Plastidial glycolysis in developing Arabidopsis embryos



This article is corrected by:

  1. Errata: Erratum Volume 186, Issue 2, 558, Article first published online: 25 March 2010

Author for correspondence:
Alison M. Smith
Tel.: +44 (0)1603 450622


  • During oilseed embryo development, carbon from sucrose is utilized for fatty acid synthesis in the plastid. The role of plastidial glycolysis in Arabidopsis embryo oil accumulation was investigated.
  • Genes encoding enolases (ENO) and phosphoglyceromutases (PGlyM) were identified, and activities and subcellular locations were established by expression of recombinant and green fluorescent protein (GFP)-fusion proteins. Mutant Arabidopsis plants lacking putative plastidial isoforms were characterized with respect to isoform composition and embryo oil content.
  • In the developing embryo, ENO1 and ENO2 account for most or all of the plastidial and cytosolic ENO activity, respectively, and PGLYM1 accounts for most or all of the plastidial PGlyM activity. The eno1 and pglym1 mutants, in which plastidic ENO and PGlyM activities were undetectable, had wild-type amounts of seed oil at maturity.
  • It is concluded that although plastids of developing Arabidopsis embryos have the capacity to carry out the lower part of the glycolytic pathway, the cytosolic glycolytic pathway alone is sufficient to support the flux from 3-phosphoglycerate to phosphoenolpyruvate required for oil production. The results highlight the importance for oil production of translocators that facilitate interchange of glycolytic intermediates between the cytosol and the plastid stroma.


During seed development in oilseed rape and its close relative Arabidopsis thaliana, a large fraction of the sucrose entering the developing embryo is converted to acetyl CoA, the immediate precursor of the fatty acids from which lipids are synthesized. In mature oilseed rape and Arabidopsis seeds, lipid reserves account for c. 40% of the weight (O’Neill et al., 2003). However, in spite of the importance of lipid synthesis in oilseed embryos, the compartmentation of the pathway that converts sucrose to acetyl CoA is incompletely understood. The initial steps – conversion of sucrose to hexose phosphates – are known to occur in the cytosol, and the final steps – synthesis of acetyl CoA from pyruvate – occur in the plastids. The glycolytic conversion of hexose phosphates to pyruvate, however, could potentially occur in either or both compartments. Recent research is uncovering a complex picture, in which some steps are thought to occur primarily in the cytosol and others primarily in the plastid. The aim of this work was to discover the importance of the plastid in the glycolytic conversion of 3-phosphoglycerate (3PGA) to phosphoenolpyruvate (PEP) on this pathway.

In developing embryos of oilseed rape and various nonphotosynthetic organs of other species, both the cytosol and the plastids contain a full glycolytic pathway (Miernyk & Dennis, 1982; Journet & Douce, 1985; Botha & Dennis, 1986; Denyer & Smith, 1988; Entwistle & ap Rees, 1988; Kang & Rawsthorne, 1994; Eastmond & Rawsthorne, 2000). Plastids isolated from developing oilseed rape embryos can take up hexose phosphates, PEP and pyruvate and convert them to fatty acids (Kang & Rawsthorne, 1994, 1996; Eastmond & Rawsthorne, 2000; Kubis et al., 2004). In Arabidopsis, genes encoding putative cytosolic and plastidial isoforms of the glycolytic enzymes, and plastid envelope translocators for hexose phosphates, PEP, 3PGA and triose phosphates are all expressed in developing seeds (White et al., 2000). In oilseed rape embryos, the capacity for glycolysis in both the cytosol and the plastid is considerably in excess of that required for maximum rates of lipid synthesis (Eastmond & Rawsthorne, 2000). Therefore, the entire glycolytic flux could potentially be accommodated in either compartment. This situation is in marked contrast with that in leaves. Chloroplasts are reported to lack detectable activity of the two glycolytic enzymes that catalyse the conversion of 3PGA to PEP, phosphoglyceromutase (PglyM) and enolase (ENO) (Stitt & ap Rees, 1979; Bagge & Larsson, 1986). Consistent with these reports, leaf metabolism is seriously disrupted by strong reduction or loss of either a cytosolic ENO (Voll et al., 2009) or a plastidial phosphoenolpyruvate/phosphate translocator (PPT; Streatfield et al., 1999). These observations imply that chloroplasts have insufficient capacity for glycolytic synthesis of PEP from triose phosphate to meet their demand for PEP.

Recent research indicates that some steps in glycolysis in oilseed embryos occur predominantly in the plastid. First, in sunflower and soybean embryos, metabolic flux analyses show that a substantial proportion of the triose phosphate used for fatty acid synthesis is made from hexose phosphates imported into the plastid (Sriram et al., 2004; Alonso et al., 2007). Second, the conversion of PEP to pyruvate (catalysed by pyruvate kinase) appears to occur predominantly inside the plastid in Arabidopsis embryos. Mutants with strongly reduced activities of plastidial pyruvate kinase have reduced rates of lipid synthesis (Andre et al., 2007; Baud et al., 2007). Third, in oilseed rape embryos, up to half of the 3PGA used for acetyl CoA synthesis is generated inside the plastid. Rubisco in embryo plastids assimilates CO2 lost in respiration and the oxidative pentose phosphate pathway, using ribulose bisphosphate (RuBP) generated via the plastidial pentose phosphate pathway. The 3PGA product enters glycolysis for conversion to acetyl CoA, and hence fatty acids (Ruuska et al., 2004; Schwender et al., 2004), enhancing the overall efficiency of the conversion of sucrose to lipid.

Taken together, these observations suggest that a significant proportion of the triose phosphate and/or 3PGA used for fatty acid synthesis is generated inside the plastid in oilseed embryos, and that most of the conversion of PEP to pyruvate also occurs in the plastid. However, the compartmentation of the flux from 3PGA to PEP is still to be discovered. Metabolic flux analyses cannot distinguish between plastidial and cytosolic fluxes for this part of glycolysis (Sriram et al., 2004; Alonso et al., 2007), and the metabolites concerned can move across the plastid envelope via either the triose phosphate/phosphate translocator (TPT; Knappe et al., 2003b; Weber et al., 2005) or the PPT (Knappe et al., 2003a). For two of the enzymes involved – PGlyM and ENO – only a very small percentage of total activity is present in the plastid (3 and 6%, respectively) in oilseed rape embryos during the phase of rapid oil synthesis, and plastidial activity declines as the rate of oil synthesis increases through development (Eastmond & Rawsthorne, 2000). Analysis of transcript abundances also suggests that cytosolic isoforms are expressed to a much greater extent than plastidial isoforms (White et al., 2000; Ruuska et al., 2002). For these reasons, it has been assumed that the flux catalysed by PGlyM and ENO occurs primarily or exclusively in the cytosol (White et al., 2000; Ruuska et al., 2002; Sriram et al., 2004). If this is the case, the provision of carbon for fatty acid synthesis in oilseed rape embryos involves major fluxes of 3PGA from the plastid to the cytosol, and of PEP from the cytosol to the plastid. However, the plastidial activities of both PGlyM and ENO are greater than would be required to account for the observed rate of fatty acid synthesis in intact embryos (Eastmond & Rawsthorne, 2000), leaving open the possibility that most or all of the glycolytic flux occurs in the plastid.

