An artificial pathway for isoprenoid biosynthesis decoupled from native hemiterpene metabolism

Abstract

Isoprenoids are constructed in nature using hemiterpene building blocks that are biosynthesized from lengthy enzymatic pathways with little opportunity to deploy precursor-directed biosynthesis. Here, an artificial alcohol-dependent hemiterpene biosynthetic pathway was designed and coupled to several isoprenoid biosynthetic systems, affording lycopene and a prenylated tryptophan in robust yields. This approach affords a potential route to diverse non-natural hemiterpenes and by extension isoprenoids modified with non-natural chemical functionality. Accordingly, the prototype chemo-enzymatic pathway is a critical first step towards the construction of engineered microbial strains for bioconversion of simple scalable building blocks into complex isoprenoid scaffolds.

Graphical Abstract

Isoprenoids comprise >55,000 natural products for which methods to access and diversify their structures are in high demand.^1–3 Ultimately, the isoprene motif plays a critical role in modulating the biological activity of isoprenoids, determines their utility as tools to study and treat human diseases, and provides the basis to develop new fuels and chemicals.^4–7 Notably, although several valuable isoprenoids have been accessed via heterologous expression,^8–11 our ability to diversify isoprenoids is extremely limited largely due to critical limitations imposed by native isoprenoid biosynthesis. Firstly, only the mevalonate (MEV) and 1-deoxy-D-xylulose-5-phosphate (DXP) pathways (Figure 1A) are known to produce the universal hemiterpene isoprenoid diphosphate building blocks, dimethylallyl pyrophosphate (DMAPP) and isopentenyl pyrophosphate (IPP).^{12, 13} The negatively charged hemiterpenes are not cell permeable, thus preventing feeding them or analogues thereof into cultures. The MEV and DXP pathways involve at least six enzymatic steps¹³ each with stringent substrate specificity and therefore offer little opportunity to diversify the structures of isoprenoids through feeding in non-natural precursors.¹⁴ As a result, while precursor-directed biosynthesis has proven a powerful approach to access diverse structures of natural products^{15, 16}—especially polyketides^17–21—by feeding non-natural building blocks, this approach has not yet been applied to isoprenoids. Furthermore, late-stage biosynthetic modification of isoprenoid scaffolds is typically limited to oxidations, often catalyzed by P450’s.^{11, 22} Secondly, terpene metabolism is highly regulated and is a burden to the carbon supply on the cell.²³ For example, the MEV pathway uses three molecules of phosphate donor (ATP) and two reducing equivalents (NADPH) for each DMAPP/IPP, while the DXP pathway requires two phosphate donors (ATP and CTP) and two reducing equivalents (NADPH) (Figure 1A).¹³ Thirdly, given that native terpenes are typically essential for maintenance of the cell, genetic modification of native hemiterpene pathways would likely be lethal.²⁴ Together, these limitations could be overcome by supplying a membrane-permeable carbon building block dedicated for a designer pathway that would function independent of native isoprenoid metabolism. A potential strategy for hemiterpene biosynthesis could start with isopentenol (ISO) and dimethylallyl alcohol (DMAA) which are converted to the required diphosphates via stepwise phosphorylation catalyzed by two independent kinases (Figure 1B). This proposed alcohol-dependent hemiterpene (ADH) pathway is completely orthogonal to the endogenous DMAPP/IPP biosynthetic machinery, such that non-natural precursors are not expected to inhibit endogenous enzymatic machinery. In addition, the ADH pathway requires only two equivalents of ATP and no other cofactors. Furthermore, an artificial pathway designed ‘bottom-up’ as a replacement for natural hemiterpene biosynthesis could leverage naturally or engineered promiscuous enzymes. In this way, the ADH pathway could enable a broad panel of easily scalable and accessible alcohols to be converted to the corresponding diphosphate, thus providing a simple strategy to probe the plasticity of downstream isoprenoid biosynthesis in vivo or in vitro. In this study, as a necessary first step to realizing this goal, we report the design and development of a prototype ADH pathway that is completely orthogonal to native hemiterpene biosynthesis. The ability of this pathway to access isoprenoids is demonstrated by coupling the ADH pathway to two different isoprenoid biosynthetic systems.

Figure 1.

