Analysis of heterologous taxadiene production in K- and B-derived Escherichia coli

Abstract

Taxa-4(5),11(12)-diene is the first dedicated intermediate in the metabolic pathway responsible for synthesizing the anticancer compound Taxol. In this study, the heterologous production of taxadiene was established in and analyzed between K- and B-derived Escherichia coli strains. First, recombinant parameters associated with precursor metabolism (the upstream methylerythritol phosphate (MEP) pathway) and taxadiene biosynthesis (the downstream pathway) were varied to probe the effect different promoters and cellular backgrounds have on taxadiene production. Specifically, upstream MEP pathway genes responsible for the taxadiene precursors, dimethylallyl diphosphate and isopentenyl diphosphate, were tested with an inducible T7 promoter system within K and B E. coli strains. Whereas, inducible T7, Trc, and T5 promoters were tested with the plasmid-borne geranylgeranyl diphosphate synthase and taxadiene synthase genes responsible for the downstream pathway. The K-derivative produced taxadiene roughly 2.5-fold higher than the B-derivative. A transcriptomics study revealed significant differences in pyruvate metabolism between the K and B strains, providing insight into the differences observed in taxadiene biosynthesis and targets for future metabolic engineering efforts. Next, the effect of temperature on cell growth and taxadiene production was analyzed in these two strains, revealing similar phenotypes between the two with 22°C as the optimal production temperature. Lastly, the effect of indole on cell growth was investigated between the two strains, showing that the K-derivative demonstrated greater growth inhibition compared to the B-derivative.

Introduction

The diterpenoid natural product Taxol (marketed by Bristol-Meyers Squibb and having the generic name of paclitaxel) possesses impressive anticancer properties and has shown efficacy against carcinomas of the ovary, breast, lung, head and neck, bladder and cervix, melanomas, and AIDS-related Kaposi’s sarcoma (Skeel 1999). Extracted from the bark of Taxus brevifolia (the Pacific Yew tree), early stage production efforts to support clinical trials resulted in extremely low yields. The sacrifice of a 100-year-old tree generated approximately 3 kg of bark, yielding approximately 300 mg of purified Taxol or the equivalent of approximately a single dose (Horwitz 1994). In 1988, a semi-synthetic method for producing Taxol from a common Yew precursor metabolite (10-deacetylbaccatin III; 10-DAB) was developed (Denis et al. 1988). A fully synthetic route to Taxol was accomplished in 1994; however, the compound’s complex molecular architecture containing 11 chiral centers and a notably rare oxetane ring complicated efforts to establish an efficient and economic synthetic production process (Nicolaou et al. 1994). Though improved semi-synthetic production methods have decreased the toll taken on the Pacific Yew natural resource, current and alternative production routes still depend on processes that are prone to issues involving process speed, scalability, and overall robustness (Frense 2007; Roberts 2007). Currently, Phyton Biotech produces Taxol and related taxane intermediates by Taxus plant cell culture at a scale of 75,000 l (Parekh et al. 2008). Taxol’s medicinal value coupled to the production challenges introduced above spurred an interest in heterologously producing the compound through more technically convenient microbial hosts.

In considering microbial Taxol biosynthesis, thought must be given to the native and heterologous components needed. Taxol biosynthesis stems from upstream isoprenoid pathway C₅ precursors, dimethylallyl diphosphate (DMAPP) and isopentenyl diphosphate (IPP; Fig. 1). In this pathway, geranylgeranyl diphosphate and taxa-4(5),11 (12)-diene (henceforth referred to as taxadiene) are the first dedicated intermediate compounds toward complete Taxol biosynthesis. Other isoprenoids have also been engineered in E. coli and Saccharomyces cerevisiae, such as the C₁₅ sesquiterpene amorphadiene (precursor to the antimalarial compound, artemisinin; Chang et al. 2007; Kizer et al. 2008; Martin et al. 2003; Newman et al. 2006; Paradise et al. 2008; Pitera et al. 2007; Ro et al. 2006; Tsuruta et al. 2009), the C₂₀ diterpenes casbene, neocembrene (Kirby et al. 2010), and the C₄₀ carotenoid lycopene (Alper et al. 2005a; 2006; 2005b; Alper and Stephanopoulos 2008; Farmer and Liao 2001; Jin and Stephanopoulos 2007; Kang et al. 2005; Klein-Marcuschamer et al. 2007; Vadali et al. 2005; Yoon et al. 2008; 2007; 2006). Within the broader context of heterologous isoprenoid production, work has predominantly focused on engineering and optimizing the precursor pathways that support biosynthesis (Martin et al. 2003; Yuan et al. 2006). More recently, combined metabolic pathway engineering and two forms of protein engineering were utilized to increase the production of the C₂₀ diterpene levopimaradiene (precursor to the plant-derived ginkogolides) over 2,600-fold to approximately 700 mg l⁻¹ through E. coli (Leonard et al. 2010).