This paper directly examines the importance of plastidial PGlyM and ENO for lipid accumulation in Arabidopsis seeds. We show that both ENO and PGlyM activities in the plastids of developing embryos are accounted for by single isoforms, ENO1 and PGlyM1, respectively. Elimination of either ENO1 or PGlyM1 results in loss of the plastidial activity of the enzyme but has no effect on seed weight and lipid content. We discuss the implications of these results for understanding the compartmentation of glycolytic flux and the importance of metabolite translocators during oil synthesis in the developing embryo.

Materials and Methods

Plant material and growth conditions

Plants were grown either in a controlled environment room (16 h light : 8 h dark, 22°C, 250 μmol photosynthetically active radiation m−2 s−1) or in a glasshouse with additional lighting to give 16 h d–1 illumination (day minimum 18°C, night minimum 15°C). Plant growth conditions and experimental design for the determination of seed lipid content were as described in Hobbs et al. (2004).

Seeds of T-DNA insertion lines for ENO1 and PGlyM1 were obtained from the Salk, the SAIL or the GABI-Kat collections as described in the ‘Results’ section.

DNA and RNA extraction, PCR and semiquantitative RT-PCR

All primers mentioned hereafter are described in Supporting Information Table S4. Genomic DNA was extracted from young leaves using the DNeasy® Plant Mini Kit (Qiagen, PCR reactions were routinely performed with the following cycling programme: 94°C for 5 min, 35 cycles of 94°C for 1 min, 55°C for 1 min (unless otherwise stated; Table S4), 72°C for 1.5 min, followed by an extension cycle at 72°C for 7 min. The reaction contained appropriate dilution of DNA template, 25 nM of each primer, 100 nM of each dNTP, 1.5 mM MgCl2, appropriate dilution of PCR buffer (Promega, and 0.5 units Taq polymerase. DNA sequencing was performed with an Abi Prism 3730XL capillary sequencer (Applied Biosystems, Foster City, CA, USA;

PCR-based screening with genomic DNA to identify individuals homozygous for T-DNA insertions was performed with gene-specific primers designed with the SIGnAL T-DNA verification primer design program ( and used in combination with the appropriate T-DNA left border (LB) primers (SalkLBa1 or SalkLBb1 for Salk lines; SailLB1 for Sail lines; and GABI_L for GABI-Kat lines). Actin primers (ActFW, ActRV) were included in each PCR reaction as internal controls. Each insertion was verified by sequencing of the corresponding T-DNA/genomic DNA junction.

Total RNA was extracted from leaves using the RNeasy Plant Mini Kit (Qiagen). RNA was treated with DNAse I (Sigma, and first-strand cDNA was synthesized with SuperScript II reverse transcriptase (Invitrogen, For semiquantitative RT-PCR, the TUBULIN2 gene (At5g62690) was used as a control. PCR conditions were as previously decribed, with 30 cycles for TUBULIN2 (primers TUB3-1, TUB5-1) and 40 cycles for ENO1 and PGlyM1, with primer pairs VA100 and EnolabR (eno1-1, eno1-2), EnolaF and EnolabR (eno1-3), or rtL1 and rtR1 (pglym1-1, pglym1-2).

Subcellular localization of Arabidopsis enolases and PGlyM1

The full-length cDNAs of ENO1, ENO2, ENO3 and PGlyM1 were amplified with primers Eno1FW and ENO1RV, VA67 and VA75, VA69 and VA76, VA101 and VA102, respectively, subcloned into pCR8/GW/TOPO vector (Invitrogen) and sequenced. These constructs were introduced into the pK7FWG2.0 binary vector (Karimi et al., 2002) with an in-frame C-terminal green fluorescent protein (GFP) via Gateway LR clonase (Invitrogen) II-mediated recombination. Recombined vectors were propagated in TOP10 Escherichia coli cells (Invitrogen) and then used to transform Agrobacterium tumefaciens GV3101 cells. Leaves of 3-wk-old Nicotiana benthamiana were infiltrated (Wu et al., 2004) and GFP fluorescence was viewed 3 d post-infiltration with a Leica SP2 confocal microscope (Leica Microsystems CMS GmbH, Mannheim, Germany) on a Leica DM IRB inverted microscope. The signals for GFP fluorescence and chlorophyll autofluorescence were collected simultaneously. The images were processed with ImageJ software (

To verify the presence of GFP-fusion proteins in N. benthamiana, infiltrated leaf material was homogenized in 100 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, 2 mM DTT, 0.1% (v/v) Triton X-100, 10% (v/v) glycerol, 1% (w/v) polyvinylpolypyrrolidone (PVPP) and protease inhibitors (plant cocktail, Sigma). After centrifugation for 30 min at 13 000g at 4°C, soluble proteins were subjected to SDS-PAGE (Laemmli, 1970), transferred on nitrocellulose membranes and probed with a monoclonal anti-GFP antibody (B-2, sc-9996; Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA). Immunoreactive bands were visualized with the ECL chemiluminescence reagent (GE Healthcare:

Expression and purification of 6-histidine tagged enolases in E. coli

The full-length cDNAs of ENO2 and ENO3 were amplified by PCR using Expand High Fidelity Taq (Roche Applied Science: with primers VA67 and VA68, and VA69 and VA70 respectively. PCR reactions were as described earlier, except that the number of cycles was reduced to 28 and the annealing temperature was modified for each individual primer-pair. Constructs were cloned into pCR8/GW/TOPO vector, sequenced and introduced by Gateway LR clonase II-mediated recombination in the pET-DEST42 Gateway vector (Invitrogen). This vector allows isopropyl-β-d-thiogalactopyranoside (IPTG)-inducible expression in E. coli of recombinant proteins with a C-terminal V5 and 6-histidine (6xHis) epitope tag. Recombined pET-DEST42 vector carrying each of the Arabidopsis ENO constructs was used to transform BL21 Star (DE3) One Shot E. coli cells (Invitrogen). Expression of recombinant proteins was induced for c. 16 h at 16°C with 0.2 mM IPTG in LB medium supplemented with 100 μg ml−1 ampicillin. 6-Histidine-tagged recombinant proteins were purified (Andriotis & Rathjen, 2006), desalted on Sephadex G25 columns equilibrated with 100 mM MOPS-NaOH (pH 7.4), 2 mM DTT, 10% (v/v) glycerol and assayed spectrophotometrically for enolase activity with 2-phosphoglycerate (2PGA; Denyer & Smith, 1988).

