The aim of this work was to identify enzymes that participate in the degradation of transitory starch inArabidopsis. A mutant line was isolated by screening leaves at the end of the night for the presence of starch. The mutant had a higher starch content than the wild-type throughout the diurnal cycle. This accumulation was due to a reduction in starch breakdown, leading to an imbalance between the rates of synthesis and degradation. No reduction in the activity of endo-amylase (α-amylase), β-amylase, starch phosphorylase, maltase, pullulanase or D-enzyme could be detected in crude extracts of leaves of the mutant. However, native PAGE in gels containing amylopectin revealed that a starch-hydrolysing activity, putatively identified as an endo-amylase and present in wild-type chloroplasts, was absent or appreciably reduced in the mutant. This is the first time that a specific enzyme required for starch degradation has been identified in leaves.
Many higher plants accumulate starch in their leaves during the day and degrade it during the subsequent night. The degradation of this transitory starch is poorly understood as there is little information concerning the enzymes involved or the fluxes they catalyse. Isoforms of endo-amylase (α-amylase;EC22.214.171.124), starch phosphorylase ( EC126.96.36.199), disproportionating enzyme (D-enzyme;EC188.8.131.52) and debranching enzyme ( EC184.108.40.206) have been localized within the chloroplasts of several species (Li et al. 1992; Lin et al. 1988b;Okita et al. 1979 ). However, extra-chloroplastic isoforms of these enzymes often constitute most of the activity in leaves, hampering the study of those within the chloroplast.
There is evidence from chloroplast preparations of spinach and pea that transitory starch can be degraded by both hydrolysis and phosphorolysis ( Levi & Preiss 1978;Stitt & ap Rees 1980;Stitt & Heldt 1981). In Arabidopsis, it is likely that hydrolysis dominates over phosphorolysis. The evidence for this view is good. First, the ratio of amylolytic activity to phosphorolytic activity in the chloroplast is 18:1 (Lin et al. 1988b). Secondly, the activity of phosphorylase in chloroplasts is much lower than the rate of starch degradation in Arabidopsis leaves ( Lin et al. 1988a , c). Thirdly, a mutation at the SEX1 locus of Arabidopsis, which eliminates the activity of a glucose transporter in the chloroplast envelope, results in the accumulation of abnormally high levels of leaf starch and significantly reduces starch mobilization at night. ( Trethewey & ap Rees 1994a). It is likely that the role of this transporter is to export glucose generated by the hydrolysis of starch, whilst the products of phosphorolytic activity are converted into triose phosphates and exported via the phosphate translocator ( Trethewey & ap Rees 1994b).
To advance our understanding of starch degradation we sought to identify important steps in the process by isolating and characterizing mutants of Arabidopsis deficient in starch degradation. A similar approach using starchless mutants has proved effective in confirming the pathway of starch synthesis in Arabidopsis leaves ( Caspar et al. 1985 ;Lin et al. 1988a )
Isolation of mutants and genetic analysis
To identify mutants with deficiencies in starch degradation we used a screen based on that described by Caspar et al. (1991) . The M2 generation of an X-ray-mutagenized population of Arabidopsis was grown for three weeks, transferred to darkness for 24 h and then screened for starch by staining with an iodine solution. The leaves of the wild-type contained no starch, but the leaves of a mutant line, designated SZ63, stained darkly for starch ( Fig. 1). Starch also accumulated in the seed coat and the anthers to a greater extent than in the wild-type (data not shown).
The phenotype of this mutant was stable over five generations. We backcrossed SZ63 to the wild-type. Five independent crosses yielded 51 F1 plants with a wild-type phenotype. Segregation of the mutant phenotype was observed in the F2 generation. Of 325 plants analysed, 261 degraded their starch whilst 64 had the mutant phenotype; a ratio of 4.1:1, close to the 3:1 ratio expected for Mendelian segregation of a single, nuclear, recessive mutation. All subsequent experiments were performed using this backcrossed line. A further two sequential backcrosses to the wild-type were performed, each followed by the re-selection of the mutant phenotype in the F2 generation. We observed no differences in the phenotype of the backcrossed lines when compared with the original mutant line.
