Synthetic Pathway for Production of Five-Carbon Alcohols from Isopentenyl Diphosphate

Abstract

Synthetic biological pathways could enhance the development of novel processes to produce chemicals from renewable resources. On the basis of models that describe the evolution of metabolic pathways and enzymes in nature, we developed a framework to rationally identify enzymes able to catalyze reactions on new substrates that overcomes one of the major bottlenecks in the assembly of a synthetic biological pathway. We verified the framework by implementing a pathway with two novel enzymatic reactions to convert isopentenyl diphosphate into 3-methyl-3-butenol, 3-methyl-2-butenol, and 3-methylbutanol. To overcome competition with native pathways that share the same substrate, we engineered two bifunctional enzymes that redirect metabolic flux toward the synthetic pathway. Taken together, our work demonstrates a new approach to the engineering of novel synthetic pathways in the cell.

INTRODUCTION

The chemical and transportation industries currently use limited nonrenewable resources to produce raw materials that could instead be synthesized from renewable resources using metabolic engineering (18). Natural biological pathways have traditionally been the source of industrially important chemicals produced using fermentation (43). Some of the regulatory mechanisms that limit production in the native hosts are circumvented by reconstructing pathways in heterologous hosts (23). To synthesize chemicals not produced by natural pathways, enzymes are overexpressed in novel combinations in order to construct synthetic biological pathways (29, 49).

One of the rate-limiting steps in the assembly of a synthetic pathway is the identification of enzymes that can catalyze new reactions of interest. Reliance on known enzymatic reactions limits the number of synthetic pathways that could be assembled, and the task of engineering an enzyme to catalyze a novel reaction remains a challenge. Strategies that incorporate directed evolution or metagenomic libraries require product-specific high-throughput screens or selections in order to identify enzymes able to catalyze a desired reaction (17). Rational and in silico protein engineering strategies generate smaller libraries to alleviate the need for high-throughput technologies but require a priori knowledge about the protein being engineered (6). Therefore, the effort to discover an enzyme that can catalyze a desired reaction is often hindered by the lack of high-throughput screening or selection and insufficient knowledge about how to rationally engineer a protein.

We provide a new framework to identify proteins able to catalyze a desired reaction on a new substrate based on models that describe pathway and enzyme evolution. The patchwork evolution model suggests that a new biological pathway is assembled by recruiting enzymes from existing pathways (16). The ability to assemble new pathways could be facilitated by the high degree of promiscuity demonstrated by enzymes. The promiscuity could be a source of divergence during enzyme evolution, since members of the same superfamily typically share the same fold or a common catalytic strategy despite dissimilarities in their sequences, substrates, and chemical transformations (19). Combining the patchwork evolution model with hypotheses that address enzyme promiscuity, our framework constructs targeted libraries from enzyme families, which could be regarded as a collection of evolutionarily related proteins able to catalyze a specific chemistry on various substrates, to discover enzymes able to catalyze the desired reactions on new substrates. Enzyme families might already serve as libraries in nature and provide an organism with a pool of enzymes from which to quickly evolve new reactions that might be useful, for example, during adaptation to xenobiotics (36).

We reasoned that enzyme families could be used in the laboratory to assemble synthetic biological pathways in order to develop new processes to convert biomass to biofuels and biomaterials. As an example, we chose to develop a metabolic pathway to produce 3-methylbutanol, 3-methyl-3-butenol, and 3-methyl-2-butenol, which have better combustion efficiencies than ethanol and research octane numbers (102, 102, and 92, respectively) that are more similar to that of gasoline than is that of ethanol, which is widely used as a gasoline oxygenate and substitute (48; Robert Dibble, Department of Mechanical Engineering, University of California, Berkeley, personal communication). 3-Methyl-2-butenol can be readily converted to citral, an intermediate in the synthesis of pharmaceuticals, vitamin A, vitamin E, some widely used carotenoids, and certain flavors and fragrances (2). 3-Methylbutanol has been produced in Escherichia coli (8) modified with the Ehrlich pathway (14) from Saccharomyces cerevisiae, but no pathway exists to synthesize either 3-methyl-3-butenol or 3-methyl-2-butenol. We used enzyme families as libraries to screen for novel enzymatic reactions and explored E. coli's metabolic potential for the assembly of a synthetic pathway to produce 3-methylbutanol, 3-methyl-3-butenol, and 3-methyl-2-butenol from isopentenyl diphosphate (IPP), the central metabolite in isoprenoid biosynthesis (21).

