Arabidopsis mutants Atisa1 and Atisa2 have identical phenotypes and lack the same multimeric isoamylase, which influences the branch point distribution of amylopectin during starch synthesis

Summary

The aim of this work was to evaluate the function of isoamylase in starch granule biosynthesis in Arabidopsis leaves. A reverse-genetic approach was used to knockout AtISA1, one of three genes in Arabidopsis encoding isoamylase-type debranching enzymes. The mutant (Atisa1-1) lacks functional AtISA1 transcript and the major isoamylase activity (detected by native gels) in crude extracts of leaves. The same activity is abolished by mutation at the DBE1 locus, which encodes a second isoamylase-type protein, AtISA2. This is consistent with the idea that ISA1 and ISA2 proteins are subunits of the same enzyme in vivo. Atisa1-1, Atisa2-1 (dbe1), and the Atisa1-1/Atisa2-1 double mutant all have identical phenotypes. Starch content is reduced compared with the wild type but substantial quantities of the soluble glucan phytoglycogen are produced. The amylopectin of the remaining starch and the phytoglycogen in the mutants are structurally related to each other and differ from wild-type amylopectin. Electron micrographs reveal that the phytoglycogen-accumulating phenotype is highly tissue-specific. Phytoglycogen accumulates primarily in the plastids of the palisade and spongy mesophyll cells. Remarkably, other cell types appear to accumulate only starch, which is normal in appearance but is altered in structure. As phytoglycogen accumulates during the day, its rate of accumulation decreases, its structure changes and intermediates of glucan breakdown accumulate, suggesting that degradation occurs simultaneously with synthesis. We conclude that the AtISA1/AtISA2 isoamylase influences glucan branching pattern, but that this may not be the primary determinant of partitioning between crystalline starch and soluble phytoglycogen.

Introduction

The major component of starch is the branched glucan, amylopectin. It is responsible for the semi-crystalline nature of starch granules and consists of chains of α-1,4-linked glucose residues connected to one another by α-1,6-linkages (branch points). The distribution of branch points within the amylopectin molecule gives rise to clusters of adjacent unbranched chains, which are viewed as crucial in allowing amylopectin to form a semi-crystalline matrix. Precisely how the starch-synthesizing enzymes determine the molecular architecture of amylopectin is an open question.

Isoamylase-type debranching enzymes (ISA, EC: 3.2.1.68) appear to play a major role in amylopectin synthesis. Mutations affecting isoamylase have been described in maize, rice and barley (Burton et al., 2002; James et al., 1995; Kubo et al., 1999), in Arabidopsis (Zeeman et al., 1998a), and in the unicellular green alga Chlamydomonas reinhardtii (Dauvillée et al., 2000, 2001a; Mouille et al., 1996). All result in striking phenotypes, reducing or abolishing starch synthesis and causing the accumulation of a soluble, more highly branched glucan (phytoglycogen). The residual starch is usually present as small granules, often with an aberrant morphology and often containing altered amylopectin. In some cases, however, amylopectin appears to be unaltered (Burton et al., 2002).

In addition to isoamylase, plants also contain a second type of debranching enzyme, limit-dextrinase (LDA or pullulanase-type, EC: 3.2.1.142). In maize and Arabidopsis, mutations that abolish activity of LDA do not cause phytoglycogen accumulation (Dinges et al., 2003; T. Delatte and S.C. Zeeman, unpublished data). However, evidence suggests that LDA may determine the extent of phytoglycogen accumulation in isoamylase-deficient mutants (Dinges et al., 2003; Kubo et al., 1999).

Models have been proposed that suggest either a direct or an indirect involvement of isoamylase in the synthesis of amylopectin (Mouille et al., 1996; Myers et al., 2000; Zeeman et al., 1998a). The direct model suggests that the synthesis of the correct branching pattern of amylopectin requires debranching enzyme activity and that, in its absence, glucans are elongated and branched to form phytoglycogen. The indirect model suggests that phytoglycogen and amylopectin are distinct products and that soluble phytoglycogen precursors, normally degraded by isoamylase activity, are elaborated into phytoglycogen when isoamylase is lacking. Both models address primarily the existence of phytoglycogen in the isoamylase mutants. More recently it has been suggested that isoamylase might also influence starch granule initiation because the starch remaining in isoamylase-deficient plants is present as numerous small granules (Burton et al., 2002; Bustos et al., 2004). However, the way in which isoamylase deficiency causes the complex phenotypes observed has yet to be fully explained.

Where studied, isoamylase has a native molecular mass of between 350 and 500 kDa (Dauvillée et al., 2001b; Fujita et al., 1999; Ishizaki et al., 1983), indicating a multimeric enzyme. In rice, isoamylase purified from developing endosperm was found to contain a single 83 kDa polypeptide, on the basis of which the enzyme was proposed to be a homomultimer (Fujita et al., 1999). In Chlamydomonas, however, mutations at two loci (STA7 and STA8) affect isoamylase activity and cause phytoglycogen accumulation (Dauvillée et al., 2001a). Mutations at the STA7 locus result in the loss of measurable isoamylase activity, whereas mutations at the STA8 locus reduce isoamylase activity and alter its native molecular mass. Recent evidence from potato also suggests that isoamylase is heteromultimeric and that two isoamylase proteins, ISA1 and ISA2, associate with each other to form a single isoamylase enzyme in vivo (Hussain et al., 2003).

The Arabidopsis genome encodes three isoamylase-like proteins (AtISA1, AtISA2 and AtISA3) together with one LDA protein (AtLDA). All four genes are widely conserved in both monocots and dicots (Hussain et al., 2003). Interestingly, amino acids thought to be required for catalysis are not conserved in the ISA2 proteins, suggesting that they may not be active debranching enzymes (Hussain et al., 2003). In the isoamylase-deficient maize, barley and rice, the ISA1 gene is affected (Fujita et al., 2003; James et al., 1995; Kubo et al., 1999), whereas in the Arabidopsis mutant dbe1-1, the expression of the ISA2 gene is affected (Zeeman et al., 1998a). Here we report a genetic approach to investigate the nature of isoamylase activity and the function of the ISA1 protein in Arabidopsis, and examine the role of this enzyme in starch metabolism.

Results

Mutations in AtISA1 and AtISA2

The three genes encoding isoamylase-like proteins annotated in the Arabidopsis genome are AtISA1 (At2g39930, GenBank accession ATAF2109), AtISA2 (At1g03310, GenBank accession AC005278) and AtISA3 (At4g09020, GenBank accession AC005359). Full-length cDNAs from these genes have been isolated (AY139980, BT000443 and AY091058 respectively). All three proteins contain predicted N-terminal chloroplast transit peptides (ChloroP; Emanuelsson et al., 1999). We obtained a line (SALK_029442) containing a putative T-DNA insertion in the coding sequence of AtISA1 and confirmed the location of the T-DNA within exon 13 using PCR (Figure 1a). Plants homozygous for the T-DNA insert (Atisa1-1) were identified from the segregating population obtained from Salk.

