Starch synthesis in Arabidopsis is achieved by spatial cotranscription of core starch metabolism genes.

Starch synthesis and degradation require the participation of many enzymes, occur in both photosynthetic and nonphotosynthetic tissues, and are subject to environmental and developmental regulation. We examine the distribution of starch in vegetative tissues of Arabidopsis (Arabidopsis thaliana) and the expression of genes encoding core enzymes for starch synthesis. Starch is accumulated in plastids of epidermal, mesophyll, vascular, and root cap cells but not in root proper cells. We also identify cells that can synthesize starch heterotrophically in albino mutants. Starch synthesis in leaves is regulated by developmental stage and light. Expression of gene promoter-β-glucuronidase fusion constructs in transgenic seedlings shows that starch synthesis genes are transcriptionally active in cells with starch synthesis and are inactive in root proper cells except the plastidial phosphoglucose isomerase. In addition, ADG2 (for ADPG PYROPHOSPHORYLASE2) is not required for starch synthesis in root cap cells. Expression profile analysis reveals that starch metabolism genes can be clustered into two sets based on their tissue-specific expression patterns. Starch distribution and expression pattern of core starch synthesis genes are common in Arabidopsis and rice (Oryza sativa), suggesting that the regulatory mechanism for starch metabolism genes may be conserved evolutionarily. We conclude that starch synthesis in Arabidopsis is achieved by spatial coexpression of core starch metabolism genes regulated by their promoter activities and is fine-tuned by cell-specific endogenous and environmental controls.

Along with the plastidial forms of PGI and PGM, there are cytosolic isozymes of PGI and PGM in Arabidopsis. Both cytosolic and plastidial PGI and PGM enzymes can transform fructose-6-phosphate (F6P) to glucose-6-phosphate (G6P) and G6P to glucose-1-phosphate (G1P), respectively. The Arabidopsis pgi1 mutant is deficient in plastidial PGI and lacks starch in mesophyll cells, but the starch synthesis in root cap cells and guard cells is not affected (Yu et al., 2000). This phenotype suggests that in the absence of plastidial PGI, cytosolic G6P can be transported into chloroplasts of guard cells and amyloplasts of root cap cells for starch synthesis, while chloroplasts of mesophyll cells cannot import G6P efficiently.
In Arabidopsis, there are two glucose 6-phosphate/phosphate (G6P/Pi) translocators (GPT1, At5g54800 and GPT2, At1g61800). While GPT1 is expressed ubiquitously during plant development, GPT2 expression is restricted to a few tissues. It has been demonstrated that ectopic expression of either G6P/Pi translocators can import cytosolic G6P to chloroplasts of mesophyll cells and can rescue the starch deficiency in mesophyll cells of the pgi1 mutant (Niewiadomski et al., 2005).
The deficiency of starch in the pgm1 mutant (Caspar et al., 1985), which has no detectable plastidial PGM activity, indicates that cytosolic G1P is not efficiently transported into plastids as the substrate for AGPase to produce ADP-Glucose (ADPG). AGPase is a tetrameric enzyme containing either four small subunits or two large and two small subunits (Li and Preiss, 1992). In Arabidopsis, two genes encode for small subunits (APS1 and a 6 pseudogene APS2) and four genes encode for large subunits (APL1-APL4) (Crevillen et al., 2005). The adg1 mutant has a mutation in the small subunit gene (designated as APS1) and the adg2 mutant has a mutation in the large subunit gene (designated as APL1). The starchless phenotype of the adg1 mutant, which lacks AGPase activity, suggests that ADPG produced in plastids is the substrate for starch synthesis (Lin et al., 1988;Wang et al., 1998).
The adg2 mutant, in which AGPase is suggested to be a homotetramer with four small subunits (Li and Preiss, 1992), has decreased starch content in leaves. While there is evidence that homotetrameric AGPases of Arabidopsis APS1 (Crevillen et al., 2003) can function in E.
coli, the APS1 homotetramer was much less sensitive to 3-PGA activation and more sensitive to Pi inhibition. It has been suggested that APL1 homologs (APL2, APL3, and APL4) may also form heterotetramer with APS1 based on the presence of specific isoforms and play roles in starch metabolism in response to various metabolic states of tissues (Crevillen et al., 2005).
Genetic evidence indicates that plant cells need cooperative actions of multiple genes to accomplish starch synthesis and degradation. Starch synthesis occurs in both photosynthetic and non-photosynthetic tissues, yet not every plant cell accumulates starch.
For instance, starch granules are present in root cap cells (Yu et al., 2000;Siedlecka et al., 2003), but are absent in other cells of the root (i.e., the part of the cap-less root designated as root proper). This tissue-specific distribution of starch must be associated with the starch synthesis machinery, suggesting that starch synthesis genes may be co-expressed in a tissuespecific manner. Coordinated expression of starch synthesis gene during the endosperm development of maize and rice has been shown (Giroux et al., 1994;Zhu et al., 2003;Ohdan et al., 2005). By Pearson correlation coefficient analysis, 30 genes whose expression is highly correlated with starch debranching enzyme genes in Arabidopsis were identified (Li et al., 2007). Of these 30 genes, 12 are involved in starch metabolism and 10 of them are known for starch degradation. Some starch metabolism genes of Arabidopsis have a similar diurnal expression pattern in leaves (Harmer et al., 2000;Smith et al., 2004), suggesting that there is a temporal coordinated regulation. In Arabidopsis, high resolution microarray expression profiles of various types of sorted root cells and root cells dissected from different www.plantphysiol.org on December 31, 2017 -Published by 7 development stages demonstrate not only developmental but also spatial co-regulation of many clusters of genes (Brady et al., 2007). However, only the steady state level of mRNA was analyzed in these previous studies, and the mechanism for the coordinated expression of these genes was not investigated. To address the regulation of starch synthesis genes, we investigated the pattern of spatial expression of several starch synthesis genes. In this study, we demonstrate that PGM1, APS1, APL1, and SS1 promoters are transcriptional inactive in cells without starch synthesis, e.g., root proper cells. Our results suggest that starch synthesis in Arabidopsis requires spatial co-expression of core starch metabolism genes, which are regulated by their promoter activities. Furthermore, starch synthesis can be fine-tuned by the accessibility of carbon sources under cell-specific and environmental controls.

