Research article

Evidence that ACN1 (acetate non-utilizing 1) prevents carbon leakage from peroxisomes during lipid mobilization in Arabidopsis seedlings

Elizabeth Allen, Annick Moing, Jonathan A. D. Wattis, Tony Larson, Mickaël Maucourt, Ian A. Graham, Dominique Rolin, Mark A. Hooks


ACN1 (acetate non-utilizing 1) is a short-chain acyl-CoA synthetase which recycles free acetate to acetyl-CoA in peroxisomes of Arabidopsis. Pulse-chase [2-13C]acetate feeding of the mutant acn1–2 revealed that acetate accumulation and assimilation were no different to that of wild-type, Col-7. However, the lack of acn1–2 led to a decrease of nearly 50% in 13C-labelling of glutamine, a major carbon sink in seedlings, and large decreases in primary metabolite levels. In contrast, acetyl-CoA levels were higher in acn1–2 compared with Col-7. The disappearance of eicosenoic acid was slightly delayed in acn1–2 indicating only a small effect on the rate of lipid breakdown. A comparison of transcript levels in acn1–2 and Col-7 showed that induced genes included a number of metabolic genes and also a large number of signalling-related genes. Genes repressed in the mutant were represented primarily by embryogenesis-related genes. Transcript levels of glyoxylate cycle genes also were lower in acn1–2 than in Col-7. We conclude that deficiency in peroxisomal acetate assimilation comprises only a small proportion of total acetate use, but this affects both primary metabolism and gene expression. We discuss the possibility that ACN1 safeguards against the loss of carbon as acetate from peroxisomes during lipid mobilization.

  • acetate non-utilizing 1 (ACN1)
  • acetyl-CoA
  • Arabidopsis
  • glyoxylate cycle
  • peroxisome
  • substrate cycle


The process of seed germination has been delineated into three steps, water absorption by the seed coat, the initiation of metabolic processes within the outer cellular layers (the endosperm of oilseeds) and the establishment of metabolic processes within the embryo leading to radical emergence from the seed coat [1]. In the oilseed Arabidopsis, the cotyledons of the embryo contain approx. 90% of the total TAG (triacylglycerol) for carbon nutrition. The endosperm contains approx. 10% of the total seed TAG, degradation of which is essential to kick-start embryo metabolism and permit normal development of the seedling [2]. Prior to seedling emergence the degradation of TAG begins in earnest to feed the rapidly developing seedling and they are completely consumed by the time the cotyledons have spread and greened completely. Fatty acids released from TAGs are transported to glyoxysomes where they are activated to their acyl-CoA esters to enter β-oxidation [3]. The β-oxidation spiral releases acetyl-CoA, which enters the glyoxylate cycle for conversion to organic acids that can be exported to mitochondria for subsequent gluconeogenesis [4] or to fuel respiration [5,6]. However, previous work on Arabidopsis mutants has revealed the enzymes critical for linking β-oxidation to primary carbon metabolism. At least one glyoxysomal isoenzyme of citrate synthase and malate dehydrogenase are essential for lipid catabolism and seedling establishment, whereas ICL (isocitrate lyase) and malate synthase are not [7].

The glyoxylate cycle and its association with β-oxidation were elucidated almost 50 years ago through feeding studies using radiolabelled acetate [8]. For exogenous acetate to be assimilated through the glyoxylate cycle, it requires transport into the peroxisome and activation to acetyl-CoA. Through the characterization of mutants resistant to fluoroacetate [9], we demonstrated previously that the ATP-binding cassette protein COMATOSE [10] and the short-chain acyl-CoA synthetase ACN1(acetate non-utilizing 1)/AAE7 (acyl-activating enzyme 7) [11,12] are the primary factors for acetate transport and activation respectively. COMATOSE appears to transport a wide variety of fatty acid-related compounds, and thus the relative importance of acetate transport remains to be elucidated. The transport and activation of fatty acids by long-chain acyl-CoA synthetases is essential [13,14], but a short-chain/acetyl-CoA enzyme activity would seem superfluous, since it resides outside the normal pathway of fatty acid degradation.

