Reverse engineering of fatty acid-tolerant Escherichia coli identifies design strategies for robust microbial cell factories

Abstract

Adaptive laboratory evolution is often used to improve the performance of microbial cell factories. Reverse engineering of evolved strains enables learning and subsequent incorporation of novel design strategies via the design-build-test-learn cycle. Here, we reverse engineer a strain of Escherichia coli previously evolved for increased tolerance of octanoic acid (C8), an attractive biorenewable chemical, resulting in increased C8 production, increased butanol tolerance, and altered membrane properties. Here, evolution was determined to have occurred first through the restoration of WaaG activity, involved in the production of lipopolysaccharides, then an amino acid change in RpoC, a subunit of RNA polymerase, and finally mutation of the BasS-BasR two component system. All three mutations were required in order to reproduce the increased growth rate in the presence of 20 mM C8 and increased cell surface hydrophobicity; the WaaG and RpoC mutations both contributed to increased C8 titers, with the RpoC mutation appearing to be the major driver of this effect. Each of these mutations contributed to changes in the cell membrane. Increased membrane integrity and rigidity and decreased abundance of extracellular polymeric substances can be attributed to the restoration of WaaG. The increase in average lipid tail length can be attributed to the RpoC^H419P mutation, which also confers tolerance to other industrially-relevant inhibitors, such as furfural, vanillin and n-butanol. The RpoC^H419P mutation may impact binding or function of the stringent response alarmone ppGpp to RpoC site 1. Each of these mutations provides novel strategies for engineering microbial robustness, particularly at the level of the microbial cell membrane.

1. Introduction

Bioproduction of fuels and chemicals at the yields, rates and titers needed for economic viability is often impacted by toxicity of the product molecule to the microbial biocatalyst (Atsumi et al., 2010; Dunlop, 2011; Van Dien, 2013). One strategy for addressing this problem is to modify the production organism so that its sensitivity to the product molecule is decreased. Strategies for this modification commonly include rational strain engineering, often guided by -omics analysis (Foo et al., 2014; Jarboe et al., 2018; Jarboe et al., 2011; Lennen et al., 2011; Sandoval and Papoutsakis, 2016), adaptive evolution in the presence of the inhibitor (Chueca et al., 2018; Jin et al., 2016; Reyes et al., 2012; Royce et al., 2015), or screening of expression libraries (Sandoval et al., 2011; Zhang et al., 2012).

Rational strain development through the design-build-test-learn iterative cycle is effective, but requires a thorough understanding of the function of all of the relevant biological parts (Guan et al., 2016; Jarboe, 2018). Alternatively, the use of natural selection is not constrained by the existing body of knowledge. A variety of -omics tools can be used in the identification of mutations and reverse engineering of evolved strains, including whole-genome sequencing, transcriptome analysis, and fluxome analysis (Atsumi et al., 2010; Chueca et al., 2018; Foo et al., 2014). However, regardless of how mutations are identified, it is important that evolved strains displaying a desirable phenotype be subjected to reverse engineering so that these clever evolutionary strategies can be incorporated into the design of other strains. Ideally, this reverse engineering goes beyond identification of the mutation and confirmation that it contributes to the phenotype, and extends to understanding of how the mutation supports the evolved phenotype. As the cost of genome sequencing has decreased, the relative focus on identification of mutations and characterization of these mutations has shifted.

Here, we describe the reverse engineering of E. coli previously evolved for increased tolerance of octanoic acid (C8) in minimal medium, where this increased tolerance was associated with a five-fold increase in fatty acid production titers (Royce et al., 2015). Fatty acids are an attractive group of biorenewable chemicals with a large and increasing market and a wide range of applications (Desbois and Smith, 2010; Korstanje et al., 2015; Lopez-Ruiz and Davis, 2014; Tee et al., 2014). They also play a role in microbial pathogenesis (Nguyen et al., 2016). Short- and medium-chain fatty acids are well-characterized in terms of their damaging effects on the microbial cell membrane (Jarboe et al., 2018; Lennen et al., 2011; Liu et al., 2013; Sherkhanov et al., 2014). Therefore, design strategies that are learned from strains evolved for increased tolerance of these fatty acids may be applicable to engineering tolerance of other membrane-damaging compounds, such as n-butanol (Fletcher et al., 2016; Reyes et al., 2012). Here, we characterize the impact of mutations acquired during evolution for their impact on increased fatty acid production and alterations in properties of the microbial cell membrane. Each of the mutations acquired by the evolved strains were found to contribute to the evolved phenotype. A mutation within RNA polymerase was also found to increase tolerance to other membrane-damaging bio-products.

2. Materials and Methods

Full materials and methods are provided with the Supplemental Data and are briefly summarized here.

2.1. Whole-genome sequencing and verification of mutations

The Illumina Genome Analyzer II platform for high throughput sequencing was used for whole genome sequencing at the Iowa State University DNA facility, using software and algorithms as previously described to identify mutations (Royce et al., 2013a). All mutations were verified by Sanger sequencing.

