Time-course experiments monitoring heterologous metabolites reveal limitation in conversion of protocatechuate to vanillate

As a starting point for understanding pathway limitations, we sought to increase carbon flux entering the heterologous portion of the vanillate pathway (Fig. 1a). We hypothesized that increasing flux would facilitate the identification of pathway bottlenecks based on changes in titer (units of mg/L) and specific yields [units of g/g_DCW (grams of dry cell weight)] of intermediate metabolites. In our previous demonstration of de novo vanillin biosynthesis, titers of all measured heterologous metabolites (protocatechuate, vanillate, protocatechualdehyde, and vanillin) were low, each less than 60 mg/L [5]. Initially, we sought to achieve increased heterologous metabolite titers by targeting endogenous precursor biosynthesis genes for overexpression given that the first heterologous enzyme (AsbF _Bt ) is reported to efficiently catalyze conversion of endogenous 3-dehydroshikimate into protocatechuate [17].

The heterologous vanillate pathway branches from endogenous aromatic amino acid biosynthesis, and efforts to improve aromatic amino acid biosynthesis in E. coli have been well-documented [18–20]. In our initial demonstration of vanillin synthesis, we harnessed one of these strategies by expressing a feedback-resistant variant of the enzyme that catalyzes the first committed step to aromatic amino acid biosynthesis (aroG*) [5]. Previous studies demonstrated two other strategies that were successfully used to increase titers of aromatic products, in these cases by increasing availability of two key aromatic precursor metabolites: phosphoenolpyruvate (PEP) from glycolysis and erythrose-4-phosphate (E4P) from the pentose phosphate pathway. PEP and E4P condense to form DAHP during the first committed step towards aromatic amino acid biosynthesis (Fig. 1a) [18–20]. One reported strategy to increase their availability is deletion of the phosphotransferase system (PTS), which is the primary means for glucose import and consumes 1 mole of PEP per mole of glucose. Growth of a PTS⁻ strain on glucose as a sole carbon source can be made viable by upregulating the gene encoding galactose permease (galP), which allows glucose entry independent of PEP consumption (PTS⁻ glu⁺). A second documented strategy to increase availability of PEP and E4P is to overexpress the genes encoding PEP synthase (ppsA) and transketolase (tktA) [18–20]. PEP synthase catalyzes the conversion of PEP into pyruvate, and transketolase catalyzes the reversible formation of E4P and xylulose 5-phosphate from fructose 6-phosphate and glyceraldehyde 3-phosphate.

To test these strategies, we engineered a PTS⁻ glu⁺ variant of the RARE strain (PTS⁻ glu⁺ RAREʹ), where the “RARE prime” designation indicates that this strain contains intact versions of two genes (dkgB and yeaE) that were targeted for deletion in the original RARE strain. Previous work found that these two genes did not substantially contribute to aldehyde reductase activity given their low native expression [5]. When the PTS⁻ glu⁺ RAREʹ host overexpressed the aroG*, asbF _Bt , and OMT _Hs genes, 132 ± 14 mg/L of vanillate was produced 72 h after induction (Fig. 1b). When the same host was used to express the ppsA and tktA genes in addition to the previously mentioned pathway genes, vanillate titers slightly decreased but cumulative titers of heterologous metabolites increased. Specifically, protocatechuate titer increased from 200 to 300 mg/L in strains expressing the ppsA and tktA genes. In addition, the specific yield of protocatechuate noticeably increased around 24 h after pathway induction when ppsA and tktA were overexpressed (Additional file 1: Figure S1). In both experiments, protocatechuate was the dominant heterologous metabolite formed on a molar basis, and this suggested that the reaction catalyzed by the O-methyltransferase was limiting. It is possible that the decrease in vanillate titers resulted either from decreased OMT _Hs expression as two additional genes were being expressed, or from the decreased rate of biomass formation during expression of the two additional genes (Additional file 1: Figure S1), or from both contributions.

