Genome concentration limits cell growth and modulates proteome composition in Escherichia coli

Abstract

Defining the cellular factors that drive growth rate and proteome composition is essential for understanding and manipulating cellular systems. In bacteria, ribosome concentration is known to be a constraining factor of cell growth rate, while gene concentration is usually assumed not to be limiting. Here, using single-molecule tracking, quantitative single-cell microscopy, and modeling, we show that genome dilution in Escherichia coli cells arrested for DNA replication limits total RNA polymerase activity within physiological cell sizes across tested nutrient conditions. This rapid-onset limitation on bulk transcription results in sub-linear scaling of total active ribosomes with cell size and sub-exponential growth. Such downstream effects on bulk translation and cell growth are near-immediately detectable in a nutrient-rich medium, but delayed in nutrient-poor conditions, presumably due to cellular buffering activities. RNA sequencing and tandem-mass-tag mass spectrometry experiments further reveal that genome dilution remodels the relative abundance of mRNAs and proteins with cell size at a global level. Altogether, our findings indicate that chromosome concentration is a limiting factor of transcription and a global modulator of the transcriptome and proteome composition in E. coli. Experiments in Caulobacter crescentus and comparison with eukaryotic cell studies identify broadly conserved DNA concentration-dependent scaling principles of gene expression.

Keywords: Caulobacter crescentus; E. coli; cell biology; cell growth; limitation; proteome allocation; ribosomes; transcription.

Figure 1.. Growth rate and genome copy number in E. coli growing in M9glyCAAT.

(A) Illustration of 1N (CRISPR interference [CRISPRi] oriC, CJW7457) and multi-N (CRISPRi ftsZ, CJW7576) cells with different numbers of chromosomes along with representative microscopy images at different time points following CRISPRi induction. Scale bars: 1 µm. (B) Plot showing representative single-cell trajectories of cell area as a function of time for the CRISPRi strains following a block in DNA replication and/or cell division. (C) Plot showing the absolute growth rate as a function of cell area for 1N (32735 datapoints from 1568 cells) and multi-N cells (14,006 datapoints from 916 cells) in M9glyCAAT. Lines and shaded areas denote mean ± SD from three experiments. This also applies to the panels below. (D) Absolute and (E) relative growth rate in 1N (32735 datapoints from 1568 cells, CJW7457), multi-N (14,006 datapoints from 916 cells, CJW7576), and dnaC2 1N (13,933 datapoints from 1043 cells, CJW7374) cells as a function of cell area in M9glyCAAT. (F) Absolute and (G) relative growth rate in 1N (13,933 datapoints from 1043 cells), 2N (6265 datapoints from 295 cells), and >2N (2116 datapoints from 95 cells) dnaC2 (CJW7374) cells as a function of cell area in M9glyCAAT.

Figure 1—figure supplement 1.. The relative growth rate of wild-type (WT) (CJW7339) cells in M9glyCAAT grown after placing them on an agarose pad.

The plot includes 61,562 datapoints from 11,907 cells. Lines and shaded areas denote mean ± SD from three biological replicates.

Figure 1—figure supplement 2.. Characterization of ploidy in CRISPR interference (CRISPRi) oriC cells.

Representative microscopy images of CRISPRi oriC cells expressing HU-CFP and ParB-mCherry with parS site at ori1 (CJW7517) in M9glyCAAT 210 min after addition of 0.2% L-arabinose. Scale bar: 1 µm.

Figure 1—figure supplement 3.. Effects of cell area overestimation on the relative growth rate calculation.

The simulated relative growth rate for increasing cell area considering different levels of cell area overestimation (0–0.2 μm²) and a constant relative growth rate of 0.014 min^–1. In the absence of an area overestimation (0 μm²), the theoretical relative growth rate remains constant as the cell area increases, consistent with exponential growth. When a constant overestimated area is added the relative growth rate scales nonlinearly with the cell area. This nonlinear increase in the relative growth rate estimation may be confused with super-exponential growth, but it is in fact the result of cell area overestimation during cell segmentation in the phase contrast channel.

Figure 1—figure supplement 4.. Relationship between growth rate and cell area in cell types of different ploidy.

(A) Absolute and (B) relative growth rate in M9glyCAAT in 1N (32,735 datapoints from 1568 cells, CJW7457), multi-N (14,006 datapoints from 916 cells, CJW7576), and wild-type (WT) (19,495 datapoints from 7440 cells, CJW7339) cells as a function of cell area. Lines and shaded areas denote mean ± SD from three biological replicates.

