We report here the results from a glasshouse trial of several transgenic sugarcane (Saccharum spp. hybrids) lines accumulating the bacterial polyester polyhydroxybutyrate (PHB) in plastids. The aims of the trial were to characterize the spatio-temporal pattern of PHB accumulation at a whole-plant level, to identify factors limiting PHB production and to determine whether agronomic performance was affected adversely by PHB accumulation. Statistical analysis showed that a vertical PHB concentration gradient existed throughout the plant, the polymer concentration being lowest in the youngest leaves and increasing with leaf age. In addition, there was a horizontal gradient along the length of a leaf, with the PHB concentration increasing from the youngest part of the leaf (the base) to the oldest (the tip). The rank order of the lines did not change over time. Moreover, there was a uniform spatio-temporal pattern of relative PHB accumulation among the lines, despite the fact that they showed marked differences in absolute PHB concentration. Molecular analysis revealed that the expression of the transgenes encoding the PHB biosynthesis enzymes was apparently coordinated, and that there were good correlations between PHB concentration and the abundance of the PHB biosynthesis enzymes. The maximum recorded PHB concentration, 1.77% of leaf dry weight, did not confer an agronomic penalty. The plant height, total aerial biomass and culm-internode sugar content were not affected relative to controls. Although moderate PHB concentrations were achieved in leaves, the maximum total-plant PHB yield was only 0.79% (11.9 g PHB in 1.51 kg dry weight). We combine the insights from our statistical and molecular analyses to discuss possible strategies for increasing the yield of PHB in sugarcane.
In recent years, the production of polyhydroxyalkanoates (PHAs) in plants has been the focus of major research efforts, as these biodegradable polyesters are considered as possible alternatives to polymers currently synthesized from non-renewable resources (Poirier and Gruys, 2002; Snell and Peoples, 2002). The simplest and most studied of the PHAs is polyhydroxybutyrate (PHB). This polyester is naturally produced in many bacterial species, including Ralstonia eutropha, where it serves as a source of carbon and reducing power. In this organism, acetyl-coenzyme A (acetyl-CoA) is converted into the polymer 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).
Previously, we have reported the generation of transgenic sugarcane lines accumulating PHB in plastids to 1.88% of leaf dry weight without any obvious deleterious effects (Petrasovits et al., 2007), in contrast with reports in other species (Lössl et al., 2003). The aims of the current study were to characterize the spatio-temporal pattern of PHB accumulation in transgenic sugarcane lines at a whole-plant level, to identify factors limiting the production of PHB in sugarcane, and to determine objectively whether PHB production in sugarcane adversely affects agronomic performance.
PHB accumulated to moderate levels in leaves during a 9-month glasshouse trial
To study the spatio-temporal accumulation of PHB in sugarcane, six PHB-positive lines, with a range of PHB concentrations from 0.026% (line B1-15) to 0.161% (line TA6) of leaf dry weight as PHB (Petrasovits et al., 2007), were subjected to a 9-month glasshouse trial with harvests every 3 months. Figure 1 lists the lines placed in the trial, and illustrates the design of the trial and the tissues that were collected at each harvest. At 3 months, the PHB concentration was measured in the roots and primary culm. PHB was not detected in the roots or culm, and these tissues were not analysed for the remainder of the trial. At 3, 6 and 9 months, the PHB concentration was measured in the leaves of the primary culm. The rank order of the lines with regard to average leaf PHB concentration remained the same throughout the trial, with a high-producing group and a low-producing group (cf. Figure 1 inset with Figure 2A–C). Statistical analysis showed no significant line-by-time interaction for leaves 1 (mature), 8 (intermediate) or 12 (young) (P-values > 0.91), indicating that the rate of change in leaf PHB concentration was similar for all lines, despite substantial differences in absolute leaf PHB concentration. The data were then analysed, irrespective of line, to assess whether there were differences in PHB concentration from harvest to harvest. There was a significant increase in PHB concentration from 3 to 6 months, and from 6 to 9 months, for leaves 1, 8 and 12 (P-values ≤ 0.003). For leaves 1, 8 and 12, the rate of change in PHB concentration reached a maximum between 3 and 6 months, and then decreased between 6 and 9 months (Figure 2A–C). This decrease was probably a result of the onset of leaf senescence because, from 6 to 9 months, the percentage dry weight for all three leaves increased substantially (results not shown). The maximum average leaf PHB concentration measured during the trial was 1.39% of dry weight in leaf 8 of line TA2 at 9 months (Figure 2B).
