Cell shape independent FtsZ dynamics in synthetically remodeled cells

The FtsZ protein is a key regulator of bacterial cell division. It has been implicated in acting as a scaffolding protein for other division proteins, being a force generator during constriction, and more recently, as an active regulator of septal cell wall production. During an early stage of the division cycle, FtsZ assembles into a heterogeneous structure coined the “Z-ring” due to its resemblance to a ring confined by the midcell geometry. While in vitro experiments on supported lipid bilayers have shown that purified FtsZ can self-organize into a swirling ring roughly the diameter of a bacterial cell, it is not known how, and if, membrane curvature affects FtsZ assembly and dynamics in vivo. To establish a framework for examining geometrical influences on proper Z-ring assembly and dynamics, we sculptured Escherichia coli cells into unnatural shapes, such as squares and hearts, using division- and cell wall-specific inhibitors in a micro fabrication scheme. This approach allowed us to examine FtsZ behavior in engineered “Z-squares” and “Z-hearts”, and in giant cells up to 50 times their normal volume. Quantification of super-resolution STimulated Emission Depletion (STED) nanoscopy data showed that FtsZ densities in sculptured cells maintained the same dimensions as their wild-type counterparts. Additionally, time-resolved fluorescence measurements revealed that FtsZ dynamics were generally conserved in a wide range of cell shapes. Based on our results, we conclude that the underlying membrane environment is not a deciding factor for FtsZ filament maintenance and treadmilling in vivo.


167
The '  Next, we wanted to know if drug-treated cells placed in deep (5 µm) rectangular 169 volumes would adapt to these shapes and effectively form 'Z-rectangles' or 'Z-170 squares' instead of 'Z-rings'. Previous work has shown that cells can adapt to 171 rectangular shapes in shallow wells, approximately 1 µm deep 19 . Here, we produced 172 quadrilateral patterns in agarose pads using silica micron pillar arrays similar to those 173 previously described 14 , with the exception that the pillars were rectangular and 5.5 ± 174 0.5 µm in height. Side lengths of the micron chambers were up to 3.5 µm in length 175 ( Supplementary Fig. S7), resulting in well volumes up to 80 µm 3 , roughly 50-fold larger 176 than the volume of a WT cell (assuming a WT cell size of 2 µm in length and 1 µm in 177 width) ( Supplementary Fig. S8). 178 Drug-exposed cells expressing FtsZ-mNeonGreen were placed in rectangular micron 179 holes and incubated at room temperature for 300 -420 minutes (longer incubation 180 times were needed due to increased well size). The cells adapted to their new shapes 181 and formed rectangular cuboids with only one "Z-square" per cell (Fig. 3a,182 Supplementary Movie SM4). Notably, FtsZ densities were observed both in the sharp 183 corners and along the sides of the rectangles (Fig. 3b, Supplementary Fig. S9). 184 Quantification of the FtsZ-mNeonGreen densities showed that they had similar 185 dimensions to those in untreated cells, with an average length of 105.4 ± 39.6 nm and 186 width of 79.6 ± 18.2 nm (n = 147) (Fig. 3c). This suggests that FtsZ filament 187 dimensions in vivo are insensitive to membrane curvature (or lack thereof). 188 To generate a fluorescent FtsZ fusion protein that could be used for both super-189 resolution STED imaging and examination of filament dynamics when grown in rich 190 media at 37 °C, we constructed a plasmid-expressed FtsZ-mCitrine fusion.  mCitrine was expressed from an IPTG-inducible, medium copy-number plasmid, 192 pTrc99a, at a level approximately equal to 30 % of total cellular FtsZ. Under these 193 conditions, FtsZ-mCitrine formed normal-looking, sharp Z-rings (Supplementary Figs. 194 S1 and S2). Cells expressing FtsZ-mCitrine were then exposed to drugs, trapped in 195 rectangular micron-sized holes, and incubated for 180 -280 minutes at room 196 temperature before gSTED imaging. We found that FtsZ-mCitrine formed filaments 197 that were 118.3 ± 41.3 nm long and 86.3 ± 22.5 nm wide (n = 162), similar to  mNeonGreen filament dimensions (Fig. 3c, Supplementary Fig. S10), indicating that 199 fluorophore choice did not influence cluster dimensions in the rings. For consistency, 200 we also imaged rectangular cells expressing FtsZ-GFP from the chromosome using 201 SIM ( Supplementary Fig. S10). All three strains tested adapted to the rectangular 202 shape, producing sharped-cornered "Z-rectangles". 203

