Global Transcription Profiling Reveals Multiple Sugar Signal Transduction Mechanisms in Arabidopsis

Effect of Nitrogen on Transcriptional Patterns

A previous microarray study using exogenous nitrate (Wang et al., 2000) revealed a relatively short list of genes that had altered transcriptional patterns. This study compared genes that were transcriptionally regulated by low (250 μM) and high (10 mM) nitrate levels when supplemented with 0.5% sucrose as a carbon source. Out of ∼5500 unique genes, 49 showed a twofold or greater change in mRNA levels. A more recent microarray analysis measuring nitrate response in Arabidopsis suggested a larger number of genes were regulated by nitrate (Wang et al., 2003). In the latter experiment, plants were grown in medium containing 0.5% sucrose, and plants were allowed to deplete their sole nitrogen source, 2.5 mM ammonium succinate, over a 10-d period before being treated with 250 μM KNO₃ for 20 min. Using a twofold cutoff and the Wilcoxon's signed rank scores of I (increase) or D (decrease), it was found that 251 genes were induced and 78 genes were repressed in root tissue, whereas in shoot tissue 76 genes were induced and two were repressed. The major differences between the two studies were that in the latter study, the roots were analyzed separately from the shoot tissue and the period of nitrogen starvation was longer. In our study, we used whole plants, in which shoot mass outweighed root mass by 22.8-fold, no carbon source was supplied, and a relatively short period (24 h) of nitrogen deprivation was used. To exclude targets with inconsistent results, we used an RCBD analysis cutoff of P ≤ 0.001 with a twofold change to filter our data. When we tested the effects of higher concentrations of nitrogen (40 mM nitrate and 20 mM ; standard for MS medium) using a sugar-free medium, only 106 and 129 genes showed greater than twofold induction or repression, respectively (Figure 1; see Supplemental Table 1 online). When an additional filter was applied to eliminate genes with expression near background levels, 24 upregulated and 37 downregulated genes were selected as nitrogen regulated. Some of the nitrogen-regulated genes were identified in the earlier microarray studies as being associated with nitrate/nitrite assimilation (Wang et al., 2000; Wang et al., 2003): among these were nitrate reductase 1 (NIA1), urophorphyrin III methylase, and ferredoxin nitrite reductase (Table 1). Markedly, two genes associated with ammonium assimilation in shoots, Asn synthetase (ASN2) and NADH-dependent Glu synthase (Temple et al., 1998; Wong et al., 2004), showed stronger upregulation in our study compared with the latter nitrate microarray study (Table 1) (Wang et al., 2003), presumably because we restored both nitrate and ammonium to our nitrogen-deprived plants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Glucose Has Profound Effects on Gene Expression Compared with Inorganic Nitrogen in 6-d-Old Arabidopsis Seedlings Predominantly Consisting of Shoot Tissue.

To remove inconsistent replicates, log₁₀ normalized signal scores were subjected to RCBD analysis (P ≤ 0.001) before twofold filtering.

We also examined the effects of nitrogen treatment when glucose was also supplied. Our results revealed that the induction of many of the previously reported nitrate-responsive genes actually required the presence of both nitrogen and sugar (Table 1), suggesting an interaction between sugar and nitrogen. The interaction is further supported by the results of cluster analysis: ∼8% of glucose-responsive genes showed altered expression when nitrogen was also present (Figure 2, clusters 1, 7, and 8). Only a few of these genes, primarily those in clusters 1 and 7, were regulated by nitrogen alone. Glucose and nitrogen appear to have synergistic effects on the induction of some genes (Figure 2, cluster 8). For example, uroporphyrin III methyltransferase (UPM1) and glucose-6-phosphate dehydrogenase (264859_at) showed modest induction in the presence of either glucose or nitrogen, but the combination of both nutrients increased expression far greater than if the effect of each nutrient were merely additive (Table 1). For UPM1, this synergistic effect was verified by RNA gel blot analysis (Figure 3C). In other examples, the regulation of gene expression occurred via an antagonistic interaction between C and N signals (Figure 2, cluster 1). A stress-related gene previously identified as SAG21 (At4g02380) was reported to be upregulated 4.5-fold by nitrogen when compared with a control containing sucrose (Wang et al., 2000); we observed that nitrogen without glucose minimally regulated SAG21 (1.3-fold) but also found that the gene was downregulated by glucose 4.6-fold when compared with a C- and N-free control (Figure 3D). When sugar and nitrogen were both available, nitrogen derepressed the glucose effect and brought transcription of SAG21 near to the C- and N-free control levels (down 1.4-fold). Of the 61 nitrogen-regulated genes showing a more than twofold expression change, more genes were found to be downregulated by nitrogen than upregulated—this is not observed in previous microarray studies. This difference is likely attributable to the presence of sugar in the earlier experiments and the use of a C- and N-free control under our conditions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Regulation of Gene Expression Orchestrated by Glucose and Nitrogen.

Cluster analysis was conducted using GeneCluster2 (Golub et al., 1999) using the genes identified in Figure 1, except those showing significant regulation by 3-OMG were removed from consideration. A self-organizing map (SOM) was generated for genes showing greater than a twofold change with expression above background/noise levels. Blue lines represent the mean expression, and the area between red lines represents the range of values within the cluster. This SOM explained 95.1% of the variance occurring in the data set. Value associated with each cluster represents the number of genes with similar behavior.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Microarray Data Validation by RNA Gel Blot and RT-PCR Analyses.

Genes chosen for analysis include glucose downregulated genes (A), glucose upregulated genes (B), a gene upregulated specifically by glucose and nitrogen (C), nitrate upregulated genes (D), and two unregulated genes (E).

Effect of Glucose on Transcriptional Patterns

In contrast with nitrogen, glucose was more potent in regulating transcription under the conditions we used (Figure 1). Of genes regulated by carbon and/or nitrogen, cluster analysis revealed that glucose altered transcription for a large portion of genes, whereas nitrogen treatment had little to no effect (Figure 2, clusters 0, 2, 3, 5, and 6). However, nitrogen could modulate the glucose effect for a smaller subset of genes (Figure 2, clusters 1, 7, and 8). Using an RCBD analysis cutoff of P ≤ 0.001 and a threefold change to filter our data, 534 and 444 genes were found to be downregulated and upregulated by glucose, respectively (Figure 4; see Supplemental Table 2 online). Nearly all types of genes were affected by glucose, ranging from stress responses and cellular metabolism to those involved in signaling/gene regulation. Possible gene functions were determined using a variety of methods, including searching gene ontologies (Rhee et al., 2003; Bard and Rhee, 2004; Camon et al., 2004; Harris et al., 2004), conducting pathway analyses (Mueller et al., 2003), and searching the literature. Our results are consistent with the findings from a recent microarray study showing that sugar regulates a broad range of gene types (Thum et al., 2004). Unlike nitrogen regulation, glucose regulation was relatively independent of nitrogen status; however, we cannot rule out a potential role for nitrogen in regulating these genes under different conditions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Glucose Regulates Genes with Diverse Functions.