Natural and engineered hemiterpene biosynthetic pathways. (A) The MEV and DXP (grey box) pathways. A branch of the MEV pathway, the archaeal MEV I pathway, is shown in blue. For full names of enzymes, see ref¹². (B) An artificial alcohol-dependent hemiterpene (ADH) pathway completely decoupled from native isoprenoid metabolism.

Inspired by the observation that several mammalian cell lines convert farnesol and farnesol analogues to the corresponding disphosphates,^25–27 it was first determined whether E. coli harbors suitable enzymatic machinery that could convert exogenously provided ISO or DMAA into a pool of hemiterpenes for isoprenoid production. To test this, a previously reported reporter system that leverages lycopene biosynthesis was used that includes genes from the CrtEBI operon, a geranylgeranyl diphosphate synthase (CrtE or IspA Y80D), and the isopentenyl diphosphate isomerase ipi (Figure 2A).²⁸ In E. coli, this reporter system expresses genes that use pools of hemiterpenes to generate the lycopene pigment, enabling quantification.^{28, 29} The intensity of the absorbance at 450–470 nm is directly associated with an increase in hemiterpene production given that the reporter system itself is not rate limiting.^30–33

Figure 2.

Lycopene reporter system and preliminary characterization of IPK. (A) The reporter system module generates lycopene from cellular hemiterpenes. Fs is expected to inhibit native hemiterpene biosynthesis via the DXP pathway but not an artificial alcohol-dependent hemiterpene pathway. (B) E. coli BL21Tuner(DE3) harboring pCDFDuet-GGPP, pACYCDuet-Lyc and pETDuet-IPK in wells of a microplate were treated with combinations of IPTG, DMAA/ISO, and Fs and visualized.

The native DXP pathway in the E. coli reporter strains supports production of lycopene independently of exogenously added DMAA/ISO. Because of this, fosmidomycin (Fs), an inhibitor of the first dedicated step in hemiterpene biosynthesis (Figure 1A),^{34, 35} was leveraged to knockdown endogenous lycopene production in order to determine whether any endogenous machinery could support conversion of DMAA/ISO to hemiterpenes (Figure 2A). Fs was added to the culture medium at sufficient concentration (0.5 μM) to inhibit the DXP pathway³⁴ but at low enough concentration growth was not significantly suppressed. In this way, the goal was to employ Fs to prevent accumulation of excess DXP-dependent hemiterpene, forcing production of lycopene solely from the exogenously fed precursors via potential unknown endogenous enzymes. Following addition of DMAA and ISO (each at 2.5 mM) and Fs (0.5 μM) to an overnight culture of E. coli BL21Tuner(DE3) harboring the lycopene reporter plasmids pCDFDuet-GGPP + pACYCDuet-Lyc, lycopene production was quantified by visual examination of the culture broth. However, no difference in lycopene production was observed upon comparison of the E. coli strain to an otherwise identical culture prepared in the absence of DMAA/ISO (Supplementary Figure S1). Thus, the E. coli native metabolism does not support conversion of the ISO/DMAA mixture to lycopene at a rate greater than the inhibited DXP-pathway. Thus, our efforts shifted to identification of enzymes that could be heterologously expressed in E. coli and function as kinase- 2 in the proposed ADH pathway.

A protein recently found in archaea, isopentenyl phosphate kinase (IPK),^{36, 37} is responsible for the generation of IPP from isopentenyl phosphate (IP) and forms a branch of the MEV pathway called the Archaeal MEV Pathway I (Figure 1A). Over-expression of IPK in E. coli could lead to improved production of lycopene from exogenously added ISO if (1) an endogenous kinase (kinase-1, Figure 1B) can convert ISO to the corresponding monophosphate, and (2) the obligatory second phosphorylation (Figure 1B) contributes to the rate-limiting step in the conversion of ISO to lycopene in E. coli. To test these assumptions, a codon-optimized gene sequence for ipk from Thermoplasma acidophilum (Supplementary Table S1) was expressed in a strain of E. coli harboring a lycopene reporter module and was incubated either in the presence or absence of supplementary DMAA/ISO, IPTG, and Fs. The expression of IPK clearly resulted in increased lycopene (compare 1–3, Figure 2B). In addition, the presence of Fs almost completely inhibited lycopene production (2 vs. 4, Figure 2B). However, an increase in lycopene production was observed when the Fs-treated E. coli cultures that expressed IPK were supplemented with DMAA/ISO (Figure 2B). This increase in lycopene production that was dependent on IPK and DMAA/ISO validates the assumption that an endogenous kinase is capable of providing IPP from DMAA/ISO and that IPK supports the second required phosphorylation.