PHB concentration in leaves of the same physiological age was substantially different across harvests
The presence of dewlaps on sugarcane leaves permits a comparison of leaves of approximately the same physiological age at different harvest times (van Dillewijn, 1952). For three of the six lines, we measured PHB concentration in leaves of the same physiological age harvested at 3, 6 and 9 months (Figure 2D–F). There was no difference in the PHB concentration in leaves harvested at 6 and 9 months. However, polymer concentrations in leaves harvested at 3 months were unexpectedly only half that of leaves harvested at 6 and 9 months.
A substantial increase in PHB yield over time was associated with an exponential increase in leaf biomass
Despite the general decrease in the rate of change in PHB concentration from 6 to 9 months for primary culm leaves 1, 8 and 12 (Figure 2A–C), there was an apparent substantial increase in total-plant PHB content from harvest to harvest (Figure 3; open symbols connected by full lines). There was no significant line-by-time interaction (P = 0.999), indicating that the total-plant PHB accumulation profiles among the lines were similar. When the data were analysed irrespective of line, there was a significant increase in total-plant PHB content from 3 to 6 months, and from 6 to 9 months (P < 0.0001). The amount of PHB on a per plant basis increased, on average, 14.9-fold from 3 to 6 months, and 5.3-fold from 6 to 9 months. This substantial increase in PHB yield over time was apparently associated with an exponential increase in leaf biomass (Figure 3; filled squares joined by broken lines).
A uniform spatio-temporal pattern of PHB accumulation in sugarcane
A comparison of Figure 2D–F, representing PHB concentration in mature, intermediate and young leaves, respectively, suggested that a concentration gradient existed from the base of the plant to the tip. Further statistical analysis confirmed that, in general, there was more PHB in older leaves than in younger leaves (Table 1). In addition to this vertical concentration gradient, there was a horizontal gradient in individual leaves. PHB concentration at the leaf tip was, on average, 2.4-fold higher than that at the leaf base [standard deviation (SD) = 1.2-fold, n = 22] (Table 2). The maximum PHB concentration measured during the trial was 1.77% of leaf dry weight for line TA6 at the tip of leaf 12 at 9 months (Table 2). When data from Figure 2 and Table 1 for two high-producing lines (TA1 and TA6) and one low-producing line (B2-16) were combined and normalized to the mean PHB concentration of leaf 8 at 9 months, a uniform spatio-temporal accumulation pattern emerged (Figure 4). This result indicates that PHB accumulation patterns were similar in lines producing very different amounts of polymer.
Table 1. Polyhydroxybutyrate (PHB) concentration gradient along the vertical axis of transgenic sugarcane lines
PHB concentration was determined for three PHB-positive lines at 3-month intervals. Means (n = 3) significantly different at *P < 0.05 and ***P < 0.001. Means followed by different letters were significantly different. Refer to Figure 1 for an explanation of leaf numbers.
Table 2. Polyhydroxybutyrate (PHB) concentration gradient along the leaf length
Means (n = 3) significantly different at *P < 0.05, **P < 0.01 and ***P < 0.001, respectively. Means marked with different letters were significantly different. Refer to Figure 1 for an explanation of leaf numbers.
