Each of four starch debranching enzymes (DBE) is distinct and highly conserved across the plant kingdom; however, the specific functions of these proteins in carbohydrate metabolism are not well understood. DBEs function in both biosynthesis and degradation of starch, and two have been shown to function as multimers in various quarternary structures that can contain one or more DBE proteins, i.e. ISA1 homomultimers and ISA1/ISA2 heteromultimers. This study characterizes potential functional relationships between the three isoamylase-type DBE proteins (ISA) of Arabidopsis using a comprehensive bioinformatics analysis and promoter fusion approach to determine tissue-, subcellular-, and temporal specificity of gene expression. The results reveal complementary sets of expression patterns, in particular that AtISA1 (known to be involved in starch biosynthesis) and AtISA2 (a non-catalytic polypeptide) are co-expressed in some conditions in the absence of AtISA3 (known to be involved in starch degradation), whereas in other conditions AtISA2 is co-expressed with AtISA3 in the absence of AtISA1 (AtISA2 and AtISA3, but not AtISA1, are co-expressed specially in root columella cells and leaf hydathodes). Thus, AtISA2 may function in starch degradation, in addition to its role in starch biosynthesis. AtISA3 and several other potential regulatory genes, starch metabolic genes, and transcription factors, are specifically induced during cold acclimation; these transcription factors are candidates for involvement of cold-induced changes in starch metabolism. Finally, bioinformatics analysis using MetaOmGraph (http://www.metnetdb.org/MetNet_MetaOmGraph.htm) identifies Arabidopsis genes of unknown function that might be involved in starch metabolism in the cold.
Synthesis and degradation of the α-glucan polymers that make up starch granules is a fundamental aspect of plant physiology. The essential catalytic mechanism for α-glycoside bond hydrolysis has been conserved broadly in the evolution of the α-amylase-related superfamily (CAZy family GH13; Carbohydrate Active Enzymes database: http://www.cazy.org/) and is used in a variety of contexts for catabolism of different α-glucans. In the plant kingdom there are four highly conserved members of this superfamily that catalyse hydrolysis of α-(1→6) glycoside bonds (Hussain et al., 2003). Three of these enzymes are similar to the prokaryotic isoamylases and one is close to the prokaryotic pullulanases. All four proteins are known or predicted to be present in plastids, coincident with the abundant α-glucan homopolymers that constitute starch granules, amylose, and amylopectin in this organelle (Zeeman et al., 1998a, b; Delatte et al., 2005, 2006). This observation suggests a role for α-(1→6) glucosidases in the breakdown of starch. Amylopectin catabolism, in particular, should require such activity because in this polymer about 3–4% of the glycoside bonds are in the ‘branched’ α-(1→6) configuration, with the remaining in the α-(1→4) configuration in ‘linear’ chains. Collectively, therefore, this group of plant enzymes is referred to as starch debranching enzymes (DBE), and the members are further classified as three isoamylase-type DBEs (ISA1, ISA2, and ISA3) and one pullulanase-type DBE (PU1).
The physiological functions of the DBEs are not fully understood, although some of them clearly have the expected catabolic roles. Pullulanase-type DBEs are well known to be involved in starch degradation during germination of monocot seeds (Burton et al., 1999). The pullulanase-type enzyme ZmPU1 has been shown to have a role in starch degradation in maize leaves, as loss of function mutations condition a moderate increase in the starch level at the end of the night (Dinges et al., 2003a). ZmPU1 is not essential for this process, however, as an extended incubation of the ZmPU1 knockout in the dark results in complete depletion of the starch store. In Arabidopsis, the AtISA3 and AtPU1 isoforms have been shown by genetic analysis to be involved in leaf starch degradation during the dark phase of the diurnal cycle (Wattebled et al., 2005; Delatte et al., 2006).
Contrary to these expected catabolic functions, genetic analysis of ISA1 and ISA2 mutants has indicated that both proteins are required for normal starch biosynthesis. Whereas loss of a catabolic function would condition starch excess owing to a deficiency in polymer hydrolysis, mutations that eliminate either ISA1 or ISA2 instead cause a reduction in granular starch content. This effect, most often coupled with the appearance of a soluble glycogen-like α-glucan polymer, phytoglycogen, which is not detected in wild-type plants, has been observed in endosperm of maize (James et al., 1995; Rahman et al., 1998; Dinges et al., 2001), rice (Kubo et al., 1999), and barley (Burton et al., 1999), in potato tubers (Bustos et al., 2004), and in Arabidopsis leaves (Zeeman et al., 1998b; Delatte et al., 2005; Wattebled et al., 2005). In Arabidopsis, this effect is mitigated by changes in other DBEs, specifically in Atisa2/Atpu1 double mutants, starch is severely reduced (Wattebled et al., 2005), while Atisa3/Atpu1 double mutants display a starch excess phenotype (Delatte et al., 2006). PU1 activity can also affect the amount of starch accumulation, in addition to its demonstrated degradative function (Dinges et al., 2003a, b; Wattebled et al., 2005). Thus, the hydrolytic activity of at least some DBEs may contribute to the anabolic pathways leading to starch accumulation. Various hypotheses have been proposed to explain these observations, including (i) a direct role of the DBEs in producing the polymers that form starch granules, (ii) an indirect role in eliminating a metabolic competitor polymer that directs glucan away from the granules and into the soluble phase, and (iii) a role of the DBEs in starch granule initiation (Mouille et al., 1996; Zeeman et al., 1998b; Myers et al., 2000; Burton et al., 2002; Bustos et al., 2004).
Taken together, these data indicate that the DBEs play roles both in the biosynthesis of starch and in its degradation. The metabolic interrelationships among the DBEs are further complicated by the fact that the isoamylase-type enzymes do not necessarily act independently of each other. Unlike the prokaryotic isoamylases, which do not have quarternary structure, the plant isoamylase-type DBEs exist as multimers (Fujita et al., 1999; Dauvillée et al., 2001; Hussain et al., 2003). In some instances, the high molecular weight complexes are homomultimers containing only ISA1, whereas in other instances ISA1 and ISA2 form a heteromultimeric complex; the ratios appear to vary among different species (Fujita et al., 1999; Hussain et al., 2003; Delatte et al., 2005). The ISA2 component of the multimeric complex is likely to have a regulatory function as opposed to a direct catalytic role, because this polypeptide is truncated relative to ISA1, and missing the catalytic domain (Hussain et al., 2003). The existence of various quarternary structure forms of ISA1 and ISA2 DBEs, together with the evidence that DBEs function in both degradation and synthesis of the glucan polymers of starch granules, led us to the general hypothesis that additional complexes of the DBEs exist, and that the complexes can contribute to distinct physiological functions.