Figure 1. Identification and analysis of mutants at the AtISA1 and AtISA2 loci. (a) Structure of the AtISA1 gene; exons are depicted as closed boxes. Sequence flanking the T-DNA left border (hatched background) indicates that T-DNA insertion in Atisa1-1 occurred in exon 13. (b) RT-PCR on RNA isolated from the wild type (odd-numbered lanes) and Atisa1-1 (even-numbered lanes) using AtISA1-specific primer combinations upstream (lanes 1 and 2), spanning (lanes 3 and 4) and downstream (lanes 5 and 6) of the T-DNA insertion site (shown in a). Lanes 7 and 8 contain control reactions using actin2-specific primers. (c) Structure of the AtISA2 gene (DBE1 locus); the single exon is depicted as a closed box. The single base pair deletion in Atisa2-1 (dbe1-1) is shown. (d) Proteins in crude extracts of leaves from the wild type (lane 1), Atisa1-1 (lane 2), Atisa2-1 (dbe1-1; lane 3) and the Atisa1-1/Atisa2-1 double mutant (lane 4) were separated by native PAGE in gels containing amylopectin. Starch-hydrolysing activities were detected by staining the gel with iodine solution after incubation for 16 h at 20°C in 100 mm Tris-HCl, pH 7.0, 1 mm DTT, 1 mm CaCl₂, 1 mm MgCl₂. D1 and D2 are isoamylase and limit dextrinase respectively, A1–A3 are α- or β-amylases. (e) Proteins in crude extracts of leaves from the wild type (lane 1), Atisa1-1 (lane 2), Atisa2-1 (dbe1-1; lane 3) were separated by SDS-PAGE, blotted onto PVDF membranes and probed with anti-AtISA1 antibodies. M, molecular weight markers, given as kDa on the left. Arrow indicates the position of the ISA1 protein.

We investigated the impact of the T-DNA insertion on expression of the AtISA1 gene using RT-PCR. Primers pairs were designed to amplify mRNA sequences corresponding to positions upstream, downstream and spanning the T-DNA insertion site. The correct mRNA sequences were amplified from the wild type. For Atisa1-1, the upstream and downstream primer combinations also amplified the correct sequences, but no product was obtained for the primer combination spanning the insertion site (Figure 1b). Using a gene-specific primer (5′ of the insertion site) and a T-DNA-specific primer, we amplified a hybrid fragment containing both AtISA1 and T-DNA sequences from mutant but not wild-type RNA (not shown). The hybrid mRNA contains a predicted stop codon 18 bp into the T-DNA-derived sequence, which would result in a truncated protein ending in loop 7 of the conserved (αβ)₈ barrel structure. The RT-PCR product from Atisa1-1 RNA with downstream primers probably represents a hybrid transcript initiated within the T-DNA itself.

The DBE1 locus was mapped to the same chromosomal location as the AtISA2 gene (Zeeman et al., 1998a). RNA gel blot analysis revealed that AtISA2 expression was reduced or abolished in this line (Zeeman et al., 1998a). To determine the nature of the mutation, we amplified the AtISA2 coding sequence and flanking regions (2 kb upstream and 1 kb downstream) from dbe1-1 genomic DNA, using PCR. Sequencing of the amplified fragments revealed a single base pair deletion in the coding sequence of AtISA2 in dbe1-1, compared with amplified wild-type DNA and with the published genomic sequence (Figure 1c). No other differences between dbe1-1 and wild-type sequences were found. The resulting frame shift introduces a stop codon, truncating the C-terminal part of the protein [in loop 8 of the (αβ)₈ barrel]. It seemed unlikely that this mutation would abolish transcription of the gene and we were able to amplify AtISA2 mRNA from dbe1-1 using RT-PCR (not shown). However, it is known that non-sense mutations can destabilize transcripts, possibly explaining why no transcript was detected previously (Zeeman et al., 1998a). Hereafter we refer to dbe1-1 as Atisa2-1.

Mutations in AtISA1 and AtISA2 abolish the same isoamylase enzyme activity

Measurement of isoamylase activity in crude extracts of higher plants is problematic due to the interference of other glucan hydrolysing enzymes. Therefore we used native PAGE to investigate whether the loss of AtISA1 expression resulted in a loss of isoamylase activity. We found that a major debranching enzyme activity present in crude extracts of wild-type leaves was missing in Atisa1-1 (Figure 1d, band D1). This enzyme is a chloroplastic isoamylase (Zeeman et al., 1998a,b). Mutations at the DBE1 locus (encoding AtISA2) also result in the loss of D1 activity (Figure 1d; Zeeman et al., 1998a). One explanation for this result is that both AtISA1 and AtISA2 contribute to the same isoamylase activity, as reported for potato (Hussain et al., 2003).

Given that the AtISA2 protein is likely to be non-catalytic (Hussain et al., 2003; also see Introduction), we investigated the expression of AtISA1 in the Atisa2-1 mutant using RT-PCR. No difference in AtISA1 mRNA abundance was observed between Atisa2-1 and wild-type plants (not shown). However, protein gel blots probed with an antibody raised against an AtISA1-specific peptide revealed that the AtISA1 protein was appreciably reduced in Atisa2-1 (Figure 1e). This suggests that in the absence of wild-type AtISA2, AtISA1 may be unstable. No AtISA1 protein was detected in Atisa1-1, as expected.

We investigated the native molecular weight of the D1 isoamylase using gel permeation chromatography. The D1 activity co-eluted with ribulose 1,5-bisphosphate carboxylase, which is known to have a native molecular mass of around 500 kDa (Figure S1). This is much larger than the predicted size of AtISA1 or AtISA2 (85 and 96 kDa respectively) and is consistent with the idea that the isoamylase D1 is a multimeric protein in vivo. Using protein gel blots probed with the AtISA1-specific antibodies, we confirmed that AtISA1 co-eluted with D1 activity in wild-type extracts (Figure S2a). We also investigated the native molecular weight of the residual AtISA1 protein in Atisa2-1. The very low levels of AtISA1 in this line made its detection impossible with the same conditions used for the wild-type extracts. However, by increasing fivefold the protein loaded the gel and using sensitive chemiluminescent methods, a protein of the correct molecular weight was just detectable (Figure S2b). This protein eluted late from the gel permeation column suggesting that, in addition to being reduced in abundance, AtISA1 has a lower native molecular weight in the absence of AtISA2.