Figure 1—figure supplement 5.. Validation of stable growth under microscope observation and absolute growth rate determination of ppGpp⁰ and ∆recA cells.

(A) Plot showing the absolute growth rate of 1N (CJW7457) cells imaged for 2 hr (dotted line). Prior to spotting cells on the agarose pad containing medium, cells were grown for 90 min in a liquid M9glyCAAT culture in the presence of the CRISPR interference (CRISPRi) inducer L-arabinose (0.2%) to allow for already-initiated DNA replication rounds to complete and for reductive divisions to ensue. The data for 1N (CJW7457) and multi-N (CJW7576) cells (presented in Figure 1C), which were grown on an agarose pad for 6 hr, are also shown for comparison. Lines and shaded areas denote mean ± SD from two biological replicates. (B) Plot showing the absolute growth rate of ppGpp⁰ (CJW7518) and ∆recA (CJW7522) cells. Lines and shaded areas denote mean ± SD from two biological replicates. Also shown are the data for 1N (CJW7457) and multi-N (CJW7576) cells from Figure 1C.

Figure 1—figure supplement 6.. Characterization of ploidy in dnaC2 cells.

(A) Representative microscopy images of wild-type (WT) (top) and dnaC2 (bottom) cells expressing HU-mCherry growing under microscope observation in M9glyCAAT. DNA replication in dnaC2 cells was blocked by growing cells at 37°C for 145 min. Arrows indicate dnaC2 cells with one (1N) or two (2N) nucleoids. Scale bars: 1 µm. (B) Graph showing the percentage of dnaC2 cells (CJW7374) with 1, 2, or >2 nucleoids after growth at 37°C. Shown are aggregated data from three biological replicates.

Figure 1—figure supplement 7.. Relationships between growth rate and cell volume across cell types of different ploidy in M9glyCAAT.

(A) Absolute and (B) relative growth rate based on cell volume in 1N (32,735 datapoints from 1568 cells, CJW7457), multi-N (14,006 datapoints from 916 cells, CJW7576), and dnaC2 1N (13,933 datapoints from 1043 cells, CJW7374) cells as a function of cell volume. Lines and shaded areas denote mean ± SD from three experiments. The growth rates are based on volume in this figure. (C) Absolute and (D) relative growth rate in 1N (32,735 datapoints from 1568 cells, CJW7457), multi-N (14,006 datapoints from 916 cells, CJW7576), and wild-type (WT) (19,495 datapoints from 7440 cells, CJW7339) cells as a function of cell area. (E) Absolute growth rate and (F) relative growth rate in 1N (13,933 datapoints from 1043 cells), 2N (6265 datapoints from 295 cells), and >2N (2116 datapoints from 95 cells) dnaC2 (CJW7374) cells as a function of cell area. (G) Cell width in WT (59,072 datapoints from 7640 cells, CJW7339), 1N (41,313 datapoints from 1585 cells, CJW7477), and multi-N (18,489 datapoints from 915 cells, CJW7563) cells as a function of cell area. Lines and shaded areas denote mean ± SD from three biological replicates.

Figure 1—figure supplement 8.. DNA-dependent growth in C. crescentus.

(A) Plot showing the absolute and (B) relative growth rate of 1N (DnaA depletion, strain CJW4823, 87 cells) and multi-N (FtsZ depletion, strain CJW3673, 181 cells) C. crescentus cells as a function of cell area. Lines and shaded areas denote mean ± SD from three biological replicates. (C) Example images of 1N and multi-N C. crescentus cells with ploidy represented by the number of ParB-eCFP foci. Scale bar: 2 µm.

Figure 2.. Lower ribosome activity explains the reduced growth rate of 1N cells growing in M9glyCAAT.