Shown are genes responding to glucose with at least threefold change after normalizing data and conducting RCBD analysis at P ≤ 0.001. Putative functions were determined using spot annotations (The Arabidopsis Information Resource; http://arabidopsis.org), gene ontology searches (http://www.geneontology.org), pathway analyses, and literature review.

Transcriptional Upregulation by Glucose Largely Requires de Novo Protein Synthesis

To confirm the results of the microarray analysis, we conducted RNA gel blot analyses and RT-PCR with a sampling of genes. RNA gel blot analyses were generally effective from genes having signal score greater than 1000, whereas genes with lower expression levels required RT-PCR for detection. The RNA gel blot/RT-PCR analyses were conducted using two sets of RNA from identically prepared plant material except that one set was treated with the protein synthesis inhibitor cycloheximide (CHX) 1 h before the 3-h pulse treatment. In plant material not treated with CHX, results from the RNA gel blot/RT-PCR consistently concurred with the results obtained from the microarray analyses (Figure 3). Some genes had enhanced expression in the presence of CHX compared with those not CHX treated (Figure 3); this is consistent with prior observations of enhanced mRNA stability upon CHX treatment (Baker and Liggit, 1993; Goda et al., 2002). The relative stabilization of some transcripts upon CHX treatment indirectly suggests that posttranscriptional modifications may be occurring. Curiously, CHX treatment did not appear to affect glucose repression; but CHX clearly diminished glucose induction (Figures 3A and 3B), even in cases where CHX stabilized transcript levels. Interestingly, hexokinase 1 and hexokinase 2, dual functional enzymes involving in sugar signaling (Jang et al., 1997; Moore et al., 2003), were no longer induced when CHX was present (Figure 3B). This suggests that glucose repression may not require de novo protein synthesis, but glucose induction appears largely to be a multistep response requiring de novo protein synthesis.

To determine whether CHX treatment disrupts glucose induction on a global scale, microarray analyses were conducted with CHX using the same plant material used for the RNA gel blot analysis. Two independent biological replicates were conducted for each experimental condition containing added CHX; each CHX replicate set was grown concurrently with a set of the non-CHX treated plants used for GeneChip analysis. The FDRs for plants treated with CHX were similar to those without CHX (see Supplemental Figures 3A to 3D online). We were primarily interested in determining how CHX affected expression of the genes regulated without CHX by glucose, nitrogen, or glucose and nitrogen, so CHX data were appended to the non-CHX data described in Figure 2, and genes showing similar expression patterns for both CHX and non-CHX treatments were identified using SOM analysis software (Golub et al., 1999). As shown in Figure 5, only 18% of glucose-inducible genes remained inducible in the presence of CHX; in contrast with glucose induction, 64% of glucose repressible genes were relatively unaffected by CHX. These results suggest that on a global scale, glucose induction is a multistep event requiring de novo protein synthesis, whereas glucose repression occurs to a large extent without de novo protein synthesis. To further analyze the effect of CHX on gene expression, we examined 85 carbohydrate metabolism–related genes out of the 978 glucose-regulated genes described in Figures 4A and 4B (Figure 6; see Supplemental Table 3 online). For a portion of glucose-inducible genes, addition of CHX reduced overall expression to near background levels, thus they could not be meaningfully analyzed by threefold filtering. On the other hand, most of the repressible genes were above background levels, revealing remarkably similar expression profiles for both CHX-free and CHX-treated plants (Figure 6). These results mirror the findings from the RNA gel blot analysis (Figure 3), indicating that the repressive effect of glucose upon transcription remains intact even when de novo protein synthesis is blocked. When the results from Figures 3 and 5 are considered together, they consistently indicate that transcriptional repression by glucose is relatively unaffected by CHX treatment. It isn't totally certain whether the loss of glucose induction upon CHX treatment is because of a direct effect on glucose regulatory mechanisms or a global reduction of expression level; however, the relative stabilization of glucose-repressed transcripts by CHX (Figures 3 and 5) suggests that the former alternative may be more likely.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Glucose Induction Often Requires de Novo Protein Synthesis.

Frequency of glucose induction versus glucose repression in the presence of CHX. Expression patterns with and without CHX were determined for the genes identified in Figure 2 using SOM software.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Expression Patterns of Carbohydrate-Related Genes Identified in Figure 4 with or without CHX.

Hierarchical average linkage clustering with correlation measure–based distance (uncentered) was used for the analysis. Red or green represents upregulation or downregulation, respectively, and gray represents either genes at background/noise levels or changes below the fold-change cutoff.

Transcription Factors Are Differentially Regulated by Glucose

Although sugars are known to have a broad effect on gene expression, it is still intriguing that a large number of transcription factors (TFs) were glucose regulated. Eighty-two glucose-responsive TFs were identified using a threefold filtering criterion; interestingly, a majority of them was downregulated. A similar trend was found when a twofold filter was applied (Figure 7A). Of the TFs identified, most were relatively unaffected by nitrogen (data not shown). Glucose affected 22 families of TFs, including bHLH, MYB, AP2, and various zinc finger–containing factors. Glucose-regulated TFs account for 8.3% (82/978) of all glucose-regulated genes; this represents relative enrichment of TF in glucose response because TFs are estimated to account for 5 to 7% of the Arabidopsis genome (Riechmann and Ratcliffe, 2000; Jiao et al., 2003). When glucose-regulated TFs are compared with the population of TFs in the Arabidopsis genome (Riechmann et al., 2000; Jiao et al., 2003), factors involved with stress responses (such as some AP2/ERF proteins) appear to be overrepresented upon glucose treatment; by contrast, relatively few developmental factors (such as MADS) appear to be glucose regulated (Figure 7B). Like other genes (Figures 5 and 6), the transcription of TFs was also affected by CHX; whereas up to 95% of the glucose induction was abolished while glucose repression was eliminated to a lesser extent (64%).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Transcription Factors Are Differentially Regulated by Glucose.

(A) Number of all genes versus transcription factors upregulated or downregulated by glucose with a twofold or threefold cutoff.

(B) Distribution comparison of glucose-regulated transcription factors with all transcription factors in the Arabidopsis genome (Jiao et al., 2003). Percentage of glucose responsive TFs is derived from the number of each category versus total number of glucose responsive TFs.

Is Sugar-Hormone Cross Talk Mediated through the Change of Hormone Biosynthesis and Perception?