Atisa1-1 and Atisa2-1 have identical phenotypes and accumulate altered glucans

Loss of D1 isoamylase caused by mutations of the AtISA2 gene results in a reduction in starch accumulation and the simultaneous production of soluble phytoglycogen in the chloroplasts of leaf mesophyll cells (Zeeman et al., 1998a). We compared the phenotypes of Atisa1-1, Atisa2-1, the Atisa1-1/Atisa2-1 double mutant and the wild type. First, plants were stained with iodine at the end of the day (Figure 2a). Wild-type plants stained a dark brown colour, typical of amylopectin. Atisa1-1 and Atisa2-1 stained a more reddish-brown (Figure 2a), as did the double mutant (not shown). This is indicative of more highly branched glucan polymers in the mutants than in the wild type. Second, we measured the starch and soluble-glucan contents of the leaves at the end of the photoperiod. We found that Atisa1-1, Atisa2-1 and the double mutant accumulated reduced amounts of starch relative to the wild type, together with large amounts of soluble glucan (Figure 2b).

Figure 2. Starch and soluble glucan content of wild-type, Atisa1-1, Atisa2-1 and Atisa1-1/Atisa2-1 double mutant plants. (a) Wild-type (middle), Atisa1-1 (left) and Atisa2-1 (right) plants were harvested at the end of a 12-h photoperiod, decolourized with hot ethanol and stained for starch with iodine solution. (b) All the leaves of individual wild-type (WT), Atisa-1, Atisa2-1 and Atisa1-1/Atisa2-1 double mutant (DM) plants were harvested and immediately frozen in liquid N₂. Insoluble (upper graph) and soluble (lower graph) glucans were extracted using perchloric acid and measured as described in Experimental procedures. Each bar is the mean ± standard error from five replicate samples. n.d., not detected.

We compared the structures of the starches and phytoglycogens from the single mutants and double mutant with the starch from wild-type plants using a combination of enzymatic digestions and high-performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD). The analyses were repeated on several batches of plants, in each of which wild-type and mutant plants were grown side by side in the growth chamber. There were small variations in the glucan structures between batches of plants, but the differences observed between wild-type and mutant plants were highly consistent. We used two different extraction methods; the first involved extracting the glucans in ice-cold buffered medium and the second involved extraction with perchloric acid (see Experimental procedures). Both types of extraction have been used in earlier studies, but the two have not been previously compared. The results from the two methods were similar for starch, but for phytoglycogen the two extraction methods yielded qualitatively different results, which affects the interpretation of the mutant phenotypes in both this and other studies.

Using ice-cold buffered extraction medium, starch and soluble glucans were extracted from leaves of the wild type, the two single mutants and the double mutant. The chain length distributions of the amylopectin from all three mutant lines were identical to each other, but differed from that of the wild type (Figure 3a,b; Figure S3). The amylopectin from the mutants had a small increase in the relative number of very short glucan chains with a degree of polymerization (dp) between 3 and 8, and a decrease in slightly longer chains (dp 9–16) compared with wild-type amylopectin. The chain length distributions of the phytoglycogen from all three mutant lines were also very similar to each other (Figure 3c; Figure S3). Compared with wild-type amylopectin, the phytoglycogen contained far more very short chains (dp 3–9) and fewer slightly longer chains (dp 11–35) (Figure 3d; Figure S3).

Figure 3. Comparisons of the chain length distributions of glucans extracted from leaves of wild-type, Atisa1-1, Atisa2-1 and the Atisa1-1/Atisa2-1 double mutant plants using ice-cold buffered extraction medium. Glucans were extracted from batches of approximately 50 plants, debranched with Pseudomonas isoamylase, and analysed by HPAEC-PAD. Peak areas were summed and the areas of individual peaks expressed as a percentage of the total. The mean ± standard errors of three independent isoamylase digests are shown. For difference plots, standard errors of the compared data sets were added together. For clarity, only the results of a representative experiment for Atisa1-1 are shown in full. The complete data set is presented in Figure S3. (a) Chain length distributions of amylopectins from the wild type and Atisa1-1. (b) Difference plots derived by subtracting the relative percentage values of wild-type amylopectin from those of Atisa1-1 (black), Atisa2-1 (red) and the Atisa1-1/Atisa2-1 double mutant (yellow) amylopectins. (c) Chain length distribution of phytoglycogen from Atisa1-1. (d) Difference plots derived by subtracting the relative percentage values of wild-type amylopectin from those of phytoglycogen from Atisa1-1 (black), Atisa2-1 (red) and the Atisa1-1/Atisa2-1 double mutant (yellow).

Figure 4 (and Figure S4) shows similar chain-length analyses performed on starch and soluble glucans extracted using perchloric acid. The results for the amylopectin from the mutants and wild type were similar to those obtained with the buffered extraction medium (compare Figures 3b and 4b; see also Figures S3 and S4). The amylopectins from all three mutant lines were similar to each other but differed from that of the wild type. However, the phytoglycogen extracted using perchloric acid contained relatively fewer very short chains and more longer chains compared with the phytoglycogen extracted in the buffered extraction medium. The likely explanation for this difference is that the phytoglycogen is susceptible to some enzymatic degradation during extraction in the buffered medium. This would not occur during extraction in perchloric acid, as enzymes would very rapidly be inactivated. Thus, we suggest that the results obtained using this latter method reflect more closely the chain length distribution of the phytoglycogen in vivo. As mentioned previously, the phytoglycogens from the single mutants and the double mutant were all very similar to one another (Figure 4d; Figure S4). However, it is striking that the magnitude of the alterations in the chain length distribution when the phytoglycogen is compared with wild-type amylopectin is quite small (Figure 4d). When compared with the amylopectin from the mutants, the difference is smaller still (Figure 4e).

Figure 4. Comparisons of the chain length distributions of glucans extracted from leaves of wild-type, Atisa1-1, Atisa2-1 and the Atisa1-1/Atisa2-1 double mutant plants using perchloric acid. Glucans were extracted from four to six individual plants. Equal quantities of starch or phytoglycogen from each plant were pooled, debranched with Pseudomonas isoamylase, and analysed by HPAEC-PAD. Peak areas were summed and the areas of individual peaks expressed as a percentage of the total. The mean ± standard errors of three independent isoamylase digests are shown. For difference plots, standard errors of the compared data sets were added together. Only the results for Atisa1-1 are shown in full. The complete data set is presented in Figure S4. (a–d) As described in Figure 3. (e) Difference plots derived by subtracting the relative percentage values of Atisa1-1 (black), Atisa2-1 (gold) and the Atisa1-1/Atisa2-1 double mutant (cyan) amylopectin from the respective values for phytoglycogen from the same lines.

