Figure 1.
Pipeline of the methods used in this work.
The identification of cells, assignment of cell type, cell stage and the estimation of cell confidence is based solely on the intensities of the RFP marker present in all strains. Please refer to the Results and Methods for descriptions of steps (i)–(vi). The cell type, stage and confidence are then used in conjunction with the GFP signal from tagged proteins in each strain in order to compute biologically interpretable features of protein expression.
Figure 2.
a) Shows the mother-bud assignment heuristic. Pairs of circular objects that reciprocally have largest and smallest sizes among neighboring areas are said to be ‘mother’ cells (indicated by M) and ‘bud’ cells (indicated by B, mother-bud pairs indicated by bidirectional arrows), unless the potential ‘bud’ cell has a smaller neighbor than itself (indicated by a unidirectional arrow). Any other cells are labelled as ‘lone’ cells (L). b) Example of low and high confidence objects. The cyan lines in each image represent the cell contours produced, and the white dots indicate the predicted bud neck position. The dashed objects represent obvious artifacts that were filtered using thresholds (See text for details). Objects on the edge of images were not automatically filtered out, but are expected to have low confidence.
Figure 3.
a) Heatmap of the mean morphological distance features for each of the 3 cell classes automatically labelled: ‘bud’, ‘mother’ and ‘lone’ (columns indicated by ‘B’, ‘M’ and ‘L’ respectively). The proteins at the two extremes are enriched in cell periphery and nucleolus proteins. b) Three examples of the morphological distances extracted from the heatmap. Although the heatmap only shows the mean, we also compute the standard deviation (error bars). c) Examples of cells from the strains indicated in b). The spread of GFP fluorescence is greater than the RFP for the first three proteins, and less than RFP for the last three.
Figure 4.
a) GFP intensity heatmap for several protein whose abundances are known to be cell-cycle dependent. b) Profiles for 3 proteins showing significantly higher expression level in large buds. ‘n’ is the number of mother-bud pairs used to infer each time series. 26 out of the 60 time points (indicated with markers) show coherent cell-stage specific deviation (permutation test, See Suppl. Figure S4). c) Examples of mother-bud pairs with the computed pixel size (pt) of the bud object (identical RFP/GFP intensity scale). The displayed cells were manually selected and then ordered by the computed bud size. Arrows indicate nuclear localization at lower intensity.
Figure 5.
Time profiles of morphological distances.
a) Top panel shows a heatmap of the morphological distances in bud and mother cells indicated as B and M, respectively. Bottom panel shows the data for two of these proteins as line graphs. The reported morphological distances are variance normalized. MCM complex subunits and Whi5 display a cell-cycle dependent subcellular location; cytoplasmic for small buds, nuclear for large buds. ‘n’ is the number of mother-bud pairs used to infer each time series. Out of the 80 timepoints for each protein, 34 for Whi5 (blue traces), and 72 for Mcm6 (red traces) show significant cell cycle variation (, indicated as dark dots). b) Examples of mother-bud pairs that were ordered by the computed bud size (pt). The GFP channel was scaled between images to more clearly illustrate the change in subcellular location.
Figure 6.
Time profile clustering result.
A heatmap with 4004 GFP-tagged strains ordered using maximum likelihood agglomerative clustering based on the time profiles of protein abundance and 5 morphological measures. Within manually selected clusters (colored bars), the fraction of proteins in the cluster that have the same subcellular localization or GO Annotation (the latter indicated with stars) is listed under Fraction. Log p-values were computed using the hypergeometric distribution to test against the null hypothesis that the cluster was drawn randomly from the protein annotations. Fold enrichment indicates the ratio of the Fraction of proteins in the cluster with each annotation compared to that in the protein collection. Nuclear proteins appear in the bud at a specific time (dashed line).
Figure 7.
Subcellular location class profiles.
a) Time series for protein abundance in buds. Nuclear proteins are the last to appear in the bud (dashed line). b) The spatial distribution of protein expression is highly variable in the growing bud cell. Organelles appear to be pushed from the bud neck at the time of the nucleus inclusion (dashed line). Note that the absence of nuclear protein in the bud leads to irrelevant variations in the morphological distance features, perhaps due to auto-fluorescence captured in the GFP channel. Actin proteins migrate from bud tip to bud neck (black traces). c) In the mother cell, organelles appear to maintain a typical distance to the bud neck, expect for the nucleus.
Figure 8.
A cluster of 91 proteins displaying time profiles with variable distances to the bud neck.
a) Heat map of the cluster displayed as in Figure 6. We observe several classes of dynamic patterns, which capture the localization to the bud neck and bud periphery. (*) 5 of the 8 subunits of the exocyst complex are found within 9 proteins. b) Examples of proteins with dynamic bud patterns. (**) The displayed GFP intensity was scaled down by 75%.