The production of edible as well as some nonedible oils in agricultural crops is of great economic importance (Murphy, 1994). The commercial interest has been enhanced in recent years because of the development of transgenic oil crops (Harwood, 1998). Such crops have been produced in order to provide speciality oils for industrial purposes (Murphy, 1994) as well as to modify traditional crops so that their storage products have more desirable properties (Ohlrogge, 1994; Voelker & Kinney, 2001).
Of the main oil crops in the world today, the oilpalm is the most important plant grown exclusively for oil. Furthermore, oilpalm is extremely productive, producing c. 5–10 t oil ha−1 compared with a yield of 1–3 t oil ha−1 for typical oilseeds (Gunstone et al., 2007). The oilpalm is a tropical crop and is of great importance for Indonesia and, particularly, for Malaysia which exports over 90% of its palm oil products (Ahmad, 2003). It is also the world's major source of edible oil (Gunstone, 2006).
Despite the economic importance of oil crops, as well as the great advances in our knowledge of their basic biochemistry and molecular biology (Harwood, 1996; Ohlrogge & Jaworski, 1997; Murphy, 2005), we still know relatively little about the regulation and/or control of lipid synthesis and accumulation (Ohlrogge & Jaworski, 1997). In fact, it has been remarked that efforts to change metabolism through the manipulation of single genes, often thought, misguidedly, to encode ‘rate-limiting’ steps, has produced disappointing results (Stitt & Sonnewald, 1995). Nevertheless, within plant lipid metabolism there have been a few experiments that provide some information about regulation and control. For example, the importance of acetyl-CoA carboxylase in controlling lipid synthesis in leaves was indicated by the levels of malonyl-CoA and other acyl-thioesters (Post-Beittenmiller et al., 1992; Ohlrogge & Jaworski, 1997) and confirmed by direct measurement of its flux control coefficient (Page et al., 1994). Increasing the activity of this enzyme by genetically transforming potato can also lead to enhanced fatty acid production, as would be predicted (Klaus et al., 2004). In developing oil crops, the relative importance of fatty acid synthesis compared with lipid accumulation has been highlighted by the effect of exogenous fatty acids on lipid synthesis in embryos of Cuphea (Bafor et al., 1990) and other crops (Bao & Ohlrogge, 1999). Moreover, stable isotope methods have been applied to the study of lipid metabolic pathways in oil seeds (Pollard & Ohlrogge, 1999) and such techniques can provide additional quantitative information about fluxes through alternative pathways of central carbon metabolism (see Ratcliffe & Shacher-Hill, 2001) including the sources of acetyl-CoA for fatty acid formation (Schwender & Ohlrogge, 2002).
In further efforts to pinpoint particular enzyme steps that may be important for regulating lipid biosynthesis, two genetic approaches have been used. First, quantitative trait loci (QTLs) that control seed oil and fatty acid composition have been examined in several important crops such as Brassica napus (Burns et al., 2003) or soybean (Csanadi et al., 2001) as well as in the experimental model plant Arabidopsis (Hobbs et al., 2004). Second, cDNA microarrays which compare wild-type with the low-lipid wrinkled 1 mutation of Arabidopsis have examined over 3500 genes in order to discover which may exert significant control (Ruuska et al., 2002).
Because of the paucity of quantitative data that would allow a more precise description and, hence, a better understanding of the regulation and control of lipid synthesis, we decided to use the technique of metabolic control analysis (MCA). With this method, it is possible to obtain a quantitative measure of the relative importance of particular parts of a metabolic pathway in controlling metabolic fluxes or metabolic intermediate concentrations under defined conditions (Fell, 1997). It is important to realize that MCA recognizes that for an enzyme, or block of reactions, even to approach being rate-limiting is extremely rare, and that all steps/parts of a metabolic pathway can contribute to control of flux, with their contributions often altering as (physiological) conditions change (Kacser & Burns, 1973; Heinrich & Rapoport, 1974). There are two important variants of control analysis, top-down (or modular) and bottom-up. In top-down control analysis (TDCA) a metabolic pathway is conceptually simplified by being divided into blocks with a unique chosen intermediate connecting the blocks of reactions. The advantage of this approach is that it provides an immediate overview of the distribution of flux control over a complex metabolic pathway (Quant, 1993) and can then be refined to allow further detailed analysis of each block. In earlier papers (Ramli et al., 2002a,b), we applied this method for the first time to lipid synthesis, performing a single-manipulation TDCA to gain an initial, broad overview of the control over total lipid biosynthesis fluxes in oilpalm and olive tissues. To achieve this, we conceptually divided the biosynthetic pathway (our analytical system) into two groups, or blocks, of reactions (those of fatty acid synthesis in the plastid (Block A) or of complex lipid assembly in the endoplasmic reticulum (Block B)) linked by a unique intermediate, cytosolic acyl-CoA. Oleate was then introduced to the experimental system to manipulate the level of the common intermediate directly and the responses of the system variables (the fluxes (JA; JB) through the blocks and the acyl-CoA levels (X)) were measured empirically.
