A Molecular and Imaging-Based Analysis of Actin-Driven Processes in Cancer Cells Von der Fakultät Energie-, Verfahrens- und Biotechnik der Universität Stuttgart zur Erlangung der Würde eines Doktors der Naturwissenschaften (Dr. rer. nat) genehmigte Abhandlung vorgelegt von Florian Meyer aus Sigmaringen Hauptberichterin: Prof. Dr. Monilola A. Olayioye Mitberichter: Prof. Dr. Hesso Farhan Prüfungsvorsitzender: Prof. Dr. Stefan Legewie Tag der mündlichen Prüfung: 9. Dezember 2025 Institut für Zellbiologie und Immunologie Universität Stuttgart 2025 Declaration of Authorship I hereby declare that this thesis was prepared by myself without illegal help. Where information, data, or illustrations have been derived from other sources, they are properly cited in the text. Eidesstattliche Erklärung Hiermit erkläre ich, dass die vorliegende Arbeit selbstständig von mir und ohne unrechtmäßige Hilfe angefertigt wurde. Verwendete Informationen, Daten und Grafiken, die nicht von mir stammen, wurden entsprechend im Text gekennzeich- net. place, date Florian Meyer III IV Contents Contents Declaration of Authorship / Eidesstattliche Erklärung III List of Figures VII Abbreviations IX Summary XIII Zusammenfassung XVII 1. Introduction 1 1.1. Spatiotemporal Control of Signaling Pathways Ensures Proper Cell Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2. On the Way to Metastasis - Migration and Invasion . . . . . . . . 3 1.2.1. From Static to Motile - Epithelial to Mesenchymal Transition 4 1.2.2. Defining the Cell-ECM Interface - Focal adhesions . . . . . 8 1.2.3. Regulating Focal Adhesions – The Proto-oncogene Src . . 10 1.3. Master of Cell Shape - The Actin Cytoskeleton . . . . . . . . . . . 12 1.3.1. Orchestrating Actin Architecture - Rho GTPases . . . . . 13 1.3.2. Conductors of Rho GTPase Dynamics – RhoGEFs and RhoGAPs . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.4. From the Plasma Membrane to Vesicles - Intracellular Trafficking 20 1.4.1. Recycling or Degradation – the Journey through Endo- cytic Compartments . . . . . . . . . . . . . . . . . . . . . 20 1.4.2. Orchestrating Endosomal Trafficking - RhoB . . . . . . . . 24 1.5. Microscopy – a Versatile Technology in Biology . . . . . . . . . . 27 1.5.1. Foundational Concepts of Confocal Fluorescence Imaging . 27 1.5.2. Making It Visible - Imaging-based Biosensors . . . . . . . 30 1.5.3. From Pixels to Information - Image Quantification . . . . . 34 1.6. Aims of this Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 37 2. Publications and Contributions 39 2.1. Reciprocal regulation of Solo and Src orchestrates Src trafficking to promote mesenchymal cell migration . . . . . . . . . . . . . . . 39 2.2. Golgi screen identifies the RhoGEF Solo as a novel regulator of RhoB and endocytic transport . . . . . . . . . . . . . . . . . . . . 52 2.3. Learning Collective Cell Migratory Dynamics from a Static Snap- shot with Graph Neural Networks . . . . . . . . . . . . . . . . . . 68 2.4. Repeat DNA methylation is modulated by adherens junction signaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 V Contents 3. Discussion 91 3.1. Orchestrating Trafficking: The Delicate Balance of RhoB Regu- lation Through DLC3 and Solo . . . . . . . . . . . . . . . . . . . 91 3.1.1. Measuring and Activating Endosomal RhoB . . . . . . . . 92 3.1.2. Molecular Regulation of DLC3 and Solo at Endosomes . . 94 3.1.3. The Src-Solo Axis: A Feedback Loop at Endosomes? . . . 95 3.1.4. Trafficking Routes Matter . . . . . . . . . . . . . . . . . . 96 3.1.5. Zooming Out: More Than DLC3, Solo, Src, and RhoB . . 97 3.1.6. Future Perspective . . . . . . . . . . . . . . . . . . . . . . 99 3.2. GNN for Collective Cell Migration: One Model to Rule Them All? 100 3.2.1. Single Snapshots to Read Cell Motility Dynamics . . . . . 100 3.2.2. Combining Biosensors with the GNN . . . . . . . . . . . . 102 3.2.3. Limitations and Chances of the GNN . . . . . . . . . . . . 103 3.2.4. Future Perspective . . . . . . . . . . . . . . . . . . . . . . 105 3.3. From Cell-Cell Contacts to the nucleus: The Epigenetic Land- scape Senses Epithelial Cell Density . . . . . . . . . . . . . . . . . 106 3.3.1. E-Cadherin Mediates DNA Methylation of α-satellite repeats 106 3.3.2. The Actin Cytoskeleton: An Underappreciated Role in Epigenetic Regulation? . . . . . . . . . . . . . . . . . . . . 107 3.3.3. Technical Limitations of Fluorescence Complementation . 108 3.3.4. Future Perspective . . . . . . . . . . . . . . . . . . . . . . 109 3.4. Conclusion and Outlook . . . . . . . . . . . . . . . . . . . . . . . 110 References 113 Publications 141 Acknowledgments / Danksagung 143 VI List of Figures 1 The invasion-metastasis cascade . . . . . . . . . . . . . . . . . . . 3 2 EMT is a dynamic process orchestrated by EMT-TFs . . . . . . . 7 3 Focal adhesions are dynamic multiprotein complexes . . . . . . . 9 4 Photoconvertible fluorophores enable quantitative analysis of protein dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 5 Actomyosin contraction and focal adhesion turnover drive cell migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 6 Rho GTPases act as molecular switches . . . . . . . . . . . . . . . 14 7 Distinct activity patterns of Rho GTPases . . . . . . . . . . . . . 16 8 Canonical Domains of DLC RhoGAP family members . . . . . . . 17 9 Endosomal trafficking is a dynamic process regulated by Rab proteins 22 10 RhoB regulates Src trafficking to the plasma membrane . . . . . . 26 11 Working principle and acquired data of a confocal microscope . . 29 12 Basic image quantification workflow . . . . . . . . . . . . . . . . . 35 VII Abbreviations AA Amino Acid AH Anillin Homology ARL Actin Regulatory Layer Arp2/3 Actin Related Protein 2/3 Complex bHLH basic Helix-Loop-Helix BiAD Bimolecular Anchor Detector BiFC Bimolecular Fluorescence Complementation CAF Cancer Associated Fibroblast CIL Contact Inhibition of Locomotion CIP Contact Inhibition of Proliferation CME Clathrin-mediated Endocytosis CNN Convolutional Neural Network CSC Cargo-selective Complex Csk C-terminal Src Kinase DH Dbl Homology DOCK Dedicator of Cytokinesis DLC Deleted in Liver Cancer DNMT DNA Methyltransferase DMGN Deep Multimodal Graph-based Network E-cadherin Epithelial cadherin ECM Extracellular Matrix EEA1 Early Endosome Antigen 1 EGF Epidermal Growth Factor EGFR Epidermal Growth Factor Receptor EMT Epithelial to Mesenchymal Transition ER Estrogen Receptor EphA2 Ephrin type-A receptor 2 ESCRT Endosomal Sorting Complexes Required for Transport F-actin Filamentous Actin FAK Focal Adhesion Kinase IX Abbreviations FMNL2 Formin-like 2 FRET Fluorescence Resonance Energy Transfer FTI Farnesyl-transferase Inhibitor FTL Force Transduction Layer G-actin Globular Actin GAP GTPase-Activating Protein GDI GDP-Dissociation Inhibitor GDP Guanosine Diphosphate GEF Guanine Nucleotide Exchange Factor GNN Graph-based Neural Network GTP Guanosine Triphosphate HER2/neu Human Epidermal Growth Factor Receptor 2 HPLC High-Performance Liquid Chromatography ILV Intraluminal Vesicle ISL Integrin Signaling Layer K8/18 Keratin 8/18 Filaments LD Leucine-Aspartic Acid LINC Linker of Nucleoskeleton and Cytoskeleton LSM Laser Scanning Microscope MBD 5-Methylcytosine Binding Domain mDia mammalian homolog of Drosophila Diaphanous MFI Mean Fluorescence Intensity MSRE Methylation-Sensitive Resctriction Endonucleases MMP Matrix Metalloproteinase MT1-MMP Membrane Type-1 Matrix Metalloproteinase MAPK Mitogen-Activated Protein Kinase MET Mesenchymal to Epithelial Transition NA Numerical Aperture N-cadherin Neural cadherin NGS Next Generation Sequencing NMII Nonmuscle Myosin II X NSCLC Non-small cell lung cancer OOD Out-of-distribution PBR Polybasic Region PDGF Platelet-derived Growth Factor PDGFR PDGF Receptor PDZ post synaptic density protein (PSD95), Drosophila disc large tumor suppressor (Dlg1), and zonula occludens-1 protein (zo-1) PDZL PDZ Ligand PH Pleckstrin Homology PI3K Phosphoinositide 3-Kinase PI Phosphatidylinositol PIP Phosphatidylinositol Phospholipid PI(3)P Phosphatidylinositol 3-phosphate PI(3,5)P2 Phosphatidylinositol 3,5-bisphosphate PR Progesterone Receptor PTM Posttranslational Modification RBD Rho-binding Domain ROCK Rho-associated Coiled-coil Kinase RTK Receptor Tyrosine Kinase SAM sterile α-motif SDCM Spinning Disk Confocal Microscope SFK Src Family Kinase SH2 Src Homology 2 SH3 Src Homology 3 SH4 Src Homology 4 SNR Signal-to-Noise Ratio SNX Sorting Nexin StAR Steroidogenic Acute Regulatory Protein START StAR-Related Lipid Transfer TF Transcription Factor TGFβ Transforming Growth Factor-β XI Abbreviations TNBC Triple Negative Breast Cancer TGN Trans Golgi Network WASH Wiskott-Aldrich Syndrome protein and SCAR homologue YAP Yes-Associated Protein ZEB1 Zinc Finger E-Box Binding Homeobox 1 XII Summary Cellular homeostasis depends on the precise integration of extracellular signal- ing cues, which are transduced inside the cell and controlled by a wide range of regulatory mechanisms. The spatiotemporal coordination of these signaling cues determines cell behavior by triggering cytoskeletal rearrangements during cell migration or establishing epigenetic modifications that drive phenotypic changes. In cancer, cellular transformation allows cells to bypass regulatory constraints and hijack these signaling pathways, thereby promoting cancer progression and metas- tasis. In fact, metastasis accounts for 90% of cancer-related deaths. Understanding how signaling pathways are regulated in time and space is therefore essential to counteract malignant transformation. This thesis examines how the actin cytoskeleton contributes to the regulation of cell behavior in breast cancer cell lines, focusing particularly on cell migration and the epigenetic landscape. Actin dynamics are controlled by Rho GTPases, whose activation and inactivation are spatiotemporally regulated by Guanine Nucleotide Exchange Factors (GEFs) and GTPase-Activating Proteins (GAPs), respectively. To investigate how actin and its regulators shape cell behavior across molecular and multicellular scales, this work combines molecular and cell biology approaches with advanced imaging, optogenetics, biosensors, and deep learning. The first study of this thesis characterized the RhoGEF Solo (also known as ARHGEF40), previously shown to regulate cell migration via RhoA and RhoC. This is the first study to identify and characterize a posttranslational modifica- tion of Solo. The proto-oncogene Src phosphorylated Solo at tyrosine 242 to modulate Solo’s RhoGEF activity and drive mesenchymal cell migration. Addi- tionally, Solo promoted trafficking of photoconverted Src to the plasma membrane. Combined with Solo overexpression and knock-down studies, these findings sug- gest that Solo coordinates Src trafficking and activation to facilitate efficient cell migration. XIII Summary The second study revealed a novel role for Solo as an antagonist of the RhoGAP Deleted in Liver Cancer 3 (DLC3 in regulating the endosomal Rho GTPase RhoB. Using a Rho GTPase activity sensor and the optogenetic Opto-Solo construct, Solo’s GEF domain was recruited to endosomes, visualizing for the first time RhoB activation at these membranes. Functional assays further demonstrated that Solo regulates Epidermal Growth Factor Receptor (EGFR) trafficking and signaling, establishing Solo’s importance for endosomal trafficking. The third, interdisciplinary study examined whether cell geometry and intercellular organization may encode information about collective migratory behavior. In collaboration with the Guo Lab (MIT, Mechanical Engineering), a Graph-based Neural Network (GNN) was developed, which successfully predicted the collective cell migratory behavior from static images. By modulating growth conditions, a wide range of migratory phenotypes was generated. The GNN outperformed predictions based on cell number and shape alone, highlighting how intracellular signaling translates into morphological changes that could be recognized by deep learning- based algorithms, offering potential for diagnostic applications. The final study focused on the interface between extracellular signals and epigenetics. Using the Bimolecular Anchor Detector (BiAD) sensor, DNA methylation at α- satellites was tracked under varying cell densities. Increased cell density reduced DNA methylation of α-satellites, mediated by the adherens junctions protein Epithelial cadherin (E-cadherin) and the actin cytoskeleton. Notably, E-cadherin- deficient breast cancer cell lines failed to respond to increased cell confluence. These findings suggest that the actin cytoskeleton and the upstream Rho GTPases may play underexplored roles in counteracting cellular transformation by maintaining genome integrity. In summary, this thesis underscores the importance of spatiotemporal signaling in shaping cell behavior. By integrating live-cell imaging and molecular perturbations, novel roles of the actin cytoskeleton and its regulators were uncovered, ranging from endosomal trafficking to epigenetic regulation. As demonstrated, advances in microscopy and the co-development of molecular tools offer great potential for biological discovery; however, these advancements also increase the volume and XIV complexity of datasets. Fully realizing the potential of microscopy-based discovery in biology will require interdisciplinary approaches that integrate imaging, molecular biology, and computational analysis to drive the development of new diagnostic and therapeutic strategies. XV Zusammenfassung Die zelluläre Homöostase hängt von der präzisen Integration extrazellulärer Sig- nale ab, die in der Zelle weitergeleitet und durch eine Vielzahl regulatorischer Mechanismen kontrolliert werden. Die raumzeitliche Koordination dieser Signale bestimmt das Zellverhalten, indem sie entweder kurzfristig zu Umstrukturierungen des Zytoskeletts während der Zellmigration führt oder langfristig epigenetische Modifikationen auslöst, die phänotypische Veränderungen bewirken. Bei Kreb- serkrankungen erlaubt die zelluläre Transformation den Zellen, regulatorische Einschränkungen zu umgehen und diese Signale für sich zu nutzen, wodurch Kreb- sprogression und Metastasierung gefördert werden. Tatsächlich sind Metastasen für 90% aller krebsbedingten Todesfälle verantwortlich. Das Verständnis, wie Signalwege zeitlich und räumlich reguliert werden, ist daher entscheidend, um einer malignen Transformation entgegenzuwirken. Diese Arbeit untersucht, wie das Aktinzytoskelett zur Regulation des Zellver- haltens in Brustkrebszelllinien beiträgt, mit besonderem Fokus auf Zellmigra- tion, Invasion und die epigenetische Landschaft. Die Dynamik des Aktinzy- toskeletts wird durch Rho-GTPasen gesteuert, deren Aktivierung und Inak- tivierung wiederum raumzeitlich durch Guaninnukleotid-Austauschfaktoren (GEFs) bzw. GTPase-aktivierende Proteine (GAPs) reguliert wird. Um zu unter- suchen, wie Aktin und seine Regulatoren das Zellverhalten auf molekularer und mehrzellulärer Ebene beeinflussen, kombiniert diese Arbeit Methoden der Molekular- und Zellbiologie mit hochauflösender Bildgebung, Optogenetik, Biosen- soren und Deep Learning. Die erste Studie dieser Arbeit charakterisierte den RhoGEF Solo (auch bekannt als ARHGEF40), welcher zuvor bereits als Regulator der Zellmigration über RhoA und RhoC beschrieben wurde. Diese Studie identifizierte und charakterisierte erstmals eine posttranslationale Modifikation von Solo. Das Proto-Onkogen Src phosphoryliert Solo an Tyrosin 242 und moduliert dadurch dessen GEF-Aktivität, was die mesenchymale Zellmigration antreibt. Zudem förderte Solo den Trans- port von photokonvertiertem Src zur Plasmamembran. In Kombination mit Solo- Überexpressions- und Knockdown-Experimenten legen diese Ergebnisse nahe, dass Solo den Src-Transport und die Aktivierung koordiniert, um eine effiziente Zellmi- gration zu ermöglichen. XVII Zusammenfassung Die zweite Studie identifizierte eine neue Rolle von Solo als Antagonist des RhoGAPs Deleted in Liver Cancer 3 (DLC3) bei der Regulation der endoso- malen Rho-GTPase RhoB. Mithilfe eines Rho-GTPase-Aktivitätssensors und des optogenetischen Opto-Solo-Konstrukts wurde die GEF-Domäne von Solo gezielt an Endosomen rekrutiert und ermöglichte so erstmals die Visualisierung der RhoB- Aktivierung an diesen Membranen. Funktionelle Analysen zeigten darüber hin- aus, dass Solo den Transport des Epidermalen Wachstums Rezeptors (EGFR) und die zugehörige Signaltransduktion reguliert, was seine Bedeutung für den endosomalen Transport unterstreicht. Die dritte, interdisziplinäre Studie untersuchte, ob Zellgeometrie und interzelluläre Organisation Informationen über kollektives Migrationsverhalten codieren können. In Zusammenarbeit mit dem Guo-Labor (MIT, Mechanical Engineering) wurde ein Graph Neural Network (GNN) entwickelt, das in der Lage war, kollektives Mi- grationsverhalten anhand statischer Zellbilder vorherzusagen. Durch Variation der Kulturbedingungen wurde ein breites Spektrum an Migrationsphänotypen erzeugt. Das GNN übertraf Vorhersagen, die lediglich auf Zellanzahl und Zellform basierten, und zeigt, wie sich intrazelluläre Signalprozesse in makroskopische Zellmuster über- setzen lassen, die mittels Deep-Learning-Algorithmen erkannt werden können – ein möglicher Ansatz für diagnostische Anwendungen. Die letzte Studie widmete sich der Schnittstelle zwischen extrazellulären Signalen und Epigenetik. Mit Hilfe des Bimolecular Anchor Detector BiAD-Sensors wurde die DNA-Methylierung an α-Satelliten unter verschiedenen Zelldichten untersucht. Eine erhöhte Zelldichte führte zu einer verringerten Methylierung dieser Regionen, vermittelt durch das Zelladhäsionsprotein E-cadherin und das Aktinzytoskelett. Brustkrebszelllinien, die kein E-cadherin exprimieren, zeigten hingegen keine Reak- tion auf erhöhte Zelldichte. Diese Ergebnisse deuten darauf hin, dass das Aktinzy- toskelett und die upstream-aktiven Rho-GTPasen eine bislang wenig erforschte Rolle bei der Aufrechterhaltung der genomischen Integrität und der Unterdrückung zellulärer Transformation spielen könnten. XVIII Zusammenfassend unterstreicht diese Arbeit die zentrale Bedeutung raumzeitlich regulierter Signalprozesse für die Steuerung des Zellverhaltens. Durch die Kombi- nation von Mikroskopie lebender Zellen und gezielten molekularen Perturbationen konnten neue Funktionen des Aktinzytoskeletts und seiner Regulatoren identifiziert werden – von endosomalem Transport bis hin zur epigenetischen Regulation. Wie gezeigt, bieten Fortschritte in der Mikroskopie und die Entwicklung neuer moleku- larer Werkzeuge großes Potenzial für biologische Entdeckungen; gleichzeitig erhöhen sie jedoch auch die Datenmenge und -komplexität. Um dieses Potenzial vollständig auszuschöpfen, sind interdisziplinäre Ansätze erforderlich, die Mikroskopie, Moleku- larbiologie und computergestützte Analyse integrieren, um die Entwicklung neuer diagnostischer und therapeutischer Strategien voranzutreiben. XIX 1. Introduction 1.1. Spatiotemporal Control of Signaling Pathways Ensures Proper Cell Function The human body consists of approximately 30 trillion cells, comprising more than 50 distinct cell types (Sender et al. 2016). To maintain tissue function and integrity, it is therefore essential that cell behavior is tightly regulated in space and time. Cellular homeostasis is achieved through biochemical cues, such as growth factors and cytokines, and biophysical cues, including signals from cell-cell interactions and tissue properties like stiffness and Extracellular Matrix (ECM) composition. Transmembrane receptors in the plasma membrane sense these extracellular inputs and, in turn, activate intracellular signaling cascades. For example, the Receptor Tyrosine Kinase (RTK) Epidermal Growth Fac- tor Receptor (EGFR) binds extracellular EGF and activates multiple intracel- lular downstream pathways, including the Ras/Mitogen-Activated Protein Ki- nase (MAPK) and Phosphoinositide 3-Kinase (PI3K)/Akt signaling cascades, to promote cell survival and proliferation (Sigismund et al. 2018). Ultimately, the integration of signaling cures in their spatial and temporal context determines cellular behavior. The integration of signaling pathways drives complex cell behavior, such as cell migration. Cell migration plays a pivotal role in physiological contexts, including embryonic development and wound healing, as well as tissue repair in adult organ- isms (Merino-Casallo et al. 2022). In chemotaxis, cells migrate along a gradient of soluble growth factors or cytokines, whereas in haptotaxis, cells migrate along a gra- dient of substrate-bound molecules. Receptors are critical in this process, as ligand binding triggers dynamic rearrangements of both cell-substrate adhesions and the cell cytoskeleton (Seetharaman & Etienne-Manneville 2020). The spatiotemporal control of these rearrangements is thus critical to mediate efficient migration while preventing inappropriate, stimulus-independent motility. 1 1. Introduction Despite tight regulatory mechanisms, cells may acquire aberrant behaviors through cellular transformation, ultimately contributing to diseases such as cancer. In the case of EGFR, overexpression or point mutations are frequently observed in cancer, leading to increased EGFR signaling and driving tumorigenesis (Sigismund et al. 2018; Levantini et al. 2022). As cancer progresses, cancer cells can hijack the migratory machinery to eventually invade distant organs and form metastases (Craene & Berx 2013). Understanding how signaling pathways are regulated in space and time to control cell behavior is therefore crucial for developing strategies to counteract malignant behavior. Microscopy stands out as a crucial and valuable technology for gaining mechanistic insights into the spatiotemporal control of signaling pathways. Advanced imaging approaches can exceed the diffraction limit of 220 nm, allowing detailed analysis of subcellular protein distribution. Notably, the combination of live-cell imaging and genetically encoded biosensors enables the real-time monitoring of signaling dynamics. This combination enables quantitative conclusions about molecular mechanisms and their roles in regulating cell behavior across microscopic and macroscopic scales. This thesis integrates molecular and cell biology techniques with biosensors and image analysis workflows to investigate the spatiotemporal regulation of the Rho GTPase RhoB, the non-RTK Src, and DNA methylation. The introduction will first explore the biological complexity relevant to this work, focusing on cell migration and invasion, the actin cytoskeleton, as well as intracellular trafficking. Throughout, selected imaging-based approaches will be highlighted to illustrate how these have provided molecular insights. Finally, the foundational concepts of microscopy will be explained, relevant biosensors will be introduced, and the process from digital images to biological information will be described. 2 1.2. On the Way to Metastasis - Migration and Invasion 1.2. On the Way to Metastasis - Migration and Invasion Epithelial cells are apical-basal polarized cells that form the epithelial tissue. They line most of the organs and reside on an underlying basement membrane (Ferrer-vaquer et al. 2010). Aberrant signaling can drive hyperproliferation, forming a carcinoma in situ, which is still locally confined by the basement membrane (see Figure 1, Craene & Berx 2013). As tumor progression advances, cancer cells can acquire mesenchymal traits through Epithelial to Mesenchymal Transi- tion (EMT). EMT is accompanied by the expression of ECM-remodelling Matrix Metalloproteinases (MMPs), which enable the cells to breach the basement mem- brane. Additionally, cancer cells hijack the migratory machinery to invade adjacent tissues, intravasate into the bloodstream, extravasate at distant sites, and ultimately establish a metastatic lesion. Figure 1.: The invasion–metastasis cascade. Epithelial cells are characterized by tight cell–cell junctions that maintain tissue integrity. Upon malignant transformation, these cells begin to proliferate, eventually forming a carcinoma in situ that remains confined by the basement membrane. Once this barrier is breached, cancer cells invade the surrounding tissue. During invasion, cells often undergo the Epithelial to Mesenchymal Transition (EMT), acquiring enhanced motility and invasive potential. Upon reaching blood vessels, they can intravasate and travel through the bloodstream to distant organs. There, they must extravasate and may revert to an epithelial phenotype through the Mesenchymal to Epithelial Transition (MET) to initiate the formation of metastatic lesions. Adapted from Craene & Berx (2013). 