Mapping temperature-sensitive mutations at a genome scale to engineer growth switches in Escherichia coli

Abstract

Temperature-sensitive (TS) mutants are a unique tool to perturb and engineer cellular systems. Here, we constructed a CRISPR library with 15,120 Escherichia coli mutants, each with a single amino acid change in one of 346 essential proteins. 1,269 of these mutants showed temperature-sensitive growth in a time-resolved competition assay. We reconstructed 94 TS mutants and measured their metabolism under growth arrest at 42°C using metabolomics. Metabolome changes were strong and mutant-specific, showing that metabolism of nongrowing E. coli is perturbation-dependent. For example, 24 TS mutants of metabolic enzymes overproduced the direct substrate metabolite due to a bottleneck in their associated pathway. A strain with TS homoserine kinase (ThrB^F267D ) produced homoserine for 24 h, and production was tunable by temperature. Finally, we used a TS subunit of DNA polymerase III (DnaX^L289Q ) to decouple growth from arginine overproduction in engineered E. coli. These results provide a strategy to identify TS mutants en masse and demonstrate their large potential to produce bacterial metabolites with nongrowing cells.

Keywords: CRISPR-Cas9 genome editing; growth-switches; metabolic valves; metabolomics; temperature-sensitive mutations.

Figure 1. CRISPR screen with 15,120 E. coli mutants identifies temperature‐sensitive mutations

1. Schematic of the CRISPR screen. 16,038 sgRNAs plus repair templates (barcodes) were designed to introduce amino acid changes in 346 essential proteins (step 1). 15,120 of the barcodes were present in the final CRISPR library (step 2). The CRISPR library was cultured at 30 and at 42°C (n = 2 replicates). Strain‐specific barcodes (sgRNA and repair template) were sequenced every 2 h to determine the composition of the library (step 3).

2. K‐means clustering of fitness scores of 8,884 strains in the CRISPR library. Time‐course data were clustered into k = 6 clusters per temperature. The fitness scores were calculated by normalizing the read counts of the barcode of each mutant to the total number of reads per sample and to the first time point. Gray curves are the moving average of the mean of two replicates. Colored lines are cluster means and shaded areas their standard deviation (blue: 4 clusters with high fitness, yellow: 1 cluster with reduced fitness, red: 1 cluster with low fitness). Dashed lines indicate a fitness score of 1.

3. Relative composition of the CRISPR library at 30 and 42°C. Blue indicates high fitness, yellow a reduced fitness, and red low fitness. The bar graph in the middle connects the 30 and 42°C data. The light blue box indicates putative TS mutants.

4. Examples of fitness score dynamics of two strains that show temperature sensitivity (MurE^W381Q and FtsQ^I74Q). Dots show data from two replicates per temperature. The lines are the moving average through the means. Blue: 30°C culture. Red: 42°C culture. Dashed lines indicate a fitness score of 1.

Source data are available online for this figure.

Figure EV1. Mutation enrichment in TS alleles and mutants with strong fitness defects

1. The bar plot shows the total number of mutants in our CRISPR library after recombination (gray), the number of putative temperature‐sensitive (TS) mutants (blue), and the number of mutants with strong fitness defects (yellow), which are strains with less than 15 reads at the start of the pooled fitness assay (t = 0 h) and strains that have a low fitness at 30°C based on cluster analysis.

2. The bar plots show the P‐values for testing an enrichment of mutations in alpha helices, beta sheets, turns, binding, and active sites among the putative TS mutants (blue bars) and strong fitness defect mutants (yellow bars). We used a one‐tailed Fisher's exact test (also see Dataset EV3) and considered conditions with P‐values < 0.05 as enriched.

3. The left bar plots show P‐values of the indicated mutations that were tested for an enrichment in the putative TS mutants (upper chart) and fitness defect mutants (lower chart) by a one‐tailed Fisher's exact test. We considered conditions with P‐values < 0.05 as enriched. The right bar plots show the %‐enrichment between the putative TS and all other mutants (upper chart) and the growth defect mutants and all other mutants (lower chart).