The hypothesis that different assembly states of isoamylase-type DBE provide biochemical specificity for catabolic and anabolic functions predicts that different subsets of gene expression patterns should exist among ISA1, ISA2, and ISA3. These patterns should correlate with the physiological state of the plant or a specific tissue. To evaluate this prediction, this study uses a combination of comprehensive bioinformatics analysis and promoter–reporter gene fusion, to characterize the tissue- and temporal-specificity of gene expression at the level of transcription and mRNA accumulation for AtISA1, AtISA2, and AtISA3. Specific pairs of genes are found to be co-expressed as distinct sets, depending on the cell type, developmental stage, or environment. AtISA1 and AtISA2 are expressed together in several cell types and tissues in which AtISA3 expression is not detectable. Conversely, AtISA2- and AtISA3-paired expression is observed in structures, such as leaf hydathodes, trichomes, styles, and root tips, where AtISA1 expression is not detectable. These data are consistent with the hypothesis that different quarternary structure assemblies of distinct complexes involving ISA1, ISA2, and/or ISA3, including ISA2/ISA3 heteromers, mediate different physiological functions that are accomplished by the same α-(1→6) glucosidase enzymatic activity. Furthermore, AtISA3, together with AtISA2 but not AtISA1, is induced during the process of cold acclimation along with a group of other genes including several known to be involved directly in starch metabolism. Finally, AtISA3 does not accumulate throughout the leaf mesophyll, and accumulates in hydathodes and root tips, both sites of IAA synthesis. One possibility is that ISA3 function in starch degradation might in part be indirect and involve plant signalling pathways.
Materials and methods
Plants (Arabidopsis thaliana ecotype Columbia) were grown in Sun Gro soil in 3 in. SQ pots and flats in a plant growth room under constant fluorescent white light at 22 °C for most experiments. For the experiment evaluating the spatial patterns of induction of DBE expression in the cold, plants were grown under short day (SD) conditions (diurnal cycle of 8 h light/16 h dark) at 22 °C until 15 d after imbibition (DAI), whereupon, the plants were treated by coldness (4 °C), bacteria infiltration, or wounding.
Construction of DBE promoter/promoter+target::GFP/GUS fusion vectors
Promoter::GFP/GUS fusion constructs were made for all four DBE genes by cloning the amplified promoter region into a binary vector. In cases where the intergenic regions are short (no more than 1 kb), several constructs containing various promoter regions were made. Promoter+target::GFP/GUS fusion constructs were made to express GFP/GUS fused to the N-terminal region of AtISA1 (or AtISA3), under the control of their promoter region. DBE promoter vectors were transformed into Agrobacterium tumefaciens strain GV3101. Construction of DBE promoter fusion vectors is described in detail under the heading ‘Supplementary Method’ in Supplementary data available at JXB online.
Arabidopsis transformation and selection
Arabidopsis plants were transformed using the floral dip method (Clough and Bent, 1998). Transformed Arabidopsis plants were selected based on bar resistance conferred by the T-DNA. T2 seeds from at least 10 independently transformed lines for each construct (a total of over 100 independent lines) were harvested for GUS or/and GFP screening.
Transgenic T2 seedlings were germinated in soil in pots. For transgenic lines containing promoter::GFP/GUS constructs, plants were harvested at various stages of development. Plant materials from a minimum of three plants were assayed for each independently transformed line; plants or organs from the same line were stained in the same tube. Organs were stained in GUS-staining solution composed of Triton/ethanol stock (Triton X-100:ethanol:water; 1:4:5), 0.5 M pH 7.0 KPO4 buffer, 10 mg ml−1 X-gluc in dimethyl sulphoxide, 0.1 M pH 7.0 potassium ferricyanide, 0.1 M pH 7.0 potassium ferricyanide (5:470:25:2:2) at room temperature overnight (Jefferson, 1987). After destaining in 70% ethanol, organs were fixed overnight in FAA (50% ethanol, 5% acetic acid, 3.7% formaldehyde) and cleared in Herr's buffer, composed of 85% lactic acid, chloral hydrate, phenol, clove oil, and xylene (2:2:2:2:1), under vacuum for 3 weeks (Herr, 1971).
Cytological techniques and microscopy
Staining patterns were observed and documented using an Olympus stereomicroscope in the Bessey Microscopy Facility (Iowa State University, Ames, IA, USA). For light microscopy, GUS-stained plant materials were dehydrated in a graded ethanol series from 30% to 95% and then in 100% ethanol for 2 h at room temperature. A graded resin infiltration used ethanol/LR White (LR White, London Resin Company, London, UK) mixtures of 2:1, 1:1, and 1:2 for 2 h each at room temperature. A final infusion with 100% LR White was performed overnight at 4 °C. Polymerization was done overnight in gelatine capsules at 60 °C. After hardening, the blocks were cut into semi-thin sections (2 μm) using glass knives fitted to a Reichert microtome Ultracut. Sections were stained with 2.0% (w/v) Safranin O for 20 s. After staining, the sections were dehydrated through an ethanol series to 100% ethanol, placed in xylene, and covered with mounting medium (Permount, Fisher) and a cover slip. Slide-mounted sections were viewed using a Zeiss Axioplan compound microscope with a Zeiss AxioCamHRc digital camera (Carl Zeiss, Inc., Thornwood, NY, USA).
Confocal laser scanning microscopy
Young leaves (3rd to 6th) from 15 DAI plants were placed in water on slides and covered with cover glass. Slides were observed and documented under a Leica TCS NT laser scanning microscope system in the Confocal Microscopy Facility (Iowa State University, Ames, IA, USA). The AR/KR laser and 488 nm/568 nm FITC/TRITC wavelengths were used to distinguish GFP signal from auto-fluorescence. Because yellowish, damaged, or dying parts of the leaf showed green colour under the confocal microscope, only the healthy, green leaves were selected to detect GFP. Digital images were processed in Adobe Photoshop 7.0 (Adobe, San Jose, CA, USA).
Qualitative comparison of leaf starch content
Wild-type seedlings grown under continuous light were boiled in 50 ml 80% (v/v) ethanol and stained in fresh I2/KI solution (10 g KI, 1 g I2 in 1.0 l water) for 5 min, rinsed in water for 1–2 h. Pictures were taken immediately.
MetaOmGraph (MOG; http://www.metnetdb.org/MetNet_MetaOmGraph.htm; Wurtele et al., 2007) was used to analyse expression patterns of DBE and starch-related genes. The experimental data and metadata from 70 experiments comprising 956 Affymetrix ATH1 microarray slides were obtained from two online microarray depositories: NASCArrays (http://affymetrix.arabidopsis.info/narrays/experimentbrowse.pl; Craigon et al., 2004) and PLEXdb (http://www.plexdb.org/; Shen et al., 2005). The data were normalized to the same range by scale normalization and biological replicates were averaged to yield 424 samples (Mentzen, 2006; Wurtele et al., 2007). These normalized data are available online in MOG.