MATERIALS AND METHODS

Oligonucleotides and DNA sequencing.

All oligonucleotides were obtained from Integrated DNA Technologies with standard purification. For the primer sequences mentioned here, see Fig. S3 in the supplemental material. DNA sequencing to confirm cloning products was performed by Quintara Biosciences.

Strain and plasmid availability.

The strains, plasmids, and plasmid sequences (in GenBank format) constructed in this study have been deposited in the private instance of the Joint BioEnergy Institute Inventory registry and will be moved to the public instance (https://public-registry.jbei.org) after publication. The strains and plasmids are physically available from Addgene (http://www.addgene.org).

Strains and plasmids.

All of the experiments were performed with E. coli JM109 {endA1 glnV44 thi-1 relA1 gyrA96 recA1 mcrB⁺ Δ(lac-proAB) e14-[F′ traD36 proAB⁺ lacI^q lacZΔM15] hsdR17(r_K⁻ m_K⁺)} or DH1 [endA1 recA1 gyrA96 thi-1 glnV44 relA1 hsdR17(r_K⁻ m_K⁺) λ⁻]. The mevalonate pathway was encoded by plasmids pMevT and pMevB (23). pMevT contains the genes atoB (acetoacetyl coenzyme A [CoA] thiolase), hmgS (3-hydroxy-3-methylglutaryl [HMG]-CoA synthase), and hmgR (truncated HMG-CoA reductase). pMevB contains the genes mk (mevalonate kinase), pmk (phosphomevalonate kinase), and pmd (mevalonate pyrophosphate decarboxylase). All genes amplified from E. coli were from MG1655.

Construction of plasmids for phosphatase and reductase libraries.

pHADX (where X is a number from 1 to 23) plasmids were constructed by amplifying hadX from the E. coli chromosome using primers HADX-F and HADX-R and cloning the product into pTrc99A using NcoI and EcoRI. pNUDY (where Y is a letter from A to M) plasmids were constructed by amplifying nudY from the E. coli chromosome using primers NUDY-F and NUDY-R and cloning the product into pTrc99A using NcoI and EcoRI. pNemA was constructed by amplifying nemA from the E. coli chromosome using primers NEMA-F and NEMA-R and cloning the product into pTrc99A using EcoRI and KpnI. PCR, restriction digestion, and ligation reactions were performed using standard cloning protocols following the manufacturers' instructions.

Screening for enzymes able to catalyze the phosphatase reaction.

pHADX (where X is a number from 1 to 23) and pNUDY (where Y is a letter from A to M) were transformed into E. coli JM109 coexpressing pMevT and pMevB and plated onto LB agar plates with ampicillin, chloramphenicol, and tetracycline. Cell cultures were grown and tested for 3-methyl-3-butenol production using gas chromatography-mass spectrometry (GC-MS) as described below, in the section on growth and GC protocols for production and detection of C₅ alcohols.

Screening for enzymes able to catalyze the reductase reaction.

pNemA was transformed into E. coli DH1, which was plated onto LB agar plates with ampicillin. Growth and experimental protocols were the same as in other experiments, except that 1 g/liter of 3-methyl-2-butenol was added to the medium after induction with isopropyl-β-d-thiogalactopyranoside (IPTG; Sigma-Aldrich). Cell cultures were grown and tested for 3-methylbutanol production using GC-MS as described below, in the section on growth and GC protocols for production and detection of C₅ alcohols.

Protein purification of HAD-like phosphatases and Nudix hydrolases and comparison of relative protein concentrations.

We amplified each enzyme of the haloacid dehalogenase (HAD)-like phosphatases using primers HADX-F and HADX-R (where X is a number from 1 to 23) and Nudix (nucleoside diphosphate linked to X) hydrolases using primers NUDY-F and NUDY-R (where Y is a letter from A to M). The PCR products were cloned into the NcoI and BamHI sites of pPro29b (22), which was transformed into JM109. An overnight culture was inoculated into a liter of LB medium with ampicillin to an initial optical density at a wavelength of 600 nm (OD₆₀₀) of 0.05. We grew the culture at 37°C until the OD₆₀₀ reached 0.6, induced it with 20 mM propionate, and grew it overnight at 30°C. The cells were pelleted and lysed using 10× BugBuster (Novagen) diluted with binding buffer (50 mM Tris-HCl [pH 7.5], 1 mM EDTA, 0.1 mM dithiothreitol) following the directions provided by the manufacturer.

