1. Biochemistry and Chemical Biology
  2. Plant Biology
Download icon

Adaptation of hydroxymethylbutenyl diphosphate reductase enables volatile isoprenoid production

  1. Mareike Bongers  Is a corresponding author
  2. Jordi Perez-Gil
  3. Mark P Hodson
  4. Lars Schrübbers
  5. Tune Wulff
  6. Morten OA Sommer
  7. Lars K Nielsen
  8. Claudia E Vickers  Is a corresponding author
  1. Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Denmark
  2. Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Australia
  3. Centre for Research in Agricultural Genomics (CRAG) CSIC-IRTA-UAB-UB, Campus UAB Bellaterra, Spain
  4. Metabolomics Australia, Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Australia
  5. School of Pharmacy, The University of Queensland, Australia
  6. CSIRO Synthetic Biology Future Science Platform, Australia
Short Report
  • Cited 0
  • Views 478
  • Annotations
Cite this article as: eLife 2020;9:e48685 doi: 10.7554/eLife.48685

Abstract

Volatile isoprenoids produced by plants are emitted in vast quantities into the atmosphere, with substantial effects on global carbon cycling. Yet, the molecular mechanisms regulating the balance between volatile and non-volatile isoprenoid production remain unknown. Isoprenoids are synthesised via sequential condensation of isopentenyl pyrophosphate (IPP) to dimethylallyl pyrophosphate (DMAPP), with volatile isoprenoids containing fewer isopentenyl subunits. The DMAPP:IPP ratio could affect the balance between volatile and non-volatile isoprenoids, but the plastidic DMAPP:IPP ratio is generally believed to be similar across different species. Here we demonstrate that the ratio of DMAPP:IPP produced by hydroxymethylbutenyl diphosphate reductase (HDR/IspH), the final step of the plastidic isoprenoid production pathway, is not fixed. Instead, this ratio varies greatly across HDRs from phylogenetically distinct plants, correlating with isoprenoid production patterns. Our findings suggest that adaptation of HDR plays a previously unrecognised role in determining in vivo carbon availability for isoprenoid emissions, directly shaping global biosphere-atmosphere interactions.

Introduction

Biogenic volatile organic compounds (BVOCs) emitted from the biosphere have significant effects on global climate and air quality (Loreto and Fares, 2013). Short-chain isoprenoids such as isoprene, a C5 hydrocarbon, contribute more than 80% of BVOCs, totalling about 650 million tonnes of carbon per year (Sindelarova et al., 2014). The vast quantity and high reactivity of emitted volatile isoprenoids affect the oxidative capacity of the troposphere (Thompson, 1992; Wennberg et al., 2018), impact the residence time of the greenhouse gas methane (Fehsenfeld et al., 1992), and contribute to air pollution through formation of secondary organic aerosols, surface-level ozone and carbon monoxide (Claeys et al., 2004; Poisson et al., 2000; Granier et al., 2000; Figure 1). The effects of isoprenoid emissions may be exacerbated by climate change and shifts in land use (Peñuelas and Staudt, 2010), warranting a better understanding of how plants accomplish and regulate these vast emissions.

Simplified scheme of the plastidic MEP pathway, important volatile isoprenoids, and their atmospheric reactions.

The MEP pathway makes IPP and DMAPP simultaneously through the action of HDR (pink box), and produces the bulk of volatile isoprenoids, contributing >80 % of total BVOCs (Sindelarova et al., 2014) . Non-volatile isoprenoids are essential and synthesised by all organisms, while volatile isoprenoid production is non-essential and highly species-dependent. The cytosolic MVA pathway contributes most sesquiterpenes (<3 % of BVOCs), but is omitted here for clarity. Emitted volatile isoprenoids are rapidly oxidised, resulting in complex atmospheric photochemistry impacting aerosol and cloud condensation nuclei formation, extension of methane residence time, ozonolysis as well as surface-level ozone formation in the presence of mono-nitrogen oxide (NOx) pollutants (Wennberg et al., 2018). BVOCs, biogenic organic volatile compounds; DMAPP, dimethylallyl pyrophosphate; DXS, deoxyxylulose synthase; IDI, isopentenyl diphosphate isomerase; IPP, isopentenyl pyrophosphate; IspS, isoprene synthase; HDR, hydroxymethylbutenyl diphosphate reductase.

All isoprenoids are made from the C5 isomers isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP) (Figure 1). Two non-homologous metabolic pathways produce DMAPP and IPP in plants: the cytosolic mevalonic acid (MVA) and the plastidic methylerythritol phosphate (MEP) pathways, the latter contributing almost all volatile isoprenoids (Pulido et al., 2012). The final step of the MEP pathway is catalyzed by the enzyme hydroxymethylbutenyl diphosphate reductase (HDR/IspH), which produces both IPP and DMAPP (Figure 1). Isoprenoid chain length is initially determined by how many units of IPP are condensed with one molecule of DMAPP, before terpene synthases and other modifying enzymes convert these intermediates into isoprenoids. The resulting compounds are classified by carbon chain length.

In plants, longer-chain isoprenoids (C15 and higher) serve many essential roles, e.g. as membrane components and parts of the photosynthetic apparatus (Pulido et al., 2012; Figure 1). Short-chain isoprenoids (C5, C10, and some C15 compounds) are volatile under physiological conditions, and their functions are generally not essential for plant survival (Vickers et al., 2009). It is currently unknown how plants control carbon allocation between short-chain and long-chain isoprenoids in the chloroplast. While the demand for essential isoprenoids (for example, photosynthetic pigments) is assumed to be relatively similar across plants (Monson et al., 2013), different species produce markedly different amounts of non-essential, short-chain volatile isoprenoids (Wiedinmyer et al., 2020). For example, some oak (Quercus) species produce vast amounts of isoprene, while closely related oaks produce little or none at all (Wiedinmyer et al., 2020). Synthesis of isoprenoids with different chain lengths requires different DMAPP:IPP substrate ratios. Much more IPP than DMAPP is needed for long-chain isoprenoid production, so presumably high relative IPP concentrations are necessary for chain elongation while an excess of DMAPP and insufficient IPP could favour short-chain isoprenoid production. Isoprene synthase (IspS) uses only DMAPP, but not IPP, as a substrate.

Volatile isoprenoid emissions can represent a significant loss of carbon; for example, up to 20% of recently fixed carbon can be emitted as isoprene in high-emitting plants (Sharkey and Loreto, 1993). Isoprene synthase (IspS) has a high Km for its substrate DMAPP (0.5–8 mM; BRENDA, 2020); despite this, it successfully competes with prenyl phosphate synthases, which typically have KM(DMAPP) values 10- to 100-fold lower (BRENDA, 2020). Similarly, monoterpene synthases, which also show lower affinity for the substrates (BRENDA, 2020), compete with downstream prenyl phosphate synthases. Hence, the relative abundance of DMAPP may determine the balance between volatile and non-volatile isoprenoids.

Here we examined HDR as a potential mechanism to provide variability in the DMAPP:IPP ratio. Previous studies in diverse organisms (Escherichia coli, the bacterium Aquifex aeolicus, red pepper chromoplasts, and cultured tobacco cells) all found that HDR produces DMAPP:IPP ratios between 1:4 and 1:6 (Rohdich et al., 2003Altincicek et al., 2002Adam et al., 2002Tritsch et al., 2010). Consequently, it has been assumed that HDR has a fixed product ratio of about 1:5. However, none of these species produce significant amounts of volatile isoprenoids (Wiedinmyer et al., 2020). Isopentenyl diphosphate isomerase (IDI) interconverts DMAPP and IPP, but the reaction is slow (Jonnalagadda et al., 2012) and IDI is rate-limiting for isoprenoid production generally, including isoprene (Vickers et al., 2014). We hypothesised that HDR enzymes from species that emit large amounts of short-chain volatile isoprenoids produce a higher ratio of DMAPP to IPP, which could support production of volatiles like isoprene.

Results and discussion

We selected HDR genes from the bacterium E. coli, Synechococcus sp. strain PCC 7002 (a photosynthetic prokaryote) and eight species from diverse taxa of the plant kingdom (Table 1). Many plants harbour more than one annotated HDR gene, some of which may be pseudogenes. Therefore, we first identified functional HDR genes by their ability to complement an otherwise lethal knockout of the ispH/HDR gene in E. coli (Altincicek et al., 2001). We found at least one functional gene from each species (Figure 2—figure supplement 1a); however, severe dose-dependent growth defects were observed when overexpressing certain HDR genes, possibly due to toxicity of prenyl phosphates (George et al., 2018; Figure 2—figure supplement 1b). This precluded accurate steady-state metabolite quantification and required alleviating toxicity by the introduction of a metabolic sink for IPP and DMAPP. Here we used a lycopene (C40 isoprenoid) biosynthetic pathway, including expression of a heterologous idi (Cunningham et al., 1994). Deoxyxylulose synthase (DXS), the primary rate-limiting step of the MEP pathway, was also overexpressed in order to achieve intracellular IPP and DMAPP concentrations above quantification limits in E. coli.

