Cell-to-cell variation and specialization in sugar metabolism in clonal bacterial populations

While we have good understanding of bacterial metabolism at the population level, we know little about the metabolic behavior of individual cells: do single cells in clonal populations sometimes specialize on different metabolic pathways? Such metabolic specialization could be driven by stochastic gene expression and could provide individual cells with growth benefits of specialization. We measured the degree of phenotypic specialization in two parallel metabolic pathways, the assimilation of glucose and arabinose. We grew Escherichia coli in chemostats, and used isotope-labeled sugars in combination with nanometer-scale secondary ion mass spectrometry and mathematical modeling to quantify sugar assimilation at the single-cell level. We found large variation in metabolic activities between single cells, both in absolute assimilation and in the degree to which individual cells specialize in the assimilation of different sugars. Analysis of transcriptional reporters indicated that this variation was at least partially based on cell-to-cell variation in gene expression. Metabolic differences between cells in clonal populations could potentially reduce metabolic incompatibilities between different pathways, and increase the rate at which parallel reactions can be performed.


Bacterial growth on AOC
We performed an additional experiment to estimate the amount of biomass that can be assimilated from AOC in our system. Four MG1655 replicate cultures were grown overnight in LB medium, shaking at 37°C, and then diluted 1 to 100 into M9 medium supplemented with 0.2% glucose. On the following day, 2 ml of each culture was pelleted, washed with M9 medium that does not contain any supplemented sugar -M9--Ø medium -and resuspended in 2.5 ml of M9--Ø medium. The cultures were grown overnight in glass culture tubes. The cultures were then diluted 1 to 100 into glass culture tubes with M9--Ø medium and grown overnight. For these cultures colony forming units (CFU) were determined immediately after dilution ('day 0') and 1 day later ('day 1'), by taking 4 samples from each tube and preparing 10--fold serial dilutions in 96--well microplates using a multi--channel pipette (10 µl was transferred into 90 µl of M9--Ø medium). From each dilution 5 µl was pipetted onto an LB agar plate. After drying of the drops at room temperature, the plates were incubated at 37°C for 9 to 13 hours. Colonies were counted for the least diluted spot that still had distinguishable colonies. We calculated an increase of 1.6 x 10 6 CFU/ml between 'day 0' and 'day 1'.

Sample preparation for NanoSIMS
Fixed bacterial samples were applied on polycarbonate filters with a pore size of 0.2 µm (GTTP, Millipore) coated with Au/Pd alloy, and stamped out to get a round filter piece with 5 mm in diameter. The filters were put in a desiccator overnight, in dark, and the following day they were washed first very quickly in sterile Milli--Q filtered H 2 0 and then with absolute ethanol to remove the residual stain. In order to preserve the fluorescence of the strains, the filters were placed on glass microscope slides, mounted in anti--fading medium and stored in dark at 4°C until analyzed with fluorescence microscopy. The anti--fading medium contained 3 parts of Citifluor (Citifluor Ltd., London, UK), and 1 part of VectaShield (Vector Laboratories, CA, USA).

NanoSIMS measurements
Using fluorescence microscopy we first selected fields with cells on Au/Pd coated polycarbonate filters [Musat et al., 2008] and marked the areas with a laser micro--dissection microscope Zeiss 200M (Scientific Center for Optical and Electron Microscopy (ScopeM) of ETH Zurich). We recorded images of green and red fluorescence, and fluorescence of the DNA stain for the marked fields (see the section 'Fluorescence microscopy' below, and 'Analysis of filters with fluorescence microscopy and NanoSIMS' in Methods). Selected fields of cells were analyzed with a NanoSIMS 50L (Cameca, Gennevilliers Cedex--France) at the Max Planck Institute for Marine Microbiology in Bremen. The filters were first washed with absolute ethanol to remove the residues of the mounting agents and after drying mounted into NanoSIMS 50L. The areas of interest were pre--sputtered with a Cs + primary ion beam of 150 pA to remove surface contamination, to implant Cs + ions in the sample and to achieve an approximately stable ion emission rate. A primary Cs + with a beam current between 1.5 and 2 pA and a beam diameter around 100 nm were rastered across the cells for analysis.
For each individual cell, secondary ion images of 1 H --, 2 H --, 12 C --, 13 C -and secondary electrons were simultaneously recorded from analysis area of 10 x 10 µm in raster size and an image size of 256 x 256 pixels with a dwell time of 1 ms per pixel. Analysis area of 20 x 20 µm in raster size and an image size of 512 x 512 pixels were applied for overnight measurements. We measured 50 planes for each position and chose each field of view such that silhouettes of single cells can be distinguished. To minimize interferences the instrument was tuned for high mass resolution (around 7000 MRP). As an internal control of NanoSIMS performance, we measured the facility's E. coli sample, which is deuterium unlabeled but 13 C--labeled.

