Probing chemotaxis activity in Escherichia coli using fluorescent protein fusions

Chemotaxis is based on ligand-receptor interactions that are transmitted via protein-protein interactions to the flagellar motors. Ligand-receptor interactions in chemotaxis can be deployed for the development of rapid biosensor assays, but there is no consensus as to what the best readout of such assays would have to be. Here we explore two potential fluorescent readouts of chemotactically active Escherichia coli cells. In the first, we probed interactions between the chemotaxis signaling proteins CheY and CheZ by fusing them individually with non-fluorescent parts of a ‘split’-Green Fluorescent Protein. Wild-type chemotactic cells but not mutants lacking the CheA kinase produced distinguishable fluorescence foci, two-thirds of which localize at the cell poles with the chemoreceptors and one-third at motor complexes. Cells expressing fusion proteins only were attracted to serine sources, demonstrating measurable functional interactions between CheY~P and CheZ. Fluorescent foci based on stable split-eGFP displayed small fluctuations in cells exposed to attractant or repellent, but those based on an unstable ASV-tagged eGFP showed a higher dynamic behaviour both in the foci intensity changes and the number of foci per cell. For the second readout, we expressed the pH-sensitive fluorophore pHluorin in the cyto- and periplasm of chemotactically active E. coli. Calibrations of pHluorin fluorescence as a function of pH demonstrated that cells accumulating near a chemo-attractant temporally increase cytoplasmic pH while decreasing periplasmic pH. Both readouts thus show promise as proxies for chemotaxis activity, but will have to be further optimized in order to deliver practical biosensor assays. IMPORTANCE Bacterial chemotaxis may be deployed for future biosensing purposes with the advantages of its chemoreceptor ligand-specificity and its minute-scale response time. On the downside, chemotaxis is ephemeral and more difficult to quantitatively read out than, e.g., reporter gene expression. It is thus important to investigate different alternative ways to interrogate chemotactic response of cells. Here we gauge the possibilities to measure dynamic response in the Escherichia coli chemotaxis pathway resulting from phosphorylated CheY-CheZ interactions by using (unstable) split-fluorescent proteins. We further test whether pH differences between cyto- and periplasm as a result of chemotactic activity can be measured with help of pH-sensitive fluorescent proteins. Our results show that both approaches conceptually function, but will need further improvement in terms of detection and assay types to be practical for biosensing.


59
Chemotaxis is the behaviour of cells to bias the direction of their motility in reaction to perceived 60 chemical gradients (1, 2). Chemotaxis by bacteria is rapid (ms-to s-scale) and does not require de 61 novo gene induction, since it is based on dynamic protein modifications and protein-protein 62 interactions (3,4). The rapidity of the chemotaxis response is potentially interesting for the 63 development of alternative biosensor assays for chemical exposure testing. One current popular 64 biosensor method relies on living bacterial cells equipped with synthetic gene circuits, which enable 6 hypothesized that the interaction between phosphorylated CheY (CheY~P) and CheZ would favor 137 binding of the split parts, inducing proper folding and eGFP fluorescence emission (Fig. 1A). Both 138 fusion proteins were expressed from a single plasmid-located operon in E. coli under control of a 139 constitutive synthetic promoter. Several promoters were tested, with an approximate difference in 140 "strength": P AA > P JJ > P II > P OO (strains 4610, 4703, 4701 and 4702, respectively)

146
We thus concluded that the foci were the genuine result of CheY-C eGFP/CheZ-N eGFP interactions and 147 not the result of spontaneous split-eGFP reconstitution and subsequent multimerization.

148
Global fluorescence intensities (i.e., averaged across the whole cell) of E. coli expressing 149 cheY-C egfp-cheZ-N egfp were not significantly higher than in the control strains (e.g., 4717, 4728, 4729, 150 4743), except for E. coli strain 4610 expressing cheY-C egfp-cheZ-N egfp from the strongest promoter 151 P AA (Fig. S2). This strain also showed the highest level of foci fluorescence, in comparison to E. coli 152 expressing cheY-C egfp-cheZ-N egfp from P OO , P JJ or P II (Fig. S2). The number of foci varied between 0-153 5 per cell (Fig. 1C). In particular E. coli strain 4610 (P AA ) showed fewer foci than strains 4701-4703 154 (P II , P OO and P JJ , respectively) and in most cells only a single (polar) focus was observed (