Starch Distribution is Similar in Arabidopsis and Rice Seedlings
To examine the distribution of starch in vegetative tissues at the end of the light period, we used Lugol's solution to stain de-pigmented plants of Arabidopsis and rice. As shown in Figure 1, starch stained with iodine as the blue-purple color was found mainly in mesophyll cells of Arabidopsis. The staining of starch was more intensive in older leaves than that in younger leaves (Fig. 1A). Within a cross section of Arabidopsis leaf, mesophyll cells at adaxial side had more starch than cells at the abaxial side at the end of the day (Fig.   1B). However, when we inverted the leaves at the end of the night and started the illumination for 12 hours, the pattern of starch accumulation was reversed (Fig. 1C). Thus, mesophyll cells towards light could accumulate more starch than those underneath cells sheltered from light. In epidermis, starch was found in guard cells and could be detected in subsidiary cells neighboring the guard cells (Fig. 1D). The root cap comprises columella cells and the outer lateral root cap cells. Starch granules were present mainly in columella cells and some lateral root cap cells; but were absent in other cells of the root, i.e., root proper ( Fig.   1E). Because root cells do not contain chloroplasts normally, starch synthesis in root cap cells would be heterotrophic. To identify cells with heterotrophic starch synthesis, we examined starch distribution in an albino mutant, ispH (Hsieh and Goodman, 2005), grown on MS medium with sucrose. Because the ispH mutant has no chlorophyll in plastids to support photosynthesis, starch synthesis must rely on exogenous carbon source. We found that starch was present in root cap cells of the ispH mutant as expected (Fig. 1F). In addition, starch was