Complementation of the enolase-deficient E. coli strain DF261

The ENO1 cDNA was cloned with Expand High Fidelity Taq using primers VA80 and VA82 to introduce a BamHI and a SmaI site at the 3′- and 5′-prime sites of ENO1, respectively. ENO2 and ENO3 were cloned using primers VA207 and VA208, and ENO3FW and ENO3RV, respectively. The ENO2 construct had a 5′- EcoRI and 3′- PstI restriction site, and the ENO3 construct had a 5′- BamHI and 3′- HindIII restriction site. These constructs were subcloned into pCR2.1-TOPO vector (Invitrogen), sequenced then ligated into appropriately restriction-digested and dephosphorylated pUC8 vector. Plasmids were propagated in TOP10 E. coli cells (Invitrogen) and used to transform eno2-E. coli DF261 cells (Hfr(PO2A), garB10, fhuA22, ompF627(T2R), fadL701(T2R), eno2-, relA1, pitA10, spoT1, rrnB-2, mcrB1, creC510) obtained from the E. coli Genetic Stock Center, Department of Biology, Yale University. Transformants were grown at 37°C in liquid minimal medium 63 (M63, Cohen & Rickenberg, 1956) containing 25 mM malate, 0.1% (w/v) casamino acids and 12.5 mM glycerol as the carbon source. To test the functionality of the Arabidopsis enolases, transformed E. coli DF261 cells were plated on ampicillin (100 μg ml−1) agar-M63 supplemented with either glucose or glycerol (both 12.5 mM) as the carbon source, at 37°C.

Site-directed mutagenesis of ENO3

PCR-based site-directed mutagenesis of ENO3 by gene splicing and overlap extension was performed as described by Horton et al. (1989). The Expand High Fidelity PCR System was used. Flanking primers were ENO3FW and VA210, which introduce BamHI and HindIII restriction sites at the 5′- and 3′-prime ends of ENO3, respectively. Internal mutagenic primer pairs VA211 and VA212, and VA217 and VA218 were used to mutate the specific cognate amino acid positions E251 and D252 of ENO3 and create the single mutant eno3E251D and the double mutant eno3E251D,D252E. These constructs were subcloned into pCR2.1-TOPO (Invitrogen) and both DNA strands were sequenced to confirm the presence of the introduced mutations. Functional complementation experiments were as described earlier.

Chromatographic separation of ENO and PGlyM isoforms

Developing siliques were powdered under liquid nitrogen then homogenized with medium A ((100 mM Bis-Tris propane-HCl (pH 6.8), 10 mM 2-mercaptoethanol, 5 mM MgCl2, 0.1% (v/v) Triton X-100, 5% (v/v) ethanediol) with 2% (w/v) PVPP and protease inhibitors (plant cocktail, Sigma)). Homogenates were centrifuged at 13 000g for 30 min at 4°C and the supernatants were desalted on Sephadex G25 pre-equilibrated with medium A. The eluates were filtered through 0.22 μm filters and loaded onto a Mono Q 5/50 GL Tricorn column (1 ml column volume) equilibrated with medium A on an AKTA FPLC system (GE Healthcare:, at 4°C. The column was washed with medium A at a flow rate of 1 ml min−1 until absorbance at 280 nm returned to baseline, then eluted with a linear gradient of NaCl in medium A (0–175 mM in 35 ml). Fractions of 1 ml were collected and assayed for enolase activity (Denyer & Smith, 1988).

For the separation of PGlyM isoforms, the method was exactly the same, except that medium A was replaced by medium B throughout (20 mM Tris/HCl (pH 7.6), 10 mM 2-mercaptoethanol, 5 mM dithiothreitol, 0.1% (v/v) Triton X-100, 5% (v/v) ethanediol, 2% (w/v) PVPP, protease inhibitors (plant cocktail, Sigma)), and the NaCl gradient was 0–1 M in 35 ml. For each eluate, 3.5 mg total protein was applied to the MonoQ column. Fractions were assayed for PGlyM activity as described by Botha & Dennis (1986), except that 3PGA was 7.5 mM.

Nuclear magnetic resonance (NMR) determination of seed lipid content

Lipid content (% w/w) of mature seeds was determined as described by Hobbs et al. (2004) using a QP20+/PC pulsed magnetic resonance spectrometer (Oxford Instruments, Abingdon, UK) operating at a proton frequency of 20 MHz (0.47 Tesla). Samples of c. 200 mg seeds were analysed in triplicate.

Bioinformatic analyses

Deduced amino acid sequences of enzymes and related proteins were obtained from the following sources: Arabidopsis, TIGR Arabidopsis thaliana Genome Annotation Database (; rice, Rice Genome Annotation Project (; poplar, JGI Populus trichocarpa Genome Project (; grape, Genoscope Centre National de Séquençage (; Physcomitrella, JGI Physcomitrella patens Genome Project (; E. coli, EcoCyc Encyclopedia of E. coli K-12 Genes and Metabolism (; and yeast, the Saccharomyces Genome Database ( All other sequences were obtained from the Protein Knowledgebase (UniProtKB) of the Universal Protein Resource ( Predicted protein sequences were compared using the automated analytical pipeline (; Dereeper et al., 2008). Alignment of multiple sequences was achieved using MUSCLE 3.7 and refined with Gblocks 0.91b. Maximum-likelihood trees were built using PhyML 3.0aLRT and rendered using TreeDyn 198.3. In these analyses (all available at parameters were set at their default values using in the ‘one click’ mode.


Enolase genes in Arabidopsis

Arabidopsis contains three ENO genes, At1g74030 (ENO1), At2g36530 (ENO2; previously named LOS2, Lee et al., 2002) and At2g29560 (ENO3) (Prabhakar et al., 2009). Their predicted translation products are polypeptides of 477, 444 and 475 amino acids, respectively (Table S1). In phylogenetic analyses, the three Arabidopsis enolases cluster closely with other plant ENO sequences. They show high similarity at the level of predicted amino acid sequence to putative enolases from rice as well as known enolases from E. coli, yeast and humans (Fig. S1).