178
Split-eGFP expression in chemotaxis deletion mutants. In order to confirm that the foci 179 detected were a consequence of interaction between CheY~P and CheZ, we introduced the P JJ -cheY-180 C egfp-cheZ-N egfp (pCRO9) construct into an E. coli mutant lacking the gene encoding the CheA 181 kinase. Deletion of CheA abolishes phosphorylation of CheY and of the CheY-C eGFP fusion protein, 182 which should cancel interactions with the CheZ phosphatase. E. coli ∆cheA (pCRO9) cells indeed did 183 not show any foci (Fig. 2C), which is in agreement with our hypothesis and confirms that the observed 184 foci in e.g., E. coli strain 4703 or 5395 must be the result of physical interaction between ( Fig. 2C), but the average number of foci per cell was statistically significantly reduced compared to wild-type E. coli strain 4703 with pCRO9 (p=0.023, pair-wise t-test; Fig. 2D). This suggested that eGFP foci could no longer be detected at all (Fig. 2C). The fliM gene encodes the motor protein with 195 which CheY~P interacts to invert flagellar rotation, but absence of FliM also prevents export of the

201
To provide additional evidence that CheY~P-C eGFP and CheZ-N eGFP also interact at the 202 flagellar motors, we constructed an E. coli MG1655 derivative strain co-expressing the CheY-CheZ 203 split-eGFP fusion proteins and a FliM-mCherry fusion protein (strain 5921, Table 1). Fluorescent foci 204 of both eGFP and mCherry were visible in individual cells, although in general the eGFP foci were a 205 bit more 'crisp' due to lower background than in case of the mCherry foci (Fig. 3A). Automated 206 segmentation of cells and foci indicated a larger average number of mCherry than eGFP foci per cell 207 ( Fig. 3B, C). eGFP foci, as before, were most frequently located at the cell poles, although 208 consistently about one-third of foci was found closer to the mid-cell (Fig. 3B). In contrast, two-thirds 209 of all FliM-mCherry foci were found along the mid-cell (Fig. 3B), suggesting an on average larger 210 number of spatially distinct flagellar motor than receptor complexes (at the resolution of regular 9 wells on top of a microscopy slide immersed in motility buffer. Foci fluorescence was recorded by microscopy imaging every minute, while focusing on cells attached to the slide, before and after 221 addition of ligand (100 µM Ser or Ni 2+ ). Fluorescence response curves of individual foci were variable 222 and mostly decreased over time, as a consequence of photobleaching (Fig. 4A). In comparison to the 223 average fitted slope of foci fluorescence decay (Eq. 1) in E. coli cells remaining in motility buffer, the 224 proportion of foci with a significantly slower decay was higher in cells exposed to 100 µM Ni 2+

231
Comparison of fluorescence intensity distributions of stable split-eGFP foci in immobilized E.

232
coli ∆cheY∆cheZ (pCRO9) cells in the wells over time (Fig. S3), showed a slight increase towards 233 brighter foci in case of cells exposed to Ni 2+ compared to motility buffer, although this was not 234 statistically significant (Fig. S4). For cells exposed to serine, the proportion of weaker foci tended to 235 increase (Fig. S4). In case of immobilized E. coli ∆cheY∆cheZ (pCRO32) producing unstable split-236 eGFP-ASV, more brighter foci appeared over time after exposure to Ni 2+ , whereas no consistent 237 changes in foci distribution appeared after exposure to serine (Fig. S4).

238
In the next experiment, instead of being attached to wells filled with motility buffer, E. coli 239 ∆cheY∆cheZ (pCRO32) cells were deposited on agarose surfaces at different distances from either a 240 source with 100 µM serine, 100 µM Ni 2+ , or in motility buffer only. Exposure to Ni 2+ led to a net 241 increase in the proportion of brighter foci compared to motility buffer, whereas exposure to serine did 242 not measurably change foci brightness' distribution ( Fig. 4C). In contrast, the overall number of foci    The goal of this work was to develop different readouts of the chemotaxis pathway of E. coli, which 278 might be used as proxies for chemotactically active cell behaviour, and which might eventually be 279 exploited for biosensing. Although other approaches have been taken, we focused here on two types of 280 reactions. In the first, we followed interactions between CheY~P and CheZ as a proxy of ligand 281 binding to the chemoreceptors in E. coli using BiFC with split-eGFPs. In the second approach, we 282 studied potential dynamic pH-changes in chemotactically active cells using the pH-sensitive 283 fluorophore pHluorin.

284
Our results demonstrated that CheY-C eGFP and CheZ-N eGFP fusion proteins were 285 functionally complementing chemotaxis in an E. coli ΔcheYcheZ deletion mutant background (Fig. 2).