In this paper we have applied a different method of analysis, that of double-manipulation TDCA. With this technique, inhibitors specific to each of the Blocks (of reactions) were used to manipulate the system fluxes independently. This is (to our knowledge) the first time such a method has been applied to lipid biosynthesis in any tissue. The results from the double-manipulation TDCA were found to be in good agreement with those found previously for single manipulation (Ramli et al., 2002b) and with a more thorough analysis of the latter using Monte Carlo simulations (MCS) and reported in this paper. These data show that more control is exerted by the fatty acid biosynthesis group of reactions (Block A) but that, nevertheless, substantial control resides with the complex lipid assembly group of reactions. Taken together, these data provide new knowledge and quantitative insight into the way in which lipid synthesis is regulated in plants and provide specific information about control of pathways in the world's major oil crop.
Materials and Methods
Oil palm (Elaeis guineensis Jacq.) seeds were obtained and callus cultures established and maintained as described previously (Ramli et al., 2002a). Cultures were subcultured routinely every 4 wk and were used for experiments 21–24 d after subculturing. As discussed previously (Ramli et al., 2002a) such calli accumulate large amounts of triacylglycerol, which is enriched in palmitate and oleate. Therefore, they are a good model system for palm fruit endosperm lipid accumulation.
Labelled Na[1-14C]acetate (sp. activity 2.11 GBq mmol−1), [U-14C]glycerol (5.51 GBq mmol−1), [U-14C]glycerol-3-phosphate (3.7 GBq mmol−1) and 1,2-di[1-14C]oleoyl-phosphatidylcholine (3.7 GBq mmol−1) were purchased from Amersham International. [1-14C]Dioleoyl-glycerol was generated from [1-14C]dioleoyl-phosphatidylcholine by phospholipase C (Bacillus cereus) digestion and separating the products by TLC. Sep-Pak C18 cartridges were purchased from Waters (Milford, MA, USA). 2-Bromooctanoate and all other chemicals and solvents used in lipid extraction and analysis were from Sigma or from BDH (Poole, Dorset, UK) and were of the best available analytical grades. Diflufenican (DFF) was a gift from Dr D. J. Cole (Rhone-Poulenc Agriculture (now Bayer Crop Science) Ongar, Essex, UK).
Oil palm calli (approx. 0.5 g FW were incubated with [1-14C]acetate (74 kBq) or (U-14C]glycerol (37 kBq) in the presence or absence of 0.02 mm bromooctanoate or 0.01 mm DFF at 30°C for 6 h. Labelling of lipids was linear within this period (Ramli et al., 2002a). Total lipids were extracted and the acyl-CoA pool analysed by separating water-soluble products through Sep-Pak C18 cartridges (Ramli et al., 2002a).
Lipid classes were separated and analysed by thin-layer chromatography (TLC) and fatty acids by gas–liquid chromatography (GLC) of their methyl esters, as described previously (Ramli et al., 2002a,b).
The microsomal fraction was prepared from calli by differential centrifugation at 4°C. Homogenates were made in a buffer of 50 mm Hepes (pH 7.2), 330 mm sorbitol, 1 mm MgCl2, 3 mm EDTA, 5 mmβ-mercaptoethanol, 0.1% BSA, 0.2% ascorbate and 1% polyvinylpyrollidine using a domestic blender and a buffer : tissue ratio (ml g−1) of 10. Following centrifugation at 5000 g for 10 min, 18 000 g for 20 min and 105 000 g for 60 min, the microsomal pellet was resuspended in 50 mm Hepes (pH 7.2), 330 mm sorbitol, 1 mm DTT using a precooled glass homogenizer. Incubations were carried out using up to 50 µg microsomal protein ml−1 of assay medium.
