We report here the production of the bacterial polyester, polyhydroxybutyrate (PHB), in the crop species sugarcane (Saccharum spp. hybrids). The PHB biosynthesis enzymes of Ralstonia eutropha [β-ketothiolase (PHAA), acetoacetyl-reductase (PHAB) and PHB synthase (PHAC)] were expressed in the cytosol or targeted to mitochondria or plastids. PHB accumulated in cytosolic lines at trace amounts, but was not detected in mitochondrial lines. In plastidic lines, PHB accumulated in leaves to a maximum of 1.88% of dry weight without obvious deleterious effects. Epifluorescence and electron microscopy of leaf sections from these lines revealed that PHB granules were visible in plastids of most cell types, except mesophyll cells. The concentration of PHB in culm internodes of plastidic lines was substantially lower than in leaves. Western blot analysis of these lines indicated that expression of the PHB biosynthesis proteins was not limiting in culm internodes. Epifluorescence microscopy of culm internode sections from plastidic lines showed that PHB granules were visible in most cell types, except photosynthetic cortical cells in the rind, and that the lower PHB concentration in culm internodes was probably a result of dilution of PHB-containing cells by the large number of cells with little or no PHB. We discuss strategies for producing PHB in mitochondria and mesophyll cell plastids, and for increasing PHB yields in culms.
Polyhydroxyalkanoates (PHAs) are polyesters formed by many bacterial species as carbon and energy reserves. PHAs are divided into two groups: short side-chain length (C4–6) and medium side-chain length (C6–16). Over 125 different hydroxyacid side-chains have been identified. Depending on the side-chain composition, PHAs exhibit physical properties ranging from hard and brittle plastics to glues, elastomeric compounds and even rubber (Steinbüchel and Valentin, 1995). These polyesters have the potential to compete commercially with a range of compounds currently made from non-renewable petrochemicals (Steinbüchel and Valentin, 1995). The biology of accumulation of the short side-chain length PHA, polyhydroxybutyrate (PHB), has been studied extensively in the eubacterium Ralstonia eutropha. In this organism, PHB is produced from acetyl-coenzyme A (acetyl-CoA) by the successive action of three enzymes [β-ketothiolase (PHAA), acetoacetyl-reductase (PHAB) and PHB synthase (PHAC)], which are encoded by the genes phaA, phaB and phaC, respectively (Peoples and Sinskey, 1989a,b). PHAs can be readily produced by fermentation, a process that incurs substantial substrate and energy costs (Poirier et al., 1992).
Plants are an alternative PHA production platform as they use CO2 and sunlight as carbon and energy sources, respectively (Poirier et al., 1992; Snell and Peoples, 2002). In plants, there is a high flux of acetyl-CoA (the precursor metabolite for PHB biosynthesis) in chloroplasts, as sites of lipid biosynthesis, and in mitochondria, where the metabolite is used for energy generation. In addition, a lower flux occurs in the cytosol, where acetyl-CoA is used for the formation of isoprenoids and secondary metabolites (Liedvogel, 1986).
Several plant species have been transformed with the R. eutropha PHB biosynthesis genes, and the enzymes have been targeted to various cell compartments (Poirier and Gruys, 2002; Snell and Peoples, 2002; Lössl et al., 2003; Wróbel et al., 2004). In Arabidopsis thaliana, cytosolic expression of phaB and phaC resulted in very low polymer concentration (0.006% of leaf dry weight; Poirier et al., 1992), whereas targeting of all three R. eutropha PHB biosynthesis enzymes to chloroplasts resulted in very high polymer concentration (40% of leaf dry weight; Bohmert et al., 2000). There are no published reports of the enzymes being targeted to mitochondria. Mitochondria could potentially be ideal targets for PHB production because of their abundance in most tissues and their high acetyl-CoA flux (Liedvogel, 1986).
