Speed Switching of Gonococcal Surface Motility Correlates with Proton Motive Force

Bacterial type IV pili are essential for adhesion to surfaces, motility, microcolony formation, and horizontal gene transfer in many bacterial species. These polymers are strong molecular motors that can retract at two different speeds. In the human pathogen Neisseria gonorrhoeae speed switching of single pili from 2 µm/s to 1 µm/s can be triggered by oxygen depletion. Here, we address the question how proton motive force (PMF) influences motor speed. Using pHluorin expression in combination with dyes that are sensitive to transmembrane ΔpH gradient or transmembrane potential ΔΨ, we measured both components of the PMF at varying external pH. Depletion of PMF using uncouplers reversibly triggered switching into the low speed mode. Reduction of the PMF by ≈ 35 mV was enough to trigger speed switching. Reducing ATP levels by inhibition of the ATP synthase did not induce speed switching. Furthermore, we showed that the strictly aerobic Myxococcus xanthus failed to move upon depletion of PMF or oxygen, indicating that although the mechanical properties of the motor are conserved, its regulatory inputs have evolved differently. We conclude that depletion of PMF triggers speed switching of gonococcal pili. Although ATP is required for gonococcal pilus retraction, our data indicate that PMF is an independent additional energy source driving the high speed mode.

with high spatial and temporal resolution. The piezo stage was actuated by a computercontrolled force feedback loop to record pilus dynamics at constant force. For this, the displacement of the bead due to a pilus retraction was allowed to reach a certain threshold. Hereafter, the displacement and thus the force was kept constant by moving the piezo stage with the bacterium relative to the optical tweezers. The movement of the bacterium (piezo stage) is a direct measure of the pilus length change.The data was analyzed with software written in MATLAB R2009b (Math Works Inc.). Raw data was recorded with 20 kHz. Each event was smoothed with a nonlinear median filter using a moving window of 1 ms. Subsequently, data was down-sampled to a resolution of 10 ms for N. gonorrhoeae and 50 ms for M. xanthus, respectively. Velocities v were calculated by the difference quotient of downsampled datasets. To account only for retraction periods, datasets were dissected into elongation (v < -150 nm/s), retraction (v > 150 nm/s) and pause periods with |v| = 150 nm/s. Retraction speed switching was identified by analyzing speed histograms and pilus length changes versus time of consecutively measured retraction events.

Simultaneous measurements of dissolved O 2 and motility.
Sealed chambers with a volume of 50µl equipped with the oxygen sensor were assembled as for twitching motility assays. PtTFPP phosphorescence and bacterial motility in brightfield mode were simultaneously monitored at the PDMS/medium interface of the oxygen sensor. PtTFPP signals were collected with a Cy3 filter set (EX: BP 535 ± 25 nm, BS: 580 nm, EM: LP 590 nm, Zeiss) and 100 ms integration time. The phosphorescence of PtTFPP is quenched by the presence of oxygen resulting in a strong effect on the viability of bacteria most likely due to generation of reactive oxygen species. To circumvent this problem the field of view (82×82 µm) was shifted by linearly scanning the sample with a step size of 200 µm. At each positions, surface motility of M. xanthus was recorded with a rate of 1 fps for a duration of 1 min followed by a single capture of PtTFPP phosphorescence. Thus, the oxygen measurement was not conducted at a single position, but monitored along a line through the entire sample. The scanning procedure was feasible because the distribution of cells was homogeneous and the response of the oxygen sensor only marginally depended on the position inside the sample.
Calibration: Phosphorescence of PtTFPP follows the Stern-Volmer equation: (1) Here, is phosphorescence intensity, intensity in the absence of oxygen, Stern-Volmer constant and [ ] the oxygen concentration as previously described (Kurre & Maier, 2012). The phosphorescence of the entire field of view was averaged for each time point and the corresponding oxygen concentration was calculated by Eq. (1) with K SV = (0.70 ± 0.09) l/µmol and I 0 = 27 896 ± 545.