Enzyme Activities of Specific Starch Synthesis Genes are Absent in Root
Genetic studies of starch deficient mutants of Arabidopsis and other species have shown that the cooperation of PGI, PGM, AGPase, and SS enzyme activities is required for normal starch biosynthesis. Cells without starch could potentially lack either a set of starch synthesis enzymes or a single key enzyme. To correlate the distribution of starch and enzymes required for starch synthesis, we analyzed the enzyme activities in extracts from leaves and roots by native activity gels. Leaf extract of wild-type Arabidopsis contained plastidial PGI, PGM, AGPase, and SS activities, which was expected for normal starch synthesis ( Fig. 2A). Root extract of wild-type Arabidopsis did not have significant plastidial PGM, AGPase, and SS activities but unexpectedly did have plastidial PGI activity ( Fig. 2A).
Root cap cells synthesize starch and should contain enzyme activities for starch synthesis.
The root extracts analyzed in our experiments included root cap cells; however, enzyme activities of root cap cells in the root tissue was proportionally very low and therefore barely detectable (e.g., PGM in Figure 2A). Both root and leaf extracts contained cytosolic isozymes of PGI and PGM. The activities of cytosolic isozymes were not correlated with the difference in starch accumulation between leaves and roots, suggesting that cytosolic PGI and PGM do not contribute directly to starch synthesis in chloroplasts. We further examined these enzyme activities present in leaf and root extracts of rice seedlings with zymograms. Based on rice genomic sequence data, there are four genes encoding plastidial PGI and two genes for cytosolic PGI. Six activity bands for PGI were present in both extracts from leaves and roots 9 ( Fig. 2B). Specific enzyme activity bands for PGM, AGPase and SS, were present in leaf extracts but were not detected in root extracts of rice seedlings (Fig. 2B). We confirmed that bands specific to leaf extracts were localized in plastids by comparison to zymograms of isolated rice chloroplasts (data not shown). Thus, the pattern of enzyme activities examined in rice was similar to that of Arabidopsis, i.e., a set of plastidial starch synthesis enzymes was present in leaf but not detected in root with the exception of PGI.

Message RNA of Specific Starch Synthesis Genes is Absent in Root
The regulation of gene expression can be achieved at many levels, including transcription, translation, and processing of mRNA and through protein turnover. To investigate the mechanism underlying the lack of starch biosynthetic enzyme activities in roots, we examined mRNA in leaves and roots of Arabidopsis by northern blot analysis with gene specific probes. As shown in Figure 3A, PGI1 was expressed in both leaves and roots and the other starch synthesis genes examined (PGM1, APS1, APL1, and SS1) were only expressed in leaves. Our results suggest that these enzymes are not synthesized in roots because the corresponding mRNAs are absent. To test if this gene co-expression pattern in leaves and roots is conserved among different plant species, we also examined the expression of genes encoding enzymes for starch synthesis in rice. By sequence comparison, starch synthesis genes of rice with homology to those of Arabidopsis were selected for the analysis of their expression in leaves and roots (Table I). Northern blot analysis showed that the transcripts of the genes were present specifically in leaves but not in roots ( Fig. 3B), with the exception of OscpPGI mRNA, which was detected in both leaves and roots. Therefore, the co-expression pattern of these rice genes was very similar to that observed in Arabidopsis.

Starch
The absence of mRNA of specific starch synthesis genes in roots could be due to absence of transcription or due to the instability of transcripts. It is possible to distinguish (650 bp) and SS1 promoter (4073 bp) fused to the β-glucuronidase (GUS) reporter, respectively (Fig. 4).
The GUS expression in transgenic seedlings with various promoter-GUS fusion constructs indicated that PGI1, PGM1, APS1, APL1, and SS1 promoters were active in mesophyll cells of leaves with higher activity in older leaves than younger leaves (Fig. 5, A-E). These GUS staining patterns were similar to that of iodine staining of leaf starch (Fig. 1A).
The GUS activity did not show distinct differences among mesophyll cells within a cross section of leaves ( , 2007). To confirm that the expression of GUS is limited to root cap cells, we also examined the GFP expression in transgenic lines carrying a P APS1 -GFP fusion construct. This analysis also showed that GFP expression was present in columella cells and lateral root cap cells (Fig. 5V), but not in root proper cells (Fig. 5W). The intensity of green fluorescence present in columella cells was higher than that in lateral root cap cells.
We compared the starch granules present in columella cells and lateral root cap cells by electron microscopy (EM). The number of plastids and size of starch granules are more and larger in columella cells than in lateral root cap cells (Fig. 6, A and B). To find out whether these plastids contain chlorophylls, we further examined these cells under a confocal microscope for auto-fluorescence emitted from chlorophyll upon the excitation of 488 nm laser light. In contrast with plastids of leaf cells, no red fluorescence was observed in root cells, confirming that plastids in root cells did not contain chlorophylls (Fig. 6, C-E).
Surprisingly, APL1 promoter activity was not detected in root, including root cap cells ( Fig. 5N). Because of this unexpected expression pattern of APL1, we examined starch granules in root cap cells of wild type and adg2 mutant that lacks a functional large subunit AGPase (Wang et al., 1997). While the appearance of starch granules of mesophyll cells in wild type and adg2 mutant was apparently different (Fig. 7, A and B), starch granules in columella cells of adg2 mutant were indistinguishable from those in wild type (Fig. 7, C-F), suggesting that APL1 is not required for starch synthesis in root cap cells.

