Effects of Genetic Variation on the E. coli Host-Circuit Interface Stefano Cardinale, Marcin Pawel Joachimiak, Adam Paul Arkin  Cell Reports  Volume 4, Issue 2, Pages 231-237 (July 2013) DOI: 10.1016/j.celrep.2013.06.023 Copyright © 2013 The Authors Terms and Conditions

Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure 1 Designing and Modeling the Host-Circuit Interface (A) (i) Model of the interaction between a host genetic variation (Δg) with heterologous gene expression (ΔE). Global effects induced by variation of cell physiology (ΔPh), mediated by growth variables (λ, r, and κ), impact expression of all heterologous genes (pink), whereas specific effects act on an individual circuit component or property (green). (ii) Schematic of the genetic probe. Promoters (arrows) and ribosomal binding sites (rectangles) of identical sequence have same color. The degree of coding (black) and protein (red) sequence identity of the reporter genes, relative to mVenus, is also indicated. (B) The expression of reporters mVenus and mCherry (Pvenus and Pcherry) is correlated with the exception of E. coli strain HB101 (ρ = 0.97). (C) Growth is linearly related to ribosomal capacity, with strain DH1 presenting an outlying ribosome level (ρ = 0.95 without DH1). (D) mCerulean expression predicted from a linear regression with variables λ, r, and a measure of lag-phase κ (R2adj = 0.94, Extended Experimental Procedures; Equation S4). (All values relative to BW25113.) n = 4–6. See also Figures S1, S2, S3, S4, S5, and S6. Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure 2 Host-Circuit Interaction in Single-Gene Knockouts (A) Pvenus, Pcherry, and Pcerulean (marker size) are well correlated across KEIO knockout strains, with Δfis as an obvious outlier (black arrow) (ρ = 0.72–0.86). In some strains, expression of all three markers (gold) or only cherry (brown text) is significantly affected (>2 SD). (B–D) A linear regression model (Extended Experimental Procedures; Equation S5) of variables λ, r, and κ (see main text) recapitulates ∼36% of variation in Pcherry (B), ∼35% of Pvenus (C), and ∼0.33% of Pcerulean (D). Text indicates outliers based on Cook’s test. (P relative to wild-type.) n = 3–4. See also Figures S7, S8, S9, and S10. Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure 3 Characterization of Host-Circuit Interference Heatmap of Z scores for E. coli knockout and wild-type expression outliers from regression and cluster analysis (Figures 2A–2D and S11). Apart from parental BW25113 and W3110, four clusters of strains are apparent: two groups with a homogenous increase (A) or decrease (B) of the measured properties, and two with an inverse variable pattern (C and D). Tables present averaged scores for each feature within each strain cluster. See also Figures S11 and S12. Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure 4 Classification of Host-Circuit Interactions Among 94 different strains of E. coli, 34 genetic backgrounds (Δg) (∼36%) influenced a genetic probe by affecting heterologous expression (ΔE) of all (pink) or individual (green) reporter genes. In particular, in 6/94 and 5/94 strains the expression of all reporter genes either increased or decreased and correlated with a growth phenotype (ΔPh) (respectively, full arrows). In 12 strains (∼13%), a growth variable affected individual genes specifically and in the same direction (two-sided white arrows), whereas for 11 (∼12%) of the strains one or two reporter genes varied without a correlated change in physiology. Finally, 16 strains (∼17%) showed variation in growth that did not lead to significant changes in reporter expression (constructed from Z scores in Figure S11 with 1 SD cutoff). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S1 Bacterial Strain Fitness, Related to Figure 1 (A and B) Bacterial growth for individual cultures is modeled with high accuracy (A, Extended Experimental Procedures; Equation S1) and curves are compared to estimate the growth rate at mid-exponential phase and to identify slow-growing strains (B). (C and D) Control growth rates for each of the 88 KEIO strains tested are plotted against growth rates in presence of expression from plasmid (C: Log-transformed values, D: z-scores, n = 3-4). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S2 Modeling Reporter Gene Expression, Related to Figure 1 Kinetic fluorescence measures of mVenus and mCherry are fitted with a system of ODEs and the rate of protein production P estimated (Extended Experimental Procedures; Equation S2) for KEIO knockout strains glnL (yellow, R2 = 0.996) and creC (cyan, R2 = 0.966) (A and C, respectively), and narL (yellow, R2 = 0.993) and qseB (cyan, R2 = 0.998) (B and D, respectively) (n = 3-4). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S3 Single-Cell Reporter Expression, Related to Figure 1 (A) Expression of mVenus (green) and mCherry (red) in various wild-type E. coli strains measured with a flow cytometer. In agreement with kinetic results, HB101 shows higher mCherry cellular amount compared to mVenus (bracket, n = 4). (B) mCerulean expression in wild-type parent strains fitted from fluorescence kinetic data (n = 4-6). (Expression relative to the mean of all measurements). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S4 Wild-Type E. coli Growth, Related to Figure 1 Growth curves of parental E. coli strains. In each plot, for comparison, the growth curve of wild-type strain BW25113 is presented (black). Table: mean length of lag phase (in minutes). Red bracket: extended lag phase in wild-type strain HB101 (see also table) (n = 4-6). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S5 Regression of Reporter Gene Expression in Wild-Type E. coli, Related to Figure 1 Protein production (P) predicted from growth related variables λ, r and κ (Equation S4) in standard laboratory E. coli for mVenus (A) and mCherry (B), plotted against measured values (from kinetic data). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S6 Regression Residuals in Wild-Type E. coli, Related to Figure 1 Studentized residuals of the regression of reporter production rate P in Figures 1D, S5A, and S5B plotted against the respective measured P values of mVenus (A), mCherry (B) and mCerulean (C). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S7 Complementation Assays, Related to Figure 2 Growth rate (A) and mVenus expression level (B) are partially restored in qseB and cheY knockout strains upon expression (complementation) of the corresponding missing gene from a plasmid (cyan: − inducer, yellow: + inducer) (B inlet, flow cytometry data) (n = 4–6). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S8 Metabolic Load, Related to Figure 2 Relationship between metabolic load (η, see Supplemental Information for details) and growth rate (A) or average reporter production rate P (B) for 88 selected KEIO single-gene knockout strains and wild-type BW25113 (n = 3–4). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S9 Single-Cell Expression in KEIO Mutant Strains, Related to Figure 2 Expression of mVenus and mCherry (green and red, respectively) measured with a flow cytometer at mid-exponential phase of growth (n = 80000). Cellular levels of mCherry reporter in the two outlier strains Δfis and ΔpfkA are higher compared to mVenus (brackets) (values are relative to wild-type BW25113) (n = 4–6). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S10 Regression Residuals for KEIO Strains, Related to Figure 2 Studentized residuals from the regression of reporter production rate P in main Figures 2B–2D, plotted against the respective P estimates (Equation S5) for mVenus (A), mCherry (B) and mCerulean (C). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S11 Cluster Analysis, Related to Figure 3 Heat map and hierarchical clustering of all knockout and parental strains based on physiological and expression Z-scores. For clustering, Ward’s minimum variance was used on Euclidean distances. Strains that showed a significant effect in reporter expression (Figures 2A–2D) are indicated by ∗ and their clusters by ∗. Bracket: large cluster of low-expressing strains. Wt: BW25113. Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions

Figure S12 Growth Curves of Outlier KEIO Strains, Related to Figure 3 Growth curves for a subset of expression outliers (Figures 2A–2D, dark yellow). In the same plot, for comparison, growth of wild-type BW25113 is presented (black, n = 3). Red brackets show extended lag phases of growth (n = 3–4). Cell Reports 2013 4, 231-237DOI: (10.1016/j.celrep.2013.06.023) Copyright © 2013 The Authors Terms and Conditions