The soluble and insoluble fractions were collected and separated on NuPAGE 4 to 12% Bis-Tris Gels (1.5 mm, 15 wells; Invitrogen) following the manufacturer's instructions. A Novex Sharp Protein Standard (Invitrogen) ladder was used. The proteins were transferred to polyvinylidene difluoride membranes using the iBlot Dry Blotting Transfer System (Invitrogen). Each membrane was blocked with 1% bovine serum albumin for 1 h, washed, and incubated overnight with 3 μl S-protein–alkaline phosphate conjugate (Novagen) in 12 ml of Tris-buffered saline (pH 7.5)–0.05% Tween 20. The following day, the membranes were washed and stained with the 5-bromo-4-chloro-3-indolylphosphate (BCIP)–Nitro Blue Tetrazolium liquid substrate system (Sigma).

Protein purification of NudB and in vitro assay for phosphatase activity.

We amplified NudB from pNUDB using primers NUDB-F and NUDB-R3 and cloned the PCR product into pPro29b using NcoI and BamHI to make pPro29b-NUDB. E. coli BLR(DE3) was transformed with pPro29b-NudB, and an overnight culture was inoculated into a liter of LB medium with ampicillin to an initial OD₆₀₀ of 0.05. We grew the culture at 37°C until the OD₆₀₀ reached 0.6, induced it with 20 mM propionate, and grew it overnight at 20°C. The cells were pelleted, suspended in binding buffer (50 mM Tris-HCl [pH 7.5], 1 mM EDTA, 0.1 mM dithiothreitol), sonicated, and centrifuged. The tagged protein was purified from the supernatant using the S·Tag thrombin purification kit (Novagen) following the manufacturer's instructions. Protein concentration was determined using the Pierce BCA protein assay kit (Thermo Scientific).

The standard reaction mixture for studying NudB kinetics contained the following in 100 μl: 50 mM Tris (pH 8.0), 10 mM MgCl₂, 1 mM DTT, 200 μM 7-methylguanosine (Sigma-Aldrich), 0.5 U/ml bacterial purine nucleoside phosphorylase (Sigma-Aldrich), 0.5 U/ml yeast inorganic pyrophosphatase (Sigma-Aldrich), and 0.5 μM phosphate sensor (Invitrogen). Phosphate release was measured using a SpectraMax M2 (Molecular Devices) by exciting at 426 nm and measuring emission at 464 nm. A standard curve comparing the concentration of P_i to fluorescence was made by incubating different concentrations of P_i to the standard mixture.

To determine NudB kinetics, 15 different concentrations of isopentenyl pyrophosphate triammonium salt solution (IPP salt; Sigma-Aldrich) ranging from 0 to 25 μM were tested. The assay was performed at 30°C with 1 μM purified NudB, and IPP salt was added to start the reaction mixture. The reaction was monitored in a SpectraMax M2 for 3 min, and the change in fluorescence was used to calculate the velocity of the reaction after normalization to the sample with 0 μM IPP salt. The change in fluorescence was converted to a change in the P_i concentration using the standard curve. SigmaPlot and its Enzyme Kinetics module were used to analyze the velocity data to calculate V_max and K_m.

Construction of fusion proteins Idi-NudB and Idi1-NudB.

The gene for Idi was amplified from the E. coli chromosome using primers IDI-F2 and IDI-NUDB-SOE-R, and the gene for NudB was amplified using primers IDI-NUDB-SOE-F and NUDB-R2. The PCR products were used as templates in a second PCR with primers IDI-F2 and NUDB-R2 to amplify the fusion protein Idi-NudB, which was cloned into pTrc99A using EcoRI and KpnI to construct pIdi-NudB. pIdi1-NudB was constructed similarly, except that Idi1 was amplified from the S. cerevisiae chromosome using primers IDI1-F and IDI1-NUDB-SOE-R, NudB was amplified using primers IDI1-NUDB-SOE-F and NUDB-R2, and Idi1-NudB was amplified using IDI1-F and NUDB-R2 in the second PCR. The 19- and 15-amino-acid linkers used to construct Idi-NudB and Idi1-NudB, respectively, were based on previous work (35).