We have named the gene affected by this mutation SEX4 (starch excess) and the allele in the line SZ63, sex4–1. The mutation was mapped by crossing SZ63 to the multiple marker mapping line N3078 and observing the segregation of the mutant phenotype with the visible characteristics of this marker line in the F2 generation. The mutation was closely linked to the central chromosome 3 marker gl1 (250 plants analysed;sex4+/gl1+ = 145, sex4+/gl1– = 65, sex4–/gl1+ = 41, sex4–/gl1– = 0).
Mutations at the GI1 locus (position 33.3 on chromosome 1) and the CAM1 locus (position 22.8 on chromosome 3) both result in late flowering and starch accumulation in the leaves ( Eimert et al. 1995 ). CAM1 maps to the same chromosome as SEX4. However, starch accumulation in the cam1 mutant only occurs at the onset of flowering and is restricted to photosynthetic tissues. Mutants at the SEX1 locus have a phenotype similar to that of SZ63 ( Caspar et al. 1991 ). The sex1 mutants lack the ability to translocate glucose across the chloroplast envelope, preventing the export of carbon in this form during starch degradation ( Trethewey & ap Rees 1994b). The SEX1 locus was assigned a chromosomal location of 12.2 on chromosome 1 ( Caspar et al. 1991 ), thus the mutation in SZ63 does not lie at this locus. This was confirmed by a complementation test between SZ63 and a sex1 mutant line (TC265). The 96 F1 plants resulting from ten separate, reciprocal crosses all had a wild-type phenotype. Furthermore, chloroplasts isolated from leaves of SZ63 were able to translocate glucose across the envelope (data not shown). This is not the case for sex1 mutants, demonstrating that mutations at the two loci cause different biochemical lesions.
To determine whether the mutant line SZ63 was able to degrade starch, we measured the starch content of the leaves of wild-type and mutant plants during the diurnal cycle. The starch content of wild-type leaves was comparable with that found in previous studies ( Caspar et al. 1991 ;Trethewey & ap Rees 1994a). At the end of the day, SZ63 plants had more than double the starch content of wild-type plants ( Fig. 2). Starch was still turned over in the mutant, but the rate of degradation was half that in the wild-type. At the end of the night, the starch content in SZ63 was significantly higher than at the end of the preceding night (Student’s t test;P≤ 0.05). This imbalance between synthesis and degradation is the likely cause of the high-starch phenotype.
We investigated the effect of the mutation in SZ63 on the rate of growth of the plants. After 6 weeks’ growth, the fresh weight of the shoots of SZ63 was significantly less than that of wild-type shoots ( Fig. 3). Similar results were obtained with the sex1 mutant of Arabidopsis ( Caspar et al. 1991 ), suggesting that the efficient mobilization of transitory starch in a diurnal cycle is important for normal growth. The rate of photosynthesis in wild-type and SZ63 plants was similar under the growth conditions used (data not shown).
Measurement of starch-metabolizing enzymes
We measured the activities of starch-degrading and starch-synthesizing enzymes in crude extracts of leaves of SZ63 and wild-type plants. We measured endo-amylase in two ways, using substrates specific for endo-amylolytic activity. The first assay employed starch azure and was saturated with β-amylase ( Doehlert & Duke 1983). The second used a specific substrate, blocked at its non-reducing end (non-reducing-end-blocked ρ-nitrophenol maltoheptaoside; Megazyme). β-amylase was assayed by measuring the production of maltose from soluble potato starch. Maltase and D-enzyme were measured with maltose and maltotriose as the respective substrates. Maltase can potentially interfere with the assay for D-enzyme as the substrate specificity of maltases can be broad ( Sun et al. 1995 ). However, the low activity of maltase compared with D-enzyme ( Table 1) and the difference in pH optima of the two enzymes (5.2 and 6.8, respectively) makes it likely that any error in D-enzyme activity will be small.
