Object-based Ratio Analysis

Transgenic reporters, such as roGFP, can be easily targeted to organelles, such as chloroplasts and mitochondria, facilitating sophisticated dissection of sub-cellular behaviour. However, these organelles are present in considerable numbers (hundreds) in most plant and fungal cells, and are highly dynamic, making it impractical to manually select individual objects, and unhelpful to consider the total pixel-population statistics described above, as these disguise individual organelle behaviour. As an alternative, it is possible to automatically segment each organelle separately, and then calculate the average ratio values on an object-by-object basis. A wide variety of segmentation algorithms can be used, but we have found a generic two-step procedure is routinely applicable. Objects are initially separated from each other into non-overlapping domains using watershed segmentation of an inverted template image calculated from the original intensity channels. Each object definition is refined using a local intensity threshold within each watershed domain, to accommodate objects with differing overall intensity. Once the objects are segmented, various morphological parameters can be measured automatically, along with the average intensity at each wavelength needed to calculate the ratio values. If the objects do not move during the time-series, object responses can be visualised as an ‘object’-time plot analogous to a kymograph. In parallel, the (non-confocal) bright-field image can be processed separately to highlight morphological features of interest, using a combination of noise filtering, projection, contrast enhancement, and manual delineation. Key landmarks can then be established to correlate physiological response with cell or organelle identity, morphology, or localised event.