles) and (A) a pore dominated surface structure, or (C) a nodule dominated surface structure. Images (B) and (D) are magnifications of the boxed areas in (A) and (C), respectively. White dotted line = trench derived from wide costa; blue ring = areola-like pore, white arrowhead = nodule.

2.12. Amino acid analysis of the insoluble organic matrices

with the NIS-Elements software (Nikon) using an EMCCD camera (Ixon Ultra 897; Andor) mounted on an inverted fluorescence microscope (NSTORM, Nikon) equipped with a 100x oil objective (CFI TIRF Apochromat, NA 1.49, WD 0.12 mm, Nikon) and an autofocus system (Nikon) at an exposure time of 100 ms and 4 frames per µm. Using the software NIS-Elements, the taken frames were combined to maximum projections displaying the highest value of each pixel in all frames of the z-stacks.

Biosilica was demineralized with 10 M NH4F (adjusted to pH 4.5 with 6 M HCl) for 1 h at room temperature. The insoluble organic matrices were pelleted by centrifugation for 7 min at 3200g, and the pellet was washed with H2O by four resuspension-centrifugation cycles. The insoluble organic matrices were incubated for 22 h at 37 °C with 1 mg/ ml Pronase in H2O. Subsequently, the Pronase solution was replaced with 1 mg/ml Proteinase K in PBS (50 mM phosphate buffer, pH 7.4, 150 mM NaCl) and incubated for an additional 22 h at 37 °C. As a control, the same amount of insoluble organic matrices were incubated for 44 h in PBS at 37 °C. After intensive washing, the samples were lyophilized and weighed. Amino acid analyses of the samples were performed by the Functional Genomics Center Zürich.

2.11. Quantification of silica formation activity in solution The proteins of the AFSC protein fraction were quantified via their phosphate content using a previously described phosphate assay (Buss and Stull, 1983). The concentration of the LCPA was determined using the 660 nm Protein Assay (Pierce) with the synthetic oligopropyleneimine dendrimer DAB-Am-16 (Sigma-Aldrich) as a standard. By comparing the intensities of LCPA bands in Coomassie-stained SDS PAGE analysis of AFSC and isolated LCPA it was determined that 100 µM AFSC (phosphate concentration) contained 1.6 µM LCPA. In a final reaction volume of 75 µl, 200 mM sodium acetate buffer pH 5.5 and the desired amount of soluble organic components (AFSC, AFSC proteins, LCPA, or a mixture of AFSC proteins and LCPA) were incubated with a final concentration of 100 mM silicic acid (freshly prepared by hydrolysis of 1 M TMOS in 1 mM HCl for 15 min). After 5 min at room temperature, silica was pelleted by centrifugation at 16,000g for 5 min, washed three times with H2O, and dissolved for 1 h at 95 °C using 100 µl 2 M NaOH. The silica concentration was determined using the β-silicomolybdate method (Iler, 1979).

2.13. Analysis of LCPA Dried LCPA were dissolved in 1% acetic acid, and mixed 1:100 with a solvent mixture consisting of 50% acetonitrile, 5% formic acid, and 45% H2O. The sample was analyzed by mass spectrometry using a static nanoelectrospray with a Nanomate Triversa ion source (Advion, Ithaca, USA). Mass spectra (MS1) were acquired in high resolution (R = 100000 at m/z 400) with an LTQ Orbitrap XL ETD mass spectrometer (Thermo Scientific, Bremen, Germany). Reductive permethylation of LCPA was performed according to a published protocol (Jentoft and Dearborn, 1983). Briefly, dried LCPA were dissolved in 50 mM sodium phosphate pH 7.0, and reacted with 150 mM sodium cyanoborohydride and 1.5 M formaldehyde (freshly prepared by heating a paraformaldehyde solution at 110 °C overnight) for 16 h. Subsequently, the sample was extracted with methanol/ chloroform (3:2), and the organic phase containing the permethylated 5

