Gene expression and growth rate are highly stochastic in E. coli. Some of the growth rate variations result from the deterministic and asymmetric partitioning of damage by the mother to its daughters. One daughter, denoted the old daughter, receives more damage, grows more slowly, and ages. To determine if expressed gene products are also allocated asymmetrically, we compared the levels of expressed green fluorescence protein in growing daughters descending from the same mother. Our results show that old daughters were less fluorescent than new daughters. Moreover, old mothers, which were born as old daughters, produced daughters that were more asymmetric when compared to new mothers. Thus, variation in gene products in a clonal E. coli population also has a deterministic component. Because fluorescence levels and growth rates were positively correlated, the aging of old daughters appears to result from both the presence of both more damage and fewer expressed gene products.

Bacterial strains, growth media, and GFP reporter

Growth experiments were performed using E. coli K12 (NCM3722 ΔmotA:frt, chromosomal:T:ptet-GFP:frt) (43), which has a chromosomal insert of constitutively expressed native green fluorescent protein (gfp), unfused to any protein and thereby with no deterministic spatial placement in the cell as a mature protein. Cells were grown in M9 minimal media (44) supplemented with 0.02 mg ml^-1 of thiamine, and 0.18 mg ml^-1 of glucose as the carbon source. Protein levels were quantified by using GFP as a reporter. Because native GFP is estimated to have a diffusion rate of 9 μm² sec^-1  (31) and E. coli cell has a mean cross sectional area of about 3 μm², the protein is rapidly dispersed throughout a cell in less than 1 s. Because our fluorescence images were taken at 20 min intervals, the distribution of GFP densities in a mother cell, and consequently also in the daughters, is not diffusion limited. Rather, the different densities result from differential production or gene expression within the cells. The strain was kindly provided by Minsu Kim (Emory University).

Cell growth and microscope slides

Cells from -80˚ C glycerol stock were streaked onto agar plates. A single colony was inoculated into M9 media and grown at 37˚ C overnight. The following day the culture was diluted 1:100 in M9 and grown for 2 hours. One µl of the culture was then pipetted onto a 10 µl M9 agarose pad. The agarose pad was then was then flipped with the bacterial side down onto a 24 x 60 mm cover glass and placed over a 25 x 75 mm single depression slide sealed with vaseline (modified from earlier methods described in (20,21,23) to fit inverted microscope). Individual cells from two different movies were followed through time lapse microscopy at 37˚ C until each grew into a micro-colony of 64 cells.

Time-lapse microscopy

Cells were imaged with an inverted microscope (Nikon Eclipse Ti-S), equipped with Nikon NIS-Elements AR control software, 100X objective  (CFI Plan APO NA 1.4), external phase contrast rings for full intensity fluorescence imaging (FITC), fluorescence light source (Prior Lumen 200) with motorized shutter (Lambda 10-B Sutter SmartShutter), and camera (Retiga 2000R FAST 1394, mono, 12 bit). Phase contrast and fluorescence images were recorded every 2 and 20 minutes, respectively.

Image quantification and analysis

Fluorescence measurements were collected by tracing cell outlines on the phase contrast images, transferring the outlines to the corresponding fluorescence frame, and quantifying density of fluorescence inside the outline. Outlines were traced manually. Blind replicate outlines, made without any awareness of cell polarity, reproduced the same results. All fluorescence images were corrected by removing outliers, subtracting background, and deconvolution to correct for diffraction scattering. The software ImageJ (NIH) was used for quantifying fluorescence densities, outlier removal, and background subtraction.  Fluorescence measurements were collected by first tracing cell outlines on the phase contrast images and the corresponding fluorescent frame was processed as following: The background of each frame was subtracted using "rolling ball" algorithm in ImageJ with ball radius 20 pixels. Noise created by heat overflow of single pixel was corrected by "remove outliers" algorithm in ImageJ with threshold intensity difference 1000 and threshold radius 0.5 pixel.  Deconvolution was accomplished by the Lucy-Richardson method in Matlab 2017b (The MathWorks, Inc., Natick, MA), (see Supplemental Material for details). Fluorescence measurements for pairs of old and new daughters were normalized by subtracting the mean of the pair’s values. To calculate elongation rates, lengths of individual bacterial cells were extracted manually from recorded time-lapse images with ImageJ. From lengths compiled over time, the elongation rate r was estimated as the slope of a linear regression of (log/length) over time. A log transformation was used because elongation rates are known to be exponential (20). All lengths were measured immediately after division and prior the next division.

Statistical tests

All comparisons were evaluated by either t-tests or randomized designs. Details of sample sizes and choices of paired, unpaired, one- and two-tailed comparisons are provided in the figure legends. Randomized designs were used when data did not conform to standard Gaussian requirements. When appropriate, values are presented as mean ± SEM (standard error of the mean).