• Open Access

A novel thiolase–reductase gene fusion promotes the production of polyhydroxybutyrate in Arabidopsis


* Correspondence (fax 617-492-1996; e-mail lkourtz@metabolix.com)


The production of polyhydroxybutyrate (PHB) involves a multigene pathway consisting of thiolase, reductase and synthase genes. In order to simplify this pathway for plant-based expression, a library of thiolase and reductase gene fusions was generated by randomly ligating a short core linker DNA sequence to create in-frame fusions between the thiolase and reductase genes. The resulting fusion constructs were screened for PHB formation in Escherichia coli. This screen identified a polymer-producing candidate in which the thiolase and reductase genes were fused via a 26-amino-acid linker. This gene fusion, designated phaA-phaB, represents an active gene fusion of two homotetrameric enzymes. Expression of phaA-phaB in E. coli and Arabidopsis yielded a fusion protein observed to be the expected size by Western blotting techniques. The fusion protein exhibited thiolase and reductase enzyme activities in crude extracts of recombinant E. coli that were three-fold and nine-fold less than those of the individually expressed thiolase and reductase enzymes, respectively. When targeted to the plastid, and coexpressed with a plastid-targeted polyhydroxyalkanoate (PHA) synthase, the fusion protein enabled PHB formation in Arabidopsis, yielding roughly half the PHB formed in plants expressing individual thiolase, reductase and synthase enzymes. This work represents a first step towards simplifying the expression of the PHB biosynthetic pathway in plants.


Polyhydroxyalkanoates (PHAs) are a broad family of biodegradable and biocompatible polymers that accumulate naturally in numerous microorganisms as an intracellular reserve of carbon (Kim and Lenz, 2001; Steinbuchel and Hein, 2001). PHAs consist of polymerized hydroxyacids, occurring as homopolymers or copolymers, and their physical properties vary with the composition of their monomer side chains (Sudesh et al., 2000). Over 130 different types of hydroxyacid monomer have been identified (Steinbuchel and Valentin, 1995), broadening the range of properties that can be obtained with these materials. The production of PHAs has been engineered into a number of prokaryotic and eukaryotic organisms, including bacteria (Taguchi et al., 2002), yeast (Marchesini et al., 2003; Terentiev et al., 2004), insect cells (Williams et al., 1996) and plants (Poirier and Gruys, 2002; Snell and Peoples, 2002).

The synthesis of the homopolymer polyhydroxybutyrate (PHB) requires the presence of three enzyme activities, thiolase, reductase and synthase, to convert acetyl-CoA to polymer (Madison and Huisman, 1999). Transformation of genes encoding these activities into plants has led to PHB production in Arabidopsis (Nawrath et al., 1994), tobacco (Lössl et al., 2003), Brassica napus (Houmiel et al., 1999), corn (Poirier and Gruys, 2002) and flax (Wrobel et al., 2004). To date, the highest yields of PHB have been produced in plastids of Arabidopsis thaliana using nuclear-encoded expression of genes encoding plastid-targeted PHB pathway enzymes (Nawrath et al., 1994; Bohmert et al., 2000). In these experiments, expression cassettes within the transformation constructs contained the same promoter, polyadenylation signal and plastid targeting signal. Although this experimental design helps to achieve coordinated expression of pathway genes for optimum product formation, previous studies have suggested that repeated use of the same promoter may result in gene silencing (De Wilde et al., 2000). Future engineering efforts to increase polymer yield and to produce healthy plants without the altered phenotypes sometimes observed in high PHB producers (Bohmert et al., 2000; Poirier and Gruys, 2002) will probably entail further engineering efforts that may require the expression of additional enzyme activities. These extra genes will add significantly to the complexity of the transformation vectors. Furthermore, the chances for gene silencing will be increased if each additional expression cassette contains its own promoter, plastid targeting signal and polyadenylation sequence. The efficiency and success of these future genetic engineering efforts will probably depend on the technique chosen for multigene expression.

Although the methods for engineering nuclear-encoded, plastid-targeted PHB production cited above have predominantly relied on the creation and crossing of individual lines expressing PHB pathway enzymes, cotransformation of multiple vectors or transformation of a single multigene expression cassette, other methods have been described for simplified, nuclear-encoded gene expression in plants (Halpin et al., 2001; Hunt and Maiti, 2001; Francois et al., 2002a). These methods include the use of polyproteins (Urwin et al., 1998; Francois et al., 2002b; El Amrani et al., 2004), internal ribosome entry sites (Urwin et al., 2002) and artificial gene fusions. Gene fusions are created by artificially fusing two or more coding sequences to form an artificial genetic fragment encoding a novel protein with multiple enzymatic activities (Bulow and Mosbach, 1991; Nixon et al., 1998). The resulting fusion proteins can, in some instances, yield an added benefit by possessing more desirable kinetic properties for a combination of sequential enzymatic reactions than equivalent mixtures of individual enzymes (Ljungcrantz et al., 1989; Lindbladh et al., 1994; Mao et al., 1995; Carlsson et al., 1996; Riedel and Bronnenmeier, 1998; Seo et al., 2000). Gene fusions have also been successfully used in whole plants to express multiple enzymatic activities from one polypeptide (Beaujean et al., 2000; Garg et al., 2002; Yilmaz and Bulow, 2002; Jang et al., 2003).

