Genome reduction improves octanoic acid production in scale down bioreactors

Abstract

Microorganisms in large‐scale bioreactors are exposed to heterogeneous environmental conditions due to physical mixing constraints. Nutritional gradients can lead to transient expression of energetically wasteful stress responses and as a result, can reduce the titres, rates and yields of a bioprocess at larger scales. To what extent these process parameters are impacted is often unknown and therefore bioprocess scale‐up comes with major risk. Designing platform strains to account for these intermittent stresses before introducing synthesis pathways is one strategy for de‐risking bioprocess development. For example, Escherichia coli strain RM214 is a derivative of wild‐type MG1655 that has had several genes and whole operons removed from its genome based on their metabolic cost. In this study, we engineered E. coli strain RM214 (referred to as WG02) to produce octanoic acid from glycerol in batch‐flask and fed‐batch bioreactor cultivations and compared it to an octanoic acid‐producing E. coli MG1655 (WG01). In batch flask cultivations, the two strains performed similarly. However, in carbon limited fed‐batch bioreactor cultivations, WG02 provided a greater than 22% boost to biomass compared to WG01 while maintaining similar titres of octanoic acid. Reducing the biomass accumulation of WG02 with nitrogen limited fed‐batch cultivation resulted in a 16% improvement in octanoic acid titre over WG01. Finally, in a scale‐down system consisting of a stirred tank reactor (representing a well‐mixed zone) and plug flow reactor (representing an intermittent carbon starvation zone), WG02 again improved octanoic acid titre by almost 18% while maintaining similar biomass concentrations as WG01.

Genome reduction is a strategy for reallocating resources away from unneeded stress responses towards production of desired products. Octanoic acid production was improved in an engineered genome‐reduced strain of Escherichia coli relative to the engineered parent strain.

INTRODUCTION

A successful bioprocess depends on coordinating strain engineering, feedstock preparation, bioreactor operation and downstream processing. In combination, these unit operations can leverage organisms to create sustainable alternatives to traditional petrochemical processes (Pfleger & Takors, 2023). However, to be economically competitive, bioprocesses must be scaled to supply sufficiently large volumes of a product while maintaining high titres, fast rates and maximized yields. Unfortunately, as bioreactor volumes increase, these core process parameters cannot always be maintained. Often, reduced performance can be traced to the physical limitations of mixing large culture volumes. Increased mixing times leads to inhomogeneous distribution of nutrients and a distribution of cellular performance states, many of which are sub‐maximal. For this reason, designing bioprocesses based solely on ideal laboratory bioreactor operation is considered a major risk (Crater & Lievense, 2018). To minimize the risk of scale‐up failure, engineers design and evaluate strains in conditions that mimic final bioprocess conditions. This concept can include utilizing the final intended nutrient feed stock in early laboratory cultivations, limiting laboratory cultivations to achievable mixing parameters at scale such as oxygen uptake rate, and evaluating strains in scale‐down systems which mimic the distribution of nutritional environments experienced by cells at large scale. These approaches can help by building strains that are better suited to the stresses at scale and maintain required production parameters throughout the scale‐up process.

One approach to reducing the impact of the reported environmental heterogeneities at scale is genome reduction. In this context, genome reduction refers to the removal of unnecessary or costly genes from the host organism to prevent wasteful expenditures during, often transient, stress responses seen at scale. Genome reduced strains may be more efficient, in that the saved energy and carbon could instead be used for either biomass accumulation, expression of heterologous genes or synthesis of more product (Ziegler & Takors, 2020). However, the exact interplay of heterologous gene burden, reduced cellular maintenance, and production heterogeneity is not always clear. Genome reduction does not show the same benefits at every scale, production phase or cultivation condition (Cordell et al., 2023; Ziegler et al., 2021). In one example, Pseudomonas putida strain EM42, a genome‐reduced strain, produced more protein than wild‐type P. putida KT2440 in batch fermentations, but also acquired additional chemical sensitivities compared to wild‐type (Amendola et al., 2024; Lieder et al., 2015). A similar genome reduced strain, Escherichia coli RM214, was designed based on the transcriptional response of E. coli MG1655 to temporary or transient glucose starvation conditions in a scale‐down reactor. Genes that were upregulated the most during starvation were removed if they were deemed to be unnecessary and unrelated to central metabolism. Deletion targets included flagella and chemotaxis gene clusters; proteins that consumed substantial cellular resources for their synthesis and operation yet were not required to survive transient starvation in a mixed reactor. In these experiments, RM214 maintained a similar maximum growth rate and biomass yield in batch‐flask cultivations when compared to wild‐type. Additionally, RM214 was found to have a lower maintenance energy when grown in a scale‐down reactor with intermittent glucose starvation in a chemostat mode. When expressing eGFP from a rhamnose inducible plasmid during intermittent glucose starvation, RM214 maintained both a greater proportion of high eGFP producing cells and a higher total eGFP yield over 28 h compared to wild‐type (Ziegler et al., 2021). While useful for evaluating growth parameters, chemostat cultivations may be less representative of the typical fed batch strategies used in industry.

