| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
|
First published online November 17, 2004; 10.1105/tpc.104.025973 © 2004 American Society of Plant Biologists
A Robot-Based Platform to Measure Multiple Enzyme Activities in Arabidopsis Using a Set of Cycling Assays: Comparison of Changes of Enzyme Activities and Transcript Levels during Diurnal Cycles and in Prolonged Darkness
|
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
; Celis et al., 2000; Michaut et al., 2003), including Arabidopsis thaliana (Wang et al., 2003; Thimm et al., 2004; AtGenExpress, www.arabidopsis.org). The sensitivity and precision with which transcript levels can be quantified will be further increased by multiplexed real-time RT-PCR (Czechowski et al., 2004). Interpretation depends on assumptions about the relation between the levels of transcripts and proteins. This is difficult to predict because it will depend on how far the transcript concentration affects the rate of translation, how rapidly the protein turns over, and whether the rate of degradation is coordinately regulated. In some cases, the response may be attenuated, and in others posttranscriptional regulation may lead to exaggerated or independent changes. For example, nitrate reductase (NIA) protein increases twofold in leaves during the first part of the light period, whereas NIA transcript decreases rapidly (Scheible et al., 1997a). This occurs because illumination stimulates the synthesis and inhibits the breakdown of NIA protein (Kaiser and Huber, 1999; Weiner and Kaiser, 1999). The large number of genes involved in protein degradation in Arabidopsis (Arabidopsis Genome Initiative, 2000) underlines the potential importance of posttranscriptional regulation.
Quantification of proteins at the proteome scale poses a major technical challenge (Greenbaum et al., 2003). Although methods are being developed (Steen and Pandey, 2002), they lag behind the techniques for expression profiling and require costly and extensive infrastructure. Enzyme activity assays require inexpensive equipment and provide quantitative data. Measurements in optimized assays that allow maximum activity (ap Rees and Hill, 1994) could act as a proxy for quantitative proteomics, at least for studies of metabolism. Throughput can be increased by transferring existing protocols to microplates and using robots (Ashour et al., 1987; Kiianitsa et al., 2003). High throughput assays of individual enzymes are used routinely in drug or herbicide discovery (Lein et al., 2004). The uncommonness of each reaction and resulting diversity of assay systems will make it logistically impossible to measure an entire enzymome, but it may be possible to create efficient platforms for subsets of enzymes by grouping them in modules that share a common detection method.
Such platforms could provide quantitative data that provide insights into the relation between transcript levels and protein levels in genomics studies. Activity measurements can also provide information about posttranslational regulation in cases where it alters the kinetic properties (Kaiser and Huber, 1994, 1999). Another potential application is in the diagnosis of genotypes or physiological states. It can also be approached via machine learning, in which a training data set from a series of defined conditions is used to generate a classification key that is used as a diagnostic tool to classify unknown samples (Quinlan, 1979; Kononenko, 2001). Machine learning can be used on data sets with a relatively small number of parameters but requires good replication and reliable quantification.
This article presents a set of robotized microplate-based activity assays for 23 enzymes from central carbon and nitrogen metabolism. In addition to allowing high-throughput analyses, most of the assays are highly sensitive, allowing very low activities to be detected or very small samples to be analyzed. The platform has been used to investigate the changes of enzyme activities during the diurnal cycle in wild-type Arabidopsis and a starchless mutant where carbohydrate storage is disturbed and during a prolonged dark treatment of wild-type Arabidopsis. These treatments allow us to ask if rapid changes in transcript levels during the diurnal rhythm lead to changes in the activity of the encoded enzyme and if larger changes of transcript levels in the starchless pgm mutant or more prolonged changes in prolonged starvation lead to more marked changes in enzyme activity. Methods are presented that integrate information about transcript levels and enzyme activity, provide insights into the quantitative relation between changes of transcript levels and enzyme activities, and highlight cases where translational regulation or protein turnover may contribute. The data sets are also used to illustrate the utility of enzyme activities for diagnosis using machine learning.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
In most cases, cycling assays (Gibon et al., 2002) were used to determine the products of the enzyme reactions. The principle is that the net rate of the cycle is a pseudo zero-order reaction whose rate (d[analyte measured]/dt = –d[precursor of this analyte]/dt) depends on the initial concentration of the metabolite being determined. Quantification is achieved by comparison with a standard curve, in which different concentrations of the metabolite in question are added in the presence of pseudo-extract (Gibon et al., 2002). A combination of stopped and cycling assays can provide 10,000-fold higher sensitivity than conventional assays. The sensitivity limit for routine measurements in leaf extracts was 0.01 to 1 nmol product per assay. This allows extracts to be diluted far more strongly in the assay, which increases the reliability of stopped assays. Interference by other components in the extract is reduced, substrates can be maintained at near-constant levels throughout the assay, product inhibition is minimized because the amounts accumulated are very low (typically below 20 µM), and interference by other enzymes is decreased.