Construction of plasmids with isomerase, phosphatase, and reductase activities.

idi was amplified from the E. coli chromosome using primers IDI-F and IDI-R and cloned into pTrc99A and pNudB using EcoRI and KpnI to construct pIdi and pNudB-s-Idi, respectively. nemA was amplified from pNemA using primers NEMA-F2 and NEMA-R2 and cloned into pIdi-NudB using BamHI and XbaI to construct pIdi-NudB-s-NemA. nemA was amplified from pNemA using primers NEMA-F3 and NEMA-R2 and cloned into pIdi1-NudB using BamHI and XbaI to construct pIdi1-NudB-s-NemA.

Verification of IPP conversion to 3-methyl-2-butenol and 3-methylbutanol.

pIdi, pNudB-s-Idi, pIdi-NudB, pIdi1-NudB, pIdi-NudB-s-NemA, and pIdi1-NudB-s-NemA were transformed into E. coli JM109 coexpressing pMevT and pMevB and plated onto LB agar plates with ampicillin, chloramphenicol, and tetracycline. 3-Methyl-3-butenol, 3-methyl-2-butenol, and 3-methylbutanol production was analyzed using GC-flame ionization detection (FID).

Growth and GC protocols for production and detection of C₅ alcohols.

Three colonies were picked for each construct and grown overnight at 37°C in 5 ml of LB medium with antibiotic. The overnight cultures were inoculated into 5 ml of EZ Rich defined medium (Teknova) with 0.2% glucose and antibiotic to an initial OD₆₀₀ of 0.05. Cultures were grown at 37°C for 3 h, induced with 0.5 mM IPTG, and grown for 20 h at 30°C. A 700-μl volume of culture was sampled, mixed with a solution of chloroform-methanol (80:20) spiked with 50 μg/ml of butanol (Sigma-Aldrich) as an internal standard, vortexed for 15 min, and centrifuged for 1 min at 13,000 × g. The chloroform layer was removed for analysis by GC.

The GC-MS data were collected in full-scan mode (m/z 50 to 300) using a Tr-Wax column (0.25 mm by 30 m, 0.25-μm film thickness; Thermo Electron) on a PolarisQ GC-MS apparatus with a TriPlus autosampler (Thermo Electron). The carrier flow rate was 1.2 ml min⁻¹, and the inlet temperature was set to 200°C. The oven program was 40°C for1.20 min, an increase from 40 to 130°C at 25°C min⁻¹, and an increase from 130 to 220°C at 35°C min⁻¹. The solvent delay was set at 3.40 min. Samples were normalized using the butanol internal standard and quantified using authentic standards. 3-Methylbutanol standards were purchased from Sigma-Aldrich. 3-Methyl-3-butenol and 3-methyl-2-butenol were purchased from the Tokyo Chemical Industry Co., Ltd.

The GC-FID data were collected using a Tr-Wax column (0.25 mm by 30 m, 0.25-μm film thickness; Thermo Electron) on a Focus GC apparatus with a TriPlus autosampler (Thermo Electron). The carrier was set at a constant pressure of 300 kPa, and the inlet temperature was set to 200°C. The oven program was 40°C for 1.50 min and then an increase from 40 to 110°C at 15°C min⁻¹. Samples were normalized using the butanol internal standard and quantified using authentic standards.

RESULTS

A synthetic pathway converts IPP into three five-carbon alcohols.

IPP is naturally synthesized from either the mevalonate or the deoxyxylulose 5-phosphate pathway and is the primary substrate of the synthetic biological pathway used to produce three C₅ alcohols (Fig. 1). In the synthetic pathway, IPP is first isomerized to dimethylallyl diphosphate (DMAPP) (reaction 1), and both IPP and DMAPP are dephosphorylated to 3-methyl-3-butenol and 3-methyl-2-butenol, respectively (reaction 2). Finally, 3-methyl-2-butenol is reduced to 3-methylbutanol (reaction 3). Reaction 1 is catalyzed by an enzyme with IPP isomerase activity. No enzymes have been discovered yet to catalyze either reaction 2 or 3, so we identified three enzyme families to screen for enzymes that might catalyze these novel reactions.

Fig 1.