Sugarcane has the potential to be a key crop for PHB production, because it has a very large biomass (mean harvest in Australia over the past 12 years of 98 tonnes/ha; Briody, 2000; Morgan, 2005), requires replanting only every 4–5 years and contains abundant sucrose reserves in the culms (up to 40% of culm dry weight; Whittaker and Botha, 1997), which could be used as feedstock in mill-associated fermentation plants. Furthermore, sugarcane industries have substantial infrastructure for the transport of biomass to mills and the co-generation of power in excess of mill requirements from the fibre remaining after sucrose extraction. In addition, for nuclear transformants, the requirement for sugarcane cultivars to be vegetatively propagated may confer superior transgene containment (Anon., 2004) compared with sexually propagated crops. Lastly, sugarcane has been metabolically engineered to produce moderate levels of a glycosylated form of the liquid crystal precursor, p-hydroxybenzoic acid (7.5% dry weight of leaves, 1.5% dry weight of culms; McQualter et al., 2005), as well as sorbitol (12% dry weight of leaves, 1% dry weight of culms; B. Fong Chong, BSES Limited, Indooroopilly, Qld, Australia, unpublished results).
The aims of this study were to assess whether PHB could be produced in different cell compartments and organs of sugarcane, and to identify the factors affecting PHB production in this crop.
PHB was detected at very low levels in cytosolic lines and was not detected in mitochondrial lines
We attempted to accumulate PHB in the cytosol, mitochondria and plastids of sugarcane. For cytosol accumulation, we used the R. eutropha phaA, phaB and phaC genes. For accumulation in the mitochondria and plastids, these genes were translationally fused to the targeting presequences of the Nicotiana plumbaginifolia ATPase β-subunit gene (Boutry and Chua, 1985) or a pea ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) small subunit (SSU) gene (Fluhr et al., 1986), respectively. Each expression cassette was driven by the constitutive maize polyubiquitin promoter (Ubi-1; Christensen et al., 1992). Following transformation and regeneration, transgenic lines were transferred to a glasshouse.
Formal proof of PHB accumulation was obtained by a high-performance liquid chromatography (HPLC) assay adapted from an earlier acid-hydrolysis-based method (Karr et al., 1983; Figure 1). Samples were acid digested, converting PHB to crotonic acid. A putative crotonic acid peak was not detected in wild-type (WT) samples (Figure 1, top panel), but was detected at 30 min in the same samples spiked with PHB (Figure 1, middle), and in samples from the transgenic population (Figure 1, bottom). The UV absorption spectra of pure crotonic acid and the reaction product detected in the transgenic samples were near identical (Figure 1, inset in bottom panel), thus confirming that the samples were PHB positive.
After 3 months of growth in the glasshouse, leaves of the transgenic lines were screened by HPLC. PHB was detected at trace levels in only two of 95 cytosolic lines, and was not detected in any of 58 mitochondrial lines, even following preparation of highly concentrated samples (Table 1). Western blot analysis confirmed that all the PHB biosynthesis enzymes were expressed in the mitochondrial lines (Figure 2A). To verify that the dicot targeting presequence used to target the PHB biosynthesis enzymes to mitochondria was functional in a monocot host, we prepared a green fluorescent protein (GFP) fusion construct. Sections from leaf, root and culm of sugarcane lines transformed with this construct were examined by epifluorescence microscopy (Figure 2B; D. J. Anderson, The University of Queensland, unpublished data). Punctate fluorescence was observed, and staining with a mitochondria-specific dye confirmed these bodies as mitochondria (Figure 2C). Therefore, it was assumed that the PHB biosynthesis enzymes were targeted to the mitochondria correctly. Although this particular targeting presequence is functional in vitro in wheat germ extracts (Dessi et al., 2003), we believe that this is the first report of its functionality in a monocot host in vivo.
Table 1. Screening for polyhydroxybutyrate (PHB) by high-performance liquid chromatography in leaves and culms of transgenic sugarcane lines
Targeted cell compartment
Lines with detectable PHB
Leaf dry weight as PHB (%)
Lines with detectable PHB
Culm dry weight as PHB (%)
NQ, not quantifiable.
