The silica-based cell walls of diatoms are outstanding examples of nature’s capability to synthesize complex porous structures with genetically controlled patterns from the nanometer scale to the range of hundreds of micrometers. Formation of the cell wall building blocks (valves and girdle bands) occurs in membrane-bound compartments, termed silica deposition vesicles (SDVs), which are unique organelles in silica-forming protists. Isolation of the SDVs has not yet been achieved, which has severely hampered the efforts to understand the mechanisms of biological silica morphogenesis. The present thesis aimed to address this shortcoming. The foundation was the development of an improved cell cycle synchronization and a fluorescence labeling method for the model diatom Thalassiosira pseudonana that enabled rapid identification of valve SDVs in a cell lysate. Correlative fluorescence and electron microscopy allowed visualizing the development of valve silica with unprecedented spatio-temporal resolution. Elemental analysis and demineralization of immature valves provided the first direct chemical evidence that silica morphogenesis is an interplay of inorganic and organic molecules inside the valve SDVs. Cryo TEM imaging of valve SDVs indicated the formation of organic patterns that precede silica depostion. From these observations, an organic biomolecule dependent, liquid-liquid phase separation based model for pore formation in the diatom T. pseudonana was proposed. The second part of this thesis was focused on the enrichment of valve SDVs from T. pseudonana and the subsequent proteomics based identification of more than 40 potential valve SDV proteins. Among these, three diatom-specific proteins contained conserved protein protein interaction domains (ankyrin-repeats) and were surprisingly predicted to be located in the cytoplasm. The fluorescent tagging of the three proteins (termed dANK1-3) confirmed their association with the valve SDVs. When the respective dank genes were knocked out by CRISPR/Cas9, the valves displayed permanent anomalies in the quantity and the pattern of ~22 nm sized pores. Double knockout mutants lacking both dank1 and dank3 were almost completely devoid of pores. The analysis of valve morphogenesis in the single and double knockout mutants revealed phenotypic changes that were consistent with the liquid-liquid phase separation based model for pore pattern formation in diatom biosilica. The work of this thesis has provided for the first time direct access to valve SDVs, which has opened entirely new possibilities for studying the composition, properties, and working mechanism of an organelle that forms a complex shaped mineral.

Kieselalgen – auch Diatomeen genannt – verfügen über die einzigartige Fähigkeit, nanostrukturierte, hierarchisch aufgebaute Zellwände aus Siliziumdioxid – auch als Biosilica bekannt – mit beispielloser Genauigkeit und Reproduzierbarkeit zu bilden. Ein tieferes Verständnis für diesen Prozess, der als “Biomineralisation“ bekannt ist, ist nicht nur auf dem Gebiet der Grundlagenforschung zu Kieselalgen sehr bedeutsam, sondern auch für die Nutzung dieser Nanostrukturierung in den Materialwissenschaften oder der Nanobiotechnologie. Nach dem derzeitigem Stand der Wissenschaft wird diese Strukturierung durch die Selbstorganisation von Proteinmustern, an denen sich das Siliziumdioxid bildet, erreicht. Um die Funktion und das Zusammenspiel einzelner Proteine, die an diesem Biomineralisationsprozess beteiligt sind, entschlüsseln zu können, ist es essentiell ihre strukturelle Organisation aufzuklären und diese mit den morphologischen Zellwandmerkmalen zu korrelieren. Die Größenordnung dieser Merkmale ist im Bereich von Nanometern angesiedelt. Mit Hilfe der Elektronenmikroskopie können diese Biosilicastrukturen aufgelöst werden, jedoch ist keine proteinspezifische Information verfügbar. Ziel dieser Arbeit war es daher, eine Technik zu etablieren, die in der Lage ist, einzelne Biosilica-assozierte