3 1. Introduction Given that metastasis accounts for 90 % of cancer-related deaths (Gupta &Massagué 2006), the formation of metastases poses a critical hallmark of cancer progression. Although the mere presence of metastases is not necessarily the direct cause of cancer-related mortality, their adverse effects on organ function and the immune system often lead to fatal outcomes (Boire et al. 2024). Moreover, the phenotypic plasticity required for adaptation to organ-specific microenvironments and the formation of a metastatic lesion fosters both immune evasion and therapy resistance (Gerstberger et al. 2023). These implications underscore the urgent need to under- stand the molecular pathways underlying metastasis in order to develop targeted and efficient therapeutic strategies. Importantly, efficient metastasis requires the activation of multiple signaling pathways, a process that is initiated in epithelial cells through EMT. 1.2.1. From Static to Motile - Epithelial to Mesenchymal Transition Epithelial cells organize into a tightly connected multicellular layer that protects underlying tissue. Cell-cell contacts, including desmosomes, tight junctions, and ad- herens junctions achieve this structural organization. Each junction type is formed through homophilic intercellular interactions between different transmembrane proteins. In desmosomes, the cadherins desmoglein and desmocollin interact with one another and are anchored to intracellular intermediate filaments, providing mechanical resistance (Thomason et al. 2010; Hatzfeld 2007). In contrast, tight junctions, composed of occludin and claudin, fulfill the “fence” and “gate” functions to separate basal and apical proteins and regulate paracellular transport, respec- tively (Garcia et al. 2018). Tight junctions are anchored to the actin cytoskeleton, similarly to adherens junctions. At adherens junctions, the extracellular domain of Epithelial cadherin (E-cadherin) mediates homophilic interactions with E-cadherin molecules on neighboring cells, thereby stabilizing cell-cell adhesion and facilitating bidirectional signal transduction across the plasma membrane (Mendonsa et al. 2018). Cytoplasmic adaptor proteins such as p120-catenin, α-catenin, and β-catenin bind to the intracellular domain of E-cadherin, connecting it to the actin cytoskeleton (Yap et al. 2015). Through this linkage, cells are able to sense their extracellular environment and activate programs 4 1.2. On the Way to Metastasis - Migration and Invasion such as Contact Inhibition of Locomotion (CIL) and Contact Inhibition of Prolifera- tion (CIP). CIL reduces migration and CIP inhibits proliferation upon cell-cell con- tact, respectively (Ribatti 2017; Huang et al. 2019), ensuring the formation of highly structured, growth-arrested and immobile epithelial cell layers. As epithelial cells should not migrate in developed organs, the reactivation of invasion and migration is critical to drive metastasis. This activation of invasion and metastasis has been previously acknowledged as one of the 14 hallmarks of cancer (Hanahan & Weinberg 2000; Hanahan 2022). The necessary pheno- typic reprogramming of epithelial cells is achieved through EMT, during which epithelial proteins are downregulated whereas mesenchymal ones are upregulated. For example, cells switch from E-cadherin to Neural cadherin (N-cadherin) via E-cadherin degradation and gene expression changes (Yilmaz & Christofori 2009, Cavallaro et al. 2002). Although this so-called cadherin switch is a central as- pect of EMT, the mesenchymal phenotype is defined by additional functional and molecular properties. An altered protein profile, cytoskeletal rearrangements, and a resulting change in cell behavior characterize the mesenchymal phenotype. In addition to the cad- herin switch, other epithelial junctional proteins such as claudin and occludin are downregulated, further weakening cell-cell contacts (Usman et al. 2021). Although N-cadherin still mediates intercellular interactions, these are weaker and can pro- mote invasion upon growth factor stimulation (Suyama et al. 2002; Kim et al. 2000). Moreover, the intermediate filament protein vimentin is upregulated, which provides mechanical protection during cell motility (Patteson et al. 2019). Rewiring of actin regulatory networks leads to the formation of actin-rich protrusions that drive cell motility (Yilmaz & Christofori 2009). In parallel, the secretion of MMPs and ECM components such as fibronectin enables active remodeling of the surrounding ECM (Lamouille et al. 2014). As a consequence of these properties, mesenchymal cells acquire a spindle-shaped morphology with a front-back polarity, and exhibit high invasive and migratory capabilities (Ferrer-vaquer et al. 2010). Although signaling pathways can initiate EMT, transcriptional reprogramming ultimately reinforces and maintains the mesenchymal state. 5 1. Introduction A set of Transcription Factors (TFs) orchestrates the transcriptional reprogramming during EMT (see Figure 2). These TFs can be induced by growth factor signaling such as EGF, Wnt signaling, and chemokines, e. g. Transforming Growth Factor-β (TGFβ) (Brabletz et al. 2021). Especially the core EMT-TFs SNAIL1 (alias Snail) and SNAIL2 (alias Slug), the basic Helix-Loop-Helix (bHLH) TF Twist, as well as Zinc Finger E-Box Binding Homeobox 1 (ZEB1) and ZEB2 play a central role during EMT, often cooperating at target genes (Peinado et al. 2007). All of them repress the expression of E-cadherin and epithelial cell-cell junction proteins, while promoting expression of N-cadherin and ECM proteins (Lamouille et al. 2014). Because EMT-TFs also exhibit non-redundant functions and tissue-dependent expression levels, the relevance of individual TFs may vary depending on tumor subtype and the specific step of the invasion-metastasis cascade (Stemmler et al. 2019; Markiewicz et al. 2021). Importantly, EMT is not a binary process, but rather a dynamic and reversible continuum, in which intermediate states can hold significant biological relevance. Cancer cells often do not transition into a fully mesenchymal state, lacking all ep- ithelial traits. Even the reversal to an epithelial phenotype, known as Mesenchymal to Epithelial Transition (MET), can confer advantages to the cancer cell. For ex- ample, E-cadherin reexpression can facilitate cell-cell interactions at the metastatic lesion and promote survival signaling (Mendonsa et al. 2018). Accordingly, cells in a partial EMT state simultaneously express epithelial and mesenchymal markers. These hybrid states allow for phenotypic plasticity and collective migration to drive metastasis and therapy resistance (Jolly et al. 2015; Nieto et al. 2016). It is therefore fundamental to distinguish between different EMT states to elucidate their specific roles during metastasis. Although methods such as immunofluorescence and flow cytometry allow for the identification of EMT markers at single-cell resolution, they are limited in capturing dynamic changes over time. In contrast, EMT reporter systems allow real-time monitoring. EMT reporter systems link fluorescent readouts to the expression of epithelial or mesenchymal state markers, offering dynamic insights into EMT. For example, Tsubakihara et al. established a system in which RFP expression was driven by the CDH1 (E-cadherin) promotor where a high signal indicated an epithelial state 6 1.2. On the Way to Metastasis - Migration and Invasion Figure 2.: EMT is a dynamic process orchestrated by EMT-TFs. The core set of Epithelial to Mesenchymal Transition (EMT)-Transcription Factors (TFs), including ZEB1, ZEB2, SNAIL, SLUG, and TWIST, represses epithelial while activating mesenchymal gene expression. Consequently, cells lose their epithelial and gain mesenchymal traits. While epithelial cells have cell-cell contacts such as adherens junctions, tight junctions, and desmosomes, these structures are lost through EMT. Additionally, cells switch from apical-basal to front-rear polarity. Together with basement membrane degradation and the formation of actin stress fibers, cells start to invade the adjacent tissue. Importantly, EMT is dynamic and can be reverted, termed Mesenchymal to Epithelial Transition (MET). Therefore, partial states can exist along the epithelial-mesenchymal spectrum. Adapted from Brabletz et al. (2021). and low fluorescence a mesenchymal state. Using this EMT reporter cell line, TGFβ was shown to control either pro-invasive or pro-stemness pathways in a mouse model (Tsubakihara et al. 2022). Depending on whether cells remained in a mesenchymal state or reverted via MET, they acquired distinct phenotypic properties (Tsubakihara et al. 2022). Similarly, Lüönd et al. employed EMT lineage tracing in breast cancer mouse models to resolve partial EMT states. Their study revealed that cells in partial EMT states were enriched in lung metastases, whereas cells undergoing full EMT were more abundant following treatment with paclitaxel, a tubulin-targeting chemotherapeutic (Lüönd et al. 2021). Although EMT-TFs dictate the cellular phenotype, increased motility and invasion also stem from cytoskeletal remodeling and signaling network rewiring, with cell adhesion sites playing a central role. 7 1. Introduction 1.2.2. Defining the Cell-ECM Interface - Focal adhesions Focal adhesions are dynamic multiprotein complexes that facilitate the attachment of cells to the surrounding ECM (see Figure 3A, B). They are organized into three distinct layers, which also partially reflect their stepwise assembly: the basal Integrin Signaling Layer (ISL), the intermediate Force Transduction Layer (FTL), and the apical Actin Regulatory Layer (ARL) (Kanchanawong et al. 2010). Within the ISL, the extracellular domains of the 24 distinct αβ-integrin heterodimers found in mammals bind to specific ECM ligands. The integrins then undergo conformational activation, leading to the formation of integrin nanoclusters (Pang et al. 2023; Kadry & Calderwood 2020). Afterwards, adaptor proteins such as talin, vinculin, and paxillin are recruited, resulting in the formation of so-called nascent adhesions (Sun et al. 2014). Paxillin is a key adaptor protein, critically involved in focal adhesion signaling, regulation, and cytoskeleton remodeling (Green & Brown 2019). Its binding to Focal Adhesion Kinase (FAK) facilitates FAK autophosphorylation at tyrosine 397 (Y397), and its catalytic activity can be further enhanced through phosphorylation mediated by the cytoplasmic kinase Src (Legerstee & Houtsmuller 2021). Protein recruitment and the activation of signaling cascades promote the maturation of nascent adhesions. Nascent adhesions mature into focal complexes and ultimately into focal adhesions through the recruitment of additional proteins, with subsequent stabilization driven by forces exerted by the ECM and actomyosin contractility (Kanchanawong & Calderwood 2023). While vinculin primarily localizes to the FTL, talin spans all three layers and both play a significant role in mechanotransduction (Kanchanawong et al. 2010). Talin directly links β-integrin with the actin cytoskeleton early during adhesion formation (Calderwood et al. 1999; Critchley 2009). Together with talin’s spring-like structure, this enables talin to function as a mechanosensor. The degree of force exerted on talin determines its conformational unfolding, exposing up to 13 distinct protein-binding sites and thereby translating mechanical input into diverse signaling outputs (Goult et al. 2021). Vinculin can then bind to newly exposed binding sites, stabilizing the actin cytoskeleton and focal adhesion (Yao et al. 2014). This maturation and following turnover of focal adhesions is a dynamic and tightly regulated process (see Figure 3B). 