In this study, we explore the impact of the genome reduced strain RM214 on microbial oleochemical production in a fed batch environment with glycerol as a substrate. We reasoned that the more efficient RM214 background may provide excess energy and potentially improve heterologous chemical production in industrially relevant cultivation modes. Glycerol as a byproduct of biodiesel, has already been used to produce microbial lipids, showing its potential as a substrate for fatty acid derived products or oleochemicals (Kosamia et al., 2020). Within this oleochemical production space, the demand for medium chain (C8‐C12 acyl chain) species is likely to only increase due to their scarcity in natural and petrochemical sources (Su et al., 2024). Microbial production of these oleochemicals, through expression of chain length specific enzymes, may prove more sustainable if scaled to industrial levels. By integrating the eight‐carbon specific thioesterase (CupTE) into the genomes of RM214 and its MG1655 wild‐type parent, we investigated the impacts of genome reduction on octanoic acid production in batch‐flask, carbon limited fed‐batch, nitrogen limited fed‐batch and a carbon starvation scale‐down reactor (Hernández Lozada et al., 2018).

EXPERIMENTAL PROCEDURES

Media and buffers

Media used based on previously reported experiments (Ziegler et al., 2021). LB miller broth used for preculture consisted of 10 g/L NaCl, 10 g/L Tryptone and 5 g/L Yeast Extract. For all experiments we used a single batch media and altered the starting glycerol or nitrogen concentrations depending on the cultivation conditions. Additionally for the fed‐batch phase in the bioreactor, we used different feeding solutions depending on the nutrient limitation desired. The base media used for flask cultivations and bioreactor preculture consisting of 10 g/L glycerol, 3.2 g/L NaH₂‐PO₄·2H₂O, 11.7 g/L K₂HPO₄, 8 g/L (NH₄)₂SO₄, 0.01 g/L thiamine hydrochloride and a 0.2% (v/v) trace elements stock solution. All STR batch bioreactor cultures consisted of 2 g/L glycerol, 3.2 g/L NaH₂‐PO₄·2H₂O, 11.7 g/L K₂HPO₄, 8 g/L (NH₄)₂SO₄, 0.01 g/L thiamine hydrochloride and a 0.2% (v/v) trace elements stock solution.

The trace element solution consisted of 4.175 g/L FeCl₃·6 H₂O acidified with concentrated HCl, 0.045 g/L ZnSO₄·7 H₂O, 0.025 g/L MnSO₄·H₂O, 0.4 g/L CuSO₄·5 H₂O, 0.045 g/L CoCl₂·6 H₂O, 2.2 g/L CaCl₂·2 H₂O, 50 g/L MgSO₄·7 H₂O and 55 g/L sodium citrate dihydrate. Stock solutions of each trace mineral were prepared separately then combined and filter sterilized for each cultivation. IPTG and thiamine HCl stocks were prepared at 1 M and 50 g/L respectively, filter sterilized and stored frozen until use.

For carbon limited cultivations, the fed‐batch solution consisted of 125 g/L glycerol, 10 g/L NaH₂‐PO₄·2 H₂O, 36.565 g/L K₂HPO₄ and 0.043 g/L thiamine hydrochloride.

For the nitrogen limited cultivation, batch phase media (NH₄)₂SO₄ was reduced to 0.6 g/L and the feed solution was supplemented with 31.9 g/L (NH₄)₂SO₄ to achieve a C/N ratio of approximately 8.5 mol/mol.

Construction of octanoic acid producing strains

Strains of E. coli MG1655 and RM214 expressing the CupTE thioesterase from a chromosomal cassette were constructed with λ‐red mediated recombineering as described previously (Figure 1B) (Hernández Lozada et al., 2018). Briefly, a helper plasmid (pMP11) was used to induce expression of the recombineering system and Cas9 prior to preparation of electrocompetent cells (Mehrer et al., 2018). Aliquots were transformed with plasmid expressing a sgRNA targeting the fadD genomic loci and a double stranded PCR product to be integrated into the fadD locus. The latter DNA comprised an IPTG induced CupTE expression cassette flanked by sequences immediately up‐ and down‐stream of fadD (see Appendix S1 for the CupTE sequence). Figure data are included in the external Appendix S1 file. Recombinant strains were plated on LB‐Agar containing 100 μg/mL ampicillin or carbenicillin to select cells maintaining the sgRNA plasmid which contained the corresponding resistance marker. Correct integration of the CupTE expression cassette in the fadD locus was verified by colony PCR (using primers rNHL153/154 targeting flanking sequences) and sequencing of the amplified product (see Figure S1 and Table 2). Supplementary figures are included in the external Figures S1–S4 file.

FIGURE 1.

(A) Overview of octanoic acid synthesis. Glycerol is converted to acetyl‐CoA in central metabolism. Acetyl‐CoA is incorporated into acyl‐ACPs as intermediates in fatty acid biosynthesis (FAB). Double arrows indicate multiple reactions. Expression of a heterologous thioesterase (CupTE) catalyses the hydrolysis of octanoyl‐ACP to octanoic acid. Octanoyl‐ACP is also a precursor of longer‐chain acyl‐ACPs that are substrates for synthesis of membrane lipids. CoA ligase (FadD) activates fatty acids including octanoic acid to acyl‐CoAs. Acyl‐CoAs are the initial compounds in the β‐oxidation (β‐Ox) pathway that liberates acetyl‐CoA units and produces NADH while catabolizing acyl‐chains. Deletion of fadD prevents free fatty acid activation and its subsequent degradation. (B) Overview of the strain construction process. (C) Minimal media (10 g/L glycerol) batch‐flask cultivations consisting of both WG01 and WG02 grown over 72‐h. Error bars represent the standard deviation of 3 biological replicates and statistical significance is marked by an asterisk representing a p‐value of less than 0.05 for a two‐sided t‐test with variance considered. See Appendix S1 for raw data.