PHB production was most successful in plastidic lines
PHB was detected by HPLC in 26 of 130 plastidic lines (population statistics listed in Table 1). The highest polymer concentration observed was 1.88% of leaf dry weight as PHB in line TA4. The spatial distribution of PHB in TA4 leaves was determined by epifluorescence and electron microscopy. Epifluorescence microscopy following staining of TA4 sections with the lysochrome, Nile Blue A, revealed punctate fluorescence in most cell types, including epidermal, subsidiary stoma, bulliform and fibre cells, as well as in bundle sheath and phloem companion cells of vascular bundles (Figure 3C). Punctate fluorescence was very weak in mesophyll cells (Figure 3C), and was absent in control sections from WT plants (Figure 3A). Electron microscopy of TA4 sections revealed electron-lucent granules in the plastids of the same cell types as identified by epifluorescence microscopy (Figure 3D–G), and an absence of granules in WT sections (Figure 3B). In TA4 sections, PHB granules were observed in bundle sheath plastids, but not in mesophyll cell plastids (Figure 3D), consistent with the Nile Blue A staining results.
PHB accumulated in culms of plastidic lines at very low levels
During the screening process, cores from basal culm internodes were assayed for PHB concentration by HPLC. Of the 26 plastidic lines positive for PHB, only two (TA4 and B3-5) showed detectable amounts of PHB in the culm, at 0.0033% of culm dry weight, or 16-fold lower than the leaf average (Table 1). To determine which fraction of 9-month-old basal internodes contained the majority of PHB, the rind (outermost millimetre) and pith distal to the epidermis were assayed. The rind PHB concentrations were 16- and 10-fold less than in similarly aged leaves for lines TA4 and B3-5, respectively, and the pith PHB concentrations were 920- and 310-fold less, respectively (Table 2). To investigate whether protein expression was limiting culm PHB production, Western blot analysis was conducted with two PHB-positive lines: TA4 which had detectable levels of PHB in the culm and TA1 which did not. Total soluble protein extracted from rind and pith fractions was probed with PHAA, PHAB or PHAC antisera. This revealed that the concentrations of the proteins were considerably lower in the pith (Figure 4). For TA1 and TA4, the pith-to-leaf concentration ratios for each of the three PHA proteins were determined by densitometry, and these three ratios were then averaged for each line. This revealed that there was no significant difference between TA1 and TA4 for the mean pith-to-leaf concentration ratio (0.44 vs. 0.39, respectively; P = 0.66 by Student's t-test). A similar result was found for the mean pith-to-rind concentration ratio (0.69 for TA1 vs. 0.51 for TA4; P = 0.30 by Student's t-test). To investigate whether there was an anatomical explanation for the difference in PHB concentration between leaf and culm, sections from basal TA4 culm internodes were stained with Nile Blue A and viewed by epifluorescence microscopy (Figure 5). Compared with leaf tissue, substantially fewer fluorescent granules per unit area were observed in the culm (cf. Figures 5H and 3C), consistent with PHB concentration data. Fluorescent granules were visible in epidermal and hypodermal cells (Figure 5D), as well as xylem parenchyma, phloem companion and sclerenchymatous fibre cells of vascular bundles (Figure 5E,F). Vascular bundles in the area corresponding to the rind fraction were more abundant than in the area corresponding to the pith fraction (Figure 5A,H). Many of the vascular bundles in the rind had large sclerenchymatous sheaths containing substantial numbers of large granules (Figure 5F). In contrast, low numbers of granules were observed in the storage parenchyma cells in the pith. These granules were appressed against the cell walls as a result of the presence of large vacuoles (Figure 5E). Low levels of punctate fluorescence were observed in sclerenchymatous cortical cells in the area corresponding to the rind fraction (Figure 5H). No punctuate fluorescence was observed in the photosynthetic, thin-walled cortical cells of the rind (cf. Figure 5D,G), or in control sections from WT culm (Figure 5A–C).
