Metabolic engineering of Zymomonas mobilis for 2,3-butanediol production from lignocellulosic biomass sugars

Investigate 2,3-BDO toxicity to Z. mobilis

First, we examined the feasibility of engineering Z. mobilis for high 2,3-BDO production by investigating the toxicity of 2,3-BDO to Z. mobilis in RMG2 medium at 30 °C using the Bioscreen C high-throughput growth measurement instrument. Our result showed that Z. mobilis tolerates higher concentration of 2,3-BDO than that of ethanol (Fig. 1). The growth rate decreased to 0.3 h⁻¹ when about 80 g/L 2,3-BDO was supplemented to the growth medium, whereas less than 45 g/L ethanol was needed to slow the growth of Z. mobilis to similar growth rate. When 80 g/L ethanol was added into the medium, the growth rate of Z. mobilis was only about one-third of that observed for the same amount of 2,3-BDO added (Fig. 1). Even when the concentration of 2,3-BDO was increased to 100 g/L, the growth rate of Z. mobilis decreased to 0.25 h⁻¹, but it was still more than half of the control without 2,3-BDO supplementation. The low toxicity of 2,3-BDO to Z. mobilis indicates that Z. mobilis has a strong potential for achieving high titers, rate, and yields for 2,3-BDO production.

Construct a minimized vector for pathway engineering

Identify and select 2,3-BDO pathway genes for heterologous 2,3-BDO pathway construction

Literature related to 2,3-BDO production was reviewed, which includes the analysis of native 2,3-BDO biosynthesis pathways in addition to the expression of heterologous enzymes for 2,3-BDO production [1, 2, 5, 10, 12, 13, 15, 16, 19, 20, 23, 25, 27, 68, 69]. The 2,3-BDO pathways from Bacillus licheniformis and Enterobacter cloacae have been successfully expressed in E. coli [13, 16, 30, 33, 70, 71], suggesting the feasibility of 2,3-BDO production in Z. mobilis. The gene combination resulting in high 2,3-BDO yields and experimental methods reported in these papers laid a solid foundation for our work. We then compiled a list of 2,3-BDO pathway genes and the promising source microorganisms. After eliminating pathogenic organisms from our list, we chose two bacterial species, B. licheniformis and E. cloacae, for the source of our three-gene 2,3-BDO biosynthesis pathway reported here.

The sequences (nucleotide and protein) of three genes (als, aldC, and bdh) were extracted and BLASTed against the Z. mobilis genome including our recently updated native plasmid sequences. BLAST results indicate that Bdh only has homologs with low similarity in Z. mobilis; therefore, heterologous bdh genes were included in our pathway engineering strategy. In addition, the reaction that Bdh catalyzes is a reversible reaction, with some enzymes preferring NADPH as a cofactor. Bdh enzymes preferring the reduction reaction from acetoin to butanediol using NADH were selected for minimizing redox imbalance and for achieving high 2,3-BDO yield.

Additional BLAST results for the other two enzymes required for 2,3-BDO production indicate that Z. mobilis possesses three homologs to Als, of which ZMO1139/ZMO1140 (IlvB/C) may form a complex contributing to amino acid biosynthesis. The homolog ZMO0687 has a high similarity to the catabolic enzyme, Als. The high similarity between Als, IlvB (ZMO1139), and Pdc (ZMO1360) based on protein sequence alignment indicates that there could potentially be competition among them for substrate (pyruvate) binding and utilization. Additionally, since Pdc in Z. mobilis is required for ethanol production, providing energy, and maintaining redox balance for robust cellular growth, strategies to shift the carbon flux toward 2,3-BDO production are required. This can be accomplished by inhibiting pdc gene expression and Pdc enzyme activity or by enhancing the performance of the heterologous Als enzyme.

Gene expression patterns of the Als homologous gene in Z. mobilis were subsequently examined using public and in-house transcriptomic datasets of microarray and next-generation sequencing (NGS)-based RNA-Seq. The result indicated that the als gene homologs are abundantly expressed under different conditions, and that ZMO0687 is also differentially expressed in cells grown from media containing different carbon sources (e.g., downregulated in xylose medium compared to glucose). Furthermore, ZMO1141, which catalyzes the formation of (R)-2,3-dihydroxy-3-methylbutanoate from α-acetolactate (hereafter referred to as acetolactate) for valine, isoleucine, and leucine biosynthesis, has an abundant transcript level of average log₂-based microarray intensity result of 13.78 compared to 14.69 for pdc (Additional file 1: Table S1B); therefore, the acetolactate decarboxylase (AldC) to be engineered should have a strong affinity for the substrate acetolactate to compete with ZMO1141, thus ensuring carbon flux toward acetoin and 2,3-BDO.