Analysis of the chain length distribution of branched glucans fails to address whether the distribution of branch points within the molecule is altered (Thompson, 2000). Therefore, we analysed the glucans from wild-type and mutant plants after first digesting the external segments of each polymer using β-amylase. β-Amylase removes maltosyl units from the non-reducing end of glucan chains to within two or three glucosyl positions of a branch point. The resulting β-limit glucans were debranched and their chain length distributions determined. This technique allows the comparison of the relative proportions of A chains (chains that carry no other chains) and B chains (chains that carry one or more other chains via a branch point) and of the lengths of internal B-chain segments (Thompson, 2000). Glucans extracted using both extraction methods (buffered medium and perchloric acid) were analysed in this way and yielded comparable results, both for the amylopectins and the phytoglycogens (Figure 5 compare a and b; compare Figures S5 and S6). This indicates that the modification to phytoglycogen structure during extraction in buffered extraction medium, detected when comparing the chain length profiles (see above; Figures 3 and 4), affects primarily the linear outer chains and not the internal branch point distribution.

Figure 5. Comparisons of the chain length distributions of wild-type and Atisa1-1 glucans after β-amylolysis of the external chains. Glucans were extracted as described in Figures 3 and 4, treated with β-amylase and debranched with both Pseudomonas isoamylase and Klebsiella pullulanase. Results obtained for Atisa1-1, Atisa2-1 and the Atisa1-1/Atisa2-1 double mutant were very similar. Note the differences and changes in scale of the y-axis. For clarity only the Atisa1-1 results are displayed. The complete data sets are presented in Figures S5 and S6. (a) β-Limit chain length profiles of glucans extracted using ice-cold buffered medium. The mean ± standard errors of three independent isoamylase digests are shown. (b) β-Limit chain length profiles of glucans extracted using perchloric acid. Values for a single set of digests is shown. Replicate digests gave very similar results. The abundance of dp 6 varied considerably between individual samples. However, this variability was not consistent between individual samples from a single genotype.

The results from the single mutants and the double mutant were very similar to one another and reveal that both amylopectin and phytoglycogen synthesized in the absence of isoamylase are different in branch point distribution from wild-type amylopectin. First, maltose and maltotriose, representing residual stubs of A-chains, dominated all the distributions but were relatively more abundant in wild-type β-limit amylopectin than in β-limit phytoglycogen from the mutants (Figure 5). This reflects a relative reduction in A-chains in phytoglycogen due to its increased branch frequency compared with amylopectin. Chains longer than maltotriose, representing internal B-chain segments, had the opposite relative abundances, as would be expected. However, the β-limit phytoglycogen was greatly enriched in the shorter residual B-chains, indicating that branch points in phytoglycogen are closer together relative to wild-type amylopectin. The chain length profiles of β-limit amylopectin from the mutants were intermediate between those of wild-type β-limit amylopectin and of β-limit phytoglycogen, However, in qualitative terms the β-limit profiles of amylopectin from the mutants resembled those of phytoglycogen more closely than of wild-type amylopectin. This shows that, in addition to an altered overall chain length distribution, the amylopectin from the mutants also has an increased branching frequency, with branches located closer together than in wild-type amylopectin.

Phytoglycogen structure is modified via degradation in vivo

Analysing the structure of the glucans present at a single time point cannot give a full picture of how these glucans are produced. Therefore, we analysed the pattern of starch and phytoglycogen accumulation throughout the day and night and investigated whether there were any differences in glucan structure at different times. All extracts for these experiments were made using the perchloric acid extraction method described above.

We determined the starch and phytoglycogen content of leaves of Atisa1-1 and Atisa2-1 and the wild type during the diurnal cycle. The results for the two single mutants were identical to each other and similar to those previously reported for Atisa2-1 (Figure 6a; Zeeman et al., 1998a). In the wild type, starch accumulated at a fairly constant rate during the day, reaching a peak value of 8 mg g⁻¹ FW and, after a brief lag, was degraded at a fairly constant rate during the night. In the mutants, starch also accumulated and was degraded at constant rates during the day and night, respectively, but the amount synthesized was less than in the wild type (2 mg g⁻¹ FW; Figure 6a). Phytoglycogen accumulation, however, occurred most rapidly in the first 8 h of the day, after which time the amount of phytoglycogen increased more slowly, if at all. The rate of phytoglycogen breakdown was fastest during the early hours of the night and after 8 h of darkness, most of the phytoglycogen had been metabolized. The total glucan accumulation in the isoamylase mutants was less than that in the wild type.

Figure 6. Starch and soluble glucan contents of wild-type, Atisa1-1 and Atisa2-1 plants. All the leaves of individual wild-type (black symbols), Atisa-1 (grey symbols) and Atisa2-1 (white symbols, no line) plants were harvested and immediately frozen in liquid N₂. (a) Starch (upper graph) and soluble glucans (lower graph) were extracted using perchloric acid and measured as described in Experimental procedures. Each point is the mean ± standard error from six replicate samples. (b) Soluble glucans were extracted from a replicate batch of plants using perchloric acid and maltose in the neutral fraction measured by HPAEC-PAD.

We investigated the cause of the reduced rate of phytoglycogen accumulation late in the day in the isoamylase mutants. First, we measured the rate of photosynthesis in the wild type and Atisa2-1 to see if the rate of carbon assimilation was altered in the mutant. This was not the case: the carbon assimilation rate was the same in both lines throughout the day (not shown). This suggests that the accumulation of phytoglycogen does not generally inhibit metabolic processes in the chloroplast. Second, we illuminated wild type and Atisa1 and Atisa2 mutant plants in a range of different light intensities for the duration of a single 12 h photoperiod to determine whether increasing the rate of carbon assimilation would increase the amount of glucan synthesized. As expected, increasing the light intensity led to increased starch accumulation in the wild type (Figure 7). In the mutants, the amount of starch synthesized also appeared to increase, whereas the amount of phytoglycogen accumulated was restricted and increased marginally, if at all, at higher levels of illumination (Figure 7).

Figure 7. Starch and phytoglycogen accumulation in leaves of wild-type and isoamylase-deficient mutants illuminated at different light intensities. Plants were grown as described in Experimental procedures until 6 weeks old. At the beginning of a photoperiod, plants were shifted to a range of different light intensities. The exact light intensity at the surface of individual mature leaves was measured and these leaves (one per plant) were harvested at the end of the 12-h photoperiod and immediately frozen in liquid N₂. Starch (upper graph) and soluble glucans (lower graph) were extracted with perchloric acid and measured as described in Experimental procedures. Wild type, grey circles; Atisa1-1, open circles; Atisa2-1, open squares; Atisa2-2 (dbe1-2) open triangles. Lines were fitted by eye.