Although the gene encoding OMT _Hs was codon-optimized for expression in E. coli, we next wondered whether OMT _Hs may be expressing poorly or whether it may have low activity in E. coli. SDS-PAGE results suggested that OMT _Hs expresses well (Fig. 2a). In addition, cultures expressing the pathway were sampled for in vitro specific OMT _Hs activity. Activity measurements revealed that, while OMT _Hs activity decreases to 34 % of initially measured activity, the enzyme is still active 48 h after induction (Fig. 2b).

Supplementation of methionine and homocysteine improve vanillate titers, strongly suggesting that availability of the co-substrate S-adenosylmethionine (SAM) limits conversion of protocatechuate to vanillate

Given that OMT _Hs appeared to be expressed and active in cells at times when conversion of protocatechuate to vanillate was not occurring, we next considered that co-substrate availability may be limiting (Fig. 3a). The co-substrate SAM fulfills a variety of cellular roles, such as serving as the primary methyl donor for all cellular methylations, which compete directly with the engineered vanillin pathway. Methionine is endogenously converted to SAM by methionine adenosyltransferase (otherwise known as SAM synthetase) encoded by metK. Although exogenously supplied SAM does not enter E. coli, exogenously supplied methionine does enter cells and is known to indirectly perturb intracellular SAM availability in E. coli [6] and in yeast [21]. Importantly, in the sole previously reported effort to produce vanillate using E. coli, it was observed that methionine supplementation improved titers [6].

We monitored vanillate titers in cultures with and without supplementation of 10 mM l-methionine at peak vanillate productivity (24 h after induction) (Fig. 3b). Although methionine supplementation did not result in significant changes in vanillate titer immediately, final vanillate titer increased twofold 72 h after induction. In addition, vanillate synthesis occurred at a relatively constant rate until the final sampling time of 72 h upon methionine supplementation. This result suggests that potential depletion of SAM and methionine pools later in the culture may be responsible for limiting conversion of protocatechuate to vanillate.

To better understand the contributions of methionine biosynthesis to the vanillate pathway, a second supplementation experiment was performed, this time with the direct precursor to methionine, l-homocysteine (Fig. 3a). The methylation of homocysteine to form methionine is reported to be problematic under oxidative conditions due to the inactivation of catalytic residues in the cobalamin-independent methionine synthase (MetE) [22–25]. As before, cultures at peak vanillate productivity (24 h) were supplemented with and without l-homocysteine. Because l-homocysteine is reported to be toxic for E. coli [26, 27], we added 2.5 mM rather than 10 mM. Once again, supplemented cultures displayed an increase in vanillate titer (up to 200 mg/L) and an increase in duration of vanillate production consistent with what was observed for l-methionine addition (Fig. 3c). The average rate of vanillate synthesis doubled from 1.3 to 2.6 mg/L h. Importantly, vanillate specific yields showed that homocysteine supplementation led to higher specific production of vanillate rather than higher cell density (Fig. 3d). These results collectively suggested that reactions upstream of homocysteine synthesis in the methionine biosynthesis pathway need to be targeted to achieve increases in vanillate production from glucose as a sole carbon source.

Unlike aromatic amino acid biosynthesis, methionine biosynthesis in E. coli and other bacteria is not well understood and is regulated at multiple transcriptional as well as post-transcriptional levels (Fig. 4a). In fact, until recently, methionine was the only essential amino acid that was not commercially produced using fermentative processes [28, 29]. The protein encoded by metJ is the primary transcriptional regulator of several genes involved in methionine biosynthesis [30–32]. Methionine titers of 910 mg/L were previously achieved using an E. coli strain that was constructed by mutagenesis with nitrosoguanidine along with selection based on resistance to methionine analogs. This strain was observed to have a mutation in metJ that decreased repression of methionine biosynthesis [33]. However, genome sequencing was not performed and it is unlikely that metJ was the only gene containing a mutation. In addition to regulation by metJ, MetA (homoserine succinyltransferase) catalyzes the first committed step in methionine biosynthesis and is reported to be inhibited by both methionine and SAM [34]. l-cysteine is also a precursor to methionine. CysE (l-serine O-acetyltransferase) catalyzes the first step of cysteine biosynthesis and is reported to display significant inhibition by cysteine [35]. Fortunately, variants of MetA and CysE that have been engineered to display desensitization to feedback inhibition (MetA* and CysE*) have been reported [34, 35]. The combination of metA* expression, metJ deletion, and other modifications have previously led to a strain that accumulated 240 mg/L of methionine [34].