Analysis of publicly available microarray data (Fig. 1) shows that ENO2 is expressed in all organs and developmental stages examined, and at levels higher than ENO1 and ENO3 in all instances except pollen. ENO1 is expressed more highly in roots, siliques and seeds than in other organs. Analysis of the expression pattern of ENO1 by semiquantitative RT-PCR and by staining for beta-glucuronidase (GUS) activity in Arabidopsis plants transformed with an ENO1 promoter-GUS fusion revealed that it is also highly expressed in the shoot apex and the trichomes of young leaves (Prabhakar et al., 2009). ENO3 is expressed at lower levels than ENO1 or ENO2 in all organs. During seed development, expression levels of ENO1 and ENO2 peak at the torpedo stage (Fig. 1). Thereafter, ENO1 transcript is almost undetectable, while ENO2 transcript abundance falls to about one-third of peak levels. ENO3 transcript is not detectable at any stage of seed development.

Figure 1.

 Expression pattern of Arabidopsis enolase genes. (a) Transcript abundances in different plant organs. Data were obtained from AtGenExpress at the Genevestigator site ( and values are means ± standard errors of three biological replicates. (b) Transcript abundances during seed development. Data were obtained as in (a). Developmental stages are as follows: 1, globular; 2, heart; 3, early torpedo; 4, torpedo; 5, curled cotyledon; 6, expanded cotyledon; 7, fully developed embryo; 8, mature embryo; 9, dry seed.

Subcellular localization of enolase isoforms

ENO2 and ENO3 are predicted to be cytosolic proteins, while ENO1 has a predicted transit peptide spanning amino acids 1–41 that putatively confers plastidial localization (Table S1). ENO1 has recently been shown to be plastidial, through analysis of the pattern of expression of a GFP fusion protein in an Arabidopsis cell culture (Prabhakar et al., 2009). Here, this analysis was extended to all three ENO gene products. Full-length ENO1, ENO2 and ENO3 fused at their C-termini to GFP, and GFP alone, were transiently expressed in N. benthamiana leaves under control of the CaMV 35S promoter. Transformed areas were examined by confocal laser microscopy. The fluorescence signal from ENO1-GFP completely overlapped that of chlorophyll autofluorescence, indicating that ENO1 is located in the chloroplast (Fig. 2a). By contrast, GFP fluorescence for ENO2-GFP and ENO3-GFP was observed in the cytosol (Fig. 2c,d). ENO2-GFP fluorescence was also observed in nuclei (Fig. 2b), as reported by Lee et al. (2002). The fluorescence signal from GFP alone was detected in the cytosol and in the nucleus (Fig. 2e). Immunoblot analysis of soluble protein extracts from infiltrated leaves, using an antiserum for GFP, confirmed the accumulation of chimeric proteins of the predicted size (Fig. 2f).

Figure 2.

 Subcellular localization of Arabidopsis enolases in Nicotiana benthamiana leaves. Infiltrated areas (mesophyll cells in (a), (c) and (d); epidermal cells in (b) and (e)) were viewed with a confocal microscope for green fluorescent protein (GFP) (left panel) and chlorophyll autofluoresence (Chl, in red, middle panel). Panels on the right are merged images of GFP and chlorophyll autofluorescence. Bars, 10 μm. (a) 35S::ENO1-GFP; (b) and (c) 35S::ENO2-GFP; (d) 35S::ENO3-GFP; and (e) 35S::GFP control. (f) Expression of ENO-GFP chimeric proteins in N. benthamiana leaves. Immunoblot analysis using a GFP monoclonal antiserum of extracts from leaves transiently expressing 35S::ENO1-GFP; 35S::ENO2-GFP; 35S::ENO3-GFP; and an empty vector (lane C). Positions of molecular mass markers are indicated (mass in kDa) on the right. The asterisk indicates bands of the size expected for ENO-GFP chimeric proteins (c. 80 kDa). For ENO1-GFP and ENO3-GFP, lanes contained 50 μg protein; for ENO2-GFP, lanes contained 5 μg protein. Equivalent loadings were confirmed by Ponceau S staining of the major protein band on the membranes (bottom panels).

ENO1 and ENO2 are functional enolases

Recombinant forms of both ENO1 and ENO2 proteins possess enolase activity in vitro (Lee et al., 2002; Prabhakar et al., 2009). Consistent with this, both proteins were able to restore the ability of the E. coli mutant strain DF261 to grow on glucose (Fig 3). This bacterial strain lacks enolase and is able to grow on glycerol but not glucose (Hillman & Fraenkel, 1975). It was previously used to test the functionality of a cDNA encoding a putative enolase from maize (Lal et al., 1991). To extend this analysis to the whole enzyme family, ENO3 was tested for in vitro activity and for the ability to complement the E. coli mutant.

Figure 3.

 Heterologous expression of Arabidopsis enolases in E. coli. (a) SDS-polyacrylamide gel of purified ENO2-6xHis (lane 1, 10 μg) and ENO3-6xHis (lane 2, 5 μg) purified from E. coli. Predicted molecular masses of ENO2-6xHis and ENO3-6xHis are 48 and 52 kDa, respectively. Masses of molecular markers (M) are given in kDa on the right. (b) Functional complementation of the eno-E. coli strain DF261. The strain was transformed with ENO1, ENO2 and ENO3 in frame with the lacZ promoter in the pUC8 expression vector. As a control, bacteria were also transformed with the empty vector (pUC8 EV). The control strain can grow on glycerol but not on glucose. Growth is restored in the presence of glucose as the carbon source when the bacteria express a functional enolase. (c) Amino acid substitutions in the catalytic site of ENO3 restore the functionality of the enzyme. Wild-type ENO3, eno3E251D and eno3E251D,D252E were cloned in frame with the lacZ promoter in the pUC8 expression vector and used to transform the eno-E. coli DF261 strain as in (b).

ENO3 with a C-terminal 6xHis epitope tag (ENO3-6xHis) was expressed in E. coli. ENO2-6xHis was also expressed as a comparison. Both yielded soluble proteins, which were purified by immobilized metal ion affinity chromatography (Fig. 3a). ENO2-6xHis was active in vitro, as reported by Lee et al. (2002). The specific activity of the purified protein was 6.5 μmol min−1 mg−1 protein and the Km with respect to 2PGA was c. 16 μM. By contrast, ENO3-6xHis had no detectable activity and was unable to complement the growth phenotype of E. coli DF261 (Fig. 3b).