2.2. Assessment of inhibitor tolerance and nutrient downshift

Overnight seed cultures were inoculated into 250 mL baffled flasks with 25 mL MOPS with 2.0 wt% dextrose and the relevant inhibitor. Unless stated otherwise, growth was performed at 37°C and 200 rpm and the media pH was adjusted to 7.00±0.05. Other inhibitors were added to the following concentrations: 10 mM hexanoic acid (C6); 600 mM NaCl; 65.6 mM levulinic acid; 200 mM citrate; 54.3 mM sodium formate; 11.9 mM hydroxybenzoate; 3.6 mM transferulic acid; 0.6% v/v n-butanol; 0.6% v/v iso-butanol; 2% v/v ethanol; 200 mM succinate; 6.6 mM vanillin; 10.4 mM furfural; 9.3% w/v glucose. Nutritional downshift was performed as previously described (Ross et al., 2013). Briefly, cells were grown to OD 0.6 – 0.8 in LB, washed in MOPS minimal medium, and resuspended in either fresh LB or MOPS minimal medium.

2.3. Fatty acid production

Strains transformed with the pJMY-EEI82564 plasmid encoding the TE10 thioesterase were grown on LB plates with ampicillin and incubated at 30°C overnight. Individual colonies were cultured in 250 mL flasks in 10 mL LB with ampicillin at 30°C on a rotary shaker at 250 rpm overnight. Seed cultures were inoculated at an approximate OD₅₅₀ of 0.1 into 250 mL baffled flasks containing 50 mL of LB with 1.5 wt% dextrose, ampicillin, and 1.0 mM isopropyl-β-D-thiogalactopyranoside (IPTG). The flasks were incubated in a rotary shaker at 250 rpm and 30°C.

Fatty acids were extracted and further derivatized from samples containing both media and cells. The fatty acid methyl esters (FAMEs) were measured with an Agilent 6890 Gas Chromatograph coupled to an Agilent 5973 Mass Spectrometer (GC-MS) at the ISU W.M. Keck Metabolomics Research Laboratory.

2.4. Extracellular polymeric substance (EPS) extraction and quantification

The total extracellular protein and polysaccharide were determined as previously described (Liang et al., 2016). Briefly, cells were grown on LB agar plates overnight, suspended in 0.85 wt% NaCl solution and quantified. The cell suspension was centrifuged at 16,300×g, at 4°C for 30 min, the supernatant was passed through a 0.45 μ filter and 90 mL of ice-cold 100% ethanol was added. The mixture was incubated at −20°C for 24 h. Then, the EPS pellet was harvested by centrifuging at 16,300×g for 30 min at 4°C, drying at room temperature and resuspension in 20 mL DI water.

2.5. Membrane characterization

Cells were grown to mig-log phase (OD₅₅₀≈1) in 25 mL of MOPS 2.0 wt% dextrose in 250 mL flasks with shaking at 37°C and 250 rpm and harvested by centrifugation at 4,500×g and room temperature for 10 min. Cells were washed twice with PBS pH 7.00±0.05 and resuspended to OD₅₅₀~1 in PBS containing 10 mM octanoic acid at pH 7.0 and then incubated at 37°C for 1 hour. For characterization of membrane permeability, the cell suspension was stained with SYTOX Green (Invitrogen, Carlsbad, CA) (Roth et al., 1997). Measurement of DPH polarization used 1,6-diphenyl-1, 3, 5-hexatirene (DPH, Life Technologies, Carlsbad, CA, USA) (Royce et al., 2013b). Measurement of cell surface hydrophobicity was performed as previously described (Rosenberg et al., 1980).

For measurement of membrane lipid composition, cells were grown to mid-log, harvested and resuspended in MOPS 2.0 wt% dextrose with or without 30 mM octanoic acid at pH 7.0, and incubated for 3 hours at 37°C. Cells were harvested and processed to recover the fatty acids, which were then derivatized and measured by GC-MS.

3. Results

3.1. Identification and timing of mutations

Evolved strains LAR1 and LAR2 were previously isolated as distinct single colonies from the same liquid culture after seventeen serial dilutions in the presence of exogenous C8 at neutral pH in defined growth medium (Royce et al., 2015). Parent strain ML115 was previously engineered from the K-12 strain MG1655 to inactivate fatty acid beta-oxidation and acetate production via deletion offadD, poxB and ackA-pta (Li et al., 2012). Mutations in LAR1 and LAR2 relative to ML115 were determined via Illumina sequencing, alignment to MG1655 as the reference genome, and verified by Sanger sequencing (Table 1). Alignment of parent strain ML115 to the reference genome revealed the presence of a 768-bp insertion sequence within lipopolysaccharide (LPS) glucosyltransferase I (WaaG). As discussed below, it seems that this mutation was implemented unintentionally during the development of ML115. LAR1 and LAR2 both have restored function of WaaG and a single amino acid change within the β’ subunit of RNA polymerase RpoC, and each has a unique mutation in the BasS-BasR two-component signal transduction system. The shared waaG and rpoC mutations are most likely due to the fact that these strains share a common ancestor.

Table 1:

Identification and timing of mutations acquired during evolution for C8 tolerance in minimal media at pH 7.0. Evolution of ML115 to strains LAR1 and LAR2 was previously described (Royce et al., 2015) over the course of 17 sequential transfers, with LAR1 and LAR2 both isolated as single colonies from the same final liquid culture. Timing of mutations was determined by PCR and restriction analysis of samples archived during the sequential transfers.

