D R A F T Polar accumulation of pyoverdin facilitates rapid exit from stationary phase

,

the interplay between the redox status of iron (with effects 23 on bio-availability), and cellular demands that are balanced 24 against potentially lethal toxic effects. For example, in aerobic 25 environments, where bio-availability of iron is low and cellular 26 demand high, cells risk harm from hydroxyl radicals liberated 27 via the Fenton reaction. One role of pyoverdin is as protectant 28 against intracellular oxidative stress (10). But pyoverdin has 29 roles beyond iron-specific chemistry: pyoverdin chelates other 30 metals, including magnesium, zinc and gallium, with roles in 31 homeostasis and detoxification of these and other metals (11). 32 Challenges arise from the need to understand the ecological 33 relevance of numerous diverse forms of pyoverdin (12), the 34 plethora of ferri-pyoverdin receptors (13,14), and the various 35 roles of pyoverdin in shaping interactions with eukaryotic hosts 36 (15-17).

37
Because pyoverdin is a soluble extracellular product, it 38 is widely assumed to be equally available to all members of 39 a community (18)(19)(20). However, recent work shows that its 40 distribution is subject to cell-level control. In one study py-41 overdin producers retained an environment-dependent fitness cells, reducing loss into the environment (22). Furthermore, 48 Pseudomonas aeruginosa cells tune periplasmic concentrations 49 of pyoverdin in order to protect against oxidative stress (10). 50 Here we use time-lapse fluorescence microscopy to study 51 the relationship between P. fluorescens SBW25 cells and py-52 overdin. Recognizing the importance of spatial structure and 53 contributions therefrom to µm-scale features of microbial as-54

Significance Statement
Bacteria secrete extracellular products that enable nutrients to be obtained from the environment. A secreted product of significance to medicine, agriculture and biotechnology is the iron-chelating siderophore, pyoverdin, produced by members of the genus Pseudomonas. By analyzing the behavior of single cells we show that on cessation of cell division, pyoverdin accumulates at cell poles and does so in periplasmic space created by cytoplasmic shrinkage. The behavior is ecologically relevant: on encountering growth-permissive conditions, cells liberate pyoverdin and exit stationary phase with minimal delay. Our study connects a general biophysical stress response to a molecule of physiological and ecological importance.  (21) where fitness assays of pyoverdin producing SBW25 and a nonproducing pvdS defective mutant yield contrasting results depending on the culture medium. In both cases the environment is unstructured and ancestral SBW25 producer cells are rare, inoculated at 1%. A fitness advantage to nonproducing cells in KB was previously reported, but the reverse in CAA (21). In KB (top) pyoverdin nonproducing cells rarely showed evidence of accumulation of pyoverdin, whereas (bottom) this was common in CAA-cultured cells where accumulation is visible at the pole. Images were obtained from 3 µl samples of these experiments imaged under fluorescent light (excitation 390 nm / emission 475 nm) to visualize pyoverdin. All scale bars correspond to 10 µm. B) Fluorescence (390/475 nm) time-lapse images of a growing microcolony of SBW25 on an agarose pad of succinate minimal medium (SMM). Images represent selected time points including, respectively: the initial inoculum, exponential growth, end of exponential growth, and end of time-lapse acquisition (18 h total) C) Mean fluorescence intensity along the long axis of cells in a growing microcolony, when the last generation of cells is born (left) and at the end of acquistion (t = 18 h, right). Black dotted lines represent the fluorescence profile of individual cells, the red line represents a smoothed mean of all cells. Between 0 h and 18 h accumulation of pyoverdin is evident, especially at the pole.

D R A F T
To accurately characterize subcellular patterns of pyoverdin, 78 time-lapse fluorescence images of ancestral SBW25 were ob-79 tained in defined succinate minimal medium (SMM), where 80 succinate acts both as carbon source and weak iron chela-81 tor. During exponential phase and early stationary phase, 82 SBW25 cells showed a phenotype typical of fluorescent Pseu-83 domonas with pyoverdin being homogeneously distributed in 84 the periplasm. In late stationary phase fluorescent pyoverdin 85 foci appeared at the cell pole, indicating accumulation of the 86 siderophore. Polarization of pyoverdin was evident by qual-87 itative observation (Fig. 1B) and was verified after image 88 analysis and segmentation by superposing fluorescence profiles 89 at different time points (Fig. 1C).