286
Chemo-attraction of E. coli expressing CheY-C eGFP and CheZ-N eGFP was slightly less steep than that 287 of E. coli MG1655 WT, both in presence or absence of native CheY/CheZ (Fig. 2). This suggests that

294
We found that eGFP foci colocalize both with the chemoreceptors and with the flagellar motors ( Fig.   295 3), both of which are known sites for CheY~P interaction (44). Expression of CheY-C eGFP and CheZ-296 N eGFP in a Δtsr E. coli mutant background showed a decrease in the average number of foci per cell, 297 but did not abolish foci formation altogether. This suggests that Tsr partially stabilizes CheY~P-CheZ 298 interaction (Fig. 2D). No foci were detected in E. coli ΔfliM, but this is not (only) the result of direct 299 absence of motor proteins stabilizing the interaction to CheY-C eGFP. In absence of FliM the anti-complexes (Fig. 3). Hence, we conclude that the CheY~P/CheZ interactions can take place both at the 307 motor and at the receptor. The latter is consistent with theory because CheY and CheZ are localized at 308 the receptor, as shown by Sourjik and Berg (44), but CheY~P, once phosphorylated by CheA, has to 309 diffuse to the motor to induce inversion of flagellar rotation (27) and may therefore further interact 310 with CheZ at the motor.

311
Although the levels of CheY~P in the cell are dynamic, since dependent on the binding of 312 attractants or repellents to the receptors, and further on CheZ constantly dephosphorylating CheY~P, 313 the CheY~P-CheZ-reconstituted-eGFP foci were rather stable. The cause for this is most likely the 314 stability of the reconstituted eGFP itself. By photobleaching cells in motility buffer in comparison to 315 motility buffer with addition of nickel ions (a strong repellent), we could see a trend that eGFP 316 photodecay was counteracted by renewed foci formation (Fig. 4A, B; Fig. 6A). In cells exposed to 317 serine we did not detect any significant changes of foci intensity compared to cells in motility buffer 318 only, but this might be the result of the stability of split-eGFP-CheY-CheZ foci, which do not 319 dissociate faster upon addition of attractant. When looking at the distribution of foci intensities among 320 individual surface-attached cells in the wells, there was again a trend that Ni-exposed cells 321 accumulated brighter foci over time (Fig. S4). The effects were more clear when using E. coli 322 expressing the split-eGFP appended with the ASV-destabilization tag (strain 5430, Table 1) (49). In 323 this case, foci are more rapidly degraded and the appearance of new brighter foci as in Ni 2+ -exposed 324 cells can be more easily distinguished (Fig. S4). This was further confirmed by placing E. coli strain 325 5430 cells on agarose surfaces with gradients in Ni 2+ -or serine-concentration, which led to brighter 326 foci appearing in Ni 2+ -exposed cells and an overall decrease of the number of foci in cells exposed to 327 serine (Fig. 4C, D). These results would be in agreement with what one would conjecture from the 328 expected changes in CheY~P levels (Fig. 6A). Both the foci brightness and the number of foci per cell 329 thus reflect temporal chemotaxis pathway activation through modification of the CheY~P levels by the of CheY~P and repellents leading to an increase (Fig. 6A).
thus prevents capturing much of the dynamic nature of CheY~P/CheZ interactions, which was 335 improved by deployment of the ASV-tagged eGFP. However, the downside of the ASV-tagged split-336 eGFP is that its turnover is higher, and fluorescent foci on average become much weaker and more 337 difficult to detect. Fluorescent foci detection is optimal on surface immobilized cells, but the  enabling optimal measurements. Interestingly, the pHluorin emission ratio increased in cells close to 367 the serine source compared to an empty source, whereas the TorA-pHluorin emission ratio decreased 368 closed to the serine source (Fig. 5D). The inverse response in the cyto-versus periplasm, suggests that 369 the pH of the cytoplasm increases whereas that of the periplasm decreases in chemotactically-active 370 cells (Fig. 6B). The fact that both signals were opposite indicates that they were not an artifact simply 371 due to the cell accumulation close to the source. Calculation of the cytoplasmic and periplasmic pHs 372 from the pH-calibrated pHluorin fluorescence ratios (Fig. 5B) indicated that, in absence of attractant, 373 the equivalent cytoplasmic pH corresponds to 7.8±0.16, which is 0.3 pH-unit higher than measured 374 elsewhere (40, 53), and that of the periplasm to 7.3±0.12 (Fig. 6B). A lower pH in the periplasm is in 375 general agreement with a net outside proton gradient across the cytoplasmic membrane in actively 376 respiring cells (53). In contrast, in presence of an attractant, reactive cells showed an increase of 0.3 377 pH units in the cytoplasm (from pH 7.8 to 8.1) and a decrease of 0.2 units in the periplasm (from pH 378 7.3 to 7.1) (Fig. 6B).