Fig. 1.

E. coli isoprenoid metabolism with taxadiene in the context of Taxol biosynthesis

Recent efforts in our groups have developed a strain of E. coli capable of producing taxadiene at a titer of over 1 g l⁻¹ in a fed-batch bioreactor (Ajikumar et al. 2010). The strain was developed using a multivariate-modular pathway engineering approach, in which the taxadiene biosynthetic pathway was divided into upstream MEP (precursor) and downstream (biosynthesis) modules, which were individually varied to identify the best expression parameters for taxadiene production. The study was conducted within MG1655-derived strains of E. coli, and it was shown that low expression of both the upstream and downstream modules generated a local optimum in taxadiene production. Moreover, there was an extremely complex, non-linear response of taxadiene production with respect to the expression of the upstream and downstream compartments under the conditions tested. These results emphasize the challenges with respect to Taxol production through microbial routes including: (1) the expression of foreign plant biosynthetic genes, (2) the assembly of multi-step and hybrid metabolic pathways, (3) improving precursor supply and pathway optimization, and (4) the introduction of unforeseen or unwanted cellular events. The introduction of a metabolic pathway can disrupt the normal cellular physiology and regulatory mechanisms of the heterologous host, and predicting the subsequent effects are often quite difficult due to the complex nature of these evolved systems.

To further characterize the E. coli taxadiene system, we analyzed cellular phenotype and production under multiple genetic and environmental conditions. Specifically, this study compares and heterologous taxadiene production in engineered variants of E. coli JM109(DE3) and BL21(DE3). Motivating this analysis and comparison were the differences in taxadiene production observed between these two hosts. We reasoned that further characterizing these strains would yield clues about the underlying metabolic and cellular effects for the differences observed. After similarly modulating the upstream and downstream parameters for taxadiene biosynthesis, we conducted a transcriptomic study to provide a systems-level profile of gene expression between the strains tested for production. The differences at the transcript level between the K and B strains were analyzed to understand expression profiles that could account for the cellular differences in taxadiene production. Next, the effect of temperature on cell growth and taxadiene production was analyzed between these two strains, revealing similar phenotypes between the two. Last, the effects of indole toxicity on cell growth were analyzed between the two strains since this compound had previously demonstrated a negative impact on final E. coli taxadiene levels.

Materials and methods

Reagents and chemicals

The reagents and chemicals used in this study were purchased from ThermoFisher Scientific (Waltham, MA, USA) or Sigma-Aldrich (St. Louis, MO, USA). Polymerase chain reaction (PCR) primers were synthesized by Eurofins MWG Operon (Ebersberg, Germany). TaKaRa LA Taq^™ DNA polymerase was from Clontech Laboratories/Takara Mirus Bio (Madison, WI, USA).

Gene, plasmid, and strain construction

Codon optimized versions of a crtE gene (coding for a geranylgeranyl diphosphate synthase (GGPPS)) from Taxus canadensis (GenBank accession code AF081514; Hefner et al. 1998) and a txs gene (coding for a taxadiene synthase (TXS)) from T. brevifolia (GenBank accession code U48796; Wildung and Croteau 1996) were synthesized and constructed previously (Ajikumar et al. 2010). Standard molecular biology techniques were then used to generate the plasmids presented in Table 1 (Sambrook and Russell 2001).

Table 1.

