Our rationale here is that we can get insights into the biological role of YgjD by following the cellular response of its depletion on the single cell level and with high temporal resolution. We diluted cultures of the conditional lethal P
ara
-ygjD mutant TB80 onto pads of solid LB medium that either contained L-arabinose (inducing ygjD expression) or D-glucose (repressing ygjD expression) and used time-lapse microscopy to follow single cells growing into microcolonies, taking an image every 2 or 4 minutes. The images were analyzed with the software "Schnitzcell" [18]. The growth rate and cellular morphology of the P
ara
-ygjD strain grown in the presence of L-arabinose was similar to the wild type grown under the same conditions (Figure 1a and 1c, and Additional file 1 - movie 1 and Additional file 2 - movie 2).
Additional File 1: Movie 1. TB80 (ppGpp+) growing on LB agar with 0.1% L-arabinose. 100 frames (one frame per two minutes) were compressed into 10 seconds. The scale bar is 5 μm in size (same in all movies hereafter). (MOV 596 KB)
Additional File 2: Movie 2: MG1655 growing on LB agar with 0.4% glucose. 100 frames (one frame per two minutes) were compressed into 10 seconds. (MOV 1 MB)
A shift of the P
ara
-ygjD strain to glucose lead to the depletion of YgjD. This depletion is based on two effects. First, transcription of ygjD stops after the shift to glucose. Residual L-arabinose that remains in the cells from growth under permissive conditions is rapidly metabolized. Lack of L-arabinose turns the transcriptional activator (AraC) of the Para promoter into a transcription repressor. In addition, glucose metabolism causes depletion of the cellular co-inducer cyclic AMP. Together these effects lead to effective repression of ygjD transcription in TB80. After termination of de novo ygjD mRNA synthesis the amount of YgjD in each cell declines, because the mRNA and the protein are diluted through cell division, and degraded by cellular nucleases and proteases, respectively [20]. The rapid cessation of transcription from Para after the shift to glucose was evident in control experiments with a strain that expressed the green fluorescent protein (GFP) from the arabinose promoter (Additional file 3 - Figure S1)
YgjD depletion leads to a change in cell size homeostasis
Time-lapse microscopy revealed that YgjD depletion lead to a gradual phenotypic transition in division and growth. Single cells that were transferred from permissive conditions to solid pads of LB medium with glucose first continued to divide regularly, forming microcolonies in which the number of cells initially increased exponentially. Then, after about four divisions, cell division slowed down and stopped (Figure 1b and Additional file 4 - movie 3). Analysis of the time-lapse images (Additional file 4 - movie 3) showed that, during this transition, cells size decreased (Figure 2a). This indicates a disturbance in of cell size homeostasis [19] - that cells divide before their cell size doubled.
Additional File 4: Movie 3. TB80 (ppGpp+) growing on LB agar with 0.4% glucose. 150 frames (one frame per two minutes) were compressed into 15 seconds. This movie was used to extract the growth dynamics shown in Figure 2 and 3. (MOV 3 MB)
We used elongation rates of single cells and the time interval between two divisions to analyze the change in cell size homeostasis during YgjD depletion. Since we were interested in how these parameters changed during depletion, we separated data from different cell generations of the depletion process. The first cell that is founding a microcolony is generation 0; this cell divides into two cells of generation 1, which divide into four cells of generation 2, and so on (also see Additional File 5 - Figure S2). To avoid comparisons between cells that are in different phases of their cell cycle, we only used cell size measurements (and later fluorescence intensities) of cells immediately before division. Also, to avoid incomplete and biased sampling, we removed data from above generation 6.
This analysis revealed that the small size of cells depleted for YgjD was a consequence of two effects: first, the rate of elongation (cell length increase over time) decreased (Figure 3a). Second, cells did not respond to the decrease in elongation rate by adjusting the frequency at which they divided; the interval between two cell divisions remained initially constant. As a direct consequence, cell length at division decreased continuously (Figure 2a).
The phenotype induced by YgjD depletion was specific, and depletions of other essential genes lead to different cellular morphologies. We analyzed time-lapse images of the depletion of three other essential genes (dnaT, fldA and ffh). Depletion of each protein resulted in cellular phenotypes that were different from each other and from YgjD when depleted (Additional file 6 - Figure S3; also see Additional Files 7, 8 and 9 - movies 4, 5 and 6). Also, the effects of YgjD depletion were different from the consequences of exposure to two antibiotics that we tested: we followed wildtype E. coli cells exposed to the translational inhibitors kanamycin and chloramphenicol at minimum inhibitory concentration (2.5 μg/ml for chloramphenicol, 5 μg/ml for kanamycin), and observed no decrease in cell size (Additional file 10 - Figure S4, and Additional Files 11 and 12 - movies 7 and 8).