Measurement of transmembrane pH gradients with cFSE.
The fluorescence of cFSE was measured in flow cells during twitching motility assays with a Yellow GFP emission filter (HQ 535 ± 30 nm, AHF) and a 100 ms integration time. Excitation of pH-insensitive wavelength (~ 440 nm) was achieved by a Cyan GFP excitation filter (D 436 ± 10 nm, AHF) and a 450 nm beam splitter (450 DCLP, AHF), excitation of pHsensitive wavelength (~ 490 nm) by a Yellow GFP excitation filter (HQ 500 ± 20 nm, AHF) and a 515 nm beam splitter (Q 515 LP, AHF) (Fig. 6a). Supplementary figure S1a illustrates the fluorescence signals of a single cell after excitation at 490 nm and at 440 nm, respectively. Via automated multipoint acquisition fluorescence of 50-150 single cells were captured for each excitation wavelength at at least 6 different positions in the sample. Single cell fluorescence was analysed by routines written in MATLAB. Following the analysis of Lo et al. (2), the total single cell fluorescence intensities and were calculated by where is pixel intensity, average background intensity around the cell (window 30×30 pixels) and a well-defined threshold to account mostly for intracellular fluorescence.
was calculated by fitting a Gaussian function, ( ) ( ( ) ⁄ ), to the pixel intensities in the window around the cell. The threshold was defined by ( ) , where is maximum pixel intensity and the smallest I for which and ( ) . The final measure was the ratio ⁄ . The latter had to be calibrated. Therefore, the PMF was depleted by 30 µM nigericin and 5 µM valinomycin (incubation time ~ 30 min) to guarantee that pH ex = pH in . Under these conditions, the calibration curve shown in supplementary Fig. S1b was obtained. This calibration curve could be used to determine pH at external pH ranging from pH ex =6.0 up to pH ex =7.8.

Measurement of transmembrane pH gradients with individual pHluorin expressing cells.
The cells expressing pHluorin were immobilized on the polystyrene spin-coated glass slides and observed in the sealed chambers with a volume of 50µl. Fluorescence of fixed cells was observed using excitation wavelengths of 410nm (excitation increases as pH increases), achieved by BrightLine HC 406/15 excitation filter (AHF) and 470nm (excitation decreases as pH increases), achieved by a Yellow GFP excitation filter (HQ 500 ± 20 nm, AHF) and a 515 nm beam splitter (Q 515 LP, AHF). Emission was observed at Yellow GFP emission filter (HQ 535 ± 30 nm, AHF) and a 100 ms integration time. Via multipoint acquisition fluorescence single cells were captured for each excitation wavelength at a minimum of 6 different positions in the sample.
Single cell fluorescence was analysed by routines written in MATLAB. Windows (30x30 pixels) with single bacteria were found. Using the isodata algorithm (Ridler & Calvard, 1978) on the combined fluorescence of both the 410nm and 470nm excited window, an automatic threshold was found and the same pixel mask was assigned to both wavelengths. The average fluorescence inside the pixel mask was obtained for each excitation and corrected for background intensity, which was calculated by fitting a Gaussian function on the pixel window. In a separate sample the same analysis was done on wild-type cells, which do not express pHfluorin, to obtain an estimate of the auto-fluorescence of the cells. After deduction of this auto-fluorescence estimate, the corrected intensity values were used to compute the 410/470 fluorescent ratio of the cell inside the window. The standard curve was obtained by fitting a sigmoid function to the fluorescent ratios obtained with a collapsed ΔpH. This fit was then used to convert the fluorescent ratios to internal pH values.
For calibration (Fig. S2a) cells were preincubated for 10 min with 10mM EDTA to increase membrane permeability for nigericin. Afterwards ΔpH was collapsed by the addition of 50 μM nigericin and 40 mM methylamine hydrochloride with the incubation time ~ 50 min at 37°C.

Measurement of transmembrane pH gradients with pHluorin expressing cells using fluorescence spectroscopy.
Fluorescence excitation spectra of radiometric pHluorin in different pH-adjusted buffers were determined with a spectrophotometer (Perkin Elmer LS55) (Fig. S2b). The buffers used had the concentration of 30mM MES (pH5.5), MOPS (pH6, pH6.5), and HEPES (pH7, pH 7.5) . The pH was determined after enriching each of these buffers with 11mM glucose and 8mM sodium pyruvate. For calibration, pHluorin expressing cells were adjusted to an OD 600 10.0 and sonicated (Fig. S2c). Samples with defined pH were excited at between 350nm and 500nm and the fluorescence emission was collected at 510nm. Spectra in Fig. S2b were normalized to their isobestic point at 428 ± 5nm. For background correction, fluorescence collected from N400 cells in the respective buffers was substracted. For determination of pH in , pHluorin expressing cells were treated as described but not sonicated. Data from three independent experiments were averaged.