Gene	Gene Mutation	Strain	Protein Mutation	Polypeptide/Enzyme	Timing of Mutation
waaG	768 bp IS removed	LAR1 LAR2	restoration of MG1655 sequence	LPS glucosyltransferase I	Between transfers #2 and #3
rpoC	A1256C	LAR1 LAR2	H419P, in very close proximity to ppGpp binding site 1, a part of the stringent response mechanism	RNA polymerase subunit β’	Complete by transfer #13
basR	G82T	LAR1	D28Y, within response regulator receiver domain	DNA-binding transcriptional dual regulator BasR	After transfer #15
basS	27 bp deletion	LAR2	Deletion of amino acids 285 – 293, within histidine kinase domain	sensory histidine kinase BasS	After transfer #15

The order and timing of these mutations were determined by PCR and restriction digestion of samples taken periodically during the sequential transfers (Figure S1). The restoration of waaG occurred first, and relatively quickly, with only the ML115 version of waaG being observed at the end of the second transfer and only the restored version of waaG (waaG^R) being observed at the end of the third transfer. During these transfers, C8 was being supplied at a concentration of 10 mM. By the thirteenth transfer, the concentration of exogenous C8 had been increased to 30 mM, and the single base pair mutation in rpoC was present in all cells. The mutations in basS and basR in strains LAR2 and LAR1, respectively, occurred after the fifteenth transfer, at which time 30 mM C8 was still being used as the selective pressure.

WaaG adds the first glucose of the outer core of LPS (Yethon et al., 2000). The mutation acquired by the evolved strains within waaG resulted in a restoration of the wild-type MG1655 sequence. The insertion sequence in waaG in ML115 not only abolished WaaG function, but also likely altered the expression of downstream genes waaPSBOJ. The deletion of waaG has previously been reported to result in a truncated LPS core and loss of flagella and pili (Parker et al., 1992) as well as altered cell surface hydrophobicity, outer membrane permeability, and biofilm formation (Wang et al., 2015).

The single base pair mutation from adenine to cystosine at position 1256 in rpoC results in the amino acid substitution from proline to histidine at position 419 in the RpoC protein. This mutation was reported, though not characterized, in our previous publication (Royce et al., 2015) and was also detected in strains evolved for tolerance of octanoic acid and glutaric acid (Lennen et al., 2018; Lennen et al., 2019). The close proximity of the H419P substitution to ppGpp binding site 1 in RNA polymerase (Ross et al., 2016) (Figure 1) is striking, and the effect of a proline substitution at position 419 could be expected to alter the conformation and function of this binding site. The stringent response alarmone ppGpp binds to two sites in E. coli RNA polymerase (site 1 and site 2) and alters transcription from a large number of promoters during the stringent response to stress conditions. Site 1 has been previously characterized biochemically, genetically, and structurally (Ross et al., 2016; Ross et al., 2013; Zuo et al., 2013), and contributes to the stringent response (Ross et al., 2016; Sanchez-Vazquez et al., 2019). Residues in both RpoC (including R417) and the adjacent RpoZ (omega) subunit participate in ppGpp binding to site 1 and are required for its effects on transcription. These residues span the junction between two mobile modules of RNA polymerase (core and shelf), and ppGpp binding has been proposed to restrict the relative motion of the modules, thereby affecting transcription (Ross et al., 2013; Zuo et al., 2013).

Figure 1.

Views of a section of the crystal structure of E. coli RNA polymerase showing location of RpoC H419 adjacent to binding site 1 for ppGpp (adapted in PyMOL from PDB 4JKR (Zuo et al., 2013)). (A) RpoC H419 (yellow spheres) is adjacent to ppGpp binding site 1. ppGpp: red spheres. Portions of the RNAP subunits RpoC (β’, light pink); RpoB (β, light blue), RpoZ (ω, teal), and RpoA (α, grey) are shown in cartoon form. Residues shown biochemically and genetically to be required for ppGpp function at site 1 (Ross et al., 2013) are shown as blue spheres (RpoC R417, K615, R362, D622, Y626), and teal spheres (RpoZ A2, R3, V4). (B) RpoC H419 and ppGpp binding site 1 are located 30A from the RNAP active site. View is rotated from that in (A) to show the active site Mg²⁺ (magenta sphere). Other colors are as for (A).

The two evolved strains contain different mutations within the BasS-BasR two-component regulatory system. LAR1 has a single amino acid change, from aspartic acid to tyrosine, within the response regulator receiver domain of BasR. LAR2 has an in-frame deletion of nine amino acids from the histidine kinase domain of BasS. The BasS-BasR two-component system senses and responds to changes in environmental conditions related to metals (Ogasawara et al., 2012).

3.2. All three mutations are required for LAR1-level C8 tolerance

It has been previously demonstrated that evolved strain LAR1 has significantly increased tolerance to exogenously supplied fatty acids relative to its parent strain, ML115 (Royce et al., 2015). To determine the contribution of each mutation to C8 tolerance, we systematically reconstructed the LAR1 mutations in parent strain ML115 and investigated the basS mutation from LAR2 (strains YC001-011, Table 2). When the mutations were implemented in the order in which they occurred, sequentia restoration of the evolved phenotype was observed.