90
By examining cell division throughout the time-lapse se-91 ries it was possible to explore the dynamic accumulation of 92 pyoverdin. The corresponding frequency of polarized cells 93 ( Fig. 2A) was tracked by classifying segmented cells automati-94 cally as "accumulated" and "non-accumulated" using a machine 95 learning algorithm (Supplementary Materials and Methods). 96 After inoculation onto microscope slides, cells undergo a pe-97 riod of acclimation without division. No polar accumulation 98 of pyoverdin was observed during this phase. As division be-99 gins, age of cells in microcolonies decreases until it reaches a 100 minimum that marks exponential phase. Pyoverdin continued 101 to be evenly distributed. Note that the small frequency of 102 cells identified as "accumulation" in the plot falls within the 103 range of classification error. Finally, the population enters 104 stationary phase and cells continue ageing without division. 105 Accumulation of pyoverdin at cell poles increased within the 106 first few hours and was evident in a majority of cells by 18h 107 (duplicate experiments are given Fig. S1).

108
While the time-averaged state of microcolonies exposes 109 population-level dynamics of polarization -namely, that ac-110 cumulation happens in stationary phase -it obscures the 111 behavior of individual cells. To specifically analyze single 112 bacteria, cells were grouped according to division status over 113 three generations: F0 for initial inoculum (cells where birth 114 could not be identified but division was observed), F1 for cells 115 in exponential phase (born from the division of F0 cells and 116 later underwent division) and F2 for the daughter cells of F1 117 that entered stationary phase and remained constant for the 118 final hours of the experiment. Growth measurements from 119 F1 cells, and pyoverdin accumulation measurements from F2 120 cells, corroborate -despite some variability in the onset of 121 accumulation -that accumulation of pyoverdin is incompatible 122 with active cell division. Cells either elongate, or accumulate 123 fluorescence (pyoverdin) at the cell pole. This result holds 124 for both old and new poles (Fig. 2B). Curiously, pyoverdin 125 accumulates preferentially at the new pole, but not exclusively, 126 and sometimes distinct foci are present at both (Fig. S2) 127 Stationary phase marks cessation of cell division due to 128 nutrient depletion. On an agarose pad, entry into stationary 129 phase is highly variable, with access of cells to nutrients and 130 oxygen depending on position and proximity to neighbour-131 ing cells. To study pyoverdin accumulation under controlled 132 conditions cells were exposed to two stressors that abruptly 133 arrest cell division, but by different mechanisms: 1, an iron 134 chelating agent (2,2'-dipyridil (DP)) and; 2, a protein synthesis-135 disrupting antibiotic (tetracycline) (Fig. S4A). Both stressors 136 induced accumulation at the pole (Fig. 2C).

137
Precisely because polarization appears when cells are 138 starved and/or stressed, processes associated with cell death, 139 D R A F T such as cell wall damage, are potential elicitors. We thus asked

206
While the genes involved in extraction of ferric iron and 207 translocation into the cytoplasm have not been characterized in 208 P. fluorescens, a recent study in P. aeruginosa suggests poten-209 tial candidates for polarization including a siderophore-binding 210 periplasmic protein (9). Deletion of the homologs of these 211 genes in SBW25 PFLU_2048, PFLU_2041, and PFLU_2043 212 did not disrupt accumulation of pyoverdin (Fig. S6). Thus, 213 neither aggregation of pyoverdin at trafficking points across the 214 periplasm, nor accumulation via the ferripyoverdin complex 215 recycling system, are involved in polarization.