In this study, we chose to fuse the thiolase and reductase genes in order to simplify PHB expression through the development of a single gene encoding the first two enzymes of this pathway. The challenges associated with the creation of active gene fusions, such as promoting proper folding of active sites while minimizing oligomerization and proteolysis (Bulow and Mosbach, 1991), are compounded for the thiolase–reductase fusion by the fact that both enzymes function catalytically as homotetramers (Davis et al., 1987; Ploux et al., 1988). Complex structures can theoretically form upon fusion of multimeric proteins (Bulow and Mosbach, 1991), and previous fusions of a dimeric protein with a tetrameric protein (Carlsson et al., 1996; Kim et al., 2000b) have provided some indication of the formation of complex structures. In order to increase our chances of isolating an active gene fusion, we developed a method to generate a library of fusions with different lengths of linker sequences by joining the thiolase and reductase domains with a randomly ligated core linker sequence. The simplification of the PHB pathway and the results obtained with this novel gene fusion in recombinant Escherichia coli and transgenic Arabidopsis are described.


Fusion of thiolase and reductase coding sequences and characterization in E. coli

The challenges associated with the expression of the three-gene PHB biosynthetic pathway in plants encouraged us to simplify the pathway by creating a novel thiolase–reductase gene fusion (Figure 1). For this purpose, the stop codon of the thiolase gene from Ralstonia eutropha, phaA, was removed, and the gene's coding sequence was fused to the amino terminus of the R. eutropha reductase gene, phaB (Peoples and Sinskey, 1989b) with linkers that varied in length and composition (illustrated in Figure 1). The resulting library of gene fusions was transformed into a synthase-expressing E. coli strain, and screened for polymer production via plate assays. DNA sequencing of the highest PHB producer identified by this screen revealed that the thiolase and reductase genes of this candidate were fused with a 78-base pair (bp) in-frame linker sequence (Figure 1). This gene fusion, designated phaA-phaB, was chosen for further characterization via Western blotting, enzyme assays and polymer accumulation assays.

Figure 1.

Schematic representation of the thiolase–reductase gene fusion. The thiolase gene (phaA) contains a start codon and is translationally fused to the reductase gene (phaB) via a 26-amino-acid linker designated L5-4. The BamHI site used to introduce the linker is shown in italics. The structure of a single linker unit created by annealing and ligating oligonucleotides L5A and L5B is shown in parentheses. The BsaWI restriction site used as a diagnostic tool to identify plasmids containing the linker is shown by the shaded box. The eight repeats of the annealed L5A and L5B linkers within linker L5-4 are alternately underlined and shown with a line above.

Western blot analysis of E. coli cells expressing pTRC(AB), a plasmid containing the phaA-phaB gene fusion, yielded a single 69-kDa band detectable with antibodies against thiolase (Figure 2A) and reductase (Figure 2B) proteins. This molecular weight corresponds to the predicted mass of the PhaA-PhaB fusion protein. Individual thiolase and reductase proteins were not observed in these samples. Collectively, these results confirm that the phaA-phaB gene fusion encoded a fusion protein. As expected, expression in E. coli of the positive control plasmid pSU(A + B), in which the thiolase and reductase proteins are expressed individually, yielded bands corresponding to individual thiolase and reductase proteins (Figure 2A,B). Thiolase, reductase and fusion proteins were not detected in E. coli strains expressing the negative control plasmid pNEB193.

Figure 2.

Characterization of the PhaA-PhaB fusion protein expressed in bacteria. (A, B) Whole cell extracts were prepared from Escherichia coli XL10 cells containing either the control plasmid pNEB193, plasmid pSU(A + B) expressing the individual thiolase and reductase genes or plasmid pTRC(AB) containing the phaA-phaB gene fusion. Proteins (2 µg) were separated by sodium dodecylsulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and analysed by Western blotting with antibodies against the thiolase (A) or the reductase (B) proteins. (C, D) Enzyme assays and polymer accumulation assays were performed as described in ‘Experimental procedures’. (C) Thiolase and reductase activities are shown for E. coli BL21 strains expressing the phaA-phaB fusion gene in pTRC(AB), the individual thiolase and reductase genes in pSU(A + B) or the pTRCN vector alone. The data represent the average and standard error of at least four trials assayed in triplicate. (D) Polyhydroxybutyrate (PHB) accumulation data of cells expressing a PHA synthase gene on pSU(C) and the phaA-phaB gene fusion, the individual thiolase and reductase genes or the control vector pTRCN. The data represent the average and standard error of three trials.

Enzyme assays of the E. coli strain expressing the PhaA-PhaB fusion protein yielded activities of 10.2 and 1.0 units/mg total protein for thiolase and reductase, respectively (Figure 2C). These values were three-fold and nine-fold less than those observed when the proteins were expressed individually in strain BL21/pSU(A + B). Strain BL21/pTRCN did not yield significant thiolase or reductase activity (Figure 2C). In polymer accumulation assays, strain BL21/[pTRC(AB) + pSU(C)], expressing the PhaA-PhaB fusion and PHA synthase, produced 63 mg PHB per gram dry cell weight for a dry weight (dwt) PHB yield of 6.3%, 5.6-fold less PHB than the control strain BL21/[pTRC(A + B) + pSU(C)] expressing thiolase, reductase and synthase as individual proteins (Figure 2D). As expected, no polymer was produced in the control strain BL21/[pTRCN + pSU(C)].