Table 1. Comparison of the maximum catalytic activites of starch-degrading enzymes in crude extracts of leaves of wild-type and mutant SZ63 Arabidopsis
Activity(nmol min–1 g–1 fresh weight)
The values are the means of measurements made on four independent extracts ± SEM.
a Activity is measured as the stimulation of the incorporation of 14 C from [14 C]glucose-1-phosphate by phosphorylase a into glucan as μmol min –1 g–1 fresh weight (Smith, 1988).
We detected no reduction in the activities of the starch-degrading enzymes in crude extracts of leaves of SZ63 compared with those of the wild-type ( Table 1). In fact, the activities of β-amylase, D-enzyme, pullulanase and starch phosphorylase were all statistically significantly higher in SZ63 than in the wild-type (Student’s t test;P≤ 0.01). We detected a small (25%) but significant reduction (P≤ 0.05) in the activity of soluble starch synthase in leaves of SZ63 relative to wild-type, but no difference in the activity of starch-branching enzyme.
Native PAGE and activity staining
We used native PAGE to separate the enzymes of starch degradation. This approach has the advantage of separating different isoforms of each type of enzyme. We used two types of gel; the first contained amylopectin and enabled the detection of starch-hydrolysing enzymes ( Hill et al. 1996 ). The second contained glycogen, and, after incubation with glucose-1-phosphate, allowed the detection of isoforms of starch phosphorylase ( Steup 1990).
Using native gels containing amylopectin, tentative identification of specific enzymes of starch degradation is possible. The products of hydrolysis of amylopectin by different enzymes stain different colours with iodine ( Kakefuda & Duke 1984). For example, endo-amylases create clear colourless bands, β-amylases create brown bands and the action of debranching enzymes results in pale blue bands. We emphasise that identification on this basis alone cannot be considered conclusive.
Five major bands were visible when crude extracts of wild-type leaves were run on amylopectin-containing native gels ( Fig. 4a,b): three putative amylases (A1, A2 and A3) and two putative debranching enzymes (D1 and D2). Numerous other minor bands were visible when the gels were incubated for longer. On gels of crude extracts of SZ63 leaves, we consistently found that the clear colourless band A2 was absent or appreciably reduced. In some gels, a minor brown band migrated close to the A2 band. This band was unaffected by the mutation in SZ63 ( Fig. 4c).
To investigate whether the reduction in the amylase A2 resulted from the accumulation of high levels of leaf starch, we performed three control experiments. First, a sex1 mutant line was used as a control. This line has a similar high-starch phenotype to SZ63, resulting from an inability to export glucose from the chloroplast during starch degradation ( Trethewey & ap Rees 1994b). The A2 band, reduced in SZ63, was present in the sex1 mutant at an intensity equal to that of the wild-type ( Fig. 4c). Second, extracts of wild-type leaves were made in the presence of starch granules purified from SZ63. Starch granules equal to five times the amount found in an equivalent mass of SZ63 leaves were added to the extraction medium prior to homogenization of the wild-type leaves. This had no effect on the extraction of the amylase A2 ( Fig. 4d). Third, equal amounts of wild-type and SZ63 leaves were homogenized together to determine whether there were any specific inhibitors of this amylase in SZ63 tissue. We found no evidence for the inhibition of amylase A2 in this experiment ( Fig. 4e). These controls suggest that the deficiency of this enzyme activity is not a secondary effect of the high-starch phenotype and may be the primary lesion in the mutant line.
The substrate specificity of the amylase A2 was investigated by running native gels as described above, then electroblotting the proteins onto a second gel containing either β-limit dextrin or the chromogenic substrate Red Pullulan (Megazyme; specific substrate for pullulanase). These gels show that the amylase A2 is capable of hydrolysing β-limit dextrin to give a clear colourless band ( Fig. 5a), but has no action upon Red Pullulan ( Fig. 5b). These results suggest an endo-amylolytic action.