Although acetyl-CoA is a metabolite pivotal to central metabolism as a biosynthetic precursor, defining the roles that acetyl-CoA synthetases play in generating acetyl-CoA remains elusive [15]. Through their characterization of the acetyl-CoA synthetase mutant acs1, Lin and Oliver [16] estimated that this enzyme is responsible for up to 90% of the total ACS (acetyl-CoA synthetase) activity in older seedlings. Accordingly, they postulated that ACS1 may be an important means by which acetate is recycled in Arabidopsis. However, ACS activity also resides within the cytosol, which is greater than either plastid or peroxisome-associated activities [12]. Several objectives toward ascertaining the importance of ACN1/AAE7 in developing seedlings were to determine the effects of eliminating this step on metabolite labelling patterns, metabolite levels and gene expression profiles using the acn1–2 mutant. We observed changes to the labelling of the sink metabolite glutamine and in the levels of both metabolites and transcripts. This suggests an important metabolic role of ACN1/AAE7 during lipid mobilization, which may subsequently alter transcriptional responses. These results highlight the still unknown complexity of carbon utilization and partitioning and its regulation during seedling establishment. We discuss our data as a foundation of mathematical modelling efforts to explain the metabolic differences between acn1–2 and Col-7, which suggest a role of ACN1/AAE7 to prevent carbon loss from peroxisomes during lipid mobilization [17].


Plant material and growth conditions

All seeds were surface sterilized and imbibed in the dark at 4 °C for 3 days before sowing on to agar plates. For all experimental conditions, seeds were germinated at 20 °C at 70 μmol of photons · m−2 · s−1 constant illumination. Standard agar medium plates contained 0.8% agar and half-strength MS salts [18]. The medium, prior to addition of agar and subsequent autoclaving, was adjust to pH 5.7 with 0.1 M KOH. In order to obtain conclusive results regarding the behaviour of the mutant compared with wild-type seedlings from the metabolite and transcript profiling, it was necessary to obtain equally developed seedlings. Owing to the more rapid development of acn1–2 [9] {6–8 h before Col-7 to attain PGS (principal growth stage) 0.7 [19]} we selected batches of seedlings with approximately the same average fresh weight in addition to visually staging seedling development. The average fresh weights of 50 seedlings from plates used for RNA extraction were 8.8±1.2 (n=8) and 8.9±1.3 (n=8) for Col-7 and acn1–2 respectively. All seedling batches used for metabolite quantification and feeding studies were selected to fall within these weights.

[13C]acetate and [13C]glutamine quantification

Seedlings were washed from standard agar medium plates using distilled water, rinsed well and ~1 g of seedlings was bubbled with air in a sterile solution of half-strength MS salts, 20 mM sucrose and 50 mM Mes, pH 5.7. After 1 h, [2-13C]-labelled sodium acetate was added to a concentration of 4 mM and the seedlings incubated for designated lengths of time. Chase experiments were performed by transferring seedlings to an air-bubbled solution of 4 mM unlabelled acetate 2 h after adding the [2-13C]acetate, and continuing the incubations for either 2 or 4 h. At each sampling time, the seedlings were removed from the solution, rinsed thoroughly with distilled H2O and frozen in liquid nitrogen. Metabolites were extracted according to Weckwerth et al. [20] and freeze-dried overnight. The residue was resuspended in 550 μl of an aqueous solution containing 18% (v/v) 2H2O, 1.75 μg/ml streptomycin, 3.5 μg/ml NaN3 and 8 mM 13C-formate, which served as an internal standard. The pH was not adjusted. Standard broadband, proton-decoupled, Fourier transform NMR 13C spectra were acquired on a Bruker AC250 spectrometer operating at 62.9 MHz. The interval between 90° pulses was 2 s, and the spectral width and data points were 16000 Hz and 8 K respectively.

Metabolite quantification

Approx. 0.6 g of seedlings was washed from the surface of the agar Petri dishes with distilled sterile water into a filtration unit. Once the water had passed through, the seedlings were washed in 10 ml more water, weighed and immediately frozen in liquid nitrogen. The time from opening the Petri dish to freezing the sample was approx. 4.5 min. Metabolites were extracted according to Weckwerth et al. [20] and metabolite quantification by 1H-NMR was conducted as described by Moing et al. [21]. Briefly, the dried extracts were resuspended in 400 mM phosphate buffer (pH 6.0) in 2H2O and analysed at 500.162 MHz on a Bruker spectrometer (Bruker Biospin Avance). Special care was taken to allow absolute quantification of the individual metabolites through addition of EDTA sodium salt solution (5 mM final concentration in the NMR tube) to improve the resolution and quantification of organic acids such as malate and citrate, adequate choice of the NMR acquisition parameters (pulse angle 90°, relaxation delay 10 s) and use of an electronic reference calibrated with glucose, fructose, glutamine and glutamic acid sodium salt solutions as described previously [21]. Individual metabolites were identified using published data [21], acquisition of NMR spectra of reference compounds under exact solvent conditions and spiking extracts with reference compounds. They were quantified using the metabolite mode of AMIX software (Bruker Biospin v. 3.5.6) based on the number of protons comprising the corresponding resonance.