Expression Profile of Arabidopsis Starch Metabolism Genes
Cooperative actions of multiple enzymes are required for starch metabolism in plant cells. Tissue-specific distribution of starch in plants must be associated with the starch metabolism machinery, implicating that starch metabolism genes may be co-expressed in a tissue-specific manner. Because we only examined the expression pattern of a limited number of starch synthesis genes, we extended our analysis to other starch metabolism related genes (Supplemental Table S1 (Fig. 8D, degenerate bases at a defined position were bracketed) was found in 83.3% (20/24) Group I genes (Supplemental Table S3), but only 48.1% (13/27) in Group II genes and 31.2% ± 6.7% (n=10)

DISCUSSION
Starch metabolism is a complex biological process requiring the function of multiple enzymes not only for synthesis but also for degradation. For cells containing chloroplasts, e.g., mesophyll cells, starch is synthesized from the carbon source fixed during photosynthesis. Enzymes required for starch metabolism must therefore be expressed in phototrophic cells. The representative enzymes for starch synthesis (PGI, PGM, AGPase and SSI) examined in our study were indeed expressed in mesophyll cells. Furthermore, data from expression profile of starch metabolism genes, including genes with roles in starch degradation, also support this prediction (Fig. 8C).
The transcription activities of starch synthesis genes showed a developmental expression pattern as the older leaf cells had higher expression level than the younger leaf cells (Fig. 5, A-E). This pattern coincided with the starch accumulation level detected by iodine staining (Fig. 1A), suggesting that the starch synthesis enzymes may be the limiting factors. On the contrary, the GUS activity did not show distinct difference within a cross section of leaves ( Fig. 5X); however, starch was accumulated to a higher level in cells towards light (Fig. 1, B and C). This pattern suggested that limited substrates derived from photosynthesis are available for starch synthesis in those cells shielded from light under a normal growth condition. The relative GUS activities in guard cells versus subsidiary cells (Fig. 5, F-J) and columella cells versus lateral root cap cells (Fig. 5, L, M, O, and V) are correlated with the starch accumulation level in these cell types (Fig. 1, D PGI enzymes, which are consistent with our native gel analysis (Supplemental Fig. S2). In addition, we analyzed PGI activities present in extracts of whole leaf or epidermis of wild type and pgi1 mutant upon heat inactivation treatment (Supplemental Table S4). We found that heat inactivation percentage was higher for extracts from the wild type leaves (37.4%) than that of the pgi1 mutant (2.2%), indicated that pgi1 mutant is deficient in the plastidial PGI which is heat labile (Supplemental Fig. S2). If there are additional PGI enzymes specifically expressed in those cells with starch in the pgi1 mutant (e.g., epidermis), the percentage of heat inaction of PGI activity would be changed notably. The heat inactivation percentage for PGI activity remained similar for the extract from the epidermal cells of wild type (38.8%) to that of whole leaf extract (37.4%) and for the extract from epidermal cells of the pgi1 mutant (2.7%) to that of whole leaf extract (2.2%). These results indicated that epidermal cells of pgi1 mutant did not contain additional PGI activities.
Our data suggested that heterotrophic starch synthesis in the ispH and pgi1 mutant relied on cytosolic G6P imported to their plastids as the substrate.  reflect its biological role in pathways other than starch synthesis, e.g., the production of NADPH through the oxidative pentose phosphate pathway (ap Rees, 1980(ap Rees, , 1985.
Unexpectedly, APL1 was not expressed in root cap cells which could synthesize starch. It has been shown that the adg2-1 mutant, which has a mutation in the APL1, has AGPase formed as a homotetramer with four small subunits (Li and Preiss, 1992 Correlation analyses of transcript profiles could help identify genes which are functionally related. By bioinformatics analysis, we found that starch metabolism genes could be clustered into two groups based on their spatial expression patterns. Expression of Group I genes, involved in both starch synthesis and degradation pathways, had a direct correlation with the spatial distribution of starch in plants (Fig. 8B), suggesting that at least some of these genes play important roles as core enzymes in starch metabolism. For example, in addition to APS1, PGM, and SS1, the spatial patterns of co-expressed ISA1 and ISA2 during plant development (Li et al., 2007) are fully consistent with their functions in starch metabolism as the proteins encoded by these two genes form a heteromultimeric complex. It has been shown that SBEI, SBEIIb, and starch phosphorylase form a protein complex involved in starch synthesis of wheat plastids (Tetlow et al., 2004), suggesting that proteins derived from co- various forms of heterotetramers would respond to substrates and allosteric effectors for starch synthesis and carbon partition in a tissue-specific manner (Crevillen et al., 2003). On the other hand, Group II genes could exert function other than the basic starch metabolism.
Presumably, these genes enable the plants to adjust starch metabolism in response to different environments, e.g., response to stress conditions (Sparla et al., 2006).
Our promoter-reporter analysis of several starch synthesis genes ( Figure 5) indicated that these genes were co-regluated spatially by transcriptional activity of the promoters. Both genes related to starch synthesis and degradation pathways were found in Group I. We suggest that a common mechanism for the transcriptional co-regulation of Group I genes exists, presumably by activation and/or repression, via specific cis-elements in the gene

