Dynamic regulation of metabolic flux in engineered bacteria using a pathway-independent quorum-sensing circuit

Abstract

Metabolic engineering of microorganisms to produce desirable products on an industrial scale can result in unbalanced cellular metabolic networks that reduce productivity and yield. Metabolic fluxes can be rebalanced using dynamic pathway regulation, but few broadly applicable tools are available to achieve this. We present a pathway-independent genetic control module that can be used to dynamically regulate the expression of target genes. We apply our module to identify the optimal point to redirect glycolytic flux into heterologous engineered pathways in Escherichia coli, resulting in titers of myo-inositol increased 5.5-fold and titers of glucaric acid increased from unmeasurable to >0.8 g/L, compared to the parent strains lacking dynamic flux control. Scaled-up production of these strains in benchtop bioreactors resulted in almost ten- and fivefold increases in specific titers of myo-inositol and glucaric acid, respectively. We also used our module to control flux into aromatic amino acid biosynthesis to increase titers of shikimate in E. coli from unmeasurable to >100 mg/L.

Figure 1

Characterization of a QS-circuit to dynamically modulate a target gene of interest (GOI). (a) Schematic of the engineered circuit containing a library of promoter and RBS combinations to vary EsaI expression and AHL production rate, and differentially trigger downregulation of any GOI. Stronger EsaI expression leads to earlier downregulation of the GOI. QS circuit components (esaI, esaR170V) are integrated in the genome. All promoter/RBS combinations utilized to drive esaI are described in Supplementary Table 1. (b) Representative fluorescence profiles for all strains containing GFP under control of the QS-circuit. Switching time and rate vary among strains containing different EsaI expression cassettes. GFP is carried on a medium-copy plasmid to elicit sufficient fluorescence for detection. Fluorescence profiles for all 31 strains tested are depicted, with strain numbers indicated for clearly distinguishable profiles. (c) Switching OD for all strains containing GFP under the control of the QS circuit. Switching OD was determined by ascertaining the cell density at the point of maximum fluorescence for a given strain. Constitutive GFP expression in the absence of EsaI (AG2681, green bar), basal GFP expression in the wild-type MG1655 host without the QS circuit (+ GFP, right red bar), and wildtype MG1655 alone (MG1655, left red bar) were also tested. Switching OD for all strains was plotted against predicted strength of the corresponding promoter-RBS combination driving esaI, obtained from Mutalik et al (Supplementary Table 1) (left inset). All strains with corresponding EsaI cassettes were also rank-ordered based on the measured switching OD (actual rank, early to late switching), and were compared to their corresponding place in the rank-ordered list of switching times based on predicted EsaI strengths from Mutalik et al (predicted rank, high to low expression) (right inset). The ranks predicted from switching OD correlated well with the previously predicted ranks, with less agreement at lower ranks (higher EsaI strengths, earlier switching times). Error bars indicate s.d. of triplicate cultures.

Figure 2

QS-based valve controlling Pfk-1 expression regulates cell growth and flux into central carbon metabolism. All circuit elements are genomically integrated. (a) G6P flux is split between three possible branches that include (1) glycolysis, (2) pentose phosphate pathway (PPP), and (3) MI production through a heterologous pathway. Pgi is a reversible enzyme that converts G6P to fructose-6-phosphate (F6P). Pfk-1 catalyzes an irreversible reaction to direct flux further down glycolysis, and was the gene of interest for dynamic control. Zwf is the entry-point into the PPP and is deleted from our strains, while INO1 is the first step towards heterologous formation of MI from G6P. (b) Growth profiles show that strains with higher EsaI expression grow slower. Relative expression strengths for the strains are provided in Supplementary Tables 1 and 2. IB2275 does not grow as well and shows a lower final OD₆₀₀. (c) Pfk-1 activity profiles in crude lysates indicate that weaker EsaI expression generally trends with slower decrease in activity over time. IB2275 and IB1379 showed constitutive activity levels over all time points. EsaI-containing strains displayed high Pfk-1 activities initially, but eventually decreased to below the levels in IB1379. Data are also provided in tabular format (Supplementary Table 5). Error bars denote s.d. of triplicate cultures.

Figure 3

Functionality of QS-based dynamic regulation in multiple culture media. (a) Endpoint MI titers in MOPS minimal medium with 10 g/L glucose plotted against Pfk-1 activity at a given time (14 hrs post inoculation). An optimum, which yields a 20% titer boost over IB1379, appears that correlates to the most suitable EsaI expression strength (L19S). Corresponding Pfk-1 activity profiles are provided in Supplementary Figure 3. (b) MI titers from production trials in MOPS minimal medium with 10 g/L glucose and 0.2% casamino acids. Maximum titer was higher than in medium without casamino acids, with a maximum titer boost of 83% from the best strain (L19S) over IB1379. Pfk-1 activity profiles are provided in Supplementary Figure 4A. (c) Endpoint MI and acetate titers in T12 medium containing 10 g/L glucose. Strains that dynamically downregulated Pfk-1 faster led to no acetate accumulation and resulted in higher titers (L24S and L19S) with neutral culture pH. Strains with no dynamic downregulation (IB2275 and IB1379) or a very slow decline in Pfk-1 activity (L31S) led to high acetate production, low MI titers and low pH at the end of the batch culture. Pfk-1 activity profiles are provided in Supplementary Figure 4B. *p < 0.05 relative to titers from IB1379 in a two-tailed t-test. Error bars denote s.d. of triplicate cultures.

Figure 4

Glucaric acid production using the QS valve at the G6P branchpoint. (a) Schematic of the glucaric acid production pathway from the G6P branchpoint. MI is converted to glucuronic acid by MIOX, which is then converted to glucaric acid by Udh. uxaC and gudD encode native E. coli enzymes that metabolize glucuronic and glucaric acid, respectively, and were deleted to prevent unwanted metabolite consumption. The first step of the pentose phosphate pathway (PPP), zwf, was also deleted. (b) Glucaric acid and acetate production in strains with varying EsaI expression cassettes. The best producer of glucaric acid is L19GA, which had previously shown the best MI production. With slower (L31GA) or no (IB1379GA) Pfk-1 downregulation, high acetate but no glucaric acid was detected, as seen previously with MI production in T12 medium. Error bars denote s.d. of triplicate cultures.

Figure 5

Shikimate production through the aromatic amino acid (AAA) biosynthesis pathway using the QS valve. (a) Shikimate is an intermediate in the AAA biosynthesis pathway, whose accumulation competes with production of the AAA. To enable accumulation of shikimate without supplementing with costly AAA, AroK was dynamically downregulated using the QS valve. Native copies of aroK and aroL, the shikimate kinase isozymes, were deleted. (b) Shikimate titers in strains with various EsaI expression cassettes. With constitutive AroK expression, as in wildtype MG1655 or in a QS strain lacking EsaI (AG2310), there is no accumulation of shikimate. With differential EsaI expression, an optimal strain emerges that maximizes the production of shikimate to >100 mg/L. Error bars denote s.d. of triplicate cultures.