Samples for fatty acid measurements were sown on to agar plates and tissue samples harvested after the same length of time under the growth conditions specified above. Samples of 50 seeds or 25 seedlings were collected by forceps, weighed in screw-cap tubes and immediately frozen in liquid nitrogen. Fatty acids were measured using the method of Browse et al. [22]. Acetyl-CoA levels were determined by HPLC according to Larson et al. [23]. Acetyl-CoA extraction efficiency and relative quantification was confirmed using the extraction protocol of Tumaney et al. [24].

Transcript profiling and data analysis

Total RNA was extracted from approx. 0.5 g of PGS 0.7 seedlings. The RNA was sent to the GARNet transcript profiling service at Nottingham University for quality analysis, dye labelling and hybridization to the Affymetrix ATH1 GeneChip™ arrays. Twelve arrays were produced with one target hybridized to each chip. Three replicates of two sample types were prepared of Col-7 and acn1–2. The data pre-processing was carried out at the GARNet transcript profiling service where poor hybridization events were noted and quantification of hybridization intensity was performed. Further data mining was carried out at Bangor University using GeneSight™ version 4.1 (BioDiscovery). The raw expression profiling data are available from the GARNet transcript profiling service ( Microarray expression data and all associated experimental metadata were presented to comply with MIAME (Minimum Information About a Microarray Experiment) guidelines [25].

Identification of differentially expressed genes between acn1–2 and Col-7 was conducted using GeneSight™ version 4.1 (BioDiscovery). Unless stated otherwise, data normalization was performed using the pre-set ‘Affy’ data transformation sequence. A dataset was created in GeneSight™ to identify gene expression ratios between Col-7 and acn1–2 that were consistent throughout the microarrays. First, three groups were created that gave gene expression ratios for each acn1–2 replicate compared with one of the three Col-7 replicates. For example, group 1 comprised expression ratios calculated for every pair between acn1–2 (1) and Col-7 (1), acn1–2 (2) and Col-7 (1), and acn1–2 (3) and Col-7 (1). To do this, signal intensities in this dataset were transformed, except that replicates were not combined as the final step of the pre-set ‘Affy’ transformation process. This kept the data within two categories each containing the minimum three replicates needed to conduct the ANOVA (P<0.05) to establish the set of consistent ratios between acn1–2 and Col-7 (GeneSight™ Users Manual v. 4.1, 2003). This set of 14256 genes represented the core set of genes common to each group from which DE (differentially expressed) genes were determined.

We adopted a combined approach of using a discretionary ‘fold’ cut-off level in conjunction with further statistical tests to determine the group of DE genes. The value at which genes showed DE was first identified based on reiterative hierarchical clustering of genes at various fold-change thresholds. The determined fold cut-off level was used in conjunction with a t test (P<0.05) to identify a statistically confident group of genes that showed DE between acn1–2 and Col-7. For each group, the log2 differences in gene expression of the 14256 genes were calculated between acn1–2 and Col-7 replicates. DE genes were identified in each group using a bootstrapping algorithm [26]. Through a reiterative process of decreasing the fold cut-off value while maintaining 95% confidence limits, we obtained a set of 201 genes showing DE between acn1–2 and Col-7 at a cut-off threshold of 1.4-fold. Twenty of 20 times, hierarchal clustering of this set of genes resulted in complete separation of acn1–2 from Col-7. The hierarchical clustering was employed to determine the optimal number of windows to create 2D-SOMs (two-dimensional self-organizing maps) to explore the structure of the data.


Global acetate consumption appears normal

We had previously reported that the acn1–2 mutant appeared to be less tolerant of acetate during the prolonged exposure provided through germination and emergence on standard agar medium plates containing acetate [9]. We hypothesized that the lack of acetate assimilation resulted in accumulation that would lead to acidification of the cytosol and ultimately death of the seedlings. Accordingly, the relatively high tolerance of the mutant to butyrate compared with acetate reflected the greater importance of ACN1 in activating acetate, and that butyrate may be activated by another enzyme. In previous [2-14C]acetate feeding studies on similarly staged seedlings, substantial label appeared in non-carbohydrate compound classes at near wild-type levels, except for the ethanol insoluble and organic acid fractions [12]. This indicated that acetate metabolism was not completely abolished, but it was unclear as to how much it was reduced in the mutant. The relatively higher level of label in the organic acid fraction seen in the mutant from that study may have come from higher levels of un-metabolized free acetate indicating reduced assimilation. In order to test this hypothesis, we compared the uptake and utilization of acetate by acn1–2 compared with Col-7. Using NMR, we quantified the resonance corresponding to the C-2 carbon of [2-13C]acetate-fed seedlings. It was necessary to conduct the feeding experiment in the presence of sucrose for observable levels of acetate to be taken up. Not only was sucrose promoting uptake, it was also exerting a protective effect on the seedlings. Seedlings fed acetate alone had a barely detectable 13C signature after even 8 h, and seedlings rapidly deteriorated as observed by their tendency to fall apart during harvest. This was not unexpected as the toxic nature of free acetate on Arabidopsis seedlings and plant cells has already been reported [27]. The degree of deterioration at any given time was comparable between mutant and Col-7 during incubation times up to 24 h. Having sucrose present permitted us to extract and estimate free acetate levels for up to 24 h after feeding started. There was no difference in labelled acetate levels between acn1–2 and Col-7 (Figure 1). Isotopically labelled acetate increased from undetectable to maximum levels of 40–50 μmol/g of fwt (fresh weight of tissue) within 2 h. Levels remained elevated for 24 h with Col-7 and acn1–2 having 29±11 and 41±18 μmol/g of fwt respectively. This shows that labelling of the endogenous acetate pool reaches steady state within 2 h, and that we were measuring the capacity of the entire seedling to take up acetate and not just a specific tissue. To test if acetate was being metabolized, samples were chased with isotopically normal acetate and the fate of labelled acetate followed. Acetate had decreased within 2 h to approx. 10 μmol/g of fwt. The fall in labelled acetate was similar for both acn1–2 and Col-7. In the subsequent 2 h, only a small further decrease was observed.