Table 2:

Restoration of waaG and mutation of rpoC and basR in parent strain ML115 reproduces the phenotype of evolved strain LAR1 during challenge with exogenous C8. Cells were grown in minimal media at 37°C with 1.5 wt% dextrose with an initial pH of 7.0. Growth measurements were taken hourly. Shading from black to red indicates the degree of reconstitution of the evolved strain genotype. Values are the average of three replicates with the associated standard deviation.

Strain	Gene				0 mM C8		10 mM C8		20 mM C8
	waaGR	rpoCH419P	basR*	basS*	Specific Growth Rate (/hr)	OD550 at 24 hr	Specific Growth Rate (/hr)	OD550 at 24 hr	Specific Growth Rate (/hr)	OD550 at 24 hr
LAR1	•	•	•		0.58±0.00b	3.01±0.04	0.58±0.00b	2.20±0.07b	0.57+0.00	1.75+0.03
ML115					0.60±0.01a	3.17±0.07	0.15±0.00a	0.29±0.01a
YC001	•				0.53±0.01a,b	2.9±0.2	0.39±0.01a,b	1.57±0.03a,b
YC005	•	•			0.57±0.01a	3.1±0.1	0.55±0.02b	2.20±0.02b	0.51+0.00a	1.70+0.07
YC010	•	•	•		0.57±0.00	2.98±0.04	0.57±0.01b	2.13±0.01b	0.56±0.01	1.74±0.02
YC002		•			0.65±0.01a,b	3.00±0.02	0.23±0.01a,b	0.38±0.02a,b
YC003			•		0.60±0.00a	2.61±0.02a,b	0.15±0.00a	0.28±0.01a
YC004				•	0.60±0.01	2.74±0.02a,b	0.16±0.00a	0.31±0.01a
YC006	•		•		0.45±0.01a,b	3.0±0.1	0.35±0.01a,b	1.57±0.07a,b
YC007	•			•	0.52±0.01a,b	3.0±0.1	0.35±0.01a,b	1.30±0.08a,b
YC008		•	•		0.67±0.00a,b	2.88±0.05	0.16±0.00a	0.17±0.02a,b
YC009			•	•	0.54±0.00a,b	2.34±0.09a,b	0.20±0.01a,b	0.21±0.01a,b
YC011	•		•	•	0.47±0.01a,b	2.98±0.04	0.39±0.00a,b	1.53±0.03a,b

“•” indicates that gene is present in the evolved form.

^ap ≤ 0.0038 relative to LAR1,

^bp < 0.0038 relative to ML115

In the presence of 10 mM exogenous C8, the parent strain showed a four-fold lower specific growth rate (p ≤ 0.0038) and 10-fold lower 24 hr OD relative to LAR1. Restoration of WaaG (waaG^R) (strain YC001) more than doubled the specific growth rate and final OD, though both metrics were still significantly lower than LAR1. Replacement of rpoC with rpoC^H419P in the parent strain with restored WaaG (strain YC005) resulted in a growth rate and final OD that were statistically indistinguishable from LAR1. Thus, this characterization in the presence of 10 mM C8 gives the initial impression that the evolved strain phenotype can be completely attributed to waaG^R and rpoC^H419P and that the basR mutation is not required. However, further characterization showed that all three mutations (strain YC010) were required for reproduction of the evolved strain growth rate in the presence of 20 mM C8. Specifically, the parent strain with only waaG^R and rpoC^H419P, but still encoding the wild-type basR (YC005), had a significantly lower growth rate relative to LAR1. Upon replacement of the genomic wild-type basR with basR* to generate strain YC010, the specific growth rate in the presence of 20 mM exogenous C8 was statistically indistinguishable from the evolved strain.

Characterization of the individual mutations in the presence of 10 mM C8 also provided insight into their role in the evolved phenotype. As described above, restoration of WaaG resulted in an increase in specific growth rate and final OD relative to ML115, but still significantly lower than LAR1. Implementation of only the rpoC^H419P mutation (strain YC003) did significantly increase the specific growth rate relative to ML115 in the absence of C8, but did not impact the growth rate during C8 challenge or the 24-hr OD in either condition. Implementation of only the basR* (YC003) or basS* (YC004) mutation resulted in no significant difference relative to parent strain ML115 in the presence of 10 mM C8.

Implementation of these mutations also resulted in significant changes in specific growth rate and 24-hr OD₅₅₀ even in the absence of C8 (Table 2). Specifically, implementation of rpoC^H419P in ML115 (strain YC002) significantly increased the growth rate relative to both ML115 and LAR1. Also, implementation of waaG^R and basS* or basR* without also conferring the rpoC^H419P mutation (strains YC006 and YC007) resulted in a decrease in the specific growth rate relative to both ML115 and LAR1.

These results demonstrate that the order of combination of mutations is important for assessing their contribution to the evolved phenotype and that each of the three general mutations acquired during evolution contribute to the evolved phenotype.