Alignment of the AtENO protein sequences with that of yeast enolase (Fig. S2) showed that residue E211 of the yeast enzyme, known to act as a general acid in the reaction mechanism, is conserved in the same context in AtENO1 and AtENO2 but is substituted by an aspartate residue in AtENO3 (D252). A search in other plant genomes with the consensus sequence KKKYGQDATNVGDE211GGFAPNIQENKEGLELL (yeast numbering) showed that the same aspartate substitution occurs in a predicted ENO3-like protein in rice (Os03g15950), and in two predicted ENO proteins in grape, although not in predicted ENO proteins in poplar or the moss Physcomitrella (Fig. S2). However, in both AtENO3 and the predicted proteins from grape, the consensus sequence GDEGG is changed to GEDGG, so the retention of E in an adjacent position may allow it to act as a general acid in the reaction mechanism. In rice, the corresponding sequence is GDDGG, but in this case E > D may be a functionally conservative substitution and the new acidic reside may still be able to fulfil the required role. It is not known whether ENO3-like sequences from plants other than Arabidopsis encode catalytically active enzymes. However, substitution of E211 by glutamine in the yeast enolase results in a dramatic reduction of kcat (Poyner et al., 1996).

To test the effects of these substitutions of presumed catalytically important residues, we expressed the wild-type ENO3 (containing the sequence GEDGG) and the mutant versions eno3E251D (containing the sequence GDDGG, as in rice Os03g15950) and eno3E251D,D252E (containing the sequence GDEGG as in AtENO2) in the enoE. coli strain DF261. Although bacteria expressing the wild-type ENO3 could not grow on glucose as the sole carbon source (Fig. 3b,c), bacteria harbouring eno3E251D showed limited growth, and the expression of eno3E251D,D252E fully complemented the growth phenotype of E. coli DF261 on glucose-containing media (Fig. 3c). Thus the lack of detectable activity of AtENO3 is attributable, at least in part, to substitutions of two residues in the active site. Overall these data suggest that, of the three Arabidopsis ENO genes, ENO1 and ENO2 encode functional enolases and account for all the enolase activity in Arabidopsis embryos.

Identification of eno1 mutants

To discover the importance of plastidial enolase in lipid accumulation in Arabidopsis seeds, three T-DNA insertion lines were identified from the Salk (Alonso et al., 2003) and GABI-Kat (Rosso et al., 2003) collections in which the expression of ENO1 was predicted to be affected. These were Salk_151513 (eno1-1), GABI_227_D02 (eno1-2) and GABI_308_H02 (eno1-3). Flanking sequence tags (FSTs) corresponding to left T-DNA borders (LB) were isolated by PCR from the putative mutants and sequenced to confirm the insertion sites. These were –73, +655 and –51 nucleotides relative to the ATG start codon in eno1-1, eno1-2 and eno1-3, respectively (Figs. 4a,S3). To investigate the impact of the T-DNA insertions on expression of ENO1, primer pairs were designed to amplify cDNA sequences corresponding to positions flanking the T-DNA insertion site. The correct cDNA sequences were amplified from wild-type ENO1 plants. No transcript was detectable in homozygous eno1-1 and eno1-2 (Fig. 4b).

Figure 4.

 Identification of T-DNA insertion lines for plastidial enolase. (a) Exon-intron structure and T-DNA insertion positions in the Arabidopsis (Col-0) gene ENO1 (At1g74030). Exons are depicted as closed boxes, introns are open boxes and the 5′- and 3′-UTRs are grey boxes. The solid line represents the promoter region of ENO1. Scale bar is 500 nucleotides. T-DNA insertions in eno1-1 (−73) and eno1-2 (+655) are indicated by inverted arrowheads and identified by the corresponding database nomenclature. The structure of the T-DNA insertion in the eno1-3 allele is shown (not to scale). LB and RB are the left and right borders of the T-DNA of vector pAC161, respectively. The CaMV 35S promoter is shown as an open arrow. Arrows show the approximate position and direction of primers used in (d). (b) RT-PCR analysis of lines carrying the eno1-1 and eno1-2 alleles. Total RNA was extracted from leaves of seedlings of mutant (−/−) and corresponding syngenic wild-type (+/+) plants and subjected to an RT-PCR assay with primer pairs flanking the insertion site. The TUBULIN2 gene (At5g62690) was used as a loading control. (c) RT-PCR analysis of a line carrying the eno1-3 allele. Total RNA was extracted from leaves of eno1-3 and corresponding syngenic wild-type (WT) plants and subjected to an RT-PCR assay. The TUBULIN2 gene was used as a loading control. Controls were run on the same gel at the same time. (d) PCR analysis of genomic DNA to determine the orientation of the T-DNA insertion in the eno1-3 allele. PCR reactions were performed with the primer combinations indicated on the right of the gels as shown in (a). M, DNA markers. *, 1000 bp marker. (e) RT-PCR analysis of ENO2 and ENO3 transcripts in the eno1 mutant lines. Total RNA was extracted from seedlings of mutant (−/− in eno1-1, eno1-2; 35S in eno1-3) and corresponding syngenic wild-type (+/+) plants and subjected to an RT-PCR assay with primer pairs specifically amplifying fragments of the ENO2 and ENO3 cDNA. The TUBULIN2 gene was used as a loading control.

Transcript abundances of ENO1 were higher in the eno1-3 mutant than in the syngenic wild-type line (Fig. 4c). The vector (pAC161) used in the generation of this line has a CaMV 35S promoter immediately inside the T-DNA right border (RB) ( A PCR-based assay was used to determine the orientation of the insertion (Fig. 4d). Gene-specific primers amplified the correct ENO1 genomic DNA fragment only in wild-type plants and not in eno1-3 (owing to disruption of the ENO1 genomic sequence by the T-DNA). A combination of gene-specific primers together with either RB- or LB-specific primers showed that the T-DNA was integrated in the 5′LB–[T-DNA]–RB3′ direction, close to and upstream of the 5′-end of ENO1. Sequencing of the RB-FST verified that the CaMV 35S promoter was located 30 nucleotides upstream of the ENO1 ATG start codon. Thus, in eno1-3, insertion of the T-DNA resulted in the deletion of c. 20 nucleotides between −51 and −30, and their substitution by the T-DNA such that positioning of the CaMV 35S promoter adjacent to the ATG start codon resulted in increased expression of the ENO1 gene (see Ülker et al. (2008) for further discussion of gene activation in insertion lines). RT-PCR analysis showed that transcript abundances of ENO2 and ENO3 in the eno1 mutants were the same as in wild-type plants (Fig. 4e). Thus, interpretation of any phenotypes in the eno1 mutants is unlikely to be complicated by changes in expression of the other ENO isoforms.