TABLE 2.

Strains, Plasmids, and Primers used in this study.

Strain or plasmid	Genotype	Source
MG1655	F, λ, ilvG, rfb‐50, rph‐1	Ziegler et al. (2021)
RM214 (Wild‐type)	MG1655 Δflk ΔfliA ΔfliC ΔflgNMABCDEFGHIJKL ΔfliEFGHIJKLMNOPQR ΔflhEABcheZYBRtaptarcheWAmotBA ΔcspD ΔaldA ΔgatABCDR ΔuhpTCBA ΔyeeL ΔflxA	Ziegler et al. (2021)
WG01	MG1655 ΔfadD::trc‐CpFatB1.2‐M4–287	This study
WG02	RM214 ΔfadD::trc‐CpFatB1.2‐M4–287	This study
pMP11	pKD46 with constitutively expressed Cas9 and an aTc gRNA targeting the ColE1 origin	Hernández Lozada et al. (2018) and Mehrer et al. (2018)
pgRNA	Constitutively expressed sgRNA targeting fadD	Hernández Lozada et al. (2018) and Mehrer et al. (2018)
rNHL153	TGTCAGATGTTTGCTGAGGGTAAA	Hernández Lozada et al. (2018)
rNHL154	ATAAACCCAGGCTGTCCAGTTC	Hernández Lozada et al. (2018)

Characterization of the scale‐down bioreactor system

To mimic the large‐scale heterogeneities, we performed experiments in a bioreactor system composed of a stirred tank reactor (STR) coupled to a plug flow reactor (PFR). The two reactors, linked by a loop connection, provided a perfectly mixed zone (STR) and a stressed zone (PFR), mimicking the heterogeneities faced by cells in large‐scale tanks. A fraction of the cell culture in the STR was pumped to the PFR using a fixed pump rate, allowing a defined mean residence time in the PFR. These cultivations were set as described previously (see Figure S2 for more detail) (Ziegler et al., 2021). Two probes were installed to monitor the dissolved oxygen (DO) tension at the inlet and the outlet of the PFR. Preliminary experiments using cell‐free media were used to identify a combination of pump and aeration rates that provided a mean residence time of 100 s in the PFR. Here, conductometers, located at the PFR inlet and outlet were used to detect changes in conductivity when pulses of 3 M KCl were added to the STR. The average residence time (τ), and its variance (σ ²) were calculated from the conductance records according to (Levenspiel, 2012):

$$\tau =\frac{{\sum }_i^{n-i}{t}_i\bullet c_i\bullet ∆t_i}{{\sum }_i^{n-i}{c}_i\bullet ∆t_i}$$ (1)

$${\sigma }^2=\frac{{\sum }_i^{n-i}{t}_i^2\bullet c_i\bullet ∆t_i}{{\sum }_i^{n-i}{c}_i\bullet ∆t_i}-{\tau }^2$$ (2)

where c is the conductivity signal, that is related to the tracer concentration at the time point t at the measuring interval i. The Bodenstein number (Bo) was also assessed to characterize the degree of back‐mixing in the PFR which represents the ratio of the convective mass transfer to the diffusive mass transfer in a flowing fluid:

$$\mathit{Bo}=\frac{2{\tau }^2}{{\sigma }_{\tau }^2}$$ (3)

The results obtained are reported in Table 1.

TABLE 1.

Setup and results obtained by the system characterization.

Pumping rate (s−1)	Aeration (vvm)	Mean residence time (τ PFR , s)	Bodenstein' number (Bo)	Variance (σ2, S 2)
100	0.25	102.8 ± 1.2 a	27.2 ± 1.9 a	808

^a Mean of three replicates and the relative standard deviation.

A Bo >10 is generally associated with a plug‐flow behaviour of the system, this was calculated according to (Octave, 1998) and (George et al., 1993). A high Bodenstein number indicates that convective transport dominates over diffusive transport. Conversely, a low Bodenstein number indicates that diffusive transport is more significant, meaning that the species spreads out due to diffusion more than it is carried by the flow. In our case, the Bo number of 27 indicates that the back mixing along the PFR has a minor importance, estimated to be less than 5%, compared to the convective flow. This means that samples taken in compartments along the PFR mimic the cellular exposure to the stress for the time interval that equals convective flow. Since the flow behaviour is intermediate (or real) we estimated the amount of tracer recovered by evaluating the function F(t) at 0.95, corresponding to a time of about 127 s. This means that approximately 95% of the tracer exits the system at τ _PFR  + 24 s:

$$F\left(t\right)=\frac{{\sum }_{t_i\le t}{C}_i∆t_i}{\sum C_i∆t_i}$$ (4)

Pre‐growth conditions, flask, batch and fed‐batch cultivation

For flask cultivations, pre‐cultures were inoculated from a freezer stock and grown in 5 mL of LB‐miller media overnight at 30°C. Experimental cultures were inoculated from the pre‐culture to an OD₆₀₀ of 0.05 and induced with 1 mM IPTG 3 h after inoculation. Flasks were grown at 30°C shaking at 250 rpm for 72‐h before harvesting for fatty acid extraction and cell dry weight (CDW) evaluation.

