Pyphe, a python toolbox for assessing microbial growth and cell viability in high-throughput colony screens

  1. Stephan Kamrad
  2. María Rodríguez-López
  3. Cristina Cotobal
  4. Clara Correia-Melo
  5. Markus Ralser  Is a corresponding author
  6. Jürg Bähler  Is a corresponding author
  1. University College London, Institute of Healthy Ageing, Department of Genetics, Evolution and Environment, United Kingdom
  2. The Francis Crick Institute, Molecular Biology of Metabolism Laboratory, United Kingdom
  3. Charité Universitaetsmedizin Berlin, Department of Biochemistry, Germany
11 figures, 1 table and 5 additional files

Figures

Figure 1 with 2 supplements
Data processing workflows using pyphe.

Pyphe is flexible and can use several fitness proxies as input. In a typical endpoint experiment, plate images are acquired using transmission scanning and colony sizes are extracted using pyphe-quan…

Figure 1—figure supplement 1
Image analysis with pyphe-quantify, described in Appendix 1.

All 144 images from the image time course of 57 wild strains in replicates in 1536 format were analysed with pyphe-quantify batch and with gitter using remove.noise=TRUE and inverse=TRUE settings. (A

Figure 1—figure supplement 2
Spatial normalisation with pyphe-analyse, described in Appendix 2.

(A) Placement of two 96 grids in opposite corners of 1536 plate maximises grid coverage and only uses 1 in 8 positions for normalisation purposes. (B) Left: Mean uncorrected colony sizes across …

Figure 2 with 3 supplements
Normalisation strategies for growth curves and endpoints.

(A) Growth curves of 57 wild S. pombe strains (average of approximately 20 replicates each) before (top) and after (bottom) correction. Corrected colony sizes describe the fitness relative to the …

Figure 2—figure supplement 1
57 wild strains on different carbon and nitrogen sources.

Left: Heatmaps showing raw and corrected colony sizes of maximum slopes and endpoints. Centre: Correlation of timepoints with each other (large heatmaps), correlation of each timepoint with maximum …

Figure 2—figure supplement 2
57 wild strains on different carbon and nitrogen sources (cont’d).

Left: Heatmaps showing raw and corrected colony sizes of maximum slopes and endpoints. Centre: Correlation of timepoints with each other (large heatmaps), correlation of each timepoint with maximum …

Figure 2—figure supplement 3
57 wild strains analysis summary.

(A) Distribution of CVs and FUVs (based on 96 replicates of the control strain 972 evenly dispersed through the plate, normalised for spatial effects with grid correction) across 8 conditions. Using …

Figure 3 with 1 supplement
Phloxine B provides an orthogonal and independent fitness proxy.

(A) Relative colony sizes and redness scores after correction for 238 single gene knock-outs in 70 conditions (after quality filtering as described in Methods, three biological replicate colonies …

Figure 3—figure supplement 1
Normalisation of redness data for 238 knock-out mutants.

(A) Median redness scores across 308 plates (covering 78 different conditions). Top left: Uncorrected redness scores as obtained with pyphe-quanitfy in redness mode. There is a strong row-dependent …

Figure 4 with 2 supplements
Phloxine B staining reflects percentage of dead cells.

(A) Example of colony redness score extraction by pyphe-quantify in redness mode. From the acquired input image (i), colors are enhanced and the background subtracted (ii), colonies are identified …

Figure 4—figure supplement 1
Distribution of phloxine B intensities in ImageStream.

Plot titles indicate the gene which was knocked out. The standard laboratory strain 972 was measured in 6 biological replicates from samples obtained from different colonies.

Figure 4—figure supplement 2
Fraction of strongly stained cells depending on colony redness score.

The correlation is weaker than for the fraction of live cells only (neither burst nor strongly stained in the flow cytometer, shown in Figure 4D). This suggests that burst cells do contribute to …

Temporal dynamics of phloxine B colony redness scores.

(A) Raw redness scores over time for 96 wild-type grid colonies (dark line shows mean, shaded area shows standard deviation). The uncorrected redness increases as colonies grow as there is a …

Author response image 1
Pixel calibration is not required for accurate determination of colony sizes.

Top row: calibration functions applied to the original scanned image. The first function is a linear transformation that scales the image to fill the entire 8bit range. We apply this to images in …

Author response image 2
Examples of raw growth curves obtained with pyphe setup.

Shown are 12 growth curves from the first row of a 1536 plate of 57 S. pombe wild strains (same data as Figure 2 in manuscript) analysed with pyphe-quantify in timecourse mode.

Author response image 3
Image conversion to jpg has negligible impact on results.

Each image of a growth curve consisting of 145 images (shown on x-axis) was analysed in the original tiff format and in the converted jpg, using gitter (right) and pyphe-quantify in batch mode …

Author response image 4
The analysis shows the number of replicates required with scan-o-matics and with pyphe in order to achieve the same statistical power.
Author response image 5
Subgroup analysis of colony staining.

We divided the data into two groups and computed the correlation separately. Both groups still show clear correlation (0.41 and -0.33) which is incompatible with the claim that the method allows a …

Author response image 6
The fraction of live cells (neither burst nor strongly stained in flow cytometer, left panel) better explains the colony redness score than the fraction of strongly stained cells only (right panel).

This suggests that burst cells do contribute to staining in the colony (while being unstained in the flow cytometer). Note that the correlation breaks down for colonies with higher redness scores …

Tables

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Strain, strain background (Schizosaccharomyces pombe)57 S. pombe wild strainsJeffares et al., 2015JBxxxThese strains were identified as a set of most diverse strains from the overall
collection
Strain, strain background (Schizosaccharomyces pombe)238 S. pombe knock-out strainsBioneer and (Sideri et al., 2015)Pombase gene IDs and namesThe original library obtained from Bioneer was made prototrophic by crossing
with suitable strain. Genes were selected to cover GO functional categories and include unknowns.
Chemical compound, drugPhloxine BSigmaCat# P2759Prepared as a 5 g/L (1000x) stock in water and stored at 4°C in the dark.
Software, algorithmPypheThis publicationPyphe provides the following tools: pyphe-scan, pyphe-scan-timecourse, pyphe-quantify, pyphe-analyse, pyphe-interpret, pyphe-growthcurvesVersion 0.95 was used for preparation of this manuscript.
OtherScannerEpsonV800 Photo

Additional files

Supplementary file 1

Corrected maximum slopes and endpoints for 57 wild strains in 8 conditions.

https://cdn.elifesciences.org/articles/55160/elife-55160-supp1-v1.csv
Supplementary file 2

Relative redness scores and colony sizes for 238 knock-out mutants.

https://cdn.elifesciences.org/articles/55160/elife-55160-supp2-v1.csv
Supplementary file 3

Differential fitness of 238 knock-out mutants in conditions with and without phloxine B.

https://cdn.elifesciences.org/articles/55160/elife-55160-supp3-v1.csv
Supplementary file 4

ImageStream classification counts for mutants.

https://cdn.elifesciences.org/articles/55160/elife-55160-supp4-v1.csv
Transparent reporting form
https://cdn.elifesciences.org/articles/55160/elife-55160-transrepform-v1.docx

Download links