Informatics-based co-expression of DBE transcripts
In vitro studies of potato isoamylases (Hussain et al., 2003) and also genetic analysis of Arabidopsis isoamylases mutants (Delatte et al., 2005) indicate that AtISA1 and AtISA2 may function as a dimer. An independent line of evidence consistent with the co-function of AtISA1 and AtISA2 could be the co-expression of these genes in specific cells, and across particular developmental stages and environmental perturbations. Furthermore, such expression patterns might also provide information about the function of this complex. To examine DBE expression across a wide variety of conditions, Arabidopsis DBE mRNA accumulation was evaluated in a co-normalized public microarray dataset (Mentzen, 2006). A high correlation coefficient of the mRNA accumulation profile across multiple conditions may reflect a high correlation in function. mRNA accumulation patterns were determined across 72 microarray experiments (over 1000 individual microarrays), representing a wide range of developmental conditions, and environmental and genetic perturbations using MOG (http://www.metnetdb.org/MetNet_MetaOmGraph.htm; Wurtele et al., 2007), a software developed to plot and analyse large sets of data (Fig. 1A). Pearson correlation coefficients of accumulation of the DBE mRNAs were determined (Table 1).
In these experiments, each DBE mRNA accumulation pattern was evaluated independently for its correlation with the 22 746 mRNAs represented on the Affymetrix Arabidopsis ATH1 chip (Fig. 1A). The microarray data analyses indicate a high co-expression of AtISA1 (At2g39930) with AtISA2 (At1g03310); the Pearson correlation coefficient between the two genes is 0.74, and AtISA2 is the most closely correlated gene to AtISA1 among the other 22 745 genes on the ATH1 chip (see Supplementary Table S2 at JXB online). AtISA2 and AtISA3 (At4g09020) are also co-expressed, with a Pearson correlation coefficient of 0.58. Under most conditions, accumulation of AtPU1 (At5g04360) mRNA is much lower than that of AtISA.
Figure 1A shows that AtISA1 and AtISA2 are co-expressed under specific conditions and/or plant tissues or stages of growth (e.g. the first two true leaves), but not all conditions. Similarly, AtISA2 and AtISA3 are co-expressed under some conditions (e.g. siliques with seeds stages 4 and 5). Under some conditions all three AtISA genes are co-expressed. The mRNA accumulation profiles of AtISA genes of Arabidopsis plants grown under SD conditions also show that AtISA1 and AtISA2, and AtISA2 and AtISA3 are co-expressed over the diurnal cycle (L Li, C Foster, ES Wurtele, unpublished results). Both AtISA1 and AtISA2 mRNAs have a large accumulation peak in the middle of the light period and a smaller peak during the dark period. AtISA1 mRNA accumulation decreases gradually after its peak in the middle of the light. By contrast, AtISA2 mRNA maintains a high accumulation until the onset of darkness and then decreases dramatically. AtISA3 has a single peak of accumulation at the beginning of the dark phase. The overlap of high accumulation of AtISA2 and AtISA3 transcripts near the beginning of the dark period is consistent with a possible co-function of these proteins. A similar timing of co-expression of AtISA1 with AtISA2, and AtISA2 with AtISA3 also occurs in plants grown under a longer period of light (12 h light/12 h dark) (microarray data of Smith et al., 2004).
Informatics-based co-expression of DBE transcript with other Arabidopsis genes
The microarray data analysis indicates a strong correlation of the expression of certain DBE genes with others in the Arabidopsis genome. Starch biosynthesis and degradation is a highly regulated process in which multiple enzymes and regulatory genes are involved. Many of these are not yet identified, or the functions are not fully understood. Genes with similar physiological functions, such as participation in the same biochemical pathway might have highly correlated mRNA accumulation profiles. Transcriptional co-expression analysis has been used to identify genes that are involved in complex metabolic processes and provide indications of function (e.g. Gibon et al., 2004; Fatland et al., 2005; Koo et al., 2006; Osuna et al., 2007). Thus, correlation analyses of transcript profiles could help identify genes which are functionally related. Further bioinformatics analysis using MOG (Wurtele et al., 2007) identifies that among the 22 746 genes represented on the Affymetrix Arabidopsis ATH1 chips, there are 30 genes whose expression is highly correlated with all four DBE genes (Pearson correlation coefficient ≥0.50) (see Supplementary Table S3 at JXB online). Of these 30 genes, 12 are implicated in starch metabolism. These are: a putative sugar transporter (At4g04750), AtBAM7 (At4g00490), AtISA2, AtPU1, AtMEX1 (At5g17520), AtAAM3 (At1g69830), AtGWD1 (At1g10760), AtGWD3 (At5g26570), AtPHS1 (At3g29320), AtPHS2 (At3g46970), AtDPE1 (At5g64860), and AtDPE2 (At2g40840). Except for the putative sugar transporter and AtISA2, these genes have previously been found to be involved in starch degradation (Tetlow et al., 2004; Delatte et al., 2005; Lu et al., 2005; Smith et al., 2005; Wattebled et al., 2005).
AtISA3 transcript levels show particularly high Pearson correlation coefficient values with five genes known to be involved in starch degradation, specifically AtSEX4 (At3g52180), AtPHS2, AtGWD1, AtDPE2, and AtGWD3 (values of 0.82, 0.81, 0.80, 0.79, and 0.76, respectively) (Fig. 1B; Table 2) (Yu et al., 2001; Lu and Sharkey, 2004, 2006; Baunsgaard et al., 2005; Niittylä et al., 2006). The co-expression of AtISA3 with genes that encode starch-degradative enzymes is consistent with what Smith et al. (2004) observed during a 12-h-light/12-h-dark diurnal cycle. The sixth highest correlated gene is AtBE3 (At2g36390) (Pearson correlation coefficient 0.74). So far this BE is known to function only in biosynthesis (Dumez et al., 2006), as opposed to the degradative function of the other five genes most highly correlated with AtISA3. After this group of six there is a drop off in the correlation coefficients to a value of 0.65 for the next most highly correlated gene (see Supplementary Table S2 at JXB online).
Starch metabolic genes are highlighted in green, heat shock genes in grey highlight, and cold-regulated genes in blue. Correlation coefficients are highlighted: ≥0.85, maroon; ≥0.75, orange; ≥0.65, yellow; ≥0.55, light yellow.