Expression of the phaA-phaB gene fusion in Arabidopsis

Based on the ability of the phaA-phaB gene fusion to enable polymer formation in an E. coli host which expressed the synthase, transformation vectors for the introduction of plastid-targeted phaA-phaB and PHA synthase genes into plants were constructed. This simplified plant transformation vector utilizes one less set of promoters, signal peptides and terminators than the positive control plant transformation vector pCAM(C + A + B) that contains individual expression cassettes for plastid-targeted synthase, thiolase and reductase genes. The chloroplast was chosen as the site of product formation due to previously described successes at producing high levels of PHB within this organelle (Bohmert et al., 2000). Transformation of plasmids pCAM(C + AB) and pCAM(C + A + B) into Arabidopsis generated 89 and 111 transgenic plants, respectively, that were resistant to the antibiotic selection agent. These plants were examined for the presence of thiolase and reductase proteins by Western blotting (Figure 3). The phaA-phaB gene fusion was expressed as a 72-kDa fusion protein that could be detected with both anti-thiolase (Figure 3A, lane 4) and anti-reductase (Figure 3B, lane 1) antibodies. This protein is slightly larger than the 69-kDa PhaA-PhaB protein expressed in E. coli (Figure 3A, lane 1) due to the presence of 24 extra amino acids from the early mature region of the small subunit of rubisco and a three-amino-acid linker. Plants transformed with control plasmid pCAM(C + A + B) expressed individual thiolase (Figure 3A, lane 5) and reductase (Figure 3B, lane 2) proteins as expected. Interestingly, the anti-thiolase antibodies also detected a 75-kDa protein in wild-type and transgenic Arabidopsis extracts (Figure 3A, lanes 3–5). The identity of this band is unknown, but its presence in the wild-type control (Figure 3A, lane 3) and absence in bacterial extracts (Figure 3A, lanes 1–2) suggests a cross-reacting plant protein that may be an endogenous thiolase homologue.

Figure 3.

Western blots of crude extracts prepared from transgenic Arabidopsis plants probed with antibodies raised against thiolase (A) or reductase (B) proteins. (A) Lanes 1–2, 0.2 µg protein extract from Escherichia coli cells expressing pTRC(AB) or pSU(A + B), respectively; lanes 3–5, 2 µg of extract from wild-type, pCAM(C + AB) or pCAM(C + A + B) lines of Arabidopsis, respectively. (B) Lanes 1–3, 2.5 µg protein extract from pCAM(C + AB), pCAM(C + A + B) and wild-type lines of Arabidopsis, respectively.

Fluorescence microscopy of Nile Blue-stained leaf slices allowed the visualization of granular inclusions within the leaves of polymer-producing plants. Plants transformed with the synthase and the phaA-phaB gene fusion expression construct pCAM(C + AB) yielded fluorescent spots (Figure 4A), whereas plants transformed with the phaA-phaB-only expression construct did not yield spots (Figure 4B), consistent with the spots being polymer granules. Analysis of leaf material by transmission electron microscopy provided further evidence of granule formation within the leaves of plants expressing the synthase and the phaA-phaB gene fusion. Granular inclusions were visible in chloroplasts of plants transformed with the synthase and phaA-phaB gene fusion expression construct (Figure 4C). Plants transformed with just the phaA-phaB gene fusion did not contain these inclusions (Figure 4D).

Figure 4.

Analysis of transgenic Arabidopsis plants via microscopy. (A, B) Fluorescence microscopy of leaf tissue stained with Nile Blue. Tissue was obtained from Arabidopsis plant C + AB #8 expressing PHA synthase and the PhaA-PhaB fusion protein and producing 0.55% dry weight (dwt) of polyhydroxybutyrate (PHB) (A), or plant AB #10 expressing the PhaA-PhaB fusion protein alone (B). (C, D) Electron microscopy of leaf tissue obtained from Arabidopsis plant C + AB #9 producing 2.9% dwt PHB (C), or plant AB #10 (D). PHB granules are marked with an arrow in (C). Bar = 0.5 µm.

Western blot analysis with anti-thiolase and anti-reductase antibodies suggested that there was some correlation between higher levels of PhaA-PhaB expression and increased yields of polymer. A plant which did not produce PHB did not contain detectable PhaA-PhaB fusion protein (Figure 5A,B, lane 3), whereas a plant producing 0.5% dwt PHB expressed low levels of the fusion protein (Figure 5A,B, lane 4). Likewise, plants producing 1.3–1.8% dwt PHB expressed significantly higher levels of PhaA-PhaB (Figure 5A,B, lanes 5–7) than a plant producing 0.5% dwt PHB (Figure 5A,B, lane 4). A similar relationship between enzyme expression level and PHB production was also observed in plants transformed with the pCAM(C + A + B) construct. Plants expressing the highest levels of reductase (Figure 5C, lanes 6–8) produced PHB levels of 5.1–6.9% dwt, whereas plants producing lower amounts of reductase (Figure 5C, lanes 3–5) accumulated 1.2–3.0% dwt PHB. Although high levels of thiolase expression were observed in all the pCAM(C + A + B) plants tested, the highest PHB producers appeared to express slightly more thiolase than the low producers (Figure 5D). Signals for the AB fusion and individual thiolase or reductase proteins were not detected in wild-type plants (Figure 5A–D, lane 2).

Figure 5.

Analysis of protein expression in Arabidopsis plants producing variable amounts of polyhydroxybutyrate (PHB). (A, B) Western blots of plants transformed with plasmid pCAM(C + AB) probed with antibodies raised against reductase (A) or thiolase (B). Lanes 1 and 2 contain 1 µg Escherichia coli XL10/pTRC(AB) and 5 µg wild-type Arabidopsis extract, respectively. Lanes 3–7 contain 5 µg protein extract from plants that accumulated PHB at levels of 0, 0.5, 1.3, 1.5 or 1.8% dry weight (dwt), respectively. (C, D) Western blots of plants transformed with plasmid pCAM(C + A + B) probed with antibodies raised against reductase (C) or thiolase (D). Lanes 1 and 2 contain 1 µg E. coli XL10/pSU(A + B) and 7.5 µg wild-type Arabidopsis extract, respectively. Lanes 3–8 contain 7.5 µg protein extract from plants that accumulated PHB at levels of 1.2, 1.6, 3.0, 5.1, 6.0 or 6.9% dwt, respectively.