Plasmids constructed in this study

Plasmid name	Description
pQE30	bla; pBR322 ori (Qiagen)
pTrcHis2B	bla; pBR322 ori (Invitrogen)
pACYCDuet-1	cat; p15a ori (Novagen)
pQE-TXS-GGPPS	bla; T5prom-txssyn-crtEsyn-T5term; pQE30 background
pTrc-TXS-GGPPS	bla; Trcprom-txssyn-crtEsyn-Trcterm; pTrcHis2B background
pACYCDuet-TXS-GGPPS	cat; T7prom-txssyn-crtEsyn-T7term; pACYCDuet-1 background

The “syn” subscript refers to codon-optimized, synthetically constructed genes

Table 2 presents the strains constructed to support taxadiene biosynthesis. E. coli strains JM109(DE3) and BL21(DE3) were used as the original hosts for YW22 and YW23, respectively. For the construction of YW22 and YW23, a sequence consisting of T7_prom-dxs-T7_term-T7_prom-idi-T7_term-T7_prom-ispDF-T7_term was inserted into the araA locations of JM109(DE3) and BL21(DE3) using λ-Red recombination (Datsenko and Wanner 2000), respectively, as described previously (Wang and Pfeifer 2008). All integrants and knockouts were verified by diagnostic PCR. All strains were stored at −80°C in LB medium supplemented with 10% (v/v) glycerol.

Table 2.

Strains constructed in this study

Strain name	Description
JM109(DE3)	F− endA1, recA1, gyrA96, thi, hsdR17 (rk−, mk+), relA1, supE44, λ−, Δ(lac-proAB), [traD36, proAB, lacIqZΔM15] (DE3) (Promega)
BL21(DE3)	F− ompT hsdSB (rB−, mB−) gal dcm (DE3) (Novagen)
YW22	JM109(DE3); araA:: T7prom-dxs-T7term-T7prom-idi-T7term-T7prom-ispDF-T7term
YW23	BL21(DE3); araA:: T7prom-dxs-T7term-T7prom-idi-T7term-T7prom-ispDF-T7term

Small-scale cultures

The strains were inoculated in 2 ml of production medium in 16×100 mm culture tubes to an OD_600nm=0.1 and cultured for 5 days at 22°C and 250 rpm. The culture medium used consisted of 5 gl⁻¹ yeast extract, 10 gl⁻¹ tryptone, 10 gl⁻¹ sodium chloride, 15 gl⁻¹ glycerol, 3 ml l⁻¹ 50% (v/v) antifoam B, 100 mM HEPES, and was adjusted to pH 7.60 with 5 M NaOH (Lau et al. 2004). Selection was maintained with 100 mg l⁻¹ carbenicillin, 50 mg l⁻¹ kanamycin, or 34 mg l⁻¹ chloramphenicol when necessary. Expression of lacI-repressed genes was accomplished by induction with 100 μM isopropyl β-d-1-thiogalactopyranoside (IPTG). For the temperature modulation studies, cultures were inoculated to OD_600nm=0.1, grown at 37°C until OD_600mn≈0.6, then induced with 100 μM IPTG and transferred to a shaking incubator at 12°C, 17°C, 22°C, 27°C, 32°C, or 37°C. For sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis, a 1-ml aliquot of cells at the end of the culture period was centrifuged, resuspended in Tris-EDTA buffer, lysed by sonication, and the lysate clarified by centrifugation. The supernatant, representing the soluble fraction of the total cell protein, was analyzed by 4–20% Bis–Tris SDS-PAGE followed by Coomassie staining.

Transcript preparation and analysis

YW22(pACYCDuet-TXS-GGPPS) and YW23(pACYCDuet-TXS-GGPPS) were grown as described above in the “Small-scale cultures” section. In late-stage exponential growth (approximately 48 h), total RNA was first prepared with the RNeasy Protect Bacteria Mini Kit (QIAGEN; Venlo, The Netherlands) and then purified by spin column with the RNeasy Mini Kit (QIAGEN). Extracted total RNA (in triplicate) was sent to the Tufts University Computational Genomics Core Facility for analysis, amplification, labeling, hybridization to the GeneChip^® E. coli Genome 2.0 Array (Affymetrix; Santa Clara, CA, USA), and scanning. The GeneChip^® E. coli Genome 2.0 Array contains 10,208 probe sets for E. coli strains K-12 MG1655, CFT073, O157:H7-EDL933, and O157:H7-Sakai. The six Affymetrix CEL files containing intensity information for each probe were exported to the MATLAB^® (The MathWorks^™; Natick, MA, USA) Bioinformatics Toolbox. Microarray data were processed using the RMA procedure for background adjustment (Irizarry et al. 2003). Normalized intensities for each probe set and each strain were averaged, and Student’s t test was utilized to calculate the p value between the YW22 and YW23 strains. Probe data were then filtered to contain only probes specific for E. coli K-12 MG1655. Statistically significant differences between the two strains were considered for cases with p<0.01, while differentially expressed genes were considered for cases with a greater than twofold difference in log₂ values.