Spatial expression patterns of DBE genes
The co-expression patterns revealed in microarray data analyses of AtISA1 with AtISA2, and AtISA2 with AtISA3, led us to investigate the spatial expression of the DBE genes in more detail. Recombinant DNA constructs were made that harbour the GFP/GUS reporter genes under the control of the AtISA1, AtISA2, AtISA3, or AtPU1 promoter regions (Fig. 2). AtISA1 and AtISA3 both have long intergenic regions upstream of the translational start site, and in these instances approximately 1.5 kb was included in the promoter fusion constructs ISA1p and ISA3p (Fig. 2A, C). AtISA2 and AtPU1 have relatively short intergenic regions upstream of their coding sequences. Two promoter fusions were made for AtISA2: ISA2p1 containing the intergenic region up to the terminus of the upstream gene At1g03300 (0.89 kb) and ISA2p2 containing the preceding 1.5 kb that includes some of the At1g03300 coding sequence (Fig. 2B). For AtPU1, promoter construct PU1p1 contained only the intergenic region (0.30 kb), PU1p2 contained the complete coding sequence of the upstream gene At5g04350 (0.73 kb), and PU1p3 extended further beyond the upstream gene to contain a total of 1.41 kb (Fig. 2D).
These DBE promoter::GUS constructs were transformed into Arabidopsis. For each of the seven constructs, at least 10 independent transgenic lines in the T2 generation were screened for GUS expression. The lines were assessed at various stages of development from germination to senescence using at least five plants per independent transgenic line. Plants were grown under continuous light for most of these developmental studies in order to avoid possible fluctuations in the level of GFP/GUS accumulation associated with the diurnal cycle. The GUS accumulation data reported here was reproducibly observed in the great majority of the independent transgenic lines. The patterns of GUS accumulation were indistinguishable in the transgenic lines containing either of the two AtISA2 promoter::GUS constructs ISA2p1 or ISA2p2 (Fig. 2B). Thus, a fully functional promoter for AtISA2 is located within the upstream 878 bp sequence. The level of GUS signal from the longer promoter ISA2p2 was slightly lighter, however, than the shorter promoter construct (data not shown), and data from this shorter construct ISA2p1 is presented here.
None of the three AtPU1 promoter::GUS constructs (a total of 30 AtPU1 T2 independent transgenic lines were tested) directed detectable GUS expression in the plant under any conditions studied, so the developmental function of these putative promoters could not be evaluated. A possible explanation is that the functional AtPU1 promoter region encompasses more than the 1.41 kb present in the longest tested promoter fusion PU1p3 (Fig. 2D). Another consideration is that AtPU1 is expressed at the level of steady-state mRNA content to a lower degree than any other DBE (Fig. 1A), possibly to the extent that the AtPU1 promoter does not support expression of GUS to a detectable level.
Under continuous light, expression of AtISA2 and AtISA3 is first evident at 2 DAI in hypocotyls and root tips, prior to detection of any AtISA1 signal (Fig. 3A; see Supplementary Table S4 at JXB online). AtISA1 expression first is detected at 3 DAI in the hypocotyl and cotyledon (data not shown). AtISA2 and AtISA3 expression is also detected earlier than that of AtISA1 in seedlings grown under a SD diurnal cycle (data not shown). Iodine staining of seedlings shows starch is first detected in root columella cells at 1 DAI (data not shown), in root columella cells and hypocotyl at 2 DAI (Fig. 3U), and in cotyledon hydathodes at 3 DAI (Fig. 3U). The expression of AtISA2 and AtISA3, is spatially, and temporally coincident with starch accumulation in these regions. Indeed, the first accumulation of starch in the developing seedlings is in root columella cells (Fig. 3U).
As the seedlings grow, AtISA1 and AtISA2 are co-expressed in mesophyll cells of cotyledons and emerging leaves (Fig. 3B), and later in the shoot apex (Fig. 3C); this pattern is maintained throughout development. In addition, both AtISA1 and AtISA2 are co-expressed in the vasculature in cotyledons (Fig. 3B, D), rosettes (Fig. 3E, F), and cauline leaves (data not shown) throughout expansion and maturation except that as leaves mature, AtISA1 expression in petiole and major veins was reduced (Fig. 3D, F).
To reveal the expression of AtISA genes in the leaf vasculature and other tissues better, semi-thin cross-sections were made from LR White-embedded GUS-stained samples (Fig. 3G, H). AtISA1 and AtISA2 expression is localized within vascular bundles in all differentiating and differentiated cells of both xylem and phloem, and in bundle sheath cells (Fig. 3G, H). AtISA1 has particularly strong expression in the bundle sheath cells (Fig. 3G, H).
Although AtISA1 and AtISA2 expression is often coincident in the cotyledon and leaf tissues, this is not always the case. AtISA1 is expressed throughout the leaf epidermis, whereas AtISA2 does not appear to be expressed in these cells (Fig. 3G). Similarly, throughout leaf development AtISA2 is expressed in whole petiole (Fig. 3B), the base of the main vein (Fig. 3E), in hydathodes (Fig. 3E, F), and in trichomes (Fig. 3I), whereas AtISA1 expression is conspicuously absent from all of these tissues.
AtISA3 expression in leaves differs from that of AtISA1 or AtISA2. By contrast to the general distribution of AtISA1 and AtISA2 throughout much of the leaf, AtISA3 expression is more limited, occurring in the cotyledon tip, the leaf edge, and associated vasculature (Fig. 3B, D–F). Analysis of the GUS activity pattern in leaf mid-sections also fails to detect AtISA3 expression, whereas AtISA1 and AtISA2 expression is clearly observed in mesophyll, epidermis, and vasculature cells (Fig. 3G, H). AtISA3 is expressed along with AtISA2 in leaf hydathodes and trichomes, both of which lack AtISA1 expression (Fig. 3B, E, F, I).
Because AtISA2 and AtISA3 are strongly co-expressed in hydathodes, and the histology of Arabidopsis hydathodes has not been described, DBE-hydathode associated expression was examined in more detail. In addition to the passive release of water by transpiration, many plants actively secrete water at hydathodes (termed guttation) (Esau, 1977), as a result of positive pressure built up in roots during the night. Arabidopsis hydathodes have been shown to secrete salt and amino acids in guttation droplets (Pilot et al., 2004). Figure 3K shows that Arabidopsis has hydathodes with characteristic ‘active hydathode’ structures including vein endings (vein/vascular tissues that are typical of some types of hydathodes that contain xylem but not phloem; Esau, 1977) and tightly packed epithem cells (epithems are sometimes chloroplast-free, aerenchymatous parenchyma; Mauseth, 1988), surrounded by a sheath and one or more water pores (stomates that are usually permanently open, not capable of opening and closing movements; Esau, 1977). In ‘active’ type hydathodes, water passes through the epithem cells toward the pore, and minerals may be reabsorbed (mediated by transfer cells) before the water is forced out through the pore(s) (Esau, 1977). AtISA2 and AtISA3 are expressed in hydathodes at all stages of leaf development and in all hydathode cell types, as well as in epidermis, vasculature, and the mesophyll cells adjacent to the hydathode (Fig. 3B, E, F, J, K). AtISA1 expression is not detected in any hydathode cell at any development stage.