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

shaped rings (i.e. areolae pores; blue rings in Fig. 1B and C). Each areolae pore encloses multiple cribrum pores, thereby constituting a hierarchical pore pattern. SEM analysis revealed that the patterns of wide and narrow costae, cribrum pores, and fultoportulae are present on both the distal side (Fig. 1C) and the proximal side of the valve (Fig. 1D). However, on the proximal side silica bridges between the costae are entirely absent, and thus areolae pores are lacking. Following demineralization of the C. cryptica biosilica using ammonium fluoride, we examined the structures of the valve-derived organic microplates. Energy-dispersive X-ray (EDX) analysis confirmed that the organic microplates are free of silica (Fig. S1A and B). In TEM analysis of the organic microplates, triangular sectors were clearly visible (white triangle in Fig. 2A). Each triangular sector contained many small electron-transparent circles (diameters ∼14 nm; red arrows in Fig. 2B and D), and groups of these were fully enclosed by electrondense walls (blue rings in Fig. 2B), thus closely resembling the hierarchical arrangement of cribrum pores and areolae pores in the valve biosilica (see Fig. 1B above). In a substantial number of organic microplates the electron dense walls were absent from large areas or were even entirely lacking (Fig. 2C and D). Presumably, the structural variations among the organic microplates from different cells indicate differences in organic composition. This might have been caused by inhomogeneities during the ammonium fluoride treatment that led to variations in the solubilization of organic matrix components. To image the organic microplates under fully hydrated conditions, AFM measurements were conducted. This analysis again confirmed the presence of the triangular sectors (white triangle in Fig. 3A and C). Neighboring sectors were bordered by 100–350 nm wide trenches (white dotted lines in Fig. 3B and D) that were 20–40 nm lower than the interior of the triangles. Regarding the morphology of the interior of the triangles, two different structural types were observed. One type exhibited pores that were about 50–70 nm wide and thus resembled the areola pores (blue circles in Fig. 3B). The other type exhibited 10–60 nm sized nodules (white arrowheads in Fig. 3 D) that were much more densely packed inside the triangular sectors rather than in the trenches, and hierarchical pore patterns appeared to be absent. We assume that the two structural types represent either the two different sides of an organic microplate or they represent organic microplates with somewhat different organic composition. A biosilica-associated chitin meshwork has previously been identified in the diatom T. pseudonana and is believed to be involved in biosilica formation (Brunner et al., 2009). Using fluorescent probes chitin has been located to diatom girdle bands (Durkin et al., 2009) and girdle band associated organic microrings (Scheffel et al., 2011) in T. pseudonana. To investigate the presence of chitin in the organic microplates of C. cryptica, a GFP fusion protein bearing a chitin-binding domain (CB-GFP) was employed. CB-GFP has previously been used to localize chitin in mollusk shells (Weiss and Schönitzer, 2006; Nudelman et al., 2007). In the C. cryptica microplates, the CB-GFP fusion protein binds to the rim rather than the interior surface (Fig. S2A). After chitinase treatment of the organic microplates no binding of CB-GFP was observed (Fig. S2B). These results show that the rim of the organic microplates and/or organic microring fragments attached to the microplate rim contain chitin, whereas chitin is entirely absent from most of or even the entire organic microplate. Protease treatment of the insoluble organic matrices lead to a 60% decrease in dry mass, which indicated that proteins are the main components of the organic microplates and microrings of C. cryptica. Amino acid analysis revealed that glycine (17 mol-%), serine (15 mol-%), asparagine/aspartic acid (9 mol-%) and glutamine/glutamic acid (8 mol-%) constitute almost 50% of the standard amino acids of the insoluble organic matrices (Table S1).

Fig. 4. SDS-PAGE analysis of biosilica-associated organic components from C. cryptica. AFSC contains the biomolecules that became soluble after dissolving the biosilica with ammonium fluoride. SSMC contains the biomolecules that were extracted from the ammonium fluoride insoluble organic matrices by treatment with SDS. (A) 6% Tris-Tricine SDS-PAGE followed by staining with “Stains All”. The blue colored bands in “Stains All” stained gels are indicative of polyanionic proteins (32). The arrows point to protein bands that are present in both the AFSC and SSMC. (B) 16% Tris-Tricine SDS-PAGE followed by staining with Coomassie Blue. The bracket indicates the apparent molecular mass range of actual and proposed LCPA molecules in the AFSC and SSMC extracts, respectively.