(A) RpsB-msfGFP fluorescence concentration in 1N (6542 cells, CJW7478) and multi-N (10,537 cells, CJW7564) cells as a function of cell area. Lines and shaded areas denote mean ± SD from three experiments. (B) Relative protein concentration of different ribosomal proteins in 1N (SJ_XTL676) and multi-N (SJ_XTL229) cells by tandem-mass-tag (TMT)-mass spectrometry (MS). 1N-rich cells were collected 0, 120, 180, 240, and 300 min after addition of 0.2% arabinose, while multi-N cells were collected after 0, 60, and 120 min of induction. Blue and cyan represent two independent experiments. Only proteins with at least four peptide measurements are plotted. (C) Apparent diffusion coefficients (D_a) of JF549-labeled RspB-HaloTag in wild-type (WT) (32,410 tracks from 771 cells, CJW7528), 1N (848,367 tracks from 2478 cells, CJW7529), and multi-N cells (107,095 tracks from 1139 cells, CJW7530). Only tracks of length ≥9 displacements are included. 1N cells are color-binned according to their cell area while multi-N cells contain aggregated data for ~2–10 µm² cell areas. (D) D_a in WT cells fitted by a three-state Gaussian mixture model (GMM): 77 ± 1%, 20 ± 1%, and 3.2 ± 0.5% (± standard error of the mean [SEM]) of the ribosome population, from the slowest moving to the fastest moving (32,410 tracks from 771 cells). (E) Example WT and 1N cells where active (red, slow-moving) and inactive (gray, fast-moving) ribosomes are classified according to the GMM. (F) Active (slow-moving) ribosome fraction in individual WT (237 cells) and 1N (2453 cells) cells as a function of cell area. Only cells with ≥50 tracks are included. Lines and shaded areas denote mean and 95% confidence interval (CI) of the mean from bootstrapping. (G) Same as (F) but for WT (237 cells) and multi-N (683 cells) cells. (H) Absolute growth rate of 1N and multi-N cells (Figure 1C) as a function of cell area was overlaid with the total active ribosome amount (calculated from A, F, and G). Lines and shaded areas denote mean and 95% CI of the mean from bootstrapping. All microscopy data are from three biological replicates. msfGFP, monomeric superfolder green fluorescent protein.

Figure 2—figure supplement 1.. Diffusive characteristics of labeled ribosomes in rifampicin-treated wild-type (WT) cells.

Plot showing the probability density of apparent diffusion coefficients (D_a) of JF549-labeled RpsB-HaloTag in WT cells (CJW7528) treated with 200 µg/mL rifampicin for 30 min. Only tracks of length ≥10 are included. Also shown is D_a fitted by three-state Gaussian mixture model (GMM) (mean ± standard error of the mean [SEM]). Data are from three biological replicates.

Figure 2—figure supplement 2.. Diffusive characteristics of labeled ribosomes in 1N cells as a function of cell area.

Plots showing the probability density of apparent diffusion coefficients (D_a) of JF549-labeled RpsB-HaloTag in 1N cells (CJW7529) in M9glyCAAT at different cell areas. Also shown is D_a fitted by three-state Gaussian mixture model (GMM). Only tracks of length ≥10 are included. Data are from three biological replicates.

Figure 3.. RNA polymerase (RNAP) activity is reduced in 1N cells growing in M9glyCAAT.

(A) RpoC-YFP fluorescence concentration in 1N (3580 cells, CJW7477) and multi-N (5554 cells, CJW7563) cells as a function of cell area. Lines and shaded areas denote mean ± SD from three experiments. (B) Relative protein concentration of core RNAP subunits and σ⁷⁰ in 1N-rich (SJ_XTL676) and multi-N (SJ_XTL229) cells by tandem-mass-tag (TMT)-mass spectrometry (MS). 1N-rich cells were collected 0, 120, 180, 240, and 300 min after addition of 0.2% L-arabinose, while multi-N cells were collected after 0, 60, and 120 min of induction. (C) Apparent diffusion coefficients of JF549-labeled RpoC-HaloTag in wild-type (WT) (91,280 tracks from 1000 cells, CJW7519), 1N (175,884 tracks from 1219 cells, CJW7520) and multi-N cells (186,951 tracks from 1040 cells, CJW7527). Only tracks of length ≥9 displacements are included. 1N cells are binned according to cell area while multi-N cells contain aggregated data for ~2–15 µm² cell areas. (D) D_a in WT cells fitted by a three-state Gaussian mixture model (GMM): 49 ± 4%, 49 ± 4%, and 2 ± 0.1% (± standard error of the mean [SEM]) of the RNAP population, from the slowest moving to the fastest moving (91,280 tracks from 1000 cells). (E) Example WT and 1N cells where active (red, slow-moving) and inactive (gray, fast-moving) RNAPs are classified according to the GMM. (F) Active RNAP fraction in individual WT (854 cells) and 1N (1024 cells) cells as a function of cell area. Only cells with at least 50 tracks are included. Lines and shaded areas denote mean ±95% CI of the mean from bootstrapping (three experiments). (G) Same as (F) but for WT (854 cells) and multi-N (924 cells) cells. (H) Total amount of active RNAP in WT, 1N, and multi-N cells as a function of cell area (calculated from A, F, and G). Also shown is a linear fit to multi-N data ($f\left(x\right)=4.16\bullet {10}^4\bullet x$, R² 0.98). Lines and shaded areas denote mean and 95% CI of the mean from bootstrapping. All microscopy data are from three biological replicates.