The top PHB-producing plant isolated upon transformation of pCAM(C + AB) contained 6.4% dwt PHB, slightly more than half of the 11.5% produced by the best plant transformed with pCAM(C + A + B). A higher percentage of Arabidopsis C + AB plants appeared to accumulate 0.5–1.0% dwt PHB than did plants transformed with the C + A + B construct (Figure 6). The average amount of PHB produced in plants transformed with pCAM(C + AB) was 0.8% dwt PHB with a standard error of 0.9%. This value was slightly lower than the average amount of 1.1% dwt PHB with a standard error of 1.5% recorded in plants transformed with the pCAM(C + A + B) construct. More C + A + B plants accumulated PHB levels in excess of 1% dwt than did C + AB plants (Figure 6). This difference in PHB yield is probably not due to the variable number of copies of the PHB genes, as segregation analysis of T2 plants generated from four top PHB-producing C + AB and C + A + B T1 plants revealed that 75% of these lines contained one T-DNA insert, whereas 25% expressed two copies of the PHB genes. Collectively, these results suggest that, although the individual thiolase and reductase enzymes slightly outperform the thiolase–reductase fusion with respect to in planta PHB production, the PhaA-PhaB fusion protein is functional, and represents a first step towards the simplification of the PHB production pathway in plants.

Figure 6.

Comparison of polyhydroxybutyrate (PHB) production in T1Arabidopsis C + AB and C + A + B transgenic lines. ‘Non-quantifiable’ represents samples that contained trace amounts of PHB that could not be accurately quantified due to the small sample size and/or the 0.25 mg PHB detection limit observed with our butanolysis/gas chromatography method. Eighty-nine T1 plants transformed with plasmid pCAM(C + AB) were generated from a single transformation event and were analysed in this study. A comparable transformation with the plasmid pCAM(C + A + B) generated the 111 T1 C + A + B plants analysed in this study.


The expression of fusion proteins is an attractive alternative for pathway engineering in plants. Fusion proteins allow the coordinated expression of component enzymes. In addition, the close proximity of active sites within fusion proteins has been suggested in some cases to increase reaction rates by allowing the transfer of metabolites between sites before re-equilibration with the surrounding environment can occur (Bulow and Mosbach, 1991). The concept that fusion proteins promote substrate channelling is controversial (Pettersson et al., 2000), but gene fusions clearly allow the simplification of plant transformation vectors by eliminating one or more sets of promoter and terminator sequences, thereby reducing the complexity of vector construction and possibly the potential for gene silencing (De Wilde et al., 2000). Several examples of artificial fusion proteins have been successfully expressed in plants. For example, the monomeric α-amylase was fused with the tetrameric glucose isomerase enzyme and expressed in potato tubers (Beaujean et al., 2000). Fusion of the tetrameric glycine betaine protein to choline dehydrogenase, an enzyme whose native structure has not been described, led to the successful production of glycine betaine in tobacco (Yilmaz and Bulow, 2002). In addition, a fusion protein consisting of trehalose-6-phosphate synthase and trehalose-6-phosphate phosphatase monomers was shown to catalyse trehalose accumulation in rice (Garg et al., 2002; Jang et al., 2003).

In this study, coding sequences of thiolase and reductase, two tetrameric enzymes that collectively convert acetyl-CoA to R-3-hydroxybutyryl CoA for polymerization by PHA synthase, were fused to simplify expression of the PHB pathway in plants. This approach was chosen in the hope that the presence of reductase activity on the same polypeptide as the thiolase enzyme might direct more carbon towards polymer synthesis. Previous studies have shown that the first reaction of the PHB biosynthetic pathway, condensation of acetyl CoA by thiolase to produce acetoacetyl CoA, is thermodynamically unfavourable, and the subsequent conversion of acetoacetyl CoA to 3-hydroxybutyryl CoA by reductase helps to drive the reaction towards polymer synthesis (Masamune et al., 1989).

A library of thiolase and reductase genes fused via random linkers was screened for polymer production in a synthase-expressing E. coli strain. The vast majority of candidate enzymes encoded by the gene fusions did not catalyse the formation of high levels of PHB, and exhibited decreased solubility. These results may be attributable to the complex structures formed by the fusion of two homotetramers. Previous studies have suggested that the native structure of the individual component enzymes can promote the formation of complex structures with the fusion protein in vivo that can be accompanied by decreased solubility and a loss of activity (Bulow and Mosbach, 1991). The gene fusion candidate phaA-phaB, which was designated the most promising candidate for plant-based expression based on its performance in bacterial PHB accumulation assays, exhibited nine-fold less reductase and three-fold less thiolase activity than individual thiolase and reductase enzymes. Surprisingly, this decreased enzymatic activity was accompanied by only a five-fold decrease in PHB production in E. coli and, on average, a 30% decrease in PHB accumulation in Arabidopsis. These data, in conjunction with the finding that the top polymer-producing Arabidopsis plant expressing the fusion produced two-fold less PHB than the top polymer-producing plant expressing the individual enzymes, strongly suggests that an E. coli screen of gene fusions can permit the identification of enzymes that perform well in planta.

