Rosiello, Valerio (2025) Cell gym: an automated cell-stretcher for mechanoprogramming. [Tesi di dottorato]

Documento PDF PHD_Thesis_revisioned.pdf Visibile a [TBR] Amministratori dell'archivio Download (16MB) | Richiedi una copia

Abstract

Cells are inherently subjected to mechanical stimuli arising from their microenvironment, which profoundly influence their morphology, function, and fate. The emerging field of mechanobiology seeks to understand how these mechanical cues are sensed, transmitted, and converted into biochemical signals—a process that bridges the extracellular matrix (ECM) to the nucleus through an integrated mechanical continuum involving focal adhesions, the cytoskeleton, and the LINC (Linker of Nucleoskeleton and Cytoskeleton) complex. Within this framework, the nucleus is not only a repository of genetic information but also a dynamic mechanosensitive organelle capable of translating physical forces into changes in chromatin organization and gene expression. These insights have paved the way for the nascent discipline of Mechanomedicine, which aims to employ controlled mechanical stimulation as a therapeutic and diagnostic approach complementary to conventional pharmacological interventions. Despite significant advances, a critical challenge in mechanobiology lies in identifying the quantitative relationship between external mechanical stimuli and nuclear response, as well as in controlling these deformations in a reproducible manner. Commercial cell-stretching devices—whether pneumatic or electromechanical—typically lack integration with microscopy systems and provide limited flexibility in defining or monitoring mechanical profiles. Consequently, real-time imaging and quantitative feedback of nuclear and cytoskeletal dynamics during stimulation are difficult to achieve. To address these limitations, this work presents the design, realization, and validation of a novel open-source electromechanical cell stretcher, fully integrated with a confocal microscope and capable of applying precisely controlled, symmetric, and programmable strain profiles to living cells while enabling continuous live imaging. This system, named “Cell-Gym”, combines mechanical precision, software-driven control, and compatibility with image-based feedback, enabling both static and dynamic mechanical loading while maintaining environmental stability within an incubator. The device was conceived as a symmetric biaxial electromechanical stretcher actuated by stepper motors coupled to double-threaded lead screws, ensuring mirror-like deformation of the PDMS chamber and preventing field-of-view displacement during microscopy. This symmetric configuration allows the same region of interest to remain within the focal plane throughout the stretching process—an essential feature for time-resolved experiments. The chamber design was optimized for uniaxial deformation up to 30%, using biocompatible PDMS substrates of tunable stiffness and incorporating nanogrooved membranes (700 nm ridge/groove width, 250 nm depth) to impose controlled cellular alignment along the stretching axis. The open-source control platform based on Arduino-driven DM542 drivers and modular software permits full programmability of strain magnitude, frequency, and waveform, thus allowing future integration with machine-learning-based feedback control systems for adaptive stimulation. The experimental investigation focused on quantifying nuclear deformation in NIH-3T3 fibroblasts subjected to controlled uniaxial strain, using live fluorescence imaging of Hoechst-stained nuclei. Morphological descriptors including nuclear area, major and minor axes, aspect ratio (AR), and orientation angle were extracted before and after stimulation to characterize the mechanical transfer from substrate to nucleus. A first series of step-response experiments revealed a proportional relationship between substrate strain and nuclear deformation, with an effective transmission gain of approximately 0.15 at 30% strain. The results demonstrated that nuclear deformation is strongly influenced by cellular alignment: when cells were randomly oriented on flat substrates, heterogeneous and incoherent nuclear responses were observed; conversely, nanogrooved substrates promoting uniform cell alignment along the stretching axis yielded more homogeneous deformations and reduced population variance. This confirmed that geometric coherence of the actin cytoskeleton is a prerequisite for consistent nuclear mechanotransduction. Additional experiments clarified the role of strain direction, slope, and substrate chemistry. Orthogonal stretching relative to cell orientation produced smaller deformations and localized nuclear invaginations, likely resulting from mechanical decoupling within the LINC complex. Reducing the ramp slope (i.e., slower strain application) attenuated the amplitude of nuclear deformation, revealing a viscoelastic