Proteine mit Nanometer-Präzision zu lokalisieren. Um dieses Ziel zu erreichen, wurde Einzelmoleküllokalisationsmikroskopie (single-molecule localization microscopy, kurz: SMLM) beispielhaft in der Kieselalge Thalassiosira pseudonana etabliert. Die Position verschiedener Biosilica-assoziierte Proteine innerhalb des Biosilicas und nach dessen chemischer Auflösung wurde mit einer hohen räumlichen Auflösung bestimmt. Um quantitative Ergebnisse zu erhalten, wurde ein Analyse-Workflow entwickelt, der grafische Benutzeroberflächen und Skripte für die Visualisierung, das Clustering und die Kolokalisation von SMLM Daten beinhaltet. Um optimale Markierungen für SMLM an Biosilica-eingebetteten Proteinen zu finden, wurde ein umfassendes Screening von photo-schaltbaren fluoreszierenden Proteinen durchgeführt. Diese wurden als Fusionsproteine mit Silaffin3, einem Protein, welches eng mit der Biosilica-Zellwand assoziiert ist, exprimiert. Es konnte gezeigt werden, dass nur drei von sechs Kandidaten funktional sind, wenn sie in Biosilica eingebettet sind. Silaffin3 konnte indirekt mittels SMLM mit einer Lokalisationsgenauigkeit von 25 nm detektiert werden. Dies erlaubte es, seine strukturelle Organisation aufzulösen und Silaffin3 als eine Hauptkomponente in der Basalkammer der Fultoportulae zu identifizieren. / Diatoms feature the unique ability to form nanopatterned hierarchical silica cell walls with unprecedented accuracy and reproducibility. Gathering a deeper understanding of this process that is known as “biomineralization” is vitally important not only in the field of diatom research. In fact, the nanopatterning can also be exploited in the fields of material sciences or nanobiotechnology. According to the current understanding, the self-assembly of protein patterns along which biosilica is formed is key to this nanopatterning. Thus, in order to unravel the function of individual proteins that are involved in this biomineralization process, their structural organization has to be deciphered and correlated to morphological cell wall features that are in the order of tens of nanometer. Electron microscopy is able to resolve these features but does not provide protein-specific information. Therefore, a technique has to be established that is able to localize individual biosilica-associated proteins with nanometer precision. To achieve this objective, single-molecule localization microscopy (SMLM) for the diatom Thalassiosira pseudonana has been pioneered and exploited to localize different biosilica associated proteins inside silica and after silica removal. To obtain quantitative data, an analysis workflow was developed including graphical user interfaces and scripts for SMLM visualization, clustering, and co-localization. In order to find optimal labels for SMLM to target biosilica-embedded proteins, a comprehensive screening of photo-controllable fluorescent proteins has been carried out. Only three of six candidates were functional when embedded inside biosilica and fused to Silaffin3 – a protein that is tightly associated with the biosilica cell wall. Silaffin3 could be localized using SMLM with a localization precision of 25 nm. This allowed to resolve its structural organization and therefore identified Silaffin3 as a major component in the basal chamber of the fultoportulae. Additionally, co-localization studies on cingulins – a protein family hypothesized to be involved in silica formation – have been performed to decipher their pattern-function relationship. Towards this end, novel imaging strategies, co-localization calculations and pattern quantifications have been established. With the help of these results, the spatial arrangement of cingulins W2 and Y2 could be compared with unprecedented resolution. In summary, this work has laid ground for quantitative SMLM studies of proteins in diatoms in general and contributed insights into the spatial organization of proteins involved in biomineralization in the diatom T. pseudonana.