The DBE gene expression patterns are also diverse with respect to reproductive structures. AtISA1, AtISA2, and AtISA3 are all expressed in flower sepals and pedicels, and in stigma, throughout the maternal tissues of the ovules (integument) and ovary wall (Fig. 3L, M). Cell-specific and development-specific co-expression of AtISA1 with AtISA2, and also AtISA2 with AtISA3 extends to the flowers. AtISA1 and AtISA2 are co-expressed in the filament (Fig. 3M) whereas AtISA2 and AtISA3 are co-expressed in the style (Fig. 3M). Only AtISA2 expression is detectable in the pollen (data not shown) and the receptacle (Fig. 3L, M). No DBE expression was detectable in the petal or in the maternal-derived tissues of anthers at any stage of development (Fig. 3M). In young developing seeds and embryos (1–3 d after flowering, only AtISA2 is highly expressed (Fig. 3N). During silique development, AtISA1 is predominately expressed in the silique wall, AtISA2 is predominately in developing ovules, and AtISA3 expression decreases in the silique wall (Fig. 3N, O).
In the roots, AtISA1 is expressed in the epidermis at all stages of development (Fig. 3P–R). AtISA2 and AtISA3 are co-expressed in the primary and lateral root tips (Fig. 3P, Q, S, T, and data not shown), specifically in the root cap, columella, and peripheral cap, to a lesser extent in the root meristem region, but not in the epidermis. Only AtISA2 is expressed in the root vasculature at the site of lateral root initiation, in the phloem, pericycle, endodermis, and cortex (Fig. 3P, T). As the lateral root develops, expression of AtISA2, and to a lesser extent AtISA3, becomes detectable throughout the root cortex and increases as the root matures (Fig. 3R); AtISA2 only is also expressed in mature root vasculature (Fig. 3R, T).
Subcellular localization of ISA1 and ISA3 proteins
AtISA1 and AtISA2 are presumed to be plastidial proteins based on the observations that the major isoamylase-type DBE activity present in Arabidopsis chloroplasts is dependent on function of both the AtISA1 and AtISA2 genes (Zeeman et al., 1998a, b; Delatte et al., 2005). The promoter fusion approach used in the current study allows direct determination of the subcellular location of AtISA1. Promoter fusion construct ISA1ptar is similar to the previously described construct ISA1p, with the inclusion of the first 110 amino acids of the AtISA1 primary translation product at the amino terminus prior to the GFP/GUS reporter (Fig. 2A). Fifteen independent transgenic lines containing the ISA1ptar fusion construct were generated, and the subcellular location of the fusion protein was visualized by GFP fluorescence in leaf mesophyll cells (Fig. 4). The GFP fluorescence pattern precisely overlaid the chlorophyll auto-fluorescence pattern that identifies chloroplasts (Fig. 4B). The GFP fluorescence generated from the control construct ISA1p, which lacks any part of the amino terminus of AtISA1, is distributed throughout the mesophyll cells and does not correlate with plastids (Fig. 4D). These data demonstrate directly that the first 110 amino acids of AtISA1 serve as a plastid targeting peptide. Similar analysis was performed on AtISA3, with results showing that the first 122 amino acids serve as a targeting peptide that directs GFP to leaf mesophyll chloroplasts in intact plants (Fig. 4C). These data from transgenic plants confirm a previous report that demonstrated localization of AtISA3::GFP in chloroplasts when the protein was transiently expressed in Arabidopsis protoplasts (Delatte et al., 2006).
ISA3 is a member of a cluster of genes induced during cold acclimation
Because AtISA3, AtGWD1, and AtGWD3 are highly correlated in expression and are all involved in starch degradation, and because the physiological functions of AtGWD genes are the subject of considerable investigation (Mikkelsen et al., 2004; Kötting et al., 2005; Yano et al., 2005), the 22 746 gene dataset was analysed and other genes that are co-expressed with this group and conditions under which they are co-expressed were identified. Forty-eight genes are most highly co-expressed (Pearson correlation coefficient ≥0.5) with all three genes (AtISA3, AtGWD1, and AtGWD3) (see Supplementary Table S5 at JXB online). Of these 48 genes, 14 are starch metabolic genes [AtBE3, AtBAM2 (At2g32290), AtISA3, AtSEX4, AtISA2, AtPU1, AtMEX1, AtAAM3, AtGWD1, AtGWD3, AtPHS1, AtPHS2, AtDPE1, and AtDPE2], all of which have been previously reported to function in starch degradation with the exception of AtISA2, AtBE3, and AtBAM2 (Yu et al., 2001; Lu and Sharkey, 2004, 2006; Baunsgaard et al., 2005; Delatte et al., 2005; Smith et al., 2005; Wattebled et al., 2005; Dumez et al., 2006; Lu et al., 2006; Niittylä et al., 2006). Also correlated with AtISA3, AtGWD1, and AtGWD3 are a group of six cold-regulated genes [AtCOR314-TM2 (At1g29390), AtLTI6A (At4g30650), AtCOR15B (At2g42530), AtKIN1 (At5g15960), AtCOR15A (At2g42540), and AtCOR414-TM1 (At1g29395)], two heat shock genes, three circadian genes, three kinase/phosphatase genes, and 20 genes of unknown or unclear function (see Supplementary Table S5 at JXB online).
To further examine relationships among expression of the cold-regulated and 14 starch metabolic genes, a pair-wise matrix of Pearson correlation coefficients among these 22 genes that highly correlate in expression with AtISA3, AtGWD1, and AtGWD3, is shown in Table 2. AtISA1 is also included in this matrix for comparison. There appears to be a broad correlation between genes involved in starch degradation and the cold-regulated genes. These analyses, together with experiments demonstrating AtGWD1 is induced in the cold (Yano et al., 2005), led us to further investigate whether AtISA3 might have a function in cold acclimation.
Cis-acting motifs in the DBE and GWD gene promoters, along with a comprehensive set of starch metabolism genes, were evaluated using the bioinformatics search programs Athena (O'Connor et al., 2005) and Plant Care (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/). The results are shown in Supplementary Tables S5 and S6 at JXB online. Cis-acting elements known in other genes to mediate response to cold are present multiple times in many starch metabolic genes. Response to cold is in part mediated by ABA (abscisic acid) (Eckardt, 2001), and ABA signalling motifs occur in all four DBE genes, AtGWD1, and AtGWD3. AtISA3 contains six ABA elements, four different cold-responsive promoter motifs, including the LTRE (low temperature-responsive element) and a DREB1A/CBF3 element (dehydration-responsive element/C-repeat-binding) (see Supplementary Table S6 at JXB online). The latter motif interacts with the DREB1A/CBF3 transcription factor, which is itself induced in cold environments, and is known to be important in freezing tolerance (Maruyama et al., 2004). In comparison, AtISA1 possesses six ABA elements but no known cold response-specific promoter motifs. The AtISA2 and AtGWD1 promoters each have a single cold response element in addition to their one and five ABA motifs, respectively.
