Metabolic Engineering of Actinobacillus succinogenes Provides Insights into Succinic Acid Biosynthesis

ABSTRACT

Actinobacillus succinogenes, a Gram-negative facultative anaerobe, exhibits the native capacity to convert pentose and hexose sugars to succinic acid (SA) with high yield as a tricarboxylic acid (TCA) cycle intermediate. In addition, A. succinogenes is capnophilic, incorporating CO₂ into SA, making this organism an ideal candidate host for conversion of lignocellulosic sugars and CO₂ to an emerging commodity bioproduct sourced from renewable feedstocks. In this work, we report the development of facile metabolic engineering capabilities in A. succinogenes, enabling examination of SA flux determinants via knockout of the primary competing pathways—namely, acetate and formate production—and overexpression of the key enzymes in the reductive branch of the TCA cycle leading to SA. Batch fermentation experiments with the wild-type and engineered strains using pentose-rich sugar streams demonstrate that the overexpression of the SA biosynthetic machinery (in particular, the enzyme malate dehydrogenase) enhances flux to SA. Additionally, removal of competitive carbon pathways leads to higher-purity SA but also triggers the generation of by-products not previously described from this organism (e.g., lactic acid). The resultant engineered strains also lend insight into energetic and redox balance and elucidate mechanisms governing organic acid biosynthesis in this important natural SA-producing microbe.

IMPORTANCE Succinic acid production from lignocellulosic residues is a potential route for enhancing the economic feasibility of modern biorefineries. Here, we employ facile genetic tools to systematically manipulate competing acid production pathways and overexpress the succinic acid-producing machinery in Actinobacillus succinogenes. Furthermore, the resulting strains are evaluated via fermentation on relevant pentose-rich sugar streams representative of those from corn stover. Overall, this work demonstrates genetic modifications that can lead to succinic acid production improvements and identifies key flux determinants and new bottlenecks and energetic needs when removing by-product pathways in A. succinogenes metabolism.

INTRODUCTION

Microbial production of commodity and specialty chemicals from renewable resources offers a promising, sustainable route to establish a viable bioeconomy. Four-carbon dicarboxylic acids have long been identified as particularly interesting precursors for bioproducts (1). Succinic acid (SA), in particular, exhibits properties analogous to petrochemically derived maleic anhydride, a primary precursor to 1,4-butanediol (BDO), unsaturated polyester resins, lubricating oil additives, and an array of other products. The similarity of SA to maleic anhydride, coupled with its natural occurrence as a by-product in microbial fermentation, has made it an attractive target chemical. SA also presents potential for a series of unique product suites, including the biodegradable polyester polybutylene succinate (PBS) (2, 3). Elucidation of the mechanisms governing SA biosynthesis in relevant industrial hosts offers a means to develop hypothesis-driven strain engineering strategies aimed at flux enhancement.

There have been numerous reports of successful SA metabolic engineering pursuits in conventional industrial biocatalysts (for a complete review, refer to references 4, 5, and 6). For example, Escherichia coli has been extensively engineered for SA biosynthetic enhancement via improved substrate transport, increased carbon flux, and removal of by-product biosynthesis (4, 7–9). Similar strategies have been employed in an array of additional hosts, including Corynebacterium glutamicum and Saccharomyces cerevisiae (5, 10, 11). In addition to the aforementioned metabolic engineering strategies, hosts encoding a canonical tricarboxylic acid (TCA) cycle require knockout or inactivation of succinate dehydrogenase, which catalyzes conversion of SA to fumarate, to terminate the oxidative pathway of the TCA cycle at SA under aerobic conditions (8). Additionally, dual-phase aeration (aerobic/anaerobic) is often employed to achieve high-titer fermentations in these engineered strains (4).

Development and deployment of anaerobic strains with high native flux to SA, such as Actinobacillus succinogenes strain 130Z, offer an appealing, alternative foundation for effective biocatalysts. In particular, A. succinogenes is a Gram-negative, capnophilic, facultatively anaerobic, biofilm-forming bacterium with the capacity to convert a broad range of carbon sources to SA as a primary fermentative product, achieving among the highest reported SA titers and yields to date (12). Prior studies have identified core metabolic and SA biosynthetic pathways and optimized fermentation engineering strategies in this organism, taking advantage of its unique incomplete TCA cycle, which natively terminates at SA (Fig. 1) (13–21). A. succinogenes mutants with inactivated pyruvate-formate lyase (PFL) have been generated, effectively eliminating formate biosynthesis (14). Identification and manipulation of additional strain engineering targets offer a means of further enhancing SA flux and resultant productivity; however, to date, there have been few reports of successful metabolic engineering approaches in A. succinogenes, in part due to the limited tractability of the organism; thus, mechanisms governing flux to SA remain to be fully elucidated.

FIG 1.

A. succinogenes metabolism and genetic modifications. Gray arrows indicate native flux from pentose and hexose sugars. Green arrows indicate overexpression targets. Red arrows indicate genetic knockout targets. 13DPG, 3-phospho-d-glyceroyl phosphate; 2PG, d-glycerate 2-phosphate; 3PG, 3-phospho-d-glycerate; 6PGC, 6-phospho-d-gluconate; AC, acetate; ACALD, acetaldehyde; ACCOA, acetyl-CoA; ACTP, acetyl phosphate; DHAP, dihydroxyacetone phosphate; E4P, d-erythrose 4-phosphate; ETOH, ethanol; F6P, d-fructose 6-phosphate; FDP, d-fructose 1,6-bisphosphate; FOR, formate; FUM, fumarate; G3P, glyceraldehyde 3-phosphate; G6P, d-glucose 6-phosphate; GLC, d-glucose; MAL, l-malate; OAA, oxaloacetate; PEP, phosphoenolpyruvate; PYR, pyruvate; R5P, alpha-d-ribose 5-phosphate; RU5P, d-ribulose 5-phosphate; S7P, sedoheptulose 7-phosphate; SUCC, succinate; XU5P, d-xylulose 5-phosphate; XYL, d-xylose; XU, d-xylulose; LAC, d-lactate; pflB, pyruvate formate lyase; ackA, acetate kinase; PCK, phosphoenolpyruvate carboxykinase; MDH, malate dehydrogenase; FUM, fumarase.