Table 1
Genetic information and volatile isoprenoid emission profiles for species studied in this work.

Key: blank cell indicates species has not been tested, or genome sequence (or other information) not available; Y indicates significant emissions of isoprene or isoprenoids have been detected, or gene/transcript has been identified; N indicates significant emissions of isoprene or isoprenoids have NOT been detected, or gene/transcript has NOT been identified; MTs, monoterpenes; IspS, isoprene synthase; TPS, terpene synthase.

EmissionsGene/transcript*
KingdomPhylum/CladeCladeGenus, speciesCommon NameHDR protein accession numberE. coli construct Genbank IDComplements?Isoprene (C5)MTs (C10)IspSShort chain TPSReference
PlantaeAngiospermsEudicotsRicinus communiscastor bean plantXP_002519102.1MH605331yesNYNYWiedinmyer et al., 2020Kadri et al., 2011Xie et al., 2012)
PlantaeAngiospermsEudicotsPopulus trichocarpablack cottonwood1ACD70402MH605329yesYYYYWiedinmyer et al., 2020Tuskan, 2006)
2PNT41333.1MH605330no
PlantaeAngiospermsEudicotsPrunus persicapeachXP_007199828.1MH605326yesNYNYWiedinmyer et al., 2020Verde et al., 2013)
PlantaeAngiospermsEudicotsEucalyptus grandisflooded gum1XP_010028563.1MH605323yesYYYYWiedinmyer et al., 2020Myburg et al., 2014
 2XP_010047332.1MH605324no
PlantaeAngiospermsEudicotsTheobroma cacaocacao treeXP_007042717.1MH605333yesNYNYWiedinmyer et al., 2020Argout et al., 2008
PlantaeAngiospermsEudicotsArabidopsis thalianathale cressAEE86362.1MH605322yesNYNYSharkey et al., 2005Chen et al., 2004Bohlmann et al., 2000
PlantaeAngiospermsMonocotsElaeis guineensisoil palmXP_010909277.1MH605325yesYYWiedinmyer et al., 2020; Wilkinson et al., 2006
PlantaeGymnospermsPinophytaPicea sitchensisSitka spruce1ACN40284.1MH605327yesYYYWiedinmyer et al., 2020Hayward et al., 2004
 2ACN39959.1MH605328yes – toxic
BacteriaCyanobacteriaSynechococcus sp. PCC 7002SynechococcusACA98524.1MH605332yesNNN
  1. * Identified from data/genomes available on NCBI (https://www.ncbi.nlm.nih.gov/) and literature search (references noted).

    † Whether protein expression was able to functionally complement an E. coli ΔispH knockout in this study.

  2. ‡ Also known as Populus balsamifera ssp. trichocarpa.

A spectrum of DMAPP:IPP ratios was observed, ranging from almost exclusive IPP production (Picea sitchensis HDR1) to almost exclusive DMAPP production (Populus trichocarpa and Ricinus communis, Figure 2a). A control without HDR overexpression (labelled (-) in Figure 2a) showed a DMAPP:IPP ratio of ~1.5 to 1 in our experimental setup, serving as a reference point. Overexpressing the E. coli HDR shifted the ratio slightly towards IPP, in agreement with previous reports (Rohdich et al., 2002). However, HDR enzymes from species known to emit volatile isoprenoids produced considerably more DMAPP - a noteworthy exception being P. sitchensis HDR1 (PsHDR1, Figure 2a).

Figure 2 with 3 supplements see all
DMAPP:IPP ratio and isoprene production with different HDR enzymes.

(a) In vivo ratio of DMAPP:IPP measured via LC-MS/MS in E. coli overexpressing HDR genes from different species, in the genetic context of dxs and lycopene biosynthetic pathway overexpression. Filled circles and squares indicate that the HDR source species natively emits C5 or C10 isoprenoids. Open symbols indicate no emission, and no symbol indicates no data or conflicting data. (b) Isoprene production in E. coli when the HDR enzymes shown in panel (a) are overexpressed with dxs and an isoprene synthase. (c) Comparison of DMAPP:IPP ratios between selected HDRs co-expressed with dxs and with expression of either lycopene or isoprene as the metabolic sink. (d) Comparison of DMAPP:IPP ratios in E. coli overexpressing Picea sitchensis (Ps) HDR1 or HDR2 in the context of dxs and lycopene biosynthetic pathway overexpression. (e) Isoprene production in E. coli overexpressing P. sitchensis HDR1 or HDR2 along with dxs and an isoprene synthase. (f) The maximum specific growth rate (µmax) of E. coli expressing selected HDRs in the context of dxs and lycopene biosynthetic pathway overexpression, with or without induction of HDR expression by addition of IPTG. All data shown as mean ± SD from > 3 biological replicates; (-) indicates the control strain without HDR overexpression.

These values do not represent direct product ratios of the examined HDRs due to the presence of the heterologously expressed lycopene pathway and idi. However, they show that product ratios vary up to 40-fold between HDRs, and that the assumed fixed 1:5 DMAPP to IPP ratio is in fact an exception, rather than the rule. Using LC-MS proteomics, we tested whether the observed phenotypes were influenced by differences in expression of the native E. coli HDR, IDI, or the plasmid-encoded lycopene production pathway. We found no difference in protein levels in any of the HDR overexpression strains compared to the no HDR overexpression control (one-way ANOVA, p>0.05), except for the anticipated increase in E. coli HDR in the respective overexpression strain (Welch’s ANOVA, Dunnett’s post hoc test p<0.005; Figure 2—figure supplement 2). Because no shared proteotypic peptides exist across all heterologous HDRs, quantitative comparison of HDR protein levels across strains is not possible. However, we confirmed that all tested HDRs were strongly overexpressed (Figure 2—figure supplement 2c), and that there was no correlation between HDR abundance and the DMAPP:IPP ratio (rho = −0.488, data not shown). Taken together, these data demonstrate that different HDR enzymes produce vastly different DMAPP:IPP ratios, with some plant HDRs producing a ratio significantly shifted towards more DMAPP than previously recognized.

To test whether an increased in vivo DMAPP:IPP ratio would favour isoprene production, we replaced the lycopene pathway with an overexpressed isoprene synthase (IspS) as a metabolic sink. A high DMAPP:IPP ratio was indeed closely associated with isoprene production (Figure 2b). To confirm that differences in DMAPP:IPP ratios are robust when changing from lycopene (C40) to isoprene (C5) production, we compared selected HDR product ratios with both downstream metabolic sinks (Figure 2c). While the absolute values shifted towards DMAPP (left y-axis; lycopene requires 6 IPP and 2 DMAPP) or IPP (right y-axis; isoprene is made only from DMAPP) depending on downstream requirements, the relative difference between HDRs remained similar, demonstrating that our experimental setup captures representative differences between the enzymes.

Isoprene was not produced in the presence of Theobroma cacao, Arabidopsis thalianaor E. coli HDR (all species that do not emit short-chain isoprenoids), presumably because the available DMAPP was insufficient for IspS to compete with downstream enzymes (Figure 2b). All HDRs from isoprenoid-emitting species enabled isoprene production, supporting our hypothesis. Interestingly, a high DMAPP:IPP ratio and high isoprene production was also observed with HDRs from P. persica and R. communis, species that emit some monoterpenes but not isoprene (Wiedinmyer et al., 2020; Kadri et al., 2011). PpHDR and RcHDR have high (>87%) sequence identity with HDR proteins from high isoprene-emitting species P. trichocarpa and Hevea brasiliensis, respectively (Figure 3), but R. communis and P. persica do not have an isoprene synthase (Table 1).

Phylogenetic tree of HDR proteins from land plants, the cyanobacterium Synechococcus and Escherichia coli.

Where known, each species’ C5 (isoprene) and C10 (monoterpenes) emission spectra are shown (Wiedinmyer et al., 2020). High DMAPP-producing HDR proteins (from P. trichocarpa, R. communis and P. persica) cluster together based on high sequence similarity. Homologues within species, such as P. trichocarpa, tend to be highly similar; except for in gymnosperms where two separate groups of likely paralogous HDRs exist. Proteins analysed in this study are highlighted in bold. The Asterids clade is collapsed for clarity. Tree generated from BLAST sequence alignment with A. thaliana HDR against all land plants, using maximum likelihood phylogeny. Empty symbol, no volatile emission; filled symbol, volatile emission; no symbol, no or conflicting data available.