The platform uses three cycling systems (Table 1, Figure 1). Glycerol-3-P and dihydroxyacetone-P are measured by a glycerol-3-P dehydrogenase/glycerol-3-P oxidase cycle (Gibon et al., 2002), NADP+ and NADPH by a glucose-6-P dehydrogenase/phenazine methosulfate (G6PDH/PMS) cycle (Nisselbaum and Green, 1969), and NAD+ and NADH by an alcohol dehydrogenase/phenazine ethosulfate (ADH/PES) cycle (Bernofsky and Swan, 1973). Reactions were stopped using HCl, except when the direct or indirect product was a reduced coenzyme, in which case NaOH was added. Oxidized NAD(P)+ is stable at low pH and labile at high pH, and reduced NAD(P)H is labile at low pH and stable at high pH (Nisselbaum and Green, 1969). This combination allows a wide range of enzyme activities to be measured, including all NAD(P)-dependent dehydrogenases and all activities that generate ATP or UTP because the latter can be coupled to glycerol-3-P formation. In some cases, one of these metabolites is the direct product of the enzyme reaction. Otherwise, coupling enzymes are included to convert a product into one of these metabolites. The modular organization allows many different enzyme assays to be fed into a pipeline with just three downstream cycling systems, which simplifies logistics and increases throughput.
|
|
) was modified by measuring glucose via a glucose oxidase/peroxidase coupling system that oxidizes dihydroxyphenoxazin into the fluorescent dye resorufin. An analogous approach could be used for many other enzymes because a range of oxidases and fluorigenic substrates are commercially available.
A small number of enzymes required dedicated assays. We have not yet found a procedure that combines high sensitivity and high throughput for enzymes that produce ADP from ATP (e.g., Gln synthetase; GS). GS activity was high enough to use an existing assay, in which ADP is measured by coupling it to NADH formation via pyruvate kinase (PK) and lactate dehydrogenase (Scheible et al., 1997a). DTT is often required to stabilize enzymes like NIA but interferes with the G6PDH/PMS and ADH/PES cycling systems because it reduces methylthiazolyldiphenyl-tetrazolium bromide (MTT) and the peroxidase-dependent oxidation of dihydroxyphenoxazin. NIA had to be assayed using an existing protocol that measures nitrite formation (Scheible et al., 1997b, 1997c). The assay for ferredoxin-dependent glutamate synthase (Fd-GOGAT) contains a mix of reductants (DTT to stabilize Fd-GOGAT and dithionite to drive the reaction). Furthermore, the product (glutamate) has to be measured in the presence of a vast excess of 2-oxoglutarate and Gln. To achieve this, glutamate dehydrogenase (GDH) was coupled to diaphorase and MTT to displace the equilibrium toward glutamate oxidation. Before determining glutamate, an excess of N-ethylmaleimide was included to remove DTT and dithionite because they would otherwise reduce MTT to its formazan. Fortunately, GDH was not inhibited by N-ethylmaleimide.
All of the standard assays contained optimized levels of substrates and, where necessary, activators, pH, and ionic conditions to allow maximum activity. In specific cases, it is also possible to use a second assay system with suboptimal levels of specific substrates or effectors to detect posttranslational regulation. As an example, posttranscriptional regulation of NIA was monitored by comparing activity in the presence of excess Mg2+ to stabilize a complex between the phosphorylated protein and the inhibitory 14-3-3 protein and selectively assay the active nonphosphorylated protein, and in the presence of excess EDTA to trap Mg2+ and allow the phosphorylated form to be released and regain activity to determine maximal activity (Kaiser and Huber, 1999).