It was reasoned that over-expression of the E. coli gene product putatively responsible for DMAA/ISO phosphorylation would result in improved production of hemiterpenes. In a preliminary attempt to identify a suitable enzyme that could act as ‘kinase-1’ (Figure 1B), a set of 12 soluble alcohol kinases from E. coli, S. cerevisiae and A. thaliana, in addition to PhoN, a class-A non-specific acid phosphatase from Shigella flexneri (Supplementary Table S2) were cloned, expressed in E. coli, subjected to immobilized-metal affinity chromatography, and analyzed for their ability to phosphorylate DMAA/ISO by LC-MS analysis. PhoN was included in this set as it has previously been shown to phosphorylate various alcohols in vitro.^{38, 39} Notably, while none of the kinases displayed the desired activity (data not shown), mass ions consistent with DMAPP (calculated 165.0317 m/z, [M-H]⁻; observed 165.0318 m/z, [M-H]⁻) and IPP (calculated 165.0317 m/z, [M-H]⁻; observed 165.0317 m/z, [M-H]⁻) were detected in the presence of PhoN. Another potential candidate for kinase-1 is a membrane-associated diacylglycerol kinase (DGK) from Streptococcus mutans which is known to display undecaprenol-kinase activity.⁴⁰ Given that DGK is membrane-bound, it was tested in vivo. In parallel, PhoN was also tested in vivo to ensure that its activity could contribute to isoprenoid biosynthesis. Accordingly, PhoN and DGK (Supplementary Table S1) were each cloned into pETDuet-IPK and tested for their ability to support isoprenoid production in an E. coli lycopene strain by extraction and HPLC-based quantification of the pigment (Figure 3A and Supplemental Figure S2). Remarkably, the prototype PhoN-IPK system was capable of converting DMAA/ISO to lycopene in titers of ~150 mg/L in E. coli after 12 hours post-induction (Figure 3A). These titers are comparable to that of an optimized engineered DXP pathway (24 mg/L)⁴¹ and heterologous production of the MEV pathway (102 mg/L) in E. coli.³¹ As expected, the activity was largely dependent on the presence of PhoN given that 17-fold less lycopene was produced at 26 hours post-induction when PhoN is absent. In addition, wild-type DGK could support lycopene production in good yields, although these were 4-fold lower than that with PhoN after 24 hours.

Figure 3.

Lycopene titers supported by engineered E. coli strains. (A) Lycopene titers are shown at 0, 12, 26, or 48 h post induction, in the presence or absence of DMAA/ISO, using E. coli NovaBlue(DE3) harboring pAC-LYCipi and the indicated plasmid with Fs treatment. Values are the average of three replicates. Error bars are the standard deviation. (B) Substrate dependence of lycopene titers determined at 0, 12, 26, or 40 h post induction with the strain harboring pETDuet-PhoN-IPK + pAC-LYCipi with Fs treatment. Values are the average of three replicates. Error bars are the standard deviation.

Next, to determine whether the lycopene titers were dependent on the concentration of the exogenously provided alcohol substrates, a series of lycopene assays were carried out at various concentrations of DMAA/ISO using the strain harboring pETDuet-PhoN-IPK/pAC-LYCipi in the presence of Fs (Figure 3B). After 24 hours, the prototype strain produced ~60 mg/L lycopene at 0.5 mM each of DMAA/ISO, ~100 mg/L at 2.5 mM each of DMAA/ISO, and ~190 mg/L at 5 mM each of DMAA/ISO (Figure 3B). At concentrations of 10 mM, DMAA/ISO supported a maximum lycopene titer of ~20 mg/L after 24 hours.