3.3. WaaG^R and RpoC^H419P are sufficient for LAR1-level fatty acid production

The characterization of growth in the presence of exogenous C8 demonstrates that restoration of WaaG function, the single amino acid change in RpoC^H419P, and alteration of the BasS-BasR two-component regulatory system all contribute to the increased C8 tolerance of LAR1. However, since the goal of increasing C8 tolerance is to increase fatty acid production, we also assessed the impact on fatty acid production in rich media (Figure 2A). Fatty acid production was enabled via the expression of the Anaerococcus tetradius thioesterase (TE10), which primarily produces octanoic acid (Jing et al., 2011).

Figure 2.

Restoration of waaG (waaG^R) and mutation of rpoC (rpoC^H419P) reproduce the increased fatty acid titer of evolved strain LAR1. All strains contain plasmid pJMY-EEI82564 encoding the A. tetradius thioesterase (TE10). Strains were grown in LB with 1.5 wt% dextrose at 30°C, 250 rpm with 100 mg/L ampicillin and 1.0 mM IPTG. (A) Fatty acid titer after 72 hours. (B) Strain growth during fatty acid production. Values are the average of three biological replicates, with error bars indicating one standard deviation. A titer of 1 g/L corresponds to approximately 7 mM.

While restoration of WaaG^R increased fatty acid titers, a much more dramatic increase was observed when combined with the RpoC^H419P mutation. Specifically, while the parent strain only produced 80 mg/L of fatty acids over 72 hrs, restoration of WaaG^R increased that value more than 2-fold to 188 mg/L (p = 0.004), and subsequent replacement of the wild-type RpoC with RpoC^H419P (YC005) further increased the titer to 780 mg/L (approximately 5 mM), a nearly 10-fold increase relative to the parent, comparable to the 783 mg/L produced by evolved strain LAR1. Consistently, the addition of BasR* to ML115+waaG^R+rpoC^H419P (YC010) did not further increase the production titers (data not shown). The lack of impact observed for the BasR* mutation is similar to the growth rate characterization (Table 2), in that the effect of BasR* was apparent only in the presence of 20 mM exogenous C8, but not 10 mM C8. The dramatic increase in fatty acid titer for strain YC005 relative to YC001 demonstrates the impact of the rpoC^H419P mutation on fatty acid production.

These differences in fatty acid titer cannot be solely attributed to growth of the production strain. For example, restoration of waaO^R resulted in a 4-fold increase in OD₅₅₀ during fatty acid production relative to the parent, but the fatty acid titers only increased by slightly more than 2-fold (Figure 2B). In contrast, there was no significant difference in the OD of the ML115+waaG^R and ML115+waaG^R+rpoC^H419P strains over the course of fatty acid production, despite a 4-fold difference in fatty acid titers.

3.4. Each mutation contributes to membrane changes

The cell membrane plays a vital role in microbial tolerance, particularly in the production of biorenewable fuels and chemicals (Jarboe et al., 2018; Lennen and Pfleger, 2013; Luo et al., 2009; Qi et al., 2019; Sherkhanov et al., 2014; Tan et al., 2017; Tan et al., 2016). It is also known that the cell membrane is vulnerable to damage by short- and medium-chain fatty acids (Lennen et al., 2011; Royce et al., 2013b; Royce et al., 2015) and other appealing bio-products (Lian et al., 2016). Our previous characterization of LAR1 and ML115 demonstrated alteration of the cell membrane in terms of integrity, fluidity, and lipid tail distribution (Royce et al., 2015). Thus, in addition to identifying which mutations contribute to increased C8 tolerance and increased fatty acid production, here we also assessed their contribution to these altered membrane properties (Figure 3).

Figure 3.

Each of the mutations contributes to changes in the cell membrane. Cells were assessed after challenge with exogenous C8 at pH 7.0 and 37°C. Data for the waaG^R strain is shown twice to support comparison of strains. (A) Membrane integrity was assessed via permeability to the SYTOX nucleic acid dye. (B) Membrane rigidity was characterized via DPH polarization. (C) Average length of the membrane lipid tails. (D) Cell surface hydrophobicity. Membrane permeability, rigidity and hydrophobicity were assessed after challenge with 10 mM C8. Average lipid length was assessed after challenge with 30 mM C8.

Evolved strain LAR1 showed drastically increased membrane integrity during exogenous C8 challenge relative to parent strain ML115, as evidenced by a decrease in permeability to the SYTOX nucleic acid dye (Figure 3A), as previously reported (Royce et al., 2015). Characterization of single and combined mutants demonstrates that restoration of WaaG, the first mutation acquired during adaptive laboratory evolution, is responsible for the increased membrane integrity of LAR1. This is consistent with previous reports that deletion of waaG in E. coli decreased outer membrane integrity (Wang et al., 2015). Implementation of only the rpoC^H419P mutation in ML115 (YC002) did not increase membrane integrity, but implementation of only the basS* or basR* mutations did, though not to the level observed for only waaO^R. Thus, the first mutation that occurred during evolution of LAR1 corrected the problematic loss of membrane integrity in the presence of exogenous C8. In strains YC003 and YC004, mutation of the BasS-BasR system also impacts membrane integrity, but this effect is only observed in the absence of a functional WaaG.