Optimization and Validation
This section describes the optimization and validation of the assay for fumarase. The same strategy was applied to all assays. Fumarase catalyzes the hydration of fumarate to malate. Malate dehydrogenase is included as a coupling enzyme to oxidize malate to oxaloacetate and simultaneously reduce NAD+ to NADH. Because oxaloacetate exerts strong product inhibition on malate dehydrogenase, it was removed by including citrate synthase and an excess of acetyl-CoA. The assay is stopped with an excess of NaOH, the nonreduced NAD+ destroyed by heating at 95°C for 5 min, the extract neutralized with HCl, and NADH determined using the ADH/PES cycle system. Activity was measured by comparing the signals with the complete assay buffer (Vmax) and when a substrate (here fumarate) was omitted (Vblank).
The first step in assay validation was to optimize the amount of plant material included in the assay. This exerted a dramatic effect on the apparent activity of fumarase (Figure 2A): a 20,000-fold dilution of fresh weight (w/v) in the reaction mixture was needed to reach linearity. Activity will be severely underestimated in conventional assays, which measure the accumulation of NADH via the absorbance change at 340 nm or the decrease of fumarate via the absorbance change at 280 nm (Bergmeyer, 1987). To check the reliability of the assay procedure, different amounts of pure malate were added to the extract. Recoveries of 
90 and 100% were obtained in Vmax and Vblank conditions across a wide range of spiked malate concentrations (Figures 2B and 2C). The reaction was stopped after different times to check linearity with time. For all of the enzymes except Fd-GOGAT, the reaction was linear at least for 20 min, which was used for routine determinations (data not shown). Fd-GOGAT reaction was found to be linear for only 10 to 15 min, probably because atmospheric oxygen reoxidized the assay medium. Concentrations of substrates, ions, and pH were chosen to ensure maximal activity was detected. In cases where activators were known from the literature (not appropriate for fumarase), they were included and optimized. Stability of activity through a freeze-thawing cycle was checked by snap-freezing samples in liquid nitrogen and storing them for at least 24 h at –80°C before rethawing to assay activity. In the case of fumarase, <10% of the initial activity was recovered (Table 1). This study compares enzyme activities in comparable tissues. In cases where activities are compared between different tissues, it is strongly advised to conduct recombination experiments in which powder from the two tissues is mixed in a 50:50 ratio, extracted, and assayed to check that the activity in the mixed extract is similar to the arithmetic mean of the activities in extracts from the two tissues.
|
Sample Flow
Handling procedures and assay protocols were designed to optimize sample flow, and spreadsheets were used to allow rapid and automatic processing of the data (see Methods). Assay of one enzyme in 20 samples requires 2 h using conventional hand pipettes, including extraction, assay, and calculations, allowing one person to measure three to four different activities in 20 samples in a single working day. Throughput is increased fivefold by a four-tip pipetting robot, allowing, for example, three to four activities to be measured in 100 samples per day. We are presently transferring the platform to a 96-channel robot and expect a further twofold to 10-fold increase in throughput. The major bottleneck is the grinding and weighing of aliquots before extraction. This manual step may be difficult to automate because the sample must be kept deeply frozen to avoid loss of activity and be very finely homogenized to allow full extraction. Tissue-permeabilizing extraction media like those that are used to extract RNA and metabolites are unsuitable for enzymes because they denature proteins.
Global Comparison of Diurnal Changes of Transcript Levels and Enzyme Activities during a Diurnal Cycle in Wild-Type Arabidopsis
In a first application, 23 enzyme activities were determined in five replicate samples collected at 2-h intervals throughout a diurnal cycle in Col-0 wild-type Arabidopsis rosettes. In parallel, RNA was prepared every 4 h from three biological replicates and used for expression profiling with Affymetrix ATH1 chips. The TranscriptScavenger module from MapMan (Thimm et al., 2004) was used to identify and extract the data for genes that are members of the families for the 23 enzymes in our platform (in total 84 genes; see supplemental data online for a list). The averaged values for the diurnal changes of enzyme activities and the transcript levels for each member of the corresponding gene family are shown for NIA, fumarase, Fd-GOGAT, ADP glucose pyrophosphorylase (AGPase), and GLDH in Figure 3 and for the other 18 enzymes in the supplemental data online. The diurnal cycle in wild-type Col-0 is shown in the second panel from the left. The results will first be analyzed with respect to the overall relation between transcript levels and enzyme activities and later with respect to the response of specific enzymes.