Together, these data suggest that HDR from different plant species has adapted to produce differing ratios of DMAPP to IPP, and that an increased DMAPP:IPP ratio is an important prerequisite for production of isoprene and perhaps other non-essential, short-chain isoprenoids. Our data indicate that a high DMAPP:IPP ratio is a necessary, but not a sufficient requirement for volatile isoprenoid emission. This places HDR at a key junction in the evolution of isoprene emission, a trait that appeared and disappeared several times across the plant kingdom (Dani et al., 2014).

Picea sitchensis (Sitka spruce) is a coniferous gymnosperm that emits both isoprene and monoterpenes (Hayward et al., 2004), but contrary to our expectation PsHDR1 produced the highest relative amount of IPP and showed very low isoprene production in E. coli (Figure 2a and b). Recently, the HDR from another gymnosperm, Ginkgo biloba (GbHDR1), was shown to produce an even lower DMAPP to IPP ratio in vitro (Shin et al., 2017). Most sequenced gymnosperms have two or more HDR isoforms which fall into two distinct classes based on sequence similarity (Kim et al., 2008; Figure 3). Interestingly, transcriptional studies (Celedon et al., 2017Kim et al., 2009) suggest that gymnosperm Type II HDRs are particularly abundant at the site of monoterpene-rich resin formation and are generally expressed at higher levels than Type I HDRs (Celedon et al., 2017) (such as PsHDR1 and GbHDR1). It was therefore tempting to speculate that HDR adaptation in gymnosperms has resulted in paralogues with complementary functions: Type I HDRs, which primarily produce IPP, show basal expression throughout the plant, and are important for long-chain isoprenoid production; and Type II HDRs, which primarily produce DMAPP and are expressed where short-chain isoprenoids are made. This prompted us to investigate the Type II HDR from P. sitchensis (Figure 2d–f).

PsHDR2 failed in our initial complementation assay (data not shown), most likely due to toxicity as no metabolic sink was present for IPP/DMAPP. Indeed, even in the presence of a sink, overexpression of PsHDR2 reduced E. coli growth rate by about 50% (Figure 2f), a level of toxicity exceeding that of other high DMAPP-producing HDRs. Interestingly, PsHDR2 produced a > 10 fold excess of DMAPP over IPP, while PsHDR1 had a ratio shifted towards more IPP (DMAPP:IPP = 0.447 +/- 0.19; Figure 2d). PsHDR2 also enabled higher isoprene production than PsHDR1 (Figure 2e), albeit at a lower yield than the other high DMAPP-producing enzymes, which is most likely an effect of the high toxicity in E. coli. The complementary product ratios of PsHDR1 and PsHDR2 strongly suggest functional specialization of these genes, making them paralogues in P. sitchensis.

While many plants encode more than one HDR gene (Figure 3), these homologues are often closely related and thus likely arose from relatively recent large-scale genome duplications (Saladié et al., 2014). In gymnosperms, the two HDR homologues are phylogenetically more distant (Figure 3) and likely define functionally specialised paralogues. Hence, we propose that two different strategies might have been employed to adapt HDR to isoprenoid production spectra: either using a single HDR and shifting the DMAPP:IPP ratio to allow production of specific isoprenoid profiles (Figure 2a), or having two functionally distinct HDRs each dedicated to the synthesis of one isomer (Figure 2d). Whether adaptation of HDR is a result of a change in the demand for DMAPP, or whether it is a driver of its release as isoprene and other volatile isoprenoids, is a fascinating question that remains to be answered.

The discovery of HDR enzymes with different product ratios has important implications for heterologous production of industrially valuable isoprenoids such as biofuels, fragrances and pharmaceuticals (Vickers, 2015) in engineered microorganisms. We have shown that only certain HDR enzymes enable production of isoprene in our engineered E. coli, and our data indicate that the choice of HDR is important to ensure availability of DMAPP and IPP at appropriate relative concentrations to achieve balanced pathway flux towards the product of interest and to avoid DMAPP toxicity. The presented LC-MS/MS method for separation and absolute quantification of the two isomers (Figure 2—figure supplement 3) proved crucial for our discovery, and will enable a deeper understanding of the processes regulating isoprenoid biosynthesis in nature and biotechnology.

Demands from downstream metabolism may determine IPP and DMAPP requirements, and could form an evolutionary driver for enzymatic activities that impact their ratio. Our data suggest that the adaptation of HDR to generate different DMAPP:IPP ratios allows for the production of large amounts of short-chain isoprenoids in certain species or tissues. Our findings illuminate the molecular mechanism underlying how plants emit isoprene and suggest a central role for HDR in determining the spectrum of isoprenoids produced by plants, including isoprenoid BVOCs. Unravelling the mechanism by which plants distribute carbon between volatile and non-volatile isoprenoids will help resolve the complex interplay between BVOC emissions, land-use management and climate change.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Gene (Escherichia coli)ispH/HDRNCBI ‘Gene’Gene_ID:944777; EcoGene:EG11081; ECK0030; lytBhydroxymethylbutenyl diphosphate reductase
Strain, strain background (Escherichia coli)Escherichia coli WATCCATCC:9637obtained from L. Nielsen lab, Australia
Genetic reagent (Escherichia coli)E. coli W∆cscR, lacZ::PtDXS, arsB::PaISPSThis paper and PMID: 21782859 (Arifin et al., 2011)knockout of cscR, knock-in of PtDXS and PaISPS
Genetic reagent (Escherichia coli)E. coli WΔcscR, lacZ::MVA, ∆ispHThis paper and PMID: 11115399 (Campos et al., 2001)knock-in of MVA pathway, knockout of ispH
Genetic reagent (Populus trichocarpa)DXSNCBI ‘Reference Sequence’XP_006378082.1Deoxyxylulose phosphate synthase, gene was truncated for expression in E. coli
Genetic reagent (Populus alba)ISPS(del2-52,A3T,L70R,S288C)Patent WO2012058494 (Beck et al., 2011)Isoprene synthase (Genbank:EF638224) variant, truncated and mutated
Recombinant DNA reagentpLacZ-KIKO(cm) plasmidPMID: 23799955 (Sabri et al., 2013)Addgene:46764used to integrate PtDXS into the genome
Recombinant DNA reagentpArsBKIKO(cm) plasmidPMID: 23799955 (Sabri et al., 2013)Addgene:46763used to integrate PaISPS into the genome
Recombinant DNA reagentpT-HDR plasmidsThis paperderived from pTrc99aall HDR genes were cloned into this expression vector
Recombinant DNA reagentpAC-LYC04PMID: 7919981 (Cunningham et al., 1994)
Recombinant DNA reagentRicinus communis HDR expression plasmidGenbankMH605331HDR protein XP_002519102.1
Recombinant DNA reagentPopulus trichocarpa HDR 1 expression plasmidGenbankMH605329HDR protein ACD70402
Recombinant DNA reagentPopulus trichocarpa HDR 2 expression plasmidGenbankMH605330HDR protein PNT41333.1
Recombinant DNA reagentPrunus persica HDR expression plasmidGenbankMH605326HDR protein XP_007199828.1
Recombinant DNA reagentEucalyptus grandis HDR 1 expression plasmidGenbankMH605323HDR protein XP_010028563.1
Recombinant DNA reagentEucalyptus grandis HDR 2 expression plasmidGenbankMH605324HDR protein XP_010047332.1
Recombinant DNA reagentTheobroma cacao HDR expression plasmidGenbankMH605333HDR protein XP_007042717.1
Recombinant DNA reagentArabidopsis thaliana HDR expression plasmidGenbankMH605322HDR protein AEE86362.1
Recombinant DNA reagentElaeis guineensis HDR expression plasmidGenbankMH605325HDR protein XP_010909277.1
Recombinant DNA reagentPicea sitchensis HDR 1 expression plasmidGenbankMH605327HDR protein ACN40284.1
Recombinant DNA reagentPicea sitchensis HDR 2 expression plasmidGenbankMH605328HDR protein ACN39959.1
Recombinant DNA reagentSynechococcus sp. PCC 7002 HDR expression plasmidGenbankMH605332HDR protein ACA98524.1
Commercial assay or kitAstec Cyclobond I2000 chiral HPLC columnSigma Aldrich20024ASTHPLC column used for IPP/DMAPP separation
Chemical compound, drugIsopreneSigma AldrichCat. # I19551
Chemical compound, drugIsopentenyl pyrophosphateSigma AldrichCat. # I0503
Chemical compound, drugDimethylallyl pyrophosphateSigma AldrichCat. # D4287
Chemical compound, drug(±)-Mevalonic acid 5-phosphateSigma AldrichCat. # 79849
Chemical compound, drugMevalonolactoneSigma AldrichCat. # M4667
Software, algorithmCLC Main WorkbenchQiagenRRID:SCR_000354
Software, algorithmiTOLPMID: 27095192 (Letunic and Bork, 2016)https://itol.embl.de/Interactive Tree of Life