8 1.2. On the Way to Metastasis - Migration and Invasion Figure 3.: Focal adhesions are dynamic multiprotein complexes. (A) Focal adhesions are organized into three distinct layers. The basal Integrin Signaling Layer (ISL) includes αβ-integrin heterodimers that interact with ECM components and associate intracellularly with proteins such as FAK and paxillin. The apical Actin Regulatory Layer (ARL) comprises F-actin and associated regulatory proteins, e. g. VASP and α-actinin. These two layers are interconnected by the Force Transduction Layer (FTL), which includes the mechanotransducer proteins talin and vinculin. Adapted from Kanchanawong & Calderwood (2023). (B) Still images from timelapse videos of ZMEL Paxillin-EGFP, revealing the lifetimes of paxillin at focal adhesions. The left displays the whole cell, while the magnified panels correspond to the region marked by the grey box. Red dotted circles indicate the same paxillin-positive puncta over time, from initial assembly to disassembly. Adapted from Xue et al. (2022). Considering the highly dynamic nature and size of focal adhesions, live-cell mi- croscopy remains the most effective method for quantitatively analyzing their turnover. Legerstee et al. employed photoconvertible fluorophores attached to pax- illin to assess the dynamics and spatial distribution of paxillin in various types of focal adhesions (Legerstee et al. 2019). Photoconvertible fluorophores change their excitation and emission spectra in response to light-induced structural changes (see Figure 4A). By inducing photoconversion in a spatially confined area, specifically at a single focal adhesion, the spatiotemporal distribution of this defined protein subpopulation can be monitored. Although technically challenging, this approach yielded novel insights: photoconversion of paxillin revealed a subpopulation of 9 1. Introduction Figure 4.: Photoconvertible fluorophores enable quantitative analysis of protein dynamics. (A) During live cell imaging, photoconvertible fluorophores can be photoconverted in a spatially confined region, in this case at a single focal adhesion. As unbound proteins will diffuse away, this technique allows the dynamics of stably bound protein subpopulations to be assessed. (B-D) Photoconvertible paxillin at focal adhesions. Overview (B) and magnified view of the boxed region showing (C) total and (D) photoconverted paxillin before, immediately after photoconversion, and several minutes after photoconversion. Adapter from Legerstee et al. (2019). stably bound paxillin that was not apparent in the unconverted population (see Figure 4B-D). Using this technique, the dynamics and distribution of multiple focal adhesion proteins within a focal adhesion were mapped, revealing the existence of nanoclusters and showing that the subcellular position of focal adhesions affects their dynamics (Legerstee et al. 2019). Combining live-cell microscopy with ad- vanced tools such as photoconvertible proteins thus provides a powerful approach to investigate protein subpopulations and their dynamics. This level of resolu- tion is crucial for understanding how specific proteins contribute to the turnover of focal adhesions. 1.2.3. Regulating Focal Adhesions – The Proto-oncogene Src The intracellular non-RTK Src is a member of the Src Family Kinases (SFKs). In general, SFKs comprise a kinase domain, a phosphotyrosine-binding Src Homol- ogy 2 (SH2), a polyproline-binding SH3 domain, as well as a SH4 domain (Shah et al. 2018). The SH4 domain contains unique residues that mediate Src-specific localization, e. g. to endosomes, while the SH2 domain facilitates recruitment to adhesion sites (Sandilands et al. 2007; Chu et al. 2014; Kasahara et al. 2007a). Src binds via the SH2 domain to phosphorylated FAK and in turn further phos- 10 1.2. On the Way to Metastasis - Migration and Invasion phorylates and activates FAK (Sadowski et al. 1986). As the FAK/Src complex phosphorylates a wide range of focal adhesion proteins, it acts as a key regulator of focal adhesion architecture. Within minutes, focal adhesions are rapidly assembled and disassembled (see Figure 3B). Src activity is a critical determinant of focal adhesion stability and thus of cell migration. In cooperation with FAK, Src promotes the formation of elongated adhesions, cell spreading, and sustained protrusive activity (Karginov et al. 2014). Conversely, inhibition of Src activity stabilizes focal adhesions and impairs cell migration (Kasahara et al. 2007a; Deramaudt et al. 2011). Although Src is important for both the assembly and disassembly dynamics of focal adhesions, its precise role can also be context-dependent. Chen et al. showed that Src regulates focal adhesion assembly and disassembly specifically when grown on a micropatterned substrate presenting the ephrinA1 ligand. Mechanistically, ephrinA1-bound Ephrin type-A receptor 2 (EphA2) activates intracellular Src, which is subsequently recruited to focal adhesions to promote cell migration (Chen et al. 2018). The cellular surroundings and perturbations of Src activity levels can therefore influence the migratory capabilities of cells. Src is frequently overexpressed in cancers, including Triple Negative Breast Can- cer (TNBC). TNBC lacks expression of Estrogen Receptor (ER), Progesterone Receptor (PR), and Human Epidermal Growth Factor Receptor 2 (HER2/neu) (Brenton et al. 2005), rendering targeted therapies against hormone receptors and HER2/neu ineffective (Derakhshan & Reis-Filho 2022). In breast cancer patients, Src overexpression correlates with significantly decreased median survival (Elsberger et al. 2010). In this context, Src contributes to enhanced proliferation and migra- tory phenotypes. For instance, Src can cooperate with RTKs such as EGFR to drive proliferation and invasion (Kim et al. 2013). Furthermore, Src overexpression alone is sufficient to increase cell migration in the TNBC cell line MDA-MB-231 (Mayoral-Varo et al. 2021). 11 1. Introduction 1.3. Master of Cell Shape - The Actin Cytoskeleton Migration is regulated through a dynamic interplay between focal adhesions and the actin cytoskeleton. This process can be simplified into a cycle comprising four distinct steps (see Figure 5A-B; Weißenbruch & Mayor 2024). First, cells form actin-rich protrusions at the cell front, known as the leading edge. Here, actin polymerization, through the conversion of Globular Actin (G-actin) into Filamentous Actin (F-actin), is driven by the actin nucleator Actin Related Pro- tein 2/3 Complex (Arp2/3), which promotes formation of branched actin and lamellipodia (Bugyi & Carlier 2010). In the second step, cells establish new ad- hesions to engage via integrins with the substratum, anchoring the leading edge and enabling forward movement of the cell body. During cell body contraction, Figure 5.: Actomyosin contraction and focal adhesion turnover drive cell migration. (A) Cell migration proceeds through a cyclical process. First, actin-rich protrusions form at the leading edge and new adhesion sites to the ECM are established. Following cell body contraction, the trailing edge is retracted, allowing the cell to move forward. Adapted from Weißenbruch & Mayor 2024. (B) MCF10A cells expressing the actin probe LifeAct-GFP. Upon stimulation with EGF, the cell undergoes the protrusion, maximum extension, and retraction phase. Adapted from Gagliardi et al. (2015). 12 1.3. Master of Cell Shape - The Actin Cytoskeleton the actin cytoskeleton is crucial to generate actomyosin contractility as part of stress fibers. They consist of anti-parallel actin filaments and the motor protein Nonmuscle Myosin II (NMII), mediating the contractile force (Narumiya et al. 2009). The nucleation and polymerization of these actin filaments are driven by the formin mammalian homolog of Drosophila Diaphanous (mDia) (Goode & Eck 2007). Actomyosin contractility is critical not only for adhesion maturation and the stabilization of lamellipodia (Giannone et al. 2007; Kalappurakkal et al. 2019), but also for generating the tension necessary to contract the cell body (Ridley 2015). In the final step of the cycle, adhesions at the trailing edge must be disassembled to allow retraction of the rear and enable continued forward migration. This cycle proceeds and repeats, driving persistent cell movement. Intriguingly, during cell migration, the actin cytoskeleton has emerged as an impor- tant regulator of gene expression. The nucleus often needs to be deformed to enable cellular movement through confined environments, a process facilitated by the actin cytoskeleton (Friedl et al. 2011). The actin cytoskeleton is mechanically linked to the nuclear envelope through the Linker of Nucleoskeleton and Cytoskeleton (LINC) complex (Lombardi et al. 2011). By altering nuclear shape, actin can modulate the nuclear access of transcription factors, chromatin structure, and ultimately regulate gene expression (Sankaran et al. 2019; Le et al. 2016). Moreover, the state of the actin cytoskeleton can directly influence the expression of actin regulatory proteins and, consequently, affect cell motility. This regulation is mediated by the relative availability of G-actin and F-actin, which is sensed by specific G-actin/F-actin- binding proteins. These proteins can translocate to the nucleus to regulate the transcription of target genes (Olson & Nordheim 2010). Altogether, the actin cytoskeleton serves multiple roles during cell migration that need to be tightly controlled in a spatiotemporal manner. This spatiotemporal regulation is achieved through the action of different upstream regulators. 1.3.1. Orchestrating Actin Architecture - Rho GTPases The Rho family of GTPases belongs to the Ras superfamily of small GTPases and functions as a key regulator of the actin cytoskeleton. As such, Rho GTPases are involved in mechanotransduction, cell migration, and vesicular trafficking (Bui et al. 13 1. Introduction Figure 6.: Rho GTPases act as molecular switches. GTP-bound, active Rho GTPases are recruited to membranes, where they activate their downstream effectors involved in vesicular traf- ficking, gene transcription, and cytoskeletal remodeling. GAPs stimulate the intrinsic hydrolysis activity of Rho GTPases, leading to hydrolysis of GTP to GDP and thereby inactivating the Rho GTPase. In the inactive state, Rho GDIs can bind and sequester the Rho GTPase in the cytosol, permitting further reactivation. For reactivation, GEFs promote the release of GDP leading to the binding of GTP, thus activating the Rho GTPase again. Adapted from Olayioye et al. (2019). 2019; Ridley 2015; Olayioye et al. 2019). The 20 members, which are subdivided into eight subfamilies, can be further classified based on their molecular regulation. Atypical Rho GTPases lack intrinsic GTPase activity and are therefore locked in an active state that is regulated by e. g. Posttranslational Modifications (PTMs) (reviewed in Huang et al. 2024). In contrast, classical Rho GTPases act as molecular switches, characterized by their inactive Guanosine Diphosphate (GDP)- bound and active Guanosine Triphosphate (GTP)-bound states. GTPase-Activating Pro- teins (GAPs) stimulate the intrinsic GTPase activity, leading to hydrolysis of GTP to GDP, thus inactivating the Rho GTPase (see Figure 6). GDP-Dissociation Inhibitors (GDIs) can then bind and sequester inactive GDP-bound Rho GTPases, thereby preventing their reactivation. Conversely, Guanine Nucleotide Exchange Factors (GEFs) decrease the GTPase affinity to GDP promoting the exchange of GDP to GTP and thereby inducing Rho downstream signaling. The different Rho GTPases then regulate distinct actin patterns, enabling the spatiotemporal coordi- nation of cytoskeletal remodeling during cellular processes. 