Synthetic pathway converts IPP to 3-methyl-3-butenol, 3-methyl-2-butenol, and 3-methylbutanol. IPP is produced by coexpression of the mevalonate pathway genes (atoB, hmgS, hmgR, mk, pmk, and pmd) in E. coli. Three different reactions convert IPP into three different C₅ alcohols. Reaction 1 catalyzes the isomerization of IPP to DMAPP. Reaction 2 catalyzes the dephosphorylation of IPP into 3-methyl-3-butenol and DMAPP into 3-methyl-2-butenol. Reaction 3 catalyzes the reduction of 3-methyl-2-butenol to 3-methylbutanol. The synthetic pathway competes with native pathways in E. coli that also use IPP and DMAPP as substrates, such as those for quinone and cell membrane synthesis (in red).

IPP phosphatase activity is discovered by screening two enzyme families.

We identified the HAD (5) and Nudix (26) superfamilies as likely to contain enzymes that can catalyze the dephosphorylation of IPP and DMAPP from the table of mechanistically diverse superfamilies provided by Glasner et al. (11). The prevalence of both superfamilies in all three domains of life (bacteria, archaea, and eukaryotes) suggests that their members are used extensively throughout evolution to catalyze mechanistically diverse reactions on various substrates, and both superfamilies of enzymes have the catalytic residues required to recognize and dephosphorylate the phosphate moieties of IPP and DMAPP. We limited our search to the family of 23 HAD-like phosphatases in E. coli (20) within the HAD superfamily as the first library and the family of 13 Nudix hydrolases in E. coli (24) within the Nudix superfamily as the second library.

We overexpressed each enzyme from the two libraries and screened for the ability to catalyze the conversion of IPP to 3-methyl-3-butenol in E. coli that expressed the mevalonate pathway genes. GC was used to detect 3-methyl-3-butenol production. A control that expressed the mevalonate pathway genes in the absence of any phosphatases produced a detectable quantity of 3-methyl-3-butenol. The background production indicates that the catalytic activity that converts IPP to 3-methyl-3-butenol exists in E. coli. Of the 36 enzymes screened, overexpression of two HAD-like phosphatases (HAD4 and HAD10) and five Nudix hydrolases (NudB, NudF, NudI, NudJ, and NudM) led to a >2-fold increase in the production of 3-methyl-3-butenol compared to that of the control (Fig. 2). NudF, whose isozyme from Bacillus subtilis was identified as having prenyl alcohol biosynthesis activity (46), demonstrated an almost 3-fold improvement in production over the control. The overexpression of NudB led to the greatest conversion of IPP to 3-methyl-3-butenol in vivo, at 5-fold more C₅ alcohol than the control.

Fig 2.

Enzymes catalyzing reaction 2 are identified from two libraries. (A) The family of 23 HAD-like phosphatases from E. coli (black bars) was screened for an enzyme able to catalyze the dephosphorylation of IPP to 3-methyl-3-butenol. Overexpression of HAD4 and HAD10 led to a >2-fold increase in 3-methyl-3-butenol production compared to the control (white bar). (B) The family of 13 Nudix hydrolases from E. coli (black bars) was screened for an enzyme able to catalyze the dephosphorylation of IPP to 3-methyl-3-butenol. Overexpression of NudB, NudF, NudI, NudJ, and NudM led to a >2-fold increase in 3-methyl-3-butenol production compared to the control (white bar).

The production of 3-methyl-3-butenol did not correlate well with the concentration of the phosphatase produced by overexpression. To determine the phosphatase concentration produced by the cell, each of the HAD-like phosphatases and Nudix hydrolases screened was tagged with an N-terminal S·Tag (see Fig. S2 in the supplemental material). Even though overexpression of NudB led to the highest 3-methyl-3-butenol production, its protein concentration was lower than that of many of the Nudix hydrolases tested. Similarly, the protein concentration of HAD4 and HAD10 was not greater than that of any other HADs, even though these two HAD-like phosphatases produced more 3-methyl-3-butenol than the other members of the family.

IPP phosphatase activity is verified in vitro.