Kieselalgen – auch Diatomeen genannt – verfügen über die einzigartige Fähigkeit, nanostrukturierte, hierarchisch aufgebaute Zellwände aus Siliziumdioxid – auch als Biosilica bekannt – mit beispielloser Genauigkeit und Reproduzierbarkeit zu bilden. Ein tieferes Verständnis für diesen Prozess, der als “Biomineralisation“ bekannt ist, ist nicht nur auf dem Gebiet der Grundlagenforschung zu Kieselalgen sehr bedeutsam, sondern auch für die Nutzung dieser Nanostrukturierung in den Materialwissenschaften oder der Nanobiotechnologie. Nach dem derzeitigem Stand der Wissenschaft wird diese Strukturierung durch die Selbstorganisation von Proteinmustern, an denen sich das Siliziumdioxid bildet, erreicht. Um die Funktion und das Zusammenspiel einzelner Proteine, die an diesem Biomineralisationsprozess beteiligt sind, entschlüsseln zu können, ist es essentiell ihre strukturelle Organisation aufzuklären und diese mit den morphologischen Zellwandmerkmalen zu korrelieren. Die Größenordnung dieser Merkmale ist im Bereich von Nanometern angesiedelt. Mit Hilfe der Elektronenmikroskopie können diese Biosilicastrukturen aufgelöst werden, jedoch ist keine proteinspezifische Information verfügbar. Ziel dieser Arbeit war es daher, eine Technik zu etablieren, die in der Lage ist, einzelne Biosilica-assozierte Proteine mit Nanometer-Präzision zu lokalisieren. Um dieses Ziel zu erreichen, wurde Einzelmoleküllokalisationsmikroskopie (single-molecule localization microscopy, kurz: SMLM) beispielhaft in der Kieselalge Thalassiosira pseudonana etabliert. Die Position verschiedener Biosilica-assoziierte Proteine innerhalb des Biosilicas und nach dessen chemischer Auflösung wurde mit einer hohen räumlichen Auflösung bestimmt. Um quantitative Ergebnisse zu erhalten, wurde ein Analyse-Workflow entwickelt, der grafische Benutzeroberflächen und Skripte für die Visualisierung, das Clustering und die Kolokalisation von SMLM Daten beinhaltet. Um optimale Markierungen für SMLM an Biosilica-eingebetteten Proteinen zu finden, wurde ein umfassendes Screening von photo-schaltbaren fluoreszierenden Proteinen durchgeführt. Diese wurden als Fusionsproteine mit Silaffin3, einem Protein, welches eng mit der Biosilica-Zellwand assoziiert ist, exprimiert. Es konnte gezeigt werden, dass nur drei von sechs Kandidaten funktional sind, wenn sie in Biosilica eingebettet sind. Silaffin3 konnte indirekt mittels SMLM mit einer Lokalisationsgenauigkeit von 25 nm detektiert werden. Dies erlaubte es, seine strukturelle Organisation aufzulösen und Silaffin3 als eine Hauptkomponente in der Basalkammer der Fultoportulae zu identifizieren.:1 INTRODUCTION 1 1.1 Diatoms – a model system for biomineralization 3 1.2 Imaging of biosilica and associated organic components 8 1.3 Single-molecule localization microscopy (SMLM) 10 2 METHODS & METHOD DEVELOPMENT FOR SMLM DATASETS 17 2.1 Super-resolution reconstruction 19 2.2 Tools for SMLM resolution estimates 21 2.3 Voronoi tessellation for noise-removal and cluster estimation 25 2.4 Tools for SMLM cluster analysis 27 2.5 Coordinate-based co-localization 32 2.6 PairRice – A novel algorithm to extract distances between cluster pairs 33 2.7 SiMoNa – A new GUI for exploring SMLM datasets 35 3 RESOLUTION OF THE SMLM SETUP TESTED WITH DNA ORIGAMI NANOSTRUCTURES 41 3.1 DNA origami as a length standard 42 3.2 Global resolution estimates 44 3.3 Local resolution estimates 47 3.4 Conclusion 53 4 EVALUATION OF PHOTO-CONTROLLABLE FLUORESCENT PROTEINS FOR PALM IN DIATOMS 55 4.1 Selecting PCFPs to minimize interference with the diatom autofluorescence 56 4.2 Screening results for cytosolic and biosilica-embedded PCFPs 58 4.3 The underlying conversion mechanism 61 4.4 Conclusion 63 5 IMAGING THE SIL3 MESHWORK 65 5.1 Analyzing protein layer thickness using tpSil3-Dendra2 65 5.2 Imaging the valve