Microsomal incubations were carried out with [U-14C]glycerol 3-phosphate in 1 ml total volume at 30°C unless otherwise stated. Final concentrations of constituents were 35 mm Hepes-NaOH (pH 7.2), 300 mm sorbitol, 0.5% BSA, 200 µm glycerol-3-phosphate (containing 0.1 µCi [U-14C]glycerol-3-phosphate), 75 µm palmitoyl-CoA, 100 µm oleoyl-CoA and up to 100 µg microsomal protein. Incubations were for 30 min after which lipid extraction and analysis were as previously. The above conditions were optimal and the reaction was linear with time and protein concentrations.
Acetyl-CoA carboxylase was assayed by following the incorporation of radioactivity from [14C]bicarbonate into malonyl-CoA as described by Herbert et al. (1996). Fatty acid synthase was measured with [2-14C]malonyl-CoA as described by Ashton et al. (1992). Zero-time incubations were used as controls.
Phospholipid:diacylglycerol acyltransferase (PDAT) and diacylglycerol:diacylglycerol acyltransferase activities were measured using microsomal fractions and [1-14C]dioleoylphosphatidylcholine or [14C]dioleoylglycerol substrates, respectively. Incubations and analysis of products by TLC were essentially as described by Stahl et al. (2004).
Analysis of the system
The system used for the ‘wet’ top-down analysis is briefly discussed in the Introduction. To analyse this system, we carried out two separate independent manipulation experiments each using an enzyme inhibitor specific for one of the blocks (i.e. DFF for Block A and 2-bromooctanoate for Block B). It should also be noted that, by choosing the cytosolic acyl-CoA pool as the intermediate, the metabolic pathway to triacylglycerol is divided into two subcellular locations with fatty acid synthesis in the plastids and lipid assembly in the cytosolic compartment (Roughan & Slack, 1982). For analysis, we manipulated the activity of the fatty acid synthesis reactions (Block A) using DFF and measured the response of enzymes of the complex lipid assembly reactions (Block B), by measuring fractional changes in [1-14C]acetate incorporation into total lipids (δJB/JB), to the resulting fractional changes in the cytosolic acyl-CoA pool (δX/X). This allowed us to calculate (by substituting the appropriate values into Eqn 1) the group elasticity of the enzymes in Block B in relation to x as:
when DFF is added to manipulate Block A directly. The second step involved manipulation of enzyme activity within the complex lipid assembly pathway using 2-bromooctanoate (Bromo). By similar methods, this manipulation allowed us to calculate, using Eqn 2, the group elasticity of Block A in relation to x as:
when Bromo is added to manipulate Block B directly. From these two sets of group elasticities, we calculated the two group flux control coefficients for Block A and Block B over total lipid synthesis, using Eqn 3 or Eqn 4, respectively:
(Eqn 3) (Eqn 4)
and the group concentration coefficients using Eqn 5 or Eqn 6.
(Eqn 5) (Eqn 6)
The error sensitivity coefficient is defined as the fractional change in error in the final calculated result for a given fractional change in error of an experimental data point:
(σz is the error in the experimental term and σy is the error in the calculated term).
Monte Carlo simulations (MCS)
Crystal Ball 7.2 (Academic Edition), by Decisioneering (Denver, CO, USA), was used in combination with Microsoft Excel (Microsoft Office XP) running under Microsoft Windows 2000 Professional to generate the simulated values.
The simulations were generated assuming that each experimental value was normally distributed, with a mean equal to the mean of the experimental repeats, and that the standard deviation was equal to the SEM of the experimental repeats.
For each data point, 1000 observations of the standardized normal distribution were generated and then converted to simulated observations of the data using the formula:
(in this context, X is the simulated data point, σ is the observed SEM, µ is the experimental mean and x is the observation from the standard distribution; Ainscow & Brand, 1998). Therefore, 1000 sets of control coefficients were generated for each of the two studies (see later, Table 7).