Figure 3—figure supplement 1.. Diffusive characteristics of labeled RNA polymerases (RNAPs) in rifampicin-treated cells.

Plot showing the probability density of apparent diffusion coefficients (D_a) of JF549-labeled RpoC-HaloTag in CJW7519 cells treated with 200 µg/mL rifampicin for 30 min. Also shown is D_a fitted by three-state Gaussian mixture model (mean ± standard error of the mean [SEM]). Only tracks of length ≥10 are included. Data are from three biological replicates.

Figure 3—figure supplement 2.. Determination of the relative Rsd concentration in 1N-rich and multi-N cells as a function of cell area.

Plots showing the relative protein concentration of Rsd in 1N-rich (SJ_XTL676) and multi-N (SJ_XTL229) cells, as determined by tandem-mass-tag (TMT)-mass spectrometry (MS). 1N-rich cells grown in M9glyCAAT were collected 0, 120, 180, 240, and 300 min, while multi-N cells were collected at 0, 60, and 120 min after 0.2% L-arabinose induction of CRISPR interference (CRISPRi). Different shades of gray represent two independent experiments.

Figure 4.. RNASelect and EUB338 concentration measurements in 1N and multi-N cells.

(A) Images of representative cells from a mixed population of 1N (CRISPR interference [CRISPRi] oriC) and multi-N (CRISPRi ftsZ) cells. Strains CJW7457 and CJW7576 carrying HU-mCherry were used for the SYTO RNASelect staining experiment, whereas DAPI-stained strains SJ_XTL676 and SJ_XTL229 were used for the EUB338 ribosomal RNA (rRNA) fluorescence in situ hybridization (FISH) experiment. (B) Concentration distribution of SYTO RNASelect (3077 cells for each population from five biological replicates) and EUB338 (1254 cells for each population from three biological replicates) in 1N and multi-N cells. (C) The average 1N/multi-N SYTO RNASelect and EUB338 concentration ratio (gray bar) calculated from five and three biological replicates (white circles), respectively. (D) RNASelect and EUB338 concentration ratios as functions of cell area (mean ± SD from five and three biological replicates, respectively). Single exponential decay functions were fitted to the average ratios (R²>97%) for each indicated reporter. All concentration comparisons or ratio calculations were performed for equal numbers of 1N and multi-N cells and overlapping cell area distributions (see Materials and methods and Figure 4—figure supplement 1).

Figure 4—figure supplement 1.. Cell sampling to match cell size distribution in mixed populations of 1N and multi-N cells.

(A) Cell area distributions from mixed CJW7457 (CRISPR interference [CRISPRi] oriC) and CJW7576 (CRISPRi ftsZ) populations stained with SYTO RNASelect (aggregated data from five biological replicates) before and after cell area sampling (cell area bins with less than 50 cells were removed from the analysis). (B) Cell area distributions from mixed SJ_XTL676 (CRISPRi oriC) and SJ_XTL229 (CRISPRi ftsZ) populations stained with the EUB338 fluorescence in situ hybridization (FISH) probe (aggregated data from three biological replicates) before and after cell area sampling (cell area bins with less than 25 cells were removed from the analysis).

Figure 4—figure supplement 2.. Comparison of RpoC-HaloTag-JF549 labeling between 1N and multi-N cells.