14 1.3. Master of Cell Shape - The Actin Cytoskeleton Among the Rho family of GTPases, the Rac, Cdc42, and Rho subfamilies are the best characterized and are known to regulate distinct actin structures. As early as 30 years ago, these three subfamilies were linked to adhesion sites and specific actin- rich structures. At the leading edge, Cdc42 was shown to induce filopodia, which consist of parallel actin bundles, whereas Rac promoted the formation of branched actin in lamellipodia (Nobes & Hall 1995). In contrast, Rho was associated with stress fiber formation, contributing to actomyosin contractility at the trailing edge (Nobes & Hall 1995). However, live-cell imaging studies using Rho GTPase activity biosensors have since led to a paradigm shift, revealing a more dynamic and coordinated regulation of Rho GTPases across the cell. Fluorescence Resonance Energy Transfer (FRET)- based RhoA, Cdc42, and Rac1 biosensors (see section 1.5.2) have demonstrated that all three Rho GTPases are active near the plasma membrane at sites of cell protrusions. Notably, active Rac1 does not colocalize with mature focal adhesions, and RhoA activity zones can span broader regions compared to others (see Figure 7A-B; Müller et al. 2020). These overlapping Rho GTPase activity zones enable functional crosstalk, which is especially important for coordinating cell migration. Using a FRET-based RhoA biosensor, active RhoA was shown to localize at the protruding edge of migrating cells, where it negatively regulates Rac (El-Sibai et al. 2008). Mechanistically, this is accomplished by RhoA/Rho-associated Coiled- coil Kinase (ROCK)-dependent activation of the RacGAP ARHGAP22, which inhibits Rac activity (Sanz-Moreno et al. 2008). This highlights a dual role for RoaA during cell migration. On the one hand, RhoA drives mDia1-dependent actin polymerization at the leading edge, which supports Rac and Arp2/3-driven lamellipodia and focal complex formation (Machacek et al. 2009; Lee et al. 2015; Zaidel-Bar et al. 2007; Bugyi & Carlier 2010). Afterwards, FAK/Src/RhoA/ROCK signaling facilitates actomyosin contractility, promoting the maturation of focal complexes into focal adhesions (Kalappurakkal et al. 2019). On the other hand, RhoA-driven actomyosin contractility facilitates the contraction of the trailing edge and forward movement of the cell (Ridley 2015). Although RhoA is therefore critical in several steps of protrusion formation, excessive RhoA activity and hence 15 1. Introduction Figure 7.: Distinct activity patterns of Rho GTPases. The activity of Rac1, RhoA, and Cdc42 in mature focal adhesions of REF52 fibroblasts spreading on fibronectin was visualized using FRET sensors in combination with mCherry-paxillin. While active Rac1 does not colocalize with mature focal adhesions, both RhoA and Cdc42 do. In contrast to active Cdc42, RhoA activity spans a broader band near the plasma membrane compared to the more localized activity of Cdc42. Importantly, all three Rho GTPases exhibit activity along the plasma membrane. (A) Whole-cell images. (B) Magnified views of the boxed regions. Adapted from Müller et al. (2020). contractility can inhibit lamellipodia dynamics and cell migration (Petrie & Yamada 2012). A finely tuned balance between the different Rho GTPases is therefore fundamental to ensure proper control of cell migration. This balance is critical, as the Rho family consists of several Rho GTPases. The classical Rho family comprises RhoA, RhoB, and RhoC, which share approx- imately 85 % sequence identity, with key differences located in the C-terminal region (Wheeler & Ridley 2004). This region contains the CAAX-motif that can be prenylated to facilitate membrane recruitment and downstream signaling (Al- lal et al. 2000). While all three Rho GTPases can undergo geranylgeranylation, RhoB is unique in its ability to alternatively undergo farnesylation (Adamson et al. 1992). Consequently, RhoB localizes not only to the plasma membrane but also to endosomal compartments, where it is involved in the regulation of membrane traf- 16 1.3. Master of Cell Shape - The Actin Cytoskeleton Figure 8.: Canonical Domains of DLC RhoGAP family members. Members of the Deleted in Liver Cancer (DLC) RhoGAP family comprise three functional domains: an N-terminal sterile α-motif (SAM) domain, a central catalytic RhoGAP domain, and a C-terminal StAR- Related Lipid Transfer (START) domain. Adapted from Frey et al. (2025). ficking (see section 1.4.2). In contrast, RhoC shares functional overlap with RhoA in regulating actomyosin contractility and lamellipodia but also exhibits unique functions, particularly during amoeboid migration in 3D environments (Lou et al. 2021). In this context, RhoC, but not RhoA, has been shown to regulate the formin Formin-like 2 (FMNL2) and promote actomyosin and bleb-driven type of migration (Lämmermann & Sixt 2009; Kitzing et al. 2010). The spatiotemporal control of Rho GTPases is achieved through RhoGAPs and RhoGEFs. 1.3.2. Conductors of Rho GTPase Dynamics – RhoGEFs and RhoGAPs The Deleted in Liver Cancer (DLC) RhoGAP Family The Deleted in Liver Cancer (DLC) RhoGAP family is a great example of how the loss of Rho GTPase regulators affects cell behavior. As their name suggests, members of this RhoGAP family are frequently deleted in liver cancer but also inactivated in various other cancer types (Frey et al. 2025). The three family members DLC1, DLC2, and DLC3 share a conserved structural organization consisting of a centrally located RhoGAP, an N-terminal sterile α-motif (SAM), and a C-terminal StAR-Related Lipid Transfer (START) domain. The conserved RhoGAP domains exhibit GAP activity towards the Rho subfamily and, to a lesser extent, the Cdc42 subfamily in vitro (Braun & Olayioye 2015; Frey et al. 2025). The N-terminal SAM domain mediates protein-protein interactions and has been shown to function as an auto-inhibitory domain, modulating the RhoGAP activity of DLC1 (Kim & Bowie 2003; Kim et al. 2008; Joshi et al. 2020). The C-terminal START domain, typically involved in lipid binding and transfer between membranes (Wirtz 2006; Alpy & Tomasetto 2005), has been implicated in the regulation of DLC1, however, the molecular mechanism and possible involvement in lipid transfer is yet to be unraveled (Frey et al. 2025). 17 1. Introduction In addition to their conserved domains, DLC proteins contain motifs that facilitate the recruitment to specific subcellular sites, including focal adhesions and adherens junctions. The PDZ Ligand (PDZL) motif of DLC3 is bound by the PDZ domain of the polarity protein Scribble, whereby DLC3 is recruited to adherens junctions where DLC3 inhibits Rho-ROCK signaling and supports junctional integrity (Hendrick et al. 2016). Moreover, DLC1 is recruited through the Leucine-Aspartic Acid (LD) motif to talin and FAK at focal adhesions to promote RhoGAP activity (Li et al. 2011), and a similar motif is found in DLC3 (Durkin et al. 2007). At focal adhesions, upon loss of DLC1, the breast cancer cell line MCF7 showed increased focal adhesion assembly and stress fiber formation, ultimately promoting migration in a Dia1-dependent manner (Holeiter et al. 2008). DLC proteins therefore act as important tumor suppressors by regulating Rho GTPase signaling and maintaining epithelial properties. The RhoGEF Solo As inactive Rho GTPases have a high affinity to GDP and GDP is more abundant in cells than GTP, they rely on RhoGEFs for their activation. In fact, changes in Rho GTPase activity due to hyperactive upstream signaling are frequently observed in cancer (Cervantes-Villagrana et al. 2023). It is therefore important to identify how RhoGEFs contribute to oncogenic signaling. RhoGEFs can be divided into two distinct groups: the Dedicator of Cytokinesis (DOCK) and the Dbl family, each named after its founding member. While DOCK family members show GEF activity towards Rac and/or Cdc42, Dbl family members can additionally target the Rho family (Meller et al. 2005; Goicoechea et al. 2014). All Dbl RhoGEFs harbor the conserved catalytic Dbl Homology (DH) domain and an adjacent Pleckstrin Homology (PH) domain, in addition to other protein domains enabling their specific function (Rossman et al. 2005). The RhoGEF investigated in this thesis was Solo. The Dbl family member Solo (alias ARHGEF40 or Scambio) contains a DH/PH domain (Curtis et al. 2004). PH domains have been shown to increase GEF activity and/or facilitate membrane recruitment (Aghazadeh et al. 2000; Snyder et al. 2001; Ferguson et al. 1995), however, the functional relevance of this domain 18 1.3. Master of Cell Shape - The Actin Cytoskeleton for Solo function remains unknown. In contrast, Solo’s DH domain was shown to elicit RhoGEF activity towards RhoA and RhoC (Curtis et al. 2004; Abiko et al. 2015). Although Solo can interact with constitutively active Rac and Cdc42, Solo does not exhibit RhoGEF activity towards them (Curtis et al. 2004). In line with its specificity for RhoA and RhoC, Solo has been implicated in several cellular processes involving actomyosin contractility. For example, the reorientation of endothelial stress fibers in response to cyclic stretch depends on Solo-RhoA signaling (Abiko et al. 2015). Solo also localizes to hemidesmosomes, where it regulates Keratin 8/18 Filaments (K8/18) organization through RhoA/ROCK signaling, mediates traction force generation, and contributes to tensional-force induced stress fiber formation (Fujiwara et al. 2016; Fujiwara et al. 2018). Having these important regulatory functions, Solo was shown to affect cell migration. In Non-small cell lung cancer (NSCLC) cells, Solo promoted directed cell migration through RhoA (Gu et al. 2022). In contrast, Solo decelerated collective but not single cell migration of epithelial MDCK cells (Isozaki et al. 2020). Together, these findings place Solo as a central regulator of cell-substrate adhesion and cell migration. 19 1. Introduction 1.4. From the Plasma Membrane to Vesicles - Intracellular Trafficking Receptors and signaling molecules must first reach the plasma membrane before they initiate downstream signaling cascades. To regulate signaling output, receptors are eventually internalized through endocytosis. Likewise, the endocytosis of cell-cell and cell-matrix molecules regulates respective adhesion sites (Mellman & Yarden 2013). Distinct molecular mechanisms can initiate the endocytosis of receptors from the plasma membrane. Among these, Clathrin-mediated Endocytosis (CME) and caveolar-type endocytosis rely on clathrin or caveolin in combination with other proteins to drive membrane invagination and fission (Doherty & McMahon 2009). Both mechanisms were first observed in thin-section electron microscopy (Roth & Porter 1964; Yamada 1955), highlighting the historical and ongoing importance of microscopy to study biological processes in their spatial context. Following endocy- tosis, receptors enter the endocytic pathway within endosomes. 1.4.1. Recycling or Degradation – the Journey through Endocytic Compartments Endosomes are intracellular, membrane-bound compartments involved in trafficking cargo to and from the Trans Golgi Network (TGN), targeting endosomal cargo for degradation, or recycling it back to the plasma membrane. To achieve this, cargo is trafficked through distinct endocytic compartments, including early and late endosomes, recycling endosomes, and lysosomes, which can undergo fusion and hence exchange their cargo proteins (Salzman & Maxfield 1988). Early endosomes represent the first compartment of the endocytic pathway. Early endosomes serve as the initial sorting station for endocytosed receptors and receptor-bound ligands. With an intraluminal pH of around 6.5 (Fuchs et al. 1989), early endosomes facilitate the dissociation of receptor-bound ligands. Following dissociation, receptors may be rapidly recycled to the plasma membrane via re- cycling endosomes, while dissociated ligands accumulate through inward-budding in Intraluminal Vesicles (ILVs) (Hu et al. 2015). For example, the transferrin receptor releases its ligand in early endosomes, allowing for the rapid recycling of the transferrin receptor to the plasma membrane (Elkin et al. 2016). As early 20 1.4. From the Plasma Membrane to Vesicles - Intracellular Trafficking endosomes mature into late endosomes and eventually lysosomes, their intralu- minal pH progressively decreases, ultimately falling below a pH of 5.0 (Saftig & Klumperman 2009). This acidification serves two main functions. First, the low pH facilitates the proteolytic reactions of nucleases, proteases, and lipases, thereby conferring the degradative capability of lysosomes (Jeger 2020). Second, different receptor-ligand complexes show varying pH-dependent dissociation constants, which in turn determine the receptor’s intracellular trafficking route (Mellman 1996). For instance, in contrast to the transferrin receptor, the mannose-6-phosphate receptor requires the more acidic environment of late endosomes for ligand release, after which the receptor is trafficked to the TGN (Elkin et al. 2016). The differential trafficking routes are partially due to the varying molecular identity and associated protein machinery of each endosomal compartment. The unique identity of endosomal compartments is mediated through Rab proteins and the local Phosphatidylinositol Phospholipid (PIP) composition (see Figure 9). Similar to Rho GTPases, Rab GTPases cycle between GEF/GAP-mediated ac- tive GTP- and inactive GDP-bound states to drive downstream signaling. Their C-terminal geranylgeranyl groups enable membrane association, and their local- ization to specific endosomal compartments is determined by interaction with distinct effectors and the local PIP composition (Stenmark 2009). The ER-derived Phosphatidylinositol (PI) is distributed throughout the cell by membrane traffick- ing. Once at its subcellular compartment, PI is modified by local PI kinases and phosphatases, giving rise to specific PIPs. By defining the lipid environment, this dynamic process known as PI conversion adds membrane identity to endosomal compartments (Di Paolo & Camilli 2006; Balla 2005). Through these mechanisms, Rab GTPases and PI signaling establish and maintain the molecular identity and functional specificity of endosomal compartments. As shown for early endosomes, the Rab and PIP identity is interconnected. On early endosomes, Rab5 recruits PI3K to the membrane leading to Phosphatidylinositol 3- phosphate (PI(3)P) enrichment (Christoforidis et al. 1999). Together with active Rab5, the tethering molecule Early Endosome Antigen 1 (EEA1) then binds the generated PI(3)P via its PI(3)P-binding FYVE domain (Simonsen et al. 1998). Importantly, endosomal identity is dynamic. Endosomal maturation from early to 21 1. Introduction Figure 9.: Endosomal trafficking is a dynamic process regulated by Rab proteins. Rab proteins confer membrane identity to distinct endosomal compartments, for example, Rab4 and Rab5 are associated with early endosomes, Rab11 with recycling endosomes, and Rab9/Rab7 with late endosomes. Importantly, various Rab domains can be found on the same endosomal compartment, forming microdomains that facilitate cargo sorting and endosomal maturation. Adapted after Stenmark (2009). late endosomes includes changes in PIP composition and associated Rabs, termed Rab conversion. Here, the maturation is mediated by Rab5-dependent recruitment of a Rab7 GEF, which in turn recruits and activates Rab7 and its effectors, ulti- mately inactivating Rab5 and promoting a switch from PI(3)P to PI(3,5)P2 (Rink et al. 2005; Huotari & Helenius 2011). It is important to note that different Rab proteins and PIPs coexist on the same endosome in spatially distinct microdomains. Microscopy studies of fluorescently labeled Rab4, Rab5, and Rab11 showed different Rab domains that did not intermix and showed biochemically distinct characteris- tics in transferrin trafficking (Sönnichsen et al. 2000). The endocytic compartments thus form a highly interconnected and dynamic network. 22 1.4. From the Plasma Membrane to Vesicles - Intracellular Trafficking Distinct Rab domains, their downstream effectors, and endosome architecture influence the trafficking of molecules along the endosomal compartments. During endosomal maturation, proteins targeted for degradation are sorted into ILVs by the Endosomal Sorting Complexes Required for Transport (ESCRT) complex (Eden et al. 2012). Upon fusion with lysosomes, ILVs and other intraluminal proteins are then degraded (Elkin et al. 2016). Conversely, transmembrane receptors targeted for recycling can be enriched in tubular extensions of endosomes. This process is driven by the retromer complex, which consists of Sorting Nexin (SNX) family proteins and the Cargo-selective Complex (CSC), which mediate membrane curvature and cargo selection (Burd & Cullen 2014). In a first step, the CSC is recruited to early endosomes through SNX-mediated PI(3)P binding and to late endosomes by interaction with Rab7 (Burd & Cullen 2014). At the same time, cargo selection for recycling is guided by either sorting signals within the cargo molecule or by associated proteins such as SNX family members (Hsu et al. 2012). Eventually, CSC, the cargo, and the Wiskott-Aldrich Syndrome protein and SCAR homologue (WASH) complex drive tubule formation through actin polymerization (Temkin et al. 2011; Gomez & Billadeau 2009). In addition to tubule formation, actin also regulates endocytic trafficking, which is regulated by several Rho GTPases. Several Rho GTPases localize to endosomal compartments. However, it must be distinguished whether a Rho GTPase is a cargo protein or an actual regulator of trafficking. Although Rac1 and Cdc42 have regulatory functions in the secretory pathway (Olayioye et al. 2019), they rather constitute cargo proteins within the endocytic pathway. Here, their activation on Rab5-positive early endosomes and the subsequent trafficking to the plasma membrane were shown to be dependent on the small GTPase Arf6 (Palamidessi et al. 2008; Osmani et al. 2010). In contrast, RhoD has been shown to regulate endosomal trafficking actively. RhoD controls the endosomal formin hDia2C, leading to endosomal actin polymerization, thereby reducing endosome motility (Murphy et al. 1996; Gasman et al. 2003). Moreover, Nehru et al. demonstrated that RhoD, Rab5, and the Rab5 effector Rabankyrin-5 control the trafficking of the PDGF Receptor (PDGFR). Here, despite the loss of RhoD or Rabankyrin-5 induced mislocalization of early endosomes, it was not 23 1. Introduction investigated whether the decreased PDGFR activity was due to prolonged PDGFR internalization dynamics or altered trafficking routes (Nehru et al. 2013). Although some other Rho GTPases, including TCL, RhoG, and the atypical RhoBTB, have also been implicated in trafficking regulation (Olayioye et al. 2019), their regulatory networks and functions during trafficking in collectively 17 publications on PubMed1 remain elusive. As of today, RhoB stands out as the best-characterized Rho protein involved in endosomal trafficking. 1.4.2. Orchestrating Endosomal Trafficking - RhoB All members of the Rho family can be geranylgeranylated, whereas RhoB can additionally undergo farnesylation (Adamson et al. 1992). These post-translational modifications define distinct RhoB pools: geranylgeranylated RhoB localizes to late endosomes and farnesylated RhoB to the plasma membrane (Wherlock et al. 2004). The functional importance of a balanced distribution of RhoB to distinct membranes was demonstrated using a Farnesyl-transferase Inhibitor (FTI). Inhibition of RhoB farneslyation through FTI led to an increased pool of geranylgeranylated RhoB on endosomes. The elevated endosomal pool led to retention of the internalized EGFR in endosomal ILVs by preventing the subsequent transfer into lysosomes (Wherlock et al. 2004). Molecularly, EGFR internalization recruits the RhoB GEF Vav2, and downstream RhoB activation then dictates the intracellular trafficking of EGFR from early endosomes to lysosomes (Gampel & Mellor 2002; Gampel et al. 1999). On endosomes, active RhoB promotes Dia1/mDia2-dependent actin polymerization, regulating their motility (Fernandez-Borja et al. 2005; Wallar et al. 2007). Therefore, increased RhoB activity enhances actin polymerization, which slows endosomal maturation and progression towards lysosomal degradation, thereby prolonging endosomal signaling. Similarly, RhoB regulates cell migration through β1-integrin trafficking. RhoB depletion decreased β1-integrin surface levels, thereby promoting focal contact turnover and lamellipodia dynamics, and thereby impaired directional cell migration (Vega et al. 2012). 1Query: ("Trafficking") AND ((RhoG) OR (TCL) OR (RhoBTB)). last accessed 21.05.2025 24 1.4. From the Plasma Membrane to Vesicles - Intracellular Trafficking RhoB not only regulates receptor trafficking but also controls the spatial avail- ability of other signaling molecules. For example, the delivery of Cdc42 and Rac to the plasma membrane upon PDGF stimulation was shown to be RhoB depen- dent, thereby promoting cell migration (Huang et al. 2011). Moreover, nuclear localization of Akt is RhoB dependent, and RhoB mediates resistance to EGFR inhibition through Akt signaling (Adini et al. 2003; Calvayrac et al. 2017). Thus, balanced RhoB activity is fundamental for the proper trafficking and localization of signaling molecules. Despite the importance of RhoB in various signaling pathways, only a few GEFs and GAPs targeting RhoB are known. GEF-H1 was shown to regulate RhoB, being also involved in the trafficking of Src (Kamon et al. 2006; Arnette et al. 2016). ARHGEF10 and p190BRhoGAP control tight junction integrity in endothelial cells specifically by balancing RhoB activity (Khan et al. 2021; Pierce et al. 2017). In general, the balanced activity of RhoGEFs and RhoGAPs is necessary to facilitate spatiotemporal control of Rho GTPases. Work included in this thesis identified a GEF-GAP pair that regulates endoso- mal RhoB. The RhoGAP DLC3 had previously been shown to regulate Golgi morphology as well as EGFR trafficking and signaling: loss of DLC3 impaired EGFR degradation and prolonged downstream Akt and Erk signaling (Braun et al. 2015). Similarly, depletion of DLC3 trapped the ECM-degrading Membrane Type-1 Matrix Metalloproteinase (MT1-MMP) in early endosomes, promoted its fast recycling to the plasma membrane, and thereby increased matrix degrada- tion capabilities (Noll et al. 2019). Using the Golgi morphology as a readout, the opposing RhoGEF was sought, whose depletion would rescue the Golgi pheno- type. As presented in the second publication of this thesis, Solo was identified as this RhoGEF followed by further characterization (Lungu et al. 2023). Strikingly, Solo regulated the trafficking of Src from the perinuclear region to the plasma membrane (Meyer et al. 2025). One of the best-studied examples of RhoB-dependent trafficking is that of Src. The trafficking of Src and its subcellular localization are coupled to its activity (see Figure 10A-B). As observed using confocal microscopy, Src accumulates in its 25 1. Introduction Figure 10.: RhoB regulates Src trafficking to the plasma membrane. (A) Active Src, indicated by red pY416 staining, localizes to the cell periphery, whereas inactive Src (green) accumulates in the perinuclear region. (B) Upon RhoB depletion, Src fails to localize to the cell periphery and remains in its inactive state in the perinuclear region. Adapted from Sandilands et al. (2004). inactive state in the perinuclear region. Upon PDGF treatment, RhoB-positive vesicles containing Src traffic to the plasma membrane in an actin-dependent manner via Scar1 and/or mDia2 (Sandilands et al. 2004). During this transport, Src becomes activated (Sandilands et al. 2004), leading to a characteristic gradient of inactive Src in the perinuclear region and active Src at the plasma membrane. Other SFK members, such as FYN and YES, also rely on the trafficking to the plasma membrane. However, because FYN and YES are palmitoylated, their trafficking is independent of RhoB (Sandilands & Frame 2008). The example of Src highlights the critical role of RhoB in ensuring proper spatial and temporal signaling. 26 1.5. Microscopy – a Versatile Technology in Biology 1.5. Microscopy – a Versatile Technology in Biology As demonstrated by various examples in the previous sections, visualizing cells and proteins is essential for uncovering molecular mechanisms and signaling pathways involved in cell migration and intracellular trafficking. As cells contain multiple compartments and time-dependent changes can occur in the matter of seconds, microscopy enables to follow dynamics and short-lived observations in living cells. Although the technical capabilities of confocal imaging have advanced over time, the fundamental principles have remained unchanged. Importantly, the quality and method of image acquisition determine whether reliable observations, accurate quantification, and meaningful conclusions can be achieved. 