In this study, we developed genetic tools for gene overexpression and/or markerless gene knockout in A. succinogenes and employed in silico metabolic modeling to evaluate SA biosynthetic regulation in this promising host. We find that the knockout of competitive carbon pathway targets leads to higher-purity production of SA. Concurrently, upregulation of the reductive branch of the TCA cycle enhances flux to SA, resulting in titer, rate, and yield enhancements when it is cultivated on mock hydrolysate sugar streams rich in xylose. These modifications reveal a finely tuned energetic and redox system and previously unidentified mechanisms of secondary organic acid biosynthesis. Additionally, the resultant strains present promising metabolic engineering strategies for the economically viable and sustainable production of SA from pentose-rich sugar streams.

RESULTS

Genetic tool development.

Prior metabolic engineering efforts in A. succinogenes have employed a positive selection strategy that leveraged the organism's glutamate auxotrophy to generate a series of genetic knockout mutants (14). Here, we report the development of an alternative set of positive selection tools via the conventional utilization of antibiotic resistance markers. Electroporation of A. succinogenes with linear PCR fragments containing genomic homology regions flanking an antibiotic resistance marker enabled homologous recombination-mediated chromosomal integration and gene disruption (Fig. 2); flanking markers with loxP sites enables antibiotic marker removal and recycling via expression of Cre recombinase. Concurrent deployment in combination with a modified version of the previously developed pLGZ920 expression vector (22), encoding a chloramphenicol resistance gene (Fig. 2), enables facile gene overexpression in wild-type and knockout mutant backgrounds. Using these tools, a series of strains with altered carbon flux, through central carbon metabolism and fermentation pathways, were generated to examine alterations in SA biosynthesis (Table 1).

FIG 2.

Example of a knockout cassette, ΔpflB (top) and plasmid vector for the overexpression of SA pathway genes (bottom). A.s. Ori frag, DNA fragment required for replication in Actinobacillus succinogenes.

TABLE 1.

Bacterial strains and plasmids

Strain or plasmid	Description	Reference or source
Strains
Escherichia coli
DH5-α	F− Φ80lacZΔM15 Δ(lacZYA-argF) U169 recA1 endA1 hsdR17 (rK− mK−) phoA supE44 λ− thi-1 gyrA96 relA1	Invitrogen
Actinobacillus succinogenes
130Z	Wild type
ΔpflB::bla	130Z derivative; contains loxP-bla-lox loxP integration replacing pflB; Ampr	This study
ΔpflB	ΔpflB::bla derivative; contains one loxP site	This study
ΔackA	130Z derivative; contains ackA knockout	This study
ΔpflBΔackA	130Z derivative; contains pflB and ackA knockouts	This study
Plasmids
pLGZ920	E. coli-A. succinogenes shuttle vector; Ampr; contains PpckA	22
pL920Cm	pLGZ920 derivative; Cmr; Cmr gene under the control of Pbla	This study
pPCK	pLGZ920 derivative; Ampr; pckA under the control of PpckA	This study
pMDH	pLGZ920 derivative; Ampr; mdh under the control of Pmdh	This study
pFUM	pLGZ920 derivative; Ampr; fum under the control of PpckA	This study
pPMF	pLGZ920 derivative; Cmr; pckA under the control of PpckA; mdh under the control of Pmdh; fum under the control of Pfum	This study
pLCreCm	pLGZ920 derivative; Cmr; cre under the control of PpckA	This study

Knockout of competitive carbon pathway components.

Wild-type A. succinogenes splits carbon flux between two main fermentative pathways, an SA-producing reductive C₄ pathway and a number of C₃ pathways that produce acetic acid, formic acid, and ethanol as by-products (18). Production of SA from glucose and CO₂ requires a net consumption of reducing agents; therefore, biosynthesis of these additional compounds serves, in part, to generate cellular energy and reducing power. However, from a process perspective, these by-products also serve as competitive carbon sinks, potentially reducing carbon yields and/or flux to SA. Additionally, the presence of organic acids other than SA reduces separation efficiency and product purity in downstream recovery of SA from fermentation broth, serving as a negative technoeconomic cost driver.

To examine the effects of the removal of competitive carbon pathways in A. succinogenes, we targeted knockout of genetic components encoding the biosynthetic machinery of the C₃ pathway products acetic and formic acid. Pyruvate formate lyase (pflB), which catalyzes the reversible conversion of pyruvate and coenzyme A (CoA) into formic acid and acetyl-CoA, and acetate kinase (ackA), which catalyzes conversion of acetyl-phosphate and ADP to acetate and ATP (Fig. 1), were chromosomally knocked out to generate “markerless” mutants (ΔpflB and ΔackA, respectively) (Table 1). We refer to these deletions as markerless but point out that, while the drug resistance marker was removed, a residual “scar” remains in the chromosome at the site of the deletion as a result of the recombination event between the loxP sites. Additionally, a pflB ackA double mutant was generated via iterative chromosomal integration and marker removal (strain ΔpflB ΔackA).