(A) Phase contrast (left) and RpoC-HaloTag-JF549 fluorescence (right) images of two representative cells from a mixed population of 1N (CRISPR interference [CRISPRi] oriC, CJW7520) and multi-N (CRISPRi ftsZ, CJW7527) cells. (B) Cell area distributions from mixed 1N and multi-N cell populations in which RpoC-HaloTag was stained with the JF549 dye (aggregated data from two biological replicates) before and after cell area sampling. (C) Distributions of RpoC-HaloTag-JF549 signal concentration for 1N and multi-N cells (748 cells for each population from two biological replicates). (D) Average 1N/multi-N RpoC-HaloTag-JF549 concentration ratio (gray bar), calculated from two biological replicates (white circles) after sampling the same number of cells per biological replicate and cell area bin for each population (panel B). (E) Plot showing the concentration ratio of fluorescently labeled RpoC derivatives between 1N and multi-N cells vs. the cell area. Squares show mean ± full range of RpoC-HaloTag-JF549 signal concentration ratios from two biological replicates, whereas gray circles indicate the mean signal concentration ratio of RpoC-YFP (from data shown in Figure 3A) for comparison. A linear regression (red dashed line) was fitted to the average RpoC-HaloTag-JF549 ratios. A cell area of ~2.8 μm² corresponds to a 1N/multi-N RpoC-HaloTag-JF549 concentration ratio equal to 1.

Figure 4—figure supplement 3.. EUB338 staining comparison between fast (M9glyCAAT) and slow (M9gly) growing populations.

The time of dCas9 induction was adjusted to obtain comparable cell area distribution between 1N and multi-N cells (see Materials and methods). The cell areas were then sampled to achieve a perfect match in distributions for fair comparison between the two populations. (A) Relative cell area and EUB338 concentration distributions from wild-type MG1655 exponentially growing populations in M9glyCAAT or M9gly. Data from two biological replicates (rep) are shown (M9glyCAAT rep 1: 5867 cells, M9glyCAAT rep 2: 9728 cells, M9gly rep 1: 8301 cells, M9gly rep 2: 3766 cells). The iso-contour plots (nine levels above the 25th density percentile) include data from both biological replicates per nutrient condition. (B) Representative fields of the EUB338 fluorescence in stained fixed cells that were grown in M9glyCAAT (top) or M9gly (bottom). For comparison, the two fields of view have the same dimensions, and the fluorescence is scaled the same.

Figure 5.. Mathematical modeling of DNA limitation.

(A–C) Plots comparing simulation results of model A (solid lines) with experimental data points (dots) and averages (open squares) in the M9glyCAAT condition. The multi-N and 1N cells are indicated as blue and yellow, respectively: (A) The relation between the absolute growth rate ($\frac{dA}{dt}$) and cell area ($A$). (B) The relation between the active RNA polymerase (RNAP) fraction and cell area. (C) The relation between the active ribosome fraction and cell area. (D) Diagram showing how the fractions of active RNAPs and ribosomes change with DNA concentration (colored from yellow to blue). Simulated results (filled dots) are based on model A. Experimental data (points with 2D error bars: 95% CI) from multi-N and 1N cells were combined and shown in the same plot. (E) Plot showing the effect of DNA limitation (using the ordinary differential equation [ODE] model A) on the decay of DNA concentration, mRNA concentration, and relative growth rate in 1N cells. Each quantity was normalized to their value at normal cell size (cell area = 2.5 µm²).

Figure 5—figure supplement 1.. Comparison between experimental results from the M9glyCAAT condition and simulation results using model B.

(A–C) Plots comparing simulation results of model B (solid lines) with experimental data (dots) and averages (open squares). The multi-N and 1N cells are indicated as blue and yellow, respectively: (A) The relation between the absolute growth rate ($\frac{dA}{dt}$) and cell area ($A$). (B) The relation between the active RNA polymerase (RNAP) fraction and cell area. (C) The relation between the active ribosome fraction and cell area. (D) A two-dimensional diagram showing how the fractions of active RNAPs and ribosomes change with DNA concentration (colored from yellow to blue). Experimental data (with 2D error bars: 95% CI) from multi-N and 1N cells were combined and shown in the same plot.

Figure 5—figure supplement 2.. Model A-based simulations examining the effects of varying the rates in either mRNA synthesis or mRNA degradation on the relative growth rate of 1N cells as a function of cell area.

(A) Plot showing the decay of DNA concentration (black) and of the relative growth rate (blue) in 1N cells when the rate of bulk mRNA synthesis ($r_1$) increases or decreases by 10-fold. Each quantity was normalized to its value at normal cell size (cell area = 2.5 µm²). (B) Same as (A) except mRNA degradation rate $\delta$ increasing or decreasing by 10-fold.