For appropriate function, the membrane should be neither too fluid nor too rigid. It has been previously demonstrated that exogenous C8 increases membrane fluidity (Royce et al., 2013b) and that engineered strains with increased membrane rigidity have an increase in C8 tolerance and production (Tan et al., 2016). Previous characterization of LAR1 showed significantly lower membrane fluidity, and thus higher membrane rigidity, than the parent strain (Royce et al., 2015), as evidenced by higher 1,6-diphenyl-hexa-1,3,5-triene (DPH) polarization values. Here, characterization of single and combined implementation of our mutations showed that, as with the alteration of membrane permeability, the restoration of WaaG functionality is sufficient to account for the difference in ML115 and LAR1 rigidity (Figure 3B).

Changes in the relative distribution of the various membrane lipids in stressful conditions have been widely reported (Liu et al., 2013; Royce et al., 2013b; Venkataramanan et al., 2014) and targeted changes to this distribution have been found to be effective in improving tolerance and sometimes improving production (Jarboe et al., 2018; Lennen and Pfleger, 2013; Luo et al., 2009; Sandoval and Papoutsakis, 2016; Sherkhanov et al., 2014). We have previously described the altered membrane lipid distribution of LAR1 relative to ML115, with the conclusion that the average lipid length was consistently higher in LAR1 across a range of conditions (Royce et al., 2015). Thus, the membrane lipid distribution for the various strains characterized here is presented in terms of average lipid length. Characterization of the single and combined mutants showed that restoration of functional WaaG (YC001) contributed to, but did not fully account for, the increase in average lipid length. However, expression of RpoC^H419P in parent strain ML115, either as the only implemented mutation (YC002) or in conjunction with waaG^R (YC005) or with waaG^R and basR* (YC010), fully accounted for the increase in average lipid length (Figure 3C).

A loss of membrane integrity and perturbation of the membrane fluidity indicate problems with the membrane function. Contrastingly, cell surface hydrophobicity can range widely without any apparent detrimental impact on cell health (Liang et al., 2016). While the membrane composition, in terms of phospholipid heads, lipid tails and proteins, is a substantial driver of membrane integrity and fluidity, hydrophobicity is influenced by various other proteins and sugars (Liao et al., 2015). We have previously observed that increased cell surface hydrophobicity is associated with increased fatty acid production by E. coli (Chen et al., 2018). Here we report that evolved strain LAR1 also differs from parent strain ML115 in that is has a substantially larger cell surface hydrophobicity (Figure 3D).

Reproduction of this increase in cell surface hydrophobicity requires the combined implementation of waaG^R, rpoC^H419P and basR* (YC010). When only the waaG^R mutation was expressed in ML115, there was no change in hydrophobicity (Figure 3D). Expression of only rpoC^H419P, basR* or basS* in ML115 also did not reproduce the evolved strain value, but combination of waaG^R and rpoC^H419P (YC005) resulted in a hydrophobicity value higher than the value observed for any of the single mutants. The presence of all three mutations, waaG^R, rpoC^H419P and basR*, reproduced the evolved strain value.

These results demonstrate that each of the mutations identified in LAR1 contributes to at least one of the alterations in membrane properties.

3.5. Restoration of WaaG^R dramatically impacts EPS sugar production

Since waaG encodes lipopolysaccharide (LPS) glucosyltransferase I, which adds the first glucose of the outer core of LPS (Yethon et al., 2000), we sought to determine the overall effect on extracellular polymeric substances (EPS). Restoration of waaG (waaG^R) decreased the production of the two major EPS, polysaccharides and proteins. Parent strain ML115, which encodes the disrupted form of waaG, produced approximately 2.8 μg EPS polysaccharides per 10⁸ cells (Figure 4A), which is nearly an order of magnitude higher than the approximately 0.3 μg per 10⁸ cells previously observed for a set of 77 environmental E. coli isolates (Liang et al., 2016). Restoration of WaaG via gene replacement with waaG^R in ML115 (YC001) resulted in a more than 10-fold decrease in EPS sugar production (Figure 4A).

Figure 4:

Restoration of WaaG affects (A) production of extracellular polysaccharides, (B) colony morphology, and (C) production of flagella.

*indicates a significant difference (p ≤ 0.01) from LAR1

It is expected that parent strain ML115, encoding only the disrupted form of waaG, should only be able to produce the inner core of LPS while strains encoding the restored waaG^R gene should produce complete LPS (Ren et al., 2016). The colony morphology of ML115, LAR1, and ML115+waaG^R (YC001) clearly differ (Figure 4B). Previous characterization of a waaGPBI deletion mutant described a mucoid colony morphology (Parker et al., 1992), consistent with our observations for ML115 (Figure 4B), but not for LAR1 and ML115+waaG^R (Figure 4B). The deletion of waaG has previously been reported to result in a truncated LPS core and loss of flagella (Parker et al., 1992). This is consistent with TEM imaging of our strains, in that flagella are visible for LAR1 but not for ML115 (Figure 4C).