Wild-type and mutant A. succinogenes strains were cultivated on mock biomass hydrolysate containing a 60-g/liter sugar stream rich in xylose as a carbon source over a 96-hour batch fermentation. The single and double mutants displayed a growth defect, demonstrating an additional 9-hour lag phase relative to that of the wild type (Fig. 3A). Growth rates (μ⁻¹), calculated from the maximum slope before cell aggregation occurred, were 0.48, 0.15, 0.21, and 0.17 for the wild-type, ΔackA, ΔpflB, and ΔpflB ΔackA strains, respectively. An associated delay in the onset and rate of glucose and xylose consumption was also observed, though complete sugar consumption was ultimately achieved in all strains following 96 h of cultivation (Fig. 3B and C). Concerning the other sugars present in the mock biomass, ∼0.4 g/liter arabinose (from an initial 6.5 g/liter) and ∼2.4 g/liter galactose (from an initial 3.5 g/liter) remained at the end of the fermentation in the four strains, demonstrating a degree of preferential carbon utilization as previously observed (data not shown) (23). Delayed growth and sugar consumption also coincided with delayed onset of SA biosynthesis relative to that of the wild type (Fig. 3D). SA titer, yields, and productivities (Fig. 3I, J, and K) decreased significantly in the ΔackA and ΔpflB ΔackA strains compared to that of the wild type (see Table S1 in the supplemental material). Although overall productivities were similar between the ΔpflB and wild-type strains, the maximum instantaneous productivity was significantly lower in the ΔpflB strain (Fig. 3K) due to the initial lag in bacterial growth and, thus, organic acid production. The ΔpflB strain has a reduced effect on metabolism, compared to that of the other strains, due to the partially redundant action of pyruvate dehydrogenase in carrying flux to acetyl-CoA. As shown in Fig. 3G, strains lacking acetate kinase activity compensated for the loss in reducing power by accumulating pyruvate, which resulted in a net production of 2 mol of NADH per mol of glucose. As this is much lower than the 4 mol of NADH produced per mol of glucose with acetate and CO₂ as the final products (Fig. 1), more carbon was drawn to the oxidative metabolic branches, resulting in a lower SA yield. These results indicate that the removal of heterofermentative pathways is insufficient to enhance carbon flux to SA under the cultivation conditions examined here.

FIG 3.

Evaluation of different fermentation and metabolic parameters in wild-type A. succinogenes (130Z) compared to those of the knockout strains ΔackA, ΔpflB, and ΔpflB ΔackA. Scatter plots present bacterial growth (A), utilization of xylose (B), and utilization of glucose (C). (D to H) Acid production: SA (D), acetic acid (AA) (E), formic acid (FA) (F), pyruvic acid (PA) (G), and lactic acid (LA) (H). The bar graphs show SA titers (I), yields and metabolic yields (J), and overall and maximum instantaneous productivities (K). SA “yield” is the ratio of SA (g/liter) and the sugars consumed (g/liter) at the end of the fermentation. SA “metabolic yield” is calculated as the yield considering the dilution factor at the end of the fermentation. The overall productivity was calculated at 96 h in all cases (since sugars were mostly consumed at that time). The maximum instantaneous productivity was observed at 12 h for the control and 48 h for the knockout strains.

Despite delayed growth and onset of SA biosynthesis, effective reduction in acetate and formate was observed in all three mutants (Fig. 3E and F), and nearly complete elimination of acetate and formate was observed in the ΔackA and ΔpflB single mutants, respectively, the latter of which is in agreement with previously reported pflB mutant generation (14). Interestingly, reduction of acetate was observed in ΔpflB mutants, and conversely, reduction in formate was observed in ΔackA mutants, indicative of interrelated and/or interdependent energetic and redox regulation of these biosynthetic pathways. Further, despite removal of the canonical acetate kinase-mediated biosynthetic pathway, acetic acid accumulation was still observed in the ΔpflB ΔackA strain, indicating that an alternative route to acetic acid biosynthesis exists in A. succinogenes. Acetyl-CoA synthetase (acs)-mediated acetic acid biosynthesis, which offers a direct route from acetyl-CoA to acetic acid, does not appear to explain these findings; there is no evidence for the presence of an acs homolog in A. succinogenes based upon genomic query (data not shown). However, A. succinogenes does encode an acetaldehyde dehydrogenase, catalyzing acetaldehyde to acetate conversion with generation of reductant, likely accounting for the low-level acetate production observed in the ΔackA strain. A. succinogenes notably carries a series of genes which may confer acetaldehyde biosynthetic capacity, including aldehyde dehydrogenase (catalyzing direct conversion of acetyl-CoA to acetaldehyde), deoxyribose aldolase, (catalyzing formation of d-glyceraldehyde 3-phosphate and acetaldehyde from 2-deoxy-d-ribose-5-phosphate), and serine hydroxymethyltransferase, which has been shown to possess threonine aldolase activity (interconversion of threonine into glycine and acetaldehyde) in other microbes.

Additional pyruvate accumulation was observed in both ΔackA and ΔpflB single mutants, indicating a flux bottleneck upon removal of the acetate and formate carbon sinks, respectively (Fig. 3G). Conversely, the ΔpflB ΔackA strain showed pyruvate accumulation profiles similar to that of the wild type and yet ultimately displayed complete pyruvate reassimilation capacity following 24 h of cultivation. Notably, this pyruvate reassimilation coincided with lactate production (Fig. 3H); no lactic acid accumulation was observed in the wild-type or single mutant strains, indicating that the double mutant channels more reductant to lactic acid than the wild-type and engineered strains. As lactic acid production involves the oxidation of NADH, this additional product likely explains the reduced SA yield observed in the double-knockout strains. Surprisingly, ethanol was not detected in any fermentation, even though its production also serves to regenerate NAD⁺.

Overexpression of the SA biosynthetic machinery.