Table 7. Estimation of group flux control coefficients and associated errors via Monte Carlo simulation of experimentally derived paired elasticity data from (a) double-manipulation top-down metabolic control analysis (TDCA) or (b) reported (Ramli et al., 2002b) single-manipulation TDCA, in oil palm callus
|Pseudo standard deviation||0.03||0.03|
Like Ainscow & Brand (1998), who pioneered application of MCS to experimental TDCA data, when setting up our simulations we assumed: that the initial experimental error was normally distributed; and that the individual data points were independent of each other.
We determined the mean and the median of the final calculated coefficients and compared these with the averaged values of the experimental data as a test for the normality of the distribution of the calculated results. Further, for a normal distribution, the pseudo-standard deviation (calculated as the limits of the calculated results that encompass 68% of the results around the median, divided by two) and the standard deviation should be the same (see later, Table 7).
Probability density curves were plotted against the relevant median values to visualize, characterize and compare the error distribution associated with each set of coefficients calculated from the MCS (data not shown).
Results and Discussion
To perform a valid analysis by double-manipulation TDCA, it is essential that any changes in carbon flux through the reactions of Block A do not affect flux through Block B other than through the chosen intermediate, cytosolic acyl-CoA. Similarly, an inhibitor affecting the reactions of Block B should not affect directly those of Block A. Thus, identifying inhibitors with a specificity for only one of the two Blocks of reactions was critical.
Use of DFF to inhibit the fatty acid biosynthesis block of reactions
In our experiments we were able to use DFF as an inhibitor of Block A (the fatty acid biosynthesis reactions). Diflufenican has been shown to be an inhibitor of fatty acid synthase in a number of plant systems, apparently through its action on the reductase partial reactions (Ashton et al., 1992), despite the failure of this compound to inhibit the purified enoyl reductase of Escherichia coli Type II fatty acid synthase (Heath et al., 2001). In Table 1 it can be seen that increasing concentrations of diflufenican inhibited the incorporation of radioactivity from [1-14C]acetate into total lipids. Labelling of the system intermediate, the acyl-CoA pool, was also reduced, approximately to the same extent as total lipids (Table 1). It is also worth mentioning that, in plants, [1-14C]acetate essentially only labels the acyl chains of glycerolipids and, therefore, can be used as a rather specific precursor of their fatty acids (Roughan & Slack, 1982). The utility of [14C]acetate for radiolabelling experiments is well recognized despite the fact that the acetyl-CoA pool will usually be generated from other sources in vivo.
Table 1. Diflufenican (DFF) reduces total lipid labelling from [1-14C]acetate in oil palm callus cultures
|0||9.8 ± 0.3||1.3 ± 0.1|
|10||8.1 ± 0.2 (83%)||1.2 ± 0.2 (92%)|
|50||6.3 ± 0.4 (64%)||0.9 ± 0.1 (69%)|
These experiments also identified the concentrations of diflufenican which were appropriate to use in the flux control experiments bearing in mind that it is preferable to use inhibitor concentrations that cause only partial reduction in metabolism, so as not to perturb the system unduly (Fell, 1997).
When the labelling of lipids was examined in more detail it was found that similar, partially inhibitory, concentrations (10–50 µm) of diflufenican did not alter either the pattern of radioactive fatty acids or of complex lipids (Table 2). In the incubation period used, the palm callus cultures formed radioactive palmitate and oleate mainly with smaller amounts of other 18C fatty acids. Intermediates of the Kennedy pathway (Kennedy, 1961), such as phosphatidate (PA) and diacylglycerol (DAG) together with the final product, triacylglycerol (TAG), were well-labelled while the major nonchloroplast membrane lipids (phosphatidylcholine and phosphatidylethanolamine) were also significant components. Because the radioactivity from [1-14C]acetate enters palm callus glycerolipids as fatty acyl groups from the acyl-CoA pool (Ramli et al., 2002a), these experiments also show that, under the experimental conditions, DFF had no effect on lipid assembly (Block B reactions).