We developed an assay to detect the phosphatase reaction with IPP as a substrate to verify NudB's ability to catalyze reaction 2 in vitro. The assay measures pyrophosphate release kinetics in real time using a coupled enzyme assay that produces a fluorescent output (13). We purified NudB using an N-terminal S·Tag. We determined the apparent V_max of NudB to be 0.63 nmol/min/mg and its apparent K_m to be 4.7 μM for IPP (see Fig. S1 in the supplemental material). NudB's apparent K_m is comparable to that of the E. coli IPP isomerase Idi, but its apparent V_max is 1,000-fold lower than that of Idi (12). The poor kinetics of NudB for IPP exemplifies the natural evolution of an enzyme, which generally begins with a protein having low affinity for a new substrate or low rates of catalytic action on the bound substrate (32). However, we recognize that the N-terminal S·Tag might also affect the activity and apparent kinetic values measured. Overall, the in vitro data support the in vivo finding that NudB binds IPP and catalyzes the phosphatase reaction to convert it to 3-methyl-3-butenol.

3-Methyl-2-butenol reductase activity is discovered from one enzyme family.

The reduction of xenobiotics with carbonyl groups commonly uses NADPH as a cofactor (30). We identified a family of NADPH dehydrogenases that we hypothesized could catalyze the reduction of 3-methyl-2-butenol to 3-methylbutanol from the list of enzyme superfamilies presented by Todd et al. (42). These dehydrogenases are part of the flavin mononucleotide-oxidoreductase superfamily called old yellow enzyme (OYE) (44). The OYE family is represented in both prokaryotes and eukaryotes, and several members are associated with xenobiotic metabolism (45), suggesting that natural systems exploit the propensity of these enzymes to catalyze reduction reactions on novel substrates. We limited our search to the OYE family in E. coli, which can reduce α/β-unsaturated carbonyl functionalities.

NemA (28), the only documented member of the OYE family from E. coli that fit our criteria, was overexpressed when there was 1 g/liter of 3-methyl-2-butenol in the medium, and catalysis of reaction 3 was confirmed by measuring 3-methylbutanol production using GC. The overexpression of NemA led to 3-methylbutanol production, and the conversion of 3-methyl-2-butenol to 3-methylbutanol was 40% after 24 h. The control that did not overexpress any enzyme did not produce any 3-methylbutanol. Even though 3-methyl-2-butenol lacks the α/β-unsaturated carbonyl functional group present in most of the substrates of enzymes in the OYE family, overexpression of NemA still demonstrated increased 3-methylbutanol production from 3-methyl-2-butenol.

Idi-NudB fusion overcomes competition with native pathways for IPP and DMAPP.

We overexpressed E. coli IPP isomerase (Idi) to catalyze reaction 1 and NudB to catalyze reaction 2 separately in E. coli that also overexpressed the mevalonate pathway in order to synthesize 3-methyl-2-butenol from IPP. However, coexpression of Idi and NudB decreased 3-methyl-3-butenol production by 5-fold compared to overexpression of NudB by itself, and no 3-methyl-2-butenol production was observed (Fig. 3A). No 3-methyl-3-butenol or 3-methyl-2-butenol production was observed when Idi was overexpressed by itself.

Fig 3.

3-Methyl-3-butenol, 3-methyl-2-butenol, and 3-methylbutanol are produced from IPP. (A) No C₅ alcohol production was observed when Idi was overexpressed by itself compared to the controls (99A, NudB). Coexpression of NudB and Idi produced 5-fold less 3-methyl-3-butenol (gray bars) than the positive control. (B) Overexpression of the fusion proteins Idi-NudB and Idi1-NudB increased C₅ alcohol production by >2-fold compared to coexpression of NudB and Idi, and 3-methyl-2-butenol production (white bars) was observed. Production of 3-methylbutanol (black bars) was observed when NemA was coexpressed with either Idi-NudB or Idi1-NudB. No production of either 3-methyl-2-butenol or 3-methylbutanol was observed when only NemA was overexpressed.

We hypothesized that a novel protein to restore 3-methyl-3-butenol production and synthesize 3-methy-2-butenol from IPP could be engineered by fusing Idi and NudB together joined by a peptide linker to construct a single bifunctional polypeptide (Idi-NudB). Idi-NudB consists of two functional domains, each of which catalyzes a different reaction (isomerization or dephosphorylation) and is part of a consecutive reaction (conversion of IPP to DMAPP and DMAPP to 3-methyl-2-butenol). This fusion protein mimics natural bifunctional enzymes, such as Arabidopsis lysine-ketoglutarate reductase/saccharopine dehydrogenase (50), human 5-amino-4-imidazolecarboxamide ribonucleotide transformylase/IMP cyclohydrolase (3), Abies grandis abietadiene synthase (31), and lysine-ketoglutarate reductase/saccharopine dehydrogenase from developing soybean seeds (27). We joined Idi and NudB with a peptide linker because linkers can control favorable and unfavorable interactions between adjacent protein domains (47).