Figure 1 Acetate utilization by acn1–2

Symbols: circles, Col-7; squares, acn1–2. Filled symbols represent acetate levels after [2-12C]acetate chase. Approx. 1 g samples of 3-day-old seedlings (PSG 0.7) seedlings were fed with 4 mM [2-13C]acetate plus 20 mM sucrose in incubation medium for the durations indicated. Chase experiments were performed by removing seedlings at 2 h and transferring them to a solution with 4 mM isotopically normal acetate plus 20 mM sucrose (arrow). Tissue samples were extracted with 80% ethanol. Acetate was quantified using NMR by comparison of the intensities of the 13C resonances of acetate and the internal standard formate. Symbols and error bars represent the means±S.D. respectively of values from three independent experiments. According to Student's t tests, the differences between values pre- and post-chase at 4 h for both genotypes were significant with P<0.01. There were no significant differences between 4 and 6 h for post-chase values for either genotype.

Disruption of the flow of carbon from acetate into glutamine

Under the nitrogen-replete conditions that we used, the predominant sink for label was glutamine, and it appeared in all three carbons, C-2, C-3 and C-4, in the NMR spectra. We determined the levels of glutamine based on each of the three labelled carbon atoms in the same extracts used for acetate quantification in Col-7 and acn1–2 (Figures 2A–2C). In both Col-7 and acn1–2, label in C-2 appeared to reach steady state within 2 h. In Col-7, the amount of label in C-3 and C-4 continued to increase up to 4 h after feeding, but reached steady-state levels in acn1–2 by 2 h. At 4 h, labelling in C-3 and C-4 were maximal in Col-7 (results not shown). By 4 h, there was an approx. 50% reduction in the amount of label appearing in the C-3 and C-4 carbons in acn1–2. There was only a slight reduction in label at position C-2 for acn1–2. The differences in the amount of total label in each carbon and the timing in reaching apparent steady state suggested that the glutamine pool size may be smaller in acn1–2 than in Col-7. Upon chasing with isotopically normal acetate, the levels of label in glutamine remained level or rose in the wild-type over the following 2 h. However, for acn1–2, there was no change or a slight decrease in labelled glutamine post-chase, which showed that the flow of label from acetate to glutamine was affected.

Figure 2 Time-course quantification of glutamine by 13C-labelled carbon resonances

Symbols: circles, Col-7; squares, acn1–2. Darkened symbols represent measured glutamine levels after chasing with [2-12C]acetate (arrow). The experimentation is briefly described in the legend to Figure 1. (AC) C-2, C-3 and C-4 positions of glutamine respectively. The values represent the means±S.D. of three independent experiments. According to Student's t tests, 2 h pre-chase values between genotypes for all C positions were not significant with P<0.05. At 4 h, the pre-chase, between-genotype values were 0.3, 0.06 and 0.07 for C-2, C-3 and C-4 respectively. All post-chase, between-genotype values were significant with P<0.01, except for C-2 at 4 h (P=0.054).

Depression of major metabolite levels in acn1–2

The timing and degree of glutamine labelling suggested that the pool size was less in acn1–2 than in Col-7. To confirm our conclusion, the levels of glutamine and other metabolites were quantified by 1H-NMR. We quantified 27 distinct resonances corresponding to a variety of known and unknown metabolites, including soluble carbohydrates, amino acids and organic acids (Table 1). Acetyl-CoA levels were determined by HPLC [23]. In general, the absolute levels of amino acids quantified on an nmol/g of fwt basis were of the same order as those observed in seeds of the Arabidopsis ecotype Wassilewskija [28] or in seed germination series for other species, such as the legumes soybean and lupin [29].

