Article Review (4): Image-based characterization of the cytoskeleton

Article: Alhussein G, et al. A spatiotemporal characterization method for the dynamic cytoskeleton. Cytoskeleton 2016; 73: 221–232.

Summary prepared by Lauren Lanier for discussion on 11-16-17:

This article presents a method for the quantification of spatiotemporal cytoskeletal patterns using a technique studied in depth by Dr. Haidekker known as fractal analysis. To begin, the article gives a concise background on the cytoskeleton, a supporting structure of all cells which integrates the outward nucleus intracellular signaling and the inward transmission of extracellular mechanobiological cues from a cell’s immediate microenvironment. The cytoskeleton is directly involved in virtually all cellular processes from maintaining cell shape and migration to more subtle functions such as intracellular signaling. Until now, there has been a significant gap between qualitative and quantitative understanding of cytoskeletal function. Live imaging technology has been used to study important subtle events of the cytoskeleton to further our understanding of its role, but as of now there is a lack of sufficient computational tools to accurately quantify the massive amount of output spatial and temporal data that comes from live imaging. The author then delves into ways previous studies have attempted quantifying time-dependent cytoskeletal changes and their limitations before introducing the method planned to be used in this study: fractal analysis. According to the article, fractal analysis “measures a fractal dimension (Df), a real number that represents pattern complexity, and inner self-similarity as calculated on different spatial scales” and the authors hope that fractal analysis can be sensitive enough to sufficiently detect subtle physiological changes of the cytoskeleton. Before launching into the methods of the study, the author finishes the introduction with an explanation of how this use of fractal analysis is novel for cytoskeleton analysis in that it is not limited to differentiating cytoskeletal arrangements of cells and focuses on human dermal fibroblasts. The researchers claim to have developed a fully automated method to quantify characteristics of the dynamic cytoskeleton from the live imaging data that is proven sensitive enough for accurate analysis, yet robust enough to accommodate noise.

The article then gives a detailed account of the methods. To summarize, human dermal fibroblasts were cultured, transfected, then seeded onto multiwell plates with PA gels of varying stiffness and human Collagen Type I or Fibronectin. Time lapse live-imaging dataset acquistion was competed using an Axiovert 200 inverted fluorescence microscope with a motorized translation stage and an incubator to keep the samples at physiological conditions. A beam splitter was used to simultaneously acquire separate images of the nucleus and the cytoskeleton and imaging was done for samples treated with Cytochalasin D (CytoD) to induce cytoskeletal disruption and quantify its affects. Finally, the author explains the automated data processing methods used in MATLAB, as well as how the images were split into interrogation windows for analysis and how Df and GVI (Grey Value Intensity) was measured in each interrogation window and as a whole cell image.

The author ends the article with an in-depth review of the results of the study, including Df values with varying interrogation window size, as well as fractal analysis of the same images as a whole. Ultimately an interrogation window size of 10 pixels proved ideal for analysis. Secondly, it was determined that Df does not become significantly inaccurate until an addition of 40%+ noise to the image or a resolution poorer than 1.93 micrometers/pixels. Third, a significant temporal decline of Df values was seen for CytoD treated cells. Finally, Df evaluation was used to show how substrate type and substrate stiffness affect cytoskeletal response (i.e. the plates that were glass vs. PA at varying pressure “stiffness”). Whole cell Df evaluation showed significant and higher differences between glass and PA substrates of high pressure (50 kPa) than Pa with lower pressure (25 kPa), and the amount of actin in the cells increased with increasing stiffness/pressure.

Overall, the researchers were pleased with their algorithm’s ability to pick up significantly different Df values, control for noise, and use it to analyze the application of the cytoskeletal drug, Cytochalasin D. Fractal analysis has proven to be a valid method to study the sensitive temporal changes in the structure of the whole-cell actin cytoskeleton network through the measurements of Df and GVI. In future studies, the authors hope to explore the biological implications (i.e. the meanings behind) the Gf and GVI values obtained, now that they have proven to be significantly relevant to detecting change in actin cytoskeletons for cells culture on substrates of differing stiffness. Eventually, the authors believe this method could become the premier method for high throughput screening of kinematic cytoskeletal rearrangement.

One thought on “Article Review (4): Image-based characterization of the cytoskeleton

  1. Post Discussion Summary:

    We began the discussion with a summary of the paper, particularly what the cytoskeleton is, some of its functions, and the importance of imaging its changes in real time. I gave a brief overview of the methods of the mathematical automatic image processing done, particularly what the certain values that were calculated are, such as Df and GVI.

    We discussed why it might be important to develop an automatic image processing system and the reasons are two-fold. First, if the entire image analysis is automated and leads to a quantifiable value for the image, there is no need for analysis by personnel, which could likely lead to subjective qualitative analysis rather than objective quantification. Secondly, if the entire process is automatic, this will significantly reduce time spend analyzing image, whereas analysis by personal can take a long time.

    We identified some strengths and weaknesses in the paper. Particularly, the strengths determined were, as described above, the importance of developing an automated system, as well as the biological relevance real-time analysis of the cytoskeleton could have. The clearest weakness with this paper is certainly the fact that as of now there is no direct correlation between the Df (quantified value) and what it means biological in terms of cytoskeleton changes/analysis. Until this value is proven biologically relevant in helping determine something beneficial/critical about the cytoskeleton, this image-processing system may render unnecessary. This lead to the future implications of the study, particularly determining if there is a biological meaning to Df values and if so, what that is. If this can be determined, this method could be used for other aspects of the cell and become a premier technique for high throughput screening of kinematic cytoskeletal rearrangement.

    Main questions from the class concerned the effect of noise on Df values, and Dr. Haidekker gave a brief explanation of the way the researchers account for noise in their mathematical operations.

Leave a Reply

Your email address will not be published. Required fields are marked *