As an alternative means of enhancing flux to SA, we next examined overexpression of the SA biosynthetic machinery. Wild-type A. succinogenes channels carbon through the reductive branch of the TCA cycle (Fig. 1), proceeding through oxaloacetate, malate, and fumarate intermediates via the activity of phosphoenolpyruvate carboxykinase (PCK), malate dehydrogenase (MDH), fumarase (FUM), and fumarate reductase, respectively. Importantly, PCK, in part, facilitates the capnophilic activity of A. succinogenes via carboxylation of phosphoenolpyruvate to oxaloacetate. A series of strains overexpressing reductive TCA pathway genes were generated to examine the relative flux impact of these biosynthetic components.

Initially, the three single-gene overexpression strains were compared. The three strains, overexpressing PCK (pPCK), MDH (pMDH), and FUM (pFUM), increased titers (31.3, 34.2, 32.6 g/liter, respectively) and metabolic yields (0.65, 0.71, and 0.67 g/g, respectively) compared to those of the wild-type strain (30.6 g/liter and 0.51 g/g, respectively) (Fig. 4A, B, and C). In addition to the three single-gene overexpression strains, we also generated a strain overexpressing three components of the reductive TCA branch, PCK, MDH, and FUM (pPMF). Considering the significant SA titer enhancement conferred by MDH gene overexpression, the pMDH strain was further compared with wild type and pPMF (Fig. 4D to K). The scatter plots for pPCK and pFUM (whose titers and yields were lower than those observed for pMDH) are presented in Fig. S1 in the supplemental material (with better graphical clarity than in Fig. 4D to K). Growth rates and sugar utilization profiles for the wild-type, pMDH, and pPMF strains were similar (Fig. 4D, E, and F). Additionally, initial onsets of SA biosynthesis were similar for all strains (Fig. 4G). However, the pMDH and pPMF strains displayed a significantly increased SA accumulation capacity relative to that of the wild type, with the pMDH overexpression strain demonstrating the highest titer of 34.2 g/liter, an 11.8% titer enhancement (Fig. 4A). In addition to enhanced SA titers, these two overexpression strains also exhibited higher final titers (≥1 g/liter) of acetic acid than did the wild type (Fig. 4H), indicating that acetic acid biosynthesis is a key component in achieving higher SA concentrations. Strains with enhanced SA flux likely balance the additional need for reducing power with the production of acetate, as the production of acetic acid and CO₂ results in the highest yields of NADH per mol of glucose (Fig. 1). The formic acid concentration was significantly higher in the wild-type strain than in the overexpression strains during the entire cultivation (Fig. 4I). In addition, a greater reduction in formic acid titer was observed in the pPMF overexpression strain relative to that of the wild type thereafter, with levels similar to those found in both the ΔackA and ΔpflB ΔackA strains (Fig. 3F). This indicates that the strain preferentially utilizes PFL over pyruvate dehydrogenase, and further oxidizes formate to CO₂ later in the fermentation. Minimal pyruvate accumulation (<1 g/liter) was observed in overexpression strains (Fig. 4J). Conversely, wild-type A. succinogenes accumulated nearly 6-fold higher pyruvate at 58 h of cultivation. Lactic acid production was observed only in the pPMF strain at low levels and at a single time point (Fig. 4K). Ethanol was not produced by any of the strains.

FIG 4.

Evaluation of different fermentation and metabolic parameters in wild-type A. succinogenes (130Z) and overexpression constructs such as the single-gene overexpression strains pFUM, pPCK, and pMDH and the triple-gene overexpression strain pPFM. The scatter plots show bacterial growth (A), utilization of xylose (B), and utilization of glucose (C). (D to H) Acid production, SA (D), acetic acid (AA) (E), formic acid (FA) (F), pyruvic acid (PA) (G), and lactic acid (LA) (H), for the wild-type, pMDH, and pPFM strains. The scatter plots corresponding to pMDH and pPFM strains are shown in Fig. S1 in the supplemental material. The bar graphs show (I) SA titers, (J) yields and metabolic yields, and (K) overall and maximum instantaneous productivities for the five strains. SA “yield” is calculated as the ratio of SA (g/liter) and the sugars consumed (g/liter) at the end of the fermentation. SA “metabolic yield” is calculated as the yield but considering the dilution factor at the end of the fermentation. The overall productivity was calculated at 96 h in wild type, pMDH, and pPFM and at 72 h (and marked with an asterisk) in pFUM and pPCK. The maximum instantaneous productivity was observed at 12 h for the control, 25 h for pMDH and pPFM, 21 h for pFUM, and 24 h for pPCK.

Titer and yield enhancements were significantly higher for the pMDH and pPMF overexpression strains than for the wild-type strain (Fig. 4A and B; see also Table S1 in the supplemental material). However, overall and maximum instantaneous productivities were similar for the wild-type and engineered strains. Flux redirection, therefore, is more effective in A. succinogenes by pulling carbon toward reduction pathways via gene overexpression, as opposed to forcing carbon into reduction pathways via markerless gene knockout of the oxidative branch of metabolism. Interestingly, expression of MDH alone generated the maximum SA titer, yield, and production rate (Fig. 4), indicating the necessity for and sufficiency of biosynthetic enhancement and implicating oxaloacetate to malate conversion as the rate-limiting step in the reductive TCA cycle of A. succinogenes. Carbon and electron balance analyses further validate these findings (see Tables S2 and S3 in the supplemental material), demonstrating nearly complete closure of top-performing strains, with equivalent product-to-biomass distribution ratios and nearly complete metabolite identification.

Concurrent SA biosynthesis gene overexpression and competitive carbon pathway knockout.