Separation of ENO isoforms and effect of the eno1 mutations on ENO activity

MonoQ anion exchange chromatography of tissue extracts was used to separate the ENO isoforms. ENO activity in extracts of developing siliques of wild-type plants eluted from the column in two peaks: a major peak at c. 60 mM NaCl (P1) and a minor peak (P2) at c. 100 mM NaCl (Fig. 5a). By contrast, activity from leaves eluted as a single peak, in the same position as P1 from siliques (Fig. 5b). To provide information about the identity of the peaks, the elution profile of ENO activity from developing siliques of the los2 mutant was examined. This mutant carries a recessive mutation in the ENO2 gene resulting in reduced ENO2 specific activity (Lee et al., 2002). In los2 extracts the size of the P2 activity peak was the same as extracts of wild-type siliques from the same genetic background (C24), but the size of the P1 peak was significantly reduced (Fig. 5c). This indicates that P1 is attributable to ENO2, and P2 to ENO1.

Figure 5.

 Separation of Arabidopsis ENO isoforms and effect of eno1 mutations on enolase activity. Soluble extracts from leaves or developing siliques were applied to a MonoQ anion exchange column by fast protein liquid chromatography (FPLC). After washing, the column was eluted with a gradient of increasing NaCl concentration (dashed line). Fractions (1 ml) were assayed for enolase activity (solid line). The x-axis represents an elution volume of 35 ml. For comparisons of extracts from eno1 and syngenic wild-type plants, the same amounts of protein from the two extracts were analysed (1.5 mg in (a) and (e); 8–10 mg in other experiments). (a) Elution profile of ENO activity from developing siliques of wild-type (Col-0) plants. P1 and P2 indicate the two activity peaks resolved by chromatography. No activity was found in the unbound fraction during sample loading and column washing. This result is typical of several independent experiments. (b) Elution profile from leaves of wild-type (Col-0) plants. (c) Elution profiles from siliques of a los2 mutant and a syngenic wild-type plant (both in the C24 background). (d) and (e) Elution profiles from siliques of eno1-1 and eno1-2 plants, respectively. (f) Elution profiles from siliques of eno1-3 plants. The two profiles are from two independent experiments, using siliques from different plants. Recoveries of enolase activity as percentages of the activity applied to the column were: (a) 79%, (b) 84%, (c) 65% for C24, and 82% for los2, (d) 79%, (e) 87%, (f) 80% (closed symbols) and 86% (open symbols).

The proposed subcellular location of isoforms was supported by the impact of the eno1 mutations on ENO activity. In extracts of developing siliques of eno1-1 and eno1-2, activity eluted from the MonoQ column as a single peak, in the same position as peak P1 in extracts of wild-type siliques. The minor peak P2 was absent (Fig. 5d,e). By contrast, the activity in P2 in eno1-3 extracts was strongly enhanced (Fig. 5f) relative to that in wild-type extracts. These data are consistent with loss of the plastidial ENO1 activity in the eno1-1 and eno1-2 mutants, and up-regulation of expression of ENO1 in the eno1-3 mutant. There was relatively little difference between wild-type and all three eno1 mutant lines in the cytosolic ENO activity (P1 peak).

Phenotypic characterization of eno1 mutants

Plants lacking or overexpressing ENO1 (the eno1-1 and eno1-2 mutants or the eno1-3 mutant, respectively) grew at the same rate as syngenic wild-type plants in both controlled-environment and glasshouse conditions, and no difference in vegetative and reproductive morphologies was observed (Fig. S4). Across three, independently grown batches of plants there was no consistent difference in lipid content of mature seeds between the mutants and their corresponding wild-types (Table 1).

Table 1.   Total lipid content of mature seeds of eno1 and pglym1 mutants and the syngenic wild-type (+/+) lines, determined by nuclear magnetic resonance (NMR) spectroscopy
  1. aThree independent batches of plants were analysed. Batch 1 was grown in a controlled-environment room. Batches 2 and 3 were grown in a glasshouse during June–July and October–December, respectively.

  2. bValues are the mean of nine biological replicates for batches 1 and 2 and six biological replicates for batch 3 (±SE).

  3. cNot determined.

  4. dMutant values marked with a Roman numeral are statistically significantly different (Student’s t-test) from syngenic wild-type values within the same batch. (i), = 0.0015; (ii), = 0.0023; (iii), = 0.031; (iv), = 0.039; (v), = 0.032.

eno1-1−/−b35.5 ± 0.4 (i)d0.6 ± 0.3 (iii)36.1 ± 1.4
+/+33.1 ± 0.741.4 ± 0.435.6 ± 0.7
eno1-2−/−35.3 ± 0.640.6 ± 0.4 (iv)35.7 ± 0.6
+/+34.9 ± 0.541.8 ± 0.433.8 ± 0.7
eno1-3See Fig. 4c ndc26.4 ± 0.4Nd
+/+ nd26.0 ± 0.6Nd
pglym1-1−/−36.5 ± 0.4 (ii)38.0 ± 0.7 (v)36.6 ± 0.4
+/+35.3 ± 0.439.7 ± 0.336.1 ± 1.2
pglym1-2−/−35.4 ± 0.738.4 ± 0.835.4 ± 0.8
+/+36.1 ± 0.735.8 ± 1.334.6 ± 0.4

Identification and phenotypic characterization of mutants deficient in plastidial PGlyM

We next investigated the importance for fatty acid biosynthesis of conversion of 3PGA to 2PGA via plastidial PGlyM. A genome-wide search for PGlyM-encoding sequences in Arabidopsis identified six putative PGlyM genes whose products clustered closely to PGlyMs from other sources (Tables S2, S3, Note S1 and Fig. S5). Two of the six putative PGlyMs (At1g22170 and At1g78050) have predicted transit peptides spanning amino acids 1–48 that may confer plastidial localization (Table S3). Interrogation of publicly available microarray data showed that At1g22170 transcript is readily detectable through much of embryo development, including the period of rapid lipid accumulation, and is substantially more abundant than At1g78050 transcript throughout this developmental period (Fig. S6). The subcellular localization of the At1g22170 product (designated PGlyM1) was verified by transiently expressing full-length PGlyM1 fused at its C-terminus to GFP in N. benthamiana leaves. The PGlyM1-GFP fluorescence signal overlapped that of chlorophyll autofluoresence. This indicates that PGlyM1 is a plastidial protein (Fig. 6a). Immunoblot analysis of soluble protein extracts from infiltrated leaves, using an antiserum for GFP, confirmed the accumulation of chimeric protein of the predicted size (Fig. 6b).