3.6. RpoC^H419P impacts tolerance of other inhibitors and possibly the stringent response

As part of the global transcription machinery, RpoC is involved in all transcription events. Replacing RpoC with RpoC^H419P in ML115 (YC002), without implementation of any other mutations, was observed to significantly increase the specific growth rate both in our control condition and in the presence of 10 mM exogenous C8 (Table 2), to increase the average membrane lipid length (Figure 3C) and increase the cell surface hydrophobicity (Figure 3D). To gain further insight into the applicability of this mutation to other bio-production scenarios, we compared the growth of ML115 expressing the restored form of WaaG (YC001) to ML115 expressing WaaG^R and RpoC^H419P (YC005) in the presence of a variety of inhibitors (Figure 5A). These experiments were done in the presence of the restored form of waaG (waaG^R) in order to increase similarity to other E. coli strains.

Figure 5:

RpoC^H419P impacts tolerance to a variety of inhibitors, possibly by affecting interaction of RpoC and ppGpp. Unless otherwise indicated, cells were grown at 37°C in MOPS minimal media containing 2.0 wt% dextrose and the indicated inhibitor and with an initial pH of 7.0.

(A) Replacement of rpoC with rpoC^H419P in ML115+waaG^R impacts tolerance to a variety of inhibitors. This analysis compares strains YC001 and YC005.

(B) The RpoC site 1 mutation also impacts tolerance relative to the corresponding isogenic control strain. Data for RpoC^H419 is reproduced from Figure 5A, comparing strains YC001 and YC005. Data for the site 1 mutant compares previously characterized strain RLG14536 lacking site 1 (RpoC R362A, R417A, K615A, RpoZ Δ2-5) to its corresponding control RLG14535. The indicated p-values compare the magnitude of the increase in specific growth rate due to rpoC^H419P to the increase in specific growth rate due to the site 1 null mutation (1-2+).

(C) RpoC^H419P does not delay recovery from nutrient downshift. ML115+waaG^R with either the wild-type version of rpoC (YC001) or rpoC^H419P (YC005) was grown at 30°C in LB and then washed and resuspended in either LB or MOPS minimal growth medium.

The presence of RpoC^H419P relative to wild-type RpoC was observed to increase the specific growth rate by more than 25% in the presence of furfural, vanillin, octanoic acid (C8), hexanoic acid, n-butanol and citrate (Figure 5A). Growth rates in the presence of moderate thermal stress (42°C) and low pH (5.5) were observed to decrease by more than 25% (Figure 5A). Thus, the RpoC^H419P mutation confers a growth benefit in the presence of many, but not all, inhibitory molecules and conditions.

The decrease in thermotolerance is especially intriguing, given the previous reported association of thermotolerance and the stringent response, as mediated by the alarmone ppGpp. Specifically, ppGpp has been shown to accumulate following heat shock (Abranches et al., 2009) and strains deficient in ppGpp production have increased sensitivity to heat shock (Yang and Ishiguro, 2003). The mutated residue in RpoC^H419P is very close to residue 417, which has been reported to be a component of the site 1 binding site for ppGpp on the RNA polymerase complex (Ross et al., 2016; Ross et al., 2013). Visualization of H419 within the existing structural model (Zuo et al., 2013) indicates that this amino acid does not directly contact ppGpp or interact directly with the active site (Figure 1). However, proline substitutions can be disruptive to local structure, and this one could potentially alter the conformation of residues directly contacting ppGpp, thereby altering binding or function of ppGpp indirectly.

Modifications to the RNA polymerase complex that eliminate site 1 (RpoC R362A, R417A, K615A and RpoZ Δ2–5) have been previously described (Ross et al., 2016). Characterization here of this site 1 null mutant and the corresponding control (RLG 14535) supports the possible role of site 1 in C8 tolerance. Specifically, the site 1 null mutant had increased tolerance to C8, isobutanol and n-butanol, as evidenced by an increase in the specific growth rate (Figure 5B). However, the magnitude of the increase in growth rate in the presence of C8 or n-butanol relative to the corresponding control was not as large for the site 1 mutant as was observed with RpoC^H419P (p < 0.001).

The site 1 mutant was previously demonstrated to have delayed recovery from a nutrient downshift from rich medium to minimal medium relative to the corresponding wild-type control (Ross et al., 2016). Specifically, the wild-type control strain had a lag time of approximately 3 hours, while the site 1 mutant had a lag time of approximately 6 hours (Ross et al., 2016). Here, we observed that the strain expressing the H419P mutation (YC005) did not show this delayed recovery from a similar nutrient downshift relative to the wild-type control (YC001), with both strains having a lag time of approximately 5.5 hours (Figure 5C).

These results suggest that the H419P mutation may affect binding or function of ppGpp at site 1, but may also have features distinct from previously-characterized mutations. It is possible that both the differing magnitude of the growth rate changes and the differing recovery from nutrient downshift are due to other genetic differences in the strain with the H419P mutation and the strain with the characterized ppGpp Site 1 mutations (e.g., fadD, poxB, ackApta). However the RpoC^H419P mutation is an intriguing strategy for possibly increasing production of other bio-products beyond C8.

4. Discussion

Here, we demonstrate a framework for characterization of evolved strains, with identification of genetic modification strategies that may be applicable to improved microbial performance in other conditions. Not only did we confirm that each of the known mutations contribute to the phenotype of the evolved strain, we were also able to demonstrate the impact of individual mutations on cell physiology (Figure 6). The restriction analysis used here to assess the timing of the mutations would not have detected other mutations within the heterogenous population. However, the demonstration that the three mutations characterized here are sufficient for recreation of the evolved strain phenotype indicates that all of the important mutations were identified. This also demonstrates that evolutionary studies involving only a small number of evolved genomes are still capable of contributing to the design, build, test and learn metabolic engineering design cycle.