Lastly, we examined the effects of combinatorial engineering, incorporating pPMF overexpression into all three mutant backgrounds (strains ΔpflB, ΔackA, and ΔpflB ΔackA). Considering the similar performances of the pPMF and pMDH strains, pPMF was selected for further combinatorial engineering to ensure maximum flux through the reductive branch of the pathway in combination with knockouts in the oxidative branch. As with the above-described knockout mutant strains (Fig. 3), a growth defect and concurrent delay in SA onset were observed (Fig. 5A). The growth rates (μ⁻¹) were 0.48, 0.10, 0.14, and 0.13 for the wild-type control, as well as the ΔackA, ΔpflB, and ΔpflB ΔackA strains carrying pPMF, respectively. A uniform lag in xylose consumption was observed for all the engineered strains relative to that of the wild type (Fig. 5B). Glucose consumption was also delayed in all three strains, with pPMF overexpression in the ΔpflB background displaying the most severe glucose consumption defect (Fig. 5C). Despite this decreased rate of glucose consumption, the ΔpflB strain carrying pPMF displayed the highest final SA titer among the combinatorially engineered strains (Fig. 5D). Regarding maximum SA production rates, the wild type produced SA at 1.75 g/liter/h between 8.5 and 12 h, followed by the ΔackA strain carrying pPMF, which produced SA at rates of 0.98 g/liter/h between 48 and 56 h (Fig. 5D). The ΔpflB and ΔpflB ΔackA strains carrying pPMF produced SA at maximum rates of 0.38 g/liter/h between 24 and 34 h and 56 and 72 h, respectively (Fig. 5D). Combined, these data indicate that xylose consumption is a sufficient driver for SA biosynthesis and further support a critical role of acetate coproduction in SA biosynthesis in A. succinogenes.

FIG 5.

Evaluation of different fermentation and metabolic parameters in wild-type A. succinogenes (130Z) compared to those of the ΔackA, ΔpflB, and ΔpflB ΔackA strains carrying pPMF. The scatter plots show bacterial growth (A), utilization of xylose (B), and utilization of glucose (C). (D to H) Acid production: SA (D), acetic acid (AA) (E), formic acid (FA) (F), pyruvic acid (PA) (G), and lactic acid (LA) (H). The bar graphs show SA titers (I), yields and metabolic yields (J), and overall and maximum instantaneous productivities (K). SA “yield” is calculated as the ratio of SA (g/liter) and the sugars consumed (g/liter) at the end of the fermentation. SA “metabolic yield” is calculated as the yield but considering the dilution factor at the end of the fermentation. The overall productivity was calculated at 96 h in the wild-type strain and at 144 h in the rest of the strains. The maximum instantaneous productivity was observed at 12 h for the control, ∼72 h for the ΔackA and ΔpflB strains carrying pPMF, and 98 h for the ΔpflB ΔackA strain carrying pPMF.

The acetic acid production rate dramatically lagged in the mutant backgrounds, as observed in mutants in the absence of MDH overexpression. Notably, acetic acid accumulation was again observed in all ackA mutant backgrounds, further supporting an alternative route to acetic acid in A. succinogenes (Fig. 5E). As observed in mutant and overexpression strains described above, a major formate reduction was also observed in all combinatorial strains (Fig. 5F). Similarly, an initial reduction (or complete elimination) in pyruvate was observed in combinatorially engineered strains relative to wild type, although the ΔpflB strain carrying pPMF accumulated pyruvate following 24 h of cultivation to a higher level than that observed for the wild type. In fact, the wild-type strain displayed some pyruvate reassimilation capacity, whereas the ΔpflB strain carrying pPMF did not (Fig. 5G). Lactic acid was generated in both the acetate kinase and double mutant backgrounds with pPMF overexpression (Fig. 5H). This notably differs from the mutant backgrounds with native SA biosynthetic machinery intact (no overexpression of reductive TCA components), demonstrating a fine-tuned interplay between organic acid biosynthesis and TCA cycle flux; knockout of both ackA and pflB is required for lactate production in the absence of additional flux alterations, whereas ackA knockout is sufficient for lactate production when concurrent TCA component overexpression is employed.

Final SA titers were slightly enhanced (∼4%) in the ΔpflB strain carrying pPMF relative to that of the wild type, but increases were not significant; thus, those increases had no impact on yield (Fig. 5I and J; see also Table S1), again underscoring the critical role of overexpression of reductive TCA components in enhancement of SA accumulation. However, productivity was significantly decreased in all combinatorially engineered strains. Overall productivity was calculated at the time that all the sugars were totally consumed, corresponding to 96 and 144 h for the wild-type and combinatorially engineered strains, respectively. Therefore, although the SA titers were similar at 144 h (Fig. 5I), the productivity was lower for the engineered strains (Fig. 5K). A similar effect was observed in the maximum instantaneous productivity. In this case, the wild-type production peaked at 12 h, while single mutant backgrounds peaked at 72 h, and the double mutant background peaked at 98 h; thus, these productivities were significantly lower in the engineered strains.

DISCUSSION

Biocatalytic production of industrial platform chemicals from renewable substrates offers a promising means of generating petrochemical alternatives. Biological production of SA from lignocellulosic substrates, in particular, has the potential to displace a significant fraction of petroleum-derived maleic anhydride, serving as a potentially high-volume functional replacement in the production of an array of biopolymers and biomaterials. To this end, we have targeted the development of biocatalysts with enhanced SA biosynthetic capacity via metabolic engineering strategies focused upon removal of alternative carbon sinks and/or enhanced biosynthetic pathway flux. The resulting strains demonstrate a number of favorable characteristics, including enhanced titers and yields of SA, as well as reduction of alternative fermentative products, all of which will serve to enhance bioprocess economics (24). Notably, overexpression of MDH was both necessary and sufficient to enhance significantly the biosynthetic capacity. It can be also noted that higher SA titers have been previously reported from similar streams with the native species (23). The current study did not aim to optimize bioreactor cultivation to achieve high SA titers but, rather, was intended to provide a deeper understanding of A. succinogenes metabolism, pave the way for further genetic modifications, and report yield enhancements from pentose sugar substrates by engineered strains relative to the native strain.