Table 2. Accumulation of polyhydroxybutyrate (PHB) in 9-month-old basal culm internodes of two plastidic lines producing PHB in leaves
Dry weight as PHB (%)
Attempted production in the cytosol and mitochondria
We were able to detect trace amounts of PHB in only two of 95 cytosolic lines, a result consistent with previous work in Arabidopsis (Poirier et al., 1992). However, in the study of Poirier et al. (1992), only two of three R. eutropha PHB biosynthesis enzymes were used: the reductase and the synthase, but not the ketothiolase. Although plants have endogenous ketothiolase activity (Poirier et al., 1992), it was unclear whether PHB accumulation in the Arabidopsis lines described by Poirier et al. (1992) was limited by inadequate ketothiolase activity. However, when we expressed all three R. eutropha PHB biosynthesis genes in the cytosol, very low levels of PHB accumulated (Table 1). Recent evidence has indicated that a decrease in acetyl-CoA levels in the cytosol adversely affects plant development (Fatland et al., 2005). Hence, in this study, PHB-biosynthesis-associated depletion of acetyl-CoA levels in the cytosol may have adversely affected regeneration, resulting in only two lines being recovered with trace levels of PHB. To test this hypothesis, the PHB biosynthesis genes could be placed under the control of a tightly regulated inducible promoter, and the effect of inducing gene expression at different stages of plant development could be observed. We were unable to detect PHB in mitochondrial targeted lines. To our knowledge, this is the first published report of attempted PHB biosynthesis in these organelles. Given that the PHB biosynthesis enzymes were expressed and presumably targeted to mitochondria correctly (Figure 2), it is unclear why the polymer was not produced. Further experimentation is required with mitochondrial lines to determine whether PHB biosynthesis is possible in this organelle. Firstly, confirmation that PHAA, PHAB and PHAC proteins and activities are enriched in mitochondrial fractions would show that the proteins are targeted correctly and are functional following post-targeting processing. Secondly, the supply of pyruvate as respiratory substrate to isolated intact mitochondria might determine whether mitochondrial PHB production is possible in vitro. Thirdly, mitochondrial PHB production may be lethal. To test this hypothesis, expression in transgenic lines of mitochondrial-leader:PHB-biosynthesis-gene fusions could be placed under the control of the same inducible promoter as discussed above for further testing of cytosolic PHB production.
Moderate PHB production in plastid-targeted lines
In plastid-targeted lines, PHB accumulated in leaves to concentrations that would have been deemed to be only moderate in a previously reported population of PHB-accumulating Arabidopsis lines (Bohmert et al., 2000; Table 1). This may in part be due to a marked difference between the species with regard to the diversity of leaf-cell plastids accumulating polymer. In the Arabidopsis lines, PHB granules accumulated in plastids of all leaf cell types (Bohmert et al., 2000), whereas, in the sugarcane lines reported here, PHB granules were present in plastids of bundle sheath cells, but were seemingly absent in plastids of mesophyll cells (Figure 3C,D). Similar results have been reported previously for PHB-accumulating maize lines (Poirier and Gruys, 2002). The low level of PHB in sugarcane and maize mesophyll is apparently not a result of poor promoter activity in this cell type, as the maize Ubi-1 promoter used in this study can drive high-level expression of the Escherichia coliβ-glucuronidase gene in mesophyll cells of transgenic sugarcane (P. Joyce, BSES Limited, Indooroopilly, Qld, Australia, unpublished results) and maize (Rasco-Gaunt et al., 2003). Furthermore, low PHB accumulation in mesophyll has been observed in transgenic maize using other, mesophyll-active promoters (Poirier and Gruys, 2002). A possible reason for the lack of PHB in mesophyll cells in this study, and in that of Poirier and Gruys (2002), is that the dicot Rubisco SSU leader sequences used to target the PHB biosynthesis enzymes may not target to mesophyll cell plastids of C4 monocots. Although Zhong et al. (2003) showed that expression in maize of a pea Rubisco SSU:PHAC fusion resulted in PHAC being associated with the exterior of mesophyll plastids, they did not demonstrate uptake of PHAC by the mesophyll plastids. Further reporter-gene work is required to demonstrate conclusively that dicot Rubisco SSU leader sequences target efficiently to mesophyll plastids in C4 monocots.