Figure 6.. Scaling of the total active RNA polymerases (RNAPs), total active ribosomes, and growth rate with cell area during genome dilution in nutrient-poor media.

(A) Plot showing the total amount of active RNAPs (calculated by multiplying the total amount of RNAPs by the fraction of active RNAPs from Figure 6—figure supplement 1A and G) in wild-type (WT) (CJW7339) and 1N (CJW7457) cells grown in M9gly as a function of cell area. Also shown is a linear fit to WT data ($f\left(x\right)=3.99\bullet {10}^4\bullet x$, R²=0.90). Shaded areas denote 95% CI of the mean from bootstrapping. All data are from three biological replicates. (B) Same as (A) but for cells grown in M9ala (calculated from Figure 6—figure supplement 1B and H). The linear fit for WT data is $f\left(x\right)=3.21\bullet {10}^4\bullet x$, R²=0.95. (C) Plot showing the total active ribosome amount of 1N and multi-N cells grown in M9gly as a function of cell area. The total amount of active ribosomes was calculated by multiplying the total amount of ribosomes by the fraction of active ribosomes (from Figure 6—figure supplement 2A and G). Also shown is a linear fit to WT data ($f\left(x\right)=2.99\bullet {10}^4\bullet x$, R²=0.97). Lines and shaded areas denote mean and 95% CI of the mean from bootstrapping. All data are from three biological replicates. (D) Same as (C) but for cells grown in M9ala (calculated from Figure 6—figure supplement 2B and H). Here, the linear fit to the WT data is $f\left(x\right)=1.90\bullet {10}^4\bullet x$, R²=0.99. (E) Absolute growth rate in 1N (50,352 datapoints from 973 cells) and WT (80,269 datapoints from 12,544 cells) cells in M9gly. The linear fit for WT data is $f\left(x\right)=6.50\bullet {10}^{-3}\bullet x$, R²=0.99. (F) Absolute growth rate in 1N (71,736 datapoints from 909 cells) and WT (63,367 datapoints from 6880 cells) cells in M9ala. The linear fit for WT data is $f\left(x\right)=4.05\bullet {10}^{-3}\bullet x$, R²=0.97. Lines and shaded areas denote mean ± SD from three biological replicates.

Figure 6—figure supplement 1.. Characterization of RNA polymerase (RNAP) diffusion and active fraction in poor media conditions.

(A) Plot showing the RpoC-YFP fluorescence concentration in 1N (CJW7477) cells grown in M9gly (three experiments) as a function of cell area. Lines and shaded areas denote mean ± SD between biological replicates. (B) Same as (A) but for 1N cells grown in M9ala (three experiments). (C) Plot showing the probability densities of apparent diffusion coefficients (D_a) of JF549-labeled RpoC-HaloTag in wild-type (WT) (CJW7519) and 1N (CJW7520) cells grown in M9gly. Only tracks of length ≥10 are included. 1N cells were binned according to cell area. (D) Same as (C) but for cells grown in M9ala. (E) Plot showing the probability density of D_a of JF549-labeled RpoC-HaloTag in WT cells (CJW7519) grown in M9gly. Also shown is D_a fitted by three-state Gaussian mixture model (GMM) (mean ± standard error of the mean [SEM]). (F) Same as (E) but for cells grown in M9ala. (G) Plots showing the fraction of active RNAPs in individual WT (CJW7519) and 1N (CJW7520) cells (dots) grown in M9gly as a function of cell area. Only cells with ≥50 tracks are included. (H) Same as (G) but for cells grown in M9ala.

Figure 6—figure supplement 2.. Characterization of ribosomal diffusion and active fraction in poor media conditions.

(A) Plot showing the RpsB-msfGFP fluorescence concentration in 1N (CJW7478) cells grown in M9gly (from three experiments) as a function of cell area. Lines and shaded areas denote mean ± SD between the biological replicates. (B) Same as (A) but for 1N cells grown in M9ala (from two experiments). (C) Plot showing the probability density of apparent diffusion coefficients (D_a) of JF549-labeled RpsB-HaloTag across cell areas for wild-type (WT) (CJW7528) and 1N (CJW7529) cells grown in M9gly. Only tracks of length ≥10 are included. 1N cells were binned according to cell area. (D) Same as (C) for cells grown in M9ala. (E) Plot showing probability density of D_a of JF549-labeled RpsB-HaloTag in WT cells (CJW7528) grown in M9gly. Only tracks of length ≥10 are included. Also shown is D_a fitted by three-state Gaussian mixture model (GMM) (mean ± standard error of the mean [SEM]). (F) Same as (E) but for cells grown in M9ala. (G) Plot showing the fraction of active ribosomes as a function of cell area for individual WT (CJW7528) and 1N (CJW7529) cells (dots) grown in M9gly. Only cells with ≥50 tracks are included. Shaded areas denote 95% confidence interval (CI) of the mean from bootstrapping. (H) Same as (G) but for cells grown in M9ala. msfGFP, monomeric superfolder green fluorescent protein.