In addition to the applied biological significance of these findings, a series of fundamental observations have also emerged through these metabolic engineering efforts. Notably, the interplay between formic and acetic acid biosynthesis in redox and energetic balance has been highlighted, and biosynthesis of the latter has been implicated as a key contributor to SA productivity. Additionally, we have observed acetic acid biosynthetic capacity in A. succinogenes mutants lacking a canonical acetate kinase gene, indicating the use of an alternative biosynthetic route, likely mediated via acetaldehyde dehydrogenase. We also identified genetic and/or flux modifications leading to coproduction of lactic acid. McKinlay et al. reported, that A. succinogenes does not produce lactate and that its genome encodes a single lactate dehydrogenase, which might be involved in oxidation rather than lactate generation (13). The finding in the current paper underscores the fine-tuned nature of this organism's metabolic regulation.

We further note that a number of the observations in this study are applicable to batch fermentation processes. Considerations associated with growth lags and delays in the onset of SA production may largely be mitigated by employing a continuous fermentation process, which has successfully been employed by multiple investigators for this microbe (25, 26). Indeed, though acetate production proved critical for maintaining growth rate integrity and early onset SA biosynthetic capacity, a continuous process may benefit from deployment of the ΔpflB combinatorial strain carrying pPMF, wherein steady-state SA production is relatively unaffected compared to that of the wild type, yet heterofermentative products are significantly reduced, enabling more facile inline separations. To this end, future studies will target examination of engineered biocatalysts in a continuous fermentation configuration, similar to our previous work (26).

There is also significant opportunity for additional strain engineering; modification of the pentose phosphate pathway, recently shown to be altered under active SA biosynthesis (27), may offer a logical upstream target for flux enhancement. Overexpression of phosphoglucose dehydrogenase in wild-type and pflB mutant backgrounds resulted in limited biosynthetic enhancement (see Fig. S3 in the supplemental material); complete pathway overexpression may be required. It is also worth noting that the results presented in the current study might be highly correlated with the carbon source employed. On this note, Guettler et al. patented a recombinant A. succinogenes strain that exhibited enhanced SA production via overexpression of the glucose-6-phosphate dehydrogenase (G6PDH) from glucose or sorbitol (28). However, in the present study, no productivity enhancement was observed via G6PDH overexpression when employing xylose-rich streams (see Fig. S3). Additionally, alternative engineering strategies targeting generation of reductant and/or ATP may offer a means for substituting for biosynthesis of heterofermentative by-products. Complementary multilevel omic analyses of the above strains also offer a means to elucidate additional metabolic alterations and strain-engineering targets. Lastly, we note that our prior results have indicated a significant yield enhancement when using real hydrolysate compared to that when mock hydrolysate is used (23); future analyses will examine yield enhancements on biomass-derived sugar streams.

MATERIALS AND METHODS

Construction and validation of a core-carbon metabolic model of A. succinogenes 130Z.

A core-carbon metabolic model of A. succinogenes strain 130Z was created following the genome annotation provided in reference 13. The model consists of 51 intracellular metabolites, 20 extracellular metabolites, and 89 mass-balanced reactions. Pathways for the metabolism of glucose, xylose, galactose, and arabinose were included based on annotated pathways (13) and stoichiometry from the MetaCyc database (29). The model was used to infer the unmeasured CO₂ fixation and production from the distribution of observed products and known pathway stoichiometry. This accounting allowed the overall carbon mole balance (see Table S2) and electron balance (see Table S3) to be calculated for the control, pPMF, and pMDH strains. To analyze both bacterial biomass and carbon content in the latter strains (wild-type, pPMF, and pMDH strains), the fermentation broths and the bacterial biofilms from the vessel and impellers were recovered separately at the end of the fermentation and centrifuged at 10,000 rpm for 15 min. The supernatants were discarded, and the cells were washed twice with distilled water. Then, the cells were freeze-dried and weighed. Carbon content of dry-cell biomass was measured with a LECO TruSpec CHN determinator using high-temperature combustion in pure oxygen, followed by infrared (IR) analysis of H₂O and CO₂. The gas was then reduced and scrubbed to measure nitrogen content via a thermal conductivity detector. Mass balances did not account for the presence of yeast extract or corn steep liquor in the culture medium.

Microorganism and growth conditions.

For seed cultivation, A. succinogenes 130Z (ATCC 55618) (12) was routinely cultivated anaerobically in 50 ml of tryptic soy broth (TSB) supplemented with 10 g/liter glucose (Fluka Analytical, India) at 37°C and 120 rpm. Serum bottle cultures were inoculated at an optical density at 600 nm (OD₆₀₀) of 0.1 with plate-harvested biomass.

Plasmid and strain construction.

Q5 Hot Start high-fidelity polymerase and 2× master mix (New England BioLabs, MA) and primers synthesized by Integrated DNA Technologies (IDT) were used in all PCR amplifications for plasmid construction. Plasmids were constructed using conventional restriction enzyme subcloning or Gibson Assembly master mix (New England BioLabs, MA) following the manufacturer's instructions. Plasmids were transformed into competent Zymo 5-alpha E. coli (Zymo Research, CA) following the manufacturer's instructions or electroporation into A. succinogenes. Transformants were selected on LB or TSB plates containing 10 g/liter glucose supplemented with either 50 μg/ml chloramphenicol or 50 μg/ml ampicillin and grown at 37°C. The sequences of all plasmid inserts were confirmed using Sanger sequencing performed by GENEWIZ, Inc.