Chemicals and reagents

Request a detailed protocol

Isoprene (Cat. No I19551), IPP (Ca. No I0503), DMAPP (Ca. No D4287), Isopropyl β-D-thiogalactoside (IPTG, Cat. No I6758), (±)-Mevalonic acid 5-phosphate (MVA-P, Ca. No 79849) were purchased from Sigma Aldrich. Mevalonate (MVA) was prepared from (±)-mevalonolactone (Sigma Aldrich, Cat. No M4667) through base-catalyzed hydrolysis (Campos et al., 2001). Ammonium acetate was purchased from Sigma Aldrich (Ca. No 73594–25 G-F). Acetonitrile hypergrade for LC-MS LiChrosolv (Ca. No 1000292500) and Methanol hypergrade for LC-MS LiChrosolv (Ca. No 1060352500) was purchased from Merck Millipore. Milli-Q water was generated via a Merck Millipore Integral 3 water purification system.

Gene, plasmid and E. coli strain construction

Request a detailed protocol

E. coli Top10 (Cat. No C404050, Thermo Fischer Scientific) was used for cloning. For all other experiments, E. coli W (ATCC 9637) with a knock-out in the csc operon (E. coli W∆cscR Arifin et al., 2011) was used. Plant HDR chloroplast targeting peptides were predicted using the ChloroP 1.1 server (http://www.cbs.dtu.dk/services/ChloroP/). Genes were truncated to remove chloroplast targeting peptides, codon-optimised for E. coli (http://idtdna.com/CodonOpt) and synthesised by Integrated DNA Technologies (Singapore). All plant genes were placed under control of the IPTG-inducible trc promoter in a pTrc99-derived (Amann et al., 1988) vector, generating the pT-HDR series of plasmids. The DXS gene from Populus trichocarpa (Genbank Accession No. XP_006378082.1) was integrated into the genome using the pLacZ-KIKO(cm) vector (Sabri et al., 2013). The chloramphenicol resistance gene was removed from the genome using pCP20 (Datsenko and Wanner, 2000). The resulting strain (E. coli W∆cscR, lacZ::PtDXS) was transformed with each of the pT-HDR plasmids and pAC-LYC04 (Cunningham et al., 1994) for IPP and DMAPP measurements. For isoprene production experiments, an engineered ISPS gene from Populus alba (Genbank Accession No. EF638224) was integrated into the genome of E. coli W∆cscR, lacZ::PtDXS using pArsBKIKO(cm). Apart from removal of the chloroplast-targeting sequence, this gene was also engineered to contain three mutations to enhance specific activity: ISPS(del2-52,A3T,L70R,S288C) (Beck et al., 2011).

Bacterial growth media

Request a detailed protocol

LB medium contained 10 g/L tryptone, 5 g/L yeast extract and 10 g/L NaCl. TB medium contained 12 g/L tryptone, 24 g/L yeast extract, 0.4% (v/v) glycerol, 2 mM MgSO4, 1 mM thiamine, 17 mM KH2PO4, 7.2 mM K2HPO4. Where indicated, media were supplemented with 1 mM mevalonate and 1 mM L-arabinose for induction of the MVA pathway operon, or with 0.2% (w/v) glucose or 0.1 mM IPTG for repression or induction of the trc promoter. All cultures were grown at 37°C with 250 rpm shaking unless stated otherwise.

Complementation of the ispH/HDR knockout mutant in E. coli

Request a detailed protocol

A partial MVA pathway under control of the arabinose-inducible PBAD promoter (Campos et al., 2001) was cloned into a pLacZ-KIKO(cm) vector and integrated into the E. colicscR genome. This strain (WΔcscR, lacZ::MVA) was used to knock out ispH using recombineering (Datsenko and Wanner, 2000), making growth dependent on supplementation with mevalonate and arabinose. Each pT-HDR plasmid was transformed into this strain and tested for its ability to grow in the absence of mevalonate and arabinose.

Growth rate measurements

Request a detailed protocol

Cells were grown in LB medium; glucose, mevalonate or IPTG were added where indicated. Precultures were grown at 37°C with 250 rpm shaking in 96-well plates (Corning, Cat No. CLS3799) until stationary phase. Cultures were diluted to a starting optical density (OD600) of 0.05 and the growth was monitored in a microplate reader (BioTek ELx808) at 37°C with 700 rpm double-orbital shaking, measuring OD600 every 10 min. All bacterial cultures for quantification of specific growth rates, metabolites and isoprene were grown at least in biological triplicates (from 3 single colonies of the same strain), and means +/- standard deviations are shown.

Fermentations for metabolite measurements

Request a detailed protocol

Strains harbouring the different pT-HDR plasmids and pAC-LYC04 were grown for determination of IPP and DMAPP concentrations. Chloramphenicol (30 mg L−1) and ampicillin (250 mg L−1) were added to the media for plasmid maintenance. Precultures were grown in LB medium as described above. A culture volume of 10 ml of TB medium was inoculated with an overnight preculture in 100 ml baffled flasks to a starting OD600 of 0.05. Protein expression was induced with 0.1 mM IPTG at an OD600 of 0.5. When an OD600 of 5 was reached (exponential growth phase in TB medium), cultures were harvested for metabolite quantification.

Quantification of IPP and DMAPP

Request a detailed protocol

Intracellular metabolites were quenched and extracted using a method adapted from Bongers et al. (2015). To harvest, the equivalent of 1 ml of culture of an optical density of OD600 = 5 was centrifuged at 4°C for 20 s at 13,000 x g, the supernatant was discarded and the pellet snap-frozen in liquid nitrogen. The pellet was resuspended in 95 µl of 90% acetonitrile (v/v) in water and metabolites were extracted by vortexing for 10 min at room temperature. Cell debris was removed by centrifugation at 4°C for 15 min at 13,000 x g. Extracts were transferred into HPLC vials, 5 µl internal standard (MVA-P) was added at a final concentration of 16 µM for analysis using liquid chromatography tandem mass spectrometry (LC-MS/MS).

LC-MS/MS data were acquired on an Advance UHPLC system (Bruker Daltonics, Fremont, CA, USA) equipped with a binary pump, degasser and PAL HTC-xt autosampler (CTC Analytics AG, Switzerland) coupled to an EVOQ Elite triple quadrupole mass spectrometer (Bruker Daltonics, Fremont, CA, USA). Separation of the structural isomers IPP and DMAPP was achieved by adapting a method from Köhling et al. (2014), by injecting 5 μl onto an Astec Cyclobond I2000 chiral HPLC column (250 mm ×4.6 mm; 5 μm particle size) (Sigma Aldrich) with an injection loop size of 2 µL. The column oven temperature was controlled and maintained at 35°C throughout the acquisition and the mobile phases were as follows: 50 mM aqueous ammonium acetate (eluent A) and 90:10 (% v/v) acetonitrile:purified water (eluent B). The mobile phase flow rate was maintained at 600 μL/min and was introduced directly into the mass spectrometer with no split. The mobile phase gradient profile was as follows: Starting condition 100% eluent B, 0.0–1.0 min: 100% B to 25% B, 1.0–22.0 min: 25% B, 22.0–22.5 min: 25% B to 0% B, 22.5–23.0 min: 0% B, 23.0–24.0 min: 0% B to 100% B, 24.0–30.0 min: 100% B. The mass spectrometer was controlled by MS Workstation 8.2.1 software (Bruker Daltonics) using electrospray ionization operated in negative ion mode. The following parameters were used to acquire Multiple Reaction Monitoring (MRM) data: spray voltage: 3.0 kV, cone temperature: 350°C, cone gas flow 20, probe gas flow: 50, nebulizer gas flow: 50, heated probe temperature: 350°C, exhaust gas: on, CID: 1.5 mTorr. The MRM scan time was set to 1000 ms for DMAPP and IPP, and 200 ms for MVA-P with standard resolution for all transitions. The collision energy (CE) was optimised for each transition. The quantifier was m/z 245.0 → 79 (CE: 16 eV) and qualifier m/z 245.0→ 159 (CE: 16 eV) for both DMAPP and IPP. For the internal standard MVA-P the quantifier was m/z 227.0 → 79 (CE: 24 eV) and qualifier m/z 227.0→ 97 (CE: 13 eV). Initial retention times (RT) were 14.1 min (MVA-P) 19.2 min (DMAPP) and 23.6 min (IPP) but shifted to less retention as the column presumably deteriorated during the runs. For quality control (QC) and to ensure correct peak integration a 1 μM standard DMAPP/IPP mix was injected every 12th sample. The RTs decreased in a linear fashion from the first 1 μM QC standard to the last QC standard (n = 52) with 0.024 min, 0.044 min, and 0.061 min per injection for MVA-P, DMAPP, and IPP respectively (R2 = 0.990, R2 = 0.991, R2 = 0.989). Analytes were integrated manually.