A variety of genetic screens have repeatedly identified genes involved in ABA biosynthesis or response or ethylene perception as being critical for sugar signaling. Loss-of-function of ABA1 (Arenas-Huertero et al., 2000), ABA2 (Arenas-Huertero et al., 2000; Cheng et al., 2002), ABA3 (Arenas-Huertero et al., 2000), ABI4 (Arenas-Huertero et al., 2000; Huijser et al., 2000; Laby et al., 2000; Rook et al., 2001; Arroyo et al., 2003), ABI5 (Arenas-Huertero et al., 2000; Brocard et al., 2002; Arroyo et al., 2003; Brocard-Gifford et al., 2003), and ABI8 (Brocard-Gifford et al., 2004) causes tolerance to developmental stresses caused by exogenous sugar. A considerable amount of genetic evidence also supports an interaction between sugar and ethylene signaling pathways (Zhou et al., 1998; Gazzarrini and McCourt, 2001; Gibson et al., 2001; Rolland et al., 2002; León and Sheen, 2003). Whereas ctr1 is less sensitive to high concentration of glucose during early seedling development, etr1 and ein2 show enhanced developmental arrest caused by sugar (Zhou et al., 1998; Gibson et al., 2001; Cheng et al., 2002; Arroyo et al., 2003; Price et al., 2003). These studies raise a possibility that genes involved in sugar-hormone cross talk might be transcriptionally regulated by glucose. Using a twofold filtering criterion, none of the ABA-related genes previously associated with sugar sensitivity were selected as being glucose-regulated under our conditions (Figure 8A; see Supplemental Table 4A online). However, several ethylene biosynthetic and signaling genes were repressed by glucose, including CTR1 (Figure 8A) and genes associated with 1-aminocyclopropane-1-carboxylate metabolism (Figure 8B; see Supplemental Table 4B online). Notably, three ethylene biosynthetic genes were downregulated approximately twofold by glucose in the presence of CHX (Table 2). Interestingly, two critical TFs involved in ethylene signaling, EIN3 and EIL1, were also repressed by glucose (Table 2). Together, these results suggest that the transcriptional repression of ethylene biosynthesis and perception may be an early event during glucose signaling. Many of the same ethylene-related genes were recently found to be induced upon prolonged light deprivation without exogenous sugar (Thimm et al., 2004). Because the degradation of EIN3 and EIL1 protein is also enhanced by glucose (Yanagisawa et al., 2003), glucose signaling is likely to affect the ethylene response at multiple levels.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Glucose Affects Expression of Ethylene and Stress Associated Genes.

Shown are hierarchical average linkage clustering analyses. Red or green represents upregulation or downregulation, respectively, and gray represents either genes at background/noise levels or no changes with specified cutoff.

(A) Nutrient response of genes implicated in sugar signaling based on genetic studies (León and Sheen, 2003; Gibson, 2004). ABI4 is not included because expression levels were near background/noise levels. None of these genes showed a more than twofold change in the presence of CHX; however, CTR1, EIN3, and EIL1 were repressed by glucose more than 1.5-fold in the presence of CHX (Table 2).

(B) Nutrient response of hormone biosynthetic genes. Filtering criteria were relaxed to twofold for CHX-treated plants.

(C) Numerous heat shock proteins are affected by glucose.

(D) Other stress-associated genes are highly glucose-responsive.

Only a few other hormone biosynthetic genes with an unknown role in plant sugar response were significantly (threefold) regulated by glucose. The ABA biosynthetic gene 9-cis-epoxycarotenoid dioxygenase (NCED1) (Seo and Koshiba, 2002) was upregulated by glucose (Figure 8B), although NCED1 also responded to the osmotic control 3-OMG. Nitrilase 3, which is involved in indole-3-acetic acid biosynthesis (Kobayashi et al., 1993), was more specifically upregulated by glucose, whereas two genes associated with jasmonic acid biosynthesis, allene oxide synthase (Simpson and Gardner, 1995) and 12-oxophytodienoate reductase (Schaller et al., 2000), were glucose induced and repressed, respectively. Further studies are needed to determine whether any of these genes are critical components of the sugar response.

Sugar and Stress Response

Another intriguing result is that many stress-related genes are induced by sugar (Figure 8C; see Supplemental Table 4C online). Heat shock proteins are molecular chaperones that assist in the proper conformation of proteins and are strongly upregulated when an organism undergoes a stress (such as heat) that can cause protein misfolding (Larkindale and Knight, 2002). RNA gel blot analysis reveals that cytosolic heat shock protein 70 (At-hsc70-3) was clearly upregulated upon glucose or glucose/nitrogen treatment, but regulation of this gene may not be a primary response to glucose because CHX prevented glucose induction of transcription (Figure 3B). The upregulation of an hsp90 is particularly intriguing because hsp90s have been shown to act as buffers in the expression of genes, revealing otherwise hidden phenotypes when hsp90 protein levels become insufficient (Queitsch et al., 2002; Rutherford, 2003). It's possible that the glucose induction of heat shock genes is merely the result of increased metabolic activity. However, other genes associated with stress responses (Knight and Knight, 2001), including those associated with ABA-mediated response, inositol metabolism, and Ca²⁺ regulation, are also largely upregulated by glucose (Figure 8D; see Supplemental Table 4D online). Conceivably, the classic glucose-sensitive phenotype might be partly attributed to a stress response, causing the typical sensitive phenotype except in cases where the normal ABA or ethylene response pathways are altered. The heat shock protein-related stress response is likely to be an indirect event, though, because most of these stress-related genes are no longer regulated by glucose when de novo protein synthesis is blocked.

Transport Proteins Are Largely Regulated According to Function

For many of the functional categories shown in Figure 4, glucose treatment appeared to cause a mixed response, where some genes were upregulated and other genes of similar function were downregulated. One class of glucose-regulated genes where discernable patterns were more evident was the transport proteins. Data in this analysis were filtered after RCBD analysis (P ≤ 0.001) using a threefold cutoff. Regulation of transport proteins by glucose appeared to be largely determined by function. Other than one gene with glucose phosphate antiporter activity, genes associated with monosaccaride transport, peptide transport, and purine transport were consistently repressed by glucose (Figure 9; see Supplemental Table 5 online). When we analyzed the expression of 13 glucose-responsive monosaccharide transporters—among which sugar transport protein1 (STP1), STP4, STP13, and STP14 were previously identified (Williams et al., 2000)—10 were downregulated at least twofold by glucose, including all four known STPs. Although only STP1 and STP4 are known to be high affinity transporters and STP1 activity is highly repressed by exogenous glucose (Sherson et al., 2003), the results here raise the possibility that the other eight genes might also be high affinity transporters with a low K_m, where the transcription of these genes may be feedback repressed via the relatively high exogenous glucose level (167 mM). Our findings are consistent with the models in yeast and humans that sugar homeostasis is controlled by sugar transporter activities coupled with sugar-mediated transcriptional regulation (Rolland et al., 2001). Unlike stress-related proteins, glucose repression of monosaccharide transporters is relatively unaffected by CHX (Figure 10). Although more than a dozen of glucose responsive monosaccharide transporters were identified, only one disaccharide transporter, SUC2, showed a modest glucose response in our experiment. This is in agreement with the finding that sucrose-specific signaling pathway was used in transcriptional regulation of sucrose transporter (Choiu and Bush, 1998).

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Nutrient Response of Various Transporters.