Construction of markerless knockouts in A. succinogenes 130Z.

To knock out pflB and ackA in the genome of 130Z via homologous recombination, a knockout cassette was constructed (Fig. 2). DNA fragments, including 1.4-kb and 1.5-kb up- and downstream regions of target genes and an ampicillin resistance gene (bla) flanked by loxP sequences, were individually amplified by PCRs using primer sets indicated in Table 2. Three fragments were assembled using a Gibson Assembly Cloning kit (New England BioLabs, MA). The cassette was used to transform the A. succinogenes host using electroporation, and integrants were selected on TSBG plates supplemented with ampicillin. Integration was verified by PCR analysis and DNA sequencing. The ampicillin resistance gene in the pfl KO integrants was removed by transformation with a plasmid expressing Cre recombinase (pL920CreCm). A plasmid curing procedure was performed to further remove the pL920CreCm from the Amp-sensitive KO integrants, involving daily subculturing of integrants in liquid TSBG for over 25 generations followed by plating on TSBG agar and identifying the Amp-sensitive and Cm-sensitive colonies.

TABLE 2.

Primers used in this study

Primer	Sequence (5′ to 3′)	Usage
YC-1	CGTTAACCGTGGGAATCAGTTTGTTAGGAATG	pflB up fragment; Gibson Assembly
YC-2	CGAAGTTATTGTAATACTTCCTTTTGCTAGTATTGATAATGAAATCCTGTAAG	pflB up fragment; Gibson Assembly
YC-3	CCATAACTTCGTATAATGTATGCTATACGAAGTTATTTGGGGTAACGTAATAAAAATG	pflB down fragment; Gibson Assembly
YC-4	TCTCTCCTTCGCGGAATAAAATATCCACTTC	pflB down fragment; Gibson Assembly
YC-5	CTAGCAAAAGGAAGTATTACAATAACTTCGTATAATGTATGCTATACGAAGTTATAATTCTTGAAGACGAAAGGGCCTCGTG	bla-loxP fragment; Gibson Assembly
YC-6	CGTATAGCATACATTATACGAAGTTATGGGGTCTGACGCTCAGTGGAACGAAAACTC	bla-loxP fragment; Gibson Assembly
YC-7	GCAGCAATAGAGGAAACACGGTTTG	pckA fragment
YC-8	GGATTTGGTACCGTGCCGGCGGCCTAATAACCTG	pckA fragment, KpnI site underlined
YC-9	CGAACCGAAGCGTTCCTGCGCGAGTAACGC	mdh fragment; blunt-end ligation
YC-10	GCTTCCCATTAATCAAACGGCGG	mdh fragment; blunt-end ligation
YC-11	CTCAAACAAACCGTGTTTCCTCTATTGCTGC	pLGZ920 to linearize for mdh cloning
YC-12	AATTATCAATGAGGTGAAGTATGACATTTCGTATTGAAAAAGACAC	fum CDS fragment; Gibson Assembly
YC-13	GAATTCGAGCTCGGTACCCGGGGATCCCTGACCGTCTTCGGTGAATACTGATATAG	fum CDS fragment; Gibson Assembly
YC-14	ACTTCACCTCATTGATAATTTAAAATTAAAAATCC	pLGZ920 to linearize for fum cloning
YC-15	GGATCCCCGGGTACCGAGCTCGAATTCACTG	pLGZ920 to linearize for mdh or fum cloning
YC-16	CGGTACCGAGCTCGAATTCACTGGCCGTCG	pPCK to linearize for mdh and fum cloning
YC-17	GTCATCTTAACAGGTTATTAGGCCGCCGGCA	pPCK to linearize for mdh and fum cloning
YC-18	CCTTCGGCCGGCCCTGCCGTTTCGGAAAACTCACGCTTTACCCG	fum fragment; FseI site underlined
YC-19	CCTTCGGCCGGCCCATAAAGAATCCAAGATAAACGAATTGGC	fum fragment; FseI site underlined

Construction of gene overexpression strains of A. succinogenes 130Z.

All plasmids were constructed based on an E. coli-A. succinogenes shuttle vector pLGZ920 (ATCC PTA-6140 [22] and its derivative pL920Cm [Fig. 2]). All overexpression plasmids are listed in Table 1. PCR using primers (Table 2) was used to generate all the DNA fragments. Restriction enzyme digestion or Gibson Assembly techniques were incorporated in the cloning of the genes. For single-gene overexpression constructs (pPCK, pMDH and pFUM), genes encoding PEP carboxykinase (pckA), malate dehydrogenase (mdh), and fumarase (fum) from A. succinogenes 130Z were independently cloned in pLGZ920. The promoter of pckA (PpckA) was used to drive the expression of pckA and fum whereas mdh was under the control of its own promoter. In the three-gene overexpression constructs, pPMF, pckA, mdh, and fum were under the control of their respective native promoters.

Electroporation of A. succinogenes.

Plasmids and/or linear DNA cassettes were transformed into A. succinogenes via electroporation using Gene Pulser Xcell (Bio-Rad, CA). Electro-competent cells were prepared by harvesting the exponential-growth-phase cells (optical density at 600-nm wavelength = 0.4 to 0.5) and centrifuging at 4°C, 3300 × g for 10 min. Cells were washed twice in ½ volume of ice-cold 15% glycerol and concentrated 100× to 150× before electroporation. One hundred microliters of competent cells was mixed with 2 to 15 μl DNA in a 0.2-cm gap cuvette and electroporated at 2.5 kV, 25 μF, and 600 Ω. One-milliliter TSBG was added to the cuvette after electroporation. Cells were transferred to a microtube and incubated at 37°C for 1 h before plating on TSBG agar plates supplemented with appropriate antibiotics. Plates were incubated at 37°C for 1 to 2 days or until colonies were observed.