FtsZ dynamics in rectangular-shaped cells 205
In order to examine the dynamics of FtsZ in rectangular cells, we performed time-lapse 206 imaging on cells expressing either FtsZ-mCitrine or FtsZ-GFP. Although a few 207 fluorescence spots were abnormally bright and immobile (~ 1 spot / 5 cells, with a 208 maximum of 2 spots in one cell) (Fig. 4b, Supplementary Movie SM7. Red arrow), the 209 majority of FtsZ densities were highly dynamic ( Fig. 4a-b, Supplementary Movies SM5 210 -SM6). Note that the bright, immobile spots were excluded from treadmilling analyses. 211 Close inspection of time-lapse sequences suggested that FtsZ bundles in rectangular-212 shaped cells may be able to treadmill, even in right-angled corners 213 Supplementary Movie SM8). The average speed of FtsZ-mCitrine densities in 214 rectangular cells with perimeter lengths up to 13 µm (more than four times the 215 9 the measured treadmilling speed of FtsZ-GFP in rectangular cells (25.3 ± 11.3 nm/s, 217 n = 122) (Fig. 4d), large cylindrical cells (30 ± 18 nm/s, Fig. 1m) and untreated cells (~ 218 25 nm/s) 13,14 . 219 To determine whether the dynamics of FtsZ subunit exchange are affected by changes 220 to circumferential length and shape, we collected FRAP measurements on FtsZ 221 bundles in rectangular-shaped cells ( with mean t1/2 recovery times of 9.85 ± 2.58 s (n = 24) and 9.15 ± 2.55 s (n = 22) for 224 FtsZ-mCitrine and FtsZ-GFP, respectively (Fig. 4f). This suggests that subunit 225 exchange from the cytoplasmic FtsZ pool is independent of circumference length and 226 membrane curvature. The data thus far indicate that the maintenance and dynamics 227 of FtsZ filaments are preserved in both large Z-rings and Z-rectangles of varying size. 228 229

FtsZ dimensions and dynamics in heart-shaped cells 230
To examine whether FtsZ could literally be (at) the heart of cell division, we engineered 231 micron pillar arrays that were heart-shaped ( Supplementary Fig. S11). Heart shapes 232 were chosen because they would sculpt cells in such a way that highly curved, straight, 233 and angled membrane segments would be present within a single cell. Drug-treated 234 E. coli cells expressing cytoplasmic GFP, FtsZ-mNeonGreen or FtsZ-mCitrine were 235 sculptured into hearts as described above (Fig. 5a). Perhaps not surprisingly, 236 quantification of 155 individual FtsZ densities from the heart-shaped cells revealed 237 dimensions similar to those in round and rectangular cells (129 ± 44 nm long and 84 238 ± 9 nm wide) (Fig. 5b). We also found that the average speed of FtsZ-mCitrine in heart-239 shaped cells (22 ± 10 nm/s, n = 44) was essentially the same as that in untreated cells 240 For about one-third of the heart-shaped cells, we noticed bright spots of internalized 242 FtsZ-FP signal that accumulated close to the cell center ( Figure 5c, green arrowhead). 243 Although we couldn't distinguish whether these were true FtsZ clusters or aggregated 244 protein, cytoplasmic clustering of FtsZ in WT cells have previously been reported 12 . 245 Furthermore, although most hearts had FtsZ-FP signal spanning the full perimeter of 246 the cell, approximately 20 % were only "half full" (Fig. 5d, left). We do not fully 247 understand the underlying reason for this, however it is unlikely due to image focus or 248 cell tilt issues, as every cell was scanned in the z direction prior to imaging. 249 Nevertheless, when we subjected the heart-shaped cells to FRAP, fluorescence 250 recovery rates were equal for both full and half-full hearts ( Fig. 5d), with mean t1/2 251 recovery times of 7.1 ± 1.1 s (n = 24) and 6.9 ± 0.9 s (n = 9), respectively ( Fig. 5e). 252 253