Shown are hierarchical average linkage clustering analyses. Red or green represents upregulation or downregulation, respectively, and gray represents either genes at background/noise levels or no changes with specified cutoff.

• Download figure
• Open in new tab
• Download powerpoint

Figure 10.

Multiple Sugar Signaling Pathways Revealed by the Regulation of Sugar Transporters.

Glucose has profound effects on the expression of monosaccharide transporters. By contrast, only one disaccharide transporter is affected by glucose, consistent with the idea that disaccharide transporters are uniquely regulated by disaccharides (Choiu and Bush, 1998).

Conversely, genes associated with transporting ions, water, and inorganic metabolites, such as nitrate, phosphate, and sulfate, were generally upregulated upon glucose treatment (Figure 9). The upregulation of these transporters is probably associated with an increase in general metabolism caused by the readily available sugar. This notion is supported by the loss of glucose induction of these genes in the presence of CHX (Figure 9). Other metabolic genes, such as ribosomal proteins, detoxification proteins, and DNA or RNA modifying proteins, were also generally upregulated (data not shown), confirming earlier findings that metabolic activity is increased when sugar becomes available (Thum et al., 2004).

Glucose Regulates Genes Related to Carbohydrate Metabolism

Although Arabidopsis is an oilseed, starch is used in the vegetative stage as a reserve for carbon. Starch synthesis typically requires starch synthase, starch branching enzyme, and glucose-1-phosphate adenylyltransferase (Fernie et al., 2002). Amylases are also involved in starch metabolism. An Arabidopsis isoamylase mutant has been shown to reduce the accumulation of starch while increasing the accumulation of the water-soluble polysaccaride phytoglycogen (Zeeman et al., 1998). Recently, transgenic potato (Solanum tuberosum) with antisense expression of Arabidopsis chloroplast-targeted β-amylase has been shown to overaccumulate starch in leaves and reduce starch breakdown during dark treatment (Scheidig et al., 2002). When we treated dark-adapted seedlings with glucose, many of the genes associated with starch biosynthesis were upregulated compared with the carbon-free control (Figure 6). A starch synthase, the 1,4 α-glucan branching enzyme SBE2.2 precursor transcript, three glucose-1-phosphate adenylyltransferase genes (including APL3), and an isoamylase-like gene (255070_at) were all significantly upregulated upon glucose treatment. Many of these same genes were shown to be downregulated upon prolonged exposure to darkness in the absence of sugar (Thimm et al., 2004). RNA gel blot analysis confirms induction of a putative starch synthase (Figure 3). Two β-amylases were upregulated, indicating that starch catabolism is also taking place. Collectively, these results suggest that glucose is a critical signal for starch metabolism; this is consistent with the findings of Thum et al. (2004).

Patterns of expression from other genes associated with sugar metabolism reveal a more complex regulatory mechanism. Trehalose has been shown to induce APL3 expression and thus promote starch synthesis in source tissues; trehalose has also been shown to serve as a stress protection metabolite (Goddijn and van Dun, 1999; Fritzius et al., 2001; Eastmond et al., 2002; Elbein et al., 2003). Trehalose-6-phosphate synthase 1 (TPS1) catalyzes the first step in trehalose biosynthesis (Eastmond et al., 2002). Curiously, we find that one trehalose-6-phosphate synthase-like protein (TPS5, 245348_at) was induced by glucose, whereas three other putative trehalose-6-phosphate synthases, TPS8 (Figures 4 and 7B; 264339_at), TPS9 (263019_at), and TPS10 (264246_at), were strongly repressed by glucose (Figure 6). Whereas other TPS genes were not affected by sugar, the differential regulation of TPS genes by glucose is likely an advantage for adaptation, where differential expression within the same gene family upon a given stimulus results from concurrent spatial- and temporal-specific controls (Eastmond and Graham, 2003). Likewise, a large set of UDP-glucose glucosyltransferases were variably regulated by glucose (Figure 6). UDP-glucose glucosyltransferases are involved in a wide range of functions ranging from regulating phytohormone activity to making macromolecules more soluble (Wetzel and Sandermann, 1994; Jones and Vogt, 2001; Lim et al., 2002). Together, these results suggest that many of the effects glucose has on sugar metabolism do not constitute global responses; rather, the glucose response appears to be targeted to fulfill the specific requirements during growth and development.

Exogenous Glucose Is More Effective Than Nitrogen in Regulating Genes Associated with Nitrogen Metabolism

Sugars and inorganic nitrate are important signaling molecules for adjusting nitrogen and reduced-carbon utilization within a plant (Coruzzi and Bush, 2001; Coruzzi and Zhou, 2001; Forde, 2002; Stitt et al., 2002). Carbon and nitrogen have matrix effects, where genes associated with nitrogen assimilation are upregulated when reduced carbon is abundant and downregulated when reduced carbon is scarce or organic nitrogen is abundant (Coruzzi and Zhou, 2001). When we examined the genes associated with nitrogen assimilation and amino acid metabolism identified in a recent study (Thimm et al., 2004), exogenous glucose appeared to regulate these genes much more profoundly than nitrogen (Figure 11; see Supplemental Table 6 online). Glucose tended to upregulate genes associated with amino acid biosynthesis and downregulate genes related to amino acid catabolism. This concurs with findings from Thimm et al. (2004), where amino acid breakdown was enhanced and biosynthesis was inhibited when plants were exposed to prolonged darkness without exogenously supplied sugar. There were some notable exceptions to the trend we observed. Two glutamate dehydrogenase (GDH) genes, which are involved in ammonium utilization and detoxification, and glutamine-dependent asparagine synthetase (ASN1), associated with the storage and/or transport of nitrogen from sources to sinks, were both downregulated by glucose—GDH has been implicated in regulating carbon-nitrogen status (Stitt et al., 2002) and ASN1 has been previously demonstrated to be tightly regulated by sugars (Lam et al., 1998). RNA gel blot analysis of ASN1 and GDH2 confirms the prior observations (Figure 3A). Like putative starch synthase and putative trehalose-6-phosphate synthase (TPS8), ASN1, GDH2, and tat binding protein (similar to an aminotransferase) have enhanced expression in the presence of CHX (Figure 3A). By contrast, genes associated with assimilation of inorganic nitrate, including NIA1 and ferredoxin-nitrite reductase, were strongly upregulated in the presence of exogenous inorganic nitrogen, even without the presence of exogenous sugars. An earlier study demonstrated that the application of sugars such as sucrose could induce NIA1 expression upon carbohydrate deprivation when nitrate was present (Cheng et al., 1992). Our data show the complementary result, where a nitrogen source is required before a sugar-like glucose can induce NIA1 (Table 1). As predicted by the matrix effect model, the availability of sugar did promote the transcription of most genes involved in nitrogen bioaccumulation in our study. However, for NIA1, exogenous nitrate and sugar are required before induction can occur.

• Download figure
• Open in new tab
• Download powerpoint

Figure 11.

Genes Associated with Nitrogen Metabolism Are Predominantly Regulated by Glucose.

