Dynamic single-cell NAD(P)H measurement reveals oscillatory metabolism throughout the E. coli cell division cycle

Abstract

Recent work has shown that metabolism between individual bacterial cells in an otherwise isogenetic population can be different. To investigate such heterogeneity, experimental methods to zoom into the metabolism of individual cells are required. To this end, the autofluoresence of the redox cofactors NADH and NADPH offers great potential for single-cell dynamic NAD(P)H measurements. However, NAD(P)H excitation requires UV light, which can cause cell damage. In this work, we developed a method for time-lapse NAD(P)H imaging in single E. coli cells. Our method combines a setup with reduced background emission, UV-enhanced microscopy equipment and optimized exposure settings, overall generating acceptable NAD(P)H signals from single cells, with minimal negative effect on cell growth. Through different experiments, in which we perturb E. coli's redox metabolism, we demonstrated that the acquired fluorescence signal indeed corresponds to NAD(P)H. Using this new method, for the first time, we report that intracellular NAD(P)H levels oscillate along the bacterial cell division cycle. The developed method for dynamic measurement of NAD(P)H in single bacterial cells will be an important tool to zoom into metabolism of individual cells.

Figure 1

Influence of exposure energy and exposure interval on E. coli growth and NAD(P)H fluorescence intensity. (a) Growth rate of E. coli at different exposure energies and three exposure intervals. Between 4 and 10 hour after loading, E. coli’s growth rate at each exposure setting was determined from two replicates with at least 20 cell tracks in total, using exponential-fitting to cell sizes. Minimal medium with 5 g/L glucose was used as culture. For each replicate, the median growth rate is shown. Each symbol represents data from an independent experiment. Number of tracked cells in Fig. 1 can be found as Supplementary Table 1. (b) Fluorescence intensity from NAD(P)H autofluorescence and from background obtained with 5 different exposure energies and 10-min exposure interval. Every bar represents data from 2 independent experiments with at least 20 cell tracks in total. For each track between 4 to 10 hours, fluorescence intensities were obtained from inside of cells (denoted as signal) and surrounding area with no cells (denoted as background). The dynamics of the background intensity with time is shown in Supplementary Fig. S1b. In each replicate, signal and background intensities from single tracks were both averaged and the mean of both replicates are shown as bars for corresponding exposure settings. Values of signal intensity at every exposure settings are listed. At 9 μJ, the ratio of NAD(P)H intensity over background intensity was low compared to the higher exposure energies, suggesting that here the NAD(P)H signal was too low to be accurately distinguished from the camera noise. Linear fitting of background and signal intensity (excluding the 9 μJ group) are shown as dashed lines. (c) Decrease of growth rate by 365 nm-light exposure for E. coli growing in minimal medium with 2 g/L fumarate, 5 g/L glucose and 5 g/L glucose supplemented with 0.5% casamino acids. Cells were exposed with and without 365 nm-light in each growth medium in three replicates and the growth rates of at least 25 cell tracks were obtained for each medium in each replicate. 22 μJ/10 min was used as exposure program. In each replicate, the mean growth rate of un-exposed cells (GR₀) was obtained and the decreases of exposed cells’ growth rate from GR₀ were calculated and averaged, as ΔGR. The mean of ΔGR from all three replicates is shown with error bars indicating standard deviation, whereas the mean of GR₀ in all three replicates is shown with error bars indicating one standard deviation.

Figure 2

Changes of autofluorescence upon metabolic perturbations demonstrate that the observed fluorescence signals stem from NAD(P)H. (a) Supplying glucose to starved E. coli increased the fluorescence intensity. Cells were cultured in minimal medium without carbon source for 5 hours (gray shaded region) after having been loaded into the microfluidic device before glucose supply. The arrival of glucose-containing medium was indicated by the descent of fluorescence dye intensity. Note, due to experimental variances, between individual experiments it can take different amounts of time until the new medium reaches cells after a medium switch. 30 cells were tracked and their fluorescence intensity, the median of their fluorescence intensity and cell size are shown as squares, solid lines and dash lines, respectively. Four images illustrate the observed upshift of autofluorescence in NAD(P)H-channel, with corresponding images in the bright field. For a quantification of the fluorescence signal, see Supplementary Fig. S3. (b) Adding 0.4 mM H₂O₂ to glucose-grown E. coli reduced the fluorescence intensity. Cells were growing in minimal medium for 3.5 hours (gray shaded region) before 0.4 mM H₂O₂ was added into the flow-channel as indicated with arrow. The increase of the fluorescence dye intensity indicates arrival of the new medium at the cells. 48 cells were tracked and their fluorescence intensity, the median of their fluorescence intensity and cell size are shown as squares, solid lines and dash lines, respectively.

Figure 3

NAD(P)H levels show an oscillatory pattern throughout the E. coli cell division cycle. (a) Median of spline-normalized NAD(P)H levels along division cycles using actual and fixed division events. Trajectories were determined from 159 cells from three independent experiments, in which E. coli was grown in glucose minimal medium and imaged at 45 μJ every 10 minutes for 10 hours. 887 division cycles were identified and were used as actual division events while 60, 70, 80, 90, 100, 110 min were, respectively, used as fixed division times in the data processing (see Materials and Methods). Three median curves generated with actual division events in three experiments (including 10% of the adjacent division cycles) were plotted as solid lines, while median curves from the data of all three experiments were generated with fixed division events and showed as gray lines. Supplementary Fig. S4h shows an alternative normalization with a linear function fitted to the first and last intensity points of every division cycle (see Methods and Materials). (b) After processing with normalization and interpolation, trajectories of NAD(P)H levels in 887 division cycles from all three experiments were clustered with k-means (4 clusters, separated with thin black lines, performed with Matlab). The color scale indicates the NAD(P)H level relative to that at initial and ending points of division cycle. The upper two clusters (n = 575, 63% of all samples) have an increased NAD(P)H level in the middle of the cycle. Supplementary Fig. S4i shows the clustering results on the basis of the alternative normalization shown in Supplementary Fig. S4h.