promoters. A motif with a signature of [ATC][AC][CGT][ATC]AAAGN[AC][GCA][ATC]
was identified in promoter regions of Group I genes (Fig. 8D). Modification of this conserved sequence located upstream of APS1 gene ectopically switched on its promoter activity in some root cells where starch biosynthetic genes are not normally active. Although more analyses are required, our preliminary results suggest that this conserved cis-element may be a target site for transcription repression. We also analyzed genome-wide expression profile with clustering parameters shown in Supplemental Figure S1. The 373 genes (Group I associated genes, see Supplemental Table S5) with similar expression patterns to Group I starch metabolism genes were identified. The genes related to transcriptional regulation were found within Group I associated genes by gene ontology analysis. Some of them may be involved in the regulation of Group I gene expression. Identification of transcription factors and other cis-elements conserved in Group I genes would provide clues for future study on spatial co-regulation, and ultimately provide a strategy to turn on or increase starch synthesis in tissues that make little starch, to increasse starch production for food or biofuels.

Plant materials and molecular analysis
Plants were grown in potting soil at 23°C under a 12 hr/12 hr day-night cycle with 100 μmol quanta m -2 sec -1 fluorescent light. For root tissue, plants were grown hydroponically in water (rice) or in B5 medium with 2% sucrose (Arabidopsis). The rice albino mutant (T58504) used in this study was isolated from a T-DNA insertion mutagenized population.
The albino mutants (ispH and T58504) were grown on MS medium with 2% sucrose.
Standard cloning and northern blot analysis were performed as described (Sambrook and  Table I. The amplified fragments were cloned into pZero2.1, and confirmed by DNA sequencing. Probes used for northern blot analysis were gene specific, except that OscpPGI. For construction of promoter-reporter fusion clones, a 1.7kb fragment of PGI1, a 2.2kb fragment of PGM1, and a 0.65kb fragment of APS1 containing the promoter region were amplified by PCR with designed primers and subcloned in front of the β-glucuronidase (GUS) reporter of a binary T-DNA vector, pPZP212GUS. A 3.9kb fragment of APL1 promoter and a 4.3kb fragment of SS1 promoter were subcloned from genomic clones to pPZP212GUS digested with XbaI/BamHI and HindIII/SmaI, respectively. A 1.8kb fragment of APS1 containing the promoter region was subcloned from a genomic clone in front of the GFP reporter of pPZP212GFP. P APS1m -GUS was contructed from P APS1 -GUS by a PCRdirected mutagenesis with a primer 5' ATTGTATCTAGAGCTCGAAGCACTATGTTTAC 3'. The promoter-reporter constructs were transformed into Arabidopsis by vacuum infiltration with Agrobacterium (GV3101) carrying the binary constructs. Transgenic plants were selected for kanamycin resistance on MS medium, then transferred to pots, and confirmed by Southern blot analysis. For each transgenic constructs, more than 10 independent transgenic lines were examined.