Fluorescence microscopy
For measuring fluorescence of the strains cultivated in mini--chemostats we acquired phase contrast and fluorescence images by first applying cells on a 1.5% agarose pad using a cavity slide as described in [Bergmiller et al., 2011]. We harvested 5 ml of bacterial culture from mini--chemostats, and concentrated the cells by immediately spinning them down at 4°C. The pad was inoculated with 1 µl of the concentrated bacterial culture, and the slide was then mounted onto an inverted fluorescence microscope (Olympus IX81), equipped with a cooled CCD camera (Olympus XM10), 100x oil objective, and X--Cite 120PV fluorescence lamp (Lumen Dynamics Group Inc., Canada). Fluorescence images were acquired with a lamp intensity of 50% and exposure time of 300 ms for both GFP filter (Chroma U--N41001: BP 460--500 nm, BA 510--560 nm, DM 505 nm) and RFP filter (Olympus U--MSWG2: BP 480--550 nm, BA 590 nm LP, DM 570 nm). DAPI filter (Olympus U--MNUA2: BP 360--370 nm, BA 420--460 nm, DM 400 nm) was used only for the analysis of NanoSIMS filters. (BP: excitation filter; BA: barrier filter; DM: dichromatic mirror) Bleed--through estimation In order to quantify the fraction of bleed--through from the GFP signal to the mCherry signal, and vice versa, we measured fluorescence of single reporter system based on the plasmids pGFP [Refardt et al., 2013] and pRFP. pRFP has the same sequence as pGFP except the gfp gene is replaced by the mCherry gene. We grew overnight wild--type strain MG1655 harboring one of these plasmids in M9 salts minimal medium containing 1 mM MgSO 4 and 0.1 mM CaCl 2 , and supplemented with 3 mM D--glucose, 3 mM L--arabinose and 100 µg/ml of ampicillin (AppliChem). The overnight cultured were diluted 100--fold and grown for 2 hours with 250 µM IPTG (Promega). The fluorescence was measured with the lamp intensity set on 50% and exposure time of 100 ms for both filter sets. Red fluorescence bleed--through factor = red fluorescence (pGFP) / green fluorescence (pGFP) Green fluorescence bleed--through factor = green fluorescence (pRFP) / red fluorescence (pRFP) Prior of computing bleed--through factors, fluorescence values were corrected for background and autofluorescence. The mean values for each bleed--through factor were used in subsequent analysis.

Analysis of fluorescence images taken from agarose pads
Recorded images were analyzed with modified Matlab analysis package Schnitzcells [Young et al., 2011], and data was extracted with custom--made Matlab scripts. First, cell outline was determined by phase contrast for images taken from agarose pads. After that, red and green fluorescence signals for each cell were extracted, and values reported in the text are mean values of fluorescence for each cell, defined as total fluorescence divided by the cell area [Young et al., 2011]. Background correction. Background fluorescence was determined for each image by quantifying the fluorescence in an area containing no cells. This quantity was subtracted from fluorescence values as determined for cells. The fluorescence is presented in arbitrary units (A.U.), meaning that for every recorded image, its background region had fluorescence of 0 A.U. in each fluorescence channel. Autofluorescence correction. We measured autofluorescence of wild--type MG1655 on agarose pads. The mean autofluorescence values for green and red fluorescence were subtracted from values of the cells of reporter strains. Bleed--through correction. Red fluorescence values were corrected for the bleed--through factor, real red fluorescence= red fluorescence -(green fluorescence * factor), for every analyzed cell. Bleed--through to green fluorescence channel was negligible. Cell length. We used segmentation (phase contrast) images of fluorescence microscopy analysis to infer differences in cell length across different conditions, three replicates per each condition. Cell length in carbon--limited chemostats was 2.20 ± 0.025 µm (mean ± standard error of the mean), and in nitrogen--limited chemostats 3.11 ± 0.060 µm, measured for strain NN114 under defined setup of the fluorescence microscopy. Average variation in cell length was comparable between two conditions, CV of 0.28 for carbon--limited chemostats and 0.30 for nitrogen--limited, carbon--excess chemostats.

Measurements for estimation of model parameters
This section describes how the maximum growth rates and yields on glucose and arabinose presented in Model Table 1 and Model Table 3 were estimated (see Supplementary Information S2 File, 'Mathematical Model'). Populations of strain NN114 were grown overnight at 37°C in M9 medium supplemented with either 0.6 mM Glc or 0.7 mM Ara, corresponding to 0.01% of each sugar. The next day cultures were diluted 20--fold into fresh medium. After 2.5 hours of growth the cultures were washed twice with M9 medium without sugar and then inoculated to an A 600 of 0.0001 into fresh M9 medium with either 1.1 mM of glucose or 1.3 mM of arabinose, corresponding to 0.02% of each sugar. Individual wells of a 96--well plate were filled with 200 µL of these cultures, and growth was measured in a BioTek Eon Microplate Spectrophotometer. Resulting A 600 measurements were background subtracted and the maximum growth rate was determined by fitting an exponential line with base e to data between A 600 = 0.03 and A 600 = 0.15. The growth yield was determined by quantifying the cell concentration in the distribution and have the same medians, we used non--parametric tests: Kolmogorov--Smirnov test and Mann--Whitney U test for two independent datasets, and Kruskal--Wallis test for more than two independent datasets. To quantify variation in the sugar assimilation and generation time, we used coefficient of variation (CV), i.e. standard deviation divided by the mean.