FtsZ-"rings" form in complex cell shapes 254
To explore if cell geometry plays a role in Z-"ring" formation, we set out to remodel 255 cells into other complex shapes. Even though highly complex-shaped bacteria occur 256 in nature, such as star-shaped bacteria 26 , we wanted to test whether rod-shaped E. 257 coli cells would allow themselves to be drastically remodeled. Using micron pillars of 258 various shapes, we produced holes in agarose pads such that drug-exposed cells 259 could be sculptured into complex shapes, such as pentagons, half-moons, stars, 260 triangles and crosses (Fig. 6a, middle row. Supplementary Fig. S11). The cells 261 conformed remarkably well to these shapes, forming sharp boundary angles < 70° 262 ( Fig. 6a, Star). After we confirmed that cells could adapt to these complex shapes, we 263 placed cells expressing FtsZ-mCitrine into the micron holes, allowed for reshaping to 264 occur, and then imaged the cells using STED nanoscopy. Cells of all tested shapes 265 produced easily recognizable FtsZ-"shapes" at midcell (Fig. 6a, bottom row). Cells, both bacterial and eukaryotic, have the ability to adapt remarkably well to their 274 local environments 27-31 , reverting to their original shapes after stress 32,33 and dividing 275 with striking midcell accuracy even when remodeled into irregular cell shapes 27,30 . In 276 bacteria, the tubulin homologue FtsZ assembles into a ring-like structure at midcell 277 and is responsible for overall maintenance of the cell division machinery 5,6 . The 278 general dynamics and organization of the FtsZ-ring have been shown to be quite 279 similar across many bacterial species 11,[13][14][15]17,[34][35][36][37] . Common to these species is 280 confinement of the FtsZ-ring to a circular geometry at midcell. Strikingly, when purified 281 FtsZ (together with its membrane anchor FtsA) is placed on supported lipid bilayers, it 282 assembles into a dynamic, swirling ring-like assembly with a diameter resembling that 283 of wild-type E. coli cells (approximately 1 µm), hinting at an intrinsically preferred FtsZ-284 ring curvature 6,38 . 285 In this study, we characterized FtsZ midcell accumulation and dynamics in cell shape-286 determining environments by 'looking through the Z-ring' along the long-axis of cells. 287 We observed normal-looking FtsZ-rings in cells with diameters three times the size 288 found in WT cells. However, this might not be surprising, considering only ~ 30 % of 289 the pool of FtsZ molecules are in the ring of WT cells at any given point in time 39 . 290 Quantification of FtsZ dimensions revealed little variation between different cell 291 shapes, such as squares, pentagons, triangles and stars (on average 123 x 80 nm, 292 length x width, respectively, and summarized in Table 1), suggesting that local 293 membrane geometry has minimal influence on FtsZ cluster dimensions. Compared to 294 untreated cells, rectangular and heart-shaped cells with perimeter lengths more than 295 four times that of a WT cell exhibited similar overall dynamics of FtsZ, as FtsZ-FP 296 fluorescence densities treadmilled at the same average velocity and FtsZ subunit 297 exchange occurred at similar rates ( Table 2), independent of cell shape and size. 298