As an additional test of the ability of the designed prototype strain to provide an isoprenoid building block, the PhoN-IPK pathway was used to provide DMAPP which was then transferred to L-Trp using the prenyltransferase (PTase) FgaPT2⁴² by providing an additional plasmid that expressed the PTase (Figure 4A). First, an in vitro reaction using purified FgaPT2 and a DMAPP standard confirmed production of a single product by HPLC and LC-MS analysis of the product mixture (Figure 4B), the mass of which (273.1591 m/z, [M+H]⁺) was consistent with the expected regioselectively mono-alkylated product, dimethylallyl-L-Trp (DMAT, calculated 273.1598 m/z, [M+H]⁺). In addition, using purified PhoN, IPK, and FgpaPT2 in a one-pot reaction, the same DMAT product was detected when DMAA was included in the in vitro reaction mixture (Figure 4B), confirming the ability of the ADH pathway to produce the required prenyl donor for FgaPT2. To test the ability of the in vivo system to provide the hemiterpene and couple it with the PTase, DMAA/ISO were fed into E. coli that harbored the ADH module (pETDuet-PhoN-IPK) in conjunction with a plasmid that harbored the PTase (pCDFDuet-FgaPT2), and after protein expression, HPLC confirmed the presence of the expected product (DMAT) in the culture media with an elution time indistinguishable from that produced by both in vitro reactions (Figure 4B). As expected from the in vitro reactions, an extracted-ion chromatogram of the culture media revealed a single peak corresponding to the mass ion for DMAT (Figure 4C). By using the in vitro FgaPT2-catalyzed conversion of L-Trp and DMAPP to DMAT as a calibration standard (the reaction went to completion), the ADH-FgaPT2 pathway in E. coli was judged to support the production of DMAT at ~20 mg/L. Furthermore, according to HPLC analysis, DMAT was not detected in negative controls that lacked either the alcohol or FgaPT2. Consistent with this observation, HR-MS analysis revealed that DMAT was present in 100-fold lower in abundance in the negative controls as compared to the positive reaction (data not shown). Together, this data confirms that the artificial ADH module is required for high-level production of the prenylated product.

Figure 4.

Tryptophan prenylation via the artificial alcohol-dependent hemiterpene pathway. (A) Reaction scheme of the ADH pathway coupled to the prenyltransferase, FgaPT2. (B) HPLC chromatograms showing peaks corresponding to DMAT: (i) in vitro FgaPT2 reaction with synthetic DMAPP; (ii) in vitro reaction with purified PhoN, IPK, and FgaPT2; (iii) culture media from in vivo bioconversion with the ADH module. The chromatograms are scaled equally. (C) Extracted-ion chromatogram showing a single peak corresponding to DMAT in the culture media of E. coli Rosetta(DE3) pLysS+pETDuet-PhoN-IPK+pCDFDuetFgaPT2 after treatment with DMAA/ISO, and IPTG.

In summary, an artificial hemiterpene biosynthetic pathway dependent on the exogenous addition of DMAA/ISO was developed. The prototype ADH pathway performed similarly to previously established routes that depend on building blocks from primary metabolism. It is expected that ribosome binding site and/or promoter engineering can be leveraged to optimize the productivity of the pathway further. Notably, although previous synthetic biology efforts have for example constructed a new entry point into the DXP pathway⁴³ and provided new routes to DXP,^{44, 45} to the best of our knowledge, the ADH pathway described here is the first to transform scalable simple precursors directly into the required pyrophosphates and couple them to isoprenoid biosynthesis. This provides a simple strategy to provide isoprenoids in good yields given that only two enzymes and DMAA/ISO need to be provided. Indeed, in the absence of ISO/DMAA, there was insufficient endogenous DMAPP in E. coli to support high level production of the prenylated tryptophan, even when Fs was not used to inhibit the native DXP pathway. Given the cost of prenol, it is unlikely that the strategy described here will be used to access isoprenoids at industrial scale. It has however been designed primarily as a future discovery tool that potentially enables for the first time the in vivo biosynthesis of hemiterpene analogues, and by extension, non-natural isoprenoids. For example, PhoN displays a broad specificity in vitro, and this is expected to extend to the in vivo system here. Furthermore, several features of PhoN have been previously targeted by enzyme engineering, including shifting its pH optima for neutral media⁴⁶ and improving its kinase activity with concomitant reduction in phosphatase activity.^{47, 48} Similarly, although the promiscuity of IPK is largely under-explored, its substrate interacts with the enzyme active site through electrostatic forces dictated by the phosphate portion of the substrate, while the remaining alkyl portion of the substrate is simply sterically accommodated.⁴⁹ Indeed, the substrate specificity of IPK has been expanded to include geranyl- and farnesylphosphate.⁵⁰ An expanded set of non-natural hemiterpenes provided by the prototype or engineered ADH pathway could be coupled with downstream enzymes to probe the promiscuity and utility of isoprenoid biosynthesis. For example, it is expected that this precursor-directed approach to non-natural isoprenoids will be readily extendible to natural product scaffolds that include L-Trp and/or other aromatics given the previously reported promiscuity of aromatic PTases.^51–54 Subsequently, the ADH pathway may enable the production of prenylated and terpene natural products with non-natural alkyl groups expanding upon the limited chemical diversity afforded by nature.