Quantification of membrane potential Δψ.
Due to the membrane potential Δψ, a cationic dye such as TMRM equilibrates across the membrane according to Boltzmann's law: where ⁄ is ratio of intracellular and extracellular TMRM concentration, charge of TMRM, Boltzmann's constant and T absolute temperature. The ratio ⁄ could not be directly derived from raw fluorescence data ⁄ because diffraction of single cell fluorescence would lead to an underestimation of . Therefore, a correction factor ( ⁄ ) was derived from a three dimensional convolution model (Lo et al., 2007): Experimental point spread function: For the convolution model the experimental point spread function ( ) was determined by single 20 nm fluorescence microspheres (FluoSpheres, Molecular Probes, Ex. 580 nm, Em. 605 nm), which were fixed to a cover glass. The motorized stage allowed for precise scans in z-direction with a step size nm and µm regarding the centre of the fluorescence spheres. With a pixel size of 160nm a cubic grid of the was obtained. For the convolution model the was background corrected and normalized. Photobleaching was negligible. Supplementary Fig.  S3a shows the at the centre of the sphere and at extreme z positions.
Convolution model: N. gonorrhoeae cells appear as mono-or diplococci. The diameter of these spherical cocci is 0.5-1.0 µm determined by transmission electron microscopy (3). Therefore, gonococci were modelled by a sphere with 750 nm in diameter (monococcus) or with two spheres side by side (diplococcus), which stuck to a cover glass. Because a microscopic image had a pixel size of 160 nm, the radius of the sphere was only about two pixels in size. In this model, the TMRM concentration ( ) is distributed as follows: where indicates the corresponding voxel in a cubic grid with an edge length of 160 nm. Using the experimental , a modelled blurred image ( ) of ( ) at midcell position was calculated by convolution: This image was the input for Eq.
(2) to calculate the internal fluorescence with . The external fluorescence was determined by the averaged pixel intensity with . Supplementary Fig. S3b illustrates the real intensity profile ( ) before diffraction, the modelled blurred image ( ), and a binary image of all pixels above threshold defined by Eq. (2). By varying the ratio ⁄ , the correction factor as a function of was calculated by Eq. (4) (Supplementary Fig. S3c). Finally, Eq. (3) enabled the determination of the transmembrane potential for a measured ratio ⁄ ( Supplementary Fig. S3d). Due to the marginal difference between the curves for modelled mono-and diplococci, the transmembrane potential of single gonococci was analysed with an averaged correction factor ( ⁄ ) without considering appearance of cells. The goodness of this convolution model and the fact that single cell fluorescence was collected from moving cells during twitching motility assays should influence the uncertainty of Δψ measurements.
To account for these error sources, the error of Δψ was estimated by the standard deviation and not, as usual, by the standard error.
Experimental procedure: Fluorescence signals were collected with a Cy3 filter set (Excitation: BP 535 ± 25 nm, beam splitter: 580 nm, emission: LP 590 nm, Zeiss) 30 min after loading with TMRM (conducted in a flow cell as for twitching motility assays). Via automated multipoint acquisition 100-200 single cells were captured with 100 ms integration time at midcell position at ten different positions in the sample. For each cell, the total single cell fluorescence was determined by Eq.
(2) and averaged to obtain 〈 〉. Furthermore, the average background intensity 〈 〉 of all cells was calculated, which was proportional to the external fluorescence . To account for membrane bound TMRM fluorescence , cells were treated with 50 µM CCCP for 30 min to deplete and a second multipoint acquisition was performed to acquire 〈 〉. In a last step, TMRM was flushed out by thoroughly washing with RAM and the average background 〈 〉 was measured. Then, was determined by 〈 〉 〈 〉. Finally, the internal fluorescence due to free TMRM was estimated as

Supporting figures
Supporting Fig. S1: Measurement of pH using cFDA-SE. a) Fluorescence intensity F of a single diplococcus for emission at 525nm and excitation at 490nm (left) and 440nm (right). b) Calibration curve of cFSE. Ratio of fluorescence intensities F 490nm /F 440nm as a function of extracellular pH in the presence of 5µM valinomycin and 30µM nigericin (each data point averaged over 500-1000 cells).
Supporting Fig. S2: Calibration of pHluorin expressing cells. a) Calibration curves of pHluorin expressing cells using single cell analysis. Each data point is averaged of 140 -500 cells. b) Spectrum of pHluorin expressing gonococci after lysis at varying pH. Average of 3 independent data sets. c) Calibration curves of lysed pHluorin expressing cells using a fluorescence spectrometer. Average and standard deviation of three independent data sets. Please note that the filter set used for microscopic single cell analysis recorded a different ratio as the fluorescence spectrometer.
Supporting Fig. S3: a) Experimental point spread function (PSF) at z = -2.4 µm (below focus plane), z = 0 µm (midcell position) and z = +2.4 µm (above focus plane). Scale bar 2 µm. b) Convolution model for single cell fluorescence. Top: real intensity profile I r of a diplococcus at midcell position modelled by two spheres with 750 nm in diameter. Pixel size 160 nm. Mid: I m is corresponding intensity profile after convolution with experimental PSF. Bottom: I mask is a binary image illustrating all pixels above the intensity threshold defined by Eq. (3). c) Correction factor S versus measured ratio of internal and external TMRM fluorescence F in /F ex for modelled monococci (black crosses) and diplococci (gray crosses). d) Corresponding transmembrane potential Δψ versus F in /F ex calculated by Eq. (4) and (5).