Construct heterologous 2,3-BDO pathway in Z. mobilis and optimize flask fermentation conditions

The gene expression cassettes containing the sequences of three codon-optimized 2,3-BDO pathway genes (als, aldC, and bdh) from B. licheniformis and E. cloacae as well as wild-type version from E. cloacae were then designed and synthesized (Table 2). As discussed above, Z. mobilis has three als gene homologs, but does not have the aldC and bdh genes needed for 2,3-BDO production. Additionally, since Z. mobilis Pdc is a very abundant and efficient enzyme with low Km (<0.31 mM) and high k _cat (>100 1/s), a strong als gene may be needed to divert carbon flux from ethanol production to 2,3-BDO. We, therefore, constructed several pathways with either two genes (aldC and bdh) or three genes (als, aldC, and bdh). These genes were assembled using the Biobrick-based metabolic pathway engineering strategy and cloned into the minimized shuttle vector pEZ (Table 1). The results indicated that the three-gene pathways produced more 2,3-BDO than did the two-gene pathways (Table 2; Fig. 2).

Strain	Plasmid	Promoter	Species	Enzymes	Codon optimization	BDO (g/L)
8b-GS6	pEZ-GS6	Wild-type	E. cloacea	AldC, Bdh	No	1.42
8b-GS2	pEZ-GS2	Pgap	E. cloacea	AldC, Bdh	Yes	3.69
8b-BC3	pEZ-BC3	Pgap, Peno	B. lincheniformis	AldC, Bdh	Yes	2.88
8b-BC4	pEZ-BC4	Wild-type	E. cloacea	Als, AldC, Bdh	No	1.9
8b-BC5	pEZ-BC5	Peno, Pgap	E. cloacea	Als, AldC, Bdh	Yes	5.0

Culture condition: 40 mL RMG8 in 125-mL flask, 33 °C, 120 rpm

EC E. cloacea; BL B. licheniformis

Three constructs (vector control pEZ, two-gene pathway plasmid pEZ-GS2, and three-gene pathway plasmid pEZ-BC5) were further compared in two strain backgrounds of 8b or 22C (Table 1). The introduction of heterologous als gene significantly increased 2,3-BDO production by shifting carbon flow away from ethanol production (Fig. 2).

Clearly 2,3-BDO production could be significantly enhanced if the native ethanol production pathway was impeded by knocking out key ethanol production genes of pyruvate decarboxylase (pdc) and alcohol dehydrogenase (adh), which has been demonstrated in S. cerevisiae that 2,3-BDO production increased in Pdc-deficient mutants [5, 7–10]. We have attempted to knock out or knock down the pdc gene to divert carbon to 2,3-BDO production without success, a result consistent with previous reports that pdc is essential for native Z. mobilis and cannot be completely knocked out. Furthermore, our efforts testing pdc knockout in 2,3-BDO production strains, such as 9C-BC5 and 9C-BC11, were also not successful. We will continue to work on blocking the route from pyruvate to ethanol to maximize the potential 2,3-BDO titer with balanced redox using similar strategies as reported in yeast [5, 6, 72–74].

To optimize the 2,3-BDO production, different flask fermentation conditions were first investigated using the three-gene pathway construct in the 22C background (22C-BC5) and using the empty vector pEZ as a negative control (Table 1; Fig. 3a). Results indicate that 2,3-BDO can be produced using mixed sugar conditions (glucose, or glucose and xylose), as well as from DMR-pretreated biomass (Fig. 3b). Growth conditions that varied the shaking speed and medium volume affecting oxygen dispersion impacted acetoin and 2,3-BDO production (Fig. 3b). For example, acetoin titers rose above 10 g/L when the shaking speed was increased to 200 rpm with 40 mL medium in the 125-mL baffled flask (Fig. 3b).

