Comparison of orthologous cyanobacterial aldehyde deformylating oxygenases in the production of volatile C3-C7 alkanes in engineered E. coli.

Aldehyde deformylating oxygenase (ADO) is a unique enzyme found exclusively in photosynthetic cyanobacteria, which natively converts acyl aldehyde precursors into hydrocarbon products embedded in cellular lipid bilayers. This capacity has opened doors for potential biotechnological applications aiming at biological production of diesel-range alkanes and alkenes, which are compatible with the nonrenewable petroleum-derived end-products in current use. The development of production platforms, however, has been limited by the relative inefficiency of ADO enzyme, promoting research towards finding new strategies and information to be used for rational design of enhanced pathways for hydrocarbon over-expression. In this work we present an optimized approach to study different ADO orthologs derived from different cyanobacterial species in an in vivo set-up in Escherichia coli. The system enabled comparison of alternative ADOs for the production efficiency of short-chain volatile C3-C7 alkanes, propane, pentane and heptane, and provided insight on the differences in substrate preference, catalytic efficiency and limitations associated with the enzymes. The work concentrated on five ADO orthologs which represent the most extensively studied cyanobacterial species in the field, and revealed distinct differences between the enzymes. In most cases the ADO from Nostoc punctiforme PCC 73102 performed the best in respect to yields and initial rates for the production of the volatile hydrocarbons. At the other extreme, the system harboring the ADO form Synechococcus sp. RS9917 produced very low amounts of the short-chain alkanes, primarily due to poor accumulation of the enzyme in E. coli. The ADOs from Synechocystis sp. PCC 6803 and Prochlorococcus marinus MIT9313, and the corresponding variant A134F displayed less divergence, although variation between chain-length preferences could be observed. The results confirmed the general trend of ADOs having decreasing catalytic efficiency towards precursors of decreasing chain-length, while expanding the knowledge on the species-specific traits, which may aid future pathway design and structure-based engineering of ADO for more efficient hydrocarbon production systems.

Introduction

As a unique trait in the bacterial kingdom, all photosynthetic cyanobacteria characterized to date harbor the capacity to produce medium chain-length hydrocarbons in their metabolism (Coates et al., 2014). Although the biological role of the hydrocarbon products is not entirely understood, the biosynthetic pathways and the key enzymes associated with the different catalytic steps have been elucidated. The most common cyanobacterial pathway is based on the enzyme aldehyde deformylating oxygenase (ADO) (Coates et al., 2014), which catalyzes the conversion of fatty aldehyde substrates, generated by fatty acyl ACP reductase (Aar), into Cn-1 hydrocarbons and formate [see review (Marsh and Waugh, 2013)]. Since the discovery of the biosynthetic route (Schirmer et al., 2010), there has been significant scientific interest towards ADO, due to the prospects in developing microbial over-production platforms for renewable replacements of petroleum-derived products. The endogenous pathways produce predominantly C15-C17 alkanes and alkenes, and the reported total levels are about 0,1% of dry biomass (Coates et al., 2014; Tan et al., 2011). While various strategies have been implemented to engineer the native pathways in cyanobacteria for higher yields, many heterologous pathway variants have been assembled in heterotrophic host Escherichia coli (E. coli) to produce alkanes of different lengths, from C9-C17 products (Schirmer et al., 2010; Harger et al., 2013; Akhtar et al., 2013; Howard et al., 2013; Choi and Lee, 2013) towards volatile compounds heptane (C7) (Akhtar et al., 2013; Kallio et al., 2014) and propane (C3) (Kallio et al., 2014; Menon et al., 2015; Zhang et al., 2016). Although the reported yields have remained marginal as compared to any industrially relevant process, the studies have provided detailed molecular-level insight into the essential limiting factors of the hydrocarbon pathways; stringent native regulatory systems, competing endogenous pathways, and unoptimal catalytic properties of the associated enzymes which compromise the output.

ADO has been extensively characterized in respect to catalytic mechanism, kinetic properties and structure. ADO is a soluble ferritin-like di-iron oxygenase which catalyzes the C1-C2 bond cleavage of aldehyde substrates into a corresponding Cn-1 alk(a/e)ne and formate via a multi-step free-radical mechanism (Paul et al., 2013; Rajakovich et al., 2015; Jia et al., 2015). ADO uses molecular oxygen as a cosubstrate, and the reaction results in the incorporation of one oxygen atom into formate while the other atom is reduced to H2O (Warui et al., 2011). The reaction is dependent on an external reducing system to provide altogether four electrons for the (i) activation of the di-iron center for oxygen binding, (ii) hemiacetal radical formation, and (iii) generation of the product complex (Paul et al., 2013; Rajakovich et al., 2015). Characteristically, ADO has relatively broad substrate specificity towards different aldehydes, with the highest activity towards the native linear C16 and C18 precursors (Khara et al., 2013; Bao et al., 2016; Zhang et al., 2016). This feature is accompanied by a low apparent catalytic turnover rate with reported kcat values below 0.02 s−1 in different contexts in vitro and in vivo (Khara et al., 2013; Andre et al., 2013; Bao et al., 2016). Kinetic characterization has also revealed that the affinity towards the aldehyde substrates is poor, especially for the shorter precursors, with KM values in the millimolar range (Bao et al., 2016). As the intracellular concentration of potentially toxic aldehyde intermediates must be maintained relatively low, ADO is expected to function below saturating substrate concentrations, and consequently, the apparent turnover rates are very low. This makes the ADO-catalyzed step one of the primary limiting factors forestalling the development of efficient hydrocarbon pathways, which cannot be effectively compensated by simple enzyme over-expression strategies alone. It has been postulated that the performance of ADO is associated with the labile nature of the di-iron center involved in the catalysis (Jia et al., 2015), suggesting that additional information on mechanistic structure-function studies could enable strategies of further enhancing the ADO activity.