For scale‐down bioreactor cultivations, pre‐cultures were inoculated from a freezer stock scrape and grown in 5 mL of LB‐miller media overnight at 30°C. After 16 h, the whole preculture was inoculated into a 2 L flask containing 200 mL of minimal media and grown for 24 h. STR‐PFR bioreactor cultivation began starting OD₆₀₀ of 1. This was achieved by inoculating 1400 mL of batch minimal media with 50 mL of the preculture concentrated and resuspended with 9 g/L NaCl. Temperature was maintained at 30°C with 0.4 mL/min sparging and a starting stir speed of 200 rpm. Dissolved oxygen was controlled above 20% by increasing the stirring speed. The vessel was maintained at 0.5 bar overpressure. pH was maintained at 7.0 by the addition of 25 vol% ammonia hydroxide. Additionally, antifoam 204 was added at 10–20 μL/h to minimize foaming. Fed‐Batch cultivations began once the initial glycerol was consumed, as indicated by a spike in the dissolved oxygen tension.

The fed‐batch phase was controlled via an exponential feed based on Equation (5):

$$F\left(t\right)=\left(0.00109\right)\bullet V_0\bullet e^{0.105\bullet t}$$ (5)

where F(t) dictates the rate of feeding over time in L/h, t is time in fed‐batch phase in s and V ₀ is the starting volume of the cultivation in litres. The feeding rate equation was calculated based on estimations of cellular yield on glycerol and the desired growth rate. To calibrate pump feeding rates by mass, the density of the feeding solution was measured and found to be 1.066 kg/L. STR‐PFR cultivations were set as described previously (Ziegler et al., 2021). Depending on the size of the reactors, glycerol fed between replicates was the same or similar (see Table 2). Starting at the fed‐batch phase, cultures were induced at a concentration of 1 mM IPTG. Final points for titre, CDW, and HPLC analysis were taken upon 24 h of feeding. The plug flow reactor made use of a diaphragm pump that circulates approximately 370 mL of culture at laminar flow rates. This worked, in carbon limited condition, as carbon starvation zone. Air was sparged into the PFR at 0.037 mL/min (0.1 vvm) to maintain oxygenation above 20%. The pump rate was set to 117 1/s, leading to a residence time of approximately 100 s in the PFR. Additional fed‐batch experiments were completed with an Infors Multifors 1.4 L bioreactor system with a starting volume of 700 mL of media. Inoculation followed the same pre‐culture noted above with a starting OD of 1 and the fed‐batch feeding rate adjusted for the smaller volume. Gassing was maintained at 0.4 L/min with stirring maintained at 250 rpm and increased as needed to maintain DO above 40%. pH control was maintained at 7.2 with 25 vol% ammonia hydroxide. Antifoam was added manually as needed via a syringe. The fed‐batch equation used was the same as noted above (Equation 5). For nitrogen limited cultivations pH control was maintained with 4 M sodium hydroxide instead of ammonia hydroxide. Analysis to confirm consistent growth rates once the feed was started is included in Figure S3.

Determination of octanoic acid, glycerol, acetate, optical density and cell dry weight and carbon balance

Samples for optical density (1 mL) and analyte analysis (5 mL) were taken throughout the culture except for batch flask cultivations which were sampled at 24, 48 and 72 h at 3 mL each. Optical density was measured through dilution of the culture with 9 g/L NaCl to the linear range of a spectrophotometer (Amersham Biosciences/GE Healthcare, Amersham, United Kingdom). Dry well weights were measured in triplicate at the end of each cultivation by centrifuging and washing 10 mL of cell culture in a glass tube with 10 mL of 9 g/L NaCl. Cell pellet weight was noted after drying for 48 h in a 105°C oven. Conversion to carbon moles was based on the literature value from the following reference (Duboc et al., 1999).

For 700 mL STR fermentations, octanoic acid quantification was completed with 2.5 mL of culture, acidified via 100 μL of glacial acetic acid, 25 μL of 50 g/L nonanoic acid internal standard and extracted with 3 mL of hexane to ensure extra volume and dilute the octanoic acid for GC–MS analysis. For GC–MS analysis, an Agilent FatWax UI 30 m 0.25 diameter with the following temperature protocol: 120°C hold for 2 min, increase to 140°C at a rate of 5°C/min and hold for 3 min, increase to 250°C at a rate of 20°C/min and hold for 6 min. Glycerol and acetate for baseline 700 mL STR fermentations with a Shimadzu HPLC with an isocratic method of 0.6 mL/min 5 mM sulfuric acid as a mobile phase and detection by refractive index detector. Various standards of glycerol and acetate were run at the beginning and end of the analysis to verify the linear range.