Optimize heterologous 2,3-BDO pathway to increase 2,3-BDO titer

Initial 2,3-BDO engineering efforts in introducing heterologous aldC and bdh genes into Z. mobilis resulted in the 2,3-BDO production of about 3.7 g/L, which confirms that both aldC and bdh genes are required for 2,3-BDO production in Z. mobilis. Furthermore, the addition of a heterologous als gene from E. cloacae increased 2,3-BDO titers to about 5.0 g/L (Table 2), which again upheld the assumption that an additional als gene may be needed to redirect carbon flux from ethanol production into 2,3-BDO to compete with the abundant and highly active Z. mobilis Pdc enzyme. Our data thus suggested that the bottleneck reactions to increase 2,3-BDO production could be acetolactate generation from pyruvate and 2,3-BDO production from acetoin. We hypothesize that stronger als and bdh genes, or genes encoding more active enzymes, will boost BDO production.

To ascertain whether other heterologous als genes with higher specific enzymatic activities than E. cloacae one as reported in E. coli [16] could improve carbon flux from ethanol production into 2,3-BDO, als genes from B. licheniformis (BlAls) and B. subtilis (BsAls) were synthesized and cloned. These included one codon-optimized BlAls gene and two versions of the BsAls genes (BsAls1 and BsAls2) with different codon optimization outputs using Genscript’s codon optimization algorithm. The attempt to incorporate the als gene from B. lichenformis (BlAls) under control of the strong Peno promoter along with the expression of AldC and Bdh using the Z. mobilis Pgap strong promoter was unsuccessful. Besides the possibility of cell growth inhibition by the enzyme Als from B. licheniformis and B. subtilis, this could be due to the protein burden caused by strong heterologous pathway gene expression. About 50 % of the proteins in Z. mobilis are involved in the glycolysis pathway [75]. This view is supported by a previous report that overexpression of plasmid-encoded protein led to the reduction of both glycolytic flux and growth rate due to protein burden [76]. Similarly, previous work also suggested that protein burden could be the reason for the negative effect observed for the isopropyl β-d-thiogalactoside (IPTG) induction of 2,3-BDO biosynthesis in a recombinant E. coli strain [26].

To test this speculation and avoid the potential metabolic burden caused by strong heterologous gene expression, the strong Peno promoter was tested along with the Ptet inducible promoter for the construction of additional three-gene pathway constructs. These constructs also included a common aldC and bdh gene operon from E. cloacae driven by a strong promoter Pgap [Pgap-(EcALDC-BDH)]. In total, six new 3-gene constructs with different versions of als genes from B. lincheniformis or B. subtilis were generated (BsAls1, BsAls2, and BlAls that were either driven by the strong Peno promoter or by the inducible Ptet promoter).

We further compared sugar consumption and 2,3-BDO production kinetics in glucose only or mixed sugars (glucose and xylose) with the new strain producing the highest 2,3-BDO titers, 9C-BC11 (Table 1). We found that this strain is stable, and 2,3-BDO titers were more than 13 g/L when grown in RMG8 or RMG8X6 media (Fig. 4). Notably, the addition of xylose delayed the sugar utilization of glucose and xylose; as well as cellular growth (Fig. 4). Furthermore, xylose was not completely utilized even 5 days after inoculation (data not shown). However, the concentration of 2,3-BDO in the mixed sugar fermentation remained level in stationary phase when compared to cultures grown in glucose only. In this case, levels of 2,3-BDO decreased after growth, possibly because of the reverse activity of Bdh converting 2,3-BDO back to acetoin, since decreasing 2,3-BDO titers was accompanied by in acetoin concentration increase (Fig. 4a).