Two isoforms of starch phosphorylase were resolved from crude extracts of wild-type Arabidopsis ( Fig. 6). These two bands were not visible in control gels incubated without glucose-1-phosphate. The mobility of the major isoform of phosphorylase (P1) was strongly reduced by glycogen. This is characteristic of cytosolic isoforms, and contrasts with chloroplastic isoforms which have a low affinity for glycogen ( Steup 1990). The band corresponding to the chloroplastic isoform (P2) in extracts of wild-type leaves was very faint. This finding is in agreement with that of Lin et al. (1988b) who determined that only 4% of the total phosphorylase activity in wild-type leaves of Arabidopsis was in the chloroplast. In SZ63, both bands of phosphorylase activity were more intense than in the wild-type, although the cytosolic form still comprised most of the activity. In contrast, the phosphorylase activity in the sex1 mutant did not appear to be elevated compared with the wild-type. In agreement with this, Caspar et al. (1991) found the activity of starch phosphorylase to be slightly lower in crude extracts of the sex1 mutant than in the wild-type.
The different isoforms of soluble starch synthase and of branching enzyme were compared in crude extracts of mutant and wild-type leaves using native PAGE techniques ( Edwards et al. 1995 ;Yamanouchi & Nakamura 1992). Three isoforms of starch synthase and two isoforms of branching enzyme were detected in extracts of the wild-type, none of which was visibly altered in SZ63 (data not shown).
Localization of starch-degrading enzymes
To determine the location of the putative endo-amylase A2 we prepared chloroplasts from wild-type leaves. The starch-degrading enzymes present in chloroplasts were compared with those in the filtered homogenate from which the chloroplasts were isolated. A native gel was loaded such that the homogenate and the chloroplast lanes contained the same activity of the chloroplast marker enzyme, NADP-linked glyceraldehyde-3-phosphate dehydrogenase. By comparing the relative intensity of the bands in the two lanes, it is possible to deduce which enzymes are present in the chloroplast, as the intensity of the bands due to chloroplastic enzymes should be similar in the two lanes. In contrast, the activity of extra-chloroplastic enzymes should be considerably reduced in the lane containing the chloroplast preparation relative to that containing the homogenate. The activity of the putative endo-amylase A2 was associated with the chloroplast, as were the two putative debranching enzyme activities D1 and D2 ( Fig. 7). The intensities of these three bands were similar in the homogenate and the chloroplast lanes on the gels. The activities of the other putative amylases (A1 and A3) were not associated with the chloroplast, as the intensities of these bands were reduced in the chloroplast lane compared with that of the homogenate.
We have shown that the mutant line SZ63 is deficient in an amylolytic activity (A2) present within the chloroplasts of wild-type leaves and provided evidence that this activity is an endo-amylase. Control experiments have shown that this deficiency is not due to the greater quantities of starch present in SZ63 preventing the extraction of the enzyme. Our results do not show whether the mutation in SZ63 is in the gene encoding the endo-amylase A2, and it is possible that the decrease in this enzyme results from a mutation in a different gene. However, the activities of the other enzymes of starch degradation were either unchanged or increased in the mutant, compared with the wild-type. The only decrease was in A2. We therefore argue that the high-starch phenotype of the mutant is due primarily to its lack of the endo-amylase and hence that the enzyme is required for the degradation of transitory starch to proceed at a normal rate. No other mutants have been isolated in which the alteration of a starch-degrading enzyme in leaves has resulted in a change in the metabolism of transitory starch.
The high-starch phenotype of SZ63 is similar to that of the sex1 mutants. This is not surprising as both the chloroplastic endo-amylase and the chloroplast envelope glucose transporter are required for the conversion of starch into glucose in the chloroplast and its export to the cytosol. The chloroplastic endo-amylase could either generate glucose directly, by cleaving the terminal glucose residue from the non-reducing end of a glucan chain, or indirectly by degrading starch into soluble glucans which in turn are degraded into glucose by other enzymes. In either case, the reduction or absence of the chloroplastic endo-amylase would restrict the rate of glucose production in the chloroplast and the effect would be similar to that in a mutant with the capacity to generate glucose but inability to export it to the cytosol.