Cold-responsive promoter motifs are common in the promoters of the 25 starch metabolic genes that correlate in expression with cold-regulated genes (see Supplementary Table S7 at JXB online). AtBAM7, AtISA3, AtGWD1, and AtSS1 all contain more than five cold-responsive promoter motifs. AtPHS2 and AtBAM8 respond strongly to cold induction although their promoters exhibit two known cold-responsive sequence motifs.
Numerous other promoter motifs including light- or stress-responsive elements are present in starch metabolic genes (see Supplementary Tables S6 and S7 at JXB online). Presumably, these elements enable the plants to adjust starch metabolism in response to different environments.
Three microarray experiments from the public data set that analysed the effects of different cold treatments were examined for cold responses of AtISA3 and other starch metabolic genes (Warren et al., 2002; Boyce et al., 2003). In these experiments, Arabidopsis plants were harvested after exposure to cold temperature (5 °C) for 3 h (experimenter: H Knight), at 4 °C for 24 h (experimenter: I Bramke), or at 4 °C for 10 d (experimenter: JG Warren). AtISA3, AtGWD1, and AtGWD3 expression increased at about 1 d after cold treatment and was maintained after 10 d of cold treatment (Fig. 5A) and thus these are called slow cold-responsive genes. AtISA1, AtISA2, and AtPU1 did not respond to cold temperature within 10 d (Fig. 5A). Figure 5B shows the response of 11 additional slow cold-responsive genes. This group includes all six genes previously identified as correlating in expression with AtISA3, AtGWD1, and AtGWD3 across multiple conditions (Table 2); the five others correlate with some members of this group. Taken together, the data in Figs 2B and 5A show that AtISA3, along with AtGWD1 and AtGWD3, is a member of a cluster of genes whose steady-state mRNA levels increase after sustained exposure to low temperature, i.e. 1 d or longer at 4 °C or 5 °C.
Two Arabidopsis genes potentially involved in regulating expression of AtISA3 and other starch metabolic genes were identified through these analyses. At5g48250 expression correlates with multiple starch metabolic genes including AtISA3, AtGWD1, and AtGWD3. They are co-expressed under cold conditions (Fig. 5C), and also across all the chips in the public database (Pearson correlation coefficients of 0.55, 0.65, and 0.64, respectively). The expression of this gene correlates with that of the cold-responsive genes from the set shown in Fig. 5B; the expression of these cold-regulated genes also correlates with that of multiple starch metabolic genes. A second gene of potential interest is AtGRP7/CCR2 (At2g21660, COLD CIRCADIAN RHYTHM AND RNA BINDING 2) (TAIR; http://www.arabidopsis.org/). Like At5g48250, AtCCR2 exhibits similar mRNA accumulation profiles as AtISA3, AtGWD1, and AtGWD3 under cold conditions (Fig. 5C), as well as other conditions (Pearson correlation coefficients of 0.56, 0.58, and 0.60, respectively); expression of AtCCR2 is also correlated with that of At5g48250 (Pearson correlation coefficient 0.67).
The expression of the 25 starch metabolic genes that are correlated (Pearson correlation coefficient ≥0.5) with one or more of the 11 slow cold-responsive genes is shown in Fig. 5D, E. Expression of a subset of starch metabolic genes, specifically, AtPHS2, AtBAM8 (At4g17090), AtBE3, AtT6PP (At2g22190), AtBAM7, AtISA3, AtSEX4, AtGWD1, AtSS1 (At5g24300), AtGWD3, AtAPS1 (At5g48300), and AtAPL1 (At5g19220), is increased by cold induction.
Cellular distribution of DBE expression in plants grown in the cold
The present bioinformatics analysis indicates that some DBEs are part of a cluster of genes that is regulated by cold. To visualize the spatial response to cold of AtISA1, AtISA2, AtISA3, and AtPU1 in intact plant tissues, the AtISA promoter::GUS transgenic lines were analysed first. Plants were grown under SD conditions at 22 °C until 15 DAI, when two true leaves had emerged. A subset of the plants was then transferred to 4 °C and the SD light regime was continued. Under these cold-adaptive conditions the plant growth rate was vastly reduced and the plants became dark green. Plants were stained for GUS activity at various times after the switch to the cold environment.
The expression of AtISA3 is prominent in cold-grown plants (Fig. 6). AtISA3 expression increases after 10 d in the cold, and the response is even greater after 41 d. At this time AtISA3 is uniformly distributed in the leaf mesophyll, whereas in plants grown at room temperature its expression is clearly observable in the hydathodes and leaf edge (Figs 3B, 6). AtISA2 expression also increases in the cold after 41 d, although there is no observable change in the tissue specificity. AtISA2 and AtISA3 are co-expressed in leaves after the plants become adapted to a cold environment (Fig. 6). Early in the cold response there may be a transient decrease in the expression of both AtISA2 and AtISA1 in leaves, as judged by the intensity of GUS staining after 10 d in the cold (25 DAI). AtISA1 expression is stable throughout the long period of cold acclimation; however, by contrast with that of AtISA2 and AtISA3, it does not appear to accumulate to levels higher than at room temperature.
The present bioinformatics analysis did not indicate major shifts in DBE expression in response to other environmental stresses. To evaluate this at a spatial level, induction of all four DBE genes in response to either bacterial infiltration or wounding was also examined. In neither instance was altered expression of any DBE gene observed (data not shown).
ISA1, ISA2, and ISA3 expression patterns
The bioinformatics and reporter gene analyses, taken together, revealed several distinct sets of co-expression relationships among the three AtISA genes of Arabidopsis, grouped as follows. (i) AtISA1 and AtISA2 are co-expressed in certain tissues in which AtISA3 expression is not evident, for example, most leaf mesophyll cells and vasculature, and the shoot meristem (Fig. 3C–F). (ii) AtISA2 and AtISA3 are co-expressed in specific locations in which AtISA1 expression is not evident, for example, leaf hydathodes, root tips, trichomes, and styles (Fig. 3D–F, I–K, Ob). (iii) All three isoamylase-type DBEs are co-expressed in leaf edge mesophyll, leaf edge vasculature, and sepals (Fig. 3B, D–F, L, M). (iv) AtISA1 is expressed in the apparent absence of AtISA2 or AtISA3, for example, in leaf and root epidermis (Fig. 3G, P). (v) AtISA2 is expressed in the apparent absence of AtISA1 or AtISA3, for example, in the embryo 2 d after flowering and at the site of lateral root initiation (Fig. 3G, P). This finding is unexpected considering that ISA2 is not catalytically active as an α-(1→6) glucosidase enzyme. Possibly AtISA2 has some as yet unidentified function in the absence of either AtISA1 or AtISA3. In no instance was expression of AtISA3 observed in the absence of AtISA2.