Figure 6.

 Subcellular localization of PGLYM1 and identification of T-DNA insertion lines. (a) Transient expression of PGLYM1-GFP in Nicotiana benthamiana. Infiltrated areas were viewed with a confocal microscope for green fluorescent protein (GFP, top panel) and chlorophyll autofluoresence (Chl, in red, middle panel). The panel at the bottom is a merged image of GFP and chlorophyll autofluoresence. Bars, 10 μm. A 35S::GFP vector was used as a control (see Fig. 2). (b) Expression of PGlyM1-GFP chimeric protein in N.benthamiana leaves. Immunoblot analysis using a GFP monoclonal antiserum of extracts from leaves transiently expressing 35S::PGlyM1-GFP and an empty vector (lane C). Positions of molecular mass markers are indicated (mass in kDa) on the left. The asterisk indicates a band of the size expected for the PGlyM1-GFP chimeric protein (c. 60 kDa). Lanes contained 50 μg protein. Equivalent loadings were confirmed by Ponceau S staining of the major protein band on the membranes (bottom panel). (c) Exon-intron structure and T-DNA insertions in the Arabidopsis (Col-0) gene PGLYM1 (At1g22170). Exons, introns and the 5′- and 3′-UTRs are depicted as closed, open and grey boxes, respectively; the solid line represents the promoter region. T-DNA insertions are indicated by inverted arrowheads and are identified by the allele name and the corresponding database nomenclature. (d) RT-PCR analysis of lines carrying the pglym1-1 and pglym1-2 alleles. Total RNA was extracted from leaves of seedlings of mutant (−/−) and the corresponding syngenic wild-type (+/+) plants and subjected to RT-PCR assay with primer pairs flanking the insertion site. The TUBULIN2 gene (At5g62690) was used as a loading control. Controls were run on the same gel at the same time.

To discover the importance of PGlyM1 in seed oil accumulation, two independent T-DNA insertion lines, Salk_087895 (pglym1-1) and SAIL_680_D09 (pglym1-2), were identified. The location of the T-DNA insertions was confirmed by sequencing the LB-FST. The insertions were at -250 and -38 nucleotides relative to the ATG start codons in pglym1-1 and pglym1-2, respectively (Fig. 6c).

The effect of the pglym1 mutations on PGlyM1 function was investigated. First, RT-PCR analysis showed that no full-length PGlyM1 transcript was detected in either line (Fig. 6d), and therefore both pglym1 mutants carry null alleles. Second, the impact of the pglym1 mutations on PGlyM activity was examined by MonoQ anion exchange chromatography (Fig. 7). PGlyM activity in extracts of developing siliques of wild-type plants eluted from the column in two distinct peaks: a major peak at c. 200 mM NaCl, and a minor peak at c. 500 mM NaCl (Fig. 7a). By contrast, activity from leaves eluted as a single peak, in the same position as the first peak from siliques (Fig. 7b). The elution profiles of activity from extracts of siliques of both pglym1 mutants were the same as those from leaf extracts: the second peak of PGlyM activity was absent (Fig. 7c,d). Activity in the major peak was similar to that in wild-type plants on a protein basis. These results are consistent with the idea that the minor peak corresponds to the plastidial PGlyM isoform and that the pglym1-1 and pglym1-2 mutations result in loss of this isoform.

Figure 7.

 Separation of Arabidopsis PGLYM isoforms and effect of pglym1 mutations on phosphoglyceromutase activity. Soluble extracts of leaves or developing siliques were applied to a MonoQ anion exchange column by fast protein and peptide liquid chromatography (FPLC). After washing, the column was eluted with a gradient of increasing NaCl concentration (dashed line). Fractions (1 ml) were assayed for phosphoglyceromutase activity (solid line). The x-axis represents an elution volume of 35 ml. For comparisons of extracts from pglym1 and syngenic wild-type (WT) plants, the same amounts of protein from the two extracts were analysed (3.5 mg). (a) Elution profile of PGlyM isozymes from developing siliques of WT (Col-0) plants. No activity was found in the unbound fraction during sample loading and column washing. The same profile was seen in a further, independent experiment on an extract from a different plant. (b) Elution profile of PGlyM activity from leaves of WT (Col-0) plants. (c) and (d) Elution profiles from siliques of pglym1-1 and pglym1-2 plants, respectively. Recoveries of PGlyM activity as percentages of the activity applied to the column were 87% (a), 89% (b), 82% (c), 77% (d).

Plants homozygous for the T-DNA insertions in PGlyM1 were indistinguishable from syngenic control plants under both controlled-environment and glasshouse conditions (not shown). Concentrations of seed lipids in mature seed of these mutants were very similar to those in wild-type plants (Table 1).


Of the three putative ENO genes in the Arabidopsis genome, only one –ENO1, At1g74030 – encodes a plastidial protein. ENO1 is an active enolase, and it probably accounts for plastidial enolase activity in all parts of the plant, including the developing embryo (Figs 3, 5). Mutant analysis indicates that it may account for c. 15% of the total activity in developing siliques. As reported previously (Lee et al., 2002), ENO2 encodes an active, cytosolic enolase. It probably accounts for all of the activity not attributable to ENO1. ENO3 encodes an enolase-like protein that is inactive when expressed in a soluble, full-length form in E. coli. Differences in active-site amino-acid residues between ENO3 and active enolases probably account for this lack of activity (Fig. 3). It seems likely that the protein has a function in the plant other than as an enolase. It is interesting to note that the ENO2 protein has DNA-binding activity in addition to ENO activity. It is necessary for the cold-induced gene expression involved in acclimation to low temperatures in Arabidopsis plants, probably because it acts as a negative regulator of expression of a cold-induced gene encoding a transcriptional repressor, the zinc-finger domain protein STZ/ZAT10 (Lee et al., 2002). Enolases also function as regulators of gene expression in animals and fungi. For example, human α-enolase binds to the promoter of the c-myc gene and forms part of a complex that controls gene expression. Alternative splicing of the α-enolase gene generates a smaller protein, MBP-1, that lacks ENO activity but retains DNA-binding and complex-formation capacity (Feo et al., 2000; Hsu et al., 2008). It is tempting to speculate that Arabidopsis ENO3 may also have some gene-regulatory function. For PGlyM, a single plastidial isoform (encoded at At1g22170) accounts for most or all of the plastidial activity in developing embryos. The activity and importance of the isoforms encoded by the remaining five PGLYM genes (Table S2) remain to be discovered.