relaxation behavior of the cytoskeleton–nucleus system. The chemical surface modification of PDMS substrates also exerted a significant influence: APTES silanization followed by fibronectin coating enhanced cell adhesion and actin organization, leading to larger and more stable nuclear deformations compared to surfaces with only adsorbed fibronectin. This finding suggests that the mode of fibronectin immobilization—and consequently the strength of focal adhesion anchoring—plays a pivotal role in dictating the efficiency of force transmission to the nucleus. To explore the temporal evolution of nuclear deformation beyond the static regime, the study introduced quasi-static alternating strain profiles, consisting of cyclic uniaxial deformations of low frequency and long duration (three hours). This protocol allowed the observation of cumulative and potentially adaptive nuclear responses under sustained mechanical input. Quantitative image analysis revealed an increase in population variance compared to not-stretched controls, indicative of heterogeneous mechanoadaptation. Some nuclei exhibited sustained elongation (increasing AR), whereas others displayed a reduction, evidencing the coexistence of divergent subpopulations within the same mechanically stimulated culture. This phenomenon was particularly evident on flat substrates, whereas nanogrooved and APTES-treated surfaces promoted a more coherent response, reinforcing the notion that substrate topography and chemistry synergistically regulate nuclear mechanosensitivity. A deeper analysis of nuclear mechanics under quasi-static stimulation revealed significant reorganization of the nuclear lamina (Lamin A/C) and the focal adhesion network, confirming the structural continuum that transmits mechanical cues from the ECM to the nucleus. Lamin intensity and distribution changes indicated local stiffening and remodeling consistent with sustained mechanical loading, while mature focal adhesions (as evidenced by vinculin and paxillin organization) correlated with enhanced nuclear elongation. These findings highlight how focal adhesion maturation and cytoskeletal pre-stress dictate the efficiency of mechanical signal propagation to the nucleus. Building on this experimental foundation, the final phase of the work focused on data-driven modeling of the substrate–nucleus system. Using machine-learning approaches, particularly binary decision-tree algorithms, the study aimed to predict whether a nucleus would increase or decrease its aspect ratio following stimulation, based solely on its initial morphological parameters. Six key features—nuclear area, major axis, minor axis, aspect ratio, orientation angle, and cellular shape index—were employed as input variables. The trained model achieved an accuracy close to 80%, effectively classifying nuclei into six morphologically distinct subpopulations with opposite deformation trends. Importantly, the analysis revealed that the minor axis length and cellular shape index were the most discriminative features, underscoring their mechanistic link to nuclear deformability. This predictive framework demonstrates that nuclear and cellular morphology encodes sufficient information to forecast mechanical response, paving the way for simulation-based planning of mechanical stimulation protocols without the need for real-time feedback. Overall, this doctoral work establishes an integrated experimental and computational methodology to quantitatively control, predict, and interpret nuclear deformation in living cells under well-defined mechanical stimuli. The “Cell-Gym” system provides a versatile and reproducible platform for mechanobiological research, bridging the gap between engineering design and biological function. The combination of open-source hardware, software-based microscope integration, and data-driven analysis represents a significant advance toward standardizing mechanical stimulation protocols and enabling mechanoprogramming—the deliberate modulation of cellular function and fate through mechanical inputs. The results highlight that nuclear deformation is not merely a passive mechanical consequence but an active, regulated process encoding long-term cellular memory. By elucidating how mechanical inputs can be translated into predictable nuclear outcomes, this research contributes to the foundation of Mechanomedicine, offering new perspectives for developing mechanical therapies and designing biomaterials and devices capable of inducing specific cellular responses. The approach presented herein—combining engineering innovation, quantitative imaging, and machine-learning-based prediction—lays the groundwork for future adaptive systems capable of closed-loop mechanical control of living tissues and for the broader application of mechanical principles to regenerative medicine, disease modeling, and cell reprogramming.

Downloads

Actions (login required)

Modifica documento