For 1400 mL STR batch and fed‐batch fermentations, octanoic acid was quantified by a Shimadzu, Model GC‐2010 gas chromatography equipped with an AOC‐20i autoinjector and flame ionizing detector. Octanoic acid was captured via hexane extraction from the culture. Using 1 mL of cell culture, 0.5 mL of hexane was added and vortexed for 10 min. 40 μL of nonanoic acid from a 10 g/L stock in ethanol was added as an internal standard. External standards for octanoic acid production were made by dosing in a known amount of octanoic acid from a 25 g/L stock in ethanol into 4 mL of batch media, diluted to various concentrations, and extracted the same as experimental samples. For GC‐FID analysis, an Agilent FatWax UI 30 m 0.25 diameter with the following temperature protocol: 120°C hold for 2 min, increase to 140°C at a rate of 5°C/min and hold for 3 min, increase to 250°C at a rate of 20°C/min and hold for 6 min. Glycerol and acetate were quantified with Boehringer Mannheim/R‐Biopharm Enzymatic BioAnalysis/Food Analysis for start and end points of the fermentation. Preliminary samples for the STR‐PFR cultivations were run on an HPLC for identification of primary fermentative products including pyruvate, formate, citrate, acetate, lactate and ethanol. Only acetate remained as an accumulated product during cultivation (data not shown). Note that while 10 carbon fatty acid products were observed in GC traces, their abundance was often small (<50 mg) compared to the eight carbon products. We elected to ignore these compounds in our analysis (data not shown). Total glycerol feed to the reactor was measured by either monitoring the mass of the feed bottles or using feed pump calibration values. The total feed glycerol was calculated using the density of the feed solution and the known concentration glycerol in the feed. Total carbon balance was calculated considering the balance of the inputs from starting glycerol, glycerol fed and starting optical density with the outputs of the final biomass, octanoic acid, acetate and carbon dioxide produced. Each component was converted to carbon moles and the remaining balance was marked as unknown. Assumptions regarding total carbon produced from carbon dioxide were calculated based on previously published methods (Chmiel et al., 2011). The composition of the bioreactor off gas was measured with a BlueSense gas monitor. To calculate the total carbon dioxide produced we first calculated the carbon evolution rate (CER) based on Equation (6):

$$\mathrm{CER}=-\frac{p{\dot{V}}_{\mathrm{g},\text{in}}}{V_L\mathrm{RT}}\left(Y_{\mathrm{CO}2,\text{in}}-\frac{1-Y_{\mathrm{O}2,\text{in}}-Y_{\mathrm{CO}2,\text{in}}}{1-Y_{\mathrm{O}2,\text{out}}-Y_{\mathrm{CO}2,\text{out}}}{Y}_{\mathrm{CO}2,\text{out}}\right)$$ (6)

where ${\dot{V}}_{g,\text{in}}$ is the flow of gas in, V _L is the liquid volume of the reactor, R is the gas law constant and T is temperature. Off gas data were collected on a per minute basis allowing the calculation of carbon evolved per minute by utilizing the trapezoidal integration method between timepoints. Total carbon dioxide evolved for Figure 5 was calculated by summing the carbon dioxide evolved over the entire cultivation. Note an accumulation term was considered for the carbon balance but was found to be insignificant and was not included in the data presented.

FIGURE 5.

Carbon balance of the STR and STR‐PFR cultivations including glycerol fed in the batch and fed‐batch phase, and octanoic acid, DCW, acetate and total carbon dioxide produced by the end of the cultivation. See Appendix S1 for raw data.

RESULTS AND DISCUSSION

Validation of octanoic acid production in batch flask cultivations

To build the octanoic acid producing strains, we replaced the fadD locus with an IPTG‐induced thioesterase (CupTE) expression cassette in both E. coli MG1655 and the reduced genome E. coli RM214 (Figure 1B) (Hernández Lozada et al., 2018). The markerless cassette was integrated in using λ‐red recombineering and Cas9‐mediated removal of the fadD locus (see methods) to create strains WG01 (MG1655 background) and WG02 (RM214 background) (Datsenko & Wanner, 2000). The removal of fadD and integration of a thioesterase blocks β‐oxidation, allowing accumulation of thioesterase‐generated octanoic acid in culture (Figure 1A). To confirm octanoic acid production and complete glycerol consumption, cells were cultured in minimal media for 72 h in a shake flask at 30°C. The starting 10 g/L glycerol was completely consumed by both strains, and the final octanoic acid titres were – 478 ± 5 mg/L for WG01 and 426 ± 13 mg/L for WG02 (Figure 1C and Table 3). The final biomass titre of WG02, reported as dry cell weight per litre (gDCW/L), was slightly higher (2.69 ± 0.12 g/L) than WG01 (2.42 ± 0.06 g/L) (Figure 1C and Table 3). Both strains produced acetate with WG02 accumulating more than four times more acetate than WG01 but with greater variability. With only a final 72‐h point, more subtle changed in acetate production and consumption over the flask experiment may not be captured however these data still speak to the efficiency of strains in a batch cultivation (Figure 1C and Table 3). Overall, WG02 maintained slight increase in biomass concentration compared WG01, but as has been seen in previous batch flask studies case (Ziegler et al., 2021). The RM214 background did not show any benefit to octanoic acid production and produced more acetate in this case.

TABLE 3.

Relevant process parameters for experiments conducted in this paper.