PHB accumulation in culms
Attempts to produce PHB in sugarcane culms were largely unsuccessful: yields in rind and pith fractions of plastid-targeted lines were approximately 10-fold and 300-fold lower, respectively, than that in leaves (Table 2). PHB biosynthesis enzyme concentrations were relatively low in the pith (Figure 4), indicating that poor protein expression may be limiting PHB production. However, although the PHB concentration in the rind was 10-fold lower than that in the leaves (Table 2), densitometry revealed that PHB biosynthesis enzyme concentrations in the rind were, on average, only 0.74-fold lower than those in the leaves [standard deviation (SD) = 0.19-fold, n = 6; Figure 4]. This demonstrates that there is no clear relationship between protein expression and PHB concentration in sugarcane culms, and that there are therefore other factors limiting PHB production in culms. To further investigate possible limiting factors, the spatial distribution of PHB granules in TA4 culms was determined (Figure 5). PHB granules were concentrated in vascular cells, but were substantially fewer in storage parenchyma cells distal to the epidermis and in sclerenchymatous cortical cells proximal to the epidermis. Hence, a reason for the PHB concentration being lower in culm internodes than in leaves is that, in culm internodes, the more abundant parenchyma and sclerenchyma cells with a low PHB concentration dilute the PHB concentration in the less abundant vascular cells. In addition, the rind fraction contained many vascular bundles with PHB-rich fibre sheaths. In contrast, not only did the pith distal to the epidermis have a paucity of bundles, but the bundles had no sheaths and were surrounded by storage parenchyma cells with low cytoplasmic density. These observations help to explain the 30-fold difference in PHB concentration between the rind and pith fractions. The majority of PHB granules were found in phloem companion and sclerenchymatous fibre cells of vascular bundles. There is substantial natural variation in the Saccharum genus for culm-internode fibre content (5–45% of culm-internode fresh weight; Bull and Glasziou, 1963). Hence, if increased fibre content is a result of the presence of more vascular bundles and/or bundles with larger fibre sheaths (Oworu et al., 1977), it may be possible to increase the PHB yield in culms by transforming germplasm having a high culm internode fibre content. Interestingly, PHB granules were not visible in the thin-walled cortical cells of the rind (cf. Figure 5D,G), which are C3 photosynthetic (Kortschak and Nickell, 1970). Hence, a commonality exists between plastids of two cell types in sugarcane that apparently do not accumulate PHB (leaf mesophyll and culm-rind cortical) in that they undertake the light reactions of photosynthesis. Therefore, in C4 monocots, PHB production may be impaired in plastids undertaking the light reactions of photosynthesis. Alternatively, the dicot targeting sequences used in this study and in PHB-producing maize plants (Poirier and Gruys, 2002) may not target efficiently to such plastids, and a monocot-derived targeting sequence may be required instead. In addition, it is unknown whether the maize Ubi-1 promoter used in this study is active in the thin-walled cortical cells of the rind.
In summary, we report here the accumulation of PHB to moderate levels in transgenic sugarcane. Polymer accumulated in a cell-type-dependent fashion, and PHB-accumulating lines did not have any obvious deleterious effects compared with control plants. These findings have prompted us to further investigate the spatio-temporal pattern of PHB accumulation in sugarcane, and to evaluate the effect of PHB accumulation on agronomic performance in this species (Purnell et al., 2007).
Plant material and growth conditions
Unless stated otherwise, sugarcane (Saccharum spp. hybrids, cv. Q117) was grown in 50% perlite : 50% vermiculite in a glasshouse at BSES Limited, Indooroopilly, Qld, Australia. Plants were watered every second day with fortnightly application of fertilizer.