LCPA were dried by a continuous stream of nitrogen. The residue was dissolved and analyzed by MS as described above. 3. Results 3.1. Structure and chemical composition of the organic microplates from Cyclotella cryptica The structure of C. cryptica biosilica has previously been described in detail (Reimann et al., 1963; Tesson and Hildebrand, 2010). Here we briefly summarize the characteristic structural features of the valve biosilica. In TEM analysis, the Cyclotella cryptica valve biosilica (Fig. 1A and B) is comprised of four main structural elements: costae, cribrum pores, areolae pores, and fultoportulae. A fultoportula is a tubular process through the valve. There are 12–19 fultoportulae at the rim of the valve (yellow circles in Fig. 1A, C and D) and often 1–2 fultoportulae close to the valve center (red circle in Fig. 1A). Each peripheral fultoportula is the end point of a pair of 90–350 nm wide pore-free ribs (i.e. wide costae) that merge near the valve center (yellow dotted lines in Fig. 1B–D). Between each pair of wide costae runs a row of small, circular pores, which are called cribrum pores and have diameters of ∼30 nm (orange arrowheads in Fig. 1B–D). Neighboring pairs of wide costa enclose a roughly triangular sector of the valve (white triangle in Fig. 1A). Each triangular sector is infiltrated with rows of cribrum pores (red arrows in Fig. 1B) and contains 2–5 narrow costae (30–50 nm wide) that run parallel to the wide costae and merge with these near the valve center (green dotted lines in Fig. 1B–D). Neighboring narrow costae are often linked by silica bridges thus generating irregularly

3.2. Re-mineralization of organic microplates in vitro The structural resemblance of the organic microplates with the 6

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

Fig. 5. SEM and TEM images of re-mineralized organic microplates. Re-mineralization was performed with 100 mM silicic acid in 50 mM sodium acetate pH 5.5 in the absence (A, C, E) or in the presence of AFSC (B, D, F). Images A and B are from SEM analysis and images C-F from TEM analysis. The boxed region in (C) and (D) are shown in (E) and (F), respectively. The red arrows point to cribrum-like pores and the blue rings indicate areola-like pores. The yellow dotted lines indicate regions that resemble wide costae.

during incubation with silicic acid. The Si signal in the EDX analysis of native biosilica valves (Fig. S1G and H) was much stronger than the Si signal of the re-mineralized organic microplates. This indicated that the re-mineralization conditions in vitro were unable to match the silica deposition efficiency of the in vivo process. Inspection with SEM revealed silica coated microplates containing numerous circular pores with diameters of ∼15 nm (Fig. 5A). The silica on the microplates that were incubated with both silicic acid and AFSC exhibited a hierarchical pore morphology that consisted of groups of cribrum-like circular pores that were enclosed by irregularly shaped, areolae-like pores (Fig. 5B). TEM analysis of re-mineralized organic microplates confirmed the

valve biosilica seems to support the hypothesis that insoluble organic matrices act as templates for biosilica morphogenesis. To test this, surface adsorbed insoluble organic matrices from C. cryptica were briefly incubated with silicic acid (5 min at pH 5.5) both in the presence and in the absence of the ammonium fluoride soluble components (AFSC). The AFSC of C. cryptica contained several polyanionic proteins with apparent molecular masses from ∼20 kDa to > 170 kDa and LCPA (Fig. 4). After removal of silicic acid by washing, the organic microplates were investigated by EDX analysis demonstrating the presence of a silica layer on the surface of the microplates irrespective of whether the AFSC were absent (Fig. S1C and D) or present (Fig. S1E and F) 7