Investigate fermentation conditions on 2,3-BDO productions

To further understand the impact of oxygen concentration on 2,3-BDO production and fermentation byproduct profiles as indicated by flask fermentation optimization (Fig. 3), Biostat-Q Plus fermenters (Sartorius Stedim North America Inc., Bohemia, NY) with oxygen control capability were used for our highest 2,3-BDO producing strain, 9C-BC11. As a control, shake flask fermentation was also conducted. Glucose is usually consumed within 20 h by the parental strain 9C under anaerobic conditions. Interestingly, the glucose utilization of Z. mobilis 9C-BC11 under anaerobic conditions with N₂ purging was significantly reduced with only 3 g/L glucose consumed and no discernible cellular growth after 22 h post-inoculation. In contrast, more than half of the glucose had been used under shake flask conditions, and nearly all glucose was consumed upon reaching the highest OD_600 nm value of 5.34 under microaerophilic condition of 1 % dissolved oxygen (DO) (Fig. 5a). However, glucose utilization and cellular growth were significantly decreased when 10 % DO was supplied. At 48-h post-inoculation, there was still 34 g/L glucose remaining in the media, compared to other conditions tested where all glucose had been consumed (Fig. 5a), and the glucose level and OD_600 nm values remained relatively steady out to 120 h post-inoculation (Additional file 1: Table S1C).

Under anaerobic conditions with nitrogen sparging, nearly all glucose was ultimately converted into ethanol, producing only 1.68 g/L 2,3-BDO. Increasing oxygen deliveries from 0 to 1 % DO, or 10 % DO, decreased ethanol production dramatically (Fig. 5b), and is consistent with a previous report that aerobic growth reduces ethanol production [63]. Although both 2,3-BDO and acetoin levels increased under aerobic growth, titers of 2,3-BDO and acetoin were significantly higher in the 1 % DO culture than in the 10 % DO culture (Fig. 5b, c).

Besides the major end-products ethanol, acetoin, and 2,3-BDO, the production profiles of minor byproducts of acetate, lactate, and glycerol were also affected by the oxygen levels in these cultures. More acetate was produced under aerobic conditions resulting in less ethanol production as reported in the literature [63]. Furthermore, acetate production was significantly higher in the more aerobic conditions (1 or 10 % DO supplementation) than in the microaerophilic conditions using flask fermentation or N₂-supplemented anaerobic fermentation (Fig. 5d). Additionally, more lactate was produced under aerobic conditions than in anaerobic condition as well (Fig. 5d). We had previously shown that the expression levels of the lactate dehydrogenase gene ZMO1237 were more abundant under aerobic conditions [63], and the expression of another d-lactate dehydrogenase gene, ZMO0256, was upregulated in the ethanol-treated cells during stationary phase [57]. The host strain, Z. mobilis 9C, used in this study, is a derivative of Z. mobilis 8b in which the ZMO1237 gene was inactivated [37]. Therefore, ZMO1237 may play the major role in lactate production under aerobic conditions. However, another lactate dehydrogenase gene, ZMO0256, could also be activated under aerobic conditions for lactate production, when ZMO1237 is deactivated, as shown in this study (Fig. 5d).

Intriguingly, despite the observation that more glycerol was produced under aerobic conditions (i.e., 1 or 10 % DO) than under anaerobic conditions with N₂ purging, a higher amount of glycerol was produced in shake flask fermentations (see Fig. 5c). In Z. mobilis, only one NAD⁺ is regenerated from 2,3-BDO biosynthesis compared to two NAD⁺ from ethanol fermentation. The increase of glycerol production for in shake flasks may be explained by the utilization of the glycerol pathway to maintain redox balance (NAD⁺ feedback) while also provides ATP to sustain cellular growth (Fig. 6).

The highest 2,3-BDO yield (0.16 g/g glucose) was achieved under shake flask fermentations and the highest ethanol yield (0.44 g/g glucose) was obtained under anaerobic conditions using N₂ purging (Additional file 1: Table S1C). As oxygen supplementation increased, acetoin yields also increased; however, these conditions also produced lower ethanol yields. In addition, the yields of all fermentation products (ethanol, lactate, glycerol, and acetate) were reduced with the increasing oxygen levels. For example, ethanol yields decreased from 0.46 g/g glucose under anaerobic conditions to 0.15 g/g glucose in the 10 % DO culture (Additional file 1: Table S1C). Conversely, the ratios of the heterologous products, acetoin and 2,3-BDO, to fermentation products increased from 0.04 g/g glucose under anaerobic conditions to 1.65 g/g glucose when 10 % DO was used, which reflects the higher acetoin yields and also correlated to lower ethanol yields under aerobic conditions (Additional file 1: Table S1C). We saw a similar result with shake flask fermentations. Higher acetoin production was accompanied by lower ethanol titer under more aerobic conditions (40 mL media in a 125-mL flask at 120 rpm) (Fig. 3b).