View this table:
Table 1 Comparison of major metabolite levels in acn1–2 and Col-7

On average, the glutamine level was 70% lower in acn1–2 than in Col-7, which corresponds to the labelling data. Furthermore, there was a depression of primary metabolite levels in acn1–2, which extended to all classes of metabolites that were NMR visible and quantifiable (Table 1). The change was significant at P<0.05 for every compound measured except for fumarate. Individual amino acids in acn1–2 were decreased by 67%, on average, compared with those in Col-7, and ranged from 80% less proline to 59% less leucine. Individual soluble sugars in acn1–2 were on average 63% lower than those in Col-7, with the largest decrease observed in case of rhamnose at 69%. When fumarate was excluded from the calculation, organic acids in acn1–2 were on average 58% lower compared with the levels of corresponding ones in Col-7, ranging from 70% less formate to 40% less citrate. Malate was 50% lower in acn1–2. The unknown compounds were affected similarly as the other classes, with levels between 40 and 60% lower. Fumarate showed a small, but insignificant, increase in the mutant. If one effect of the mutation is to reduce carbon flow through the glyoxylate cycle via ICL, then a potential decrease in succinate levels could have a corresponding effect on fumarate levels. If succinate was not the primary organic acid transported into mitochondria, since ICL activity appears unnecessary in light grown seedlings [6], then fumarate may be a metabolite that is unaffected by the anapleurotic/gluconeogenic functions of the glyoxylate/trichloroacetic acid cycle couple. In contrast with the behaviour of primary metabolites, the concentration of acetyl-CoA was found to be more than 2-fold greater in acn1–2 than in Col-7.

Measurements of fatty acid catabolism

One possibility for lower metabolite levels in acn1–2 may have been due to less carbon entering metabolism from lipid mobilization. We conducted a time-course analysis of the disappearance of eicosenoic acid (C20:1), which is a marker of TAG degradation [30], at three different time points, imbibed seeds and 3 and 5 days post-imbibition (Figure 3). The amount of eicosenoic acid in seeds was similar, suggesting that fatty acid and TAG synthesis and storage are not affected in developing seeds. During germination and seedling growth, virtually all eicosenoic acid was gone in both Col-7 and acn1–2 by day 5. For both genotypes, the decrease in eicosenoic acid was significant between successive time points and the difference in levels between acn1–2 and Col-7 was significant at day 3, despite the seedlings being at a slightly more advanced stage of growth.

Figure 3 Degradation of eicosenoic acid

Fatty acids were extracted from seeds at 4 days after imbibing in the cold (day 0) and 3 and 5 days after transfer to growth conditions. The values represent means±S.D. of five independent samples. According to Student's t tests, for both genotypes the levels were significantly different between days with P<0.01, and only at day 3 were the levels of acn1–2 significantly different from those in Col-7 (P<0.01).

Transcript profiling

A group of differentially expressed genes were identified in each dataset using an algorithm based on the bootstrapping algorithm developed by Kerr and Churchill [26]. Genes were assigned as being either differentially expressed or not, based on the intensity of the replicated spots. For a gene to be identified as differentially expressed between acn1–2 and Col-7, the gene expression values from all three replicated microarrays need to be identified as differentially expressed. We employed a reiterative process of decreasing the fold cut-off value while maintaining 95% confidence limits [26] for DE among replicates. In this manner, we obtained a set of 201 genes showing DE between acn1–2 and Col-7 at a cut-off threshold of 1.4-fold, which represented a statistically confident group of DE genes. Hierarchal clustering of the DE genes showed consistent segregation of seedling samples between mutant and Col-7 (see Supplementary Figure S1 available at and that the optimal number of bins for visualization of the data by 2D-SOM was nine (see Supplementary Figure S2 available at Three (bins 7–9) and six (bins 1–6) of the nine bins contained genes that were up- and down-regulated respectively (Supplementary Figure S3 available at Of the 201 DE genes, 71 were up-regulated in acn1–2, with an increase in expression from 1.8- to 27.7-fold, whereas 130 genes were down-regulated with a decrease in expression ranging from 3.8- to 52.87-fold. All the DE genes are listed according to bin number for which repetitive mapping gave the greatest proportion of instances in which a gene fell into a particular bin (see Supplementary Table S1 available at

The up- and down-regulated sets of genes were analysed for ontological patterns according to the bins assigned by the 2D-SOM [31]. A variety of genes with assigned identity were found in each cluster. Of the genes with assigned biological function based on ontology, Cluster 1 comprised genes involved in developmental processes, protein metabolism, nucleic acid-related processes and signal transduction. Cluster 2 contained genes for protein metabolism, signalling and transcription. Cluster 3 contained genes involved in response to stress or external stimuli, development, signalling and transcription. Genes from Clusters 4 and 5 were examined jointly as these genes clustered together intermittently on repeated 2D-SOM clustering, but no enrichment of a particular type of gene was noted. Similarly, genes in Cluster 6 did not show a bias towards any specific function or process. Analysis of Cluster 7 showed that the majority of genes were involved in responses to stress or external stimuli, signalling and nucleic acid related. The genes of Cluster 8 did not show enrichment in any specific ontology. Analysis of Cluster 9 showed the genes were mainly involved in transcription or development.