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

presence of porous silica in material that was exposed to silicic acid in the absence (Fig. 5C and E) or the presence of AFSC (Fig. 5D and F). The cribrum-like pores in the re-mineralized organic microplates had an average diameter of 17.5 ± 3.2 nm and thus were about half the size of the cribrum pores in the native biosilica (average diameter: 31.1 ± 4.3 nm; Fig. S3A). Regarding surface area distribution, the areola-like pores on the re-mineralized organic microplates almost perfectly matched the areola pores in the native biosilica (Fig. S3B). This result demonstrated that a diatom-like silica morphology can be generated in vitro through the combined action of biosilica-derived soluble biomolecules and insoluble organic matrices. However, the SEM images of re-mineralized microplates were devoid of silica structures that resembled wide costae (Fig. 5B). Also in TEM images the regions corresponding to the locations of wide costae exhibited relatively high electron transparency (Fig. 5D and F) indicating very low or complete absence of mineralization. When analyzing large numbers of TEM images it became apparent that hierarchically porous silica patterns were occasionally also present in material that was re-mineralized in the absence of AFSC. Furthermore, re-mineralization in the presence of AFSC did not always result in hierarchically porous silica throughout the microplate. To quantify the outcomes of the experiments, the structures of re-mineralized microplates were differentiated into two categories: “complete hierarchy” and “incomplete hierarchy”. Objects with complete hierarchy exhibited triangular sectors containing cribrum-like pores surrounded by areolae-like pores throughout the surface of the re-mineralized microplate (Fig. S4A). In objects with incomplete hierarchy, not all of the triangular sectors contained areolae-like pores (Fig. S4B) or these pores were entirely absent (Fig. S4C). Upon re-mineralization in the presence of AFSC, 70% of the objects showed complete hierarchy, whereas in the absence of AFSC it was only 34% (Fig. 6). This demonstrated that the AFSC biomolecules strongly enhance the capability of the organic microplates to generate hierarchical pore patterns upon re-mineralization in vitro. We then aimed at identifying biomolecules of the AFSC that are particularly effective in promoting the formation of hierarchically porous silica patterns in vitro. Therefore, the AFSC was subjected to gel permeation chromatography yielding an AFSC protein fraction that lacked LCPA (Fig. S5A) and an LCPA fraction that lacked AFSC proteins (Fig. S5B). In a homogeneous solution (i.e. in the absence of insoluble organic matrices) both the AFSC proteins and the LCPA exhibited silica formation activity. The silica formation activity of the total AFSC was only slightly higher (on average +15%) than the sum of the activities of the LCPA and the AFSC proteins alone (Fig. 7). This indicated that only a small amount of silica forming activity was lost during chromatography. In the presence of the ASFC protein fraction, only 12% of the re-mineralized microplates exhibited complete hierarchy (n = 60; Fig. 6), whereas this amount was 60% when using the LCPA fraction (n = 60; Fig. 6). When the AFSC was reconstituted by combining the LCPA and ASFC protein fractions, re-mineralization of the organic microplates yielded complete hierarchy in 68% of the objects (n = 79; Fig. 6). These results indicate that during re-mineralization of the organic microplates in vitro LCPA rather than the AFSC proteins are the soluble components that promote the formation of hierarchical silica pore patterns. Analysis by mass spectrometry (MS) demonstrated that the C. cryptica LCPA are composed of a mixture of molecules that contain only C, H, and N atoms and have masses ranging from 458 to 642 Da (Fig. 8A, Table S2). Molecules of such masses have previously been observed in the LCPA fractions of other diatoms, where they were shown to be putrescine- or propyldiamine-based polyamines containing linearly N-C-linked propylenimine units that are partially or fully Nmethylated (Kröger et al., 2000; Sumper et al., 2005). Fragmentation analysis by tandem mass spectrometry (MS/MS) confirmed the presence of propylenimine (PI) and N-methylated propylenimine (NPI) residues within the molecules (C. Heintze and N. Kröger, unpublished

Fig. 6. Yield of hierarchically porous silica patterns using different soluble additives in re-mineralization of organic microplates. Re-mineralization was performed with 100 mM silicic acid in 50 mM sodium acetate pH 5.5 in the presence of the indicated additives. Re-mineralized microplates containing hierarchically porous patterns in all triangular sectors exhibited “complete hierarchy” (black bars). Organic microplates that lacked hierarchically porous patterns in one or more triangular sectors exhibited “incomplete hierarchy” (grey bars). For each re-mineralization condition, the fractions of re-mineralized organic microplates exhibiting complete and incomplete hierarchy are expressed in % of the number of objects that were analyzed. For each re-mineralization condition > 50 objects were analyzed. The asterisks indicate pairs of re-mineralization results (brackets) that show statistically valid differences (p < 0.001) regarding the yields of microplates with complete and incomplete hierarchy.

Fig. 7. Silica formation activities of AFSC molecules in vitro. Components were incubated with 100 mM silicic acid in 50 mM sodium acetate pH 5.5 for 10 min. Pelleted silica was dissolved in NaOH and quantified using the molybdate assay.

8

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

Fig. 8. Analysis of C. cryptica LCPA by mass spectrometry. Mass spectra of (A) LCPA and (B) permethylated LCPA. Singly charged, monoisotopic LCPA are labeled with the corresponding m/z value. (C) Proposed chemical structures of the native LCPA molecules; 5 ≤ n + m ≥ 7. Fig. 9. Model illustrating the role of organic microplates and LCPA in morphogenesis of hierarchically porous silica. (A) An organic microplate consists of proteins and polysaccharides (light blue layer) and contains numerous organic nodules throughout the surface (white circles). The black ovals indicate the organic material that is associated with the fultoportulae. (B) Except for the nodules, all parts of the organic microplate have silica deposition activity. Therefore, the addition of silicic acid (in the absence of LCPA) leads to deposition of a thin layer of silica that is perforated by the nodules (i.e, cribrum-like pores). (C) We hypothesize that organic microplates contain docking sites for LCPA molecules around many groups of nodules. When soluble LCPA molecules are added, each docking site assumes the structure of an oval/rectangular shaped wall that encircles a group of nodules. (D) Silica deposition on LCPA loaded organic microplates is enhanced in areas containing bound LCPA and thus results in hierarchically porous silica.