Table 2. High concentrations of diflufenican (DFF) do not affect the patterns of fatty acid or lipid labelling in oil palm callus cultures
|0||26 ± 2||8 ± 4||52 ± 4||9 ± 1||5 ± tr.|
|50||24 ± 1||8 ± tr.||54 ± 3||9 ± 1||5 ± tr.|
|0||10 ± 1||8 ± tr.||nd||28 ± 3||24 ± 2||30 ± 3|
|50||10 ± 3||7 ± 1||nd||25 ± 2||28 ± 2||30 ± 5|
|0||nd||nd||14 ± 1||34 ± 2||30 ± 4||22 ± 2|
|10||nd||nd||15 ± 1||35 ± 1||26 ± 1||24 ± 1|
|0||nd||nd||45 ± 5||35 ± 1||12 ± 1||8 ± 1|
|10||nd||nd||47 ± 1||34 ± 1||11 ± 2||8 ± 2|
To confirm that diflufenican had not acted against any enzyme involved in lipid assembly, labelling of the calli was carried out with [U-14C]glycerol and of microsomal fractions with [U-14C]glycerol 3-phosphate (Table 2). In neither case was any change in lipid labelling patterns seen, confirming that diflufenican could be used as a specific inhibitor of Block A reactions. Because diflufenican had been shown previously to inhibit fatty acid synthase in plants, (Ashton et al., 1992), we assayed two of the major reactions in Block A, namely fatty acid synthase and acetyl-CoA carboxylase, directly. The data show clearly (Table 3) that DFF was able to inhibit fatty acid synthase in vitro but had no effect on acetyl-CoA carboxylase.
Table 3. Effect of inhibitors on enzyme activities in the soluble fraction from oil palm callus cultures
|None (control)||4.00 ± 0.18||5.08 ± 0.28|
|Diflufenican (DFF) (10 µm)||3.00 ± 0.21 (75%)||4.98 ± 0.50 (98%)|
|None (control)||3.96 ± 0.22||5.26 ± 0.54|
|Bromooctanoate (20 µm)||3.88 ± 0.58 (98%)||5.36 ± 0.22 (102%)|
Bromooctanoate is a specific inhibitor of lipid assembly
2-Bromooctanoate has been reported as a specific inhibitor of diacylglycerol acyltransferase (DAGAT) in mammalian systems (Mayorek & Tana, 1985) as well as plants (Weselake et al., 1991). Indeed, we have used it as such an inhibitor in callus cultures (Ramli et al., 2005).
To check on the specificity of 2-bromooctanoate in only inhibiting Block B, we carried out a series of experiments. As shown in Table 3, bromooctanoate at 20 µm (a concentration later selected for flux control analysis) had no effect on fatty acid synthase or acetyl-CoA carboxylase in vitro. By contrast, 2-bromooctanoate inhibited lipid assembly as monitored with [U-14C]glycerol and callus cultures or by the incorporation of radioactivity from [U-14C]glycerol 3-phosphate in microsomes (Table 4). In both cases, there was a significant decrease in the percentage labelling of TAG which in the simpler in vitro system was accompanied by a rise in relative labelling of DAG. This was consistent with the inhibition of DAGAT which has been shown to influence the flux of carbon for TAG assembly in several plants (see Ramli et al., 2005; Weselake et al., 2008).
Table 4. 2-Bromooctanoate reduces triacylglycerol labelling in vivo and in vitro
|Control||2.03 ± 0.31||16 ± 4||14 ± 2||44 ± 4||22 ± 2|
|+Bromooctanoate||1.59 ± 0.36||23 ± 3||7 ± 1*||48 ± 2||14 ± tr.*|
|Control||52 ± 6||46 ± 2||40 ± 2||6 ± tr.||8 ± tr.|
|+Bromooctanoate||22 ± 1||45 ± 3||40 ± 1||11 ± 1*||4 ± tr.*|
For [1-14C]acetate labelling in vivo, significant increases in the labelling of phosphatidylcholine and phosphatidylethanolamine were also seen (data not shown). This may have been caused by an increase in DAG, since both phospholipids can be formed from the latter by the phosphotransferase reaction (Harwood, 1989; Browse & Somerville, 1991).