Transcript levels of the phaA, phaB and phaC genes were highly correlated
Reverse transcription and subsequent real-time reverse transcriptase–polymerase chain reaction (RT-PCR) quantification of the transcripts of the phaA, phaB and phaC genes were undertaken with RNA purified from young leaves of 6-month-old PHB-producing lines. The relative amounts of phaA, phaB and phaC in the six PHB-positive lines are shown in Figure 5. The amount of phaA transcript correlated well with the amount of phaB and phaC transcript (r = 0.87, P < 0.0001; r = 0.86, P < 0.0001, respectively), and the amount of phaB transcript correlated well with the amount of phaC transcript (r = 0.84, P < 0.0001). These results indicate that expression of the transgenes was apparently coordinated. Coordinated transgene expression has been reported previously in transgenic tobacco (Dean et al., 1988; van Engelen et al., 1994; Mlynárováet al., 2002).
The abundance of PHAA, PHAB and PHAC correlated well with PHB concentration
Western blot analysis was conducted to investigate whether the amount of one or more of the PHB biosynthesis enzymes was limiting for PHB production in leaves. Total soluble protein isolated from the same young leaf extracts as used for real-time RT-PCR analysis was probed with antisera specific to PHAA, PHAB and PHAC. The relative amounts of PHAA, PHAB and PHAC in the six PHB-positive lines are shown in Figure 5. The amount of PHAA correlated well with the amount of PHAB and PHAC (r = 0.88, P < 0.0001; r = 0.81, P < 0.0001, respectively), and the amount of PHAB correlated well with the amount of PHAC (r = 0.85, P < 0.0001). These correlations are not unexpected given that the expression of the transgenes encoding these proteins was apparently coordinated. However, although there were good correlations among the PHB biosynthesis enzymes, and, among the transcripts of these enzymes, there were relatively poor correlations between each enzyme and its respective transcript (PHAA : phaA r = 0.59, P = 0.0096; PHAB : phaB r = 0.43, P = 0.072; PHAC : phaC r = 0.53, P = 0.024; Figure 5). Mlynárováet al. (2002) described a similar situation, and suggested that systematic errors in two different analytical methods (in this case, real-time RT-PCR vs. Western blot analysis) can mask correlations between data sets generated by the respective methods.
When the amount of each protein was plotted against PHB concentration, reasonable and significant correlations were observed for each protein (PHAA r = 0.78, P = 0.0001; PHAB r = 0.69, P = 0.0016; PHAC r = 0.74, P = 0.0005). However, because of the collinear relationship among PHAA, PHAB and PHAC, the data were analysed further. The combined variation among the three proteins explained 65.7% of the variation in PHB concentration (multiple correlation coefficient, P = 0.001). Individually, the proteins accounted for 61.4%, 47.3% and 54.7% of the variation in PHB levels, respectively. When the correlations between each of the proteins and PHB were corrected for the correlations among the three proteins, only the PHAA correlation was near significant [partial correlation coefficient (rp) = 0.470, P = 0.0659; PHAB rp = −0.176, P = 0.515; PHAC rp = 0.335, P = 0.205]. This result indicates that PHB production in young sugarcane leaves may have been limited by the PHAA concentration.
Differences in agronomic performance were independent of PHB accumulation
To assess whether PHB-accumulating sugarcane incurred any growth penalty, the height of the primary culm and the total aerial biomass were measured every 3 months. In addition, sugar concentrations of culm internodes were determined at 3 and 9 months (Figure 6). Significant differences (P < 0.05) among PHB-producing lines and control lines for these variables were independent of PHB accumulation, demonstrating that PHB accumulation to 1.4% of leaf dry weight did not cause an agronomic penalty. Between 6 and 9 months, the primary culm stopped elongating, but biomass accumulation in the same period was linear, because of profuse tillering. An analysis of aerial biomass components revealed that significant differences (P < 0.05) among lines for leaf-to-culm biomass ratios and the abundance of tillering were also independent of PHB accumulation (Table 3).
Table 3. Plant biomass accumulation for several polyhydroxybutyrate (PHB)-positive, green fluorescent protein (GFP)-positive and wild-type (WT) lines
Biomass was separated into leaf and culm at 3 and 6 months, and primary culm leaf and primary culm, and tiller leaf and tiller culm at 9 months. Means (n = 3) significantly different at *P < 0.05, **P < 0.01 and ***P < 0.001, respectively. Means marked with different letters were significantly different. Total aerial biomass at each time point is plotted in Figure 6.