The selected genes were normalized and subjected to RCBD analysis (P ≤ 0.001) and showed a more than twofold transcriptional change.

Global Transcriptional Response to Carbon and Nitrogen in Arabidopsis

This study indicates that glucose affects the transcription of a relatively large proportion of the Arabidopsis genome. Whereas some genes are directly regulated by glucose, others are likely affected indirectly by altered metabolic activities induced by glucose (Figure 12). The scale of the transcriptional change is comparable to the sugar response in other eukaryotes. Depending on the stringency of the filter criteria and the number of genes tested, previous microarray analyses have shown that sugar deprivation significantly alters expression for ∼3.8% of the genes in a Drosophila array (fourfold filtering), whereas up to 27% of the genes on a yeast array (twofold filtering) were affected by glucose starvation (DeRisi et al., 1997; Lin et al., 2002; Zinke et al., 2002; Boer et al., 2003). By contrast, the role of inorganic nitrogen is somewhat less pronounced, where although exogenous nitrogen did modify expression of ∼8% of the glucose-responsive genes (Figure 3), very few genes were regulated by nitrogen alone. This is consistent with an earlier conclusion that regulation of nitrate reductase in tobacco (Nicotiana tabacum) becomes insensitive to nitrate or nitrogen metabolite regulation when sugar levels drop below a certain threshold (Klein et al., 2000). The recent study by Wang et al. (2003) revealed additional nitrogen-regulated genes that were not identified under our conditions. Their study differs from ours in many respects. In particular, their growth medium contained sucrose, and they used a shorter exposure with lower concentration of nitrate. The relatively high concentration of and used in our experiments may have repressed the expression of some nitrate responsive genes—for instance, NRT2.1, a high-affinity nitrate transporter whose nitrate induction is repressed by (Gansel et al., 2001; Glass et al., 2002), was not induced under our conditions. Also, the plant material was grown to a different developmental stage and had a longer time period to deplete internally sequestered nitrogen reserves. However, the key difference between their study and ours was that their analysis measured expression changes in shoot and root independently, whereas in our whole plant samples the mass of shoot tissue outweighed root tissue by more than 20-fold. As a result, our whole plant samples behaved remarkably similarly to their shoot tissue. In fact, the differences in gene expression between samples treated with glucose and glucose/nitrogen in our experiment were comparable to the changes seen for their control versus nitrate treatment in shoots because sucrose was used as basal ingredient in all treatments of their experiment (Table 1). It's likely that the glucose response occurs mainly in the shoot and the nitrate response occurs mainly in the root. The higher proportion of glucose-regulated genes seen in our study is probably attributable in part to the overrepresentation of shoot tissue. It's intriguing that a few of the carbohydrate metabolic genes we identified as being glucose regulated were identified in their study as being nitrate regulated in root. Depending on the temporal or spatial conditions, it's possible that both carbon and nitrogen deprivation might regulate genes such as phosphoglycerate mutase (262180_at) and trehalose-6-phosphate synthase (TPS9 and TPS10). Indeed, genes associated with glycolysis and the pentose phosphate pathway have been implicated in nitrogen assimilation and metabolism (Weber and Flugge, 2002; Wang et al., 2003); yet in our study, nitrogen didn't alter expression of any of these genes unless glucose was also present. However, there were considerably more carbohydrate-related genes regulated by glucose, including many associated with starch metabolism, that were unaffected by nitrogen. To date, relatively few genes associated with carbon metabolism have been found to be induced upon addition of nitrate in either Arabidopsis or tomato (Lycopersicon esculentum) (Buckhout and Thimm, 2003).

• Download figure
• Open in new tab
• Download powerpoint

Figure 12.

A Proposed Model Summarizes the Metabolic and Signaling Roles of Glucose.

Arrows pointing upward are induction and those pointing downward are repression.

It's possible that the difference in scale between the glucose and inorganic nitrogen responses may be partly attributable to the control mechanisms needed to maintain adequate nutrient levels. When plants are grow in the wild, inorganic nitrogen is usually the nutrient that most limits growth (Forde, 2002). Under normal conditions where plants are not starved for carbohydrate, plants appear to be adapted to assimilate nitrate from available environmental sources (Martin et al., 2002). This is evident in split-root experiments, where a nitrate-treated root half is upregulated for nitrate assimilation compared with a nitrate-deprived root half even when the treatments for both root halves are maintained over a long period (Forde, 2002). Circadian rhythm and the availability of sugar are also known to affect regulation of nitrogen assimilation genes (Cheng et al., 1992; Harmer et al., 2000; Martin et al., 2002). Nevertheless, the degree of coordination necessary for nitrogen assimilation may be relatively simple, so fewer genes would require regulation when an inorganic nitrogen source becomes available. Reduced carbon is a critical starting material for most biosynthesized molecules, and the energy needed to make sugars is available from the environment only during the day when light is present (Winter and Huber, 2000). To maintain adequate levels of reduced carbon as the availability of light varies, plants use newly synthesized sugars when light is present and rely on breakdown of starch reserves during the night (Schleucher et al., 1998). To maintain homeostasis and to take advantage of opportunities when sugar can be made, genes would require precise regulation, coordinating the assimilation of CO₂ as well as the synthesis/mobilization of starch (Huber et al., 1993; Quick, 1996); coordination also would be required to transfer reduced carbon from source to sink tissues (Quick, 1996). Additionally, circadian rhythms affect genes associated with sugar utilization and homeostasis (Harmer et al., 2000), and our data show that once the nitrogen is assimilated, glucose is a key regulator of organic nitrogen metabolism. Consequently, the relative complexity of the response needed to maintain sugar homeostasis may mean that relatively more genes require transcriptional regulation. The difference in complexity for controlling sugar and nitrogen levels may provide a partial explanation why a greater proportion of the genome was identified as being glucose regulated rather than nitrogen regulated when we examined global expression patterns. However, this snapshot observation may not fully represent the true response to either nutrient. Plants may respond differently to exogenous and endogenous supplies of sugar, and it's possible that the full response to inorganic nitrogen may not be fully observed within a 3-h timeframe. Alternatively, organic nitrogen sources might be more effective signals in gene regulation; in this study, it isn't clear to what extent the applied inorganic nitrogen is being assimilated to organic forms. A time-course experiment with detailed metabolite profiling would provide insights in the full effects of nitrogen provision. Gene expression profiling using mutants with elevated endogenous sugar levels may address whether exogenously supplied and internally assimilated sugar cause distinct signaling events.