Additional File 7: movie 4: Depletion of FldA from growing cells. A Para-fldA conditional lethal mutant was shifted from 0.1% arabinose to an agar pad with 0.4% glucose. FldA is essential for isoprenoid biosynthesis [44], and as the movie shows, depletion of FldA leads to lysis of cells. 80 frames (one frame per four minutes) were compressed into 8 seconds. (MOV 199 KB)
Additional File 8: movie 5: Depletion of Ffh from growing cells. A Para-ffh conditional lethal mutant was shifted from 0.1% arabinose to an agar pad with 0.4% glucose. Ffh protein is part of the signal recognition particle translocation system, that cotranslationaly sequesters proteins into or across the cytoplasmic membrane [45]. Depletion resulted in visible intracellular aggregates, followed by elongation and cell lysis. 120 frames (one frame per two minutes) were compressed into 12 seconds. (MOV 5 MB)
Additional File 9: movie 6: Depletion of DnaT from growing cells. A Para-dnaT conditional lethal mutant was shifted from 0.01% arabinose to a 0.4% glucose containing agar pad. Depletion resulted in filament formation, which is in agreement with "unbalanced" growth upon abrogation of DNA replication. dnaT (and the following gene dnaC) is part of the "primosome" and is crucial for initiation of DNA replication. 100 frames (one frame per four minutes) were compressed into 10 seconds. (MOV 1 MB)
Additional File 11: movie 7: Growth of E. coli MG1655 on 2.5 μg/ml chloramphenicol. E. coli MG1655 was precultured in LB medium and transferred to an agar pad containing 2.5 μg/ml chloramphenicol. 100 frames (one frame per four minutes) were compressed into 10 seconds,. (MOV 629 KB)
Additional File 12: movie 8: Growth of E. coli MG1655 on 5 μg/ml kanamycin. E. coli MG1655 was precultured in LB medium and transferred to an agar pad containing 5 μg/ml kanamycin. 60 frames (one frame per four minutes) were compressed into 6 seconds. (MOV 609 KB)
For reference, we also analyzed images of growing microcolonies of wildtype E. coli MG1655 cells on LB medium supplemented with glucose. This experiments confirmed cell size homeostasis, as expected from normally growing cell: cells divided close to the moment when they had doubled their size, and small fluctuations in cell elongation rates were compensated for by adjustments in the time of cell division (Figure 2b and 3b).
The transition towards smaller cell size is controlled
What kind of disturbance of cell size homeostasis is induced by depletion of YgjD? We considered two possibilities. First, it is possible that the control that couples cell division to cell size is lost, so that cells divide in an uncontrolled way, irrespective of their size. Second, it is conceivable that cell division remains coupled to cell size, but the target size that a cell needs to reach before initiating division decreases over time.
If the decrease in cell size is the result of a controlled transition towards smaller cells, one would expect that, during the transition, the cell elongation rate and the timing of cell division would still be linked, but that this link would change quantitatively over time. In fact this is what we observed when we analyzed each generation of cells during the depletion process separately (inserts Figure 3a and 3b). Within a given generation the time interval between divisions and the rate by which a cell elongated was negatively correlated: cells that grew faster than the average of their generation tended to initiate division more quickly; cells that grew more slowly initiated division later. This suggests that cell growth and the timing of cell division are still linked within each generation in the depletion process, but that this link changes quantitatively over successive generations.
This analysis has, however, an important limitation: cells within a given generation are not independent from each other. Some of these cells are more closely related, because they derive from the same mother or grandmother. This can lead to spurious correlations between traits; in our case, this effect could lead to artificial correlations between cell elongation rates and interdivision intervals. This problem of relatedness in lineage trees is known from phylogenetic studies, where it is referred to as phylogenetic dependence [21]. In the context of phylogenetic studies, these dependencies can be resolved by analyzing differences between independent pairs of species, rather than calculating correlations on the basis of the whole phylogenetic lineage [21].
We used a variation of this approach to get an unbiased view on the relationship between cell growth and the timing of cell division: for each generation, we analyzed pairs of cells emerging from the same cell division, and calculated the difference in growth rates and in the time to division for each pair. We refer to two cells emerging from the same division as 'sisters' (thereby ignoring that these two cells have cell poles of different ages, [22, 23]). The differences for all sister pairs represent independent data points, and we can use them to calculate the correlation between cell growth and time to division in an unbiased way.
The independent contrast analysis confirmed our earlier conclusions: comparing each cell to its sister cell, we found that cells that grew slower than their sisters also displayed a longer interval between cell divisions, while faster growing cells divided earlier. This manifests as a negative correlation between the difference in cell elongation rate and the difference in interdivision intervals between two sisters (inserts Figure 3c and 3d; see also Additional File 13 - Figure S5). This is consistent with the interpretation that, during YgjD depletion, the timing of cell division remained coupled to a given cell size - and that the target cell size declined.