4. Discussion

results). Therefore, the 14 Da mass differences between neighboring peaks were expected to represent methylation isoforms of polyamine molecules of the same chain length with each polyamine chain occurring in up to 9 methylation isoforms. Indeed, after eliminating the Nmethylation isoforms by reductive permethylation only 5 different molecular species were detected in MS analysis (Fig. 8B). The masses are consistent with permethylated putrescin-based molecules containing five (500 Da), six (571 Da) and seven (642 Da) NPI residues, and with permethylated propyldiamine-based molecules containing six (557 Da) and seven (628 Da) NPI residues. The proposed chemical structures of the native (i.e. non-permethylated) LCPA molecules from C. cryptica are shown in Fig. 8C.

The present study reports for the first time the in vitro synthesis of diatom-like hierarchically porous silica patterns using diatom biomolecules. To the best of our knowledge, there has been only one previous published attempt on reconstituting silica formation in vitro using insoluble organic matrices from diatoms. In these experiments, organic microrings from T. pseudonana were incubated for 10–120 min with silicic acid in the presence or absence of a synthetic oligo-propyleneimine dendrimer (Scheffel et al., 2011). Filamentous silica patterns rather than porous silica structures were obtained in these experiments both in the presence and absence of the dendrimer (Scheffel et al., 2011). The re-mineralization of organic microrings or organic 9

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

areola-like pores could be investigated in future research by correlative TEM and super-resolution fluorescence microscopy of organic microplates that have been re-mineralized in the presence of fluorescently labeled LCPA. If involved in areola pore formation, the fluorescently tagged LCPA should accumulate predominantly at the areola walls. Previously, circumstantial experimental evidence for an involvement of LCPA in morphogenesis of the areolae pores has been provided. It was demonstrated that the diatom T. pseudonana produced valve biosilica with aberrant morphology when grown in the presence of 1,3-diaminopropane (DAP) (Frigeri et al., 2006). In DAP treated cells, morphogenesis of valve biosilica was significantly impaired in the z-direction resulting in a widespread absence of areolae pores. DAP is an inhibitor of ornithine decarboxylase, which is suspected to be involved in LCPA biosynthesis (Frigeri et al., 2006; Michael, 2011). However, it was not evaluated whether LCPA levels were reduced in the DAP treated cells. Additionally, it cannot be excluded that the effect of DAP on silica morphology was rather indirect by affecting the general physiology of the cell due to low levels of spermine or spermine-derived metabolites other than LCPA. A key question is to what extent the in vitro process of hierarchical pore pattern formation actually reflects the mechanism of biosilica morphogenesis in vivo. Previous SEM results from analysis of developing C. cryptica valves (Tesson and Hildebrand, 2010) are consistent with a templating mechanism for the biogenesis of areolae pores, which appear to be formed en bloc inside the SDV (note that cribrum pores were not resolved in the SEM images of reference 29). Wide costae are formed first during valve development in vivo (Tesson and Hildebrand, 2010), yet such silica structures were absent in re-mineralized organic microplates. This may at least partially account for the relatively low Si signal intensity in EDX analysis. The reduced mineralization capability of the reconstituted system may be due to the physicochemical differences between the in vivo and the in vitro process. In contrast to in vitro re-mineralization, silica biogenesis occurs in spatial confinement (SDV lumen) and likely under conditions of molecular crowding, which both strongly affect mineralization processes (Rao and Cölfen, 2016). The silica precursor in diatoms is currently unknown, but likely it is different from the freshly prepared silicic acid that was used in our experiments (Kinrade et al., 2002; Gröger et al., 2008). Furthermore, the organic components used for the reconstitution experiments were only those that remained tightly associated with the mature biosilica and thus likely do not represent the complete set of molecules involved in silica biogenesis inside the SDV. Additionally, different types and amounts of biomolecules may enter the SDV at different stages of silica morphogenesis, and thus the mixture of biomolecules that we isolated from the mature biosilica will likely not represent the optimal mixture required for silica morphogenesis. For example, it was previously demonstrated that in a homogeneous mixture of LCPA and silaffins the yield and morphology of in vitro produced silica was strongly dependent on the type and relative amount of silaffin that was added (Poulsen and Kröger, 2004). Considering the many discrepancies between the conditions of in vitro and in vivo silica formation, it seems rather remarkable that the in vitro produced hierarchical pore pattern quite closely resembles the native biosilica morphology of C. cryptica. We ascribe the robustness of the in vitro silica formation process to the relatively stable pattern of areolae walls and nodules. Our discovery that diatom-like, hierarchically porous silica patterns can be generated in vitro through the interaction of an insoluble organic matrix (organic microplate) with LCPA seems to support the model for template-directed silica morphogenesis in the SDV lumen (see Introduction). However, it is currently unknown whether the entire organic microplate is present inside the SDV during silica biogenesis or whether it is entirely or parts of it added to the valve biosilica after completion of morphogenesis. Therefore, elucidating the role of microplates in biosilica morphogenesis will require high spatial resolution imaging of the development of their structures during different stages of biosilica morphogenesis. This should in principle be possible through