1.5.1. Foundational Concepts of Confocal Fluorescence Imaging Coupling a protein of interest with a fluorophore enables the visualization of its specific intracellular localization and is fundamental for studying dynamic cellular processes. Fluorophores are characterized by distinct energy levels (Monks 2022). Upon absorption of a photon, the fluorophore transitions from the ground electronic state S0 to the first excited state S1, which harbors higher energy. As the molecule undergoes vibrational relaxation, the fluorophore eventually returns to the S1 state, thereby emitting a photon. During this process, the energy of the emitted photon is reduced, resulting in a longer wavelength than that of the absorbed photon. For example, the excitation and emission maxima of EGFP are 488 nm and 507 nm, respectively2. By filtering excitation and emission light appropriately, background signals can be further reduced, and image contrast can be enhanced. This ensures that only the protein of interest is visualized, highlighting a major advantage of fluorescence imaging. In combination with confocal microscopy a high spatial resolution can be achieved. The foundational concept of confocal microscopy is that both the illumination and detector focus the same focal spot in XY and Z position of the specimen (see Figure 11A). Light from a monochromatic laser is focused with mirrors, lenses, and the objective to illuminate a specific focal point in the specimen. In Laser Scanning Microscopes (LSMs), the laser beam is guided by motorized mirrors to 2https://www.fpbase.org/protein/egfp/ 27 1. Introduction scan hundreds of focal points along the X- and Y-axis of the sample. Fluorescence light emitted from the sample is then collected by the objective and routed to the detector. Additional filters within the light path, such as dichroic mirrors, play a key role in separating the emission signal from brighter excitation and background light. The dichroic mirror reflects the excitation wavelength from the laser towards the specimen while allowing emitted light of longer wavelengths to pass towards the detector. Before reaching the detector, emitted light passes through a pinhole aperture. This diaphragm removes out-of-focus light from planes above and below the focal plane by excluding higher-order components of the Airy disc. Light exclusion is a critical step, because the illuminated point is, in reality, rather a volume than a point: it is not absolutely restricted in XY size and the sample is excited along the Z-axis (D’Antuono 2022). While the pinhole effectively improves optical sectioning, it also reduces the total signal, thereby decreasing the Signal-to-Noise Ratio (SNR) (Weisshart 2014). By imaging multiple time points, combining different Z-planes to gain axial resolu- tion, and using different excitation and emission wavelengths to visualize distinct cellular components, confocal microscopy offers a broad applicability (see Fig- ure 11B). However, the more complex the imaging setup, the longer the total exposure time, which can lead to photobleaching and reduce signal intensity. This is especially critical with living samples, as prolonged exposure can induce photo- toxic effects. Therefore, several key aspects need to be considered when selecting an appropriate imaging approach. Depending on the scientific question, two key aspects need to be considered: the type of sample and the required resolution. On the one hand, fixed cells are stained with fluorescently labeled antibodies, and weak signals can be enhanced by increasing the laser intensity. However, fixation only captures a static snapshot and can alter the distribution of proteins. In contrast, live-cell imaging allows the study of process dynamics but requires fluorescently-tagged proteins and/or live-cell stains. On the other hand, the necessary resolution needs to be considered. For this, Nyquist’s sampling criterion can be used as a helpful guideline, which states that at least two to three pixels together should resolve the feature of interest: lower resolution results in information loss (undersampling), while excessively high resolution does 28 1.5. Microscopy – a Versatile Technology in Biology Figure 11.: Working principle and acquired data of a confocal microscope. (A) During confocal imaging, both the illumination and detector are focused on the same focal volume within the specimen. This is achieved through a combination of lenses and the objective. Dichroic mirrors, such as the beamsplitter, and additional filters ensure that only fluorescence emitted from the fluorophore reaches the detector. A pinhole placed in front of the detector excludes out-of-focus light, ensuring a high spatial resolution (Jonkman et al. 2020). (B) In addition to acquiring single-wavelength 2D frames, microscopes enable the excitation and detection of multiple wavelengths, allowing the visualization of different proteins within the same cell. By imaging living cells over time, multidimensional time-lapse can be obtained. Compared to widefield imaging, confocal imaging offers higher axial resolution, making it suitable for generating 3D stacks that include the Z-dimension (Sanderson 2022). not provide additional insights but increases data size (oversampling) (Rachid 2022). Proper sampling ensures efficient data acquisition without compromising scientific validity. To select an adequate objective, the Rayleigh criterion and Abbe’s resolution criterion can be used to estimate the minimum resolvable distance between two points. Accordingly, the lower the wavelength and the higher the Numerical Aperture (NA) of the objective, the higher the resolution (Sarkar 2022). Based on the imaging conditions and necessary resolution, different confocal micro- scopes offer distinct advantages (Elliott 2020; Brzostowski & Sohn 2021). Classic LSMs typically feature an adjustable pinhole size, allowing for minor resolution enhancement. However, due to their slower acquisition speed, they are less suited for live-cell imaging compared to Spinning Disk Confocal Microscopes (SDCMs). SDCMs employ a spinning disk with multiple pinholes, which increases depth 29 1. Introduction penetration and acquisition time and thereby reduce phototoxicity in living cells. A limitation is pinhole crosstalk that can introduce artifacts and slightly reduce resolution. Instead of using a single detector, Airyscan utilizes an array of 32 detec- tors functioning as separate pinholes.This configuration enables lateral resolution of up to 120 nm, compared to 220 nm of a conventional confocal microscope (Wu & Hammer 2021; Weisshart 2014). Compared to a single pinhole, this reflects the strong signal of a large pinhole while generating the resolution of a small pinhole (Reilly & Obara 2021). Due to this, Airyscan supports fast acquisition and low phototoxicity while still maintaining high lateral and axial resolution. However, airyscan generates up to 32 x more data, which can affect data management and slow down subsequent image processing and quantification. The advancements in confocal live-cell imaging also enable the use of sophisticated biosensors to visualize protein activity in real-time within living cells. 1.5.2. Making It Visible - Imaging-based Biosensors The use of biosensors in living cells offers a unique opportunity to assess protein activity and localization with high spatial and temporal resolution. This section outlines general principles, advantages, and limitations, with a focus on Rho GTPase biosensors as well as the Bimolecular Anchor Detector (BiAD) technology, which was used in this thesis to probe the epigenetic landscape. Visualizing Rho GTPase Activity – Rho Biosensors Several tools are available to measure Rho GTPase activity. A common approach in- volves retreaving active Rho GTPases from whole-cell lysates using the Rho-binding Domain (RBD) of a downstream effector (Koh et al. 2022). While this approach allows Rho isoform-specific activity measurements, it indicates only population-wide Rho GTPase activity and lacks spatiotemporal resolution. Similarly, antibodies that bind to the GTP-loaded Rho GTPase can additionally be used in immunoflu- orescence to reveal subcellular localization. However, because such antibodies are rare and require fixed cells, this approach is rather limited in its applicability and cannot capture dynamic activity changes. 30 1.5. Microscopy – a Versatile Technology in Biology To address the dynamics, RBDs can be fused to fluorophores to create live-cell relocation sensors. These reversibly bind to GTP-loaded Rho GTPases and dy- namically accumulate at sites of activity. One example is the C-terminus of Anillin, which includes the Anillin Homology (AH) and PH domains tagged with GFP (GFP-AHPH) (Budnar et al. 2019; Yap et al. 2015). Relocation sensors enable the detection of endogenous Rho GTPase activity, thereby eliminating the need to overexpress exogenous Rho proteins. However, they do not distinguish between RhoA, RhoB, and RhoC and therefore lack isoform specificity. For isoform-specific analysis of Rho GTPases, FRET-based biosensors provide a precise alternative. These biosensors consist of a specific Rho GTPase and an RBD, each fused to one of two fluorescent proteins with overlapping spectral emission and absorption profiles (Donnelly et al. 2014). When the donor fluorophore is excited by a photon, energy transfer to the acceptor fluorophore occurs only if the two are within 10 nm (Andrews et al. 2015; Verma et al. 2023). Consequently, FRET and the corresponding emission from the acceptor only occur if the active Rho GTPase interacts with the RBD. By coupling this interaction-dependent FRET signal to the active conformation of Rho GTPases, FRET sensors provide a dynamic, quantitative readout for Rho activity (Martin et al. 2016). The ratiomeric nature of FRET imaging, based on two independent channels, improves the robustness against photobleaching and differences in expression levels. Although FRET biosensors allow isoform-specific measurements, they require overexpression of exogenous Rho GTPases; the readout reflects the activity of the introduced protein rather than the endogenous Rho GTPase pool. Additionally, FRET probes may displace endogenous signaling proteins, potentially disrupting native Rho GTPase signaling (Donnelly et al. 2014). This limitation could be addressed by the use of smaller fluorescent tags. Fluorescence complementation offers a potential strategy to reduce perturbations introduced by chimeric Rho GTPases. In this approach, a fluorophore is split into two (as in Bimolecular Fluorescence Complementation (BiFC)) or more fragments, which must reconstitute into the native structure to regain fluorescence capability (Kerppola 2006). For example, the tripartite Split-GFP relies on the reconstitution of three fragments. A key advantage of this approach is that two of the fragments, 31 1. Introduction GFP10 and GFP11, are only 18 and 20Amino Acids (AAs) long, and are thus presumed to minimally interfere with protein interactions and activity (Cabantous et al. 2013). Based on this design, endogenous tagging of Rho GTPases in combina- tion with the overexpression of a correspondingly tagged RBD could enable the monitoring of Rho activity at the endogenous level. However, the RBD may still perturb downstream signaling. Since reconstitution in both BiFC and tripartite Split-GFP systems is irreversible (Cabantous et al. 2013), this irreversibility could outcompete endogenous effectors and potentially shut down Rho signaling entirely. Although reconstitution in the Split-GFP system is tightly regulated by the addi- tion of recombinant GFP1-9 (Castillo et al. 2023), this addition happens on the entire population, limiting the ability to image pre- and post-addition to only a few concurrent cells per well. Finally, as fluorescence complementation can reduce fluorophore brightness, the Split-GFP system also relies on an additional nanobody to enhance signal intensity (Koh et al. 2022; Castillo et al. 2023). The development of more straightforward systems with brighter fluorophores that address these limitations could lead to a promising new generation of Rho GTPase activity sensors. The principle of fluorescence complementation is also employed in the BiAD sensor to probe the epigenetic landscape. Visualizing the Epigenome – The BiAD Technology Epigenetics refers to heritable and reversible mechanisms of gene regulation, pri- marily mediated by DNA methylation and histone modifications. DNA is wrapped around an octamer of histones H2A, H2b, H3, and H4 forming the nucleosome (Kornberg 1974). Nucleosomes can further condense into supramolecular structures such as the 30 nm fiber or a chromatid of a chromosome, collectively referred to as chromatin (Luger et al. 2012). The spatial organization of chromatin is a critical determinant of DNA accessibility, thereby directly influencing gene expression and cell identity (Deans & Maggert 2015). DNA Methyltransferases (DNMTs) catalyze the addition of methyl groups to cytosines within CpG dinucleotide-rich CpG islands. At promotors, DNA methylation locally reduces the negative electrostatic potential of the DNA, which can impede binding of TFs and subsequently lead to gene silencing (Kulis & Esteller 2010; Dantas Machado et al. 2014). The regulatory function of DNA methylation also extends beyond promotor regions at enhancers or repetitive elements. 