The decreased yield of polymer observed upon expression of the PhaA-PhaB fusion in both bacteria and plants may be due to problems with proper folding of the thiolase and reductase active sites, possibly altering the activities of the enzymes. The random linkers used to construct the library of thiolase–reductase fusions were designed to preferentially encode amino acids (serine, threonine, glycine) abundant in linkers of naturally occurring proteins (Agros, 1990). Despite this precaution, conformational constraints may still exist. For instance, the short region on the N-terminus of the reductase protein, encoding the NADPH-binding fold (Peoples and Sinskey, 1989a), is now located in the middle of the PhaA-PhaB fusion protein. Suboptimal folding of this domain could account for the finding that the reductase activity (nine-fold decrease) was compromised to a greater extent than the thiolase activity (three-fold decrease) in PhaA-PhaB. Previous studies with a β-galactosidase–galactose dehydrogenase fusion have suggested that longer linkers can facilitate proper folding, thereby allowing increased activity (Carlsson et al., 1996). The 26-amino-acid linker of the phaA-phaB gene fusion is considerably longer than the average linker length of 6.5 amino acids observed in natural proteins (Agros, 1990). Active fusions with longer linkers were not observed in this study, possibly because longer linkers have previously been associated with an increased risk of oligomerization (Carlsson et al., 1996) and/or proteolysis (Bulow and Mosbach, 1991; Seo et al., 2000). Some inclusion bodies were observed during expression of the fusion in recombinant E. coli suggesting possible oligomerization. This finding is not surprising as thiolases and reductases naturally exist as homotetramers (Davis et al., 1987; Ploux et al., 1988), and may be expected to form complex protein structures (Bulow and Mosbach, 1991). Possible reduced oligomerization of the PhaA-PhaB fusion in Arabidopsis due to lower expression levels may be one explanation for the enhanced performance of the fusion in plants relative to that observed in bacteria. Directed evolution technologies (Turner, 2003) could be used to optimize the phaA-phaB fusion for increased product formation. These technologies have successfully optimized the stability of a novel N-carbamylase–d-hydantoinase fusion enzyme (Kim et al., 2000a).

Previous studies have noted that high levels of PHB production are accompanied by a stunted chlorotic plant phenotype (Bohmert et al., 2000; Poirier and Gruys, 2002). Interestingly, this study revealed that the presence of the thiolase–reductase fusion did not result in an amelioration of this phenotype when compared with plants expressing individual thiolase and reductase enzymes that produced similar levels of polymer. Preventing this phenotype in future genetic engineering efforts will require a greater understanding of carbon and cofactor pools that might be drained upon PHB formation. Metabolic profiling (Sumner et al., 2003; Trethewey, 2004) studies could provide guidance for additional genetic engineering strategies to correct the phenotype. The use of gene fusions could facilitate these experiments by reducing the number of genes to be expressed, simplifying pathway expression.

The next logical step for the simplification of PHB pathway expression in transgenic plants is the creation of a three-gene fusion in which all of the enzyme activities for PHB production are contained on one polypeptide. Ljungcrantz et al. (1990) created a three-gene fusion of galactose dehydrogenase, β-galactosidase and galactokinase that displayed reduced activity of each of the component enzymes. The four subunits of the E. coli glucose phosphotransferase system were also fused to yield an active fusion protein (Mao et al., 1995). Although a three-gene fusion for PHB synthesis would probably need to be optimized by directed evolution technologies, it would simplify the expression of the PHB pathway in plants such that only one gene, promoter, targeting signal and polyadenylation sequence would be required to express the complete PHB production pathway.

Experimental procedures

Strains, media, oligonucleotides and plasmids

Dehydrated LB Broth (Miller) and Bacto 2x YT were purchased from Difco Laboratories (Detroit, MI, USA). Antibiotics were added to the culture media as required at the following concentrations: chloramphenicol (25 µg/mL); ampicillin (100 µg/mL); gentamycin (30 µg/mL); and kanamycin (50 µg/mL). pTRC plasmids were derived from pTRCN, a low copy pBR322 derivative that encodes a gene for ampicillin resistance (Gerngross et al., 1994). pSU plasmids were low copy pACY184 derivatives that encode a gene for chloramphenicol resistance (Bartolome et al., 1991). The oligonucleotides used in this study are listed in Table 1. Plasmid constructs were verified by restriction digests. Constructs generated via polymerase chain reaction (PCR) were also checked by DNA sequencing.

Table 1.  Sequences of oligonucleotides used in this study
NameDNA sequence


Plasmid pTRCAB11 encodes a fusion protein consisting of the R. eutropha thiolase, a 6-bp linker encoding a BamHI site and the R. eutropha reductase (Figure 1). The plasmid was constructed using the following multistep procedure. The thiolase gene was amplified with its initiation codon, but without a termination codon from plasmid pAeT413, a derivative of plasmid pAeT41 (Peoples and Sinskey, 1989c), using primers A1-FKpn and A1-FBam. The isolated PCR product was digested with KpnI and BamHI and cloned into the KpnI and BamHI sites of pTRCN, forming plasmid pTRCAF. The reductase gene was amplified from pAeT413 with primers B1-LBam and B1-LXba. The PCR product was digested with BamHI and XbaI and cloned into the BamHI and XbaI sites of pTRCN to create pTRCBL. The BamHI/XbaI phbB gene fragment from pTRCBL was cloned into pTRCAF to produce plasmid pTRCAB11.