We found that the starch content of SZ63 leaves at the end of the night was significantly higher than at the start of the preceding day ( Fig. 2). It therefore seems likely that the high-starch phenotype is established gradually by the accumulation of a fraction of the starch made during each day. This contrasts with findings for the sex1 mutant, in which starch accumulation and starch degradation were balanced ( Trethewey & ap Rees 1994a). Despite the high starch content of the leaves, the amount of starch accumulated during the day in SZ63 was less than in the wild-type, indicating a shift in photosynthetic partitioning away from starch. Furthermore, the activity of soluble starch synthase was reduced in SZ63 compared to the wild-type. Similar observations have been made with the sex1 mutant ( Caspar et al. 1991 ;Trethewey & ap Rees 1994a)
Despite the deficiency in chloroplastic endo-amylase, the total endo-amylase activity in the leaves of SZ63 was similar to that of the wild-type. Another endo-amylase has recently been reported from Arabidopsis leaves ( Kakefuda & Preiss 1997), but its location within the cell was not determined. This endo-amylase has a high mobility in native gels and was subject to diurnal variation, distinguishing it from the endo-amylase A2, which exhibits neither of these properties. The total β-amylase activity was appreciably higher in SZ63 than in the wild-type, a characteristic of Arabidopsis mutants with lesions in starch metabolism ( Caspar et al. 1989 ). Our native gels revealed a putative β-amylase (A3), the identity of which was confirmed by our failure to detect this enzyme when blotted into a gel containing β-limit dextrin. However, both our data and those of previous studies ( Monroe & Preiss 1990) indicate that this enzyme is extra-chloroplastic.
Endo-amylases are widely regarded as the only enzymes capable of initiating the degradation of intact starch granules ( Beck & Ziegler 1989). Our data show that SZ63 is capable of some starch degradation, implying either that there is a second endo-amylase present in chloroplasts, or that the mutation reduces rather than eliminates the amylase A2. An alternative hypothesis is that other enzymes may attack starch granules in vivo. There is some evidence that starch phosphorylase from pea chloroplasts is able to degrade carefully prepared, intact starch granules from leaves ( Kruger & ap Rees 1983a). The mutant line SZ63 has a three-fold higher starch phosphorylase activity in its leaves than the wild-type ( Table 1), probably resulting from increased activities of both the cytosolic and chloroplastic isoforms ( Fig. 6). It is possible that in SZ63, the increased phosphorylase activity enables the phosphorolysis of starch partly to compensate for the reduction in hydrolysis.
A mutant population of Arabidopsis thaliana L. (Heynh.), race Columbia, was generated by X-ray mutagenesis. Unless otherwise stated, plants were grown in a growth room at 20°C, 75% relative humidity and a 12 h photoperiod. A coconut-fibre-based compost was used, mixed with 0.25% (v/v) gamma BHC dust insecticide. The quantum irradiance was 90 μmol quanta m–2 sec–1 for the first 3 weeks and 170 μmol quanta m–2 sec–1 thereafter. Plants for enzyme assays and native gels were grown in a greenhouse with minimum temperatures of 15 and 10°C during the day and night, respectively. A peat- and sand-based compost supplemented with Dimlin™ fungicide was used. Wild-type and mutant plants were harvested after 4–5 and 5–6 weeks, respectively. At these ages they were at equivalent developmental stages.
Enzymes were from Boehringer-Mannheim UK (Lewes, East Sussex, UK), except β-amylase and maltase which were from Megazyme (Bray, Ireland), who also supplied pullulan, Red Pullulan™ and the Ceralpha kit. Radiochemicals were from Amersham International (Amersham, Bucks, UK).
Measurement of starch
Samples comprising all the leaves of individual plants (100–300 mg fresh weight) were killed in 20 ml boiling 80% (v/v) ethanol. The insoluble material was homogenized in an all-glass homogenizer and extracted a further three times with 20 ml boiling 80% (v/v) ethanol. The insoluble material was resuspended in water, autoclaved at 121°C (114 kN m–2), and the glucose released after incubation with amyloglucosidase and α-amylase was determined ( Hargreaves & ap Rees 1988).