Taxadiene analysis

Culture aliquots were extracted with an equal volume of ethyl acetate, followed by vortexing for 20 s and centrifuging at 10,000 rpm for 10 min. Taxadiene analysis was performed on a Shimadzu (Kyoto, Japan) QP5050A GC-MS using splitless injection. Separation was performed on a Restek (Bellefonte, PA, USA) Rtx^®-XLB column (30 m×0.25 mm ID, 0.25 μm). The column was initially held at 50°C for 1 min, then it was heated to 320°C using a gradient of 8°C min ⁻¹, and it was finally held at that temperature for 2 min. Mass spectrometry was conducted in single ion monitoring mode scanning for mass to charge ratios of 107m/z, 122m/z, and 272m/z, as determined previously (Jennewein et al. 2001). Taxadiene eluted at 22.21 min and quantification was accomplished by using a five-point calibration curve (5, 10, 25, 50, and 100 mg l⁻¹) created with a purified taxadiene standard.

Organic acid analysis

The aqueous phase of a culture aliquot was analyzed on an Agilent (Santa Clara, CA, USA) 1100 series high-performance liquid chromatography (HPLC) coupled to a refractive index detector. A 20-μl sample of the clarified culture supernatant (by centrifugation at 10,000 rpm for 10 min) was applied to a Bio-Rad (Hercules, CA, USA) Aminex^® HPX-87H Ion Exchange (300×7.8 mm, 9 μm) column. The isocratic analysis used a solvent with a composition of 9.5 mM H₂SO₄ held at a flow rate of 0.3 ml min⁻¹. These conditions were identified by using an iterative stochastic search HPLC optimization program based on the compounds anticipated to be present in the culture medium (Dharmadi and Gonzalez 2005). A five-point standard calibration curve was created and used for quantification of glycerol, pyruvate, acetate, ethanol, succinate, formate, and lactate. The elution order was as follows: pyruvate (16.7 min), succinate (22.7 min), lactate (24.2 min), glycerol (25.1 min), formate (26.8 min), acetate (29.1 min), and ethanol (41.3 min).

Microplate growth assay

To more accurately determine the intrinsic growth properties of YW22(pACYCDuet-TXS-GGPPS) and YW23 (pACYCDuet-TXS-GGPPS) and analyze the effects of exogenous indole on growth, cultures were conducted in a 96-well-plate format. Sterile, round-bottom 96-well plates were filled with 200 μl production medium (described previously) containing 34 mg l⁻¹ chloramphenicol and 100 μM IPTG. Exogenous indole was supplemented at concentrations of 0, 10, 25, 50, 100, 200, or 300 mg l⁻¹. A stab of glycerol stock was inoculated into LB medium supplemented with 34 mg l⁻¹ chloramphenicol and grown overnight at 30°C and 250 rpm. The wells of the 96-well plate were then inoculated to OD_600nm≈0.05. Well plates were inserted into a Molecular Devices (Sunnyvale, CA, USA) VERSAmax microplate spectrophotometer preheated to 30°C and incubated for 16 h with mixing. Every 10 min, mixing stopped, and the absorbance at 600 nm was measured for all wells.

Parameter estimation

Data were imported into MATLAB^®, and background absorbance was subtracted from the raw data and then fit to a logistic population model (Bailey and Ollis 1986).

$$X(t)=\frac{X_0{e}^{\mu t}}{1-\frac{X_0}{X_{sQt}}\left(1-e^{\mu t}\right)}$$ (1)

Eq. 1 describes the cell density, X, as a function of time, t, with the following constant parameters (X₀, the initial cell density; X_sat, the stationary phase cell density; and μ, the specific growth rate). A non-linear least-squares regression of the data with the logistic population equation was undertaken in MATLAB^® (using the “nlinfit” function) to determine the X₀, X_sat, and μ parameters. The 95% confidence intervals for the cell-density estimates and the estimated parameters were examined for accuracy.