We reasoned that during the period of glucan synthesis, glucan-degrading enzymes could be acting on phytoglycogen, limiting its accumulation through turnover. To test this hypothesis, we first measured the levels of maltose, a known intermediate of starch breakdown in Arabidopsis leaves (Zeeman et al., 2004). The maltose contents increased throughout the day in the isoamylase mutants, whereas in the wild type the levels remained low until the onset of starch degradation at night (Figure 6b). During the night, the amount of maltose in the mutant declined faster than in the wild type, presumably because the glucan reserves are exhausted more rapidly (Figure 6a). Next, we examined the structure of phytoglycogen at different times during the day and night in wild-type and Atisa1-1 plants. The phytoglycogen structure synthesized early in the day (4 h time point) had fewer chains less than dp 11 (and notably very few chains smaller than dp 6) than the phytoglycogen present at the end of the day (12 h; Figure 8c). Together, these data suggest that some external chains are being degraded (e.g. by β-amylase), resulting both in the production of the very short chains of phytoglycogen and the observed maltose. In contrast, the chain length profiles of the amylopectin present in the wild type after 4 h and at the end of the day (12 h) were similar to one another (Figure 8a). Intriguingly, the chain length profiles of the amylopectin from the mutant exhibited small changes similar in nature to that of the phytoglycogen (Figure 8b). The results from both experiments support the idea that glucans in the mutant, especially the phytoglycogen, are subject to the action of degradative enzymes during the day, whereas in the wild type relatively little such daytime glucan degradation occurs.

Figure 8. Comparisons of the chain length distributions of glucans extracted from leaves of the wild type and Atisa1-1 at different times of the diurnal cycle. Glucans were extracted using perchloric acid after 4 h of light (4 h), at the end of the photoperiod (12 h) and after 4 h of darkness (16 h). Values for a single experiment are shown. A second experiment with another batch of plants yielded similar results. (a) Chain length distributions of wild-type amylopectin at 4, 12 and 16 h time points. The difference plots are derived by subtracting the relative percentage values of 4 h from 12 h (filled circles), and of 12 h from 16 h (open circles). (b) Chain length distributions of Atisa1-1 amylopectin at 4, 12 and 16 h time points, as described for (a). (c) Chain length distributions of Atisa1-1 phytoglycogen at 4, 12 and 16 h time points, as described for (a).

We also compared the chain length profiles of glucans extracted from the leaves harvested after 4 h of darkness (16 h) with those extracted at the end of the day. Phytoglycogen contained many more very short chains less than dp 8, compared with the phytoglycogen present at the end of the day, indicating that the phytoglycogen pool as a whole is subject to gradual, progressive degradation. The amylopectin from the wild type, however, showed no such differences suggesting that, during degradation, amylopectin at the surface of the granule is completely broken down, whilst that within the granule remains untouched. Again, results intermediate between wild-type amylopectin and phytoglycogen were obtained with the amylopectin from Atisa1-1, which exhibited a small increase in chains less than dp 8 between the 12-h and 16-h time points.

Tissue-specific patterns of glucan accumulation in Atisa1-1 and Atisa2-1

Leaf starch is synthesized primarily from photoassimilates in the mesophyll cells and previous investigation of Atisa2-1 focussed on this tissue (Zeeman et al., 1998a). We used transmission electron microscopy (TEM) to survey the different cell types in the mutants because other cells, such as the bundle sheath cells surrounding the vasculature (starch sheath; Müller-Röber et al., 1994; Rook et al., 2001), also contain starch. The results for Atisa1-1 and Atisa2-1 were identical. Many glucan structures were visible in the mutants, ranging from starch granules with a normal appearance through to phytoglycogen, including many aberrant granules and, in some cases, material that was difficult to define either as insoluble starch or soluble phytoglycogen (Figure 9). However, a striking, tissue-specific pattern was observed. The majority of mesophyll chloroplasts in the mutants contained very small amounts of starch in the form of tiny granules, often irregular in outline (Figure 10, compare a with b and c). Most of the space was occupied by phytoglycogen, causing the chloroplasts to be distended compared with those of the wild type. In guard-cell chloroplasts of the mutants, more starch and less phytoglycogen were present compared with the mesophyll cells (e.g. Figure 9f). Remarkably, the plastids of epidermal cells in the mutants (both upper and lower epidermis) contained starch but no apparent phytoglycogen and were indistinguishable from the corresponding cells of the wild type (Figure 10f,g). Chloroplasts in the subtending palisade cells of the mutant were typical of the mesophyll. Sieve element companion cells and the adjacent bundle sheath cells also contained starch but no apparent phytoglycogen (Figure 10h,i) and chloroplasts in some of the outermost bundle sheath cells contained predominantly starch, but in the form of many irregular granules (Figure 10i).

Figure 9. Transmission electron micrographs of a range of glucan structures from plastids of Atisa2-1 leaf cells. Leaf samples were harvested 9 h into a 12-h photoperiod and prepared for TEM as described in Experimental procedures. Bars = 200 nm. (a) Epidermal cell plastid. (b, d, e) Chloroplasts from bundle sheath cells. (c, g, h, i, j) Chloroplasts from mesophyll cells. (f) Chloroplast from stomatal guard cell. All the corresponding cell types of the wild type accumulated regular starch granules similar in appearance to those in (a) and (b) and none of the structures shown in (c–j) (see also Figure 10).

Figure 10. Starch- and phytoglycogen-accumulating phenotypes of different cell types in wild-type and Atisa1-1 and Atisa2-1 (dbe1-1) leaves. Leaf samples were harvested 9 h into a 12-h photoperiod and prepared for TEM as described in Experimental procedures. The pattern of starch and phytoglycogen accumulation was the same in Atisa1-1 and Atisa2-1. Bars = 2 μm. (a–c) Chloroplasts from palisade cells. Note the increased size of the mutant chloroplasts. (d–f) Epidermal cell plastids (above) and the chloroplasts from subtending palisade cells (below). pal, palisade cell; epi, epidermal cell. (g–i) Cells surrounding the vasculature. Arrows indicate the chloroplasts of the mutants that contain predominantly starch. sm, spongy mesophyll cell; bs, bundle sheath cell; cc, sieve element companion cell.