Different approaches have been taken to overcome the limitations and to enhance the performance of the hydrocarbon pathways for better productivity and yield. Enzyme over-expression has been used to compensate for poor catalytic activity of ADO and to improve precursor availability in the native cyanobacterial hosts (Hu et al., 2013; Wang et al., 2013; Kaiser et al., 2013), while removal of endogenous pathways competing for the pathway intermediates has been shown to improve the flux towards the target hydrocarbon products (Kallio et al., 2014; Sheppard et al., 2016; Song et al., 2016). Optimization of the redox system to ensure sufficient flow of electrons to ADO has been accomplished by overexpressing ferredoxin (Fdx) and NADPH/ ferredoxin/flavodoxin-oxidoreductase (Fpr) (Kallio et al., 2014) or by engineering an ADO-Fdx fusion protein for higher activity (Wang et al., 2014). In addition, structure-based enzyme engineering of ADO has been utilized to alter the substrate preference towards specific aldehyde precursors (Khara et al., 2013; Hayashi et al., 2015; Zhang et al., 2016; Bao et al., 2016; Sheppard et al., 2016), thus enhancing the relative yield of the corresponding hydrocarbon products. Despite the advances, the efficiency of the enzymatic conversion remains to be insufficient to enable commercial applications, and alternative strategies are needed to identify and alleviate the key metabolic bottlenecks.

In order to gain further insight into the existing natural diversity of ADOs, the aim of the study was to characterize orthologous enzymes from different sources and to evaluate enzyme-specific differences in catalytic performance in heterologous biosynthetic context in vivo. The approach was to assemble a set of parallel biosynthetic hydrocarbon pathways in E. coli with a selection of over-expressed ADOs, a carboxylic acid reductase (CAR), and a reduction system based on cyanobacterial ferredoxin PetF (Fdx) (Fig. 1). To avoid primary complications with substrate availability, the strategy was to use externally supplemented fatty acid precursors of different lengths, with the focus specifically on volatile C3-C7 end-products. Such in vivo comparison has not been conducted earlier, and is expected to provide information for further structure-based evaluation, and identification of enzyme-specific features associated with substrate specificity or catalytic activity.

Fig. 1

Schematic representation of the hydrocarbon production system assembled in E.coli. The C4, C6 and C8 fatty acid substrates supplemented one at a time into the culture medium (left pane) are taken up by the E. coli cells. Inside the cells the fatty acid precursors are converted into aldehydes by heterologously expressed carboxylic acid reductase (CAR), assisted by the maturase phosphopantetheinyl transferase (ppt), and further into Cn-1 alkanes by the aldehyde deformylating oxygenase (ADO) (middle pane). The reduction of ADO is accomplished by the combined action of ferredoxin (Fdx) and NADPH: Ferredoxin/Flavodoxin-oxidoreductase (Fpr). The endogenous aldehyde reductases ahr and yqhD, which compete with ADO for the aldehyde intermediates, have been deleted from the production strain. The resulting C3, C5 and C7 hydrocarbon products diffuse out of the cells and into the headspace of the closed reaction vials (right pane), and are quantitated from the gas phase by GC-MS against commercial standards for propane, pentane and heptane, respectively.

Materials and methods

Reagents and enzymes

Standard molecular biology procedures and commercial kits (Qiagen) were used for DNA plasmid isolation and manipulation. All the used enzymes were purchased from New England BioLabs or from Thermo-Scientific Fermentas. All chemicals used in this study were purchased from Sigma-Aldrich if not mentioned otherwise.

Cell strains and growth conditions

Escherichia coli strain DH5α was used as the host in expression plasmid assembly. The E. coli strain BL21(DE3)∆yjgB ΔyqhD (Kallio et al., 2014) which lacks the two endogenous aldehyde reductases yqhD and ahr competing with ADO for the aldehyde precursors, was used for the quantitative production experiments. The E. coli cells were routinely cultivated in Luria-Bertani (LB) medium at 37 °C in a shaker at 150–200 rpm or on the solid LB plates containing 1.5% (w/v) agar. The details of the production cultures are described below.

Expression plasmid construction for heterologous production of alkanes in E. coli

All plasmids, genes and organisms used in this study are listed in Table 1.

Table 1

The host organism, plasmids and genes used in this study.

Host								Reference
E. coli BL21 (DE3)ΔyqhDΔahr								(Kallio et al., 2014)
Plasmid		Genes to be expressed					Reference
pET-FPC		fdx, ppt, car					This study
pACYC-fpr		fpr					This study
pCDF-ado (N. punctiforme)		ado: Nostoc punctiforme PCC 73102					This study
pCDF-ado (P. marinus)		ado: Prochlorococcus marinus MIT9313					This study
pCDF-adoA134F(P. marinus)		ado: Prochlorococcus marinus MIT9313					This study
pCDF-ado (Synechococcus)		ado: Synechococcus sp. RS9917					This study
pCDF-ado (Synechocystis)		ado: Synechocystis sp. PCC 6803					This study
Genes	Function			Origin			GenBank ID
fdx	Ferredoxin I			Synechocystis sp. PCC 6803				AAB72025.1
ppt	Phosphopantetheinyl transferase			Bacillus subtilis				X65610.1
car	Carboxylic acid reductase			Mycobacterium marinum				ACC40567.1
fpr	NADPH: ferredoxin/flavodoxin-oxidoreductase			Escherichia coli K−12				AAB03056.1
yqhD	Aldehyde reductase			Escherichia coli BL21(DE3)				ACT44688.1
ahr	Aldehyde reductase			Escherichia coli BL21(DE3)				AAA97166.1