Results

In this study, taxadiene production was accomplished in and compared between two commonly used lineages of E. coli known to show significant differences in cellular metabolism and overall cellular phenotype (Yoon et al. 2009). Biosynthetic genes were expressed on plasmids with three promoters of different strengths (T7, Trc, and T5). Whereas, numerous genes within the MEP pathway (dxs, idi, ispD, and ispF) were over-expressed using T7 promoters from the chromosome of either E. coli JM109(DE3) or BL21(DE3). Strain design was based upon the genotype described previously for E. coli carotenoid biosynthesis (Alper et al. 2005a, b; Yuan et al. 2006). Because lycopene and taxadiene were derived from the same metabolic precursors, this initial strain design was chosen as a template for subsequent engineering. These plasmids and strains are detailed in Tables 1 and 2, respectively. We were particularly interested to observe any differences in taxadiene production between these two hosts given differences we have previously observed with these strains during heterologous polyketide biosynthesis (Wu et al. 2010).

Figure 2a presents E. coli taxadiene production as a function of strain and the promoter strength used for the downstream pathway. Across all three strengths of the downstream module, taxadiene production in YW22 is roughly 2.5-fold that of YW23. The differences with respect to upstream MEP pathway strength within each strain were negligible (for YW23, ANOVA p=0.585). Expression of the MEP pathway genes from a T7 promoter within the JM109(DE3) or BL21(DE3) chromosomes improved specific taxadiene titers. Of the improvements observed, YW22(pACYCDuet-TXS-GGPPS) had the highest specific taxadiene titer at 0.502 mg gDCW⁻¹. The results indicate that the over-expression of the isoprenoid biosynthetic pathway genes and expression of the heterologous genes from a low-copy, T7-based plasmid was best for taxadiene production, as is consistent with our previous studies (Ajikumar et al. 2010).

Fig. 2.

Recombinant parameter optimization. The downstream pathway was expressed under T7 (pACYC), T5 (pQE), and Trc (pTrc) promoters on multi-copy plasmids, while the upstream pathway was expressed from T7 promoters localized to the E. coli chromosomes of JM109(DE3) or BL21(DE3), resulting in YW22 and YW23. Error bars represent ±1 SD of three to six replicates (n=3–6); b SDS-PAGE analysis of heterologous protein levels in the soluble fraction of total cell protein for the three promoter systems tested in YW22 and YW23

A transcriptomics study (global transcript profiling) was undertaken to better understand why the K strain produces more than twice as much taxadiene with qualitatively comparable or lower expression of the heterologous genes (as seen in Fig. 2b, particularly for the pTrc plasmid). Total RNA was extracted from YW22 and YW23 during late-stage exponential growth at 22°C (approximately 48 h) in the production medium previously described and analyzed using the GeneChip^® E. coli Genome 2.0 Array. After adjustment, normalization, and filtering, 348 of the 4070 MG1655 genes were differentially expressed (log₂(YW22/YW23)>±2) at a statistically significant level (p value <0.01; as can be seen in the scatter plot in Fig. 3a). Of these 348 genes, 243 were upregulated in YW22 while 105 were upregulated in YW23.

Fig. 3.

Microarray data comparing late-stage exponential growth gene-expression between YW22(pACYCDuet-TXS-GGPPS) and YW23(pACYCDuet-TXS-GGPPS). a A volcano plot shows the relationship between p value and differential expression. Points in the upper right quadrant correspond to genes upregulated greater than twofold in YW22(pACYCDuet-TXS-GGPPS) at a statistically significant level (p<0.01) while points in the upper left quadrant correspond to genes that are upregulated greater than twofold in YW23(pACYCDuet-TXS-GGPPS) at p<0.01. b A plot showing the relative expression level of the genes in the isoprenoid MEP pathway for both YW22(pACYCDuet-TXS-GGPPS) and YW23(pACYCDuet-TXS-GGPPS). The dxs, idi, and ispDF genes were previously assessed for gene expression using qPCR (Ajikumar et al. 2010). Error bars represent ±1 SD of three replicates (n=3). Asterisk indicates statistically significant results (p<0.05), while dagger indicates p<0.01