Spatio-temporal pattern of PHB accumulation in sugarcane
We propagated six transgenic sugarcane lines accumulating PHB in plastids (Petrasovits et al., 2007) in a 9-month glasshouse trial, with harvests every 3 months (Figure 1). We applied robust statistical analyses to determine the pattern of accumulation in PHB-producing sugarcane. We found a vertical gradient throughout the plant, the PHB concentration being lowest in the youngest leaves and increasing with leaf age (Figure 2 and Table 1). In addition, there was a horizontal gradient along the length of a leaf, with the PHB concentration increasing from the youngest part of the leaf (the base) to the oldest (the tip) (Table 2). These gradients are consistent with PHB being a non-metabolized carbon sink (Snell and Peoples, 2002). Two interesting findings from our analyses were that the rank order of PHB production among lines did not change over time (cf. Figure 1 inset and Figure 2A–C), and, moreover, there was a uniform spatio-temporal pattern of PHB accumulation among the lines, despite the large differences in absolute PHB concentration (Figure 4). These results provide the rationale for an early screening programme. In our earlier work, PHB was detected in only 20% of sugarcane lines transformed with the PHB biosynthesis genes targeted to plastids, and the mean polymer concentration of the PHB-positive transgenic population was low (0.053% of leaf dry weight as PHB; SD = 0.064% of leaf dry weight as PHB; Petrasovits et al., 2007). Hence, we expect that our current transformation strategy for PHB accumulation in plastids will generate high-producing lines at a very low frequency, requiring many lines to be screened to identify the highest producing line. Our results indicate that an early screening programme could help to identify the highest producing line: early in development, high-producing lines could be selected, whereas low-producing lines could be discarded before they consume further resources. Preliminary results from our laboratory indicate that early screening is possible at the tissue culture stage (L. Petrasovits, unpublished results). Another interesting finding from our analyses, and one requiring verification in a field trial, was that, at 3 months, the PHB concentration in mature, intermediate and young leaves was only half that of leaves of the same physiological age harvested at 6 and 9 months. It is unclear whether the factors apparently limiting PHB production in the 0–3-month interval were developmental or environmental, or both. A possible developmental factor is that PHB accumulation in leaves was limited following vegetative propagation via culm cuttings, because considerable resources were used by the growing lateral bud to establish a root system and culm.
Strategies for increasing total-plant PHB yields in sugarcane
The total-plant PHB content for the highest yielding line of this study, TA2, at 9 months was only 0.79% (11.9 g of PHB per total aerial dry weight of 1.51 kg; Figure 3 and Table 3). It has been estimated that commercially viable PHB production in transgenic crops would require 15% of total aerial dry weight to be PHB, without a biomass yield penalty (Bohlmann, 2006). Our results indicate two strategies for potentially increasing total-plant PHB yields. First, expression of the PHB biosynthesis enzymes should be increased, as there was a good correlation between the amount of these enzymes and PHB concentration (Figure 5). It follows that, for this strategy to succeed, the negative effects of PHAA expression must be successfully mitigated (Bohmert et al., 2002; Kourtz et al., 2005; Lössl et al., 2005; Ruiz and Daniell, 2005). Second, the proportion of the plant that is leaf should perhaps be increased when selecting sugarcane germplasm for PHB production, as a large proportion of plant biomass was culm (Table 1), and PHB concentration in this tissue was very low (Petrasovits et al., 2007). Moreover, the substantial increase over time in total-plant PHB yield was associated with an exponential increase over time in leaf biomass (Figure 3). Petrasovits et al. (2007) have suggested two additional strategies to increase whole-plant PHB yield: achieving PHB production in mesophyll cells and increasing culm fibre content. Of the strategies listed above and by Petrasovits et al. (2007), those most likely to boost PHB yields are increasing the expression of the PHB biosynthesis enzymes, increasing the proportion of the plant that is leaf, and producing PHB in mesophyll. In addition, if it is not possible to select germplasm with a substantially higher leaf-to-culm biomass ratio, high-level PHB production in the culm will be necessary. We believe that high-level PHB production in the culm will require problems with mitochondrial PHB production (Petrasovits et al., 2007) to be overcome.