The genes coding for ADO from Synechocystis sp. PCC 6803 (GenBank ID: WP_010873444), Nostoc punctiforme PCC 73102 (GenBank ID: WP_012408400) and Synechococcus sp. RS9917 (GenBank ID: ZP_01080370) were ordered as synthetic fragments codon-optimized for E. coli (GenScript) in pUC57 with the His-tag coding regions as shown in Table 2, flanked by the restriction sites NcoI and AvrII. The NcoI-AvrII fragments were subcloned from the pUC57 into the commercial plasmid pCDF-Duet (Novagen) to obtain the corresponding final constructs pCDF-ado (N. punctiforme), pCDF-ado (Synechococcus) and pCDF-ado (Synechocystis) (Table 1). The codon-optimized genes encoding for the ADO from Prochlorococcus marinus (GenBank ID: WP_011130600) and the corresponding variant A134F were obtained from pET-28b(+) -based expression plasmid (Das et al., 2011) and a modified derivative (Khara et al., 2013), respectively. The genes were subcloned (NcoI-BamHI) from pET-28b(+) into the pCDF-Duet to obtain the final constructs pCDF-ado (P. marinus) and pCDF-adoA134F (P. marinus) (Table 1) with the His-tag coding regions as shown in Table 2.

Table 2

The sequence around the 5′ region of the ado expression constructs (Table 1) showing the RBS (yellow highlight), start codon (red), His-tag coding region (underline) and the beginning of the ado sequence (grey).

In addition to the alternative pCDF-ado constructs, the engineered system constituted of two other plasmids which allowed the simultaneous expression of the remaining four enzymes required in the alkane pathway (see Fig. 1). Ferredoxin (from Synechocystis), phosphopantetheinyl transferase (from Bacillus subtilis) and carboxylic acid reductase (from Mycobacterium marinum) were expressed from pET-FPC (Table 1). The construct pET-FPC was prepared by subcloning the fdx fragment (NcoI-HindIII) from pCDF-PPC (Kallio et al., 2014) to replace tesA in pET-TPC (Akhtar et al., 2013). NADPH: ferredoxin/flavodoxin-oxidoreductase (from E. coli K-12) was expressed from pACYC-fpr (Table 1). The construct pACYC-fpr was prepared by removing petF from existing pACYC-petF-fpr (Kallio et al., 2014) by preparative digestion (HindIII-NcoI), followed by blunting with Mung bean nuclease and re-ligation.

Evaluation of performance of the different ADO orthologs in hydrocarbon biosynthesis in vivo

Cultivation of the production strains (Table 1) was based on a method optimized earlier (Kallio et al., 2014); 5 ml overnight pre-cultures were grown in Luria-Bertani (LB) medium at +37 °C, and used to inoculate 50 ml main-cultures in Terrific Broth (TB) medium at +37 °C. The cultures were carried out in the presence of 100 µg ml−1 ampicillin, 50 µg ml−1 spectinomycin (omitted from the main culture), 20 µg ml−1 streptomycin and 34 µg ml−1 chloramphenicol to maintain selection pressure for the three expression plasmids. Enzyme over-expression was induced by the addition of β-d−1-thiogalactopyranoside (IPTG; 1 mM) at OD600 0.5, and cultivation was continued at +30 °C for 4 h. The cells were then pelleted (+4 °C, ~4700 g, 10 min) and re-suspended into 20 ml of 100 mM potassium phosphate buffer (pH 7.5) containing 25 g l−1 glucose and 1 mM IPTG. In order to provide substrates for hydrocarbon biosynthesis, the cultures were supplemented with 5 mM butanoic acid, 5 mM hexanoic acid or 5 mM octanoic acid for the production of propane, pentane or heptane, respectively. Samples of 0.5 ml from the re-suspended cells were transferred to 2 ml gas-tight GC vials, sealed and incubated in a rotating shaker at RT for 30 min, 1.5 h, 2.5 h, 4 h and 10 h. Three parallel samples were prepared for GC-MS analysis in each timepoint and timepoints 30 min, 1.5 h, 2.5 h and 4 h were analyzed in a separate experiment from the 10 h timepoint.

GC-MS analysis of the volatile alkanes

Alkanes were analyzed from the headspace of sealed reaction vials using a 7890 A GC (Agilent) and 5975 C inert MS (Agilent) system with a Supelco Equity-1 (300 °C, 30 m x 0.32 mm x 1 µm) (propane) or Agilent HP-INNOWax (260 °C, 30 m x 0.32 mm x 0.5 µm) (heptane and pentane) column as described previously (Kallio et al., 2014).

Propane, pentane and heptane eluted at ~1.2 min, ~1.2 min and ~1.4 min, respectively, and were quantified by relating the GC-MS peak areas (m/z=43, 43 and 57, respectively) to corresponding commercial standards. The propane standard was a 1% mixture (v/v) in N2 gas (AGA, Finland). The concentration series for pentane (Figure SI2) and heptane (Figure SI3) were obtained by diluting liquid standards in acetone and filled to 0.5 ml with TB medium in sealed 2 ml reaction vials, followed by 30 min incubation at +30 °C before headspace GC-MS analysis. Four replicates were analyzed in all cases, and the values were used for calculating the averages and standard deviations.