Strain	WG01	WG02	WG01	WG02	WG01	WG02	WG01	WG01	WG02
Cultivation Conditions	Batch‐flask	Batch‐flask	Carbon limited fed‐batch bioreactor	Carbon limited fed‐batch bioreactor	Nitrogen limited fed‐batch bioreactor	Nitrogen limited fed‐batch bioreactor	Carbon limited fed‐batch bioreactor	Carbon limited fed‐batch bioreactor	Carbon limited fed‐batch bioreactor
Replicates	3	3	3	3	3	3	2	2	2
Stress	NA	NA	NA	NA	NA	NA	NA	PFR Carbon Starvation	PFR Carbon Starvation
Starting volume (mL)	50	50	700	700	700	700	1400	1400	1400
Glycerol Base media (g/L)	10	10	2	2	2	2	2	2	2
Glycerol fed (g)	NA	NA	9.67 ± 0.20	9.73 ± 0.00	9.14 ± 0.20	9.24 ± 0.01	20.44 ± 0.22	20.45 ± 0.64	20.57 ± 0.32
Octanoate (mg/L)	478 ± 5	425 ± 13	838 ± 28	873 ± 46	846 ± 85	991 ± 125	927 ± 10	1017 ± 9	1238 ± 29
gDCW/L	2.42 ± 0.06	2.69 ± 0.12	3.76 ± 0.09	4.86 ± 0.10	3.62 + 0.10	4.07 + 0.06	3.64 ± 0.24	3.72 ± 0.08	4.02 ± 0.12
Acetate (g/L)	0.06 ± 0.02	0.25 ± 0.18	0.62 ± 0.08	<0.1	0.61 ± 0.1	0.24 ± 0.06	1.47 ± 0.11	1.05 ± 0.14	0.33 ± 0.03
Yps (mol/mol)	0.031 ± 0.000	0.027 ± 0.001	0.039 ± 0.002	0.040 ± 0.002	0.040 ± 0.003	0.048 ± 0.006	0.039 ± 0.001	0.043 ± 0.001	0.052 ± 0.001
Yxs (g/g)	0.24 ± 0.01	0.27 ± 0.01	0.25 + 0.01	0.33 + 0.01	0.24 ± 0 0.00	0.27 + 0.00	0.23 ± 0.00	0.23 ± 0.01	0.25 ± 0.00

Note: octanoic acid, gDCW/L, and acetate are endpoint values and yield calculations include base media glycerol in their calculations. Y_xs includes a correction for the starting OD₆₀₀ of the cultures. Error represents standard deviation for experiments of three replicates and absolute error for experiments with two replicates. See Appendix S1 for raw data.

Genome reduction improves biomass yield in carbon limited fed‐batch cultivations

To further evaluate RM214 and wild‐type strains in a more industrially relevant scenario, we cultivated each strain in bioreactors while applying a carbon limited exponential feeding strategy (Figure 2A). Upon complete consumption of the 2 g/L glycerol in the batch phase, an exponential feed was initiated, and the cells were induced by the addition of IPTG. During the fed‐batch phase, WG01 was fed 9.73 ± 0.00 grams of glycerol and WG02 was fed 9.67 ± 0.26 g of glycerol (Table 3). After the fed‐batch phase, we observed octanoic acid titres of 873 ± 46 mg/L for WG02 and 838 ± 28 mg/L for WG01 (Figure 2C and Table 3). Additionally, WG02 generated ~22% higher biomass titre compared to WG01 (4.86 ± 0.10 vs. 3.76 ± 0.09 gDCW/L) (Figure 2C and Table 3). Interestingly, a biomass increase was not noted in previous chemostat experiments comparing eGFP‐producing RM214 and wild‐type strains, however these experiments were completed with glucose as a substrate (Ziegler et al., 2021). Finally, WG02 accumulated less acetate (<0.1 g/L) than WG01 (0.62 ± 0.08 g/L) (Figure 2C and Table 3). These data indicate that in a carbon limited fed‐batch, the primary benefits of genome reduction are a minor increase in octanoic acid titre, reduced acetate accumulation, and increased biomass yield on glycerol.

FIGURE 2.

(A) Experimental setup for the carbon limited fed‐batch. (B) OD₆₀₀ of WG01 and WG02 throughout the cultivation with a dotted line to separate the Batch and Fed‐batch phases. (C) Final octanoic acid, acetate, and biomass (gDCW/L) titre after 24 h of an exponential feed. Error bars represent the standard deviation of 3 biological replicates and statistical significance is marked by an asterisk representing a p‐value of less than 0.05 for a two‐sided t‐test with variance considered. See Appendix S1 for raw data.

Genome reduction alters product yield and reduces acetate formation in nitrogen limited fed‐batch cultivations

Given WG02 had a higher biomass yield on glycerol, we reasoned that an alternative nutrient limitation, such as nitrogen, could potentially shift carbon and energy flux to octanoic acid or potentially other byproducts rather than biomass (Rajpurohit & Eiteman, 2022). Informed by previous nitrogen limitation experiments, we maintained a C:N ratio of approximately 8.5 mol/mol during the cultivation to ensure nitrogen limited growth (Löffler et al., 2017). During the fed‐batch phase, WG01 and WG02 strains were fed 9.14 ± 0.20 and 9.24 ± 0.01 g of glycerol respectively. The final octanoic acid titre for WG02 (991 ± 125 mg/L) was higher than WG01 (828 ± 85 mg/L) (Figure 3C and Table 3). WG02 acetate titres were again reduced relative to WG01 (Figure 3C and Table 3). The final CDW for WG02 (4.07 ± 0.06 gDCW/L) and WG01 (3.62 ± 0.10) was reduced compared to the carbon limited fed‐batch by ~16% and ~4% respectively compared to the carbon limited fed‐batch (Figures 2C and 3B,C and Table 3). Overall comparing carbon and nitrogen limited fed‐batch strategies, WG02 had a small benefit to biomass yield, but primarily improved octanoic acid production, although with greater variability under nitrogen limitation. WG01 in contrast, maintained a similar octanoic acid titre (less than 1% difference) and biomass production to the corresponding carbon limited fed‐batch (Figures 2 and 3).

FIGURE 3.