Deregulation of SAM biosynthesis through deletion of metJ and expression of metA* and cysE* improves vanillate titers

We hypothesized that vanillate titers might increase if methionine biosynthesis were modified by (i) deleting metJ; and (ii) expressing metA* and cysE* (Fig. 4a). Given our observation of slower growth in the PTS⁻ glu⁺ RAREʹ strain relative to the RARE strain, we decided to delete metJ in both strains and compare performance. Upon overexpression of aroG*, asbF, OMT, ppsA, and tktA, the RARE ∆metJ host resulted in higher titers of protocatechuate and vanillate than the PTS⁻ glu⁺ RAREʹ ∆metJ host (Fig. 4b). However, protocatechuate titers remained higher than vanillate titers in both cases. As previously noted, metJ is only one of many simultaneous modes of methionine biosynthesis regulation. Given feedback-resistance at the entrance to the pathway, one could envision the metJ deletion strain performing slightly worse because of increased expression of downstream genes with minimal flux entering the pathway. Given that the RARE ∆metJ host performed better than the PTS⁻ glu⁺ RAREʹ ∆metJ host, we decided to continue with RARE ∆metJ for remaining experiments.

To further understand the limitations in methionine biosynthesis, we next supplemented the RARE ∆metJ host harboring the pathway with one of three amino acids: l-methionine, l-cysteine, and l-aspartate. Because cysteine and aspartate are precursors to methionine (Fig. 4a), our goal was to investigate whether reaction steps downstream of their biosynthesis were problematic. In this case, we added 10 mM of each amino acid at the time of induction (0 h) to see whether the time of supplementation would affect pathway kinetics. In this experiment, we also included the RARE host (with metJ intact) expressing the pathway in order to better understand the potential effect of metJ deletion. Performance of the RARE and RARE ∆metJ hosts supplemented with methionine was similar, with roughly 280 mg/L vanillate produced in just 24 h (Fig. 4c). Little additional vanillate formed after the first 24 h, suggesting that methionine supplementation at induction might be more effective than supplementation at 24 h. The decrease in the rate of vanillate formation suggested that all of the methionine added initially may have been depleted within 24 h. Addition of l-cysteine or l-aspartate did not improve vanillate titers, which supported the notion of next focusing on the reactions catalyzed by MetA and CysE. Interestingly, biomass formation occurred more rapidly than usual during l-aspartate supplementation, suggesting that uptake of this amino acid occurred readily but was diverted to other cellular purposes besides methionine biosynthesis (Additional file 1: Figure S2). On the other hand, addition of 10 mM l-cysteine significantly decreased titers and biomass formation. Cysteine has been shown to inhibit the growth of several E. coli strains by inhibiting threonine deaminase and therefore isoleucine synthesis [36].