Evaluation of alcohol production of the ADO-expressing strains

For quantitating butanol, hexanol and octanol produced by the parallel strains, the reaction cultures were prepared as described in Section 2.4 for alkanes. At 10 h reaction timepoint the cell suspensions were mixed with acetone (1:1 for C4 and 1:19 for C6/C8) and the soluble phase was subjected to GC-MS analysis. The GC-MS set-up was the same as for heptane and pentane, operated in the split-less mode (1.4 ml min−1 He flow, injector at 250 °C, 2 min at 40 °C then to 250 °C with 20 °C min−1, hold at 250 °C for 2 min). The amount of butanol (tR ~5.4 min), hexanol (tR ~7.0 min) and octanol (tR ~8.5 min) was calculated based on the GC-MS peak areas (m/z=56) of the corresponding commercial standards. The standard concentration series (0–150 mg/L were prepared in water from 1 g/L stock solutions (1:10 acetone: water)), incubated at RT for 2 h under rotation, and analyzed to obtain the standard curves for butanol (Figure SI4), hexanol (Figure SI5) and octanol (Figure SI6).

Analysis of the ADO expression levels in different strains

For evaluating ADO levels in the cells, the expression strains were cultivated as described for the alkane production experiments until the re-suspension of the cells into 100 mM potassium phosphate buffer (pH 7.5), supplemented with 25 g l−1 glucose, 1 mM IPTG and 50 µg ml−1 kanamycin. The cell suspensions were then adjusted to the same cell density (OD600 nm), and 6 ml aliquots were incubated in 15 ml screw capped tubes at RT in a rotational mixer for 0 h, 4 h or 10 h. After incubation, the cells were pelleted (+4 °C, ~4700 g, 10 min) and stored at –20 °C until analysis.

The His-tagged ADO enzymes were extracted by Ni2+ affinity chromatography using Ni-NTA Spin Columns (Qiagen) according to manufacturer's protocol for protein purification under native conditions from E. coli lysates. The bound ADO enzymes were eluted from column with 3×200 µl elution buffer (50 mM NaH2PO4; 300 mM NaCl; 500 mM imidazole; pH 8.0) and glycerol was added to final concentration of 40% before storage at –20 °C. The ADO expression levels and stability were evaluated using the Bradford dye-binding method (Bradford, 1976) and commercial sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (12% Mini-PROTEAN® TGX™, Bio-Rad).

The protein concentrations estimated by Bradford were normalized to the sample OD600 nm values in order to compare samples containing different amount of cells. The SDS-PAGE analysis was used to confirm the purity of the protein samples and to verify the ADO expression levels calculated based on the Bradford analysis.

Computational methods used for amino acid sequence alignment, enzyme structure analysis and mRNA secondary structure prediction

The amino acid sequences of the ADO orthologues were compared by using the Basic Local Alignment Search Tool BLAST web server (http://blast.ncbi.nlm.nih.gov/Blast.cgi) (Altschul et al., 1990). Sequence alignment was prepared using Clustal Omega multiple sequence alignment tool from the EMBL-EBI web pages (http://www.ebi.ac.uk/Tools/msa/clustalo). The secondary elements were included into sequence alignment using ESPript web server (http://espript.ibcp.fr/ESPript/ESPript) (Robert and Gouet, 2014). The structural analysis of the S. elongates PCC 7942 ADO crystal structure with bound C16 fatty alcohol (PDB code 4RC5) was done with PyMOL 1.7.4.5 (Schrödinger, LCC). The secondary structures of ADO mRNA transcripts (either the first 170 bp from the transcription start site of the T7 promoter or the whole transcript from the T7 promoter to the T7 terminator) were predicted with the Predict a Secondary Structure Web Server (http://rna.urmc.rochester.edu/RNAstructureWeb) using the default settings of the server (Reuter and Mathews, 2010).

Results

Sequence comparison of ADO orthologues

Native orthologous ADO enzymes from Nostoc punctiforme PCC 73102 (N. punctiforme), Synechocystis sp. PCC 6803 (Synechocystis), Synechococcus sp. RS9917 (Synechococcus) and Prochlorococcus marinus MIT9313 (P. marinus) and the P. marinus variant A134F were selected as the targets for the in vivo comparison. Alignment of the native amino acid sequences revealed 48% −68% sequence identities between the ADOs, while the overall structure, active site topology, and catalytic mechanism were expected to be rather conserved between the enzymes. The highest similarity was observed between the ADOs from N. punctiforme and Synechocystis (amino acid sequence identity 75%), whereas the ADO from Synechococcus appeared to be more diverged (amino acid sequence identity 40–41%), and was included in the comparison specifically as an outlier. The 17 amino acid residues within 4 Å distance from the bound C16 fatty alcohol ligand in the crystal structure of S. elongatus PCC 7942 (Jia et al., 2015), were all conserved between the enzymes from P. marinus, N. punctiforme and Synechocystis (Fig. 2). The ADO from Synechococcus, in comparison, carried two conservative substitutions (I17V and I104V; numbering according to the amino acids visible in the A chain of S. elongatus PCC 7942 crystal structure with PDB code 4RC5) within the region (Fig. 2). Further comparison of the sequence alignment (Fig. 2) against the crystal structure of S. elongatus PCC 7942 ADO revealed that the remaining substitutions were mainly located on the outer surface of the Synechococcus enzyme (Figure SI1).