Absence of an observable agronomic penalty in PHB-producing sugarcane
The production of PHB in planta has been associated with growth penalties. Transgenic tobacco lines accumulating PHB to 0.10% of leaf dry weight appeared chlorotic and were male sterile (Lössl et al., 2003). In contrast, the transgenic sugarcane reported here, accumulating PHB to 1.4% of leaf dry weight, did not suffer any observable agronomic penalty. Differences among PHB-accumulating lines and control lines with regard to primary culm height, total aerial biomass and culm-internode sucrose concentration were independent of PHB accumulation (Figure 6). In addition, metabolomic analysis of young leaves from three PHB-producing lines did not reveal any major effects from PHB accumulation to 0.14% of leaf dry weight (results not shown; analysis by O. Fiehn, MetaProfile, Berlin, Germany). Another line too immature for the glasshouse trial subsequently accumulated PHB to approximately 2.5% of leaf dry weight without observable adverse effects (line TA4; Petrasovits et al., 2007). Although these collective results are promising, the true utility of sugarcane as a PHB production platform will only be ascertained by analysing lines with greater leaf PHB concentrations, namely ≥ 3% of leaf dry weight, which was the minimum amount of polymer associated with growth retardation in Arabidopsis (Bohmert et al., 2000).
Plant material and growth conditions for the glasshouse trial
Sugarcane (Saccharum spp. hybrids, cv. Q117) culm cuttings were sourced from six transgenic lines with a range of leaf PHB concentrations (Petrasovits et al., 2007). These lines were generated from separate biolistic transformation experiments, and hence were independent transformants. The culm cuttings from these lines, together with culm cuttings from three tissue-cultured, wild-type clones and three green fluorescent protein (GFP)-positive lines, were planted in 20-L pots containing premium garden soil. On 31 January 2003, the pots were arrayed in a glasshouse at Woodford, Qld, Australia, in a randomized complete-block design with three harvests of clonal siblings at 3 (24 April 2003), 6 (25 July 2003) and 9 months (10 October 2003). Plants were watered automatically twice daily and fertilizer (Peter's Professional; Scotts, Baulkam Hills, Vic., Australia) was applied weekly to run-off through the watering system at a rate of 0.9 g/pot. Biomass and primary culm height were measured on site. For the determination of culm internode sugar concentration, internodes were collected and stored at −20 °C until processing. For the determination of PHB concentration, leaves were collected and dried at 70 °C prior to analysis. A schematic diagram of the tissues harvested is shown in Figure 1.
Determination of PHB concentration
PHB concentration was determined by high-performance liquid chromatography (HPLC) as described previously (Petrasovits et al., 2007).