We examined the chain length profiles of starch preparations from epidermal peels of wild-type and Atisa2-1 plants to determine whether the apparently normal starch granules in the epidermal cells had amylopectin similar to the wild type. Lower epidermal peels were prepared from plants at the end of the day and the ratio of starch and phytoglycogen determined. The preparation from Atisa2-1 predominantly contained starch (not shown), consistent with the TEM observations. Comparison of the amylopectin chain length distribution of these preparations revealed a difference pattern that was very similar to that obtained for the starch derived from the whole tissue (Figure 11; compare with Figure 3a,b). This suggests that the epidermal starch in Atisa2-1 has an altered amylopectin structure, even though no phytoglycogen is apparent in this tissue.

Figure 11. Comparisons of the chain length distributions of starch purified from leaf epidermal peels of wild-type and Atisa2-1 plants. Starch in the insoluble fraction obtained from peels of the lower epidermis was debranched and analysed as described in Figure 3. The difference plot was obtained by subtracting the relative percentage values of wild type amylopectin from that of Atisa2-1.

Discussion

AtISA1 and AtISA2 are both required for a multimeric isoamylase

The insertion of a T-DNA in SALK_029442 effectively eliminates the expression of the AtISA1 gene and is reflected in the loss of D1 isoamylase activity in crude extracts of leaves (Figure 1b,d). The fact that D1 is also absent in AtISA2 mutants indicates that both AtISA1 and AtISA2 are needed for this activity. The most likely explanation is that AtISA1 and AtISA2 are both components of a heteromultimeric enzyme, as is the case for the orthologous proteins in potato (Bustos et al., 2004; Hussain et al., 2003). This is consistent with the high native molecular mass of D1 (Figure S1). Hussain et al. (2003) argued that, as a result of amino acid substitutions, the Arabidopsis and potato ISA2 proteins were unlikely to retain hydrolytic activity. This conclusion was supported by in vitro analysis of the StISA2 protein, which was inactive when expressed in Escherichia coli. In contrast, the ISA1 proteins retain the consensus sequences for active hydrolases and StISA1 was active when expressed in E. coli. This suggests that in a heteromultimeric enzyme of ISA1 and ISA2, ISA1 would be the catalytic subunit.

In the absence of AtISA2, AtISA1 may or may not be active in vivo. However, our results suggest that AtISA1 is unstable in Atisa2-1, as the protein is dramatically reduced in abundance, even though its transcript is still detected. Thus, we cannot determine if the phenotype of Atisa2-1 is the same as Atisa1-1 because of the reduction of AtISA1 protein or because AtISA1 is inactive (or has altered activity/specificity) in the absence of AtISA2, or both. If AtISA1 were stable and active, then the phenotype of Atisa2-1 might be distinct from that of Atisa1-1. It is possible that the phenotype of the sta8 mutant of Chlamydomonas reflects such a situation. Chlamydomonas possesses a multimeric isoamylase enzyme and mutations at STA8 locus affect its integrity and properties, but do not abolish its activity. Interestingly, the sta8 mutant displays a mild phytoglycogen-accumulating phenotype (Dauvillée et al., 2001a), whereas the sta7 mutant, which completely lacks isoamylase activity, has a severe phenotype (Mouille et al., 1996). However, the gene encoded at the STA8 locus has yet to be identified.

The role of AtISA1 and AtISA2 in starch biosynthesis

The data presented here shed new light on the function of isoamylase and, on balance, favour a direct involvement for the enzyme in determining amylopectin structure. The starch synthesized in the absence of the AtISA1/AtISA2 isoamylase is altered in structure compared with wild-type starch. These differences were not detected in our previous study of the Atisa2-1 mutant due to the relative insensitivity of the methods used at that time (Zeeman et al., 1998a). The overall chain length distribution of amylopectin differs from that of the wild type in essentially the same way as in the isoamylase mutants of maize and rice (Dinges et al., 2001; Wong et al., 2003). Importantly, we find that the branching pattern of the amylopectin in the mutants is also altered compared with wild-type amylopectin, but resembles that of phytoglycogen. In particular, the residual B-chains of the β-limit amylopectin and β-limit phytoglycogen from the mutant have a high proportion of short chains compared with wild-type amylopectin, indicating branch points that are very close together. Thus, the function of the AtISA1/AtISA2 isoamylase may be to hydrolyse preferentially branch points that are very close to other branch points, thereby facilitating the crystallization of amylopectin. If this is the case, the proposed heteromultimeric nature of the ISA1/ISA2 isoamylase may be significant in conferring such structural specificity. It is possible to speculate that the non-catalytic ISA2 subunit could recognize one branch point and facilitate the removal of a nearby branch by the catalytic ISA1 subunit. It is important to note that alterations in the chain length profiles of the residual starch are not always observed in isoamylase-deficient plants (e.g. barley; Burton et al., 2002). In such cases, it would be interesting to determine whether there is any change in the branch point distribution.

Our data suggest that the structural differences between granular starch and soluble phytoglycogen may have been overestimated previously because phytoglycogen is prone to structural alterations both during extraction and in vivo. First, by extracting glucans in perchloric acid to rapidly inactivate enzymes, we obtained chain length profiles for phytoglycogen that were quite different to those from buffer-extracted phytoglycogen. The phytoglycogen extracted with perchloric acid had longer chains on average, indicating that during extraction in buffered medium degradation of the external chains may be occurring. Second, comparison of perchloric acid-extracted phytoglycogen from leaves of the Atisa1-1 harvested at different times of the day reveals that the phytoglycogen present early in the day has longer chains on average than phytoglycogen present later in the day. This could be caused by an increased degree of branching or by shortening of the chains through degradation. The evidence points towards the latter explanation because maltose, a known intermediate of starch breakdown (Chia et al., 2004; Niittyläet al., 2004), accumulates in the leaves of the mutant in a pattern that is reciprocal to the rate of phytoglycogen accumulation, but does not accumulate in the wild type until the onset of starch breakdown. Furthermore, the phytoglycogen accumulated early in the day has few chains shorter than dp 6, consistent with action of branching enzymes, which are not thought to transfer chains shorter than dp 6 (Nielsen et al., 2002; Rydberg et al., 2001). Thus, the relative increase in chains shorter than dp 6 in the phytoglycogen present later in the day is caused by degradation of longer chains.

We suggest that a balance between synthesis and degradation of the outer chains of phytoglycogen is achieved in the mutants during the latter part of the day. The maltose produced by degradation is presumably exported to the cytosol and metabolized (Niittyläet al., 2004). Previous experiments using a ¹⁴C pulse-labelling approach failed to detect phytoglycogen turnover in Atisa2-1 (Zeeman et al., 2002). However, these experiments were performed early in the day when phytoglycogen was still accumulating rapidly.