Preparation of fermentation seed culture and fermentation medium.

A. succinogenes strains were anaerobically grown in 150-ml sterile capped bottles containing 50 ml of TSBG and antibiotic (30 mg/liter chloramphenicol or 100 mg/liter ampicillin, as indicated in Table 1) (excluding the control strain, 130Z), and incubated overnight at 37°C and 200 rpm. Cells were inoculated in the fermentor at an initial OD₆₀₀ of 0.05. To ensure anaerobic fermentation, CO₂ was sparged overnight before bacterial inoculation.

The medium used for fermentations consisted of (per liter) 6 g yeast extract, 10 g corn steep liquor (Sigma-Aldrich, USA) (prepared as described below), 0.3 g Na₂HPO₄, 1.4 g NaH₂PO₄, 1.5 g K₂HPO₄, 1 g NaCl, 0.2 g MgCl₂·6H₂O, and 0.2 g CaCl₂·2H₂O. Corn steep liquor was prepared at a concentration of 200 g/liter (20×) and then boiled at 105°C for 15 min (25). After cooling, solids were separated, and the supernatant was autoclaved and used as nutrient source. As a carbon source, a mixture of sugars mimicking the concentration in real biomass hydrolysates (in particular, deacetylated and diluted acid pretreated hydrolysate [DDAPH]) (23) was utilized. The final sugar concentration was 60 g/liter and consisted of 6.5 g/liter glucose, 44 g/liter xylose, 3.5 g/liter galactose, and 6.5 g/liter arabinose. In addition, acetic acid (1.7 g/liter) was supplemented to the medium since DDAPH contains this acid. Appropriate antibiotic was added to the fermentor prior to inoculation.

Fermentation conditions.

Fermentations were performed in 0.5-liter BioStat-Q Plus fermentors (Sartorious) with 300 ml of medium. The pH was maintained at 6.8 by supplementing 4 N NaOH. The temperature was controlled at 37°C and the agitation at 300 rpm. During the fermentation, CO₂ was sparged at 0.1 vvm. Samples (∼1 ml) from the fermentations were taken under aseptic conditions at various time points to follow bacterial growth, sugar consumption, and the production or uptake of acids (e.g., SA, formic acid, acetic acid, lactic acid, and pyruvic acid) or ethanol. All the fermentations were performed at least in duplicate.

Analytical methods.

Bacterial growth was tracked by OD₆₀₀ in a spectrophotometer, as a measurement of cells in suspension in the fermentation broth. Growth rates (μ⁻¹) were calculated from the maximum growth slope before cell aggregation occurred. Samples were then filtered through a 0.2-μm syringe filter before placing them in high-pressure liquid chromatography (HPLC) vials to analyze carbohydrates (glucose, xylose, arabinose, and galactose), organic acids (SA, formic acid, acetic acid, lactic acid, and pyruvic acid), and ethanol. Carbohydrate HPLC analysis was performed by injecting 6.0 μl of 0.2-μm-filtered culture supernatant onto an Agilent 1100 series system equipped with a Phenomenex Shodex SUGAR 7u SP0810 20A Column 300 by 8 mm plus an anion guard column and cation guard column (Bio-Rad Laboratories) at 85°C, using a mobile phase of Nanopure water at a flow rate of 0.6 ml/min and a refractive index detector for detection. HPLC analyses for ethanol and all organic acids, except for pyruvic acid, were performed by injecting 6.0 μl of 0.2-μm-filtered culture supernatant onto an Agilent 1100 series system equipped with a Bio-Rad HPLC Organic Acid Analysis Column, Aminex HPX-87H Ion Exclusion Column 300 by 7.8 mm and a cation H⁺ guard column (Bio-Rad Laboratories) at 55°C, using a mobile phase of 0.01 N sulfuric acid at a flow rate of 0.6 ml/min and a refractive index detector for detection. Pyruvic acid HPLC analysis was performed by injecting 6.0 μl of 0.2-μm-filtered culture supernatant onto an Agilent 1100 series system equipped with a Phenomenex Rezex RFQ-Fast Acid H⁺ (8%) column 100 by 7.8 mm and a cation guard cartridge (Bio-Rad Laboratories) at 85°C, using a mobile phase of 0.01 N sulfuric acid at a flow rate of 1.0 ml/min and a diode array detector at 315 nm for detection. Analytes were identified by comparing retention times and spectral profiles with pure standards.

Calculation of SA yields, titers, and productivity.

SA yield is calculated as the ratio of SA (g) and total sugar consumption (g) at the end of the fermentation. Metabolic yield is calculated similarly to SA yield but correcting substrate and product concentrations with the dilution produced from base addition. In addition, compensation was also made for the removal of substrate and products via sampling (considering 1 ml as sample volume). SA titer (g/liter) is the SA concentration at the end of the fermentation, and values are not corrected by the dilution factor. Overall productivity (g/liter/h) is calculated as SA production (g/liter) divided by the time (h) at the end of the fermentation. The end of the fermentation is considered when total sugar concentration is close to zero. Maximum instantaneous productivity (g/liter/h) is the maximum productivity peak value observed during the fermentation course. A statistical analysis (a nonpair homoscedastic t test with a 95% confidence level) was also performed for titers, yields, and rates to ascertain the level of confidence for the differences between the wild-type and the engineered strains. In addition, as the wild-type strain was biologically replicated at least four times, the estimated precision for each parameter, considering the standard deviation for each value, was also calculated.