We next explored whether expression of the metA* and cysE* genes would improve conversion of vanillate to protocatechuate without the addition of l-methionine to the culture medium (Fig. 4d). In the cases in which metA* and cysE* were overexpressed, average vanillate titer was observed to exceed average protocatechuate titer at all time points sampled. Expression of metA*−cysE* increased final vanillate titers by 33 % (from 205 ± 16 to 272 ± 35 mg/L). To gain insight into potential upper bounds of vanillate titer obtainable by increasing SAM availability, we supplemented cultures expressing metA*−cysE* with 10 mM methionine at both induction and at 24 h after induction (Additional file 1: Figure S3). In this case, vanillate titers continued to exceed protocatechuate titers at all time points sampled. Addition of methionine at induction led to 341 ± 25 mg/L of vanillate produced in the first 24 h. This suggested that further increased SAM availability within the first 24 h could still improve vanillate production. Although the addition of an extra 10 mM methionine (20 mM methionine total) at 24 h after induction resulted in a final vanillate titer of 419 ± 58 mg/L at 72 h after induction, protocatechuate titers remained above 100 mg/L. Thus, a strategy of additional exogenous methionine supplementation is insufficient to address the limitation of protocatechuate conversion, particularly once cultures have achieved late exponential or stationary phase. Given continued observations of a decreasing rate of vanillate formation after 24 h, we also investigated potential loss of the ampicillin-resistant plasmid harboring OMT (pET-AsbF-OMT). Based on the similar number of colonies appearing on plates taken at the same time points (Additional file 1: Figure S4), we could rule out plasmid loss as an explanation for the limited conversion observed.

Orthogonal strategy of deregulating SAM regeneration through overexpression of mtn and luxS also achieves improvement in vanillate titers

To our knowledge, no previous study has sought to increase SAM availability by targeting genes downstream of SAM utilization, in what is known as the activated methyl cycle. S-adenosyl-l-homocysteine (SAH) is a co-product of the reaction along with vanillate, and SAH is a potent inhibitor of SAM-dependent methyltransferases [37, 38] (Fig. 5a). Homocysteine supplementation results presented earlier suggested that recycling of SAH back to homocysteine does not occur at a rate sufficient to maintain SAM availability. One of the two reactions required to recycle SAH is catalyzed by Mtn and converts SAH to S-ribosyl-l-homocysteine (SRH). The subsequent conversion of SRH to homocysteine is catalyzed by LuxS and coupled to the production of autoinducer AI-2, which is a quorum sensing molecule [39, 40]. In E. coli and other bacteria, AI-2 has been implicated as an inducer of biofilm formation [41, 42] and as an attractant for chemotaxis [43]. Because recycling of SAH is linked to the formation of AI-2, SAM regeneration is expected to be subject to immense regulation. A recent LC–MS study profiled the effect of deletions of mtn and luxS on intracellular concentrations of SAM, SAH, SRH, homocysteine, and methionine at different OD₆₀₀ values for E. coli MG1655 [44]. However, overexpression of these genes and the possible effect on intracellular concentrations of the same metabolites was not tested. We hypothesized that overexpression of mtn and luxS may increase SAM availability and therefore improve vanillate titers.

In eukaryotes, archaea, and non-LuxS-containing bacteria, SAM regeneration occurs differently, with conversion of SAH directly to homocysteine and adenosine catalyzed by SAH hydrolase (SAHase) [40]. We also hypothesized that overexpression of a heterologous SAHase might provide an alternative way to improve protocatechuate conversion with minimal interference to native regulation (Fig. 5a). To explore this concept, we tested expression of a SAHase from S. cerevisiae (sahH _Sc ). In all of the remaining experiments, neither metA* nor cysE* were expressed.

Strains and plasmids

Chemicals

The following compounds were purchased from Sigma: vanillic acid, protocatechuic acid (otherwise known as 3,4-dihydroxybenzoic acid), l-methionine, l-homocysteine, l-cysteine, and l-aspartate. Isopropyl β-d-1-thiogalactopyranoside (IPTG) was purchased from Denville Scientific. Ampicillin sodium salt, chloramphenicol, streptomycin sulfate, and kanamycin sulfate were purchased from Affymetrix.