Plasmid pTRCAB11 was digested with BamHI prior to dephosphorylation with shrimp alkaline phosphatase. Oligonucleotides L5A and L5B were designed to anneal and insert repeatedly in a forward or reverse fashion in the BamHI site of pTRCAB11. These oligonucleotides were phosphorylated using T4 polynucleotide kinase, annealed and ligated into pTRCAB11 to produce a library of pTRCAB11 derivatives containing linkers of various lengths and configurations. This library was transformed into MBX240, an E. coli strain that constitutively expresses the PHB synthase gene (Peoples et al., 2000). The resulting transformants were plated on to LB plates containing 1% glucose, grown overnight at 30 °C and qualitatively screened for polymer production. PHB-producing colonies, identifiable by their white opaque phenotype, were further characterized by restriction digestion with BsaWI, a restriction site uniquely present in the linker formed by annealed oligonucleotides L5A and L5B. The highest PHB-producing strains containing this site were further characterized by DNA sequencing, quantitative polymer accumulation assays and enzyme assays. The highest PHB-producing construct, designated pTRC(AB), contained the gene fusion phaA-phaB in which the thiolase and reductase genes were fused via a 26-amino-acid linker, designated L5-4, formed by the random ligation of eight copies of annealed L5A and L5B oligonucleotides in the forward and reverse directions (Figure 1).

pSU(A + B)

This was constructed by inserting a 2.3 kbp PstI fragment, containing the phaA and phaB genes from pAeT10 (Peoples and Sinskey, 1989b), into the PstI site of pSU18 (Bartolome et al., 1991).

pTRC(A + B)

This was constructed by swapping the phaA-phaB fusion gene of pTRC(AB) with the phaA and phaB genes from pAet10 using an NcoI-BsrGI digest. The correct clone was confirmed by loss of the unique BglII restriction site found in the 26-amino-acid L5-4 linker of pTRC(AB) (Figure 1). Western blotting confirmed the expression of the individual thiolase and reductase proteins and the absence of the PhaA-PhaB fusion protein.


Plasmid pCMSY106 contains a 1.7-kbp HindIII fragment encoding a hybrid Pseudomonas oleovorans/Zoogloea ramigera synthase, designated phaC (Huisman et al., 2001) in vector pUC19 (Yanisch-Perron et al., 1985). Primers 554 and 555 were used to amplify the hybrid synthase from plasmid pMSXC5cat (Huisman et al., 2001), and the resulting blunt-ended fragment was cloned into the SmaI site of pUC19.


This was constructed by amplifying the phaC gene from pCMSY106 using primers CF-Xba-RBS and CR-Bam. These oligonucleotides were designed to add a 5′XbaI site, an optimized bacterial ribosome binding site and a 3′BamHI site to phaC. The resulting fragment was inserted into the XbaI and BamHI sites of pSU19 (Bartolome et al., 1991).


This contains the 35S-C4PPDK promoter (Chiu et al., 1996), DNA encoding the signal peptide of the small subunit of rubisco from pea and the first 24 amino acids of the mature protein (Cashmore, 1983), DNA encoding a three-amino-acid linker that contains an XbaI restriction site allowing fusion of the desired transgene, and the 3′ terminator of the nopaline synthase gene (Bevan et al., 1983). This plasmid was constructed using the following multistep procedure. Oligonucleotides BamXbaNot–A and BamXbaNot–B were annealed and ligated into plasmid pUC18-35S-C4PPDKsGFPnos (Chiu et al., 1996) that had been previously digested with BamHI and NotI. The resulting plasmid was named pUC18-35S-C4PPDK–BXNP-nos. The rubisco chloroplast targeting signal and the first 24 amino acids of the mature protein were amplified from genomic DNA obtained from expanded young green leaves of Pisum sativum Progress #9 using primers PEATSC and PEATSR. The resulting 0.34-kbp fragment was cloned into the BamHI and XbaI sites of pUC18-35S-C4PPDK-BXNP-nos forming plasmid pUC18-35S-C4PPDK-P.t.s.nos. To remove the intron from the pea targeting signal, plasmid pUC18-35S-C4PPDK-P.t.s.nos was digested with SphI and MfeI. Linkers P.t.s.nointron A and P.t.s.nointron B were annealed and ligated into the SphI and MfeI sites of pUC18-35S-C4PPDK-P.t.s.nos to create pUC18-C4PPDK-rbcs-nos. The start site of the signal sequence from plasmid pUC18-C4PPDK-rbcs-nos was optimized for plant expression by changing the existing TCCATGG sequence to AAAATGG using the QuikChange® Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA) to form plasmid pUC-C4PPDK-AAA-RBS-nos.


This was derived from the pCAMBIA2300 binary vector (Centre for Application of Molecular Biology to International Agriculture, Canberra, Australia) and contains the promoter, signal sequence and terminator fragment previously described for plasmid pUC18-C4PPDK-AAA-RBS. Plasmid pCAM(RBCS-link) was constructed using the following multistep procedure. Double-stranded synthetic linkers 1 and 2 were prepared by annealing oligonucleotides 1A and 1B, and oligonucleotides 2A and 2B, respectively. The promoter, signal sequence and terminator were excised from plasmid pUC-C4PPDK-AAA-RBS-nos using unique EcoRI and XhoI sites. The resulting 1.1-kbp fragment was annealed to linkers 1 and 2 to create a promoter, signal sequence and terminator DNA fragment flanked on the 5′ end by HindIII, BstBI, PacI and XhoI restriction sites and on the 3′ end by EcoRI, PacI, AscI, AvrII and SacI sites. This fragment was cloned into the SacI and HindIII sites of the plant transformation vector pCAMBIA2300 to create pCAM(RBCS-link).