Our data provide strong evidence to support the view that plastids from different organs differ in their capacity to catalyse glycolytic flux from 3PGA to PEP. For both ENO and PGlyM, we observed two peaks of activity after chromatography of silique extracts but only one after chromatography of leaf extracts. Our analyses of mutants lacking putative plastidial isoforms of these enzymes indicate that the peak missing in leaves is plastidial. Chloroplasts are generally reported to lack detectable activities of these enzymes (e.g. pea, Stitt & ap Rees, 1979; spinach, Bagge & Larsson, 1986). Nonphotosynthetic plastids are frequently reported to contain both enzymes (e.g. castor bean endosperm, Miernyk & Dennis, 1982; developing pea seeds, Denyer & Smith, 1988; wheat endosperm, Entwistle & ap Rees, 1988) or to contain ENO, while having extremely low or undetectable activities of PGlyM (e.g. sycamore suspension culture, Frehner et al., 1990; cauliflower buds, Journet & Douce, 1985; pea roots, Borchert et al., 1993).

Our T-DNA insertion lines provided the opportunity to examine the impact of both elimination of plastidial glycolytic capacity (eno1-1, eno1-2 and pglym1) and enhancement of plastidial ENO activity (eno1-3). The mutations in ENO1 did not affect expression and the activity of the cytosolic enolase ENO2 (Figs 4, 5), and hence interpretation of phenotypes is not complicated by changes in the capacity for conversion of 2PGA to PEP in the cytosol. Neither an increase in plastidial ENO activity nor its elimination had any consistent effects on seed lipid content. Similarly, the loss of plastidial PGlyM activity in the pglym1 mutants had no effect on seed lipid content. Thus neither plastidial ENO nor plastidial PGlyM exerts measurable control over the flux of carbon to lipid in the developing embryo. If plastidial conversion of 3PGA to PEP were necessary for normal rates of fatty acid synthesis, a reduction in seed lipid content in eno1-1, eno1-2 and pglym1 mutants, and conceivably an increase in the eno1-3 mutant, would be expected.

The results show that the entire flux of carbon from 3PGA to PEP for fatty acid synthesis can proceed in the cytosol. In the pglym1 mutants, 3PGA generated in the plastid by reassimilation of carbon dioxide via Rubisco must be exported to the cytosol for conversion to 2PGA via cytosolic PGlyM. The 2PGA could then be converted to PEP via cytosolic ENO, or imported into the plastid for conversion via plastidial ENO. In the eno1 mutants, any 2PGA made in the plastid must be exported to the cytosol for conversion to PEP via cytosolic ENO. Movement of 3PGA and 2PGA across the plastid envelope could occur via the TPT (Knappe et al., 2003b; Weber et al., 2005), which has an affinity for 3PGA equivalent to that for triose phosphates, via the glucose-6-phosphate/phosphate translocator (GPT; Kammerer et al., 1998), or via the PPT. Although 3PGA is a poorer substrate for the GPT and PPT than glucose-6-phosphate and PEP, respectively (Kammerer et al., 1998; Knappe et al., 2003a), both GPT and two isoforms of PPT (AtPPT1 and AtPPT2) are expressed to a relatively high level in developing embryos (White et al., 2000; Knappe et al., 2003a). Of course, in wild-type plants, at least part of the flux from plastidial 3PGA to PEP may proceed via plastidial ENO and PGlyM.

Our results suggest a complex partitioning between cytosol and plastid of the flux of carbon from sucrose to fatty acid synthesis in developing oilseed embryos (Fig. 8). Hexose phosphates formed in the cytosol are probably partitioned between cytosolic glycolysis and imported into the plastid for generation via the pentose phosphate pathway of NADPH required for fatty acid biosynthesis and the RuBP substrate for Rubsico. Assimilation of respiratory carbon dioxide via Rubisco generates a plastidial pool of 3PGA. The conversion of 3PGA to PEP can potentially occur in either the plastid or the cytosol. This plasticity is in complete contrast to the next step in the pathway, the conversion of PEP to pyruvate: the strong phenotypes of mutants lacking plastidial pyruvate kinase show that most of the flux occurs in the plastid (Andre et al., 2007; Baud et al., 2007). At least in the eno1 and pglym1 mutants, and potentially in wild-type embryos, all of the 3PGA made in the plastid may be exported to the cytosol for conversion to PEP, and most or all of the PEP must then move back into the plastid for conversion to pyruvate. Our study thus highlights the importance of translocators in the glycolytic flux in oilseed embryos: plastid envelope translocators for 3PGA and PEP must be considered as components of the glycolytic pathway.

Figure 8.

 Subcellular locations of major reactions involved in acetyl CoA generation in developing Arabidopsis embryos. Sucrose entering the embryo cell is catabolized in the cytosol to glucose 6-phosphate (Glc 6-P). Part of the Glc 6-P enters glycolysis in the cytosol and a further portion enters the plastid via the Glc 6-P translocator (4) as substrate for the oxidative pentose phosphate pathway (not shown) and to generate ribulose bisphosphate (RuBP), the substrate for Rubisco. Rubisco reassimilates carbon dioxide lost in respiration (6), producing 3-phosphoglycerate (3PGA) in the plastid. This 3PGA can potentially leave the chloroplast (via any of three translocators) and be converted via cytosolic enolase and phosphoglyceromutase to phosphoenolpyruvate (PEP). Plastidial 3PGA may also be converted to PEP via plastidial phosphoglyceromutase (1) and enolase (2). Cytosolic PEP is imported into the plastid via the phosphoenolpyruvate/phosphate translocator (PPT) (3). Inside the plastid, PEP is converted to pyruvate for acetyl CoA synthesis via plastidial pyruvate kinase (5). Fru 6-P is fructose 6-phosphate.


We are grateful to Dr Matthew Hills (JIC, UK) for seeds of pgylm1-1 and eno1-2, Professor Jian-Kang Zhu (University of California, Riverside, USA) for the kind gift of los2 seed, Dr Bérengère Ize (JIC) for the pUC8 plasmid, Grant Calder and Dr Sylviane Comparot-Moss (JIC) for help with confocal microscopy, and Sheila Mitchell for excellent horticultural support. We thank the NASC (University of Nottingham, UK) and GABI-Kat (MPI Cologne, Germany) Arabidopsis Stock Centers and the Escherichia coli Genetic Stock Center (Yale University, USA) for biological material integral to this study. This work was supported by the Biotechnology and Biological Sciences Research Council (208/P19448) to AMS and a Core Strategic Grant to the John Innes Centre.