As an attempt to identify an overall bottleneck in the isoprenoid biosynthetic pathway itself and to identify additional differences between YW22 and YW23, we next compared the expression of the genes in the isoprenoid biosynthetic pathway (dxs, dxr, ispDEFGH, idi, ispA, ispB, and ispU; Fig. 3b). Of the 11 genes, four (dxs, ispE, ispF, and ispA) were expressed higher (p<0.05) in YW23 and one (ispU) was expressed higher in YW22 (p<0.01).

Gene expression, protein folding, and therefore resulting enzyme activity vary as a function of temperature. It was previously found in our group that this factor had significant effect on heterologous polyketide production in E. coli, being optimal at 22°C and non-existent at 37°C (Boghigian et al. 2011). As a result, a study was undertaken to determine the optimal temperature for taxadiene production. Here, a temperature down-shift from 37°C to between 12°C and 32°C (at 5°C intervals), concurrently with induction of gene expression at OD_600nm=0.6, was utilized to observe the differences in cell-density attained, taxadiene titer, substrate uptake, and by-product formation for both YW22(pACYCDuet-TXS-GGPPS) and YW23(pACYCDuet-TXS-GGPPS). Figure 4 presents the data in graphical form.

Fig. 4.

Temperature modulation study for YW22(pACYCDuet-TXS-GGPPS) and YW23(pACYCDuet-TXS-GGPPS). For both strains, as a function of temperature between 12°C and 37°C in 5°C increments, the a cell density, b specific uptake rate of glycerol, c specific production rate of acetate, d specific production rate of ethanol, and e specific taxadiene titer are plotted. Error bars represent ±1 SD of four replicates (n=4)

It has been shown previously that BL21(DE3) grows to a higher cell density than JM107 (a close relative to JM109 (DE3); Yau et al. 2008). However, there have not been many side-to-side comparisons of E. coli strains and their physiology (Phue et al. 2008; 2005; Phue and Shiloach 2004), much less in terms of metabolic engineering applications (Tseng et al. 2009; Wu et al. 2010). Here, we also observed that YW23 (the B derivative) grew to higher cell densities than YW22 (the K derivative); however, the cell-density differences achieved were not statistically significant at 22°C, 27°C, and 32°C (Fig. 4a). The specific uptake rate of glycerol increased roughly linearly with respect to temperature (Fig. 4b). Interestingly, YW23 showed lower specific uptake rates of glycerol than YW22 but grew to a higher cell density, lending itself to better utilization of the initial carbon source. Across all strains and culture temperatures, acetate was the primary by-product. The specific production rate of acetate was higher for YW22 than YW23 at all temperatures (Fig. 4c), except 22°C, also indicating that YW23 better utilizes glycerol. Lactate, succinate, formate, and pyruvate were not observed (less than 0.5 mM) in the residual culture medium for all strains and culture temperatures. Lastly, taxadiene production was observed at all temperatures except 37°C for both YW22 and YW23 (Fig. 4d). Interestingly, for each strain, the specific taxadiene titers did not vary significantly between 12°C and 27°C, with a slight decrease observed at 32°C.

In our previous study, it was shown that intracellular levels of the aromatic heterocycle indole, inversely correlated with extracellular taxadiene concentration (Ajikumar et al. 2010). In the top-producing strain, exogenous indole at levels of 100 mg l⁻¹ decreased taxadiene production roughly eightfold (to 50 mg l⁻¹) and almost abolished taxadiene production at levels of 200 mg l⁻¹. Moreover, it was shown that the highest taxadiene producer was also the most growth-inhibited by indole. To further analyze this phenomenon with respect to the two strains featured here, we undertook growth testing under various concentrations of exogenous indole (0 mg l⁻¹ to 300 mg l⁻¹) in the medium used in this study. Figure 5a shows the exponential-phase-specific growth rates of YW22 and YW23 as a function of exogenous indole concentration. The growth was not inhibited for both strains at levels of 10, 25, or 50 mg l⁻¹ (p<0.05 as compared with the untreated control). At levels of 100, 200, and 300 mg l⁻¹, the growth rates steadily declined for both strains. As can be seen by the specific growth rates normalized to the untreated control in Fig. 5b, indole inhibited the growth of YW22 greater than that of YW23. At 300 mg l ⁻ indole; the specific growth rate of YW22 was roughly 65.3% of the control, while that of YW23 was 78.5% of that respective control.