379
These results are not in immediate agreement with our initial hypothesis that increased the cytoplasm would on its turn facilitate the proton requirements by the flagellar motor for faster or continued rotation.
In conclusion, our study showed how autofluorescent proteins may be used to interrogate the 388 chemotaxis pathway in motile E. coli in gradients of attractant or repellent. BiFC with unstable split-389 eGFP parts fused to CheY and CheZ can to some extent reveal the dynamic behaviour of 390 CheY~P/CheZ foci (Fig. 6A), whereas pHluorin expressed in the cyto-and periplasm can measure 391 dynamic pH changes in chemotactically attracted cells (Fig. 6B). Both methods may be further 392 optimized and calibrated to allow quantitative chemotaxis readout. 400 Na 2 HPO 4 ·12H 2 O, 3.0 g l -1 KH 2 PO 4 , 0.5 g l -1 NaCl, 1.0 g l -1 NH 4 Cl, 4 g l -1 of glucose, 1 g l -1 of 401 Bacto TM casamino acids (BD Difco), Hutner's trace metals (54), and 1 mM MgSO 4 . The medium was 402 supplemented with 30 µg chloramphenicol (Cm) ml -1 for strains containing pSTV-based plasmids and 403 100 µg ampicillin (Amp) ml -1 for plasmids expressing pHluorin. All used strains are detailed in Table   404 1.

405
Cloning of the split-eGFP. The chemotaxis regulator protein CheY was fused (at its C-406 terminal end) with the C-terminal part of eGFP, and its phosphatase CheZ was fused (at its C-terminal 407 end) with the complementary N-terminal part of eGFP. First, the gene coding for cheY (without its 408 stop codon) was amplified by PCR from E. coli MG1655 genomic DNA using primers 130719 and 409 130720, elongated with BamHI and AatI restriction sites, respectively (Table S1). The 3'-end of egfp 410 ( C egfp) corresponding to amino acids 158-238 was amplified by PCR from plasmid pPROBE (38) 411 using primer 130724 containing an AatI restriction site and a sequence for a seven-amino-acid linker both were ligated downstream of the synthetic promoter P AA in pSTV28P AA mcs, cut with BamHI and containing an XhoI restriction site and a sequence encoding an eight-amino-acid linker (GGSGSGSR).

419
The 5'-end of egfp ( N egfp) encoding amino acids 2-157 was amplified from plasmid pPROBE using 420 primers 130726 and 130723, elongated with XhoI and HindIII restriction sites, respectively. Both PCR 421 fragments were then inserted downstream of the cheY-C egfp hybrid gene within the same operon (i.e, 422 under P AA control), by digestion with SpeI and HindIII, and ligation. A variety of derivative plasmids 423 was created, in which P AA was replaced by different weaker synthetic promoters named P JJ , P II , and sequences were flanked by EcoRI and BamHI restriction sites, to allow easy exchange (56).

426
Equivalent constructs of the P JJ -cheY-C egfp-cheZ-N egfp were produced, in which LVA, AAV and 427 ASV-instability tags were added to the 3'-end of C egfp (38) ( Table 1)

500
The number of deviant individual response curves was then counted for each condition, 501 motility buffer only, nickel or serine, if k Ni or Ser + SD k,Ni or Ser < AVE k,buffer -SD k,buffer , which corresponds 502 to a fitting curve that decreases less than the expected baseline second-order decay or if k Ni or Ser + 503 SD k,Ni or Ser > AVE k,buffer -SD k,buffer , which corresponds to a fitting curve that decreases more than the

531
To test dependency of pHluorin fluorescence on external pH, cultures were grown as above, 532 centrifuged, but resuspended in 50 µl of M63 minimal medium supplemented with 2 g l -1 casein 533 hydrolase, 20 mM sodium benzoate and buffer to the respective test pH. M63 medium consists of 0.4 534 g l -1 KH 2 PO 4 , 0.4 g l -1 K 2 HPO 4 , 2 g l -1 (NH 4 ) 2 SO 4 and 7.45 g l -1 KCl. To obtain pH 6.0, we used 50

572
Cells were imaged for eGFP foci at between 10 and 30 min after application, at three X-positions at 573 relative distances of 8 mm from each other, and five Y-positions 300 µm apart.

575
Foundation NanoTera project 20NA21-501 143082, and by financing from the Herbette Foundation

802
Note that distances are not differentiated here. Note the shift to brighter foci in Ni 2+ -exposed cells with

818
(A) Split-eGFP foci reconstituted from interacting CheY~P-C eGFP and CheZ-N eGFP both at 819 chemoreceptor (MCP) as well as motor complexes. In cells exposed to Ni 2+ , there is a tendency for 820 more and brighter eGFP foci. Cells perceiving serine tend to form less bright and fewer eGFP foci. For absence of attractant, motile E. coli cells maintain a difference of ~0.8 pH unit between cyto-and 823 periplasm. Cells accumulating in a radial gradient from a 100-µM serine source after 20 min increase 824 cytoplasmic pH (pH 8.4), possibly to sustain higher proton flux through the flagellar motors.