32 1.5. Microscopy – a Versatile Technology in Biology The functional outcome of DNA methylation depends on the cellular and genomic context. While DNA promotor methylation typically inhibits expression, DNA methylation within the gene bodies correlates with gene expression and may suppress spurious transcription initiation within active genes (Ball et al. 2009; Suzuki & Bird 2008). Moreover, DNA methylation at repetitive elements, including centromeric α-satellites, plays a critical role in maintaining genome integrity, serving both struc- tural and regulatory functions (Pappalardo & Barra 2021). As such, hypomethyla- tion of repetitive DNA elements can lead to genomic instability, thereby contributing to tumor heterogeneity in cancer (Besselink et al. 2023). Because epigenetic reg- ulation is highly context-specific, there is a need for tools that can dynamically monitor locus-specific changes of epigenetic modifications. Studying the epigenetic landscape poses technical challenges that typically re- quire relatively static, endpoint-based methods. For example, assessing DNA methylation patterns often requires cell lysis followed by High-Performance Liquid Chromatography (HPLC) or digestion with Methylation-Sensitive Resctriction Endonucleases (MSRE) and subsequent Next Generation Sequencing (NGS). Al- though these methods allow a locus-specific readout, they are population-wide and lack temporal resolution. Alternatively, fluorescently labeled 5-Methylcytosine Binding Domains (MBDs) can be used to visualize DNA methylation in living cells (Jeltsch et al. 2020). Although this enables monitoring of DNA methyla- tion changes over time at the single-cell level, MBDs alone lack locus speci- ficity and therefore indicate only global methylation patterns. Tools are thus needed that allow the dynamic, locus-specific readout of epigenetic modifications with single-cell resolution. The modular fluorescence complementation-based BiAD technology effectively fills this methodological gap, enabling the visualization of dynamic, locus-specific epigenetic changes with single-cell resolution (Lungu et al. 2017). These sensors comprise a DNA-binding anchor that targets a defined DNA sequence and a detector containing a domain that recognizes a specific epigenetic modification. Each component is fused to complementary fragments of the fluorescent protein mVenus. Fluorescence is reconstituted only when the anchor and detector are brought into proximity at a locus carrying the modification of interest. Therefore, 33 1. Introduction mVenus fluorescence reports the presence of the specific epigenetic mark at the targeted site (Lungu et al. 2017). These sensors allow to directly link the epigenetic information with information on cell signaling and physiology and were used in this thesis to monitor the DNA mehtylation of α-satellites. 1.5.3. From Pixels to Information - Image Quantification A digital image, as acquired by microscopes, is essentially a two-dimensional matrix of numbers, with each number representing the intensity at a given pixel. Due to this, image analysis follows a general mathematical workflow that incorporates op- erations such as thresholding, filtering, and Boolean logic (see Figure 12, Rebollo & Bosch 2019). Following adequate image acquisition, segmentation extracts a binary image that separates the objects of interest in the foreground from the background, also referred to as a mask. However, challenges such as background noise, incom- plete structures, or merged objects often require further refinement. Therefore, each object of interest needs to be extracted from the binary image for accurate downstream analysis. Lastly, quantitative features such as size, shape, and intensity of these objects can be extracted and represented as numerical data. These steps can be carried out using established programs such as FIJI or CellProfiler that are readily available for several systems. While more advanced deep learning and machine learning tools are also available within FIJI as DeepImageJ (Gómez-de- Mariscal et al. 2021) and Ilastik (Berg et al. 2019), others may require integration into custom pipelines written in Python or Matlab pipelines. Ultimately, quanti- tative comparison of this data across experimental conditions allows researchers to detect and interpret biological changes. It is therefore evident that accurate segmentation is essential for reliable image quantification. Images often require preprocessing to improve segmentation accuracy. For exam- ple, inhomogeneous illumination during image acquisition is often overlooked and can significantly impair segmentation. Such illumination artifacts can be reduced by subtracting reference images or applying correction filters (Smith et al. 2015). Moreover, inevitable background noise can interfere with the segmentation. To address this, classical filters such as a Gaussian kernel can be used to homogenize the intensity distribution (Pertusa & Morante-Redolat 2019), or deep learning 34 1.5. Microscopy – a Versatile Technology in Biology Figure 12.: Basic image quantification workflow. Image quantification begins with the image acquisition, which determines the dimension and modality of the digital image. Through image processing and segmentation, a binary image can be generated using various approaches, for example, thresholding based on the histogram. Although the binary image includes the objects of interest, binary image processing ensures that false-positive objects are excluded. By extracting features that describe the objects, such as size, shape, and intensity, numerical data can be obtained for subsequent analysis and comparison. Adapted from Pertusa & Morante-Redolat (2019). tools can be applied. One such tool, Noise2Void uses a Convolutional Neural Network (CNN) to denoise images (Krull et al. 2019). As Noise2Void does not rely on ground-truth data for training, it is particularly suitable for microscopy images where this data is usually unavailable. Following preprocessing, the con- trast between object and background is ideally enhanced, improving segmentation performance. Segmentation can be broadly divided into two main approaches (Caicedo et al. 2017). In the model-based approach, an algorithm with predefined parameters is applied based on visual inspection. Here, objects may be identified through, e. g., thresholding, i. e. their intensities exceed a specific value, or through edge detection using filters that highlight discontinuities (Pertusa & Morante-Redolat 2019). 35 1. Introduction Alternatively, watershed segmentation is beneficial for separating touching objects. For example, while two adjacent GFP-positive cells may not be distinguishable by thresholding alone, watershed segmentation additionally analyzes the intensity gradient of surrounding pixels. Metaphorically, if local intensity maxima are treated as peaks, “waterfalls” flowing from each peak will meet at the cell boundaries, effectively separating the two cells. In contrast, machine learning approaches offer greater flexibility and are well-suited for more complex segmentation tasks. However, achieving good model performance typically requires large training datasets composed of (manually) labeled ground- truth data. The need for such ground-truth data can hinder the implementation of machine learning approaches. To address this, projects such as the BioImage Model Zoo aim to promote the exchange of annotated imaging datasets and AI tools within the bioimaging community, lowering the entry threshold for machine learning applications. As such, the deep-learning-based segmentation tool Cellpose offers pretrained models for cell segmentation (Stringer et al. 2021). Using a human- in-the-loop approach, these models can be further adapted to individual datasets by annotating only up to 200 cells, achieving performance comparable to models trained on datasets with up to 200 000 cells (Stringer & Pachitariu 2024). Machine learning approaches are not limited to cell segmentation: the AI-driven framework SpotMAX detects and quantifies globular 3D structures (Padovani et al. 2024). Similarly, TrackMate enables both segmentation and tracking of objects over time, which is particularly relevant for time-lapse studies (Ershov et al. 2022). Machine learning based approaches, therefore, represent a valuable tool within quantification pipelines. Last but not least, the role of the scientist as a human-in-the-loop remains indis- pensable - and will likely continue to be. As long as a quantification pipeline is technically working, it will produce output data, regardless of input quality or even when flawed image operations have been applied. It is therefore critical to ensure that all steps of the workflow function correctly and that the imaging and experimental conditions are the appropriate ones to begin with. 36 1.6. Aims of this Thesis 1.6. Aims of this Thesis Cellular homeostasis relies on the integration of extracellular biochemical and biophysical cues through tightly regulated intracellular signaling cascades. Imbal- anced signaling resulting from cellular transformation can initiate tumor formation and promote cancer progression. Cell migration and invasion, critical steps in metastasis, are orchestrated by spatiotemporally controlled signaling pathways. Here, the actin cytoskeleton and upstream Rho GTPase coordinate key processes, including endocytic transport, membrane protrusion formation, and mechanical force generation. In addition, the actin cytoskeleton is linked to both focal ad- hesions and cell-cell junctions, underscoring its central role during cell motility. Therefore, identifying the involved regulatory proteins and understanding their molecular mechanisms is essential for unraveling how migration and invasion are controlled in cancer. The RhoGEF Solo is an established regulator of cell migration via activation of RhoA and RhoC. However, no post-translational modifications of Solo have been characterized to date. Based on database mining and preliminary experiments, tyrosine 242 (Y242) of Solo was identified as a candidate phosphorylation site, potentially targeted by SFKs. The aim of the first study included in this thesis, Meyer et al. (2025), was to determine the functional significance of this phosphory- lation, particularly in the mesenchymal cell state and during cell motility of breast epithelial cells. Additionally, photoconvertible c-Src-mEos, Solo overexpression, and Solo depletion were employed to investigate Solo’s role in modulating Src trafficking and activity. Building on previous findings from the Olayioye group that identified Solo as a potential antagonist of DLC3, the second study, Lungu et al. (2023), aimed to elucidate the molecular regulation of RhoB by Solo and DLC3. The molecular regu- lation was addressed using a combination of the GFP-AHPH Rho GTPase activity biosensor and the optogenetic Opto-Solo construct, which enables light-induced recruitment of Solo’s GEF domain to endosomes. This approach was complemented by overexpression and knock-down experiments to analyze endosomal RhoB pools. To evaluate the functional consequences of impaired trafficking, EGFR trafficking and downstream signaling were assessed in Solo-depleted cells. 37 1. Introduction Given that motility is partially coupled to cell shape, the third study explored whether the geometric organization of cells encodes information predictive of migra- tory behavior. In collaboration with the Guo Lab (MIT; Mechanical Engineering), a Graph-based Neural Network (GNN) was developed based on data acquired during this thesis. As presented in Yang et al. (2024), this GNN was trained to predict collective cell migratory behavior based on a single snapshot of cells. To train the model across diverse migration and shape patterns, cells were cultured under varying conditions to induce a broad range of migratory phenotypes. Finally, migrating cells also undergo changes in gene expression and chromatin orga- nization. While the mechanisms of epigenetic modifications and their impact on gene expression are well understood, the effect of extracellular cues on the epigenome re- mains elusive. The last study of this thesis, Brenner et al. (2024), aimed to character- ize confluence-dependent DNA methylation changes specifically at α-satellites. Us- ing the