Figure EV2. Fitness score dynamics and growth kinetics of an allelic series of cysS TS mutants

We reconstructed a panel of 14 cysS alleles from the CRISPR library that were putative TS mutants. Each box shows a single mutation site in cysS, and each site had more than one allele available, except for the site cysS ^L387. In each box, the left charts show the fitness score dynamics of the indicated mutant from the pooled fitness assay (Fig 1A). The dots in the left figures show data from two replicates per temperature, and the lines are the moving average through the means (blue: 30°C culture, red: 42°C culture). Dashed lines indicate a fitness score of 1. The right figures in each box show the maximum specific growth rates during plate reader growth at six different temperatures (30, 34, 38, 40, 42, and 44°C). The dots in the right figures are the mean of three replicates, and the vertical lines are the standard deviation. The dashed lines were calculated by fitting an Arrhenius‐type function to the data (R ² is the coefficient of determination, also see Appendix Fig S9). Indicated P‐values were calculated with two‐sample t‐tests (two‐tailed) comparing each the mutant strain against the unedited control strain at 42°C (Dataset EV4).

Figure 2. Reconstruction of 94 TS mutants and characterization of growth–temperature relationships

1. Relative abundance of strains in the pooled sublibrary of 250 putative TS mutants. The strain library was constructed twice (replicate 1 and 2), and dots are the mean of two technical sequencing replicates. The black lines indicate the difference between the two technical replicates. The read counts of single strains were normalized to the total number of reads. Yellow dots indicate 94 TS mutants that were isolated from the pooled sublibrary.

2. Maximal specific growth rates of 94 TS mutants and a control strain at 10 different temperatures between 30 and 44°C. Dots are the mean (n = 3 biological replicates for each temperature and each of the 95 strains). Vertical lines indicate standard deviations. Black squares and the line indicate the control strain. The box‐whisker plots show the median and 25^th/75^th percentiles. Two‐sample t‐tests (two‐tailed) were performed for each mutant against the control strain (also see Dataset EV7). At 42 and 43°C, all P‐values were below the respective indicated values.

3. Examples of maximal specific growth rates of the TS mutants AccD^V120W and FolA^M92P at different temperatures (Fig EV2 shows all 94 TS mutants). Dots are the mean, and vertical black lines indicate the standard deviation (n = 3). The black line was calculated by fitting an Arrhenius‐type function to the data (also see Appendix Fig S9).

4. Growth dynamics of the MurE^W381Q strain (green) and the HisC^I336D strain (yellow) during a temperature shift from 30 to 42°C. Dots show data from two replicates, and the lines connect the mean.

Source data are available online for this figure.

Figure EV3. Maximum specific growth rates of 94 TS mutants at different temperatures

The charts show the maximum specific growth rates μ (h⁻¹) of 94 TS mutants (and a control strain) at 10 different temperatures ranging from 30 to 44°C. The growth rates were determined from growth curves in 96‐well microtiter plate cultivations (Dataset EV7). Dots show the mean from three replicates, and black vertical lines show the standard deviation. An empirical Arrhenius‐type function was fitted to the data (black lines, also see Appendix Fig S9). The strains were sorted according to their responsiveness and switching temperature, which are parameters based on the Arrhenius‐type functions. The upper dot plot shows the switching temperatures (°C) of the strains in the columns below. The dot plot on the right side shows the responsiveness (h⁻¹ K⁻¹) of the strains in the rows. The bars indicate the medians of the responsiveness values/switching temperatures.

Figure 3. Metabolome responses of 94 TS mutants are strong, mutant‐specific, and metabolism‐wide

1. Location of TS mutants in the metabolic network. Sixty‐six TS mutations affect metabolic enzymes (yellow dots). Twenty‐two TS mutations affect nonmetabolic enzymes (green dots). Seven TS mutations affect proteins without enzymatic function (gray boxes).

2. Subset of the metabolome data. Shown are substrate metabolites that increase in 24 TS mutants (blue dots). Each strip of the dot plot shows one metabolite in all 94 TS mutants (mod. z‐score normalized). Dots are means (n = 3). Lines indicate standard deviations (calculated with error propagation). Each metabolite is the substrate for at least one TS mutant enzyme, which is indicated next to each strip of the dot plot.