Total RNA was extracted from young leaves of 6-month-old plants using Tri-Reagent (Sigma, Castle Hill, NSW, Australia) following the manufacturer's instructions, purified on a silicon matrix (RNeasy, Qiagen, Clifton Hill, Vic., Australia), and treated with 25 U of DNase on-column (Qiagen). Following elution, the RNA was quantified in a real-time PCR instrument (Rotor-gene 3000a, Corbett Research, Sydney, NSW, Australia) using an RNA-specific fluorophore (Picogreen, Invitrogen, Mulgrave, Vic., Australia), and the RNA integrity was checked by agarose gel electrophoresis. RNA concentrations were adjusted equally and 100 ng was reverse transcribed (Superscript III, Invitrogen) at 55 °C according to the manufacturer's instructions. PCR was conducted with 1 × Sybrgreen premix (Invitrogen) using primers specific to phaA (primers phaA-F4 and phaA-R10; annealing temperature, 57 °C; amplicon 164 bp), phaB (phaB-F11 and phaB-R11; 57 °C; 154 bp) or phaC (phaC-F10 and phaC-R10; 62 °C; 130 bp), or to S. officinarum 18S rRNA (So18S-F10 and So18S-R10; 62 °C; 110 bp), which was used as a normalizer (primer details are listed in Table S1, see ‘Supplementary material’). All primers were used at a final concentration of 200 nm, except for phaA-R10, which was used at 270 nm. Template was diluted 1 : 20 for phaA, phaB and phaC reactions, and 1 : 1000 (achieved via three 10-fold serial dilutions) for S. officinarum 18S rRNA reactions, and 5 µL was added to a total reaction volume of 15 µL. All template dilutions and reaction set-ups were conducted by a liquid handling robot (CAS-1200, Corbett Robotics, Brisbane, Australia). No-template controls were run for each set of reactions and no-reverse-transcription controls were run for randomly chosen samples. Reactions were conducted with three analytical replicates in a real-time PCR instrument (Rotor-gene 3000a, Corbett Research), and the data were analysed using software provided by the manufacturer. Standard curves were run for each set of reactions, and phaA, phaB and phaC data were normalized against 18S data using the two-standard-curves method (R2 > 0.95 for all standard curves). Average reaction efficiencies for each primer set were 1.64, 1.62, 1.80 and 1.12 for phaA, phaB, phaC and S. officinarum 18S rRNA, respectively. All amplicons were confirmed by agarose gel electrophoresis to be the expected size and free of primer dimers (results not shown).
Western blot analysis
Total soluble protein was extracted in parallel from the same extracts as used for real-time PCR and analysed as described previously (Petrasovits et al., 2007). Analytical replicates were generated by electrophoresing each protein sample on three separate gels. Non-saturated fluorescent signals were quantified using the area quantification method of ImageQuant software (Amersham, Baulkham Hills, Vic., Australia) with local background correction.
Determination of sugar concentration
Frozen culm internodes were homogenized in a blender. Homogenate (0.5 g) was extracted with 10 volumes of distilled water at 80 °C for 3 h, followed by repeat extraction with a further 10 volumes of distilled water for 6 h. Samples were then centrifuged and the supernatant was collected. Prior to HPLC analysis, sodium azide was added to a final concentration of 0.02%, and the sample was then passed through a 0.2-µm filter. Filtrates (20 µL) were injected on to a 300 × 8 mm Shodex SDVB gel column (Waters, Rydalmere, NSW, Australia) at 65 °C, and eluted with ultrapure water containing one drop of 1 n (1 m) NaOH per litre at a flow rate of 0.9 mL/min. Standards were known concentrations of sucrose, glucose and fructose in water, with sodium azide added to a final concentration of 0.02%.
Unless stated otherwise, all analyses were based on ranks, because data were not normally distributed and the variances for each treatment were not homogeneous. A Friedman two-way analysis of variance (anova) was undertaken, which is the non-parametric method for the analysis of a randomized complete-block design (Conover, 1999). Where significant differences were observed (P < 0.05), multiple comparisons were made using the least significant difference (LSD) method (Fisher, 1990), or Tukey's method with α = 0.05 (Tukey, 1974) if the number of comparisons was greater than three.
We thank Palmina Bonaventura and Peter Abeydeera for technical assistance, Jo Stringer and Del Greenway for help with the statistical analyses, 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 PHAA and PHAB antisera and Anthony Sinskey (MIT Department of Biology, Cambridge, MA, USA) for PHAC antiserum. This research was supported by an Australian Research Council Linkage Grant, LP0210658, to L.K.N.
M.P.P. designed the glasshouse trial, conducted real-time PCR, generated the data (except Western blot data) in all tables and figures and conducted statistical analysis of the data in all tables and figures, prepared all tables and figures, and drafted the paper. L.A.P. conducted Western blots, generated the data (except real-time PCR data) in all the tables and figures, and prepared Table S1. L.K.N. conducted statistical analysis of the data in Figure 5. L.K.N. and S.M.B. jointly supervised the work. All authors discussed the results and commented on the manuscript.