Preparation of leaf starch granules
Arabidopsis leaves (15–20 g) were harvested at the end of their photoperiod, washed and homogenized using a Polytron blender (Kinematica Gmbh, Luzern, Switzerland) in five volumes of 100 m m 3- (N-morpholino)propanesulphonic acid (MOPS), pH 7.2, 5 m m EDTA, 10% (v/v) ethanediol. The homogenate was filtered through two layers of Miracloth and a 20 μm nylon mesh, centrifuged (10 min, 4°C, 3000 g) and the pellet resuspended in 30 ml of the same medium plus 0.5% (w/v) SDS. The starch was pelleted, washed twice more with SDS-containing medium and then six times with 30 ml deionized water. The intactness of the granules was confirmed by scanning electron microscopy.
Preparation of β-limit dextrin
β-limit dextrin was prepared from potato amylopectin as described by Enevoldsen & Manners (1994). The product was purified from maltose released during the reaction by two sequential precipitations with 75% (v/v) methanol, 1% (w/v) KCl. The resultant pellet was dried and re-dissolved in deionized water. Less than 2% of the material was susceptible to further degradation by β-amylase and contamination with maltose was less than 0.2%.
Enzyme assays were optimized for crude extracts of wild-type Arabidopsis leaves with respect to pH and the concentrations of all the components of the assays. Leaves (0.25–0.5 g fresh weight) harvested midway through the photoperiod, were homogenized at 0–4°C using an all-glass homogenizer.
Endoamylase (α-amylase): Leaves were extracted in 15 m m calcium acetate, pH 6.0. Method 1 ( Doehlert & Duke 1983): the 500 μl reaction mixture contained 50 m m MOPS, pH 6.6, 500 units β-amylase (from barley), 10 mg ml–1 starch azure. After incubation at 25°C, undigested starch azure was precipitated by adding 2.5% (w/v) trichloroacetic acid and removed by centrifugation. The absorbance of the supernatant at 595 nm was measured. Method 2: the specific substrate ‘non-reducing-end-blocked ρ-nitrophenol maltoheptaoside’ (BPNPG7; Ceralpha™ kit from Megazyme) was used in 20 m m sodium acetate, pH 5.6, in accordance with the manufacturer’s instructions.
Pullulanase: Leaves were extracted in 100 m m MOPS pH 7.2, 2 m m MgCl2, 10 m m dithiothreitol (DTT), 10% (v/v) ethanediol, 50 mg ml–1 polyvinylpolypyrrolidone (PVPP). The extract was concentrated fivefold using a Microcon-10 microconcentrator (10 kDa molecular mass cut-off). The 200 μl reaction mixture contained 50 m m MOPS, pH 7.2, 10 m m DTT and 20 mg ml–1 pullulan. After incubation at 30°C, the increase in reducing ends was determined by the dinitrosalicylic acid method of Bernfeld (1955) and compared with maltotriose standards assayed in the same conditions.
β-Amylase: Leaves were extracted in 50 m m sodium acetate, pH 6.0, 5 m m EDTA, 5 m m DTT. The 500 μl reaction mixture contained 50 m m sodium acetate, pH 5.6, 5 m m EDTA, 5 m m DTT and 10 mg ml–1 soluble potato starch. The reaction was stopped by boiling for 10 min. Maltose was hydrolysed by incubation with 2 units maltase (from yeast) for 2 h at 37°C, and glucose was determined.
Starch phosphorylase: Leaves were extracted in 40 m mN-(2-hydroxyethyl)piperazine-N′-(2-ethane sulphonic acid) (HEPES), pH 7.5, 1 m m EDTA. The 1 ml reaction mixture contained 20 m m MOPS, pH 7.0, 20 m m Na2HPO4/KH2PO4, 10 m m MgCl2, 3.4 m m NAD, 1 unit phosphoglucomutase (from rabbit muscle), 1 unit glucose-6-phosphate dehydrogenase (from Leuconostoc mesenteroides), 4 μm glucose-1,6-bisphosphate and 2.5 mg ml–1 amylopectin ( Steup 1990).
Maltase (α-glucosidase) and D-enzyme (4-α-glucanotransferase): Leaves were extracted in 20 m m 2-(N-morpholino)ethanesulfonic acid (MES), pH 6.2. The 500 μl reaction mixture contained 50 m m sodium acetate, pH 5.2, 90 m m maltose for maltase ( Kruger & ap Rees 1983b) and 50 m m MOPS, pH 6.8, and 60 m m maltotriose for D-enzyme ( Takaha et al. 1993 ). Assays were stopped by boiling and glucose was determined.