216
Additional possibilities for polar accumulation are via con-217 nections to extracellular polymer synthesis and cell morphology. 218 SBW25 secretes cellulose when constructing bacterial mats at 219 the air-liquid interface (30, 35), which could trap pyoverdin 220 at the perimeter of cells. However, pyoverdin localization was 221 not altered in SBW25 ∆wsp∆aws∆mws (36) that is unable 222 to produce cellulose ( Fig. 3D.4). Finally, the rod-cell-shape 223 characteristic of Pseudomonas was considered a possible con-224 tributory factor. To test this, pyoverdin accumulation was 225 observed in a spherical ∆mreB mutant (37). Despite the 226 aberrant cell shape, this mutant displayed fluorescence foci of 227 pyoverdin after extended culture ( Fig. 3D.5).

228
Fitness effects. Polar accumulation of pyoverdin is linked to 229 cell physiology, but is unperturbed by known genetic determi-230 nants of pyoverdin biosynthesis, regulation, or transport. This 231 leads to the possibility that pyoverdin accumulation under 232 conditions of cellular stress is an accidental consequence of the 233 biophysics of cell biology and lacks ecological relevance. The 234 alternate hypothesis is one of ecological relevance: a plausible 235 explanation being that release of accumulated pyoverdin allows 236 cells to reduce the time required to exit stationary phase.

237
An ideal test would be an experiment in which time to 238 resumption of growth of an ancestral type that accumulates 239 pyoverdin is compared to a mutant unable to accumulate 240 pyoverdin, under conditions where pyoverdin accumulation 241 both occurs, and doesn't occur. If accumulation of pyoverdin 242 confers a fitness advantage, then ancestral types are expected 243 to resume growth more rapidly than non-accumulating mu-244 tants, but only after cells have first experienced conditions 245 that promote accumulation of pyoverdin.

246
Unfortunately, as evident above, no such pyoverdin-247 accumulating mutant exists, nonetheless, progress is possi-248 ble via an experiment that exploits the fact that a PvdS 249 mutant cannot accumulate pyoverdin because it is unable to 250 produce it. To this end, SBW25 and a PvdS mutant (SBW25 251 pvdSG229A(D77N) (21) (hereafter termed Pvd -) were grown 252 separately in SMM for sufficient time to ensure cells entered 253 stationary phase (24 h). A sample of cells from both popula-254 tions was then transferred to independent SMM agarose pads 255 and the time to first cell division determined. While a slight 256 delay in median time for resumption of growth was observed 257 in Pvdcompared to SBW25 the distribution of data points 258 shows significant overlap. The findings were not affected by 259 transfer of cells from overnight SMM culture to agarose pads 260 additionally supplemented with excess iron (Fig. 4A, left). 261 pad. Measurement of the time to resume growth showed little 282 difference among the competing genotypes (Fig. 4B). This is 283 consistent with the expectation that SBW25 releases accumu-284 lated pyoverdin into the (fresh) medium to the benefit of both 285 producer and non-producer alike. Notable though is the fact 286 that producer cells reap greatest benefit. When nonproducers 287 are instead inoculated in the presence of rare ancestor cells 288 (∼ 1%) their lag time again increases significantly, support-289 ing the conclusion that newly released pyoverdin underpins 290 the faster division of producers (Fig. S8). Furthermore, the 291 amount of pyoverdin accumulated during starvation seems to 292 be sufficient to support at least twice as many cells, as the lag 293 time of SBW25 producer cells was not affected by the presence 294 of nonproducers. [2] Periplasmic pyoverdin is exported into the external medium by a complex that includes the transporter OpmQ. There, it chelates insoluble iron (Fe 3+ ).

D R A F T
[3] Ferripyoverdin complexes (no longer fluorescent) are then imported back into the periplasm after binding to the receptor FpvA. This receptor is known to also bind free pyoverdin (7). In the periplasm, iron is extracted and pyoverdin is again recycled into the external medium by OpmQ. Polarization might also be determined by genes unrelated to the pyoverdin pathway: [4] SBW25 is known to secrete polymers such as cellulose that might trap pyoverdin (30).
[5] Pyoverdin could accumulate at the cell poles due to the rod shape of SBW25. D) Mutants associated to the main processes depicted in A) and their phenotype with regards to pyoverdin polarization. Mutants were grown on an agarose pad as described and fluorescence images displaying pyoverdin were taken after 16 h. The pyoverdin nonproducing mutant pvdSG229A(D77N) was co-inoculated with SBW25 to enable access of the mutant to pyoverdin. This mutant was tagged with a red fluorescent protein to aid identification.
a cellulose-degrading Bacillus isolated from a compost heap 301 on minimal medium with cellulose as sole carbon source (38).