Differential Regulation of Glucose Induction versus Glucose Repression

One unexpected finding was that distinct regulatory mechanisms appear to be controlling transcript abundance when comparing glucose upregulated and downregulated genes. Although transcript abundance can potentially result from posttranscriptional modification (Chan and Yu, 1998a, 1998b; Lam et al., 1998; Cheng et al., 1999), it is likely that some of the expression differences seen upon glucose treatment are the result of transcriptional regulation. Gene transcription is either positively or negatively regulated via the action of transcriptional activators or repressors, respectively. Both types of control proteins are typically modular, where a DNA binding domain typically tethers the regulator to the promoter DNA, whereas a functional domain causes the actual activation or repression of the gene (Ptashne and Gann, 2002). Activators typically function through the recruitment of histone-modifying and -remodeling activities, the direct contact of the regulator with components of general transcription machinery, and the interaction of the transcriptional complex with other coactivators; by contrast, transcription repressors antagonize many of these functions (Workman and Kingston, 1998; King and Kingston, 2001; Ptashne and Gann, 2002). Our results indicate that glucose affects gene transcription via two different mechanisms. The first mechanism is controlled by a process where de novo protein synthesis is not required (CHX insensitive): this is the mechanism used predominantly in glucose repression. The second mechanism, which is blocked by CHX, affects some glucose repressible genes and a large portion of glucose inducible genes. This suggests that glucose induction in plants requires multiple steps, presumably caused by the change of metabolic activities. Loss of glucose response caused by CHX may be because of the inhibition of signaling component, transcription factor, or coactivator biosynthesis, which is required for the induction/repression of certain glucose responsive genes. This possibility may be verified by linking upstream transcription factors with the cis-regulatory elements of downstream targets using an approach such as chromatin immunoprecipitation coupled GeneChip analysis (Horak and Snyder, 2002; Lee et al., 2002). There is a precedent in yeast where glucose initially regulates activators through transcriptional repression without requiring de novo protein synthesis (Johnston, 1999; Rolland et al., 2001, 2002; Schuller, 2003). Glucose initially activates the yeast Mig1 repression complex via the inhibition of Snf1 kinase activity (Rolland et al., 2002; Schuller, 2003). Mig1 transcription factor recruits corepressors Ssn6 and Tup1 to form a complex, which in turn represses a diverse array of genes including several gene family–specific transcriptional activators involved in alternative carbon usage. The repression of these activators leads to a profound repression of downstream genes whose expression is dependent on these activators. In our study, a majority of transcription factors is repressed by glucose; this raises the possibility that similar glucose repression mechanisms may be conserved in plants. Curiously, a recent microarray study found that sugar- or light-regulated transcription factors were underrepresented compared with the total number of TFs present on the microarray (Thum et al., 2004). This is contrary to our results, where expression of TFs was enriched upon glucose treatment. There are many possible explanations for the difference—1% sucrose rather than 3% glucose was used in their experiments, their sugar treatment time was 8 h as opposed to our 3-h treatment, their plant material was harvested at a later developmental stage, and the results from their 8000-gene microarray chip may not be representative of the full TF response.

Cross Talk between Ethylene, ABA, and Sugar Signaling Pathways

The ethylene signal is transmitted via a pathway that includes a transcriptional cascade, and EIN3 has been identified as a critical component within this cascade (Guo and Ecker, 2004). Recent studies have shown that ethylene enhances the stability of EIN3 and EIL1 proteins (Guo and Ecker, 2003; Potuschak et al., 2003), whereas sugar reduces the stability of these two transcription factors (Yanagisawa et al., 2003). The concerted regulation of EIN3 and EIL1 by ethylene and sugar indicates that cross talk exists between the two signaling pathways. Remarkably, we have found that the transcription of EIN3, EIL1, and CTR1 is also downregulated by glucose (Figure 8A). Consistent with our findings, reduced transcription of CTR1, a mitogen-activated kinase kinase kinase upstream of ethylene signaling transcriptional cascade, was observed in seedlings treated with 7% glucose for 3 h (Arroyo et al., 2003). Our data show that genes associated with ethylene biosynthesis are also transcriptionally repressed by glucose and that repression of three of these genes occurred in the presence of CHX. This raises the possibility that the cross talk between the glucose and ethylene signal transduction pathways may occur through the sugar-mediated transcriptional control of ethylene biosynthetic genes. Earlier findings are consistent with this possibility. Wild-type seedlings were developmentally repressed when grown on MS plates containing 6% glucose, but when seedlings were supplied with ethylene precursor 1-aminocyclopropane-1-carboxylate (ACC) in addition to the MS and 6% glucose, the glucose repression was relieved (Zhou et al., 1998; León and Sheen, 2003). Presumably, glucose repression of ACC oxidase (247774_at), ACC synthase (253999_at), and 2-oxoglutarate-dependent dioxygenase (246843_at) may reduce effective endogenous ethylene levels. Because ethylene decreases the sensitivity of seedlings to ABA (Beaudoin et al., 2000; Ghassemian et al., 2000; Gazzarrini and McCourt, 2001) and ABA represses germination and seedling development (Price et al., 2003), a decrease in ethylene caused by glucose may be the key mechanism by which glucose signaling interacts with ABA/ethylene signaling. Further experiments would be needed to confirm this premise.

By contrast, expression of ABA-related genes previously associated with glucose responsiveness was not altered by the conditions used in our study. There are several possible explanations for this observation. Previous experiments have shown that mRNA expression of ABA2 and ABI4 does not increase in the presence of 167 mM glucose until germination has occurred (Price et al., 2003). This suggests that the developmental program of the plant can potentially override the effect of sugar for these genes. Also, expression changes in ABA2 and ABI4 may constitute an indirect response and thus may not be evident upon a 3-h sugar exposure. In a time-course experiment using 7% glucose, ABI4 induction in seedlings was shown to begin primarily after 6 h of glucose exposure (Arroyo et al., 2003). A third possibility is that glucose regulation of ABA-related genes may not initially occur at the transcriptional level.

Is Glucose-Induced Stress Response a Physiological Process?

We were also intrigued by how glucose affected transcription of genes associated with stress. Previous studies have demonstrated that high concentration of exogenous glucose stunts the growth of young seedlings (Zhou et al., 1998; Gibson, 2000)—our data raise the possibility that glucose causes a stress response. The actual cellular concentration of glucose resulting from the treatment (167 mM exogenous glucose) is likely to be higher than that typically seen in vegetative plant tissue (Borisjuk et al., 2002)—perhaps there's a threshold where once a certain concentration is reached, a stress response is triggered. The stress response glucose elicits cannot be solely attributed to an osmotic event because 167 mM 3-OMG did not activate the same stress responsive genes regulated by glucose (Figure 8). The glucose stress response is also distinct from a heat stress response because light has been shown to be essential to observe a phenotypic change with heated plants (Larkindale and Knight, 2002), whereas glucose causes transcriptional and phenotypic changes without light being present (Jang et al., 1997). In any event, a glucose-induced stress response may provide an additional link for the cross talk between sugar signaling and ABA and ethylene signaling. It is known that ABA regulates plant responses when imposed with environmental stresses (Zeevaart and Creelman, 1988). However, additional experiments are needed before a linkage between ABA- and ethylene-signaling events and the glucose stress response seen here can be confirmed. Preliminary results from a microarray study indicate that some of the stress responses seen with glucose treatment are not replicated in seedlings treated with exogenous ABA (J. Price and J.-C. Jang, unpublished results). This is consistent with an earlier study showing that exogenous glucose treatment causes different signaling events than exogenous ABA treatment during germination (Price et al., 2003). However, these results don't rule out a linkage between glucose-induced stress response and ABA. In fact, it is well established that sugar can trigger changes in ABA biosynthesis and signaling (Cheng et al., 2002; Rolland et al., 2002; León and Sheen, 2003); thus, many stress responsive genes are likely coregulated by glucose and ABA. Even during germination, the response of germinating seeds to glucose has been shown to be affected by a block in ABA biosynthetic genes (Price et al., 2003). One possible strategy for dissecting the connections between ABA-related genes and a glucose stress response may be to conduct transcriptional analysis using plants having an ABA deficiency mutation such as aba2.