To obtain quantitative data, a matrix-matched internal standard calibration was used. Analyte stock solutions were prepared in 90% (v/v) acetonitrile and were diluted with blank matrix extract, extracted with 90:10 (% v/v) acetonitrile:Milli-Q water). The internal standard was added to the final HPLC vial at a concentration of 16 μM. The calibration curve ranged from 0.25 μM to 10 μM with R2 values of 0.968 and 0.981 for DMAPP and IPP, respectively. For both calibration curves a 1/x2 weighting factor was applied. Sample concentrations lower than the lowest standard were obtained through extrapolation of the calibration curve. The limit of quantification (LOQ) was approximated, using the lowest standard as reference (0.25 μM, n = 4), as 10x the signal-to-noise ratio. The LOQ estimate was 0.033 and 0.045 μM for DMAPP and IPP respectively. The 1 μM QC standard (n = 8) recovery was 85.6 (RSD 18.7%) and 93.2 (RSD 15.9%) for DMAPP and IPP respectively. Additionally, five standards with different DMAPP/IPP ratios were injected to verify the ratio accuracy. DMAPP:IPP ratios fortified were 10, 2, 1, 0.5, and 0.1, while ratios found were 11.1, 1.8, 0.96, 0.56, and 0.10 (bias ranging from −9.9 to 12.5% with a mean bias of 1.9%).

Protein quantification

Request a detailed protocol

Cells were harvested for proteomics analyses at the same time point as metabolomics samples. Cell pellets corresponding to 1 ml of cultures of an optical density of OD600 = 5 were processed according to Rennig et al. (2019), both regarding preparation of samples, the applied gradient on the CapLC system and the settings for Orbitrap HF_X mass spectrometer. Here, a total of 1 µg of peptides/sample was injected into the mass spectrometer. After acquisition the raw files were analysed using Proteome Discoverer 2.3 (P.D. 2.3) in order to identify and quantify detected proteins. The following software settings were used: Fixed modification: Carbamidomethyl (C) and Variable modifications: oxidation of methionine residues. First search mass tolerance 10 ppm and a MS/MS tolerance of 0.02 Da., trypsin as proteolytic enzyme and allowing two missed cleavages. FDR was set at 0.1%. For match between runs the ΔRT was set to 0.2 min and the minimum peptide length was set to 7. As database for the searches the E. coli W proteome (UP000008525) was used combined with a contaminant database (cRAP) and the sequences of heterologous HDRs (see Table 1) and lycopene production proteins Idi (Genbank ID AAC32208.1), CrtE (WP026199135.1), CrtI (AAA64981.1), and CrtB (WP020503292.1). Normalization of the data across samples was done with P.D. 2.3. using total peptide amount, meaning all identified peptides in the individual samples are used for normalization, while using one file as master file to which all other counts are normalized. For quantification only unique peptides were used, and for all HDR proteins, hits were manually inspected to ensure correct identification and quantification. HDR overexpression strains were compared by analysing normalized peptide counts using one-way analysis of variance (ANOVA) or Welch’s ANOVA test in case of unequal variances, respectively. Where reported, p-values were adjusted for multiple comparison testing using Dunnett’s method, n ≥ 3 biological replicates.

Isoprene production

Request a detailed protocol

The different pT-HDR plasmids were transformed into E. coli W∆cscR, lacZ::Pt-DXS, arsB::PaISPS(del2-52,A3T,L70R,S288C). All growth media contained 250 mg L−1 ampicillin for plasmid maintenance. Strains were grown in LB medium until stationary phase, then diluted in 0.5 ml TB medium containing 0.1 mM IPTG to a starting OD600 of 0.1, and grown at 30°C, with 250 rpm shaking. Cultures were grown in 20 ml sealed gas chromatography vials and isoprene was quantified after 48 hr as described previously (Vickers et al., 2015).

Sequence alignments and generation of phylogenetic trees

Request a detailed protocol

HDR protein sequences were downloaded from the results of a BLASTP search with A. thaliana HDR against land plants (taxid: 3193), manually removing identical duplicates and obvious pseudogenes (deletions or mutations in highly conserved regions). Sequences were truncated to remove N-terminal chloroplast targeting sequences and aligned using CLC Main Workbench (Qiagen). HDR phylogenetic tree (unrooted) was generated using maximum likelihood phylogeny, neighbour-joining method, WAG protein substitution model, and bootstrap analysis with 100 replicates, also in CLC Main Workbench. Phylogenetic trees were visualised using Interactive Tree of Life (iTOL) v3 (Letunic and Bork, 2016).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
    Enzyme Database
    1. BRENDA
    (2020)
    Accessed June 8, 2018.
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
    Protocols for the Production and Analysis of Isoprenoids in Bacteria and Yeast
    1. CE Vickers
    2. M Bongers
    3. SF Bydder
    4. P Chrysanthopoulos
    5. MP Hodson
    (2015)
    In: T. J McGenity, K. N Timmis, B Nogales Fernández, editors. Hydrocarbon and Lipid Microbiology Protocols: Synthetic and Systems Biology - Applications. Berlin Heidelberg: Springer. pp. 23–52.
    https://doi.org/10.1007/8623_2015_107
  50. 50
    Microorganisms in Biorefineries
    1. CE Vickers
    (2015)
    303–334, Production of Industrially Relevant Isoprenoid Compounds in Engineered Microbes, Microorganisms in Biorefineries, 26, Berlin, Heidelberg, Springer, 10.1007/978-3-662-45209-7_11.
  51. 51
  52. 52
  53. 53
  54. 54

Decision letter

  1. Joerg Bohlmann
    Reviewing Editor; University of British Columbia, Canada
  2. Ian T Baldwin
    Senior Editor; Max Planck Institute for Chemical Ecology, Germany

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Adaptation of hydroxymethylbutenyl diphosphate reductase enables volatile isoprenoid production" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Based on the reviews below, we regret to inform you that your work will not be considered further for publication in eLife. While both reviewers found the work very interesting, they also expressed some major concerns. Reviewer 1 highlights a number of factors that would need to be better controlled to support the authors conclusions. To overcome these concerns additional experiments would be required.

Reviewer #1:

The architecture of the isoprenoid pathway requires fusion of two different C5 intermediates, DMADP and IDP, to produce all but the simplest products. Since the formation of the growing prenyl diphosphate chains is initiated with the allylic substrate DMADP and followed by multiple additions of the homo-allylic IDP, different ratios of DMADP:IDP are needed for different length isoprenoid products. Thus the ratio of DMADP:IDP available could control the rate of formation of different isoprenoid classes.

The authors of this contribution have discovered that for C5 intermediates derived from the MEP pathway, the ratio of DMADP:IDP is not fixed at 1:5 by the last enzyme of the pathway (hydroxymethylbutenyl diphosphate reductase- HDR) as once believed. Different HDR enzymes in different species seem to make a range of ratios, which might be important in controlling the formation of various isoprenoids, especially the gaseous isoprene, which is a product of DMADP solely. The finding of two apparent HDR paralogs in some gymnosperms is also fascinating as it suggests that plants might regulate DMADP:IDP ratios in different ways in distinct cell types depending on the varying spectrum of isoprenoids needed. This is an exciting story, but I have two concerns with the results reported.

1) The major concern is that the authors have not really measured the direct product ratio of the enzyme in a standard way by characterizing the isolated enzyme in vitro. Instead they have measured the concentrations of DMADP and IDP in living E. coli cells in which the enzymes have been over-expressed. This is partly understandable as HDR, like other Fe-S cluster proteins, is oxygen-sensitive and considered hard to assay in vitro. However, many other processes affect the in vivo concentrations of DMADP and IDP in living cells, and so the authors might do a better job of controlling these factors to make a stronger case that the differences they see in the HDRs of various species are real.