Because triacylglycerol can also be formed by two other enzyme reactions in plants (phospholipid: DAG acyltransferase (PDAT) and DAG: DAG acyltransferase), we tested microsomal fractions for these activities (Stahl et al., 2004). However, the reactions were barely detectable in oil palm fractions and there was no inhibition of them by 0.5 mm 2-bromooctanoate (data not shown).
Although 2-bromooctanoate had no significant effect on fatty acid labelling from [1-14C]acetate with up to 3 h of incubation, longer times resulted in significant inhibition (Table 5). This was despite the lack of effect on either acetyl-CoA carboxylase or fatty acid synthase (Table 3). However, during incubations the acyl-CoA pool was increased significantly (Table 5) and we conclude that this could cause feedback inhibition of de novo fatty acid synthesis, such as has been demonstrated in other systems (Salati & Goodridge, 1996).
Table 5. Effect of 0.02 mM 2-bromooctanoate on incorporation of radioactivity from [1-14C]acetate into total lipids and the acyl-CoA pool
|3 h||3.45 ± 0.29||3.36 ± 0.45||0.92 ± 0.08||1.06 ± 0.04*|
| ||(97%)|| ||(115%)|
|6 h||6.40 ± 0.25||5.00 ± 0.71*||1.01 ± 0.09||1.19 ± 0.03*|
| ||(78%)|| ||(118%)|
Double manipulation control analysis
From the data given above, we were able to confirm that the two inhibitors each targeted only one of the Blocks of reactions directly and could, therefore, be used for double-manipulation experiments. We then carried out a series of such experiments, with three replicates in each case. Both diflufenican and 2-bromooctanoate were used at concentrations that only inhibited total labelling 10–20% and, therefore, did not put an undue strain on the overall flux – as required for MCA (Fell, 1997). The data were then used to calculate the group flux control coefficients. The theory used to derive the relationship between changes in flux and intermediate concentration and the group flux control coefficients for each block over pathway flux is described in the Materials and Methods section. Our calculation for group flux control coefficients, using Eqns 1–4, gave values of and (Table 6).
Table 6. Group coefficients for fatty acid synthesis and complex lipid assembly over lipid biosynthesis in oil palm calli calculated without (a) or with (b) Monte Carlo simulation of data from double-manipulation top-down metabolic control analysis (TDCA)
|(a) – Monte Carlo simulation|| || || || |
|Group flux control coefficients were calculated via elasticity coefficients (x̄ ± SEM, n = 5), in turn calculated from meaned experimental flux and intermediate data from each (triplicate) set of wet experiments on five independent tissue preparations. (See Eqns 1–4)||0.61 ± 0.03||0.39 ± 0.03|| || |
|(b) + Monte Carlo simulation|| || || || |
|Group control coefficients are (x̄ ± SD, n = 5) calculated from 1000 simulations of the meaned elasticity pair derived from preliminary series of 1000 simulations of meaned experimental flux and intermediate data from each (triplicate) set of wet experiments on five independent tissue preparations. (See Eqns 1–6)||0.61 ± 0.03||0.39 ± 0.03||0.25 ± 0.02||−0.25 ± 0.02|
|Error sensitivity coefficients of group control coefficients to the group elasticities (See Eqn 7):|| || || || |
Monte Carlo simulation (1000 simulations) of the meaned elasticity pair (which, in turn, was derived from a preliminary series of 1000 simulations of meaned experimental flux and intermediate data from each (triplicate) set of wet experiments on five independent tissue preparations) also gave values of 0.61 ± 0.03 and 0.39 ± 0.03 for the group flux control coefficients for Block A and Block B, respectively (Tables 6, 7).
Monte Carlo simulation of reported single-manipulation TDCA data
Previously (Ramli et al., 2002b), we carried out a single-manipulation TDCA by manipulating the intermediate (acyl-CoA) pool through the addition of exogenous oleate. In retrospect, and as a result of recent experiments, we should now apply caveats to some of the original assumptions. Thus, for example, in agreement with Ainscow & Brand (1998) we have observed that, even when errors on our experimentally measured values from a single preparation are normally distributed, the combined errors on the flux control coefficients are often skewed and cannot be assumed to be normal because of the multistep calculations involved. Therefore, we used each reported experimental flux, intermediate or elasticity value to perform a series of MCS to generate sets, each of 1000 values, of simulated ‘experimental’ data.