The overexpression of the Idi-NudB enzyme fusion increased C₅ alcohol production by >2-fold, compared to the overexpression of Idi and NudB separately, and led to 3-methyl-2-butenol production (Fig. 3B). We also fused the S. cerevisiae IPP isomerase Idi1 (40) to NudB (Idi1-NudB). The overexpression of Idi1-NudB increased 3-methyl-3-butenol and 3-methyl-2-butenol production by an additional 40% compared to that of Idi-NudB. The production of 3-methyl-2-butenol from IPP by the fusion protein indicates that the catalytic domains of Idi and NudB are functional, the single polypeptides are bifunctional, and Idi-NudB and Idi1-NudB can compete with the native enzymes in E. coli for IPP and DMAPP.

Three five-carbon alcohols are produced from IPP.

We assembled the entire synthetic pathway by cloning either idi-nudB or idi1-nudB 5′ of nemA to construct plasmid pIdi-NudB-s-NemA or pIdi1-NudB-s-NemA, respectively. We observed production of 3-methyl-3-butenol, 3-methyl-2-butenol, and 3-methylbutanol from E. coli that overexpressed the genes of the mevalonate pathway and harbored one of these two plasmids. The total amount of C₅ alcohols produced remained the same as that produced in the absence of NemA, and the amount of 3-methyl-3-butenol remained unchanged. The decrease in 3-methyl-2-butenol production observed in the presence of NemA is accounted for by the increase in 3-methylbutanol production. The production of three C₅ alcohols from IPP verifies our synthetic pathway and framework of using enzyme families to identify enzymes able to catalyze new reactions.

DISCUSSION

One major challenge in the construction of synthetic biological pathways is the identification of enzymes able to catalyze each new desired reaction in the pathway. The patchwork evolution model (16) provides a foundation for strategies that screen libraries of natural enzymes to identify those that can catalyze a novel reaction of interest. However, these strategies typically involve large libraries and high-throughput screening or selection able to recognize the desired product. For example, a metagenomic library of almost 10⁶ variants constructed from environmental DNA was screened to identify enzymes with 4-hydroxybutyrate dehydrogenase activity, whose identification was made possible by the use of growth on 4-hydroxybutyrate as a selection (15). Unfortunately, most industrially important compounds are not readily amenable to high-throughput screening and selection, which makes large libraries unwieldy. Alternative screening methods, such as phage display (25, 39) and mRNA display (34, 38), can screen 10⁹ to 10¹² variants and detect protein-protein, protein-peptide, and protein-DNA interactions. However, these techniques are hampered by practical limitations, such as whether the proteins are folded in a functional form either in vitro or when displayed on the phage, and have not been used to engineer enzymes to catalyze new chemical reactions involving small molecules.

In this work, we decreased the size of the library that needs to be screened on the basis of our understanding of how enzymes naturally evolve new activities and the body of work that categorizes enzymes into families on the basis of their mechanistic activities (7). We combined the patchwork evolution model with hypotheses that explain the role of enzyme promiscuity during protein evolution and formulated a framework that uses protein families as small, targeted libraries. We screened 36 enzymes from the HAD-like phosphatase and Nudix hydrolase families and identified 7 able to dephosphorylate IPP and DMAPP. Overexpression of NudB led to the highest 3-methyl-3-butenol production, even though its in vivo protein concentration was lower than that of most of the other enzymes screened, which suggests that the increased alcohol production seen is due to its superior kinetic properties. Furthermore, we identified a member of the OYE family able to reduce 3-methyl-2-butenol.

