Quantitative Fitness Analysis
During Quantitative Fitness Analysis (QFA) we generate cell density timecourse estimates by repeatedly photographing microbial cultures growing in regular arrays on solid agar plates. First, we quantify cell density from timecourse photographs using the image analysis tool Colonyzer (Lawless et al., 2010). Next, we fit a generalised logistic model to the cell density timecourse for each independently growing culture, using the inferred model parameters to define a quantitative measure of culture fitness. Comparisons between fitnesses can be extremely useful for comparing the health of cultures and deducing the relative effect of their genotypes in a specific environment. The computational components of QFA: culture tracking, growth curve construction, logistic model fitting, genetic interaction strength estimation, quality control and data visualisation are carried out using the qfa R package.
Spotted, 384-format cultures A cropped image of 15 out of 384 spotted cultures whose growth curves can be captured by repeated photography and image analysis during QFA.
QFA is a very useful, quantitiative alternative to spot-testing of cultures which have been manually inoculated onto solid agar. Since the experimental steps of QFA are based on traditional lab techniques, such as replica plating and spot testing, and since digital image capture by photography is inexpensive, manual QFA can be carried out quite cheaply. On the other hand, the equipment for genome-wide QFA is expensive since QFA at this scale requires significant robotic assistance, however the high-throughput screening facility at Newcastle University offers QFA as a service.
JoVE QFA video
Below is a link to a short video describing the experimental and computational components of QFA, from the Journal of Visualised Experiments. Associated with this video there is also an open-access manuscript with a detailed description of the QFA protocol.
One particularly straightforward and well documented aspect of the QFA R package is a set of functions for real-time, interactive visualisation of QFA data. Simple installation instructions for users interested in using the data visualisation tools in the QFA R package can be found here. The R package contains example datasets taken from Addinall et al., 2011. Alternatively, a web-based data visualisation tool can be found here.
By carrying out QFA on arrayed microbial cultures containing relevant gene mutations, we can use fitness estimates to infer the presence or absence of epistasis, and to quantify and rank genetic interaction strengths. For example, during screens for gene deletions interacting with telomere capping mutations, Addinall et al., 2011 inferred genetic interaction strength as the deviation of observed double mutant fitnesses from predicted double mutant fitnesses given observed single deletion fitnesses and assuming a multiplicative model of epistasis. They also calculate the statistical significance of deviations given heterogeneity in experimental observations.
Two growth curves Timelapse images (top) and growth curves quantified from images (bottom) for a wild-type strain (his3Δ) and a relatively sick strain (htz1Δ) cultured during QFA.
If arrayed cultures are grown at selected temperatures we can infer the temperature sensitivity of mutants by QFA. Addinall et al., 2011 demonstrated the use of QFA to identify gene deletions resulting in increased temperature sensitivity. These results were verified by comparison with manually curated observations listed in the Saccharomyces Genome Database (SGD). If cultures are subject to particular treatments (e.g. nutrition, or drug treatments) QFA can generate ranked lists of cultures which grow surprisingly well or badly given the particular treatment.
Example output from a genome-wide QFA screen 308 independent growth curves, together with fitness estimates from one out of 480 plates examined during a control screen of the yeast deletion collection as described in Addinall et al., 2011.
The supplementary data website for Addinall et al., 2011 contains large, genome-wide datasets suitable for analysis with this package.
cdc13-1 Fitness Plot Scatterplot comparing the output of two QFA experiments: fitnesses of strains from the yeast deletion collection (x-axis) and fitnesses of the same strains combined with a second, background mutation (cdc13-1). Dashed line is the line of equal fitness. Blue line represents expected double mutant fitness, given single mutant fitness assuming the multiplicative model of genetic independence. Points highlighted in red and green are statistically significant suppressors and enhancers of the cdc13-1 phenotype respectively.
QFA consists of four main stages: culture inoculation, timecourse photography, image analysis and fitness estimation.
The first stage of QFA requires the generation of arrays of cultures of interest. This can be carried out in a high-throughput fashion, using libraries of gene mutations (as was done in Addinall et al., 2011), however it can also be carried out in a lower-throughput fashion, with hand selected and hand inoculated cultures, growing on a single plate for example. Cultures should be diluted appropriately and inoculated onto solid agar.
There are custom, semi-automatic and fully-automatic technologies available to facilitate timecourse photography of such plates (e.g. S&P Robotics SPImager, or a custom-built alternative), however any lab should be able to capture such timecourses by fixing a consumer digital camera over a plate, and repeatedly capturing images to a computer hard-drive via USB cable. It is important to record the time at which the photographs were captured, and this is achieved most simply by embedding the current date and time into the captured image filenames.
Image analysis converts culture photographs into cell density estimates. This is usually achieved using the the Colonyzer software (Lawless et al., 2010). However the output of any image analysis tool capable of generating culture size estimates from images of rectangular arrayed cultures could be reformatted and incorporated into the QFA workflow. Colonyzer is a suite of particularly sensitive image analysis algorithms, capable of detecting signal on spotted agar plates early in the growth curve. It has been designed to eliminate lighting gradients in captured images, thereby minimising spatial bias in cell density estimates without having to use expensive, carefully optimised lighting systems.
Given cell density estimates from Colonyzer and an appropriate experimental description specifiying the genotype or ID of each culture on each plate, the treatments applied, and the time of inoculation for each plate, the qfa R package contains everything needed to generate growth curves and quantitiative fitness estimates for all cultures, together with estimates of genetic interaction strengths (for appropriately designed experiments).