The transition to decreased cell size is reminiscent of morphological changes that occur during the 'stringent response' [24, 25], a stress adaptation program that is elicited when cells encounter amino-acid or carbon-starvation [26]. The stringent response is induced by accumulation of the 'alarmone' guanosine tetra/penta phosphate ((p)ppGpp), e.g. in response to low concentrations of amino-acylated tRNAs [26]. We thus wanted to investigate this possible link to (p)ppGpp signaling more closely, and asked whether the changes in cell homeostastis upon YgjD depletion are mediated through (p)ppGpp.
Changes in cell size homeostastis are mediated through ppGpp
We constructed a strain, TB84, that is deficient in (p)ppGpp synthesis ((p)pGpp0), due to deletions of relA and spoT [26, 27], and in which expression of ygjD was again under control of Para. We followed growing microcolonies of TB84 as described above and found that the consequences of YgjD depletion were profoundly different: cell elongation rate decreased during the YgjD depletion process as for the relA+ spoT+ strain TB80 (Figure 4a). In contrast to what we observed with this (p)ppGpp+ strain, the decrease in elongation rate was compensated for by an increase in the time interval between two divisions (Additional file 14 - movie 9, and Figure 4a). As a consequence, cell size at division was not reduced, and the final cell length of depleted (p)ppGpp0 cells (TB84) was on average twice that of depleted (p)ppGpp+ cells (TB80) (Figure 4b). This is reminiscent of the elongated cells found in populations of cells depleted for YgjD by Handford and colleagues [3].
Additional File 14: Movie 9. TB84 (ppGpp0) growing on LB agar with 0.4% glucose. 200 frames (one frame per two minutes) were compressed into 20 seconds. (MOV 3 MB)
This suggests that the changes in cell size in response to YgjD depletion are mediated through the alarmone (p)ppGpp; an alternative explanation is that the absence of (p)ppGpp leads to cell elongation (as has been previously reported [27]), and that this elongation compensates indirectly for reductive fission upon YgjD depletion. Importantly, TB84 cells still ceased cell division (Additional file 15 - Figure S6). Thus, ygjD is still essential even in the absence of (p)ppGpp, and termination of cell division is not solely a consequence of a diminished cellular growth rate.
To further test the idea that ygjD depletion triggers (p)ppGpp synthesis we measured, on a single cell level during YgjD depletion, the activity of two promoters known to respond to the intracellular level of (p)ppGpp: Papt is repressed by (p)ppGpp, while Prsd is induced by (p)ppGpp [28]. We transformed TB84 with plasmids carrying transcriptional promoter-gfp fusions [29] encoding Papt-gfp and Prsd-gfp, and measured gene expression from these promoters as fluorescence intensity over consecutive cell divisions. The level of GFP expression steadily decreased in the strains where gfp was controlled by Papt (Figure 5a), and steadily increased when controlled by Prsd (Figure 5c). Furthermore, this change in fluorescence was tightly linked to the rate by which cells elongated (Figure 5b and 5d). When the same strains were grown on L-arabinose containing medium no consistent changes of fluorescence could be observed (Additional file 16 - Figure S7). These observations are consistent with the scenario that YgjD depletion induces (p)ppGpp synthesis, and thus influences promoters whose expression depends on the levels of (p)ppGpp.
Single cell analysis indicated that, in the cells depleted for YgjD, there is a link between decreased cell elongation rate and (p)ppGpp levels. Using independent comparisons between sister cells in the microcolonies undergoing YjgD depletion, we found that if a cell had a lower elongation rate than its sister, it also tended to have lower levels of GFP expressed from Papt (details not shown; for Prsd-gfp, this pattern was not observed). These data support the idea that the link between (p)ppGpp levels and the cell elongation rate is direct; for example, it is possible that high levels of (p)ppGpp cause low elongation rates [30].
Our results suggest further that YgjD depletion has two (possibly linked) effects: first, depletion triggers (p)ppGpp synthesis. Second, it leads to termination of cell division. To gain insights in which phase of the cell cycle YgjD-depleted cells are arrested we visualized the DNA-content of individual cells with DNA-staining and subsequent fluorescence microscopy (Additional File 17 - Figure S8). After YgjD depletion in (p)ppGpp+ cells (TB80), DNA was localized at midcell and filled large areas of the cell (Additional File 17 - Figure S8 b), possibly indicating that cells were unable to carry out additional cell divisions due to "nucleoid occlusion" [31]. This mechanism prevents premature cell division before chromosomes have been distributed to opposite cell halves. However, termination of cell division also manifests in a (p)ppGpp0 strain (Additional File 17 - Figure S8 c): depleted cells were elongated, and only a small fraction of the cell volume was filled with DNA. Thus, in the (p)ppGpp0 background, nucleoid occlusion alone cannot be responsible for termination of cell division. The elongated phenotype of YgjD depleted (p)ppGpp0 cells resembles filamentous cells blocked in cell division. However, since abrogating cell division is not inhibiting DNA replication or DNA segregation [32] it appears unlikely that YgjD directly affects cell division.