Dynamic bimodality of curli expression in planktonic cultures of Escherichia coli is stabilized by cyclic-di-GMP regulation

Curli amyloid fibers are major constituent of the extracellular biofilm matrix formed by bacteria of the Enterobacteriaceae family. Within Escherichia coli biofilms, curli gene expression is limited to a subpopulation of bacteria, leading to heterogeneity of extracellular matrix synthesis. Here we show that bimodal activation of curli expression also occurs in well-mixed planktonic cultures of E. coli, resulting in stochastic differentiation into distinct subpopulations of curli-positive and curli-negative cells at the entry into the stationary phase of growth. Monitoring curli activation in individual E. coli cells growing in a microfluidic device revealed that the curli-positive state is only metastable and it can spontaneously revert during continuous growth in a conditioned medium. The regulation by c-di-GMP is not required for curli gene activation or for differentiation of E. coli in subpopulations of curli-producing and curli-negative cells. Instead, we observe that c-di-GMP modulates the probability and dynamics of stochastic curli activation and enhances stability of the curli-positive state.


Introduction 31
Curli amyloid fibers are the key component of the extracellular matrix produced during biofilm 32 formation by Escherichia coli, Salmonella enterica, and other Enterobacteriaceae [1][2][3][4][5][6][7][8][9]. In E. 33 coli and S. enterica serovar Typhimurium, curli genes are organized in two divergently 34 transcribed csgBAC and csgDEFG operons that share a common intergenic regulatory region 35 [10]. Expression of these operons is under regulation of the stationary phase sigma factor σ S 36 (RpoS) and thus becomes activated during the entry into the stationary phase of growth [4, 11-37 14]. This activation is achieved by the σ S -dependent induction of the transcriptional regulator 38 CsgD, which then controls the expression of the csgBAC operon that encodes the major curli 39 subunit CsgA along with the curli nucleator CsgB and the chaperone CsgC [7,8,15]. In turn, 40 csgD expression in E. coli and S. Typhimurium is either directly or indirectly regulated by 41 multiple cellular factors that mediate responses to diverse environmental changes, including In this study we demonstrate that stochastic differentiation of E. coli csgBAC operon 68 expression into distinct subpopulations of curli-positive and -negative cells occurs during the 69 entry into the stationary phase in a well-stirred planktonic culture, and thus in absence of any 70 environmental gradients. Similar stochastic and reversible differentiation could be observed 71 among cells growing in conditioned medium in the microfluidic channel. The upstream 72 regulation by c-di-GMP is not required to establish the bimodality of curli expression, but it 73 determines the fraction of curli-positive cells and enhances the stability of curli activation.

Materials and methods 76
Bacterial strains and plasmids 77 All strains and plasmids used in this study are listed in Table S1   Macrocolony biofilms were grown as described previously [26]. Briefly, 5 µl of the overnight 112 liquid culture grown at 37°C in lysogeny broth (LB) medium (10 g tryptone, 10 g NaCl, and 5 g 113 yeast extract per liter) was spotted on salt-free LB agar plates supplemented with Congo red 114 (40 µg/ml). Plates were incubated for 8 days at 28°C.

118
Ltd., Switzerland) were done using fluorescence excitation at 483 nm and emission at 535 nm.

119
Relative fluorescence was calculated by normalizing to corresponding OD600 values of the 120 culture.

121
For fluorescence measurements using flow cytometry, aliquots of 40-300 µl of liquid bacterial 122 cultures were mixed with 2 ml of tethering buffer (10 mM KH2PO4, 10 mM K2HPO4, 0.1 mM 123 EDTA, 1 µM L-methionine, 10 mM lactic acid, pH 7.0). Macrocolonies were collected from the 124 plate, resuspended in 10 ml of tethering buffer and then aliquots of 40 μl were mixed with 2 ml 125 of fresh tethering buffer. Samples were vigorously vortexed and then immediately subjected to 126 flow cytometric analysis using BD LSRFortessa Sorp cell analyzer (BD Biosciences, Germany) 127 using 488-nm laser. In each experimental run, 50,000 individual cells were analyzed. Absence 128 of cell aggregation was confirmed by using forward scatter (FSC) and side scatter (SSC) 129 parameters. Data were analyzed using FlowJo software version v10.7.1 (FlowJo LLC, inverted microscope with a time interval of 10 min. Details of image analysis are described in 142 Supporting protocols.