microplates of T. pseudonana in the presence of LCPA and AFSC proteins had not been pursued (Scheffel et al., 2011). The present work is focused on elucidating the mechanism for formation of hierarchically porous silica patterns. Therefore, we investigated the in vitro re-mineralization of the valve-derived organic microplates rather than the organic microrings, because hierarchically porous silica patterns are only present in the valve biosilica. In our hands, the preparation of organic microplates from T. pseudonana yielded rather irreproducible results regarding the patterns of organic material in large parts of the microplate area (unpublished data). In contrast, the structures of the organic microplates from C. cryptica were very consistent between different preparation batches and well preserved throughout the microplate area. The organic microplate of C. cryptica is endowed with a nanopatterned architecture for templating both the large areola-like pores (∼100–300 nm wide) and the small cribrum-like pores (∼15 nm diameter). Although the organic microplates were sufficient for the synthesis of hierarchically porous silica patterns, the addition of biosilica-associated organic components from the soluble fraction (AFSC) strongly enhanced the efficiency of this process. In the following, we put forward a hypothesis that explains the role of LCPA in enhancing the formation of hierarchical pore patterns (Fig. 9). We hypothesize that the cribrum-like pores are templated by spherical organic particles that are present throughout the organic microplates. These particles appear as 10–60 nm sized nodules in AFM analysis (see Fig. 3C and D). In electron microscopy analysis the nodules appear as ∼ 30 nm sized, almost transparent spherical pores (see Fig. 2B and D). Similarly sized nodules that correspond to the positioning of cribrum pores in the biosilica have previously been identified in the organic microplates from the diatom S. turris (Tesson and Hildebrand, 2013). Silica deposition is prevented at each nodule, because it may lack silica forming components. As a result, patterns of silica free nodules are generated, which would appear as cribrum-like spherical pores in TEM analysis. The silica deposition activity of the C. cryptica organic microplate outside the nodules may be due to the presence of silica-forming proteins and LCPA-like molecules that are covalently linked within or non-covalently adhered to the microplates (see below). It is reasonable to assume that the areolae-like pores are templated by the oval or rectangular shaped walls that encircle groups of nodules in the organic microplates (see Fig. 2B), because their shapes, sizes, and locations closely match the characteristics of the areolae pores in biosilica. However, the majority of the organic microplates either completely lacks these oval/rectangular walls or they are only partially present. We hypothesize that LCPA molecules bind to specific docking sites on the organic microplate surface, thereby establishing the structures of silica forming oval/rectangular walls. This would explain, why re-mineralization in the presence of added LCPA yields much higher amounts of areola-like pores in the re-mineralized microplates compared to re-mineralization in the absence of added LCPA (Fig. 9). The ability of organic microplates to template, albeit with relatively low yield, areola-like pores even in the absence of added LCPA (see Fig. 6) may be due to remnant LCPA molecules that we assume to be located in the oval/rectangular walls. When the insoluble organic matrices of C. cryptica were treated with sodium dodecyl sulfate (SDS) at 95 °C, a significant amount of biomolecules were extracted (Fig. 4), yet the organic microplates and microrings remained largely intact (D. Pawolski and N. Kröger, unpublished results). Some of the SDS solubilized matrix components (SSMC) had apparent molecular masses that matched ASFC proteins (arrows in Fig. 4A), and some were in the molecular mass range of LCPA (bracket in Fig. 4B). Unfortunately, it was not possible to validate by MS whether the SSMC contained LCPA because we were unable to remove the SDS without complete loss of the low molecular mass material. However, given the similar migration behavior in SDS-PAGE and sensitivity to Coomassie staining, we regard it as likely that the SDS extractable molecules with apparent molecular masses < 10 kDa are LCPA molecules. Whether LCPA molecules are involved specifically in templating the formation of 10