The authors have over-expressed these HDRs in cells with over-expressed DXS ("to achieve intracellular IDP and DMADP concentrations above the quantification limits", DXS is supposed to be the rate-controlling step of the pathway), and with a new IDI enzyme and an additional biosynthetic pathway for the carotenoid lycopene (apparently to provide a sink for all of the excess IDP and DMADP, which are toxic). The result seems a very artificial, high flux environment for any isoprenoid synthesizing process, and it is hard to know if the ratios measured really reflect the intrinsic properties of the HDR, particularly since the expressed HDR is always mixed with the E. coli HDR.

Each HDR was expressed in the same cassette with the same promoter, but the ratios could vary depending on the level of expression of the E. coli enzyme vs. the heterologous HDR. Can the authors verify (by proteomics) that the protein levels of the tested HDRs in each experiment were the same? This might help show that the in vivo pools reflect real differences among the enzymes. One might expect that the host E. coli HDR responds in a varying way to the expression of heterologous enzymes, and so its level should be measured also. Hopefully, its contribution is also constant.

Another approach might be to down-regulate the E. coli HDR when the foreign HDR to be tested is over-expressed. This might avoid the toxic effects, and allow the foreign HDR to be tested with less competition.

The authors might additionally determine not only the ratio of DMADP to IDP, but also the absolute amounts of the two compounds to see if they are in the range expected in a cell, considering all of the genetic manipulations being carried out.

I realise that all of these suggestions require additional work. Perhaps there is another way to validate such in vivo measurements as representing the real characteristics of the enzyme when tested in vitro. Maybe an attempt could be made to try expression, purification and in vitro assay of say two of the HDRs from opposite ends of the DMADP:IDP scale.

2) One of the authors' most impressive achievements was to develop an HPLC separation scheme for IDP and DMADP. But as this procedure has never been formally published in the refereed literature, the authors should provide chromatograms with accompanying mass spectra to verify their separation. Also, the previously published retention times vary greatly from those reported here (subsection “Quantification of IPP and DMAPP”), which seems to deserve some comment.

Reviewer #2:

This manuscript describes mechanistically interesting novel data on the important role of hydroxymethylbutenyl diphosphate reductase (HDR/IspH) regulating the ratio of DMAPP and IPP as prerequisite for enhanced metabolic flux to isoprene (and monoterpenes) or long-chain terpenoids such as lycopene.

The authors have chosen an elegant molecular biological approach for their investigations: In addition to HDR genes from E. coli and Synechococcus, they investigated 8 HDR genes from plant species of originating from different taxa of the plant. In a first step, functional HDR genes were identified by complementing lethal knockout mutants of HDR/IspH in E. coli. It was found that the overexpression of some HDR genes led to growth inhibition in E. coli, which was attributed to the toxicity of accumulating prenyl phosphates.

For the main investigations, the E. coli metabolism was modified in which an overexpressed DXR gene led to an increased metabolic flow of the MEP pathway; a lycopene synthase or an isoprene synthase were incorporated to produce a metabolic sink to higher terpenoids (C40) and volatile (C5) hemiterpenes. In addition, a heterologous IPP isomerase (IDI) was incorporated to ensure sufficient IPP for the biosynthesis of longer chain terpenoids.

DMAPP and IPP were measured with directed LC-MS/MS which is state-of-the-art.

It is impressive to see how different the DMAPP/IPP ratios are adjusted by the different HDR enzymes. Enzymes from plants with naturally high isoprene emission have very high DMAPP/IPP ratios. Enzymes from organisms without isoprene emission show very low ratios, similar to HDR enzymes from gymnosperms.

Very interesting is the result shown in Figure S4 that the DMAPP/IPP ratios in vivo depend on the end product of the metabolism: higher at isoprene emission, lower at lycopene accumulation. I would suggest to show this result in the main text. I would also like to see a more detailed discussion of this observation. A point that is also a bit neglected in the discussion is the role of IDI in the fine tuning of DMAPP/IPP ratios. As rightly noted, the DMAPP/IPP ratio measured in vivo does not correspond to the (not measured) in vitro ratio, but is changed by the activity of the IDI. For isoprene emitting oaks it was shown that the activity of the IDI correlates positively with ISPS and isoprene emission (Brüggemann and Schnitzler, 2002), which indicates a rate-limiting role of the IPI as mentioned.

To understand the fine tuning of the DMAPP/IPP ratio between HDR/IspH, IDI and the sinks towards the different soluble and volatile end products additional information on IDI is necessary: i.e. what is the in vitro (in vivo) ratio of the IDI used in the present experiments? I'm wondering why no experiments without overexpressed (and knocked out) IDI have been performed. How does the absence of this enzyme (absence of overexpression) influence the in vivo ratios of DMAPP/IPP in the different systems (lycopene vs isoprene)? I hope the authors can provide this information from experiments not yet shown.

Overall a very interesting novel work demonstrating the potential role of HDR/IspH to control the ratio of C5 terpenoid intermediates central for the channeling of metabolites to the higher terpenoids/isoprenoids and or volatile compounds, such as isoprene and monoterpenes. Beside understanding the metabolic regulation within the MEP pathway in natural systems knowledge on the ratio of the enzyme products of HDR/IspH also will provide the avenue for much more efficient biotechnological system for all type of terpenoids.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your revised article "Adaptation of hydroxymethylbutenyl diphosphate reductase enables volatile isoprenoid production" for consideration by eLife.

Your article has been reviewed by one of the two reviewers who have seen the original submission. The second reviewer was unfortunately no longer available to assess the revised paper. The evaluation has been overseen by a Reviewing Editor, Joerg Bohlmann, and Ian Baldwin as the Senior Editor. The reviewer has opted to remain anonymous.

We find that paper has been substantially improved, but there are a few remaining concerns in the external reviewer's report.

Reviewer #1:

The authors have made extensive revisions and significantly improved their manuscript. In particular, the new proteomics measurements help to assure that the activity of the over-expressed HDR proteins is not affected by the native E. coli protein or the expression of the genes on the lycopene plasmid. Obtaining in vitro assay data on these proteins would clearly be the "gold standard" for comparison among them, but I can accept that this is just not yet technically attainable, and they have done what they can to increase the accuracy of their in vivo measurements. I also appreciated the chromatograms to verify the DMADP-IDP separation.

Two minor concerns that could be addressed:

1) Compared to the first submission, the DMADP:IDP ratio for Prunus persica has now decreased by 10 fold, while the ratios for other species have declined by more than 50%. Since these are the core data of the paper, such big changes do not inspire confidence in the analytical methods. Is there a reason for such variance from the previous version, such as a change in the analysis protocol or a change in calibrating factors?

2) In subsection “Isoprene production” it states that "isoprene was quantified as described previously", but the reference given is incorrect. Please correct this.

https://doi.org/10.7554/eLife.48685.sa1

Author response

[Editors’ note: The authors appealed the original decision. What follows is the authors’ response to the first round of review.]

Reviewer #1:

The architecture of the isoprenoid pathway requires fusion of two different C5 intermediates, DMADP and IDP, to produce all but the simplest products. Since the formation of the growing prenyl diphosphate chains is initiated with the allylic substrate DMADP and followed by multiple additions of the homo-allylic IDP, different ratios of DMADP:IDP are needed for different length isoprenoid products. Thus the ratio of DMADP:IDP available could control the rate of formation of different isoprenoid classes.

The authors of this contribution have discovered that for C5 intermediates derived from the MEP pathway, the ratio of DMADP:IDP is not fixed at 1:5 by the last enzyme of the pathway (hydroxymethylbutenyl diphosphate reductase- HDR) as once believed. Different HDR enzymes in different species seem to make a range of ratios, which might be important in controlling the formation of various isoprenoids, especially the gaseous isoprene, which is a product of DMADP solely. The finding of two apparent HDR paralogs in some gymnosperms is also fascinating as it suggests that plants might regulate DMADP:IDP ratios in different ways in distinct cell types depending on the varying spectrum of isoprenoids needed. This is an exciting story, but I have two concerns with the results reported.

1) The major concern is that the authors have not really measured the direct product ratio of the enzyme in a standard way by characterizing the isolated enzyme in vitro. Instead they have measured the concentrations of DMADP and IDP in living E. coli cells in which the enzymes have been over-expressed. This is partly understandable as HDR, like other Fe-S cluster proteins, is oxygen-sensitive and considered hard to assay in vitro. However, many other processes affect the in vivo concentrations of DMADP and IDP in living cells, and so the authors might do a better job of controlling these factors to make a stronger case that the differences they see in the HDRs of various species are real.