Gene constructs and plant transformation
All DNA manipulations were carried out as described by Sambrook et al. (1989). Constructs for targeting PHAA, PHAB and PHAC to plastids (Nawrath et al., 1994) were prepared by polymerase chain reaction (PCR) amplification from pUC18 clones harbouring the phaA (primers TphaF and phaR; primer details in Table S1, see ‘Supplementary material’), phaB (primers TphaF and phbR) or phaC (primers TphaF and phcR) genes translationally fused at the 5′ end to a sequence encoding a Rubisco SSU from Pisum sativum (Fluhr et al., 1986; GenBank accession number X04334). These amplicons were ligated as BamHI/blunt end fragments into the BamHI/SmaI sites of the vector pUSN [a gift from L. Pickering, Commonwealth Scientific and Industrial Research Organization (CSIRO), Brisbane, Australia], a derivative of the vector pAHC27 (Christensen and Quail, 1996), between Zea mays Ubi-1 and the Agrobacterium tumefaciens nopaline synthase terminator (Depicker et al., 1982). Constructs for accumulation in the cytosol were prepared by PCR amplification of the phaA (primers PhaF and PhaR), phaB (primers PhbF and PhbR) and phaC (primers PhcF and PhcR) genes, without the plastid targeting sequence. These amplicons were then ligated into pUSN, as described for the plastid targeting constructs. Constructs for targeting to mitochondria were prepared by excising from the plasmid psB-pma4-35S-βdel-GFP, a 252-bp BglII/BamHI fragment, which encodes the mitochondrial targeting presequence from a Nicotiana plumbaginifolia ATPase β-subunit gene (Boutry and Chua, 1985; GenBank accession number X02868). This fragment was then inserted in the correct orientation into the unique BamHI site of the constructs described for cytosol accumulation. A GFP fusion construct was prepared by replacing the phaA gene of the mitochondrial-leader:PHAA construct with the gene encoding S65T GFP (Chiu et al., 1996) from the plasmid pUbiGFP (Elliott et al., 1999). Biolistic transformation and regeneration of plants were conducted as described in Bower et al. (1996). Embryogenic callus was co-bombarded with the plasmid pEMU-Kn (Last et al., 1991), driving the expression of the Escherichia coli neomycin phosphotransferase gene, nptII, as a selectable marker, together with the three constructs for targeting the PHB pathway to either plastids, cytosol or mitochondria. A single regenerant was sourced per callus piece, and hence the lines produced in this study were deemed to be independent.
Determination of PHB concentration
PHB concentration was assayed by HPLC essentially as described by Karr et al. (1983), with the following modifications. Tissue (50 mg fresh weight) was placed in a 2-mL screw-capped microfuge tube containing two ceramic beads [φ = 1/4 in. (6.5 mm); Bio101 Inc., La Jolla, CA, USA], a pinch of acid-washed sand and 500 µL of concentrated sulphuric acid [36 n (18 m)]. Samples were homogenized for 40 s at top speed in a bead-beater (FastPrep, Bio101 Inc.). To achieve complete digestion of plant tissue, samples were incubated for 8 h at 90 °C (M. P. Purnell, unpublished results). Digested samples were diluted 100-fold prior to HPLC detection of the main reaction product, crotonic acid (trans-2-butenoic acid). Leaf extracts from PHB-producing Arabidopsis plants were used as positive controls (Bohmert et al., 2000). Standards were made from pure PHB (a gift from E. Ulian, Copersucar, Piracicaba, Brazil) digested in a WT leaf background. For cytosolic and mitochondrial lines, 250 mg of lyophilized lamina was extracted with 5 mL CHCl3 at 55 °C for 2 h, reduced to dryness under vacuum, and then digested as above. PHB concentrations were also determined by gas chromatography (GC), essentially as described by Snell et al. (2002), except that volumes were scaled for processing in microfuge tubes, and samples were homogenized in a bead-beater as described above. When identical samples were analysed in parallel by HPLC and GC, the HPLC results were, on average, 8.65% (SD = 1.44%, n = 3) higher than the GC results. HPLC was used routinely for PHB quantification because, compared with GC, it was less labour intensive and simplified sample handling and waste disposal.
Western blot analysis
Total soluble protein was isolated using Tri-Reagent (Sigma, Castle Hill, NSW, Australia) following the manufacturer's instructions. Following quantification by the Lowry method (DC Protein Assay Kit, Biorad, Regents Park, NSW, Australia), 10–20 µg of total soluble protein was separated on 4–12% polyacrylamide gradient gels (NuPage, Invitrogen, Mulgrave, Vic., Australia) and transferred subsequently on to polyvinylidene difluoride (PVDF) membrane (Hybond P, Amersham, Baulkham Hill, Vic., Australia). Total soluble protein extracted from leaves of PHB-producing Arabidopsis plants (Bohmert et al., 2000) and purified PHAC protein (a gift from A. Sinskey, MIT Department of Biology, Cambridge, MA, USA) were co-electrophoresed as positive controls. Rabbit antisera specific to PHAA, PHAB and PHAC (Gerngross et al., 1993; Nawrath et al., 1994) were used at dilutions of 1 : 10 000, 1 : 2000 and 1 : 5000, respectively. Detection was performed using goat anti-rabbit antiserum conjugated to horseradish peroxidase (Biorad; dilution, 1 : 2000) and the ECL Plus Kit (Amersham), in accordance with the manufacturers’ instructions. Non-saturated signals were generated by fluorescence scanning at 560 nm (Storm 840 phosphoimager, Molecular Dynamics, Amersham), and quantified using ImageQuant software (Amersham).