Multiple Sugar Signal Transduction Pathways Revealed by Transcriptional Control of Sugar Transporters

Glucose treatment also resulted in differential expression of sugar transporters. Such control is typical in budding yeast, where several hexose transporters are transcriptionally regulated by multiple glucose signaling pathways (Ozcan and Johnston, 1999). In yeast, some hexose transporter-like genes actually function as signaling receptors rather than actual transporters (Ozcan and Johnston, 1999; Rolland et al., 2001); it's conceivable that some putative sugar transporter genes in plants may have similar signaling functions (Lalonde et al., 1999). In yeast, at least 16 of 48 carbohydrate transporter-like genes have demonstrated transport function (Ozcan and Johnston, 1999; Mewes et al., 2002). Among the rest, Snf3 and Rgt2 have been identified as sugar sensors that can bind to glucose but are unable to transport glucose. Upon binding to glucose, the cytosolic C-terminal portions of Snf3 and Rgt2 interact with downstream signaling components, initiating a signaling cascade and ultimately causing the activation of hexose transporters. This glucose mediated transcriptional regulation controls sugar uptake in yeast (Ozcan et al., 1996, 1998; Ozcan and Johnston, 1999; Ozcan, 2002). In plants, at least 59 sugar and monosaccaride transporters have been putatively identified (Rolland et al., 2001; Mewes et al., 2002), of which 13 monosaccaride transporters and one sucrose transporter are regulated at least twofold under our conditions. Some of the plant sugar transporter-like genes are probably involved in the complex cellular functions, including the maintenance of a balanced source and sink relationship and the regulation of turgor in guard cells (Lalonde et al., 1999; Smeekens, 2000; Coruzzi and Bush, 2001; Coruzzi and Zhou, 2001; Truernit, 2001; Stadler et al., 2003). Whereas sugar transporter-like genes in plants might play more diverse roles than yeast counterparts, the possibility remains that some plant sugar transporter-like genes can act as sugar sensors resembling Snf3 or Rgt2. Although not regulated by glucose under our conditions, AtSut2 (or Suc3) has an extended intracellular domain structurally similar to Snf3 and Rgt2 (Barker et al., 2000).

In summary, our analysis revealed that glucose affected a broad range of genes not previously identified through traditional methods. Besides serving as a critical signal in assessing the general metabolic status, glucose elicits a broad stress response and significantly changes many regulatory genes, including numerous transcription factors. Under the conditions used, nitrogen appeared to have a relatively limited effect on transcriptional patterns, primarily altering expression of genes associated with nitrate assimilation. Much of the regulation for nitrogen utilization appears to be dependent on the availability of reduced carbon, for glucose was much more effective in regulating organic nitrogen metabolism. Nevertheless, nitrogen plays a critical role in modulating the effects of glucose on gene expression. This provides a molecular basis for the importance of carbon/nitrogen ratio in the control of plant growth and development. It is interesting to find that even though glucose causes repression or activation of a similar number of genes, glucose repression may be a somewhat more direct signaling event than glucose activation, which requires de novo protein synthesis. More analysis is needed to confirm the direct and indirect events caused by glucose and to identify cis-regulatory elements and trans-acting factors involved in the transcriptional activation or repression mechanisms. With the clues provided from GeneChip analyses, new avenues of inquiry may eventually dissect how metabolites like glucose and nitrogen regulate different aspects of the plant life cycle.

Preparation of Plant Material and RNA Extraction

A pilot study measuring expression of select genes was conducted to determine suitable conditions for monitoring transcriptional changes caused by sugar and/or nitrogen treatment. Because we use whole plants in our experiments and a recent study demonstrated that transcriptional changes upon nitrogen treatment occur primarily in the root (Wang et al., 2003), the mass ratio of shoot:root was measured in 10 replicates of 10 shoots or roots each to determine whether either tissue is overrepresented under the test conditions used in this study. Exogenous sugar can delay the onset of germination; so to avoid problems interpreting results from plant material at different developmental stages, the initial growth conditions were standardized to allow all plants to be at the early seedling stage. Arabidopsis thaliana seeds (ecotype Columbia-0) were surface sterilized and water imbibed in the dark for 3 d at 4°C. Seed pools were transferred to 1× MS basal salt mixture (GIBCO, Invitrogen) with B5 vitamins, 0.05% Mes, pH 5.7, and 58.4 mM sucrose. The plant material was incubated in the dark at 4°C for 3 d to break dormancy and then was transferred to light at 24°C for 5 d. Cultures were shaken at 140 rpm using an orbital platform shaker (New Brunswick Scientific, Edison, NJ) under continuous white light (100 μE m⁻² s⁻¹).

Once the plant material was uniformly germinated, the experimental conditions were applied. Germinated seedlings were washed seven times with sugar- and nitrogen-free MS to remove residual exogenous sugar or nitrogen, and the plant material was kept in the dark for all subsequent steps. The seedlings were placed in 1× MS salt mixture with B5 vitamins and 0.05% Mes, pH 5.7, but without sucrose or NH₄NO₃ and replacing KCl for KNO₃. Cultures were shaken at 140 rpm at 24°C for 24 h and then either sterile water, NH₄NO₃ and KNO₃ solutions (final concentration 20 mM each), or sugar solution (glucose or 3-OMG; final concentration 167 mM) was added to the medium of randomly selected plant cultures. CHX-treated plants were prepared identically except that 100 μM CHX was applied to seedlings 1 h before the addition of sugar and/or nitrogen. The seedling pools were treated for 3 h shaking at 140 rpm, washed with sterile water, and flash frozen in liquid N₂. The pooled plant material was used for RNA extraction to minimize the effects of variation amongst individual plants. RNA was prepared from frozen tissue using the RNeasy kit (Qiagen, Valencia, CA) following the manufacturer's protocol. The RNA was quantified and tested for quality before it was used for subsequent analyses. Four biological replicates for experiments without CHX and two biological replicates for experiments with CHX were performed. The two biological replicates with CHX were conducted simultaneously with two replicates without CHX.