Journal of Structural Biology xxx (xxxx) xxx–xxx

D. Pawolski et al.

super-resolution fluorescence microscopy of synchronized diatom transformant cells that express fluorescently tagged organic microplate proteins (Gröger et al., 2016). Such studies should be able to reveal whether the same protein patterns that are present in the isolated organic microplates are already present inside the developing SDV. In any case, the current study has demonstrated that nanopatterned organic matrices and soluble biomineralization molecules can be synergistically utilized for the synthesis of porous mineral patterns. Future in vitro studies may reveal whether this process can be adopted to produce porous patterns of non-silica minerals that have interesting optical or catalytic properties (e.g., germania, titania) for uses as chemical or optical sensor materials.

studies. J. Struct. Biol. 161, 55–63. Hamm, C.E., et al., 2003. Architecture and material properties of diatom shells provide effective mechanical protection. Nature 421, 841–843. Hildebrand, M., Lerch, S.J.L., Shrestha, R.P., 2018. Understanding diatom cell wall silicification – moving forward. Front. Mar. Sci. 5, 125. Iler, R.K., 1979. The Chemistry of Silica. Wiley, New York. Jentoft, N., Dearborn, D.G., 1983. Protein labeling by reductive methylation. Meth. Enzymol. 91, 570–579. Kinrade, S., Gillson, A.M.E., Knight, C.T.G., 2002. Silicon-29 NMR evidence of a transient hexavalent silicon complex in the diatom Navicula pelliculosa. J. Chem. Soc. Dalton Trans. 3, 307–309. Kotzsch, A., et al., 2017. Silicanin-1 is a conserved diatom membrane protein involved in silica biomineralization. BMC Biol. 15, 65. Kotzsch, A., Pawolski, D., Milentyev, A., Shevchenko, A., Scheffel, A., Poulsen, N., Shevchenko, A., Kröger, N., 2016. Biochemical composition and assembly of biosilica-associated insoluble organic matrices from the diatom Thalassiosira pseudonana. J. Biol. Chem. 291, 4982–4997. Kröger, N., 2007. Prescribing diatom morphology: toward genetic engineering of biological nanomaterials. Curr. Opin. Chem. Biol. 11, 662–669. Kröger, N., Deutzmann, R., Bergsdorf, C., Sumper, M., 2000. Species-specific polyamines from diatoms control silica morphology. Proc. Natl. Acad. Sci. U.S.A. 97, 14133–14138. Kröger, N., Sandhage, K.H., 2010. From diatom biomolecules to bioinspired syntheses of silica- and titania-based materials. MRS Bull. 279, 42993–42999. Lenoci, L., Camp, P.J., 2008. Diatom structures templated by phase-separated fluids. Langmuir 24, 217–223. Michael, A.J., 2011. Molecular machines encoded by bacterially-derived multi-domain gene fusions that potentially synthesize, N-methylate and transfer long chain polyamines in diatoms. FEBS Lett. 585, 2627–2634. Nudelman, F., Chen, H.H., Goldberg, H.A., Weiner, S., Addadi, L., 2007. Lessons from biomineralization: comparing the growth strategies of mollusc shell prismatic and nacreous layers in Atrina rigida. Faraday Discuss. 136, 9–25. Parkinson, J., Brechet, Y., Gordon, R., 1999. Centric diatom morphogenesis: a model based on a DLA algorithm investigating the potential role of microtubules. Biochim. Biophys. Acta 1452, 89–102. Poulsen, N., Kröger, N., 2004. Silica morphogenesis by alternative processing of silaffins in the diatom Thalassiosira pseudonana. J. Biol. Chem. 279, 42993–42999. Poulsen, N., Sumper, M., Kröger, N., 2003. Biosilica formation in diatoms: characterization of native silaffin-2 and its role in silica morphogenesis. Proc. Natl. Acad. Sci. U.S.A. 100, 12075–12080. Rao, A., Cölfen, H., 2016. Mineralization and non-ideality: on nature’s foundry. Biophys. Rev. 8, 309–319. Reimann, B.E.F., Lewin, J.M.C., Guillard, R.R.L., 1963. Cyclotela cryptica, a new brackishwater diatom species. Phycologia 3, 75–84. Reimann, B.E.F., Lewin, J.C., Volcani, B.E., 1966. Studies on the biochemistry and fine structure of the silica shell formation in diatoms. II. The structure of the cell wall of Navicula pelliculosa (Bréb.) Hilse. J. Phycol. 2, 74–84. Robinson, D.H., Sullivan, C.W., 1987. How do diatoms make silicon biominerals? Trends Biochem. Sci. 12, 151–154. Scheffel, A., Poulsen, N., Shian, S., Kröger, N., 2011. Nanopatterned protein microrings from a diatom that direct silica morphogenesis. Proc. Natl. Acad. Sci. U.S.A. 108, 3175–3180. Schmid, A.M.M., 1994. Aspects of morphogenesis and function of diatom cell walls with implications for taxonomy. Protoplasma 181, 43–60. Sumper, M., 2002. A phase separation model for the nanopatterning of diatom biosilica. Science 295, 2430–2433. Sumper, M., Brunner, E., 2008. Silica biomineralization in diatoms: the model organism Thalassiosira pseudonana. ChemBioChem. 9, 1187–1194. Sumper, M., Brunner, E., Lehmann, G., 2005. Biomineralization in diatoms: characterization of novel polyamines associated with silica. FEBS Lett. 579, 3765–3769. Tesson, B., Hildebrand, M., 2010. Dynamics of silica cell wall morphogenesis in the diatom Cyclotella cryptica: substructure formation and the role of microfilaments. J. Struct. Biol. 169, 62–74. Tesson, B., Hildebrand, M., 2013. Characterization and localization of insoluble organic matrices associated with diatom cell walls: insight into their roles during cell wall formation. PLOS One 8, e61675. Weiss, I.M., Schönitzer, V., 2006. The distribution of chitin in larval shells of the bivalve mollusk Mytilus galloprovincialis. J. Struct. Biol. 153, 264–277. Willis, L., Page, K.M., Broomhead, D.S., Cox, E.J., 2010. Discrete free-boundary reactiondiffusion model of diatom pore occlusions. Plant Ecol. Evol. 143, 297–306.