The spatial patterns of co-expression of AtISA1 and AtISA2 during plant development are fully consistent with the known functions of both these genes in starch production and/or accumulation, and with the evidence that the proteins encoded by these two genes form a heteromultimeric complex (Hussain et al., 2003; Delatte et al., 2005; Wattebled et al., 2005). AtISA1 and AtISA2 are co-expressed in leaf and cotyledon mesophyll, highly active sites of starch accumulation. AtISA1 and AtISA2 co-expression in the shoot meristem indicates starch metabolism may be important in these highly developmentally active cells.
By contrast, the co-accumulation of AtISA2 and AtISA3 mRNAs in the apparent absence of AtISA1 mRNA was not expected, in light of the fact that AtISA2 is known to be involved in starch production, whereas AtISA3 functions in the degradative pathway (Delatte et al., 2005, 2006; Wattebled et al., 2005). However, co-expression of AtISA2 and AtISA3 occurs in many discrete locations throughout development, and is particularly obvious in hydathodes of mature leaves (Fig. 3F, J, K) and in germinating seedling root tips (Fig. 3A). The global microarray analysis revealed high co-accumulation of AtISA2 and AtISA3 transcripts in several conditions (Fig. 1A). Considering the established catabolic role of ISA3, and the fact that ISA2 is non-catalytic (Hussain et al., 2003), one explanation for the co-expression of ISA2 and ISA3 is that they function together in the starch degradation process. Thus, ISA2 could function in both starch biosynthesis and starch degradation. Furthermore, considering that ISA2 forms a stable complex with ISA1 and this complex functions in starch biosynthesis (Hussain et al., 2003; Delatte et al., 2005; Wattebled et al., 2005), it is possible that ISA2 and ISA3 also form a protein complex, in this instance dedicated to starch degradation. Both AtISA3 (Fig. 4A; Delatte et al., 2006) and AtISA2 (Zeeman et al., 1998a, b), are localized in the leaf chloroplast, also consistent with the possibility of an ISA2/ISA3 complex. The fact that mutation of ISA2 does not condition a starch excess phenotype (Delatte et al., 2005; Wattebled et al., 2005) does not contradict the hypothesis that the protein functions catabolically, but rather indicates that the biosynthetic abnormality may dominate in the mutant, i.e. in a background of low starch, such as in Atisa2, a decrease in starch degradation might not be evident. Experiments are underway to provide biochemical evidence for a physical association between AtISA2 and AtISA3.
An unexpected finding from this study is that AtISA3 expression in leaves appears to be concentrated in the hydathodes and leaf border tissues, and is not readily observable throughout the mesophyll. Loss of AtISA3 causes a major deficiency in the ability to catabolize starch during the dark phase of the diurnal cycle (Wattebled et al., 2005; Delatte et al., 2006). Most of the starch subjected to degradation in the dark is located in the leaf mesophyll, so the fact that ISA3 transcript is not present or is present in low levels in that tissue is surprising. Leaf cross-sections including the main vein appear to lack ISA3 completely, whereas ISA1 and ISA2 are both easily detected in the same sections (Fig. 3G, H). Thus, there is a possibility that the function of ISA3 in starch degradation acts at a distance from the majority of the substrate, perhaps through a metabolic signalling pathway. The expression of genes encoding other starch-degradative enzymes, notably AtGWD1 and AtGWD3, in the hydathode, but not in most of the leaf mesophyll (A Blennow, Copenhagen University, personal communication), adds support to this hypothesis.
The expression of AtISA2 and AtISA3 is coincident with starch accumulation in root columella cells (Fig. 3U; Yamamoto et al, 2002) and leaf hydathodes (Fig. 3U), while the expression of AtISA1 was never observed in root columella cells (Fig. 3A, S) and leaf hydathodes (Fig. 3E, F, J). These results show that starch is present in tissues where there is no detected expression of AtISA1. Thus, AtISA1 might not be necessary for starch synthesis in some tissues. This consideration begs the question of how starch is produced in hydathodes and the root tip in the apparent absence of ISA1. Possibly ISA2/ISA3 could be both biosynthetic and degradative, or there could be an unknown factor providing the necessary DBE function. These observations are also consistent with the suggestion that AtISA3 might account for a remnant isoamylase-type DBE that supports starch accumulation in certain cells of Atisa1 or Atisa2 mutants (Delatte et al., 2005); the starch accumulation reported in the Atisa1 and Atisa2 mutants might be located on the leaf periphery, where AtISA3 is highly expressed (Fig. 3B, E, F).
Flowering is a high energy-demanding process. Furthermore, starch metabolism is predicted to play important roles in floral transition and tissue differentiation, flower opening, and nectar release (Corbesier et al., 1998; Bieleski et al, 2000; Thornberg, 2007). The high expression or co-expression of debranching enzymes in floral tissue and the precise spatial developmental shifts in these patterns of expression supports the hypotheses that starch hydrolysis has a critical function in providing carbohydrate and energy for flower development.
Starch is important not only for energy storage, but it also mediates osmotic and gravitropic responses. The co-expression of AtISA2 and AtISA3 in a variety of tissues may be indicative of a function in starch catabolism in response to signals or energy demands. In root columella cells of Arabidopsis and radish, starch has an apparent osmoregulatory function, as immediate degradation of starch and a reduction in starch content take place in response to increased moisture (Takahashi et al., 2003). Starch is also important in columella cells for the gravitropic response in which starch grains in the amyloplasts help to reorient the roots (Kiss, 2000). The location of AtISA2 and AtISA3 mRNAs in the columella cells suggests that the corresponding proteins might be involved in root starch degradation to control orientation and growth in response to gravity and water. Arabidopsis hydathodes are water-secreting glands (Pilot et al., 2004) which also contain starch (Fig. 3U), and the presence of AtISA2 and AtISA3 may indicate that these proteins function in the regulation of starch content by water. It is possible that hydathode starch is important in osmotic response.
A possible mechanism by which AtISA3 action in the hydathode might affect starch degradation in other tissues involves auxin (IAA) signalling. Hydathodes are a primary site for IAA synthesis (Aloni et al., 2003; Aloni and Ullrich, 2005), as are root tips (Ljung et al., 2005), and the pattern of expression of genes involved in auxin transport in leaf veins, hydathodes and trichomes, and root columella cells (Mattsson et al., 2003; Blilou et al., 2005) corresponds closely with the ISA3 expression pattern. Furthermore, IAA has been reported to reduce the expression of starch biosynthetic genes in cultured tobacco cells (Miyazawa et al., 1999) and to decrease starch accumulation in cotton (Bornman et al., 1966). From these observations, we suggest that the activity of ISA2, ISA3, GWD1, and GWD3 in hydathodes and root tip may influence auxin transport or signalling, and this in turn could regulate the decrease in starch accumulation that occurs throughout the leaf during the dark phase of the diurnal cycle.