299
In summary, our results from different shaped cells show that Z-"ring" formation and 300 dynamics are not limited to cells of a certain shape or size. This agrees with previous 301 findings, which show that internal cellular structures are maintained in cells that have 302 been reshaped into unnatural forms 19 . Our observation that FtsZ clusters conform to 303 the geometric shape of the membrane at midcell suggests that FtsZ-ring formation is 304 not affected by changes in membrane curvature. Indeed, cell shape and size are 305 important for proper cellular functions 40 , however, with the many naturally-occurring 306 shape variations of bacteria 26,41 , it is perhaps not surprising that FtsZ can adapt to 307 changing environments without compromising its own ability to maintain fundamental 308 functionality. Although our data do not explicitly show that sculptured cells can divide 309 (since downstream division proteins were inhibited), the fact that the dynamic 310 properties of FtsZ are conserved suggests that this may be possible. One particular 311 implication of this is the notion that the Z-ring can be decoupled from the division 312 process but with maintained dynamics, making treadmilling a possible requirement for 313 divisome assembly and organization in rod shaped model bacteria, as previously 314 suggested for cocci 17 . Presently, we have shown in vivo that E. coli FtsZ-ring 315 formation and dynamics are conserved, irrespective of cell shape and size.

Western blot analysis 345
Cell extracts from a volume corresponding to 0.1 OD600 units were collected for each 346 strain to be analyzed. The extracts were suspended in loading buffer and resolved by 347 SDS-PAGE gel electrophoresis. Proteins were transferred to nitrocellulose 348 membranes using a semi-dry Transfer-Blot apparatus (Bio-Rad). The membranes 349 were blocked in 5 %(w/v) milk and probed with antisera to FtsZ (Agrisera, Sweden) 350 and detected using standard methods. 351 352

Nanofabrication of micro arrays 353
Micron pillars were engineered using two different, but related, approaches. The first 354 approach was used for round and square/rectangular micron pillars, and was adapted 355 from 14,15 . Briefly, using a multi-step process similar to that described in 44 , micron-356 scale pillars were fabricated on a silicon (Si) substrate by reactive ion etching. A 357 pattern of hard-baked photoresist was created on a Si surface using UV lithography, 358 to work as a mask for etching. Subsequent etching was performed using an Oxford 359 Plasmalab100 ICP180 CVD/Etch system, with a mixture of SF6 and O2 plasma as an 360 etchant. For our process, a SF6:O2 ratio of 1:1 was optimal. After etching, the 361 remaining photoresist was removed by O2 plasma treatment. Pillar arrays (1 x 1 cm or 362 2 x 2 cm) with round pillars were engineered to contain one micron-sized pillar every 363 5 μm, with dimensions between 0.9 and 3.5 μm wide and 5.25 ± 0.75 μm high 364 ( Supplementary Fig. S3). Pillar arrays (1 x 1 cm) with square pillars contained micron-365 sized pillars approximately every 5 μm, with side lengths varying between 1.8 and 3.5 366 μm, and heights of 5.5 ± 0.5 μm (Supplementary Fig. S7). 367 To create more complex shapes, a second approach, based on electron beam 368 lithography was used. For this, the micron-scale structures were fabricated on a Si 369 substrate by a multi-step process, which was a combination of electron beam 370 lithography and reactive ion etching techniques. Similar approaches to silicon 371 patterning are described in a number of earlier works 44-47 . First, a pattern of e-beam 372 resist was created on a Si surface using e-beam lithography. A 50 nm-thick Ti layer 373 was then deposited, and a lift-off process was used to create a metal mask for etching. 374 The use of a metal mask, instead of a baked e-beam resist mask, was necessary due 375 to the high selectivity ratio required for generating structures only a few microns in 376 height. Finally, the etching process was performed as described above, using an 377 Oxford Plasmalab100 ICP180 CVD/Etch system and a mixture of SF6 and O2 plasma 378 as an etchant. For our process, a SF6:O2 flow ratio of 3:2 produced the best results, 379 with a Si:Ti etching selectivity ratio of approximately 100:1. Increased concentration of 380 O2 in the mixture has two effects: (i) it improves etching anisotropy, which is essential 381 for avoiding shape distortion from the undercut effect, and (ii) it reduces the selectivity 382 ratio, as the Si etch rate gets slower. After etching, the structures were characterized 383 using a Dektak surface profiler and SEM imaging. The micron structure arrays, which 384 contained various shapes (hearts, triangles, pentagons, half-moons and crosses), 385 were fabricated on 1 x 1 cm Si chips with inter-structure distances of approximately 5 386 μm, and structure heights of 5.5 ± 0.5 μm (Supplementary Fig. S11).