Culture conditions

A 1X M9 salt medium (Sigma) containing 6.78 g/L Na₂HPO₄∙7H₂O, 3 g/L KH₂PO₄, 1 g/L NH₄CI, and 0.5 g/L NaCl, supplemented with 2 mM MgSO₄, 0.1 mM CaCl₂, glucose, trace elements, and antibiotics was used as the culture medium for experiments in this study. The trace element solution (100×) used contained 5 g/L EDTA, 0.83 g/L FeCl₃∙6H₂O, 84 mg/L ZnCl₂, 10 mg/L CoCl₂∙6H₂O, 13 mg/L CuCI₂∙2H₂O, 1.6 mg/L MnCl₂∙2H₂O and 10 mg/L H₃BO₃ dissolved in water. This was added to a concentration of 1× to supplement the M9-glucose medium. This medium will be henceforth referred to as “M9-glu-trace.” For all experiments except the experiment corresponding to Fig. 1b, the initial glucose concentration was 1.8 %. The initial glucose concentration was 1.2 % for the experiment corresponding to Fig. 1b.

All experiments were performed in 250 ml baffled PYREX shake flasks that contained 50 ml culture volumes. Overnight cultures were grown in 3 ml in 14 ml round-bottom tubes (Corning). Experimental cultures were initiated as follows: 1 % (v/v) inoculum volumes of overnight culture in LB medium were first transferred into overnight culture in M9-glu-trace medium, and then 1 % (v/v) inoculum volumes of overnight culture in M9-glu-trace were transferred into 50 mL M9-glu-trace medium, incubated at 30 °C, and agitated at 250 rpm. The OD₆₀₀ was measured regularly during exponential growth using a DU800 UV/Vis spectrophotometer (Beckman Coulter). Cultures were induced with 0.5 mM IPTG at OD₆₀₀ values between 0.8 and 1.0. Depending on the experiment, culture medium was supplemented with either 50 mg/L ampicillin, 17 mg/L chloramphenicol, 25 mg/L streptomycin, 25 mg/L kanamycin, or combinations of the previous antibiotics to provide selective pressure for plasmid maintenance. In some experiments (Figs. 4, 5; Additional file 1: Figures S2–S4), 50 mg/L carbenicillin was used in place of ampicillin in liquid cultures to provide more stringent pressure for plasmid maintenance. All experiments were performed in biological triplicate, and results are presented as averages with error bars representing one standard deviation.

For experiments in which S-adenosyl-l-methionine precursors were supplemented, flask cultures were set up as mentioned before but with culture volumes adjusted to achieve final concentrations as follows: 10 mM l-methionine, 2.5 mM l-homocysteine, 10 mM l-cysteine, or 10 mM l-aspartate. Stocks of supplemented metabolites were pH-neutralized and sterile filtered. In these experiments, control cultures received an equal volume of sterile deionized water instead of the metabolic precursors at the time of supplementation. Supplementation times varied from occurring at induction (0 h), occurring 24 h after induction, and occurring at both induction and 24 h after induction.

Metabolite analysis

Culture samples were pelleted by centrifugation and aqueous supernatant was collected for HPLC analysis using an Agilent 1100 series instrument equipped with a diode array detector. Heterologous compounds produced in vanillin experiments were separated using a Zorbax Eclipse XDB-C18 column (Agilent) and detected using a wavelength of 280 nm. A gradient method used the following solvents: (A) 50 % acetonitrile+ 0.1 % trifluoroacetic acid (TFA); (B) water+ 0.1 % TFA. The gradient began with 5 % Solvent A and 95 % Solvent B. The setting at 20 min was 60 % Solvent A and 40 % Solvent B. The program restored the original ratio at 22 min and ended at 25 min. The flow rate was 1.0 ml/min and all vanillin pathway compounds of interest eluted within 15 min. Column temperature was maintained at 30 °C.

SDS–PAGE analysis

To determine qualitative protein expression level of OMT in the absence of other pathway gene overexpression, E. coli MG1655(DE3) was transformed with empty pETDuet-1 or pET-OMT. Single colonies from plates of each transformation were grown overnight in 3 ml of LB with 50 mg/L ampicillin. Cells were passaged into second overnight cultures by inoculating 3 ml of M9+ 1.8 % glucose with 100 μL of the overnight LB cultures. Shake flask cultures containing 50 ml M9+ 1.8 % glucose were inoculated at 1 % inoculum from overnight M9 cultures and incubated with agitation at 30 °C and 250 rpm. Cultures were induced with 0.5 mM IPTG at OD₆₀₀ values between 0.8 and 1.0. Twenty-four hours after induction, 5 ml of each culture were sampled and pelleted by centrifugation. Cell pellets were resuspended in 1 ml of 10 mM Tris–HCl at pH 8.0 and lysed using sonication. After lysis, samples were pelleted by centrifugation (16,000g, 4 °C, 20 min) and supernatant was collected as soluble lysate. The remaining pellet was resuspended in 10 mM Tris–HCl and deemed the insoluble fraction.