The present study of co-expression of DBE genes in Arabidopsis is consistent with previous studies in which it was found that in dicot leaves and tubers the predominant form of isoamylase-type DBE is the heteromultimer (Hussain et al., 2003; Delatte et al., 2005). In monocot endosperms it appears that ISA1 homomultimers predominate (Fujita et al., 1999). The homomeric and heteromeric DBE complexes might have different biological properties.
Expression of ISA2, ISA3, and other starch metabolic genes during cold acclimation
Potato tubers slowly degrade starch and accumulate sugars when exposed to cold temperatures, a process known as cold sweetening (Müller-Thurgau, 1882; Burton, 1969). Such starch degradation has been shown to provide an important mechanism that supports cold acclimation in potato and in Arabidopsis (Lorberth et al., 1998; Kaplan and Guy, 2005; Yano et al., 2005). The results presented here indicate additional factors likely to function in starch breakdown and monosaccharide and disaccharide accumulation at low temperatures. Bioinformatics analysis shows that AtISA3 is a slow cold-responsive gene, as well as other starch metabolic genes, specifically AtBAM7, AtSEX4, AtPHS2, AtDPE2, AtSS1, AtBE3, AtAPS1, AtAPL1, AtGWD1, and AtBAM8. Steady-state mRNA levels of starch-degradative genes, AtGWD1 and AtBAM8, have been shown previously to increase in response to cold (Kaplan and Guy, 2004; Yano et al., 2005; Lu and Sharkey, 2006) (see Supplementary Table S8 at JXB online; Fig. 5D, E). Induction of the starch biosynthetic genes, AtSS1, AtBE3, AtAPS1, and AtAPL1, in addition to the degradative genes, might indicate that not only starch breakdown but other aspects of starch metabolism are important for long-term acclimation to cold.
The induction of AtISA3 in response to cold in the promoter fusion analysis indicates the increase in steady-state mRNA levels can probably be attributed to increased transcription of the AtISA3 gene. The AtISA3 up-regulation shown by the GUS reporter is observed after 10 d in the cold (Fig. 6). For AtISA2, increased expression was not detected after 10 d but was clearly observed after 41 d in the cold (Fig. 6). This is consistent with the microarray data, in which an increase in AtISA2 expression was not detected, since these experiments extend only to 10 d in the cold. Thus, AtISA2 induction appears to be a longer-term adaptation to cold. The promoter analysis is also consistent with the observation that AtISA3 is expressed throughout the leaf mesophyll when it is induced in response to cold; this is a vivid contrast from its restriction to the leaf margin and hydathode restriction during growth at normal temperature. Based on these differences in tissue localization, AtISA3 appears to serve different environmentally mediated physiological roles in starch degradation. Finally, the increase and co-localization of AtISA2 and AtISA3 expression during the response to cold temperature provides further evidence consistent with these two proteins having related functions in starch metabolism.
Sequence elements that mediate transcriptional induction in response to cold are frequent in the promoters of many starch metabolic genes (see Supplementary Tables S6 and S7 at JXB online). AtISA3 contains a promoter motif known to be the binding target of the transcription factor DREB1A/CBF3 (see Supplementary Tables S6 and S7 at JXB online) (Maruyama et al., 2004). The bioinformatics analysis described in this study indicates that of the three DREB family genes, AtDREB1A/CBF3 (At4g25480), and AtDREB1C/CBF2 (At4g2540) were induced within 3 h after exposure to cold (see Supplementary Fig. S1 at JXB online). Among the genes found to respond to cold similarly to ISA3 (Fig. 5A, B) are nine that have been reported to be regulated by DREB1A/CBF3, specifically COR314-TM2, COR15A, COR15B, KIN1, KIN2, LTI6A, COR78, COR47, and LTI29 (Maruyama et al., 2004). DREB1A/CBF3 functions to provide tolerance to freezing stress by effecting an increase in sucrose accumulation (Gilmour et al., 2000; Maruyama et al., 2004). ISA3 is likely also to be regulated by DREB1A/CBF3, and to participate in this same physiological response. DREB1A/CBF3 is expressed only under cold conditions and thus could modulate expression of ISA3 only in response to cold. Therefore, other factors must regulate ISA3 expression in conditions other than induction during cold acclimation. Based on their high correlation in steady-state mRNA level with ISA3 under a wide range of conditions, we suggest two possible modulators of responses of the DBE genes. These are CCR2 [which codes for a glycine-rich RNA-binding protein that is involved in regulating cold circadian rhythm (Heintzen et al., 1997; Kreps and Simon, 1997)] and the product of locus At5g48250 [which codes for a zinc finger (B-box type) family protein, belonging to the C2C2-CO-like family (TAIR)]. These genes could be investigated experimentally as potential starch metabolism regulatory factors.
The characterizations of AtISA1, AtISA2, and AtISA3 expression in this study are consistent with the hypothesis that multiple assembly states of the isoamylase-type DBE enzyme(s) provide, at least in part, biochemical distinctions that translate into various roles in either starch synthesis or starch degradation. This hypothesis predicts that ISA2/ISA3 heteromeric enzymes may exist in plants, in addition to the ISA2/ISA1 heteromultimer that has been inferred to exist in Arabidopsis from genetic data and identified directly in other species. Tissue-specific expression of the three AtISA genes indicates the potential for cell-specificity of the proposed isoamylase-type DBE complexes. Biochemical characterization can provide more definitive tests of functional relationships between these three ISA proteins.
The following supplementary data for this article are available at JXB online:
Fig. S1. mRNA accumulation profiles of DREB family genes in response to cold.
Table S1.Expression of DBEs
Table S2.Genes highly correlated to DBEs, and GWDs
Table S3. Starch metabolic genes commonly highly correlated with all DBEs
Table S4.ISA genes are differently expressed in Arabidopsis
Table S5.Genes that are correlated to AtISA3, AtGWD1 and AtGWD3
Table S6.Promoter analysis for DBEs, AtGWD1 and AtGWD3
Table S7.Cold and light responsive promoter motifs among starch metabolic genes
Table S8.Expression of starch metabolic genes and cold regulated genes in cold
days after imbibition
starch debranching enzyme
isoamylase-type starch debranching enzyme
short day diurnal cycle of 8 h light/16 h dark
We thank Dr Jack Horner, Tracey Pepper, and Randall Den Adel in the Bessey Microscope Facility and Margie Carter in the Confocal Microscopy Facility for their help in using the equipment. We are particularly grateful to Susan Blauth, at the University of Redlands for her very helpful discussions about hydathodes.