Fig. 5.

Influence of indole on specific growth rates for YW22 (pACYCDuet-TXS-GGPPS) and YW23(pACYCDuet-TXS-GGPPS). For both strains, as a function of exogenous indole concentration, the a exponential-phase specific growth rate, and b exponential-phase specific growth rate normalized to the untreated culture are plotted. Error bars represent ±1 SD of three replicates (n=3). Asterisk indicates statistically significant results (p<0.05)

Discussion

Although a large number of genes identified through the transcriptomics analysis had no clearly assigned function (hypothetical proteins), a number of encoded enzymes involved in central metabolic pathways were identified. In YW23, both pyruvate kinase I (pykF) and phosphoenolpyruvate carboxykinase (pck) were upregulated, indicating that phosphoenolpyruvate is being produced from both pyruvate (through the action of pykF) and oxaloacetate (through the action of pck). Because pyruvate is one of the direct precursors (along with glyceraldehyde-3-phosphate) for the DXP pathway, decreasing pyruvate flux from this pathway by upregulation of these two enzymes is a likely explanation for why the taxadiene production titer in YW23 is less than half that of YW22 (Fig. 6). Interestingly, over-expression of pck in E. coli improved heterologous lycopene production roughly threefold, while over-expression of ppc (phosphoenolpyruvate carboxylase) reduced lycopene production by roughly 30% (Farmer and Liao 2001).

Fig. 6.

A simplified metabolic map of E. coli with relation to taxadiene biosynthesis. Genes in green text indicate increased expression in the B-derivative, while genes in red text indicate decreased expression in the B-derivative. Dashed lines indicate metabolic regulation (green meaning upregulated, red meaning downregulated). Dotted lines indicate other metabolic pathways. Abbreviations: G6P glucose-6-phosphate, F6P fructose-6-phosphate, F16BP fructose-1,6-bisphosphate, G3P glyceraldehyde-3-phosphate, DHAP dihydroxyacetonephosphate, PEP phosphoenolpyruvate

Another key difference between these two strains was the activity around the fructose-6-phoshate metabolite node. YW22 significantly upregulated phosphofructokinase I (pfkA) which catalyzes the conversion of fructose-6-phosphate to fructose-1,6-bisphosphate. At the same time, YW23 significantly upregulated fructose-1,6-bisphosphatase (fbp), which catalyzes the opposite reaction (Fig. 6). Although both of these enzymes are inhibited by phosphoenolpyruvate, it would appear that the expression of the glycolytic pfkA would be preferable over the gluconeogenic fbp, especially being that fructose-1,6-bisphosphate is degraded during glycolysis to yield glyceraldehyde-3-phosphate, the other direct precursor for the isoprenoid biosynthetic pathway. Over-expression of pfkA and/or deletion of fbp would likely increase the cellular pool of glycerol-3-phosphate and would therefore increase flux toward the isoprenoid biosynthetic pathway and improve taxadiene production. Moreover, this is consistent with what is known regarding the metabolic regulation of glycolysis, where the accumulation of phosphoenolpyruvate activates phosphofructokinase and inhibits fructose-1,6-bisphosphatase.

When analyzing transcript profiles for the MEP pathway, it should first be noted that ispB and ispU are also required for the native production of octaprenyl diphosphate (in menaquinol and ubiquinol biosynthesis) and undecaprenyl diphosphate (used in peptidoglycan biosynthesis), respectively. It is clear that slightly increased expression of certain genes in the MEP pathway by YW23 does not result in improved taxadiene production, leading us to believe the main difference in taxadiene phenotype between the K and B strains lies within central metabolism. It is also interesting to note that native ispA (which encodes a GPPS and FPPS) expression is approximately an order of magnitude lower than that of the next lowest expressed gene in the E. coli pathway. Normally, this situation might indicate a potential bottleneck in precursor supply; however, in this case, IspA activity may be compensated by the heterologously introduced plant GGPPS (encoded by crtE_syn) which can polymerize the C₅ DMAPP and IPP directly to the required C₂₀ GGPP.