Figure 6:

Proposed summary of how mutations in WaaG, RpoC and the BasS-BasR system impact tolerance of octanoic acid and membrane integrity, hydrophobicity, rigidity and composition.

Restoration of WaaG increased membrane integrity and increased the membrane rigidity, as evidenced by DPH polarization. This finding emphasizes the potential of cell-surface sugars and proteins as targets for engineering membrane properties and microbial robustness, consistent with previous reports using a modified version of carbon storage regulator A (CsrA) to modulate the abundance of these sugars and proteins (Jin et al., 2017). Mutations within the BasS-BasR system contribute to the increased tolerance of high C8 concentrations and also to the change in cell surface hydrophobicity. The BasS-BasR system has been previously recognized as contributing to tolerance of n-butanol (Reyes et al., 2012), and this work provides further support for utilizing this system as an engineering target.

The mutation within RpoC^H419P is able to fully account for the increase in average lipid length observed in LAR1, contributes to the increase in cell surface hydrophobicity, and when expressed in conjunction with the restored WaaG resulted in a dramatic increase in C8 production titers. This mutation also increased tolerance to a variety of other inhibitors that are relevant to economically viable bio-production, such as furfural, vanillin, and n-butanol. Comparison of the RpoC^H419P mutation to a mutation in one of the ppGpp-RpoC binding sites (site 1) shows that while the magnitude of the impact is higher for RpoC^H419P, the two mutations both conferred increased growth rate in the presence of exogenous C8. The increased heat sensitivity conferred by RpoC^H419P and the increase in specific growth rate in the absence of C8 challenge are also consistent with perturbed sensitivity of RNA polymerase to ppGpp (Table 2). Thus, the interaction of the stringent response alarmone ppGpp and RNA polymerase appears to be relevant to tolerance and production of short- and medium-chain fatty acids.

Other reports of strain evolution have presented mutations within the stringent response system as a clever strategy for increasing tolerance to biorenewable fuels and chemicals. For example, evolution of E. coli for tolerance to n-butanol, isopropanol, ethanol and 2,3-butandiol each found mutations within ppGpp synthetase I, encoded by RelA (Horinouchi et al., 2017; Horinouchi et al., 2015; Lennen et al., 2019; Reyes et al., 2012). Alcohols have been reported to interact with translation machinery (Laughrea et al., 1984; So and Davie, 1964) and stimulate production of ppGpp (Mitchell and Lucaslenard, 1980). Up to a quarter of the cell’s promoters have been shown to be affected by the stringent response (Sanchez-Vazquez et al., 2019). All of the 10 genes classified as involved in fatty acid biosynthesis initiation and elongation of saturated fatty acids showed a statistically significant decrease in transcript abundance 5 minutes after induction of relA in a strain expressing wild-type RNA polymerase (Sanchez-Vazquez et al., 2019). However, in an isogenic strain expressing RNA polymerase lacking both of the ppGpp binding sites, only one of these 10 genes showed a statistically significant perturbation in expression following relA induction (Table S4). It is possible that, like alcohols, short-chain fatty acids interact with translation machinery. Relaxing of the stringent response, such as by a mutation within RpoC, seems to contribute to the increased growth and fatty acid titers observed in evolved strain LAR1 and reconstructed evolved strain YC005.

It is possible that the original insertion that disrupted the waaG gene occurred unintentionally during the creation of ML115 from MG1655. ML115 is a triple knockout strain engineered to improve fatty acid titer by inactivation of acetate production and the β-oxidation pathway (Li et al., 2012). These gene deletions were implemented by λ Red recombineering and P1 phage transduction, methods that require multiple electroporation, plasmid curing, and phage infection steps, especially in an automated framework. As these processes are often lethal to a large portion of the cells, it may be that a change in membrane composition via the incorporation of an insertion element made a sub-population more robust to these challenges, which was then enriched during subsequent steps. The drastic change in cell properties conveyed by the waaG insertion should serve as a cautionary note to those who perform multiple transformation or transduction steps. However, mutations to waaG and its pathway are easily identifiable by mucoid colony phenotype (Figure 4B), consistent with an increase in EPS (Figure 4A). Mucoid colonies can often be present following P1 transduction (Thomason et al., 2014) and previous studies have shown that mucoid phenotypes decrease transduction efficiency (Zhang et al., 2008) and that mutations via spontaneous insertion of transposable elements often arise during these procedures (Nagahama et al., 2006).

The microbial cell membrane is a frequently-recognized engineering target when addressing microbial robustness, and here we have characterized several strategies for membrane engineering. These results also contribute to the growing body of evidence that the stringent response plays a substantial role in improving not just tolerance, but also production, of fuels and chemicals at economically viable titers.

• RpoC H419P mutation increases fatty acid tolerance and production, lipid length

• BasS-BasR mutation improves tolerance of fatty acids at higher concentrations

• WaaG impacts EPS sugar abundance, membrane permeability and membrane rigidity

• Reverse engineering identifies timing, contribution of mutations to evolved phenotype