Fig. 2

Amino acid sequence alignment of the five different ADOs compared in the study. The secondary structures and corresponding amino acid numbering shown above the alignment represent the A chain of S. elongatus PCC 7942 ADO, as derived from the crystal structure with bound C16 fatty alcohol (PDB code 4RC5) (Jia et al., 2015). The green arrows under the sequence alignment indicate the amino acid residues within 4 Å distance from the fatty alcohol ligand in the 4RC5 complex structure, and the black stars denote the residues which are different in Synechococcus ADO in comparison to all the other studied ADO enzymes. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Assembly of the heterologous hydrocarbon pathways in E. coli

The genes coding for the selected alternative ADOs were assembled in a pCDF-based E. coli expression vector (Table 1), in order to systematically compare the enzymes in a heterologous hydrocarbon production system in vivo. The constructs were identical apart from the selected ado coding regions, and each gene carried a sequence encoding an N-terminal hexa-Histidine affinity tag to allow isolation (Table 2). The rest of the heterologous hydrocarbon pathway assembled in E. coli BL21(DE3)ΔyqhDΔahr was constructed on two additional plasmids with ferredoxin (Fdx), phosphopantetheinyl transferase (ppt) and carboxylic acid reductase (CAR) over-expressed from pET and NADPH: Ferredoxin/Flavodoxin-oxidoreductase (Fpr) from pACYC (Table 1). Altogether, the pathway enabled the in vivo conversion of externally supplied C4, C6 and C8 fatty acid precursors into the corresponding Cn-1 end-products, propane, pentane and heptane, respectively, followed by quantitation by GC-MS from the headspace of the reaction vials (Kallio et al., 2014).

Production of heptane from supplied octanoic acid

GC-MS analysis confirmed that all the five parallel E. coli strains carrying the heterologous hydrocarbon pathways were able to produce C7 heptane when supplemented with 5 mM octanoic acid in the culture medium. However, there were clear differences in the total production levels as well as the production rates between the systems expressing the different ADO orthologs (Figs. 3A, B, 4). The highest amount of heptane, as measured at the 4 h timepoint, was produced by ADO from N. punctiforme (134 µmolL–1OD600 nm–1), while the yield sequentially decreased in the order P. marinus A134F (116 µmolL–1OD600 nm–1), P. marinus (112 µmolL–1OD600 nm–1) to about 60% of the maximum in Synechocystis (81 µmolL–1OD600 nm–1) (Fig. 3A). By far the lowest amount of heptane was produced by ADO from Synechococcus (11 µmolL–1OD600 nm–1), which at 4 h was under 10% of the observed maximum (Fig. 3A). The same pattern in the relative ADO activities was obtained in a separate quantitation at the 10 h timepoint (Fig. 3B), although the amount of heptane continued to increase in the case of P. marinus and N. punctiforme enzymes over the last 6 h.

Fig. 3

Alkane production yields of different ADO orthologues expressed in E. coli. The time-course analysis between 0 h and 4 h for the production of (A) heptane, (C) pentane and (E) propane, and bars representing the product accumulation at 10 h timepoint for (B) heptane, (D) pentane and (F) propane. The ADOs used in the comparison are from P. marinus (black ■), a corresponding variant A134F (red ), Synechocystis (blue ), N. punctiforme (magenta) and Synechococcus (green ). Each datapoint represents mean ±SD calculated based on three parallel reactions. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

The initial production rates of heptane (dark grey), pentane (light grey) and propane (white) by the different ADO orthologues. The production rates are depicted from the amount of alkane formed within the first 30 min after sealing the reaction vials. Each column represents mean ±SD calculated from the initial rates of three parallel reactions.

The rates of heptane production, as depicted from the amount of product accumulated in the vial headspace within the first half an hour of the reaction, also varied between the alternative ADO orthologs (Figs. 3A, 4). The relative rates followed the general pattern of the total productivity, with the ADO from N. punctiforme being the most efficient (135 µmolL–1OD600 nm–1h–1), followed by the P. marinus variant A134F (77 µmolL–1OD600 nm–1h–1) and the WT enzymes from P. marinus (67 µmolL–1OD600 nm–1h–1), Synechocystis (55 µmolL–1OD600 nm–1h–1) and Synechococcus (2.2 µmolL–1OD600 nm–1h–1).

Production of pentane from supplied hexanoic acid

All the engineered E. coli strains expressing the parallel ADO orthologs produced C5 pentane when supplied with hexanoic acid in the culture medium. Although clear differences were observed in the performance of the alternative ADOs (Figs. 3C, D), the total pentane productivity and the relative differences between the enzymes were smaller than for heptane. While the ADO from Synechocystis gave the highest pentane yield (105 µmolL–1OD600 nm–1), the strains expressing the orthologs from P. marinus (WT and A134F) and N. punctiforme all produced about 80 µmolL–1OD600 nm–1 as measured at the 4 h timepoint. The lowest amount of pentane, about 30% of the maximum, was produced by ADO from Synechococcus (29 µmolL–1OD600 nm–1). The relative ADO activities determined based on the time-course analysis (Fig. 3C) followed the pattern obtained in a separate quantitation at the 10 h timepoint (Fig. 3D). Again, for the two most efficient enzymes (from Synechocystis and N. punctiforme), the product accumulation continued over the last 6 h.

The initial rates of pentane production did not in all cases directly correlate with the final yields; The fastest product accumulation within the first 30 min was measured for the ADO from N. punctiforme (77 µmolL–1OD600 nm–1h–1), whereas the ortholog from Synechocystis with higher overall production appeared to perform less efficiently (48 µmolL–1OD600 nm–1h–1) (Fig. 4). The initial pentane production rate for the P. marinus WT ADO and A134F variant was about equal (~39 µmolL–1OD600 nm–1h–1), and lowest for the ADO from Synechococcus (12 µmolL–1OD600 nm–1h–1).

