What are the key mechanical mechanisms governing integrin-mediated cell migration in three-dimensional fiber networks?

Cell migration plays a vital role in numerous processes such as development, wound healing, or cancer. It is well known that numerous complex mechanisms are involved in cell migration. However, so far it remains poorly understood what are the key mechanisms required to produce the main characteristics of this behavior. The reason is a methodological one. In experimental studies, specific factors and mechanisms can be promoted or inhibited. However, while doing so, there can always be others in the background which play key roles but which have simply remained unattended so far. This makes it very difficult to validate any hypothesis about a minimal set of factors and mechanisms required to produce cell migration. To overcome this natural limitation of experimental studies, we developed a computational model where cells and extracellular matrix fibers are represented by discrete mechanical objects on the micrometer scale. In this model, we had exact control of the mechanisms by which cells and matrix fibers interacted with each other. This enabled us to identify the key mechanisms required to produce physiologically realistic cell migration (including advanced phenomena such as durotaxis and a biphasic relation between migration efficiency and matrix stiffness). We found that two main mechanisms are required to this end: a catch-slip bond of individual integrins and cytoskeletal actin-myosin contraction. Notably, more advanced phenomena such as cell polarization or details of mechanosensing were not necessary to qualitatively reproduce the main characteristics of cell migration observed in experiments.

Neural network surgery: Combining training with topology optimization.

With ever increasing computational capacities, neural networks become more and more proficient at solving complex tasks. However, picking a sufficiently good network topology usually relies on expert human knowledge. Neural architecture search aims to reduce the extent of expertise that is needed. Modern architecture search techniques often rely on immense computational power, or apply trained meta-controllers for decision making. We develop a framework for a genetic algorithm that is both computationally cheap and makes decisions based on mathematical criteria rather than trained parameters. It is a hybrid approach that fuses training and topology optimization together into one process. Structural modifications that are performed include adding or removing layers of neurons, with some re-training applied to make up for any incurred change in input-output behaviour. Our ansatz is tested on several benchmark datasets with limited computational overhead compared to training only the baseline. This algorithm can achieve a significant increase in accuracy (as compared to a fully trained baseline), rescue insufficient topologies that in their current state are only able to learn to a limited extent, and dynamically reduce network size without loss in achieved accuracy. On standard ML datasets, accuracy improvements compared to baseline performance can range from 20% for well performing starting topologies to more than 40% in case of insufficient baselines, or reduce network size by almost 15%.

What do cells regulate in soft tissues on short time scales?

Cells within living soft biological tissues seem to promote the maintenance of a mechanical state within a defined range near a so-called set-point. This mechanobiological process is often referred to as mechanical homeostasis. During this process, cells interact with the fibers of the surrounding extracellular matrix (ECM). It remains poorly understood, however, what individual cells actually regulate during these interactions, and how these micromechanical regulations are translated to the tissue-level to lead to what we observe as biomaterial properties. Herein, we examine this question by a combination of experiments, theoretical analysis, and computational modeling. We demonstrate that on short time scales (hours) - during which deposition and degradation of ECM fibers can largely be neglected - cells appear to not regulate the stress / strain in the ECM or their own shape, but rather only the contractile forces that they exert on the surrounding ECM. STATEMENT OF SIGNIFICANCE: Cells in soft biological tissues sense and regulate the mechanical state of the extracellular matrix to ensure structural integrity and functionality. This so-called mechanical homeostasis plays an important role in the natural history of various diseases such as aneurysms in the cardiovascular system or cancer. Yet, it remains poorly understood to date which target quantity cells regulate on the mircroscale and how it translates to the macroscale. In this paper, we combine experiments, computer simulations, and theoretical analysis to compare different hypotheses about this target quantity. This allows us to identify a likely candidate for it at least on short time scales and in the simplified environment of tissue equivalents.

Capping protein-controlled actin polymerization shapes lipid membranes.

Arp2/3 complex-mediated actin assembly at cell membranes drives the formation of protrusions or endocytic vesicles. To identify the mechanism by which different membrane deformations can be achieved, we reconstitute the basic membrane deformation modes of inward and outward bending in a confined geometry by encapsulating a minimal set of cytoskeletal proteins into giant unilamellar vesicles. Formation of membrane protrusions is favoured at low capping protein (CP) concentrations, whereas the formation of negatively bent domains is promoted at high CP concentrations. Addition of non-muscle myosin II results in full fission events in the vesicle system. The different deformation modes are rationalized by simulations of the underlying transient nature of the reaction kinetics. The relevance of the regulatory mechanism is supported by CP overexpression in mouse melanoma B16-F1 cells and therefore demonstrates the importance of the quantitative understanding of microscopic kinetic balances to address the diverse functionality of the cytoskeleton.

The evolution of radiation dose over time: Measurement of a patient cohort undergoing whole-body examinations on three computer tomography generations.

OBJECTIVES: To evaluate and compare the radiation dose and image quality of whole-body-CT (WBCT) performed on the 3rd-generation dual-source-CT (DSCT) with 2nd-generation DSCT and 64-slices-Single-Source-CT (SSCT) in a large patient cohort. MATERIAL AND METHODS: Using a monitoring and tracking software 1451, 747 and 1861 patients scanned with a one-spiral-thorax-abdomen-pelvis-CT-examination on a 3rd-, 2nd-generation DSCT and SSCT, respectively, were extracted from the PACS server. For the intra-individual analysis, 203 patients on the 3rd-generation DSCT were identified. Out of those 203 patients, 155 had the same examination on the 2nd-generation DSCT, 91 patients had the same examination on the SSCT and 43 patients had an examination on all three CT-generations. Automatic tube current modulation was active on all three CT-generations, whereas automatic tube voltage selection was only available on both DSCT-generations. Dose was recorded by the size-specific-dose-estimate-method (SSDE); signal-to-noise-ratio (SNR) and contrast-to-noise-ratio (CNR) were calculated placing a ROI on the ascending aorta/liver and the subcutaneous adipose tissue at comparable level. Image quality of axillary and mediastinal lymph nodes and adrenal glands was assessed by two experienced radiologists. RESULTS: Subjective image quality was excellent throughout all three CT-generations (p=0.38-0.98). Quantitative image quality in both DSCT generations was superior to SSCT (p<0.001). SNR and CNR in the liver parenchyma were superior in the 3rd-generation DSCT compared to the 2nd generation DSCT (p<0.001), whereas there was no difference in the aorta. In the inter-individual analysis, CTDIvol was lower by 26.9% and 44.3% in the 3rd-generation DSCT, when compared to the 2nd-generation DSCT and SSCT, respectively; SSDE was lower by 31.5% and 51% in the 3rd-generation DSCT, when compared to the 2nd-generation DSCT and SSCT, respectively. In the intra-individual comparison CTDIVol in the 3rd-generation DSCT was lower by 33% and 45%, when compared to the 2nd-gneration DSCT and the SSCT, respectively. Consequently, SSDE in the 3rd-generation DSCT was lower by 29% and by 43% when compared to the 2nd-generation DSCT and SSCT, respectively. CONCLUSION: State-of-the-art CT-equipment substantially reduce radiation dose without affecting image quality.