|
The data were statistically analyzed to uncover general features of the relation between transcript levels and enzyme activities. A formula given in Methods was used to estimate a smoothness value, which has an optimal value of zero if every data point lies on a smooth curve that moves through one maximum and one minimum per diurnal cycle, rising to one as the data points become increasingly irregular (see supplemental data online for examples of the values obtained for a set of randomly generated data points). Values between 0 and 0.05 were arbitrarily chosen as indicating a good oscillation for transcripts and enzyme activities, values ranging from 0.05 to 0.1 as indicating a possible oscillation, and values >0.1 as indicating no oscillation of transcripts and enzyme. Figure 4A indicates that almost all transcripts but only 30% of the enzyme activities show a smooth oscillation (Figure 4A). A further 43% of the enzymes show a slightly irregular oscillation. The amplitude of the change is smaller for enzyme activities, making them more susceptible to experimental noise.
|
These data indicate that whereas most transcript levels exhibit a clear and marked diurnal rhythm, the response of the corresponding activities was damped, and in some cases hardly detectable. The amplitude of the diurnal changes of transcripts may still be an underestimate for low-expressed genes. It has been inferred that hybridization kinetics on expression arrays are second order (Holland, 2002), and the linear range afforded by the Affymetrix ATH1 arrays (three to four orders of magnitude) may lead to some attenuation of the signals. Comparison of the signals for >1400 transcription factors measured by real-time PCR and the ATH1 array in Arabidopsis shoot and root samples (Czechowski et al., 2004) and in Arabidopsis seedlings in differing nitrogen nutrition (Scheible et al., 2004) reveals that the ATH1 signals in the low signal range are attenuated by a factor of three. Transcription factors are low-expressed genes, and a better quantitative agreement might be expected for genes whose transcripts are present at higher levels. A well-correlated fit line with a slope close to one was found between data obtained from ATH1 probe arrays (expressed as Log2 ratios) and real-time PCR (expressed as average cycle time differences) for a small set of more strongly expressed genes (Redman et al., 2004).
Global Comparison of Diurnal Changes of Transcript Levels and Enzyme Activities during a Diurnal Cycle in the Starchless pgm Mutant
In the light, photosynthesis leads to synthesis of sucrose that is exported to support respiration and growth in the remainder of the plant. At night, the entire plant becomes a net consumer of fixed carbon. Some photosynthate is stored in leaves as starch in the light and is remobilized at night to support leaf respiration and the continued synthesis and export of sucrose. This provides a buffer against the daily alternation of the carbon budget. The starchless pgm mutant lacks plastidic phosphoglucomutase, which is essential for starch synthesis in the leaf (Caspar et al., 1985). Sugars accumulate during the day but are rapidly depleted in the first part of the night. A recurring phase of sugar starvation during the second part of the night leads to a severe growth impairment (Gibon et al., 2004). The transcriptome of pgm at the end of the day resembles wild-type Arabidopsis, whereas the transcriptome at the end of the night resembles wild-type plants after a 6-h extension of the night (Gibon et al., 2004; Thimm et al., 2004). Diurnal changes of enzyme activities were investigated in pgm because it was expected that they would be larger than in wild-type plants. It is already known that acid invertase (Caspar et al., 1985) and SPS and AGPase activities (Caspar et al., 1985) are higher in pgm plants.
For pgm, enzyme activities were measured in five replicates. Transcript profiles were only repeated at the end of the night and at the end of the day. The reproducibility found for these data points was very good (data not shown). For transcript levels, the smoothness factor moved even closer to the ideal value of zero (Figure 5A) than in wild-type plants. This trend might be explained because the larger diurnal changes make the data less susceptible to random noise. By contrast, smoothness increased only slightly for a subset of the enzyme activities (Figure 5B). There was a clear increase in the amplitude of the diurnal changes of transcript levels (the average value rose from 35% in wild-type plants to almost 60% in pgm; Figure 5C), but the amplitude of the diurnal changes of enzyme activities was not altered (Figure 5D).
|
|
|
Figure 7I shows the results for selected enzymes. In many cases, an average regression coefficient >0.8 was obtained when changes of transcripts were plotted against the changes of enzyme activities that occurred at a later time point. Different enzymes respond with a different time delay. For example, a set of key enzymes in nitrogen metabolism (NIA, GS, Fd-GOGAT, AlaAT, and GLDH) show a lag of 12 h or longer between rapid changes in transcript levels and the corresponding changes of enzyme activity. NADP-isocitrate dehydrogenase responds similarly, whereas fumarase, pyrophosphate-dependent phosphofructokinase (PFP), and especially fructokinase respond with a shorter delay, and NADP-GAPDH shows an extremely long delay. It has been noted that AGPase shows the largest diurnal changes of all the enzymes studies. The changes are strongly delayed compared with the change of transcript levels and the regression coefficients rather poor. In most cases, a similar trend is found in wild-type plants and in pgm. Invertase showed a strikingly different response in wild-type plants and pgm. This might be because only one specific member of the invertase gene family is strongly induced in pgm (see supplemental data online).