The authors have over-expressed these HDRs in cells with over-expressed DXS ("to achieve intracellular IDP and DMADP concentrations above the quantification limits", DXS is supposed to be the rate-controlling step of the pathway), and with a new IDI enzyme and an additional biosynthetic pathway for the carotenoid lycopene (apparently to provide a sink for all of the excess IDP and DMADP, which are toxic). The result seems a very artificial, high flux environment for any isoprenoid synthesizing process, and it is hard to know if the ratios measured really reflect the intrinsic properties of the HDR, particularly since the expressed HDR is always mixed with the E. coli HDR.

This is a valid argument that we have now addressed with additional experiments. Our initial experimental setup was guided mainly by two factors:

a) being able to directly compare a relatively large number of HDR enzymes to each other, which precludes in vitro characterization of these difficult-to-express proteins, or expression and analysis in plants

and

b) using E. coli, which has a very low native flux through the MEP pathway, low MEP pathway gene expression levels and no apparent transcriptional or other regulation of MEP pathway genes in response to pathway engineering (Bongers et al., 2015). In this previous study, we established that MEP pathway gene expression in E. coli is not affected by expression of the lycopene pathway plasmid (including idi). With this in mind, our E. coli-based experimental system provides a much ‘cleaner’ background to study differences between plant HDRs compared to their native hosts.

In this revised submission, we have used LC-MS/MS proteomics to quantify proteins that could influence the in vivo ratio of DMAPP:IPP (Figure 2—figure supplement 2). We have confirmed that:

· native HDR protein abundance is the same across all strains including the negative control (except for the E. coli HDR overexpression strain, as intended), demonstrating that there is no regulation of E. coli HDR expression in response to overexpressing the heterologous HDRs.

· The same is true for the native E. coli Idi, as well as CrtE, CrtB, CrtI and HpIdi, expressed from the lycopene production plasmid. These data demonstrate that if these proteins affect the DMAPP:IPP ratio measured in our setup, their influence is constant across all strains.

· E. coli HDR peptide counts increase >50-fold in the EcHDR overexpression strain compared to all others. In the absence of a dedicated standard curve for the quantified peptides, counts cannot be assumed to increase linearly with protein abundance; however, these data strongly indicate a dramatic increase in HDR abundance upon overexpression compared to the baseline. The heterologous HDRs from other species (though not directly comparable, see comments below) show similarly strong expression levels as the overexpressed EcHDR. This further strengthens the view that the low and constant native EcHDR expression plays a minor role in determining the observed DMAPP:IPP ratios.

Expression of high DMAPP-producing HDRs causes severe toxicity in E. coli, making the co-overexpression of a metabolic sink (and, in some cases, idi) necessary to prevent cell death (Figure 2—figure supplement 1B). No growth defect was caused by overexpression of the native E. coli HDR, or by any HDRs that favour the production of IPP over DMAPP. This further supports the argument that the observed differences in DMAPP:IPP ratio cannot be sufficiently adjusted by E. coli’s native metabolism or available Idi activity.

We have now quantified both IPP and DMAPP in the manuscript (Figure 2—figure supplement 3), and calculated the intracellular concentrations of these compounds (using an accepted method for calculating cell count and volume based on culture density: Volkmer 2011, PloS One). The generated intracellular concentrations of DMAPP and IPP range between 0.5 – 8.7 μM, which may be high for E. coli, but certainly seems within physiologically relevant boundaries.

Each HDR was expressed in the same cassette with the same promoter, but the ratios could vary depending on the level of expression of the E. coli enzyme vs. the heterologous HDR. Can the authors verify (by proteomics) that the protein levels of the tested HDRs in each experiment were the same? This might help show that the in vivo pools reflect real differences among the enzymes.

Direct quantitative comparison between HDR expression levels is technically challenging because there are no proteotypic peptides that are shared between all HDRs in this set. The only way we could conceive to directly compare different HDR protein levels would be to use protein fusion tags such as His- or FLAG-tag and quantify based on those. We discarded this option because fusion tags may influence protein activity, and because (re-) cloning of high-DMAPP producing HDRs is very time-consuming due to toxicity issues. We now present LC-MS/MS-based proteomics data for all tested HDRs (Figure 2—figure supplement 2C), showing that all are highly expressed on the basis of total normalized counts and the number of peptides detected (between 28-52 peptides detected for each HDR). Using this method, it is not valid to compare counts for non-identical peptides because their different physicochemical properties affect behavior in the mass spectrometer, making results not quantitatively comparable. Since there is no single peptide shared by all tested HDRs, the presented data can only be seen as approximate indications of protein abundance. However, the data suggests that all HDRs are strongly overexpressed, and Pearson’s correlation analysis showed that there is no correlation between HDR ‘abundance’ and DMAPP:IPP ratio. This is now also explained in the main text.

One might expect that the host E. coli HDR responds in a varying way to the expression of heterologous enzymes, and so its level should be measured also. Hopefully, its contribution is also constant.

See above: EcHDR expression is constant, and relatively low unless it is specifically overexpressed.

Another approach might be to down-regulate the E. coli HDR when the foreign HDR to be tested is over-expressed. This might avoid the toxic effects, and allow the foreign HDR to be tested with less competition.

See above: E. coli HDR is expressed at low levels, and overexpressing this gene does not cause any toxicity. The toxicity is only linked to high DMAPP-producing HDR enzymes. The argument that EcHDR has a limited influence in our experimental setup is further supported by Figure 2C, showing that overexpression of EcHDR shifts the ratio slightly towards more IPP compared to the ‘no HDR overexpression’ control. This is in agreement with all previous studies showing a 1:5 DMAPP to IPP ratio for the E. coli enzyme.

The authors might additionally determine not only the ratio of DMADP to IDP, but also the absolute amounts of the two compounds to see if they are in the range expected in a cell, considering all of the genetic manipulations being carried out.

Without enhancing MEP pathway flux through engineering, we have found that metabolite measurements (isoprene and lycopene, as well as MEP pathway intermediates) are at or below the detection limit with mass spectrometry-based analytic methods.

Compared to plants, which can redirect up to 20% of recently fixed carbon into isoprenoids and produce many essential isoprenoid compounds, E. coli has minimal requirements for isoprenoids and naturally very low MEP pathway flux. Indeed, two decades of metabolic engineering of the MEP pathway have demonstrated its reluctance to sustain high flux and it remains challenging to produce high yields of isoprenoid compounds using this pathway in E. coli (DOI: 10.1093/femsle/fny079). The high demand for redox equivalents, which in plants is met by electrons coming from photosynthesis, may be one of the reasons for this. Hence, overexpressing dxs (as done in our study) is only the first step of several required to engineer high flux through the MEP pathway in E. coli. On this basis, and with our new observations that intracellular concentrations are in the low μM range, we do not think that the metabolite levels generated in this work are outside of physiologically relevant ranges.

I realise that all of these suggestions require additional work. Perhaps there is another way to validate such in vivo measurements as representing the real characteristics of the enzyme when tested in vitro. Maybe an attempt could be made to try expression, purification and in vitro assay of say two of the HDRs from opposite ends of the DMADP:IDP scale.

We feel that demonstrating the differences in HDR production ratios is most relevant in an in vivo setting with realistic reaction conditions and redox partners, where their effect on downstream isoprene production can be directly observed. Achieving suitable in vitro reaction conditions for oxygen-sensitive FeS-cluster HDRs, as well as finding a physiologically relevant redox donor is documented to be very challenging (DOI: 10.1021/ja903778d), and requires specialist facilities which we do not currently have access to. We acknowledge that data from in vitro assays of selected HDRs could be valuable in future studies, but for this study we feel that we have sufficiently improved our in vivo analysis and interpretation.

Reviewer #2:

[…]

Very interesting is the result shown in Figure S4 that the DMAPP/IPP ratios in vivo depend on the end product of the metabolism: higher at isoprene emission, lower at lycopene accumulation. I would suggest to show this result in the main text. I would also like to see a more detailed discussion of this observation.

This result might indeed be better presented in the main text, as it addresses the question raised by both reviewers of how robust and context-independent the observed differences in HDRs are. We have now included this data in Figure 2C, and discuss it in the main text (Results and Discussion paragraph four). We show that when using a different metabolic sink (isoprene synthase instead of lycopene), the relative differences between at least a subset of the tested HDRs, stay constant. Even though the absolute ratio of DMAPP:IPP shifts depending on the downstream sink, differences between the tested HDRs can be captured with both experimental setups.