Epifluorescence microscopy of stained mitochondria
Transgenic sugarcane lines expressing GFP translationally fused to the mitochondrial targeting presequence were grown hydroponically as described by Purnell et al. (2005). Young, healthy roots with several lateral root primordia were isolated and placed for 10 min in a 1 mm solution of the mitochondrial stain, Mitotracker Red CMXRos (Invitrogen). After a 15-min wash step in fresh hydroponic solution, root tips were squashed on to microscope slides under coverslips. Sections were viewed with a compound fluorescence microscope (Leitz DM-RBE, Leica Microsystems Australia, Gladesville, NSW, Australia). Mitotracker Red fluorescence was visualized using the following filter set: exciter BP515/560, beam splitter 580, emitter LP590 (N2.1, Leica). GFP fluorescence was visualized using the following filter set: exciter BP480/40, beam splitter 505, emitter BP527/30 (L5, Leica).
Epifluorescence microscopy of stained PHA granules
Fresh plant material was sectioned by hand and stained for 5 min with a 1% w/v solution of Nile Blue A (Sigma) in tap water for leaf sections and in buffered osmoticum [0.30 m d-mannitol, 0.30 m d-sorbitol, 3 mm 2-(N-morpholino)ethanesulphonic acid (MES) and 0.7 mm NaH2PO4 adjusted to pH 5.9 with 1 m NaOH] for culm sections. Sections were destained for 10 min in the same respective solutions, and then mounted on glass slides in the same respective solutions and coverslipped. Stained sections were visualized with a compound fluorescence microscope (Olympus BX50, Olympus Australia, Windsor, Qld, Australia) using the following filter set: exciter HQ545/30, beam splitter d590/20m, emitter q570lp (Chroma Technology, Brattleboro, VT, USA). For the visualization of chlorophyll autofluorescence, sections were examined using the following filter set: exciter d480/30, beam splitter 505dclp, emitter 660/50 (Chroma Technology).
Electron microscopy of stained PHB granules
Transmission electron microscopy to detect PHB granules was conducted by the Analytical Electron Microscopy Facility, Queensland University of Technology, Qld, Australia. Samples were prepared according to Bohmert et al. (2000), except that leaves were fixed in 3% glutaraldehyde and a JEOL 1200 EX electron microscope (JEOL, Tokyo, Japan) was used.
We thank Palmina Bonaventura, Peter Abeydeera, Niall Masel and Patricia Lindeman for technical assistance, and Peter Allsopp, David Anderson, Annathurai Gnanasambandam and Scott Hermann for critical reading of the manuscript. We thank Yves Poirier (University of Lausanne, Switzerland) for supplying the pUC18 clones containing phaA, phaB and phaC and for PHAA and PHAB antisera. We thank Anthony Sinskey (MIT Department of Biology, Cambridge, MA, USA) for PHAC antiserum and PHAC protein, and Karen Bohmert (Metabolix, Boston, MA, USA) for seeds from PHB-producing Arabidopsis plants. Purified PHB was kindly supplied by Eugenio Ulian, Copersucar, Piracicaba, Brazil. This research was supported by an Australian Research Council Linkage Grant, LP0210658, granted to L.K.N.
L.A.P. made the plastidic and cytosolic constructs, generated and analysed the plastidic and cytosolic transgenic lines, generated the data in Tables 1 and S1 and Figures 2A, 3B, D–G and 4, and prepared Table S1 and Figures 1 and 3. M.P.P. made the mitochondrial and GFP constructs, generated and analysed the plastidic, cytosolic and mitochondrial lines, generated the data in Tables 1 and 2 and Figures 1, 2B, C, 3A, C and 5, prepared all the tables and figures, and drafted the paper. S.M.B. generated the data in Figure 3B, D–G. L.K.N and S.M.B. jointly supervised the work. All authors discussed the results and commented on the manuscript.