Production of propane from supplied butanoic acid

When supplied with 5 mM butanoic acid in the culture medium, all the five engineered E. coli strains produced propane in the reaction vial headspace (Figs. 3E, F). In comparison to pentane and heptane, the yields of propane were lower (total production below 50% in comparison to the C5 and C7 products), and the differences in the performance of the alternative orthologs less significant (Figs. 3E, F). Apart from the ADO from Synechococcus with the lowest propane levels (7.8 µmolL−1OD600 nm–1), the other ADO orthologs produced between 30 and 40 µmolL–1OD600 nm–1 by the 4 h reaction timepoint. Unlike for pentane and heptane, the propane accumulation did not continue after this for any of the strains (Fig. 3E), and the levels remained the same when analyzed at the 10 h timepoint (Fig. 3F).

As for the initial propane production rate, the native ADOs from N. punctiforme, Synechocystis, and P. marinus performed at near equal efficiency, producing about 20−23 µmolL–1OD600 nm–1h–1 (Fig. 4). Interestingly, the initial propane productivity of the P. marinus ADO variant A134F (12.0 µmolL–1OD600 nm–1h–1) evaluated based on the first 30 min was only about half of the WT enzyme in the tested in vivo system. Again, the ADO from Synechococcus had the lowest production efficiency (7.2 µmolL–1OD600 nm–1h–1).

Expression levels and stability of the ADO enzymes

In order to correlate the hydrocarbon production efficiencies (Fig. 3, Fig. 4) to the relative amounts of the ADO orthologs expressed, the intracellular ADO levels and stability were evaluated for each of the engineered E. coli strains. The comparison was carried out by extracting the His-tagged ADO enzymes from the strains at specified timepoints after protein expression, followed by quantitative analysis by Bradford (Fig. 5A) and SDS-PAGE (Fig. 5B). The first timepoint (0 h) represented the situation at the beginning of the reaction, whereas the subsequent timepoints (4 h and 10 h), after addition of kanamycin to inhibit further protein synthesis in the cells, reflected the overall enzyme stability in vivo. The Bradford analysis (Fig. 5A) revealed that the expression levels of four of the ADOs (Synechocystis, N. punctiforme, P. marinus, P. marinus A134F) were within the same range (0.12–0.20 mgOD600 nm−1). In contrast, the amount of the ADO from Synechococcus was significantly lower, 11–20% in comparison to the other enzymes. These results were further supported by the SDS-PAGE analysis (Fig. 5B), which confirmed that (i) the purity of the enzyme preparations was high, and that the (ii) ADO levels were rather uniform between the different samples, with the exception of the ADO form Synechococcus which was barely visible on the gel. Altogether, the in vivo ADO levels remained rather constant throughout the extended incubation at RT, suggesting that enzyme stability in the heterologous host did not have any significant impact on result interpretation.

Fig. 5

The expression levels and in vivo stability of the ADO enzymes in E. coli. (A) The relative amount of His-tag purified ADOs from E. coli cells incubated at RT for 0 h (dark blue), 4 h (mid blue) or 10 h (light blue) after the expression phase, quantitated with the Bradford method. (B) SDS-PAGE gel visualizing the ADO purification fractions at different timepoints. Each column represents mean ±SD calculated based on three parallel Bradford reactions. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

ADO mRNA secondary structure analysis in silico

In order to predict and compare mRNA secondary structure formation between the ADO transcripts in silico, the homologous sequences were subjected to computational analysis of free energies of potential secondary structures. The prediction performed on the full length ADO mRNAs as well as on the first 170 bases of the transcripts revealed that the lowest free energy of the secondary structures were comparable in all cases (Table SI1), and thus failed to identify divergent structural features which could explain the low expression levels of Synechococcus ADO.

Alcohols produced by endogenous aldehyde reductases

Despite the deletion of two of the endogenous aldehyde reductases in the host genome (∆yjgB ΔyqhD), E. coli encodes for another 11 homologous enzymes which potentially compete with ADO for the aldehyde substrates (Rodriguez and Atsumi, 2014), and thus may interfere with the hydrocarbon production under evaluation. To study the possible effect of the remaining aldehyde reductases on the outcome, the five generated E. coli strains were compared for the production of octanol, hexanol and butanol in response to the supplementation of the three alternative fatty acid substrates. As the primary observation, all the strains were shown to produce alcohols from the respective fatty acids, and despite of specific differences, the overall levels were within the same order of magnitude as measured for the corresponding Cn-1 alkanes (Fig. 6). Comparison of the parallel biological replicates, as well as independently repeated trials (Figure SI7), showed that the variation between experiments was marginal, suggesting that set-up was applicable for the quantitative analysis of the alcohols. In all cases, only a fraction of the supplemented fatty acid was converted into either of the two products (~7–24%) (Table SI2), indicating that majority of the substrate was always left unreacted with the aldehyde reductases or the ADOs. It was also clear that the fatty acids were not consumed for cell growth, as increase of the OD values was not typically observed during the reaction phase (Table SI3).

Fig. 6

Alcohol production of the E. coli strains under study. The levels of (A) octanol, (B) hexanol and (C) butanol (red bars) were measured with GC-MS after 10 h incubation in response to the supplementation of octanoic acid, hexanoic acid and butanoic acid, respectively. Each bar represents mean ±SD calculated based on three parallel reactions. The corresponding production levels of the respective Cn-1 alkanes (black outlines; adapted from Fig. 3) are shown to ease direct comparison between the two products. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

As seen in Fig. 6A-C, the production of octanol, hexanol and butanol differed clearly from one another in the total levels as well as in the strain-specific patterns. The outcome did not correlate with the substrate chain-length or with the amount of the alkanes produced. Octanol (Fig. 6A) was produced at the lowest titers with very subtle differences between the strains, while the hexanol (Fig. 6B) and butanol (Fig. 6C) profiles showed considerably more variation. This indicated dependence between the production of alkanes and alcohols for the C6/C4 substrates, with maximum alcohol accumulation in the strains expressing the Synechococcus ADO.