The resulting series of meaned elasticity coefficients, with estimated errors, were then simulated again to generate the final flux control coefficients with associated errors using essentially the same method (i.e. via Eqns 3 and 4) as described in the original paper (Ramli et al., 2002b). Table 8 contains a summary of the previously reported flux control coefficients (a) and the flux control coefficients obtained from the MCS (b) of the reported elasticities directly (Table 8b, Expt 1) or via the sequential MCS of the experimental flux and intermediate data and the subsequently meaned elasticities (Table 8b, Expt 2).
Table 8. Group coefficients for fatty acid synthesis and complex lipid assembly over lipid biosynthesis in oil palm calli calculated without (a) or with (b) Monte Carlo simulation (MCS) of reported data (Ramli et al., 2002b) from single-manipulation top-down metabolic control analysis (TDCA)
| || || || |
|(a) –Monte Carlo simulation|
|Expt 1 (Oleate (400 µm))|| ||0.70||0.30|
|Expt 2 (Oleate (400 µm ))|| ||0.60||0.40|
|Expt 3 (Oleate (800 µm))|| ||0.62||0.38|
| x̄ ± SD (n = 3)|| ||0.64 ± 0.05||0.36 ± 0.05|
| x̄ ± SEM (n = 3)|| ||0.64 ± 0.03||0.36 ± 0.03|
| || || || || |
|(b) +Monte Carlo simulation values|
|Flux or concentration control coefficients are means ± SD calculated from 1000 simulations of:|| || || || |
|Expt 1 (each of three reported elasticity pairs from three separate preps) and|| || ||0.64 ± 0.06||0.36 ± 0.06|
|Expt 2 (the meaned elasticity pair derived from three preliminary series of 1000 simulations of the original meaned flux and intermediate data).||0.25 ± 0.01||−0.25 ± 0.01||0.66 ± 0.07||0.34 ± 0.07|
| x̄ ± SD|| || ||0.65 ± 0.05||0.35 ± 0.05|
|Expt 3 (Error sensitivity coefficients of the group control coefficients to the group elasticities (See Eqn 7)):|| || || || |
As a preliminary step, we simply simulated the three reported pairs of elasticities independently, assuming a normal distribution and 10% error on each, to act as a control experiment for obtaining errors on the meaned flux control coefficients (Table 8b, Expt 1). Not surprisingly, this exercise resulted in the same values for the pair of meaned flux control coefficients (;) with associated similar errors to those reported, as the underlying calculation method was the same.
The next method of calculation involved simulating the means of the three reported elasticity pairs to obtain the meaned flux control coefficients. However, use of the means of the reported elasticities to calculate the flux control coefficients ± SD, via MCS, gave skewed mean elasticity and flux control coefficients where the errors were non-normally distributed. It was only by repeated MCS of the reported flux and intermediate data that we were able to simulate the results of further ‘wet’ experiments and, therefore, to reduce the inherent errors on the meaned elasticities, resulting in normally distributed errors. By, in turn, further MCS of the meaned elasticities ± SEM (see below), we were able to obtain flux control coefficients ± SD that were normally distributed (Table 8b, Expt 2). This is confirmed by the mean and the median values being equal and the SD equalling the pseudo-standard deviation (PSD; see Ainscow & Brand, 1998 and the Materials and Methods section) values. Table 7b shows a summary of the flux control coefficient data obtained from this second MCS series.
In each case (Table 8b, Expts 1 or 2), the estimates of errors in the flux control coefficients from the MCS were greater than those obtained by wet experiment and reported in both Table 8a and our earlier paper (Ramli et al., 2002b). As the second simulation series (Table 8b, Expt 2) was based on the error in the average of the simulated elasticities, the distribution of the calculated results represents the error in the average of the control coefficients. Therefore, the SEM of the experimental values should have been comparable with the SD of the MCS-calculated results (Ainscow & Brand, 1998). (Because the reported errors were expressed as SD in our original paper, we have also given SEM values (Table 8a) for a more accurate comparison with the MCS SD data in Table 8b for the reason cited in the previous sentence.) However, the mean values (± SD) of the pair of simulated elasticities had, in turn, been recalculated from sets of original intermediate and flux data that had each been previously subjected to 1000 simulations. Therefore, the pairs of elasticities that were meaned and simulated in Expt 2 (Table 8) each carried through inherent errors from the previous simulations of the original intermediate and flux data, giving an increased and more accurate error on the final values for the pair of flux control coefficients.