This was created by amplifying the phaA-phaB fusion gene from pTRC(AB) using primers AB-F and AB-R. The resulting fragment was cloned into the XbaI and PstI sites of pCAM(RBCS-link) to create a translational fusion of the pea targeting signal with the phaA-phaB gene.


This was created by amplifying the phaA gene from pAeT10 (Peoples and Sinskey, 1989b) using primers AB-F and A-R. The resulting fragment was cloned into pCAM(RBCS-link) using the procedure previously described for pCAM(AB).


This was created by amplifying the phaB gene from pAeT10 (Peoples and Sinskey, 1989b) using primers B-F and AB-R, and cloning the DNA fragment into pCAM(RBCS-link) using the procedure previously described for pCAM(AB).


This was created by amplifying the synthase gene from pCMYS106 using primers C-F and C-R. The PCR product was digested with XbaI and NsiI. The resulting fragment was cloned into the compatible cohesive ends of the XbaI and PstI sites of pCAM(RBCS-link) to create pCAM(C).

pCAM(C + AB)

This was created by removing the phaA-phaB cassette from pCAM(AB) using the 5′BstBI and 3′AvrII sites. The BstBI site was blunted, and the resulting DNA fragment was cloned into the AvrII and blunted AscI sites of pCAM(C).

pCAM(C + A + B)

This was constructed using a two-step process. pCAM(A + B) was created by removing the phaA cassette from pCAM(A) using the 5′BstBI and 3′AvrII sites, blunting the BstBI site and cloning the resulting insert into the AvrII and blunted AscI sites of pCAM(B). pCAM(C + A + B) was created by removing the phaA and phaB cassettes from pCAM(A + B) using the 5′BstBI and 3′AvrII sites, blunting the BstBI site and cloning this insert into the AvrII and blunted AscI sites of pCAM(C).

Transformation of A. thaliana

Electrocompetent cells of Agrobacterium strain GV3101/pMP90 (Koncz and Schell, 1986) were transformed with plasmid DNA, and single colonies were isolated on LB plates containing gentamycin and kanamycin. A. thaliana Columbia Col-0 (Lehle Seeds, Round Rock, TX, USA) was grown in soil at 20 °C, 70% humidity and a 16 h light, 8 h dark cycle. Plants were transformed using an Agrobacterium-mediated floral dip procedure described by Clough and Bent (1998). Seeds from mature siliques were harvested, sterilized and spread onto selection plates containing 1/2 × Murashige Minimal Organics Medium (Life Technologies, Rockville, MD, USA), 0.7% agar, 1 × Gamborg's B5 vitamins (Sigma, St. Louis, MO, USA) and kanamycin (50 µg/mL). Plates were incubated for 2 days at 4 °C and transferred to 20 °C, 70% humidity and a 16 h light, 8 h dark cycle. After 7 days, green kanamycin-resistant seedlings were transferred to soil and incubated at the same growth conditions until plants were ready for analysis.

Production of antibodies

The production of polyclonal antibodies to the R. eutropha synthase has been described previously (Gerngross et al., 1993). Protein for production of antibodies to the R. eutropha thiolase and reductase was purified from recombinant E. coli essentially as described previously (Ploux et al., 1988; Palmer et al., 1991). Thiolase and reductase samples were purified by sodium dodecylsulphate-polyacrylamide gel electrophoresis (SDS-PAGE), stained with Coomassie blue and bands corresponding to each protein were excised for use as antigen. Polyclonal antibodies were produced by Fitzgerald Industries International, Inc. (Concord, MA, USA) in SPF (pasturella) New Zealand White rabbits. The IgG fraction was purified from the crude antibody mixture by protein A affinity chromatography.

Protein analysis

Bacterial and plant extracts for SDS-PAGE analysis were prepared as follows. E. coli XL10 cells (Stratagene) containing plasmid pNEB193 (New England Biolabs, Beverly, MA, USA), pSU(A + B) or pTRC(AB) were grown at 37 °C in 50 mL of LB supplemented with chloramphenicol or ampicillin. Cells were grown to mid-log phase prior to induction for 2 h with 1.0 mm isopropyl-β-d-1-thiogalactopyranoside (IPTG). The cells were harvested and resuspended in 1 mL of enzyme buffer [10 mm Tris-HCl, pH 8.1, 1 mm ethylenediaminetetraacetic acid (EDTA), 10 mmβ-mercaptoethanol, 20% glycerol, 0.2 mm phenylmethylsulphonylfluoride].

Leaves from transgenic Arabidopsis plants were harvested, placed in a 1.5 mL Eppendorf tube and immediately frozen in liquid nitrogen. Cell lysis buffer (100 mm Tris-HCl, pH 6.8, 10 mm EDTA, 4 mmβ-mercaptoethanol, 0.1 mm phenylmethylsulphonylfluoride) was supplemented with a broad-spectrum cocktail of protease inhibitors (Complete Mini tablets, Roche Diagnostics GmbH, Mannheim, Germany) according to the manufacturer's instructions. Two to five medium-sized leaves were homogenized in 200 µL of the supplemented buffer on dry ice with a Pellet Pestle-Motor (Kimble-Kontes, Vineland, NJ, USA).