Figure 7.. Proteome and transcriptome remodeling in 1N-rich cells.

(A) Schematic explaining the calculation of the protein slopes, which describes the scaling of the relative protein concentration (concentration of a given protein relative to the proteome) with cell area. (B) Plot showing the protein scaling (average slopes from two reproducible biological replicates, see Figure 7—figure supplement 1A and B) in 1N (x-axis) and multi-N (y-axis) cells across the detected proteome (2360 proteins). The colormap corresponds to a Gaussian kernel density estimation (KDE). (C) Plot showing the first principal component (PC1) used to reduce the dimensionality of the relative protein concentration during cell growth. The PC1, which represents the overall change in relative concentration regardless of the sign of the slope, explains 69% of the total variance considering both 1N-rich and multi-N cells. The x-axis corresponds to the log-transformed cell area, whereas the marker size shows the cell area increase in linear scale. (D) Correlation between average protein and RNA slopes across 2324 genes. The colormap corresponds to a KDE. (E) Relation between mRNA abundance (transcripts per million 60 min after CRISPR interference [CRISPRi] induction) and RNA slopes in 1N-rich cells. The colormap indicates a KDE (3446 genes in total). The binned data are also shown (orange markers: mean ± standard error of the mean [SEM], ~380 genes per bin). The Spearman correlation (ρ=–0.04) is considered not significant (NS, p-value>10^–10). (F) Correlation between RNA slopes and mRNA degradation rate from a published dataset (Balakrishnan et al., 2022) across genes. The colormap indicates a KDE (2570 genes with quantified slopes and positive mRNA degradation rates). The binned data are also shown (orange markers: mean ± SEM, ~280 genes per bin). A significant negative Spearman correlation (p-value<10^–10) is shown for mRNAs with a degradation rate above 0.7 min^–1. (G) RNA slope comparison between essential and non-essential genes in E. coli. Three different published sets of essential genes were used (Gerdes et al., 2003; Goodall et al., 2018; Hashimoto et al., 2005). The horizontal white lines indicate the inter-quartile range of each distribution. Mann-Whitney non-parametric tests justify the significant difference (p-value<10^–10) between the two gene groups (essential vs. non-essential genes).

Figure 7—figure supplement 1.. Comparison of protein and mRNA scaling between biological replicates.

(A) Correlation of protein slopes across the proteome (2360 proteins) between two biological replicates for 1N cells. (B) Same as (A) but for multi-N cells. (C) Correlation of RNA slopes across the genome (3446 mRNAs) of 1N cells between two biological replicates. The indicated Spearman correlation shown (ρ) is significant (p-value<10^–10).

Figure 7—figure supplement 2.. Comparison of our data with reference datasets.

(A) Comparison between the concentrations of 3446 RNAs present in both our dataset and that of Balakrishnan et al., 2022. The first time point (60 min) after CRISPR interference (CRISPRi) induction was used to determine the RNA abundance in our experiments. The RNA concentration was expressed as transcripts per million reads. The indicated Spearman correlations (ρ) are significant (p-value<10^–10). Each marker represents a single gene or protein and the colormap shows the kernel density estimation (KDE). (B) Correlations between RNA slopes (from 2577 genes, ~280 genes per bin) in 1N cells and the rate of transcription initiation in wild-type cells (Balakrishnan et al., 2022). The binned data are also shown (orange markers: mean ± standard error of the mean [SEM]). The p-value of ρ is not significant (p-value>10^–10). (C) Same as (B) but for protein slopes instead of RNA slopes (2084 genes, ~230 genes per bin). (D) Correlation between protein slopes in 1N cells and mRNA degradation rate in wild-type cells (Balakrishnan et al., 2022). The colormap corresponds to KDE (for 2078 genes with positive mRNA degradation rates). The binned data are also shown (orange markers: mean ± SEM, ~230 genes per bin). A significant negative Spearman correlation (p-value<10^–10) is shown for genes with a degradation rate above 0.7 min^–1. (E) Comparison of the mRNA degradation rates (Balakrishnan et al., 2022) between essential and non-essential genes in E. coli. Three different published sets of essential genes were used (Gerdes et al., 2003; Goodall et al., 2018; Hashimoto et al., 2005). The horizontal white lines indicate the inter-quartile range of each distribution. Mann-Whitney non-parametric tests justify the non-significant difference (p-value>10^–2) between the two gene groups (essential vs. non-essential genes).