Labeling of RNA Probe and Hybridization to Arabidopsis GeneChip

Labeling and hybridization of RNA were conducted using standard Affymetrix protocols by the University of California, Irvine DNA MicroArray Facility. Briefly, ATH1 Arabidopsis GeneChips (Affymetrix, Santa Clara, CA) were used for measuring changes in gene expression levels. Total RNA was converted into cDNA, which was in turn used to synthesize biotinylated cRNA. The cRNA was fragmented into smaller pieces and then was hybridized to the GeneChips. After hybridization, the chips were automatically washed and stained with streptavidin phycoerythrin using a fluidics station. The chips were scanned by the GeneArray scanner by measuring light emitted at 570 nm when excited with 488-nm wavelength light. Data from the chips were compiled using MicroArray Suite 5.0 software.

Analysis of GeneChip Data

Data from the GeneChip experiments were analyzed using MicroArray Suite 5.0 and DataMining Tool software as well as Vizard/EPCLUST (Moseyko and Feldman, 2002), GeneCluster2 (Golub et al., 1999), Q (Storey and Tibshirani, 2003), and Cluster/TreeView (Eisen et al., 1998). The MicroArray Suite 5.0 signal was the basis for all subsequent analyses. Two or four independent biological replicates were conducted with or without CHX, respectively. Control versus control scatter plots were generated to assess if the data were linearly distributed. MicroArray Suite 5.0 was used to conduct scalar normalization of the data—because the data consistently appeared to have a linear distribution after normalization with MicroArray Suite 5.0 (see Supplemental Figure 1B online), we did not employ other normalization methods typically used for nonlinear data (Bolstad et al., 2003). To identify significantly regulated genes, the log₁₀ normalized signals were subjected to RCBD analysis using a cutoff of P ≤ 0.001. The estimated false positive rate was determined using the P value, whereas the estimated FDR was estimated using the software Q (Storey and Tibshirani, 2003). The averaged filtered data were then subjected to an additional filter, which selected genes with greater than a twofold or threefold change versus the control. Threefold change filtering was selected for most analyses. Filtering criteria were primarily selected to minimize the number of false positives and to reduce the number of housekeeping genes considered, many of which are regulated twofold to threefold. Admittedly, the stringent threefold filtering will create some false negatives, as is seen with AtHXK1 (Figure 3); however, relaxing the criterion to twofold will increase the likelihood of including false positives in the data. For nitrogen metabolic genes and CHX regulated genes, filtering was relaxed to twofold change to create a more comprehensive list of regulated genes. To eliminate background noise and reduce false positives, a stringent minimum expression level difference of 140 was set. Probe sets scored as present ranged from 49.7 to 65.4% of the total probe sets. Maximal background for all chips had a signal of 112, whereas maximal raw noise was 6.9 or less.

Confirmation of Transcription Levels with RNA Gel Blot Analysis

The same batches of total RNA used in the microarray experiments were used to conduct RNA gel blot analyses on select genes of interest. RNA gel blot analysis was performed with 5 μg of total RNA/lane using standard protocols (Xiao et al., 2000). The probes for the blots were generated with PCR in the presence of radiolabeled dAT³²P. For transcripts that were undetectable by RNA gel blot analysis, RT-PCR was conducted using the one-step RNA PCR kit following the manufacturer's protocol (Takara, Madison, WI). The primer pairs used and the accession numbers for both RNA probes and RT-PCR are as follows: transporter protein (At1g16390, 262730_at) forward 5′-AACCCACACGTTTCAAT-3′, reverse 5′-AAACCCAAATGCTCGTT-3′; sugar transporter (At3g05150, 259351_at) forward 5′-ATTTCATCGGTCGGAAA-3′, reverse 5′-TGCAACGCAATTTCA-3′; glycosyl hydrolase (At5g18670, 250007_at) forward 5′-AGACGGTGAGCTGAAA-3′, reverse 5′-TGCCTGAACCTATGCT-3′; ASN1 (At3g47340, 252415_at) forward 5′-TTGCTCACTTGTACGAG-3′, reverse 5′-ATTGCTTAGCCGCCTTA-3′; glutamate dehydrogenase (At5g07440, 250580_at) forward 5′-TGGGCACTAACGCTCA-3′, reverse 5′-CCAAGAGCGCATGGAA-3′; trehalose phosphate synthase 8 (At1g70290, 264339_at) forward 5′-CCCAAGCTTGCTAATATATAGT-3′, reverse 5′-CGGGATCCGACGCGTGGAAGAGTT-3′; bZIP (At5g49450, 248606_at) forward 5′-GGCAAACGCAGAGAA-3′, reverse 5′-AGGACGCCATTGGTTG-3′; tat binding (At1g10070, 264524_at) forward 5′-TCCCCGCGGTACATGTATACATATGCTTAG-3′, reverse 5′-CGGGATCCGTAATCAGCTGGATTTAG-3′; hexokinase 1 (At4g29130, 253705_at) forward 5′-ATGCACAACGACACTT-3′, reverse 5′-TCAGAACTCCAGTGAA-3′; hexokinase 2 (At2g19860, 266702_at), forward 5′-ATGGGTAAAGTGGCAGTTGCAA-3′, reverse 5′-AATTGAACAAAGTCTCAGTAGAAG-3′; expressed protein (At1g80130, 262050_at) forward 5′-GTGGGATGTGGATGAGG-3′, reverse 5′-CTCGATAGGACATGGGT-3′; hsp81-2 (At5g56030, 248045_at) forward 5′-TACGGCTGGACTGCAA-3′, reverse 5′-GAATCAGTCTCTTGAGC-3′; hsp(At-hsc70-3) (At3g09440, 258979_at) forward 5′-CGAGAAGCTTGCTGGAG-3′, reverse 5′-GCTCATCGAAACAAGCG-3′; membrane channel (At2g28900, 266225_at) forward 5′-TGGCAGTGGACATGGG-3′, reverse 5′-AAGCGCGCCACCAAGA-3′; starch synthase (At1g32900, 261191_at) forward 5′-GGCAACTGTGACTGCT-3′, reverse 5′-GCAGCCTGACACAACA-3′; trehalose phosphate synthase 5 (At4g17770, 245348_at) forward 5′-ATGCTCCTTCTTCCGT-3′, reverse 5′-ATCAGCGTTGAGGAGT-3′; NIA1 (At1g77760, 259681_at); SAG21/LEA (At4g02380, 255479_at), forward 5′-CGGGATCCATGGCTTTAAAACATATGCA-3′, reverse 5′-CCATCGATCGCAGCTGCCTTGATTCT-3′; UPM1, (At5g40850, 249325_at) forward 5′-TCCTCCAGTATTCGGA-3′, reverse 5′-CCCTCCTTTCCTTGAAT-3′; ABA2 (AT1g52340, 259669_at) forward 5′-AAAGTGGCATTGATCACT-3′, reverse 5′-TCCTAGTCAAGCCTAGA-3′; ABI4 (At2g40220, 263377_at) forward 5′-CACCGACTAATCAACTT-3′, reverse 5′-CATCTGGACCATCTGAT-3′.