Discussion

Approach for comparing orthologous ADOs in the production of volatile alkanes

Because of the potential biotechnological applications for generating renewable replacements for petroleum-derived fuels, the cyanobacterial enzyme aldehyde deformylating oxygenase, ADO, has been extensively characterized and evaluated as part of various heterologous hydrocarbon production pathways. In this study we conducted a systematic comparison of five alternative ADO orthologs for the production of three short-chain alkanes, heptane, pentane and propane. Short-chain products have been of particular interest because of their volatility under standard atmospheric conditions (Khara et al., 2013; Kallio et al., 2014; Menon et al., 2015; Zhang et al., 2016; Sheppard et al., 2016), which could be exploited for the extraction of the compounds directly from the biotechnological process due to spontaneous separation into the culture head-space. At the same time, while there are general kinetic constraints in the catalytic performance of ADO limiting the development of efficient production systems (Bao et al., 2016; Andre et al., 2013), the capacity of the enzymes to utilize short-chain precursors remains to be especially poor. The ADO orthologs in the current work were selected based on their appearance in literature as representatives of the four major classes of cyanobacteria, as well as the overall limited information on the relative capacity to produce short chain end-products. Despite the numerous studies related to the ADOs from N. punctiforme (Schirmer et al., 2010; Warui et al., 2011; Li et al., 2012; Howard et al., 2013; Pandelia et al., 2013; Paul et al., 2013; Hayashi et al., 2015; Rajakovich et al., 2015), Synechocystis (Hu et al., 2013; Wang et al., 2013), Synechococcus (Schirmer et al., 2010; Harger et al., 2013), P. marinus (Schirmer et al., 2010; Krebs et al., 2011; Li et al., 2012; Andre et al., 2013; Khara et al., 2013; Aukema et al., 2013; Buer et al., 2014; Kallio et al., 2014; Zhang et al., 2016) and the P. marinus variant A134F (Khara et al., 2013; Menon et al., 2015; Zhang et al., 2016), the enzymes have not been directly compared in a biosynthetic context for the production of volatile hydrocarbons. For this purpose, a novel experimental set-up was constructed based on the earlier analytical system in E. coli (Kallio et al., 2014), which enabled the conversion of supplied C8, C6 and C4 fatty acid precursors into the corresponding Cn-1 alkanes in vivo (Fig. 1; Table 1). The strategy was based on the ability of E. coli to scavenge fatty acids from the medium into the cell, after which the precursor is reduced into aldehyde by the heterologous enzyme CAR, and further into the final alkane product by ADO. To ensure that the conversion would be primarily limited by ADO, as required for the comparison between the orthologs, additional enzyme components (ppt, Fdx and Fpr) were introduced into the pathway. Ppt is necessary for the optimal function of CAR (Akhtar et al., 2013), while Fdx and Fpr (Kallio et al., 2014), which have not been previously co-expressed in corresponding substrate feeding biotransformation trials, constitute the in vivo redox relay for ADO (Fig. 1). For optimizing the performance further, the host strain used for the work was E. coli BL21(DE3)∆yjgB ΔyqhD (Kallio et al., 2014) from which two genes coding for competing aldehyde reductase enzymes had been deleted. After expression of the pathway enzymes and substrate supplementation, the alkane products were quantitated from the gas-phase of the reaction vials by GC-MS, and statistically compared between the alternative strains.

Validity of the in vivo strategy for evaluating the performance of alternative ADOs

In order to critically assess the constructed in vivo system and to monitor the consistency of the obtained data, all the reactions were carried out in three parallel biological replicates and analyzed in a continuous sequence at successive timepoints. The results showed that (i) the set up enabled the detection of ADO-specific catalytic differences between the orthologs, and that (ii) the variation between replicate samples was marginal in respect to the total signal levels, thus allowing reliable numerical analysis. In addition, the data from the time-course analysis (Figs. 3A, C, E) was further verified in an independent reaction series comparing all the ADOs in parallel in a single experiment (Figs. 3B, D, F), which confirmed that the overall activity pattern was reproducible and that the individual assays could be directly compared between the separate experiments. Altogether, the experimental system appeared to be applicable for comparing the performance of the alternative ADO enzymes in a heterologous biosynthetic context in E. coli, and allowed repeatable and statistically-relevant conclusions to be drawn based on the data.

Differences between ADO expression levels and stability

To evaluate whether the observed differences in the recorded alkane production efficiencies were affected by differences in ADO expression, the amount and stability of the enzymes were studied under the conditions used for the activity assays in the heterologous host. Although there was some variation between the parallel replicates, Bradford assay (Fig. 5A) and SDS-PAGE analysis (Fig. 5B) revealed that there was no significant deviation between the ADOs from Synechocystis, N. punctiforme, P. marinus or P. marinus A134F which would affect the comparison. The only ortholog present in clearly lower amount was ADO from Synechococcus, which also shared the lowest sequence conservation amongst the enzymes. As the promoter and the ribosome binding site (RBS) region of all the constructs were identical, and the computational free energy prediction of mRNA secondary structures failed to identify obvious differences between the transcripts (Table SI1), the low abundance could reflect constraints at the enzyme-level. Majority of the conserved amino acid substitutions found specifically in the Synechococcus ADO are located on the outer surface of the enzyme (Fig. 2, SI1), and these could potentially contribute to unoptimal folding causing premature degradation. Such conformational defects could also affect the integrity of the binding site pocket and interfere with the substrate-enzyme interactions, which would explain the observed reduced capacity of the Synechococcus ADO to utilize the C8 substrate (Fig. 3A).