We examined more closely the top 20 up- and down-regulated genes (Table 2) following the assumption that the genes that demonstrate the largest change in expression levels between two samples are the pivotal genes around which the difference in biological basis revolves [32]. Of the 20 genes that showed the greatest increase in expression, 16 had a predicted biological function. These included a gene associated with the phytohormone auxin that was induced (At5g13350) as well as a number of genes possibly involved in the regulation of translation or transcription (At3g49460, At1g49920, At5g19920, At2g46780 and At2g17920). Examining the top 20 down-regulated genes in acn1–2, the identities of 16 were assignable. Eight of which were involved with early seedling development, including genes encoding LEA (late embryogenesis abundant) proteins (At2g36640, At1g32560, At4g36600 and At4g21020), a seed storage protein (At5g39110) and a protein associated with embryonic development (At5g62210). Genes associated with hormone regulators of plant development At1g48660 (auxin) and At3g02480 (ABA) and a RING-finger family transcription factor (At4g35480) were also repressed in acn1–2.

View this table:
Table 2 Comparative gene expression of Col-7 and acn1–2 from transcript profiling

The 20 most up- and down-regulated genes and acetate assimilation genes with fold DE greater than 1.4 are shown. The gene descriptions are those given in TAIR10 Genome Release. Fold DE, fold differential expression (ratio of intensity values).

Transcript levels of genes associated with the assimilation of acetate

Since the sequential action of the glyoxylate cycle, trichloroacetic acid cycle and gluconeogenesis is important for assimilating both free acetate and acetyl-CoA from β-oxidation, it was of particular interest to examine what happened with the expression of glyoxylate cycle, anapleurotic and gluconeogenic genes in acn1–2. None of the genes examined showed differential expression between acn1–2 and Col-7 when the 1.4-fold change in gene expression threshold was used in conjunction with a t test (P>0.05). However, certain genes showed DE when a simple threshold cut-off value of ±1.4-fold was applied (Table 2). Two genes, a putative cytosolic malate dehydrogenase (At5g56720) and a PCK (phosphoenolpyruvate carboxykinase) showed an increase in transcript levels in acn1–2 of 2.9- and 2.0-fold respectively. The putative PCK is not the one known to be developmentally important during early seedling establishment [30]. The genes for which transcript levels were lower in acn1–2 were malate synthase (At5g03860) [33], ICL (At3g21720) [6], a malate dehydrogenase-related protein (At3g53910), CSY3 (citrate synthase 3; At2g42790) [34], an aspartate aminotransferase (ASP1, At2g30970) [35] and ACN1/AAE7 (At3g16910) [12] in the range of 5.5- to 1.4-fold. We have shown previously that a partial transcript for ACN1/AAE7 is detectable in the acn1–2 mutant [12]; therefore, its expression in the mutant could be determined.


We have employed a combination of metabolite and transcript profiling to investigate the effects on metabolism and gene expression caused by eliminating the peroxisomal acetyl-CoA synthetase step catalysed by ACN1/AAE7. As other acetyl-CoA synthetase activities are present in Arabidopsis [12,36], it was necessary to determine if the lack of ACN1/AAE7 affected the overall assimilation of acetate. It was evident from our previous studies that acetate metabolism is not greatly compromised in acn1–2 [10,12]. However, this component of acetate assimilation represents the gluconeogenic contribution as shown by the large decrease in [14C]acetate appearing in soluble carbohydrate in the acn1–2 and comatose/acn2 mutants compared with wild type [10,12]. We conclude that the effects that we observe on metabolite levels and gene expression reflect a defect in a relatively narrow, but metabolically important, aspect of acetate assimilation.