Results 145
Bimodal curli expression is induced in planktonic culture

146
In order to characterize curli expression in planktonic culture of E. coli, we followed the 147 induction of chromosomal transcriptional reporter of csgBAC operon, where the gene encoding 148 for green fluorescent protein (GFP) was cloned with a strong ribosome binding site as a part 149 of the same polycistronic RNA downstream of csgA [28]. In our previous study of submerged 150 E. coli biofilms, this reporter showed bimodal expression both in the surface-attached biofilm 151 and in the pellicle at the liquid-air interface [28]. When E. coli culture was grown at 30°C in 152 tryptone broth (TB) liquid medium, this reporter became induced during transition to the 153 stationary phase ( Figure 1A), which is consistent with previous reports [12,14]. The observed 154 induction of curli expression occurred at similar density in the cultures with different initial 155 inoculum size. In both cases the onset of induction apparently coincided with the reduction of 156 the growth rate, which likely occurs due to depletion of amino acids in the medium and 157 induction of the stringent response [36,37], consistent with proposed role of stringent response 158 in the regulatory cascade leading to curli gene expression [18,23]. In agreement with that, 159 curli expression was strongly reduced when E. coli culture was grown in a concentrated TB 160 medium ( Figure S1A) or when TB medium was supplemented with serine ( Figure S2A).

161
Moreover, the induction of curli reporter was strongly enhanced by addition of serine 162 hydroxamate (SHX), which is known to mimic amino acid starvation and induce stringent 163 response [38] ( Figure S2A).

164
In order to investigate whether curli expression was uniform or heterogeneous across 165 planktonic E. coli population, we next measured curli reporter activity in individual cells using 166 flow cytometry. The reporter was induced only in a fraction of cells, and this bimodality of curli 167 expression became increasingly more pronounced at later stages of culture growth, reaching 168 its maximum in the overnight culture ( Figure 1B). Thus, the bimodal induction of curli gene 169 expression is observed not only in biofilms but also in a well-mixed planktonic culture. While 170 curli activation was more pronounced in a cell culture growing in an orbital shaker ( Figure 1B), 171 bimodality was also observed during culture growth in the plate reader ( Figure S1B). Notably, 172 stimulation of curli expression by SHX or its suppression by additional nutrients affected the 173 fraction of positive cells rather than their expression levels ( Figure S1B and Figure S2B).    Figure S4).

216
Also individual dgc and pdh gene knockout strains showed reduced activation of curli reporter 217 ( Figure S4). Of note, another difference with the flask culture was that the low-fluorescence 218 peak of the wildtype culture was not fully negative but apparently contained a large fraction of 219 cells with incompletely activated curli reporter, which could also be seen in ∆pdeH or ∆pdeR 220 knockouts but not in the ∆pdeH ∆dgcE ∆pdeR ∆dgcM, ∆pdeR ∆dgcM or ∆pdeH ∆dgcE strains 221 ( Figure 3B and Figure S4). Similar results were obtained even upon prolonged incubation in 222 the plate reader ( Figure S5), confirming that the observed difference with the overnight flask 223 culture was not because of the different growth stage.

224
We further tested reporter activation under growth conditions that favour biofilm formation.

233
We also grew all strains in the form of macrocolony biofilms on an agar plate [26]. Interestingly,

234
here the extent of reporter activation in the ∆pdeR ∆dgcM and ∆pdeH ∆dgcE ∆pdeR ∆dgcM 235 strains was much higher and comparable to that of the wildtype ( Figure 3D) and even individual 236 ∆dgcE and ∆dgcM knockouts showed high fraction of curli-positive cells ( Figure S7A), 237 consistent with their stronger Congo red staining compared to the ∆mlrA negative control 238 ( Figure S7B). Summarily, these results confirm that the regulation by c-di-GMP is not required 239 for (bimodal) curli activation, but also suggest that in absence of this control the fraction of 240 culri-positive cells is more sensitive to growth conditions.

243
In order to investigate the dynamics of curli activation, and the effects of c-di-GMP regulation, 244 at the single-cell level, we utilized "mother machine", a microfluidic device where growth of 245 individual bacterial cell lineages could be followed in a highly parallelized manner over multiple

557
The switch to conditioned medium is at time zero, as indicated. Shaded area is the interquartile range.

558
The number of cells in the device varies with time, but is on average n = 296 for WT and n = 522 for the

637
Growth rate drops rapidly and cells switch on curli expression after a switch to conditioned medium.

758
To track cells between time points, we applied a length conservation strategy for the cells along

765
Since cells may grow in length between time points, we also initialised a growth rate parameter 766 for each cell that biased the expected cell lengths for time point t+1 as a fold-increase in length.

767
To enable adaptation to the true growth rate, the growth rate parameter was updated by a 768 lagging average over 20 time points. To increase robustness to errors in segmentation, we 769 additionally allowed state transitions from one to many and many to one, and built a proposal sister cells could later be segmented as a single mother cell. Specifically, when two sister cells

781
-i.e., cells that were previously involved in a division event -merged into one, the label was 782 set back to that of the mother; at the next division event, the labels of the sister cells were also 783 retained. Finally, note that any outlines below a minimum size threshold of 50 pixels were 784 ignored. All errors in tracking were manually curated.

785
Cell length was estimated from cell regions as the 'major axis length' of the Matlab regionprops