(A) Experimental setup for the nitrogen limited fed batch including a feed and batch phase media with a combined molar C:N ratio of approximately 8.5:1. (B) OD₆₀₀ of WG01 and WG02 throughout the cultivation with a dotted line to separate the batch and fed‐batch phase. (C) Final octanoic acid, acetate and biomass (gDCW/L) titre after 24 h of an exponential feed. Error bars represent the standard deviation of 3 biological replicates and statistical significance is marked by an asterisk representing a p‐value of less than 0.05 for a two‐sided t‐test with variance considered. See Appendix S1 for raw data.

Genome reduction results in a higher biomass and octanoic acid yield during intermittent glycerol starvation

In a final experiment, we grew WG01 and WG02 in a scale‐down reactor mimicking an intermittent carbon starvation stress (Figure 4A and Figure S2). To first verify the differing reactor cultivation did not significantly alter cultivation outcomes, we first grew WG01 without any PFR stress. Product yields were the same and biomass yields were less than 9% different when comparing WG01 cultivations from the previous carbon limited cultivation described in Figure 2, suggesting little effect (see Table 3). For each condition, approximately the same amount of glycerol was fed to each reactor (Table 3). To confirm the scale‐down reactor was sufficiently stressing the cells to illicit a response, we next compared growth of WG01 in both an STR (with no glycerol starvation stress) and an STR‐PFR (with intermittent glycerol starvation). To our surprise after completion of the fed‐batch phase, cultivation in an STR‐PFR led to an improvement in octanoic acid titre, from 927 ± 9 mg/L to 1017 ± 10 mg/L and a reduction in acetate accumulation (Figure 4C and Table 2), despite less than a 2% difference in biomass titre. Culturing WG02 in an STR‐PFR resulted in both an increase in the final titre (1238 ± 29 mg/L) and in the gDCW/L (4.02 ± 0.12) and a reduction in acetate (0.33 ± 0.03 g/L) compared to WG01 (Figure 4B,C and Table 3). WG02 appears to be significantly limited in its final biomass compared to the unstressed carbon limited fed‐batch (Figure 2C), and instead maintains a higher octanoic acid titre, like the nitrogen limited conditions shown in Figure 3C. Additional analysis of glycerol and acetate over time is included in Figure S4 showing a very slight accumulation of glycerol (less than 0.25 g/L) for WG01 in the STR and STR‐PFR cultivations. A consequence of this may be that the carbon starvation stress is less strenuous along the PFR for WG01 if it does not completely consume glycerol during the fed‐batch phase in the STR.

FIGURE 4.

(A) Experimental setup for the carbon limited fed‐batch cultivation including both an STR and PFR reactor to simulate poor mixing conditions. (B) OD₆₀₀ of WG01 and WG02 strains throughout the cultivation with a dotted line to separate the batch and fed‐batch phase. (C) Final octanoic acid, acetate and biomass (gDCW/L) titre after 24 h of an exponential feed. Error bars represent the absolute error of biological duplicates. See Appendix S1 for raw data.

The overall carbon balance of the STR and STR‐PFR cultivations inform how the scale‐down stress and genome reduction alters cellular expenditures. During carbon starvation, WG01 appears to shift acetate accumulation primarily to CO₂ and octanoic acid production but maintained a similar biomass titre, whereas WG02 accumulated relatively little acetate, and instead produced more biomass, octanoic acid and carbon dioxide compared to wild‐type (Figure 5). The relatively lower fraction of biomass and higher fraction of CO₂ may be a result of diverting carbon to the octanoic acid product and additional CO₂ released in the initiation of fatty acid production. Given the unknown fraction of this carbon balance, it is also possible that the extraction methodology used in these experiments underestimates the total octanoic acid produced or that other byproducts were not captured in our analysis. Furthermore, it is unclear to what extent each strain relies on the glyoxylate shunt when consuming glycerol. The increased fraction of carbon dioxide may suggest that RM214 utilizes the full TCA cycle more to support increased octanoic acid production.

DISCUSSION

In comparing RM214 and wild‐type strains for octanoic acid production, we first showed that in batch flask WG02 had a reduced titre but slightly more biomass and acetate in a batch‐flask cultivation after 72‐h (Figure 1C and Table 3). When scaled to a carbon limited fed‐batch, WG02 grew faster than its wild‐type counterpart leading to a higher biomass yield and slightly improved octanoic acid titre (Figure 2B,C and Table 3). Considering this result occurred in the carbon limited fed‐batch without a carbon starvation condition, likely either the glycerol itself as a carbon source or the carbon limitation induced a stress response in the cell. If true this could lead to expression of chemotaxis and flagellar genes, as has been suggested in previous minimal media experiments (Bhatia et al., 2022; Liu et al., 2005). Additionally, the cost of continued flagellar motion could also lead to significant energy saving in RM214 (Schavemaker & Lynch, 2022). In theory, if these genes are expressed in WG01 and not WG02, the excess energy and carbon could be directed to either product formation, maintenance, or biomass. The results from Figure 2 show that in a carbon limited fed‐batch, WG02 will spend its excess resources primarily on biomass generation. An added benefit of WG02 seen in all fed‐batch cultivations is also reduced acetate accumulation (Figures 2C, 3C and 4C). Interestingly, while there have been previously reported glycerol‐based fed‐batch cultivations for fatty acid production, acetate was not noted as a major side product (Lu et al., 2008). Utilizing a nitrogen limited fed‐batch, we successfully reduced the biomass accumulation seen in Figure 2, resulting in an increase in octanoic acid titre and yield (Figure 3, Table 3). Comparing the carbon and nitrogen limitations, WG01 approximately maintained both its biomass and product yields, whereas WG02 had a significant reduction (~21%) in biomass yield and a significant increase (~16.7%) in product yield in nitrogen limited conditions, although with increased variability in the latter (Table 3). These results suggest that the previously reported maintenance benefits of RM214 are more obvious as growth rates are reduced under nutrient limited conditions. In nitrogen limited conditions, we saw benefits to octanoic acid production and a reduction of acetate accumulation.