Interestingly, we also detected small changes in the chain length profile of the mutant amylopectin at different times of the day, indicating that at least some of the starch of the mutant may be more susceptible to modification than wild-type starch (see below). During the night, glucan synthesis stops and degradation commences. The abundance of very short chains in the phytoglycogen from the perchloric acid-extracted night-time samples was even greater than in samples extracted at the end of the day. In fact, the 16-h chain length profile resembled most closely the phytoglycogen extracted using buffered extraction medium, supporting our earlier conclusions about the unsuitability of this extraction method.

A range of glucans accumulates in tissue-specific patterns

Our ultrastructural observations provide further new insights into the phenotype caused by isoamylase deficiency. First, specific cell types in the mutants displayed strikingly different phenotypes. Second, many different glucan structures were apparent within the mutant tissues as a whole.

Localized phytoglycogen and starch accumulation has been observed in the developing endosperm of isoamylase mutant of cereals (Boyer et al., 1977; Burton et al., 2002; Nakamura et al., 1997) but it was not apparent from these studies that particular cell types were behaving differently. In Arabidopsis, some cell types appear to accumulate only starch with no apparent phytoglycogen, some primarily phytoglycogen with small amounts of starch, while others appear to have an intermediate phenotype. An obvious explanation for the cell-type-specific phenotype is that other debranching enzyme isoforms (i.e. ISA3 or LDA) are highly expressed in epidermal and bundle sheath cells and can, to some extent, fulfil the same role as ISA1/ISA2. We are pursuing research to evaluate the function of these proteins in Arabidopsis and thereby test this possibility. However, it is also reasonable to suppose that other aspects of starch metabolism may be different between palisade cells (primary photosynthetic cells), guard cells (photosynthetic, but which synthesize starch at night and metabolize it during the day) and epidermal cells (largely non-photosynthetic). The complement of enzyme isoforms, both biosynthetic and degradative, may differ between cell types, as could the way in which these enzymes are regulated.

The interpretation of the chain length profiles of the glucans must be considered in the light of the range of apparent glucan structures visible by TEM (Figures 9 and 10). The results from whole-leaf preparations may represent an average of many glucan structures. However, our analysis of the starch in preparations of epidermal peels suggests that the granules in epidermal cells, although normal in appearance, are still made of altered amylopectin (Figure 11). This is important because it raises the possibility that factors other than AtISA1/AtISA2 determine whether or to what extent glucan is accumulated in an insoluble form (starch granules) or a soluble form (phytoglycogen).

Partitioning between starch and phytoglycogen

We consider it likely that other enzymes of starch metabolism influence the partitioning between starch and phytoglycogen when the ISA1/ISA2 isoamylase is lacking. Myers et al. (2000) suggested that during the synthesis of starch, crystallization prevents continued modification of the nascent, soluble amylopectin molecules by glucan-modifying enzymes. In this context, it has been suggested that disproportionating enzyme (D-enzyme) may participate in the synthesis of amylopectin, based on the altered starch-phenotype of a D-enzyme-deficient mutant of Chlamydomonas (Colleoni et al., 1999). Although there is no evidence that D-enzyme is involved in normal amylopectin synthesis in Arabidopsis (Critchley et al., 2001), the evidence presented here indicates that the glucans produced in the AtISA1/AtISA2 isoamylase mutants are subjected to appreciable secondary modification.

We suggest that the AtISA1/AtISA2 isoamylase mutants produce glucans with an altered branching pattern which delays, but does not necessarily prevent crystallization. This would render these glucans susceptible to more modification by enzymes that may or may not normally feature in the starch biosynthetic process. The proportion of glucan that remains soluble or crystallizes into granules could depend on the extent of secondary modifications and therefore on which other glucan-metabolizing enzymes are present. Such a process could explain the variation from cell type to cell type and the range of glucan structures we observe. It is also possible that crystallization of the altered amylopectin in the mutants is incomplete, compared with wild-type amylopectin, with a fraction of the glucan still accessible for modification. This might explain why the structure of the amylopectin in Atisa1-1 changes slightly, both as the day progresses and during degradation, whereas that of the wild type does not. Further analyses are underway to evaluate these hypotheses.

Experimental procedures

Plants and growth conditions

T-DNA insertion mutant lines from the Salk Institute (San Diego, CA, USA) were obtained via the Nottingham Arabidopsis Stock Centre (Nottingham, UK). Plants were grown in a controlled environment chamber. Unless otherwise stated the conditions were as follows: constant 20°C, 75% relative humidity and a 12-h/12-h light/dark cycle, with a uniform illumination of 175 μmol photons m⁻² sec⁻¹. Sown seeds were covered with a clear plastic lid and stratified at 4°C for 2 days. Lids were removed after the cotyledons fully emerged (approximately 10 days after sowing). Seeds were sown either directly onto a peat-based potting compost, or germinated first on fine-grade seed compost then transplanted into potting compost after 2–3 weeks.

Molecular methods

DNA was extracted using the GenElute^TM Plant genomic DNA miniprep kit (Sigma, Buchs, Switzerland) and RNA using RNeasy® plant mini kit (Qiagen, Hombrechtikon, Switzerland), according to the manufacturer's instructions. RNA preparations were DNAse-treated to reduce the possibility of contamination. RT-PCR was performed using cMaster^TM RT_plus PCR system kit (Eppendorf, Basal, Switzerland) according to the manufacturer's instructions. The actin2 gene (At3g18780) was used as a constitutively expressed control. Primer sequences for PCR and RT-PCR are given in Table S1.

Native PAGE and size exclusion chromatography

Native PAGE of crude extracts of leaves was performed as described previously (Zeeman et al., 1998b). Separation of native proteins by size exclusion chromatography was carried out at 4°C using a 50-ml Sepharose CL-6B column (1.3 cm diameter × 38 cm length), equilibrated with extraction buffer (Zeeman et al., 1998b). The flow rate was 0.3 ml min⁻¹ and fractions were collected at a rate of 1 every 3 min.

Design of AtISA1 antigen and immunoblot analysis

Antibodies specific to the AtISA1 protein were purchased from Sigma Genosys (Haverhill, Suffolk, UK). A 16-amino acid peptide specific to AtISA1 (CFDWEGDMHLKLPQKD) was synthesized, conjugated to keyhole limpet hemacyanin and used for rabbit immunization. Anti-AtISA1 antibodies were purified by the manufacturer via immunoaffinity purification, using a column containing the covalently bound peptide antigen. The purified antibodies were used to probe protein gel blots of crude extracts or concentrated fractions eluted from the size exclusion chromatograms. Blots were developed using Sigma-Fast^TM (Sigma) or Immun-Star^TM (Bio-Rad, Reinach, Switzerland) alkaline phosphatase reagents.