Total protein was quantified by the Bradford assay method [58] using Bio-Rad Protein Assay Dye Reagent (Cat #500-0006) and a bovine serum album (BSA) standard. A Bio-Rad 10 % Mini-PROTEAN TGX gel (Cat #456-1034) was run using the Mini-PROTEAN Tetra Cell electrophoresis apparatus. Bio-Rad Precision Plus Protein All Blue Standard (Cat #161-0373) and 10 μg of total protein for each sample was loaded on the gel. After running at 200 volts for 33 min, the gel was washed with deionized water before staining with Bio-Rad Bio-Safe Coomassie Stain (Cat #161-0786).

O-methyltransferase in vitro specific activity

To determine the specific activity of the O-methyltransferase as a function of time in the context of the pathway, E. coli MG1655(DE3) was transformed with pET-OMT-AsbF and pACYC-AroG*-PpsA-TktA plasmids. Single colonies from plates of each transformation were grown overnight in 3 ml of LB with 50 mg/L ampicillin and 17 mg/L chloramphenicol. Cells were passaged into second overnight cultures by inoculating 3 ml of M9+ 1.8 % glucose+ antibiotics with 100 μL of the overnight LB cultures. Shake flask cultures containing 50 ml M9+ 1.8 % glucose+ antibiotics were inoculated at 1 % inoculum from overnight M9 cultures and incubated with agitation at 30 °C and 250 rpm. Cultures were induced with 0.5 mM IPTG at OD₆₀₀ values between 0.8 and 1.0. At 12 h time point increments after induction, 5 ml of each culture were sampled and pelleted by centrifugation. Cell pellets were resuspended in 1 ml of 100 mM sodium phosphate buffer at pH 7.5 and lysed using sonication. After lysis, samples were pelleted by centrifugation (16,000g, 4 °C, 20 min) and supernatant containing OMT was collected.

The OMT activity assay contained the following final concentrations to a final volume of 600 µL per sample: 100 mM sodium phosphate buffer (pH 7.5), 2 mM protocatechuate, 2 mM S-adenosylmethionine, 5 mM MgCl₂, and roughly 0.1 mg/mL BSA protein equivalent of lysate (60 µL). Lysate was always added last, and immediately upon addition, the assay solution was split into two 300 µL volumes (t_i and t_f). The t_i volume was quenched with 1 % TFA, and the t_f was quenched in the same manner after 1 h. Both volumes for each sample were maintained at 30 °C. Potential debris from quenched volumes were pelleted by centrifugation (16,000g, 4 °C, 3 min) and supernatant was analyzed by HPLC. Concentrations of protocatechuate and vanillate were measured as described before. Specific OMT activity was calculated by dividing differences in vanillate concentration by total protein concentration in corresponding lysate. Specific activities were subsequently normalized to the activity at the first time point sampled (12 h). Lysate from a strain transformed with empty pET was used as a control in otherwise identical assays to ensure there was no background conversion of protocatechuate.

Calculation of maximum theoretical pathway yield

Maximum theoretical yields presented for vanillin and vanillate were calculated using an in silico genome-scale metabolic model of E. coli (iJO1366) [59] and the constraint-based reconstruction and analysis (COBRA) toolbox version 2.0 [60, 61] within the framework of MATLAB 2013a. Heterologous vanillin pathway reactions were added to the model, and uptake rates for glucose and oxygen were set to −10 mmol/gDW-h and −1000 mmol/gDW-h, respectively [62].