302
Because SBW25 is unable to degrade cellulose, it must rely 303 on the Bacillus species to obtain carbon for growth. These the phenotype (e.g., P. putida KT2440). In others, amendment 327 with an iron chelator was necessary to observe the effect (e.g., 328 P. aureofaciens U149), at the same dosage that induced polar-329 ization in SBW25, and in the specific case of P. aeruginosa 330 PAO1, very high doses of DP were required. This range of re-331 sponses could reflect the secretion of secondary siderophores by 332 some strains (12) or differences in the regulation of pyoverdin 333 production (20), and suggest that polarization is an ecologi-334 cally relevant trait that varies depending on the evolutionary 335 history of each lineage.

337
Previous work showing that the population-level distribution of 338 pyoverdin changes depending on nutrient status (21), contact 339 with neighboring cells (22), and environmental stress (10) 340 motivated our investigation. With focus on P. fluorescens 341 SBW25, and using time-resolved microscopy, we have shown 342 that pyoverdin transiently accumulates at cell poles, that 343 localization is a reversible process associated with arrest of 344 cell division, and is affected by factors such as entry into 345 stationary phase and deprivation of specific nutrients (Fig. 2). 346 Particularly significant is demonstration that accumulation of 347 pyoverdin has ecological relevance (Fig. 4). and pyoverdin showed an almost perfect overlap (3A,B). This 395 leads to the conclusion that polar accumulation of pyoverdin 396 is part of a general cellular response to starvation. Thus obser-397 vations made in E. coli with a chimeric reporter (31) appear 398 to hold for SBW25, but with our observations connecting cy-399 toplasmic shrinkage to accumulation of a biologically relevant 400 molecule. Additionally we demonstrate that accumulation 401 delivers beneficial effects on fitness.
402 Data in Fig. 4A show that cells with localized pyoverdin 403 resume growth more rapidly compared to non-accumulating 404 cells. A time-to-first cell division advantage was not evident 405 when growth-arrested cells were transferred to iron-replete 406 conditions. This demonstrates that liberation of the stock of 407 pyoverdin accumulated during growth cessation speeds the 408 time to growth resumption after growth arrest, presumably 409 through provision of available iron to growing cells. 4B.

410
Of additional interest is evidence from cell-level observa-411 tions that rare pyoverdin producers are largely unaffected by 412 the presence of Pvdmutants (Fig. S8). Such mutants have 413 often been referred to as "cheats" and thus expected to neg-414 atively impact the fitness of pyoverdin producers (so called 415 "cooperators"). Lack of detrimental impact is further evidence 416 that pyoverdin production preferentially benefits producer cells 417 and is indicative of preferential benefit to pyoverdin producing 418 cells. This finding further calls into question the fit between 419 social evolution theory and production of extracellular prod-420 ucts by microbes (20, 21). It also gels with work showing that 421 populations of siderophore-producing Pseudomonas can be 422 grown in the presence of an excess of nonproducing mutants 423 without a significant reduction in overall yield (46).

424
Recent work suggests important physiological differences 425 between acute (and unexpected) interruptions of cell division, 426 and gradual growth arrest that determines entry into station-427 ary phase (47). In the former case, sudden arrest induces a 428 disrupted state that results in large variability in cell-level du-429 ration of lag phase. In the latter case, cells implement genetic 430 programs that buffer the effects of impending starvation cre-431 ating cohesive population-level responses (47). The seemingly 432 adaptive nature of pyoverdin accumulation under stress is 433 reminiscent of -and perhaps even connected to -the capacity 434 of SBW25 to enter a semi-quiescent capsulated state upon 435 starvation. During starvation, SBW25 cells produce an excess 436 of ribosomes that allow rapid exit from stationary phase once 437  degrading Bacillus (Fig. 5) overdin fluorescence and 100 ms 20 % intensity for red fluorescence).

512
Images were taken with 63x and optovar 1.6x magnification.