The negative impact of indole is consistent with the previous finding that higher taxadiene producers were more severely inhibited at higher concentrations of exogenous indole. The exact relationship between indole and taxadiene formation is still unknown, however. The role of indole exists in tryptophan and chorismate-related metabolic pathways. For example, the degradation of tryptophan results in the generation of one equivalent indole and one equivalent of pyruvate through the action of tryptophanase (encoded by tnaA in E. coli). Being that pyruvate is a direct precursor used in isoprenoid biosynthesis, there may be a connection between indole levels and taxadiene production. Indole has also recently been implicated in antibiotic resistance at a cost of reduced cellular fitness (Lee et al. 2010). Though it is again difficult to draw a direct correlation between indole and taxol intermediate formation from the data presented in the study by Lee et al., it is nonetheless notable that indole may influence both metabolic and signaling pathways during taxadiene production.

With the development of modern “-omics” techniques such as genomics, transcriptomics, proteomics, and metabolomics, along with the methodologies of systems biology gaining popularity among both engineers and biologists, the wealth of information on an increasing number of organisms is advantageous for metabolic engineering (Durot et al. 2009; Feist et al. 2009; Medini et al. 2008). Among many other applications, DNA microarray and transcriptional profiling has shown promise for metabolic engineering applications. Particularly, it was recently used in tandem with targeted metabolite profiling to examine a severe growth defect in amorphadiene-producing E. coli (Kizer et al. 2008). As stated previously, amorphadiene is a similar compound to taxadiene, but this E. coli host was engineered with the heterologous mevalonate pathway from S. cerevisiae, which resulted in the generation of the foreign intermediate of 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA; Martin et al. 2003). The growth inhibition of this host was shown to correlate to the intracellular concentration of HMG–CoA (Pitera et al. 2007). By applying transcriptomics and targeted metabolite profiling, it was shown that high levels of HMG–CoA-inhibited type II fatty acid biosynthesis (inducing the accumulation of malonyl–CoA; Kizer et al. 2008). This inhibition of fatty acid biosynthesis altered membrane structure properties and induced the expression of a number of genes associated with osmotic, oxidative, and heat-shock stresses. By supplementing the even-chain fatty acids palmitic acid (16:0) or oleic acid (cis-Δ⁹-18:1) in the culture medium, the HMG–CoA-induced growth inhibition was alleviated (Kizer et al. 2008). In a later paper, the group replaced two of the S. cerevisiae enzymes related to HMG–CoA (an HMG–CoA synthase and an HMG–CoA reductase) with bacterial versions of these genes from Staphylococcus aureus, which improved amorphadiene production significantly (Tsuruta et al. 2009). This is one example of how the powerful tools of systems biology can be utilized to engineer a biological system from the top-down. In theory, this type of analysis could be utilized to further determine the effect of indole toxicity on the system presented in this study.

In summary, the study presented here established and compared the heterologous production of taxadiene, a highly complex small molecule and precursor to the potent anticancer molecule Taxol, in two lineages of E. coli, K and B. We showed that expression of the upstream pathway using a T7 promoter from the chromosome enabled taxadiene production, while across all promoter strengths of the downstream pathway, the K-derivative outperformed the B-derivative. We investigated this phenomenon by applying global transcript profiling to these two strains in exponential phase growth, revealing significant differences in pyruvate metabolism and central metabolism more globally that provide clues for future metabolic engineering efforts to improve taxadiene production in E. coli. It was also shown in this study that the B-derivative showed increased expression of a number of the upstream pathway genes, although this did not lead to improved taxadiene production. We then compared these two hosts across two environmental conditions (varying temperature and varying exogenous concentration of indole). Temperature had a significant effect on taxadiene production, with optima at 22°C. Indole more significantly inhibited the growth of the K-derivative than the B-derivative, which is consistent with previous observations that higher producing strains of taxadiene are more severely inhibited by this molecule.