Native gel assay for starch synthesis enzymes
Native gel assays for PGI, PGM (Caspar et al., 1985), AGPase (Yu et al., 2000) and SS1 (Fontaine et al., 1993) activities were carried out according to previously described methods with 7.5% (w/v) polyacrylamide gels. Leaf and root samples were extracted with extraction buffer (100 mM Tris-HCl pH 7.0, 40 mM β-mercaptoethanol, 10 mM MgCl 2 , 100 mM KCl, 15% glycerol, 2 mL per gram fresh tissue). The crude extract was cleared by centrifugation and the supernatant was assayed for enzyme activities. For heat inactivation, the extracts were treated at 50°C for 10 min in a thermocycler. PGI activity was assessed by following the Tek) at 25°C. The reaction mixture (0.1 ml) consisted of 0.1M Tris HCl (pH 7.0), 4.2 mM fructose-6-P, 1.3 mM NADP + , and 1 Unit/ml of glucose-6-P dehydrogenase (Sigma). PGI activity was expressed as unit per μg protein (one unit converting 1 pmole of fructose-6-P to glucose-6-P per min).

Histological detection of starch and GUS activity
Plants were de-pigmented with ethanol to remove chlorophylls, and stained with Lugol's iodine solution (6 mM iodine, 43 mM KI, 0.2 N HCl) to detect starch granules. GUS staining of transgenic seedlings was performed as described. Stained specimens were cleared and mounted with chloral hydrate clearing solution. Stained seedlings were examined under a Nikon SMZ 1500 stereo microscope. To peel epidermis from mesophyll cells, leaf samples were glued between two transparent tapes and the tapes pulled apart; an epidermal layer would adhere to the separated tapes. The epidermis on the tape was covered by GUS staining buffer and incubated in a chamber assembled from a microscope slide and a cover slid and incubated at room temperature overnight. Epidermis was examined with an Olympus BHT microscope. Images were captured with a Nikon DS-5Mc digital camera.

Confocal and transmission electron microscopic analyses
A confocal laser-scanning microscope (Zeiss LSM-510) using the 488 nm laser light for excitation was used to examine auto-fluorescence of chlorophyll (emission filter was long pass 650 nm), GFP fluorescence (emission filter was band pass 505-530 nm) and propidium iodide fluorescence (emission filter was long pass 560 nm) in cells. Roots of the transgenic plants carrying P APS1 -GFP fusion construct were stained with 10 μg/ml propidium iodide for 1 minute and mounted in water, then examined for GFP and propidium iodide fluorescence.
For electron microscopic analysis, Arabidopsis tissues were fixed in a solution of 4%

Microarray data analysis
The cell type specific expression data of roots (Birnbaum et al., 2003)

Supplemental Figures:
Supplemental Figure S1.           S1): Group I genes were mainly expressed in columella cells, but not for Group II genes (B). In addition, Group I genes co-expressed in various tissues (see Supplemental Table S2 for conditions of tissues), but not for Group II genes (C). Sequence logo of a motif statistically over-represented in promoter regions (see Supplemental Table S3) of coregulated Group I genes was shown (D). and mutant (P APS1m -GUS) were illustrated (A). β-Glucuronidase (GUS) activities present in seedlings, epidermal peels, lateral roots, and root tips of 9-day-old transgenic lines carrying P APS1 -GUS or P APS1m -GUS were shown (B).The inset in the lateral root of P APS1m -GUS showed that GUS activities were present in stele cells at lateral root base.