region using tpSil3 68 5.3 Resolution and localization parameters of tpSil3 70 5.4 Conclusion 72 6 DECIPHERING CINGULIN PATTERNS WITH CO LOCALIZATION STUDIES 73 6.1 A two-color cingulin construct for PALM-STORM 73 6.2 Steps towards PALM-STORM: screening, alignment, and imaging routine 76 6.3 Co-localization studies: quantification, clustering, and correlations 83 6.4 Conclusion 91 7 OUTLOOK 93 8 MATERIALS & METHODS 97 8.1 Microscope specifications 97 8.2 DNA origami annealing and AFM measurements 99 8.3 Diatom sample preparations 100 8.4 Fluorescence imaging conditions 102 8.5 Buffer systems 103 9 APPENDICES 105 9.1 Tables and Protocols 105 9.2 Satellite projects 112 9.2.1 Quantitative fluorescence intensity analysis of 3D time-lapse confocal microscopy data in diatoms 112 9.2.2 Applying neural networks to filter SMLM localizations 118 9.2.3 In vivo imaging at super-resolution conditions using SOFI 121 9.2.4 Quantifying chromatic aberrations in the microscope using fiducials 123 10 REFERENCES 127 / Diatoms feature the unique ability to form nanopatterned hierarchical silica cell walls with unprecedented accuracy and reproducibility. Gathering a deeper understanding of this process that is known as “biomineralization” is vitally important not only in the field of diatom research. In fact, the nanopatterning can also be exploited in the fields of material sciences or nanobiotechnology. According to the current understanding, the self-assembly of protein patterns along which biosilica is formed is key to this nanopatterning. Thus, in order to unravel the function of individual proteins that are involved in this biomineralization process, their structural organization has to be deciphered and correlated to morphological cell wall features that are in the order of tens of nanometer. Electron microscopy is able to resolve these features but does not provide protein-specific information. Therefore, a technique has to be established that is able to localize individual biosilica-associated proteins with nanometer precision. To achieve this objective, single-molecule localization microscopy (SMLM) for the diatom Thalassiosira pseudonana has been pioneered and exploited to localize different biosilica associated proteins inside silica and after silica removal. To obtain quantitative data, an analysis workflow was developed including graphical user interfaces and scripts for SMLM visualization, clustering, and co-localization. In order to find optimal labels for SMLM to target biosilica-embedded proteins, a comprehensive screening of photo-controllable fluorescent proteins has been carried out. Only three of six candidates were functional when embedded inside biosilica and fused to Silaffin3 – a protein that is tightly associated with the biosilica cell wall. Silaffin3 could be localized using SMLM with a localization precision of 25 nm. This allowed to resolve its structural organization and therefore identified Silaffin3 as a major component in the basal chamber of the fultoportulae. Additionally, co-localization studies on cingulins – a protein family hypothesized to be involved in silica formation – have been performed to decipher their pattern-function relationship. Towards this end, novel imaging strategies, co-localization calculations and pattern quantifications have been established. With the help of these results, the spatial arrangement of cingulins W2 and Y2 could be compared with unprecedented resolution. In summary, this work has laid ground for quantitative SMLM studies of proteins in diatoms in general and contributed insights into the spatial organization of proteins involved in biomineralization in the diatom T. pseudonana.:1 INTRODUCTION 1 1.1 Diatoms – a model system for biomineralization 3 1.2 Imaging of biosilica and associated organic