A point that is also a bit neglected in the discussion is the role of IDI in the fine tuning of DMAPP/IPP ratios. As rightly noted, the DMAPP/IPP ratio measured in vivo does not correspond to the (not measured) in vitro ratio, but is changed by the activity of the IDI. For isoprene emitting oaks it was shown that the activity of the IDI correlates positively with ISPS and isoprene emission (Brüggemann and Schnitzler, 2002), which indicates a rate-limiting role of the IPI as mentioned.

To understand the fine tuning of the DMAPP/IPP ratio between HDR/IspH, IDI and the sinks towards the different soluble and volatile end products additional information on IDI is necessary: i.e. what is the in vitro (in vivo) ratio of the IDI used in the present experiments? I'm wondering why no experiments without overexpressed (and knocked out) IDI have been performed. How does the absence of this enzyme (absence of overexpression) influence the in vivo ratios of DMAPP/IPP in the different systems (lycopene vs isoprene)? I hope the authors can provide this information from experiments not yet shown.

We found that IDI is absolutely required when studying any of the high DMAPP-producing HDR enzymes in E. coli (Figure 2F, and Figure 2—figure supplement 1B, and data not shown). Any attempts to clone, express or analyze these high DMAPP HDRs in the absence of IDI have resulted in mutated, nonfunctional HDR genes. We agree that the influence of IDI on the DMAPP:IPP ratio is an important question that warrants further investigation, particularly in vivo in plants which likely use regulation of IDI expression as a further mechanism to control isoprenoid production.

In the present study, the important question is whether IDI co-expression artifactually influences our results. In our revised submission we have validated that IDI expression is constant across all strains (Figure 2—figure supplement 2A and 2D). Therefore, the different HDR enzymes must be the primary driver of the 40-fold differences in in vivo DMAPP:IPP ratios between different HDRs. While the IDI enzyme must affect DMAPP and IPP concentrations to some extent, it does not appear to have a major impact on our observations.

Furthermore, our proteomics data suggest that IDI expression is overall very low, based on both the total counts, number of peptides detected (1 peptide for EcIdi, 3 peptides for HpIdi, compared to 28-50 peptides detected for each HDR protein) and the normalized rank (EcIdi ranks at around 2300 out of 2500 quantified proteins in our samples). These data have to be seen in light of the inherent limitations of non-targeted proteomics, yet they strongly suggest low relative protein abundance for both IDIs present in our system.

[Editors’ note: what follows is the authors’ response to the second round of review.]

We find that paper has been substantially improved, but there are a few remaining concerns in the external reviewer's report.

Reviewer #1:

The authors have made extensive revisions and significantly improved their manuscript. In particular, the new proteomics measurements help to assure that the activity of the over-expressed HDR proteins is not affected by the native E. coli protein or the expression of the genes on the lycopene plasmid. Obtaining in vitro assay data on these proteins would clearly be the "gold standard" for comparison among them, but I can accept that this is just not yet technically attainable, and they have done what they can to increase the accuracy of their in vivo measurements. I also appreciated the chromatograms to verify the DMADP-IDP separation.

Two minor concerns that could be addressed:

1) Compared to the first submission, the DMADP:IDP ratio for Prunus persica has now decreased by 10 fold, while the ratios for other species have declined by more than 50%. Since these are the core data of the paper, such big changes do not inspire confidence in the analytical methods. Is there a reason for such variance from the previous version, such as a change in the analysis protocol or a change in calibrating factors?

The reviewer’s observations are correct, but the reason for the difference is that we have substantially improved the analytical method. It is important to note that the ranked order of HDR proteins has not changed in the new data set (ranked in terms of their DMAPP:IPP ratios), and the improvements to the analytical method give us confidence that the DMAPP and IPP quantification is more accurate.

Author response image 1

In our earlier dataset we reported DMAPP and IPP as a ratio of their LC-MS/MS peak areas. The accuracy of these ratios was limited by the fact that we had not managed to achieve full baseline separation of DMAPP and IPP. The to some extent overlapping and tailing peaks particularly affected peak area calculations in cases with very high DMAPP (first eluting analyte) or very low IPP (for example, the IPP concentration in P. persica samples was very low).

We subsequently made substantial improvements to the analytical method. We can now report a method that results in complete baseline separation of DMAPP and IPP peaks and an appropriate internal standard. With this improved method we can now report actual concentrations for DMAPP and IPP, and calculate DMAPP:IPP ratios with much greater accuracy.

Since this is the first publication of a method for LC-MS/MS separation and absolute quantification of DMAPP and IPP, we have taken great care to describe this technically challenging method in detail, including QC steps and potential pitfalls, to ensure reproducibility in other labs.

2) In subsection “Isoprene production” it states that "isoprene was quantified as described previously", but the reference given is incorrect. Please correct this.

This reference has been corrected now.

https://doi.org/10.7554/eLife.48685.sa2

Article and author information

Author details

  1. Mareike Bongers

    1. Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby, Denmark
    2. Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Brisbane, Australia
    Contribution
    Conceptualization, Data curation, Investigation, Visualization, Methodology, Project administration
    For correspondence
    marbon@biosustain.dtu.dk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4739-3852
  2. Jordi Perez-Gil

    1. Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Brisbane, Australia
    2. Centre for Research in Agricultural Genomics (CRAG) CSIC-IRTA-UAB-UB, Campus UAB Bellaterra, Barcelona, Spain
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5632-9556
  3. Mark P Hodson

    1. Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Brisbane, Australia
    2. Metabolomics Australia, Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Brisbane, Australia
    3. School of Pharmacy, The University of Queensland, Brisbane, Australia
    Contribution
    Data curation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5436-1886
  4. Lars Schrübbers

    Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby, Denmark
    Contribution
    Data curation, Methodology
    Competing interests
    No competing interests declared
  5. Tune Wulff

    Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby, Denmark
    Contribution
    Formal analysis, Performed and analysed the proteomics work
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8822-1048
  6. Morten OA Sommer

    Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby, Denmark
    Contribution
    Resources, Funding acquisition
    Competing interests
    No competing interests declared
  7. Lars K Nielsen

    1. Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby, Denmark
    2. Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Brisbane, Australia
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition
    Contributed equally with
    Claudia E Vickers
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8191-3511
  8. Claudia E Vickers

    1. Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Brisbane, Australia
    2. CSIRO Synthetic Biology Future Science Platform, Brisbane, Australia
    Contribution
    Conceptualization, Supervision, Funding acquisition, Investigation
    Contributed equally with
    Lars K Nielsen
    For correspondence
    c.vickers@uq.edu.au
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0792-050X

Funding

Australian Research Council (DP140103514)

  • Lars K Nielsen
  • Claudia E Vickers

Novo Nordisk Foundation (NNF10CC1016517)

  • Lars Schrübbers
  • Tune Wulff
  • Morten O A Sommer
  • Lars K Nielsen
  • Mareike Bongers

Marie Skłodowska-Curie Actions (FP7-PEOPLE-2013-IOF. Project: 623679)

  • Jordi Perez-Gil

Department of Education, Australian Government (National Collaborative Research Infrastructure Strategy (NCRIS))

  • Mark P Hodson
  • Lars K Nielsen

Queensland Government (Accelerate Fellowship)

  • Claudia E Vickers

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

Research at the University of Queensland was funded by an Australian Research Council grant to LKN and CEV. Research at the Center for Biosustainability was supported by The Novo Nordisk Foundation under NFF grant number NNF10CC1016517. JPG was supported by a Marie Curie International outgoing Fellowship within the 7th European Community Framework Programme. CEV was supported by Queensland Government Smart Futures and Accelerate Fellowships. Metabolomics Australia is part of the Bioplatforms Australia network, funded through the Australian Government’s National Collaborative Research Infrastructure Strategy (NCRIS). The authors would like to thank James Behrendorff for valuable feedback on this work.

Senior Editor

  1. Ian T Baldwin, Max Planck Institute for Chemical Ecology, Germany

Reviewing Editor

  1. Joerg Bohlmann, University of British Columbia, Canada

Publication history

  1. Received: May 22, 2019
  2. Accepted: February 16, 2020
  3. Version of Record published: March 12, 2020 (version 1)

Copyright

© 2020, Bongers et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 478
    Page views
  • 71
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics
    Moitrayee Bhattacharyya et al.
    Research Article Updated
    1. Biochemistry and Chemical Biology
    Byoungsan Choi et al.
    Research Article