Comparison of Time Courses of Transcripts and Activities for a Set of 23 Enzymes in a Transition to an Extended Night in Arabidopsis Col-0
The relatively short time between the maximum and minimum of transcript levels during a diurnal cycle may prevent the enzyme activity from fully adjusting to the change in transcript level. Prolongation of the night leads within hours to dramatic changes of many transcripts for enzymes involved in central metabolism (Thimm et al., 2004). This system was used to investigate the impact of prolonged changes of transcript levels on enzyme activities.
Samples were taken 2, 4, 6, 8, 24, and 48 h into the extended night to measure transcript levels and 2, 4, 8, 12, 24, 48, 72, and 148 h into the extended night to measure enzyme activities. The results were expressed relative to the value at the end of the normal night, converted to a log2 scale to linearize the distribution and used to calculate the probability mass for each time point (Figures 8A to 8G). There were detectable changes of some transcript levels by 2 h (Figure 8A) and marked changes of most by 4 h (Figure 8B). For most enzymes, there were no large changes of activity until 48 to 72 h into the extended night (Figures 8E and 8F). The only exception was NIA, which showed a rapid decrease of activity. For every enzyme activity, a P value was calculated to reveal whether the activity at each time in the prolonged night was significantly different from the activity at the end of the normal night. Figure 8G summarizes the average and SE of the P values (n = 23 enzyme) at each time point and reveals that enzyme activities, considered as a whole, did not change significantly until the plants had been subjected to 2 d of carbon starvation.
|
NIA is subject to sophisticated transcriptional, posttranscriptional, and posttranslational regulation in tobacco (Nicotiana tabacum) leaves (see Introduction). Our results (Figure 3A) reveal a similarly complex network in Arabidopsis. Of the two genes that encode NIA in Arabidopsis, one is highly (At1g37130) and the other weakly (At1g77760) expressed in rosettes. Wild-type plants show large diurnal changes of transcripts and smaller changes of activity (amplitudes = 80 and 88% for NIA1 and NIA2 transcripts and 36% for NIA activity), with transcripts peaking at the end of the night and enzyme activity 6 h later in the middle of the light period. Compared with wild-type plants, pgm showed a slight reduction in transcript levels and a 1.5-fold to twofold reduction of activity. The diurnal changes of transcripts were less marked, but the diurnal changes of activity remained large. Prolongation of the night led to a rapid and marked decrease of NIA transcript and activity. NIA protein also decreases rapidly in response to carbon starvation in tobacco leaves (Klein et al., 2000). NIA is subject to reversible posttranslational regulation, in which phosphorylation of Ser-534 leads to binding of a 14-3-3 protein that inhibits activity and may also increase degradation of NIA protein (Kaiser and Huber, 1999; Meyer and Stitt, 2001). Posttranslational regulation was monitored using a selective assay. In wild-type plants, overall NIA activity and NIA activation are low during the night, increase rapidly after illumination, and decrease as the light period progresses. In pgm, overall activity and activation increased slowly at the start of the light period but remained high until the end of the day.
AGPase (Figure 3B) is a heterotetramer consisting of two large (AGPS) and two small (AGPB) subunits. AGPB is encoded by one gene (At5g48300; Hendriks et al., 2003) and AGPS by a four-member family, of which one is highly (At5g19220), two are weakly (At4g39210 and At1g27680), and one is very weakly (At2g21590) expressed in rosettes. Transcripts for AGPB and the major AGPS form showed weak diurnal changes, with peaks during and at the end of the light period, respectively. These results were confirmed using real-time RT-PCR (data not shown). AGPase activity was assayed in the presence of excess levels of the allosteric activator 3-phosphoglycerate and DTT to completely convert the enzyme into the active reduced form (Tiessen et al., 2002; Hendriks et al., 2003) and to ensure that the activity measurements reflected changes in AGPase protein. AGPase activity fell rapidly during the first part of the day and recovered in the last part of the light period and during the night. AGPase was the only enzyme where the changes of activity (51%) were larger than the changes in transcript levels (12 and 13% for AGPB and the major AGPS transcript). These were the largest diurnal changes of activity in the entire set of enzymes. Total AGPase activity decreased in the first part of the light period, when AGPB and AGPS transcripts were stable or rising. A similar picture emerged during diurnal changes in pgm, where the changes were even more pronounced (amplitudes of 69% for activity, 45 and 38% for transcripts for AGPB and the major AGPS family member, respectively). When wild-type plants were exposed to a prolonged night, transcripts decreased but activity stayed high.