components 8 1.3 Single-molecule localization microscopy (SMLM) 10 2 METHODS & METHOD DEVELOPMENT FOR SMLM DATASETS 17 2.1 Super-resolution reconstruction 19 2.2 Tools for SMLM resolution estimates 21 2.3 Voronoi tessellation for noise-removal and cluster estimation 25 2.4 Tools for SMLM cluster analysis 27 2.5 Coordinate-based co-localization 32 2.6 PairRice – A novel algorithm to extract distances between cluster pairs 33 2.7 SiMoNa – A new GUI for exploring SMLM datasets 35 3 RESOLUTION OF THE SMLM SETUP TESTED WITH DNA ORIGAMI NANOSTRUCTURES 41 3.1 DNA origami as a length standard 42 3.2 Global resolution estimates 44 3.3 Local resolution estimates 47 3.4 Conclusion 53 4 EVALUATION OF PHOTO-CONTROLLABLE FLUORESCENT PROTEINS FOR PALM IN DIATOMS 55 4.1 Selecting PCFPs to minimize interference with the diatom autofluorescence 56 4.2 Screening results for cytosolic and biosilica-embedded PCFPs 58 4.3 The underlying conversion mechanism 61 4.4 Conclusion 63 5 IMAGING THE SIL3 MESHWORK 65 5.1 Analyzing protein layer thickness using tpSil3-Dendra2 65 5.2 Imaging the valve region using tpSil3 68 5.3 Resolution and localization parameters of tpSil3 70 5.4 Conclusion 72 6 DECIPHERING CINGULIN PATTERNS WITH CO LOCALIZATION STUDIES 73 6.1 A two-color cingulin construct for PALM-STORM 73 6.2 Steps towards PALM-STORM: screening, alignment, and imaging routine 76 6.3 Co-localization studies: quantification, clustering, and correlations 83 6.4 Conclusion 91 7 OUTLOOK 93 8 MATERIALS & METHODS 97 8.1 Microscope specifications 97 8.2 DNA origami annealing and AFM measurements 99 8.3 Diatom sample preparations 100 8.4 Fluorescence imaging conditions 102 8.5 Buffer systems 103 9 APPENDICES 105 9.1 Tables and Protocols 105 9.2 Satellite projects 112 9.2.1 Quantitative fluorescence intensity analysis of 3D time-lapse confocal microscopy data in diatoms 112 9.2.2 Applying neural networks to filter SMLM localizations 118 9.2.3 In vivo imaging at super-resolution conditions using SOFI 121 9.2.4 Quantifying chromatic aberrations in the microscope using fiducials 123 10 REFERENCES 127

Diatoms are single-celled microalgae with silica-based cell walls (frustules) that are abundantly present in aquatic habitats, and form the basis of the food chain in many ecosystems. Many benthic diatoms have the remarkable ability to glide on all natural or man-made underwater surfaces using a carbohydrate- and protein-based adhesive to generate traction. Previously, three glycoproteins, termed FACs (Frustule Associated Components), have been identified from the common fouling diatom Craspedostauros australis and were implicated in surface adhesion through inhibition studies with a glycan-specific antibody. The polypeptide sequences of FACs remained unknown, and it was unresolved whether the FAC glycoproteins are indeed involved in adhesion, or whether this is achieved by different components sharing the same glycan epitope with FACs. Here we have determined the polypeptide sequences of FACs using peptide mapping by LC–MS/MS. Unexpectedly, FACs share the same polypeptide backbone (termed CaFAP1), which has a domain structure of alternating Cys-rich and Pro-Thr/Ser-rich regions reminiscent of the gel-forming mucins. By developing a genetic transformation system for C. australis, we were able to directly investigate the function of CaFAP1-based glycoproteins in vivo. GFP-tagging of CaFAP1 revealed that it constitutes a coat around all parts of the frustule and is not an integral component of the adhesive. CaFAP1-GFP producing transformants exhibited the same properties as wild type cells regarding surface adhesion and motility speed. Our results demonstrate that FAC glycoproteins are not involved in adhesion and motility, but might rather act as a lubricant to prevent fouling of the diatom surface.