Over the last 15 years, florescence microscopy has given us a window into the bacterial cell, providing a wealth of insights into the previously unobservable higher order organization with these small organisms. We have moved from the simple imaging of cells to the observations of single molecule dynamics (Deich et al., 2004; Kim et al., 2006), and now the development of high throughput microscopy is allowing us to conduct genome-wide screens using protein localization as our readout (Werner et al., 2009; Christen et al., 2010; Goley et al., 2010). As imaging technology further evolves, generating data both higher in detail and larger in quantity, we begin to, more than ever, confront the issue of data analysis. The light microscope is a powerful tool, and with careful calibration and choices of acquisition settings, fluorescence microscopy can provide a quantitative spatial map of proteins distribution [for technical discussions, see (Waters, 2009)]. However, unlike biochemical or genetic assays, which are easily quantified via single metrics, light microscopy presents the dilemma of yielding data that is rich in information, while posing significant challenges to interpret quantitatively. At the simplest level, we use microscopy to paint a descriptive picture of the bacterium. This allows us to state that a given component appears to be ‘polar’, ‘at midcell’, or ‘helical’. While these descriptions are useful, these observations can be made even more powerful through the use of quantitative metrics that place these localizations into the context of cell shape and relative signal intensity. A protein defined to be ‘at midcell’ is a much different definition than ‘in over 2000 cells, 74% of the protein is localized at 40% 5% of the cell length from the old cell pole’. In order to gain more exact measures, we need the development of tools that extract the quantitative meaning out of our descriptive pictures without the subjective interference of human interpretation. Most critically, for our data to have spatial meaning, it must be placed into the context of cell shape. This requires algorithms that can identify and separate out cells within dense fields of bacteria, and fit their shapes with high precision. Once this shape is determined, we require methods to analyse our intensity data in respect to this co-ordinate system. If we are examining with diffraction-limited point sources of signal, such as chromosomally bound GFP-LacI arrays, we can determine their localization with nanometre accuracy using Guassian fitting (Thompson et al., 2002; Yildiz et al., 2003), and then follow their dynamics with high spatiotemporal resolution (Kim et al., 2006; Niu and Yu, 2008; Shebelut et al., 2010). The above approaches – the segmenting of cells, subpixel assignments and spatial quantification of fluorescent signal – are common approaches in image analysis. However, in most lab settings, these techniques are difficult to implement, as they exist as scattered fragments of code written in various programming languages, and are not available in one software package. Because of this, integrating these approaches into one’s own research flow can be difficult, and students often weigh the time difference between spending months learning the programming skills required to conduct these analyses versus simply carrying out brute force manual quantifications of their data. In this issue, the Emonet and Jacobs-Wagner labs describe a new image analysis tool, MicrobeTracker, built specifically for the automated analysis of bacterial image data. This package implements all of the above computational approaches, allowing for a range of analyses to be performed. This software is freely available to the biological community, and uses an easily accessible graphical interface built on top of MATLAB, allowing for further detailed analysis. With modifications to parameter files this software can be adapted to work with a variety of organisms, even to those outside the bacterial kingdom. Perhaps the most impressive feature of MicrobeTracker is its approach to the difficult problem of cell identification Accepted 1 February, 2010. *For correspondence. E-mail ethan_garner@hms.harvard.edu; Tel. (+1) 617 432 3724; Fax (+1) 617 432 5012. Molecular Microbiology (2011) 80(3), 577–579 doi:10.1111/j.1365-2958.2011.07580.x First published online 24 March 2011
Read full abstract
Jan 1, 2013
F Emiliani + 10
Open Access