Protein concentrations were determined using the Bradford assay (Bradford, 1976) with reagent purchased from Biorad (Hercules, CA, USA) and bovine serum albumin as the standard. For SDS-PAGE analysis, all samples were diluted in 2 × SDS gel loading buffer (100 mm Tris-HCl, pH 6.8, 100 mm dithiothreitol, 4% SDS, 0.2% bromophenol blue, 20% glycerol), boiled and loaded into Novex precast Tris-glycine gels (Invitrogen, Carlsbad, CA, USA). Gels were transferred to nitrocellulose (Biorad) using a Biorad Mini-Trans Blot cell. Transfer buffer consisted of 20% methanol, 25 mm Tris-base and 200 mm glycine. Blots were blocked using 3% Biorad blotting grade non-fat dry milk and washed with TBST (50 mm Tris-HCl, pH 7.5, 150 mm NaCl, 0.05% Tween 20). Protein detection by immunoblotting was performed with the Pierce Super Signal West Femto Max Sensitivity substrate as described by the manufacturer (Pierce, Rockford, IL, USA). Antibodies were diluted in TBST supplemented with 0.3% blotting grade blocker (Biorad) prior to use. Donkey anti-rabbit horseradish peroxidase-linked secondary antibodies and hyperfilm ECL film were purchased from Amersham (Piscataway, NJ, USA).

Enzyme assays

E. coli BL21 cells (Stratagene) containing pTRCN, pTRC(AB) or pSU(A + B) were grown overnight at 37 °C in 50 mL LB supplemented with ampicllin or chloramphenicol. Cells were diluted 1 : 1000 in fresh medium, grown to mid-log phase and induced with 1.0 mm IPTG. Cells were cultured for an additional 3 h at 37 °C prior to harvest. Cell pellets were resuspended in 1 mL of enzyme buffer (10 mm Tris-HCl, pH 8.1, 1 mm EDTA, 10 mmβ-mercaptoethanol, 20% glycerol, 0.2 mm phenylmethylsulphonylfluoride) and lysed by sonication on ice. The lysate was clarified by centrifugation (microfuge, 4 °C, 10 min) and the resulting supernatant was used for enzyme assays. Thiolase activity was assayed as described previously (Palmer et al., 1991) with some modifications. The assay mixture (1 mL) contained 62.4 mm Tris-HCl, pH 8.1, 50 mm MgCl2, 62.5 µm CoA and 62.5 µm acetoacetyl-CoA (Sigma). The assay was initiated upon the addition of enzyme and the loss of acetoacetyl-CoA was measured with time (ɛ304 = 16.9 × 103 cm−1 M−1). Acetoacetyl-CoA was quantified prior to the assay in 62.4 mm Tris-HCl, pH 8.1, and 50 mm MgCl2304 = 16.9 × 103 cm−1 M−1). Reductase activity was assayed as described previously (Ploux et al., 1988).

Polymer analysis

E. coli XL10 cells containing the synthase plasmid pSU(C), and either pTRCN, pTRC(AB) or pTRC(A + B), were grown at 30 °C in LB supplemented with ampicillin, chloramphenicol and 2% glucose. Cells were grown to mid-log phase, induced with 0.5 mm IPTG and cultured for an additional 2 days at 30 °C prior to harvest. Cell pellets were lyophilized and analysed for PHB content using a butanolysis/gas chromatography method as described previously (Snell et al., 2002).

Arabidopsis leaves from mature plants (approximately 6 weeks old) were harvested, lyophilized and dried tissue (30–150 mg) was ground to a fine powder. Samples were incubated for 2 h at 25 °C in 10 mm citric acid, pH 3.0. The resulting liquid was decanted and the remaining material was lyophilized and washed as described previously (Bohmert et al., 2000) with some modifications. Samples were washed once with 1 : 1 methanol : water, twice with methanol and once with ethanol. The PHB content of the remaining leaf material was determined using butanolysis and gas chromatography as described previously (Snell et al., 2002).


Fluorescence microscopy with Nile Blue staining was performed as described previously (Poirier et al., 1992) with some modifications. Leaf tissue was sliced as thin as possible with a razor blade and fixed in 3% paraformaldehyde (Electron Microscopy Sciences, Ft. Washington, PA, USA) in 0.1 m KH2PO4, pH 8, for 3 h. Fixed samples were washed with water and stained with a previously filtered 1% Nile Blue (Sigma) solution for 5 min at room temperature. Samples were washed with water and destained with 8% acetic acid. Samples were washed an additional two times with water. Samples were viewed by fluorescence microscopy on a Zeiss Axiolab light microscope equipped with a Zeiss HBO 100 fluorescence attachment and a 20× Ph-1 lens using the following filter set: exciter, HQ545/30; beam splitter, Q570lp; emitter D590/20 (Chroma Technology, Brattleboro, VT, USA). Images were recorded with a Zeiss MC 80 DX Microscope Camera using Kodak Elite Chrome 100 film.

Electron microscopy was performed using a Philips EM 410 microscope by Nicki Watson (I.W.M. Keck Biological Imaging Facility, The Whitehead Institute, Cambridge, MA, USA) essentially as described previously (Bohmert et al., 2000).


This research was supported by a Department of Energy Industry of the Future Award (DE-FC07-011D14214). The authors acknowledge Nicki Watson (I.W.M. Keck Biological Imaging Facility, The Whitehead Institute, Cambridge, MA, USA) for performing electron microscopy, Jen Sheen (Department of Molecular Biology, Massachusetts General Hospital, MA, USA) for generously supplying the plasmid pUC18-35S-C4PPDKGFPnos, Stephanie Aquin for constructing plasmid pUC18-35S-C4PPDK-P.t.s.nos, Nii Patterson for constructing plasmid pCMSY106 and Feng Feng for preparing antibodies to the R. eutropha thiolase and reductase proteins.