Production of alcohols by the endogenous aldehyde reductases

The production of alkanes was shown to be accompanied by the accumulation of the corresponding alcohols, as expected based on the wide range of endogenous aldehyde reductases still present in the E. coli host. Generally the recorded alcohol levels were somewhat higher in comparison to the alkanes, but remained within a similar overall molar range (Fig. 6). At the same time, majority of the supplemented fatty acid substrates were not converted into either of the products (Table SI2) nor used for cell growth (Table SI3), implying that access to the substrates would not be the primary limiting factor in the systems. While the two alternative pathways appeared to have a rather insignificant effect on another in the case of the C8 precursor, as suggested by the constant octanol production levels of the parallel strains (Fig. 6A), the observed variation in the hexanol (Fig. 6B) and butanol (Fig. 6C) profiles indicate competition between the aldehyde reductases and ADO. Considering the relative output levels and the amount of available substrates, however, the concomitant formation of alcohols was not expected to alter the overall alkane production patterns to a degree which would significantly alter data interpretation or the conclusions of the study.

Substrate-specific characteristics common to all the pathways

All five ADO orthologs were shown to be active, and produced heptane, pentane and propane in the headspace of the reaction vials from the supplied C8, C6 and C4 fatty acid substrates, respectively. There were certain general substrate-specific patterns, however, which were common to the ADO enzymes. As a clear trend, the overall productivities decreased in correlation with the substrate chain length (Fig. 3). At the same time, the initial reaction rates were slower for the shorter substrates (Fig. 4), and the plateau was reached much earlier for butanoic acid in comparison to the longer fatty acids (Fig. 3). These observations are in agreement with the reported kinetic constraints of the ADO enzymes towards shorter precursors (Khara et al., 2013) (Kallio et al., 2014) (Bao et al., 2016), and reflect lower affinity (KM) and overall catalytic efficiency (apparent kcat/KM) towards substrates outside the native chain-length range. While the comparison demonstrates that ADO substrate-specific properties can be quantitatively evaluated in defined biosynthetic contexts in vivo for volatile products, it must be noted that the overall kinetics are also affected by rate of substrate diffusion into the cell, the efficiency of the previous enzymatic step by CAR (Fig. 1) and the affinity of the remaining endogenous aldehyde dehydrogenases for the precursors, which all need to be accounted for.

ADO-specific characteristics in volatile hydrocarbon production

The comparative analysis revealed clear enzyme-specific differences between the alternative ADOs both in the total productivity (Fig. 3) and the initial reaction rates (Fig. 4) for all of the evaluated substrates. Judged based on the combined data, the ADO from N. punctiforme appeared to perform better than the other orthologs. The production yields of heptane, for example, were several folds higher in comparison to the enzymes from P. marinus or Synechocystis (Fig. 3B). Also the initial production rates of the N. punctiforme enzyme were highest in most cases, especially for heptane and pentane (Fig. 4). This is in line with the reported superior activity of the N. punctiforme ADO in respect to the native products pentadecane and heptadecene (Schirmer et al., 2010), suggesting that the observed effect is not chain-length specific but related to other catalytic properties than mere substrate preference. Interestingly, the P. marinus ADO variant A134F which has previously been shown to have higher preference for shorter precursors (Khara et al., 2013; Sheppard et al., 2016) performed slightly better for the C8 substrate, while the activity towards C6 and C4 aldehydes remained at the WT levels. This may be due to variations in the local substrate concentrations between alternative experimental set-ups, which could result in differences in the observed KM -dependent catalytic efficiencies. By far the lowest activities were recorded for the system harboring ADO from Synechococcus (Fig. 3, Fig. 4). Taking into account the very low levels of the enzyme in the cells (Fig. 5), however, the specific activity and the overall kinetic parameters may not differ so much from those of the other orthologs. The inferior performance of the Synechococcus ADO has been observed already earlier (Schirmer et al., 2010), which confirms that this is not specific to the current experimental set-up, but rather an intrinsic property of the enzyme. The only clear deviations in the catalytic patterns of the ADOs were the unexpectedly high activity of Synechocystis enzyme for pentane (Figs. 3C, D), and the low activity of the Synechococcus enzyme for heptane (Figs. 3A, B). These trends were repeatable and apparent in the separate independent reactions series, and expected to reflect enzyme-specific traits on substrate preference, which cannot be predicted based on the currently existing structural and functional data.

Conclusions

In this study, five different cyanobacterial ADO enzymes were evaluated in vivo in a novel biosynthetic context in E. coli for the production of short-chain volatile alkanes, heptane, pentane and propane. The comparison, based on supplemented fatty acid precursors, provided new insight into the catalytic performance of the ADOs, with information on the relative production efficiencies and substrate specificities between the enzymes. The study revealed common substrate-specific characteristics of the ADOs, as well as clear enzyme-specific differences in the catalytic performance. The observed variation in the production of the volatile alkanes between the ADOs is not obvious at amino acid level, but the information may still provide new perspectives into the detailed structural comparison in respect to kinetic traits of the different orthologs. The results also demonstrated that endogenous aldehyde reductases of the E. coli host exhibit affinity towards the C4-C8 aldehyde substrates, which must be taken into consideration when evaluating ADO-based microbial production systems for short-chain volatile alkanes.