Bacterial strain and growth conditions

Zymomonas mobilis 8b was revived from frozen glycerol stocks for about 6–8 h in 10 mL RMG2 (20 g/L glucose, 10 g/L yeast extract, 2 g/L KH₂PO₄) at 33 °C. Other 8b derivatives used in this work are: 22C, a hydrolysate-tolerant mutant strain; 9C, an 8b derivative with both chloramphenicol and tetracycline resistance genes removed. The final spectinomycin concentration used for Z. mobilis is 200 and 50 μg/mL for E. coli.

Growth curve measurement using Bioscreen C

Growth of Z. mobilis was monitored by Bioscreen C using the 420–580 nm filter (Growth Curves USA, NJ) as described previously without shaking [88, 89]. Three replicates were used for each condition. The seed cultures used for Bioscreen C were first revived from frozen glycerol stocks overnight in RMG2, then diluted 100-fold into fresh RMG2 until it reached exponential phase. The absorbance at OD_600 nm was adjusted to 3, then added 10-μL to each well containing 290-μL medium, such that the final OD_600 nm = 0.1. Procedures for measurement, recording of final cell densities, and calculations used to correct for non-linear response at high cell densities are previously reported [88, 89].

Minimized shuttle vector construction and 2,3-BDO pathway assembly

A new minimized shuttle vector, pEZ15Asp, was designed and synthesized that includes origins of replication from both E. coli and Z. mobilis, antibiotics marker for the spectinomycin resistance gene, and multiple cloning sites for Biobrick-based pathway assembly [67].

Wild-type and codon-optimized versions of three 2,3-BDO biosynthesis pathway genes were synthesized by Genscript (NJ, USA). The genes were then assembled in different combinations using Biobrick-based approaches. Specifically, the minimized shuttle vector pEZ15Asp (pEZ) was digested with EcoRI-HF and PstI-HF (Bio-Rad, CA), which was treated using the Rapid DNA Dephos and Ligation Kit (Roche, CA) for dephosphorylation and then quantified using Nanodrop and gel electrophoresis followed by gel purification. The insert(s) were also treated with restriction enzymes of EcoRI/SpeI or XbaI/PstI and then quantified again using Nanodrop and gel electrophoresis followed by gel purification. The insert and vector were ligated using the fast-ligation kit (NEB, CA) with a molecular ratio of 3:1 for insert to vector. The ligation product (2-μL) was used for transformation into NEB C2925 competent cells. The transformants were confirmed by colony PCR using the primers of pBAD-GFP_F/R (5′TCACCAGCTCACCGTCTTTC3′ and 5′CCTGATGAATGCTCATCCGG3′) to confirm the insert size. The gene-specific primers were used to confirm that the targeted genes were cloned into the vector. Colonies showing the expected PCR bands were inoculated into LB broth supplemented with 50 μg/mL spectinomycin overnight and the plasmids were extracted and confirmed by PCR, restriction digestion, and Sanger sequencing.

Electroporation transformation and 2,3-BDO strain selection

Zymomonas mobilis or Escherichia coli cells were transformed with plasmids by electroporation (Bio-Rad Gene Pulser, 0.1-cm gap cuvettes, 1.6 kV, 200 ohms, 25 μF). Electrocompetent Z. mobilis cells were prepared by centrifuging cells from cultures that had reached an OD_600 nm of 0.4–0.6. Cell pellets were washed once in ice-cold sterile water, re-centrifuged, and washed again in 10 % glycerol. These pellets were resuspended in 10 % glycerol at a concentration approximately 1000-fold higher than the starting culture. Competent cells were stored at −80 °C as small aliquots for later use. Transformants of E. coli or Z. mobilis were selected on LB or RMG agar plates, respectively, containing appropriate antibiotics. Due to the presence of restriction/modification systems in Z. mobilis [90] that can decrease transformation efficiency, all plasmids were built in and isolated from a methylation-deficient E. coli strain, C2925 (NEB, MA), for efficient transformation into Z. mobilis 8b or its derivatives.