Micron-sized chamber production and cell growth 391
Liquefied agarose (5% w/v) in M9 minimal media (supplemented with 0.2 % glucose, 392 0.1 % casamino acids, 2 μg ml -1 thiamine, 40 μM A22 and 20 μg ml -1 cephalexin) was 393 dispersed on glass slides and the silica mold (pillar facing downwards) was placed on 394 top. The molds contained either round or rectangular pillars, or various geometrical 395 shapes, as described above. Once the agarose solidified, the mold was removed and 396 ~ 5 μl of live cell culture at OD600 0.4 -0.55 (pre-treated with 16 μM A22 for 10 -15 397 minutes) was applied on top. To allow the cells to adapt to the different shapes, slides 398 were incubated at RT or 30 °C in a parafilm-sealed petri dish together with a wet tissue 399 to prevent drying. After incubation, cells were covered with a pre-cleaned cover glass 400 (♯1.5) for live cell imaging. For STED imaging, cells were first fixed with ice-cold 401 methanol for 5 minutes and carefully rinsed with PBS prior to cover glass application. Confocal Z-stacks (focal plane ± ~ 3.5 μm) were acquired on a Leica TCS SP8 STED 424 3X system (operated in confocal mode) using predetermined optimal system settings 425 (Leica, LAS X), with 0.22 μm steps (resulting in 30-32 images per stack), and pinhole 426 size 1 AU. All imaging was performed at RT (~ 23-24 °C). 427

FRAP measurements 429
Confocal FRAP measurements were performed on a Zeiss LSM780 system using a 430 100x 1.4 NA plan Apo oil immersion objective and pinhole size 60 μm, as described 431

14
. Bleaching was performed for 0.5-0.7 s using 100% laser power applied over the 432 region of interest. Data were collected in time intervals of 1 -2 sec until steady state 433 was reached. Following background correction, and to account for overall successive 434 bleaching, the fluorescence intensity (F) of the bleached region (half a ring) was 435 normalized to the average ring fluorescence of an unbleached area of the same size, 436 for each time point (t); FNORM(t) = FBLEACHED(t)/(FBLEACHED(t)+ UNBLEACHED(t)). All data 437 were exported to Origin9 Pro and data points were fitted to the single exponential 438 function F(t) = Fend -(Fend -Fstart)* e -kt , where F(t) is the fluorescence intensity at time 439 t, Fend is the fluorescence intensity at maximum recovery, Fstart is the fluorescence 440 recovery momentarily after bleaching (at t = 0), and k is a free parameter. The recovery 441 half-time was then extracted from t1/2 = ln 2 / k. Importantly, all cells were scanned 442 from top to bottom in order to find the division plane (in which the rings reside). 443 444

Image analysis 445
Image analysis was performed using Fiji. When necessary, images were background-446 corrected using a rolling ball with radius 36. Image stacks were motion-corrected using 447 the plug-in StackReg. Kymographs were generated from time-lapse images using the 448 KymoResliceWide plugin (line width 5), from which treadmilling speeds were 449 calculated using the slope of the fluorescence trace, as previously described 38 .  Kymographs were taken along the yellow line starting at the yellow arrowhead (left 720 kymograph), or over the bright spot indicated by the red arrow (right kymograph). The 721