Even though the overexpression of an enzyme with the desired activity on the new substrate can lead to an increase in the reaction rate sufficient to produce the desired product, overexpression alone is insufficient when engineering a new combination of enzymes to work in concert and produce the desired final product. Overexpression is insufficient, because multiple native pathways often compete for a common intermediate, which is a recurring problem in metabolic engineering (41). For example, the overexpression of Idi and NudB did not lead to any significant increase in 3-methyl-2-butenol production. One explanation for this observation is that when Idi is overexpressed, the low kinetic efficiency with which NudB catalyzes the novel phosphatase reaction, compared to that of the enzymes in the native pathways that consume IPP and DMAPP, prevents any significant amount of 3-methyl-2-butenol from accumulating. We reasoned that the IPP and DMAPP produced by the heterologous mevalonate pathway in the presence of Idi were consumed by native enzymes, since in vitro analysis of NudB indicated that it is relatively slow at catalyzing the new phosphatase reaction. For example, E. coli farnesyl pyrophosphate (FPP) synthase, IspA (10), uses IPP and DMAPP to synthesize FPP, which is used in quinone and cell wall biosynthesis. Therefore, we engineered Idi-NudB and Idi1-NudB, which successfully competed with native enzymes in E. coli for IPP and DMAPP and led to 3-methyl-2-butenol production and the production of 3-methylbutanol from IPP when they were coexpressed with NemA.

The decrease in C₅ alcohol production occurred only when Idi was overexpressed, because DMAPP is required by the native pathways in E. coli to synthesize isoprenoids. Therefore, the IPP produced by the mevalonate pathway could have been channeled away from C₅ alcohol production and toward native isoprenoid production when Idi and NudB were overexpressed. Even though overexpression of the fusion proteins Idi-NudB and Idi1-NudB recovered some of the C₅ alcohol production, total production was still lower than that from the overexpression of only NudB. The difference in production might be due to some of the DMAPP produced by the Idi and Idi1 domains of the fusion enzymes being used to synthesize native isoprenoids before it could be converted to 3-methyl-2-butenol by the NudB domain. Since both Idi and NudB are E. coli proteins, it is unlikely that their solubilities would change significantly if they were fused together. Idi1 is from S. cerevisiae and is unlikely to be more soluble than Idi from E. coli. Therefore, the increase in C₅ alcohol production by Idi1-NudB over that by Idi-NudB might be explained by the fact that even though Idi and Idi1 have similar catalytic efficiencies, E. coli Idi has a lower V_max and a lower K_m than yeast Idi1 (12). Overexpression of Idi1-NudB led to higher production of 3-methyl-2-butenol than did that of Idi-NudB, suggesting that under conditions where abundant IPP is available from the overexpression of the mevalonate pathway, a higher V_max can lead to more DMAPP production, which compensates for the catalytic inefficiency of NudB for DMAPP conversion to 3-methyl-2-butenol.

The approach of fusing metabolic enzymes together to control metabolic flux in a new pathway is similar in principle to the use of protein scaffolds to control flux (9) and complements existing metabolic engineering strategies based on the control of gene expression by using gene knockouts (4), the control of promoter strength and plasmid copy number (1), and changing ribosomal binding site or promoter strength (37). The synthetic pathway we constructed provides a new process for producing 3-methyl-3-butenol, 3-methyl-2-butenol, and 3-methylbutanol from IPP. The maximum theoretical production of C₅ alcohols from glucose using the mevalonate pathway is 0.33 g/g. Overexpression of NudB achieved 55 mg/liter of 3-methyl-3-butenol from 2 g/liter of glucose, which is 8.3% of the theoretical maximum, and the overexpression of Idi1-NudB and NemA achieved 4.8% of the theoretical maximum. These titers do not account for the volatility of C₅ alcohols and their loss to the headspace during production, similar to the production of other IPP-derived compounds, such as isoprene and amorphadiene (23). Additionally, the yield from this new pathway could be improved by building on previous engineering efforts that increased isoprenoid yields (33).

The shift from nonrenewable to renewable resources for the synthesis of chemical building blocks used by various industries will require the invention of new processes to produce molecules that are functionally identical to those synthesized from nonrenewable resources. The field of synthetic chemistry facilitated the development of processes to produce novel compounds from basic laboratory materials. Similarly, metabolic engineering has the potential to produce novel compounds from renewable resources by engineering cells. To achieve that potential, new enzymatic reactions are necessary for assembling synthetic pathways, so that engineers are neither limited to producing compounds from natural pathways nor reliant on known enzymatic reactions. The framework we describe here for discovering enzymes able to catalyze new reactions not discovered in nature yet could be used to assemble a large number of synthetic biological pathways for producing chemicals with industrial applications.