For single colony isolation, transformants grown on the selective plates containing spectinomycin were further streaked on RMG plates with spectinomycin at a final concentration of 200 μg/mL (RMGSp). Following isolation, these colonies were then used for colony PCR to confirm the introduction of plasmids with correct pathway genes using the primers pBAD-GFP_F/R to check the insert size and gene-specific primers. Colonies with expected PCR bands pattern were selected and inoculated into RMGSp for preservation and further flask evaluation.

Lignocellulosic biomass sugar hydrolysate

Biomass sugar hydrolysates were prepared using DMR corn stover. The details of DMR pretreatment process were described elsewhere [42]. In brief, corn stover (INL LOT#6, harvested in Hurley County, SD) was knife milled to pass through a 19 mm (3/4-in.) rejection screen followed by dilute alkaline (NaOH at 0.1 M) deacetylation at 80 °C in a 1900-L paddle mixer to remove most of the acetyl groups and some of the lignin. The deacetylated corn stover was disc refined at a specific refining energy of approximately 100 kW/ODMT in a pilot scale Sprout 401 (36-in.) disc refiner provided by Andritz, Inc., Springfield, OH. A secondary milling was applied to the deacetylated and disc refined corn stover substrates at 100kWh/ODMT using a planetary type Szego mill to further improve digestibility. The DMR-pretreated corn stover was then hydrolyzed using cellulases from Novozyme as reported previously [41]. The hydrolysis was conducted at 15 % total solids, pH 5.2, 50 °C and completed in 5 days.

Shake flasks and Biostat-Q fermentations

Seed cultures of Z. mobilis strains harvested at exponential phase were inoculated into 125-mL shake flasks containing 40-mL media of RMG8 (RM with 80 g/L glucose) or RMG8X6 (RM with 80 g/L glucose and 60 g/L xylose) to a starting OD_600 nm of 0.1. The medium was supplemented with spectinomycin at a final concentration of 200 μg/mL. The temperature was maintained at 30 or 33 °C with a shaking speed of 120 rpm.

For BiosStat-Q Plus fermentations, strains were revived from frozen stock on RMG5 (50 g/L glucose) in 50 mL baffled flasks containing 10 mL of media, and incubated overnight at 30 °C in a shaking incubator at 180 rpm. The revived and grown culture was used to start the seed culture for fermentation. The seed cultures were then prepared in 125 mL shake flasks containing 40 mL RMG8 (80 g/L glucose) using revived cultures and inoculated at an initial OD_600 nm of 0.1. Seed flasks were incubated at 30 °C overnight in a shaking incubator at 180 rpm. Fermentations to evaluate the strains for BDO production were carried out in BioStat-Q plus fermenters with a 300-mL working volume of RMG with higher glucose concentration of 100 g/L (RMG10) to investigate whether higher carbon source could increase 2,3-BDO titer. The media were supplemented with spectinomycin at the final concentration of 200 μg/mL. The fermenters were inoculated from an overnight grown seed culture with an initial OD_600 nm value of 0.1. The fermenter operated at 30 °C, 300 rpm, and controlled at pH 5.8 with 4 N KOH.

High-performance liquid chromatography (HPLC)

The dry weight of cells at the end of fermentation was determined by centrifuging 10 mL of broth and washing the cells with double deionized H₂O. The washed cells were placed in pre-weighed aluminum dishes and dried overnight at an 85 °C incubator. The correlation between dry cell weight and OD_600 nm value was then calculated and used to estimate the dry cell weight in other time points.

Samples from the shake flasks or fermenters were taken at various time points. The samples were diluted for OD_600 nm measurement. In addition, samples were filtered through a 0.2-μm syringe filter into HPLC vials. Concentrations of glucose, xylose, 2,3-BDO, acetoin, xylitol, ethanol, HMF, furfural, lactic acid, glycerol, and acetic acid were determined from filtered sample supernatants by Agilent 1100 series HPLC (Agilent, CA) utilizing a BioRad Aminex HPX-87H organic acids column and Cation H⁺ guard cartridge (Bio-Rad, CA) operating at 65 °C. A refractive index detector was used for compound detection. Dilute sulfuric acid (0.01 N) was used as the isocratic mobile phase at a flow rate of 0.6 mL/min, following published procedures [41]. Sugar utilization, 2,3-BDO, acetoin, and ethanol titers were calculated based on the HPLC and dry cell weight data.