Acknowledgments We would like to thank Thomas Kurth (CMCB Technology Platform, TU Dresden) and Markus Günther (Biology, TU Dresden) for help with electron microscopy, and Marc Gentzel (CMCB Technology Platform, TU Dresden) for help with mass spectrometry analysis. We are grateful to Nicole Poulsen (B CUBE, TU Dresden) for critically reading the manuscript. Funding This work was supported by the Deutsche Forschungsgemeinschaft (DFG) through Research Unit 2038 “NANOMEE” (DFG grants KR 1853/ 6-2 and STE 884/13-2). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.jsb.2018.07.005. References Aitken, Z.H., Luo, S., Reynolds, S.N., Thaulow, C., Greer, J.R., 2016. Microstructure provides insights into evolutionary design and resilience of Coscinodiscus sp. frustule. Proc. Natl. Acad. Sci. U.S.A. 113, 2017–2022. Brunner, E., et al., 2009. Chitin-based organic networks: an integral part of cell wall biosilica in the diatom Thalassiosira pseudonana. Angew. Chem. Int. Ed. 48, 9724–9727. Buss, J.E., Stull, J.T., 1983. Measurement of chemical phosphate in proteins. Meth. Enzymol. 99, 7–14. De Tommasi, E., Gielis, J., Rogato, A., 2017. Diatom frustule morphogenesis and function: a multidisciplinary survey. Mar. Genom. 35, 1–18. Dimas, L.S., Buehler, M.J., 2012. Influence of geometry on mechanical properties of bioinspired silica-based hierarchical materials. Bioinspir. Biomim. 7, 036024. Durkin, C., Mock, T., Armbrust, E.V., 2009. Chitin in diatoms and its association with the cell wall. Eukaryot. Cell 8, 1038–1050. Frigeri, L.G., Radabaugh, T.R., Haynes, P.A., Hildebrand, M., 2006. Identification of proteins from a cell wall fraction of the diatom Thalassiosira pseudonana: insights into silica structure formation. Mol. Cell. Proteom. 5, 182–193. Goldberg, H.A., Warner, K.J., 1997. The staining of acidic proteins on polyacrylamide gels: enhanced sensitivity and stability of ‘‘Stains-All’’ staining in combination with silver nitrate. Anal. Biochem. 251, 227–233. Gordon, R., Drum, R.W., 1994. The chemical basis of diatom morphogenesis. Int. Rev. Cytol. 150, 243–372. Gröger, P., Poulsen, N., Klemm, J., Kröger, N., Schlierf, M., 2016. Establishing superresolution imaging for proteins in diatom biosilica. Sci. Rep. 6, 36824. Gröger, C., Sumper, M., Brunner, E., 2008. Silicon uptake and metabolism of the marine diatom Thalassiosira pseudonana: solid-state 29Si NMR and fluorescence microscopic

11