3. Pathway‐focused analysis of the metabolome data. Pathways with metabolite increases (mod. z‐score > 3) are highlighted in the heatmap. The color code indicates if one, two, or more than two metabolites increase in a particular pathway. Target pathways (yellow boxes) are defined as pathways that involve a TS mutant enzyme.

4. Number of pathways that respond per strain (each dot is one strain, shown are 95 strains), and number of TS mutants that led to responses in a pathway (each dot is one pathway, shown are 93 pathways). A “response” means that at least one metabolite increases in a pathway (mod. z‐score > 3). The box‐whisker plot indicates the median (yellow line), and the 25^th/75^th percentiles (gray box). Analysis is based on mean values of the metabolome data from three biological replicates.

Source data are available online for this figure.

Figure EV4. Correlation analysis of metabolomes of 94 TS mutants

1. The heatmap shows the Pearson correlation coefficients (PCC) between metabolite data of all pair‐wise combinations of 94 TS mutants and a control strain. The metabolite levels were measured by FI‐MS after cultivation of the strains in 96‐well microtiter plates at 42°C for 16 h (n = 3). Data of 325 metabolites (mod. z‐scores) were used to calculate the PCC values.

2. The dot plot shows all PCC values from (A) (given under “all combinations”), the PCC values from pairs of genes that are in the same metabolic pathway (given under “pathway internal combinations”), and from genes, whose proteins form a complex (given under “protein complex internal combinations”). The box‐whisker plot indicates the median (red line), and the 25^th and 75^th percentiles (each dot is a combination, shown are all 4,465 combinations, 272 pathway internal combinations, and 8 protein complex internal combinations). We tested for differences between the PCC values in the three groups using a Wilcoxon rank‐sum test (two‐tailed) and indicate the respective P‐values. Analysis is based on mean values of the metabolome data from three biological replicates.

Figure 4. MetA^F285W and ThrB^F267D function as metabolic valves and overproduce homoserine

1. Two TS mutants, MetA^F285W and ThrB^F267D, at the homoserine branchpoint. MetA catalyzes the first step in the methionine biosynthesis pathway in E. coli, and ThrB catalyzes the first step in the threonine biosynthesis pathway.

2. Biomass‐specific concentration of the pool of homoserine and threonine (LC–MS/MS could not separate homoserine and threonine). Shown are cultures of the MetA^F285W strain (red), the ThrB^F267D strain (blue), and the control strain (gray). The strains were grown in shaking flasks at the indicated temperatures. Dots show samples from two replicates (n = 2). Lines connect mean values. Lower charts show the optical density (OD) in the same cultures.

Source data are available online for this figure.

Figure EV5. Substrate production in TS mutants of enzymes

1. Schematic of biosynthesis pathways of chorismate, lysine, methionine, isoleucine, and arginine. Dots represent metabolites. Valve symbols indicate TS mutant enzymes.

2. The chart shows the natural logarithm of biomass data (OD) from shaking flask cultivations of TS mutants and a control strain at 42°C. Dots are data from two replicates, and the lines connect the means.

3. The charts show the concentrations (μmol/l) of indicated metabolites during the cultivations from (B). The concentrations were quantified in samples of the whole culture broth by LC–MS/MS. Dots are data from two replicates, and the lines connect the means.

Figure 5. Decoupling growth from arginine overproduction by DnaX^L289Q

1. Arginine overproduction strain with the TS mutation DnaX^L289Q for growth control. Dysregulation of the arginine pathway was achieved by deleting argR (removes transcriptional feedback) and inserting the ArgA^H15Y mutation (removes allosteric feedback). The arginine exporter ArgO was overexpressed from a plasmid.

2. OD of the engineered arginine overproduction strain during cultivations at 42°C in shaking flasks. Dots are samples from two replicates (n = 2), and the line connects the means.

3. Arginine concentration (μmol/l) during the same cultivation (shown in B). Arginine levels were quantified in the whole culture broth by LC–MS/MS and calibrated with an authentic arginine standard. Dots show samples from two replicates (n = 2). The volumetric arginine production rate q_V in the initial 6 h of cultivation was determined using a linear regression model (R ² = 0.99, SE: standard error of the slope q_V). The bar plot shows final arginine concentrations that were determined after 23 and 24 h.

Source data are available online for this figure.