Carbohydrate measurements and structural analyses

Leaves were stained for starch with Lugol solution (Sigma) after clearing in hot 80% (v/v) ethanol. To measure starch and phytoglycogen, samples (all the leaves of individual plants, unless otherwise stated) were harvested and immediately frozen in liquid N₂. Subsequent steps were conducted at 0–4°C. Samples were homogenized using an all-glass homogenizer in 0.7 m perchloric acid. Soluble and insoluble fractions were separated by centrifugation (3000 g, 15 min, 4°C). Insoluble material including starch was resuspended once in water to remove residual soluble glucans, and at least three times in 80% (v/v) ethanol (20°C), then finally resuspended in water and stored at −20°C. The soluble fraction, including phytoglycogen was adjusted to pH 5 by adding 2 m KOH, 0.4 m MES, 0.4 m KCl. Precipitated potassium perchlorate was removed by centrifugation (20 000 g, 15 min, 4°C) and extracts were stored at −20°C. Total glucan in each fraction was determined by measuring the amount of glucose released by treatment with α-amylase and amyloglucosidase as described previously (Hargreaves and ap Rees, 1988). Maltose in the soluble fraction was determined by HPAEC-PAD as described previously (Critchley et al., 2001).

For the extraction of starch and phytoglycogen for structural analyses, two different methods were used. The first method was that described above for starch and phytoglycogen measurement. For each sample, equal quantities of glucan from four or more individually extracted plants were pooled to yield sufficient material for structural analyses. The second extraction method was similar to that described in Zeeman et al. (1998a). Briefly, leaves (5–10 g) were homogenized using an electric blender in ice-cold buffered extraction medium (100 mm Tris, pH 7.0, 5 mm EDTA). Leaf debris was removed by filtration. Starch was removed from the filtrate by centrifugation and contamination removed by several washes with 1% (w/v) SDS. The cleaned starch was rinsed five times with water. The supernatant of the filtrate was heated rapidly to 95°C for 30 min and precipitated proteins removed by centrifugation (3000 g, 15 min, 4°C). Phytoglycogen in the supernatant was precipitated by adjusting to 75% (v/v) methanol, 1% (w/v) KCl and collected by centrifugation (3000 g, 15 min, 4°C).

Glucan samples (0.1 or 0.2 mg) were boiled for 15 min in water. After isoamylase treatment to debranching glucans (Zeeman et al., 1998a), all phytoglycogen samples and the starch samples extracted using the perchloric acid method were passed through sequential Dowex 50 and Dowex 1 mini-columns (Harley and Beevers, 1963) to remove contaminating proteins and charged compounds. The neutral glucan chains were eluted with four column volumes of water, lyophilized, and re-dissolved in a small volume of water. Starch samples extracted using buffered medium were clean and did not require this treatment. Using these samples we compared untreated with Dowex-treated samples to confirm that the Dowex procedure did not alter the results of the subsequent HPAEC-PAD analyses. For the analysis of starch in the epidermis, peels were immediately frozen in liquid N₂. Samples were homogenized in a ground glass homogenizer in buffered extraction medium (as above). In this case, starch was not separated from other insoluble material, but instead washed once in water, three times in 80% (v/v) ethanol and finally resuspended in water. Starch in the insoluble material was debranched and passed through Dowex columns as described above. The HPAEC gradient used for the separation of chain lengths was as described in Zeeman et al. (1998a).

β-Limit glucans were prepared by incubating glucan samples (0.2 or 0.4 mg) with 100 units of β-amylase (Megazyme, Bray, Ireland) for 2 h at 37°C in 10 mm MES, pH 6.0. The β-limit glucans were precipitated by the addition of 3 volumes of methanol, redissolved in water and then debranched by sequential incubations with isoamylase (Zeeman et al., 1998a) and pullulanase (Megazyme; 1 unit of enzyme for 2 h at 37°C in 10 mm Na acetate, pH 5.2). Subsequent sample treatment and HPAEC-PAD analyses were performed as described above.

Transmission electron microscopy

Transverse sections of leaves were cut with a razor blade and fixed in 3% glutaraldehyde in 0.1 m cacodylate buffer, pH 7.4 for 4 h. Tissue was washed three times in buffer, post-fixed overnight in 1% (w/v) aqueous osmium tetroxide, then dehydrated in an ethanol series of increasing 10% steps and transferred to acetone. The sections were then infiltrated and embedded in epoxy resin (Spurr's, Agar Scientific, Stansted, UK). Ultra-thin sections were cut with a diamond knife, collected on copper grids and stained sequentially with uranyl acetate and Reynold's lead citrate. Stained sections were examined and photographed in a JEOL 1200EX/B electron microscope (JEOL, Welwyn Garden City, UK).

Gas exchange measurements

Gas exchange measurements were made using a CIRAS 1 infra-red gas analyser (PP Systems, Hitchin, UK). Fully expanded, attached leaves of mature plants were carefully placed in the cuvette at the start of the photoperiod and CO₂ uptake rate was recorded at 15-min intervals.

Acknowledgements

We thank Cris Kuhlemeier for his support, Alison Smith and Cathie Martin for valuable discussions, Christopher Ball, Rebecca Alder and Andrew Chapple for assistance in growing the plants and Pierre Haldimann for help in making the measurements of photosynthetic rate. We also thank the Salk Institute Genomic Analysis Laboratory for providing the sequence-indexed Arabidopsis T-DNA insertion mutants. The work was funded by the Swiss National Science Foundation (Grant 3100-067312.01/1) and the Swiss National Centre for Competence in Research (Plant Survival).

Supplementary Material

The following material is available from http://www.blackwellpublishing.com/products/journals/suppmat/TPJ/TPJ2348/TPJ2348sm.htm

Figure S1. Estimation of the native molecular weight of Arabidopsis isoamylase D1.

Figure S2. Estimation of the native molecular weight of the AtISA1 protein in extracts of wild-type and Atisa2-1 plants.

Figure S3. Comparisons of the chain length distributions of glucans extracted from leaves of wild-type, Atisa1-1, Atisa2-1 and the Atisa1-1/Atisa2-1 double mutant plants using ice-cold buffered extraction medium.

Figure S4. Comparisons of the chain length distributions of glucans extracted from leaves of wild-type, Atisa1-1, Atisa2-1 and the Atisa1-1/Atisa2-1 double mutant plants using perchloric acid.

Figure S5. Comparisons of the chain length distributions of wild-type and Atisa1-1 glucans after β-amylolysis of the external chains.

Figure S6. Comparisons of the chain length distributions of wild-type and Atisa-1 glucans after β-amylolysis of the external chains.

Table S1 Primer sequences for PCR and RT-PCR amplifications

Supplementary references