METHODS

General

All plasmids were verified by DNA sequencing. Purifications of all DNA were performed with kits from BioBasic. Lycopene standard was purchased from Sigma Aldrich. Synthetic oligonucleotides were purchased from IDT (Coralville, IA, USA). Plasmid pAC-LYCipi was purchased from Addgene (Plasmid #53279, Addgene, Cambridge, MA, USA). Restriction enzymes were purchased from New England Biolabs (Ipswich, MA, USA). Polymerase chain reactions were conducted using Phire Hot Start II DNA Polymerase from ThermoFisher Scientific (Waltham, MA, USA).

Lycopene quantification

E. coli NovaBlue (DE3) containing pAC-LYCipi and various pETDuet constructs were grown in 250 mL LB supplemented with ampicillin (100 μg/mL) and chloramphenicol (35 μg/mL) at 37 °C overnight with shaking at 250 rpm after inoculation with 0.25 mL of the starter culture. After 5 h the OD₆₀₀ of the culture was ~0.2 at which point combinations of DMAA/ISO (in DMSO), IPTG, and Fs were added to give final concentrations of 5 mM, 1 mM, and 0.5 μM, respectively. In controls that lacked DMAA/ISO, DMSO was added to give the equivalent volume. At various time points, 600 μL of culture was removed and lycopene extracted and quantified (see the Supplemental Methods).

In vitro FgaPT2 assay

FgaPT2 reactions were run at pH 7.5 in 200 μL containing 50 mM Tris-HCl, 5 mM CaCl₂, 1 mM L-tryptophan, 2 mM DMAPP, and 40 μg of FgaPT2. The reactions were incubated at 37 °C for 1 h and then quenched by the addition of an equal volume of methanol. Reactions with PhoN-IPK generated DMAPP contained 25 mM Tris-HCl, 5 mM magnesium chloride, 1 mM L-tryptophan, 1.8 mM ATP, 30 mM DMAA, 270 ng/μL FgaPT2, 20 ng/μL IPK, and 87 ng/μL PhoN at pH 8.0 in a total volume of 50 μL. The reactions were incubated at 37 °C overnight and then quenched by the addition of an equal volume of methanol. For analytical-scale HPLC analysis, FgaPT2 reactions were followed at 269 nm using a Phenomenex Kinetex 5u EVO C18 column (250 x 4.6 mm; 100 Å) at a flow rate of 1 mL/min. A linear gradient of 20–70% acetonitrile in 0.1% aqueous trifluoroacetic acid over 20 min was used.

In vivo FgaPT2 assay

A 3 mL culture of E. coli Rosetta(DE3) pLysS pETDuet-PhoN-IPK+pCDFDuetFgaPT2 was grown overnight at 37 °C and 250 rpm in LB media containing ampicillin (100 μg/mL), chloramphenicol (35 μg/mL), and streptomycin (100 μg/mL). An aliquot (100 μL) of overnight culture were used to inoculate 10 mL cultures in TB media containing the same antibiotics as before, and those cultures were grown at 30 °C for 6 h at 250 rpm until induction with 0.5 mM IPTG (final concentration) and addition to a final concentration of 5 mM DMAA, 5 mM ISO, and 10 mM Trp. The cultures were grown for 48 h after induction. The culture supernatant was diluted 1:1 in methanol before analysis by HPLC and LC-MS.