There was no significant difference in the reported mean flux control coefficient pair ± SEM and the meaned result of our MCS (Table 8). This would suggest that, for our simple two-block system, our assumptions of normally distributed errors and of independent values of original data points, were reasonable approximations.
Concentration control coefficients
For completeness, we report the group concentration control coefficients (; ) that describe the control exerted by each of the two blocks of reactions (BlkA or BlkB) over the concentration of the intermediate, cytosolic acyl-CoA ([X]) (Tables 6 and 8b, Expt 2). They are calculated from the group elasticities according to Eqns 5 and 6. In this simple, linear, two-block system, the concentration control coefficients are equal in magnitude and opposite in sign, as a consequence of both summation and connectivity (Brand, 1996). Only in branched systems do the absolute numerical values of the concentration control coefficients become more useful and informative, as the control distribution will then be unevenly shared among the blocks (Ainscow & Brand, 1998).
Error sensitivity coefficients
The error sensitivity coefficients of the four control coefficients to the elasticities from both the double- and the single-manipulation TDCAs (Tables 6 and 8b, Expt 3), defined (as described in the Materials and Methods section, Eqn 7) as the fractional change in the errors on the final calculated control coefficients for a fractional change in the error on the experimentally determined data points, showed that the errors in both the flux and the concentration control coefficients are acutely sensitive to error in under the conditions prevailing in our experimental system. The higher values of the error sensitivity coefficients to (Table 8b, Expt 3) suggest that a more accurate measurement of the associated experimental parameters would result in more precise control coefficients. Lower values would have demonstrated that the errors on the control coefficients were relatively insensitive to the magnitude of the errors in the associated experimental terms. Therefore, we used the error sensitivity coefficients to identify which experiments should be repeated to increase the precision of and confidence in our four final control coefficients and then carried out further simulation experiments (Table 8b, Expt 2).
We found that reducing the error associated with the kinetics of the Block B reactions reduced the error on all four coefficients because had the larger error as a proportion of the determined value and the largest absolute error.
In the experiments described we have applied the technique of both single- and double-manipulation TDCA to lipid synthesis in oil palm. The results from the different types of analysis (as set out in Tables 6–8 and summarized in Table 9) show that the fatty acid synthesis block of reactions exerts considerably more control than lipid assembly under our experimental conditions. This puts a quantitative value on the two contributions, which agrees with data from other types of experiments (Bafor et al., 1990: Bao & Ohlrogge, 1999) in showing that supply of fatty acids can be a major controlling factor for lipid accumulation in some plants but less so in others (Weselake et al., 2008). These contrasting findings emphasize the difficulty of extrapolating data from one plant species to others, as we have emphasized for oil crops in the past (Ramli et al., 2005). Nevertheless, the top-down approach has also validated the concept that control is shared throughout the pathway and that the concept of a single ‘rate-limiting’ step within a pathway is too simple a description. Thus, lipid assembly is still an important contributor to overall control in oil palm and efforts to raise fatty acid synthesis (e.g. by genetic manipulation) will increasingly cause lipid assembly to become controlling.
Table 9. Summary of group flux control coefficients for fatty acid synthesis and complex lipid assembly over lipid biosynthesis in oil palm calli from two top-down metabolic control analysis (TDCA) studies
|Double-manipulation (Table 6)|
|−MCS (x̄ ± SEM)||0.61 ± 0.03||0.39 ± 0.03|
|+MCS (x̄ ± SD)||0.61 ± 0.03||0.39 ± 0.03|
|Single-manipulation (Table 8)|
|−MCS (x̄ ± SEM)||0.64 ± 0.03||0.36 ± 0.03|
|+MCS (x̄ ± SD)||0.65 ± 0.05||0.35 ± 0.05|
|+MCS (x̄ ± SD, n = 4)||0.63 + 0.02||0.37 + 0.02|
This work was supported in part by a studentship to U.S.R. from the Malaysian Palm Oil Board, for which we are grateful.