Glyceraldehyde-3-phosphate dehydrogenase: The 1 ml reaction mixture contained 50 m m glycylglycine, pH 8.2, 80 m m 3-phosphoglyceric acid (3-PGA), 20 m m MgSO4, 10 m m cysteine, 5 m m glutathione (reduced form), 2.5 m m ATP, 0.4 m m NADPH and 2 units 3-phosphoglycerate kinase (from yeast) ( Lin & Stocking 1980).
Soluble starch synthase: Leaves were extracted in 100 m m MOPS pH 7.2, 1 m m EDTA, 1 m m dithiothreitol (DTT), 10% (v/v) ethanediol, 50 mg ml–1 PVPP. The resin method of Jenner et al. (1994) was used, except that the pH was 8.3 and the amylopectin concentration was 18 mg ml–1.
Starch-branching enzyme: Leaves were extracted as described for soluble starch synthase, and the phosphorylase stimulation assay, described by Smith (1988), was used.
Chloroplasts from wild-type Arabidopsis leaves were prepared using a method based on that described by Walker et al. (1987) . All procedures were carried out at 0–4°C. Leaves (10 g) were harvested midway through the photoperiod and chopped using a Polytron blender in 50 ml extraction buffer containing 330 m m sorbitol, 25 m m MES, pH 6.5, 5 m m MgCl2, 2 m m isoascorbate. The homogenate was filtered through a pad of absorbent cotton wool between four layers of cheesecloth, pre-wetted with extraction medium. The filtrate was centrifuged at 80 g for 3 min, the chloroplast-enriched pellet resuspended in 10 ml of extraction buffer, then centrifuged once more at 50 g for 2 min to give a chloroplast pellet.
Soluble proteins from leaves harvested midway through the photoperiod were extracted as described for measuring soluble starch synthase activity. Chloroplast pellets were resuspended in 1 ml of the same extraction medium. Native discontinuous PAGE for detecting starch-hydrolysing enzymes was performed in separating gels containing 7.5% (w/v) acrylamide (30:0.8 acrylamide:bisacrylamide), 375 m m Tris (hydroxymethyl)aminomethane (Tris), pH 8.8 and 0.1% (w/v) potato amylopectin ( Hill et al. 1996 ). The stacking gel contained 3.75% (w/v) acrylamide, 63 m m Tris, pH 6.8. After electrophoresis at 4°C, the gel was washed twice with 50 ml 100 m m Tris, pH 7.0, 1 m m MgCl2, 1 m m CaCl2, 1 m m DTT for 15 min, incubated in 50 ml of this medium overnight at 20°C and stained with aqueous 0.67% (w/v) I2, 3.33% (w/v) KI.
Starch phosphorylase was detected using gels identical to those described above but containing 0.8% (w/v) oyster glycogen in place of amylopectin. After electrophoresis, the gels were washed twice with 50 ml 100 m m sodium succinate, pH 6.0, 0.05% (w/v) soluble starch, incubated in this medium plus 20 m m glucose-1-phosphate overnight at 20°C, and stained with 0.67% (w/v) I2 and 3.33% (w/v) KI ( Steup 1990). Native gels for detecting soluble starch synthase and branching enzyme were as described by Edwards et al. (1995) and Yamanouchi & Nakamura (1992), respectively.
Proteins, separated on amylopectin-containing gels, were electroblotted onto gels containing either 0.1% (w/v) β-limit dextrin or 1% (w/v) Red Pullulan ( Steup & Gerbling 1983). A third gel was placed behind the transfer gel to ensure that proteins with high mobility were detected. After blotting, gels were incubated as described for the amylopectin-containing gels. The β-limit dextrin-containing gels were developed by iodine staining.
We thank Kay Denyer, Lynda Fitzpatrick and Rod Casey for their advice and critical review of the manuscript. We acknowledge the Arabidopsis Seed Stock Centre, Nottingham, UK, for providing the mapping line N3078. This work was funded by the Gatsby Charitable Foundation.