One possibility is that it serves to recycle acetate in young seedlings. Several mechanisms for acetate production have been described [37]. For example, cysteine biosynthesis is likely to be a substantial producer of free acetate. In loblolly pine, free cysteine levels go from being essentially undetectable in dry seeds, in contrast with all other amino acids, to become the tenth most abundant free amino acid within several days of radical emergence [38]. Furthermore, the flux through cysteine would be high, since it serves as the route by which sulfur is assimilated into proteins and metabolic intermediates [39] and a variety of sulfur-containing defence compounds [40]. Therefore it is highly likely that cysteine biosynthesis with contributions from protein deacetylation provides a steady and high flux of free acetate within developing seedlings. In our study, the amount of labelled acetate did not change dramatically during the second 2 h of the chase, thereby suggesting that it continues to be produced. We have shown that plant cells have an enormous capacity to accumulate acetate if presented with an exogenous source, but levels of free acetate in untreated seedlings are very low. Thus it appears that Arabidopsis seedlings have a large capacity to turn-over free acetate. From the disappearance of label in acetate, the turnover of the acetate pool is estimated to be approx. 5.4 nmol/s per g of fwt and the entire acetate pool could turn over in approx. 8 s. It must be noted that this rate is an estimate from sucrose-fed seedlings and the rate may be different in non-treated seedlings. However, we previously reported that most free acetate assimilation is ACN1 independent in seedlings not fed with sucrose, and a large cytosolic ACS activity in seedlings may account for acetate activation [12]. Therefore we can conclude that re-assimilation of extra-peroxisomal acetate is not likely to be a major function of ACN1 in young seedlings. This function would not appear to fall on to ACS either, since acs1 does not appear to be affected by exogenous acetate in seedlings less than 10 days old [16].

The key difference between mutant and wild-type that suggests an intra-peroxisomal function in acetate cycling is the difference in acetyl-CoA levels. Greater acetyl-CoA levels in acn1–2 appear counterintuitive for an enzyme that convert acetate into acetyl-CoA. However, these results are consistent with a model of lipid mobilization where ACN1/AAE7 prevents the loss of carbon from peroxisomes as free acetate by reactivating it (Figure 4). Acetyl-CoA cannot cross the peroxisomal membrane [41], but there are a number of acyl-CoA thioesterases present in peroxisomes of Arabidopsis, some of which are active with short-chain substrates or the substrate specificity is currently unknown [42,43]. This situation would be analogous to the increase in acetyl-CoA levels observed for the heterologous expression of the Saccharomyces cerevisiae acetyl-CoA hydrolase in mitochondria of tobacco plants [44]. The explanation in this previous study was that free acetate is exported from mitochondria and reactivated in the cytosol. The lack of acetate reactivation in peroxisomes would not expect to have the severe consequences on plant growth as observed for the acetyl-CoA hydrolase-expressing plants, because sufficient acetyl-CoA would be produced by β-oxidation. Increased levels of cytosolic acetyl-CoA may even promote biosynthetic processes allowing faster development under ideal growing conditions. A similar conclusion was drawn about an aconitase isoenzyme of tomato that was hypothesized to be a factor affecting the distribution of carbon between the trichloroacetic acid cycle and sucrose biosynthesis [45]. Elimination of this aconitase resulted in reduced levels of primary metabolites and an increase in various soluble sugars, indicating a shift in carbon allocation from respiration to biosynthetic processes. The tomato aconitase mutant grew taller and produced greater fruit mass than the corresponding wild-type plant. The lack of ACN1/AAE7 may affect the rate at which acetyl-CoA enters the glyoxylate cycle from fatty acid degradation. The opposing activities of ACN1/AAE7 and a thioesterase, in conjunction with the ability of peroxisomes to import and export acetate, would constitute a true substrate cycle [46]. Metabolic rather than transcriptional control of lipid mobilization processes may explain why the expression of β-oxidation and glyoxylate cycle genes is unaffected even in situations where lipid mobilization is severely diminished [34], and why we observed an effect on the expression of very few metabolic genes. It is known that germination and lipid mobilization are regulated independently [47]. Whereas the molecular mechanisms of germination are quite well defined, the regulation of metabolic processes remains poorly understood.

Figure 4 Models to explain the role of ACN1 preventing the loss of carbon from peroxisomes

(A) The wild-type situation showing the anapleurotic/gluconeogenic function of the glyoxylate cycle. The principle flow of carbon from fatty acids out of peroxisomes is through citrate and malate. (B) Removal of ACN1 permits the exit of acetate from peroxisomes due to the action of a thioesterase, where is it reactivated in the cytosol by an acetyl-CoA synthetase activity. Ac, unesterified (free) acetate; Ac-CoA, acetyl-CoA; Cit, citrate; FA, fatty acids; Mal, malate; TE, thioesterase.