Figure 7—figure supplement 3.. Protein slopes relative to the chromosome position of their gene in 1N-rich and multi-N cells.

(A) Relationship between the absolute distance of a gene from oriC and the slope of the protein it encodes. Data from 2268 proteins are shown (colormap: Gaussian kernel density estimation), as well as binned data in six gene distance bins (open circles, average ± standard error of the mean [SEM], ~370 proteins per bin). For the multi-N cells, the Spearman correlation was not significant (ρ=NS, p-value>0.01), whereas for the 1N-rich cells, there was a significant Spearman correlation (ρ=0.23, p-value<10^–10) for genes within 1.35 Mbps of oriC (below the mid-point of the fourth gene distance bin). (B) Relationship between the absolute distance of a gene from oriC and its mRNA slope. Data from 2200 RNAs are shown (colormap: Gaussian kernel density estimation), as well as binned data in six gene distance bins (open circles, average ± SEM, ~360 proteins per bin). The Spearman correlation is not significant (ρ=–0.02, p-value>0.1).

Figure 7—figure supplement 4.. Protein slopes relative to protein ion intensity for 1N-rich cells.

The summed ion intensity of each protein was divided by the protein sequence length and the quotient was log-transformed. The locations of selected proteins are annotated.

Appendix 1—figure 1.. Estimation of DNA concentration.

(A) DNA replication pattern for different medium conditions. (B) Extraction of the genome content, cell volume, and genome concentration along the division cycle. (C) Genome copy, cell volume, and genome concentration along the division cycle.

Appendix 2—figure 1.. Interpolation and extrapolation of parameters.

The parameters of the ODE models were calculated using the fitted formulae. The obtained values are summarized in Supplementary file 9.

Appendix 3—scheme 1.. Topology of the three RNA-polymerase states and their transition fluxes.

Appendix 3—figure 1.. Comparing changes in active RNAP fraction between ODE models A and B.

The two-dimensional colormaps show the values of active RNAP fraction (α_RNAP) under different promoter and RNAP concentrations. (A) Using the formula of model A with M9glyCAAT parameters. (B) Using the formula of model B with M9glyCAAT parameters. The asterisk and filled square symbols indicate the promoter and RNAP concentration of wild-type cells and 1N cells (with 10μm² cell size), respectively. The dashed arrow indicates the trajectory of promoter and RNAP concentration under DNA-limited growth in the model.

Appendix 4—figure 1.. Comparison between initial (Ini) and optimized (Opt) parameters.

(A) Parameters used in model A. (B) Parameters used in model B.

Author response image 1.

The mRNA production rate equivalent (mRNA abundance at the first time point after CRISPRi oriC induction multiplied by the mRNA degradation rate measured by Balakrishnan et al., 2022, PMID: 36480614, expressed in transcript counts per minute) does not correlate (Spearman correlation’s p-value = 0.24) with the RNA slope in 1N-rich cells. Data from 2570 genes are shown (grey markers, Gaussian kernel density estimation - KDE), and their binned statistics (mean +/- SEM, ~280 genes per bin, orange markers).

Author response image 2.

The relative fitness of each gene (data by Hawkins et al., 2020, PMID: 33080209, median fitness from the highest sgRNA activity bin) plotted versus the gene-specific RNA and protein slopes that we measured in 1Nrich cells after CRISPRi oriC induction. More than 260 essential genes are shown (262 RNA slopes and 270 protein slopes, grey markers), and their binned statistics (mean +/- SEM, 43-45 essential genes per bin, orange markers). The spearman correlations (ρ) with p-values above 10-3 are considered not significant (NS). In our analyses, we only considered correlations significant if they have a Spearman correlation p-value below 10-10.