Finally, we evaluated the effect of an intermediate carbon starvation for WG01 and WG02 strains. Surprisingly, we found that a starvation stress did not affect biomass accumulation but instead led to a modest increase in product titre, reduced acetate accumulation, and a similar OD₆₀₀ and final DCW for the WG01 cultivation (Figure 3B,C). One possible reason for the increase in octanoic acid titre could be that cell growth is halted within the carbon starvation zone, by limiting glycerol backbones for phospholipid biosynthesis but fatty acid biosynthesis continues. Carbon starvation is known to induce the stringent response, leading to an increase ppGpp levels which decreases phospholipid production though the regulation of PlsB (Noga et al., 2020). So, fatty acyl‐ACPs may continue to be made, and potentially converted to free fatty acids by CupTE, but ultimately cannot be converted to phospholipids. Comparing WG02 to WG01 in an STR‐PFR cultivation showed a more than 17% improvement octanoic acid titre (Figure 3C) and a greater than 7% increase in biomass for WG02 (Figure 3B,C). The result supports that removal of unnecessary genes typically expressed during carbon starvation may provide a surplus of energy for alternative functions in the cell if grown in a carbon stressed condition.

Assessing the carbon balance between WG01 and WG02, with approximately 80–85% of the carbon moles accounted for, WG02 appears to produce more carbon dioxide and octanoic acid, whereas WG01 produces relatively more acetate (Figure 5). However, it is not immediately clear why acetate is produced as a byproduct, or why WG02 produces less acetate overall when grown on glycerol. Previous analysis suggested glycerol induces an acetate recycling mechanism in which poxB, along with acetate reuptake through increased expression of ACS and pta (Martínez‐Gómez et al., 2012). If additional energy is available in WG02, this may allow efficient reuptake of acetate (2 ATP equivalent via ACS or 1 ATP via ackA‐pta), which is then committed to either the TCA cycle or fatty acid production (Wolfe, 2005). Alternatively excess energy could reduce the need for acetate recycling, and in combination with increased acetyl CoA demand in fatty acid biosynthesis, less acetate is then produced from acetyl‐CoA or pyruvate. Considering future metabolic engineering efforts, LB glucose batch‐flask cultivations have shown significantly reduced acetate accumulation with the removal of poxB and ackA‐pta. However, removal of these genes has also resulted in reduced fatty acid titre in production strains (Li et al., 2012). Whether acetate production or re‐uptake, more research is needed to understand the driving factor for acetate accumulation for WG01 and WG02 during octanoic acid production. While this report focuses on comparing the efficiency of strains in various fed batch cultivations, other considerations for future work could be if genome reduction has any effect on the toxic final titres of fatty acid production. Previous reports suggest a toxic limit of octanoic acid is around 2 g/L and for mixed medium chain fatty acid products is just below 4 g/L in E. coli (Hernández Lozada et al., 2018; Wu et al., 2017). Future tolerance experiments should also consider previously reported adaptive laboratory evolution mutations and known metabolic engineering strategies in stressed fed‐batch cultivations (Chen et al., 2020; Lennen et al., 2023; Yan & Pfleger, 2020). This strategy will ensure the modifications are beneficial in a strain endogenously producing octanoic acid in industrial‐like scenarios.

CONCLUSION

Within these experiments, we found that the cost or benefit of genome reduction greatly depends on the cultivation mode. Other than a small increase in biomass yield, WG02 provides no substantial benefit when compared to WG01 in batch‐flask cultivations. However, depending on the fed‐batch bioreactor conditions, WG02 may primarily benefit either biomass yield (in a carbon limitation) or octanoic acid yield (in a nitrogen limitation). Furthermore, when exposed to intermittent carbon starvation conditions, WG02 maintained both a higher biomass and octanoic yield compared to WG01. Whether through intermittent carbon starvation stress or nitrogen limitation, we showed that genome reduction can improve heterologous product accumulation on glycerol. It is worth emphasizing however that WG02 is not superior in all cultivation conditions, stressing that future applications of genome reduced strains should always be compared to their parent strains to confirm improvements or shortfalls if any.

AUTHOR CONTRIBUTIONS

William T. Cordell: Conceptualization; investigation; writing – original draft; writing – review and editing; formal analysis; visualization. Gennaro Avolio: Conceptualization; investigation; writing – original draft; writing – review and editing; formal analysis. Ralf Takors: Conceptualization; funding acquisition; writing – review and editing; supervision; formal analysis. Brian F. Pfleger: Conceptualization; funding acquisition; writing – review and editing; project administration; formal analysis; supervision.

FUNDING INFORMATION

No funding information is provided.

CONFLICT OF INTEREST STATEMENT

The authors declare no competing interests.

Supporting information

Appendix S1

Figures S1–S4

Cordell, W.T. , Avolio, G. , Takors, R. & Pfleger, B.F. (2024) Genome reduction improves octanoic acid production in scale down bioreactors. Microbial Biotechnology, 17, e70034. Available from: 10.1111/1751-7915.70034

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.