We conducted mathematic modelling to further explore the possibility that ACN1/AAE7 is necessary to prevent acetate leakage from peroxisomes [17]. The aim was to model the situation during lipid mobilization with acn1–2 containing higher levels of acetyl-CoA than Col-7. The assumptions associated with the model were: (i) a steady-state production of acetyl-CoA from lipid mobilization, which is likely to be the case just after seedling emergence [48], (ii) a higher rate of acetyl-CoA production from fatty acid degradation than from extra-peroxisomal sources, (iii) an increase in ACN1 activity during seedling growth, (iv) an extra-peroxisomal source of acetyl-CoA synthetase activity and a peroxisomal thioesterase active with acetyl-CoA, (v) the pool of primary metabolites is fed by fatty acid degradation via steps of the glyoxylate cycle, and (vi) acetyl-CoA can be derived from the metabolite pool. Through extensive varying of parameters to mimic various conditions of metabolic fluxes, we could not obtain the observed differences in acetyl-CoA levels by forming acetyl-CoA exclusively via input from the metabolite pool. In addition, it would not seem likely that a reduced metabolite pool would generate higher acetyl-CoA levels in the mutant. However, we could produce models fulfilling these observations by having acetate leave peroxisomes and being activated by an external acetyl-CoA synthetase as long as this activity was greater than that of ACN1/AAE7. More extensive modelling of acn1–2 metabolism using a targeted label input, such as through labelled fatty acids, at different developmental stages will be required to solidify our conclusions.

From gene expression profiles taken over the course of seedling establishment we have evidence that seedlings gradually attain photosynthetic competence after germination, which is not complete until the first true leaves begin to appear [49]. This suggests that transcriptional programmes promoting development are moderated by some mechanism, which may involve catabolite repression [50] and/or genetic factors. It has been hypothesized that genetic factors present in the embryo that regulate cotyledon expansion may continue to exert their effects well into establishment [51]. Our results show that both development and gene expression are affected by elimination of ACN1/AAE7. When global patterns of gene expression in acn1–2 seedlings are compared with Col-7 at the same stage of development, we see lower transcript levels of genes that are associated with embryogenesis in acn1–2. Subsequently, the down-regulation of genes, such as the LEA proteins, may promote seedling growth like we observe with acn1–2. Genetic mechanisms preventing unimpeded (or causing suboptimal) growth have been demonstrated for the micro-organism Bacillus subtilis [52]. In general, transcript levels of acetate assimilation genes are also lower in acn1–2. This may reflect the lower rate at which fatty acids appear to be degraded in acn1–2 or could be the result of a diminished metabolic signal from a reduced primary metabolite pool. At this time it is not possible to conclude if altered gene expression is a cause or result of perturbed development, but it is evident that metabolic processes involving peroxisomal acetate/acetyl-CoA metabolism are important for normal seedling development. It is likely that ACN1/AAE7 begins to exert its influence on peroxisomal metabolism during imbibition. It is expressed in imbibed seeds, and fluoroacetate, which is activated by ACN1/AAE7, greatly delays germination when seeds are imbibed in the presence of it [12]. In conclusion, our results support the findings of Bender-Machado et al. [44] and Oliver et al. [15] that acetate and acetyl-CoA levels are tightly regulated, and disrupting peroxisomal acetyl-CoA metabolism during germination affects both metabolism and gene expression and appears to have knock-on consequences for germination and seedling development.


Elizabeth Allen prepared the manuscript, prepared all tissue extracts for the metabolite profiling and conducted the analysis of transcript data. Annick Moing directed the metabolite profiling, pre-processed the metabolite data and wrote the sections on the methodology for the metabolite analyses. Jonathan Wattis did the mathematical modelling of the effect of ACN1 on carbon flow in seedlings. Tony Larson quantified the levels of eicosenoic acid and acetyl-CoA. Mickaël Maucourt conducted the metabolite profiling of tissue extracts. Ian Graham and Dominique Rolin interpreted data and contributed to the preparation of the manuscript. Mark Hooks conducted the 13C-acetate time-course feeding experiments and transcript profiling and was overall co-ordinator of the writing of the manuscript.


The Ph.D. studentship funding for Elizabeth Allen came from the Sir Williams Roberts trust provided to Bangor University. The work was funded by the UK Biotechnology and Biological Sciences Research Council [grant number P19408].


We would like to thank Dr Martine Dieuaide at INRA-Bordeaux for helping with the 13C-NMR, the GARNet microarray facility at the University of Nottingham for carrying out the array hybridization and Professor A. Deri Tomos (Bangor University) for aiding in the supervision of Elizabeth Allen. We would also like to acknowledge the contribution of the following people to the modelling effort: Martin Berglund, Hilmar Castro, Anna Lovrics, Tal Pearson, Nikolas Popovich, Jamie Twycross, Maria Grazia Vigliotti and Minaya Villasania.

Abbreviations: 2D-SOM, two-dimensional self-organizing map; AAE7, acyl-activating enzyme 7; ACN1, acetate non-utilizing 1; ACS, acetyl-CoA synthetase; DE, differential expression/differentially expressed; fwt, fresh weight of tissue; ICL, isocitrate lyase; LEA, late embryogenesis abundant; PCK, phosphoenolpyruvate carboxykinase; PGS, principal growth stage; TAG, triacylglycerol


View Abstract