Protein gel blots were performed to check that the changes of AGPase activity reflect changes of protein. The gels were run in the absence of DTT so that the reduced (monomer) and oxidized (dimer) forms (Figure 9A) are visible (Hendriks et al., 2003). An increase of the ratio between reduced and oxidized forms has been associated with the activation of AGPase (Tiessen et al., 2002; Hendriks et al., 2003), and the strong monomerization found in pgm has been discussed elsewhere (Hendriks et al., 2003). Total AGPB protein was measured by summing the signals from both bands. The diurnal changes of AGPB protein agree well with the changes of total AGPase activity (cf. Figures 3B and 9B). This figure also illustrates that by using suitable extraction and assay conditions, it is indeed possible to override posttranslational changes and obtain a measurement of activity that reflects total protein.
|
Fd-GOGAT (Figure 3D) is encoded by one major transcript (At5g04140). Wild-type plants had marked diurnal changes of the major transcript (35%), which were followed with an 12-h delay by damped changes of enzyme activity (18%). The average transcript and activity levels in pgm were 1.6- and 1.9-fold lower than in wild-type plants. The diurnal changes were slightly larger (51% for the transcript and 20% for activity). Prolongation of the night led to a rapid and marked 50% decrease of transcript, falling to 50% by 8 h, and 10% by 48 h, whereas activity remained unaltered for at least 7 d.
Inspection of the results for other enzymes reinforces the conclusion that changes of transcript levels are accompanied by much smaller changes of enzyme activities during the diurnal cycle and to strongly damped and slow responses in a prolonged night (AlaAT, cytosolic fructose-1,6-bisphosphatase, fructokinase, hexokinase, NAD- and NADP-GAPDH, isocitrate dehydrogenase, phosphoenolpyruvate carboxylase, PFP, and transketolase). In some cases, no relation was apparent (glycerokinase and shikimate dehydrogenase). These activities were, however, very low and more difficult to assay precisely.
Comparison of Wild-Type Arabidopsis and the pgm Mutant
Compared with Col-0, pgm has twofold lower NIA and Fd-GOGAT activity (see above), slightly lower fumarase, GS, isocitrate dehydrogenase, phosphoenol pyruvate carboxylase, PFP, and transketolase activity (see supplemental data online) and twofold higher acid invertase (Caspar et al., 1985) and GLDH activity (Figure 3E).
Transcripts for GDH1 (At5g18170) and GDH2 (At5g07440) underwent extreme diurnal changes in pgm, with a 20-fold increase toward the end of the night (Figure 7E). This contrasts with wild-type plants, where there were only small diurnal changes of transcripts. The average level of the GDH1 and GDH2 transcripts across the entire diurnal cycle was 6.7- and 3.2-fold higher, respectively, in pgm. Activity in pgm showed only a slight diurnal oscillation but was 1.8-fold higher than in wild-type plants. This indicates that recurring transient peak of GLDH transcript at the end of each night leads to a quasi-stable increase of GLDH activity throughout the entire diurnal cycle. GDH transcript levels increased dramatically during the first hour of a prolonged night before relaxing to levels that were approximately fourfold higher than those found for a normal night in wild-type rosettes. The sustained increase of GDH1 and GDH2 transcript levels in a prolonged night leads to increased GLDH activity, but the response was damped by a factor of two to three and only developed after a considerable time lag. These results indicate that the increase of the GDH transcripts in the later part of the night in pgm is a response to low sugar.
A similar response was found for acid invertase (see supplemental data online). The major transcript (At1g12240) showed much larger diurnal changes (10-fold) than in wild-type plants, falling to lower levels at the end of the day and rising to threefold higher levels at the end of the night. The 10-fold diurnal changes of transcript in pgm were not accompanied by diurnal changes of activity. However, invertase activity increased 1.7-fold, compared with wild-type plants. Transcript increased threefold at the start of a prolonged night, before relaxing to a sustained twofold higher level. This was followed over several days by a slow twofold increase of activity. In both treatments, changes of At1g12240 transcript do lead to marked changes of invertase activity but only after a considerable time lag.
The Use of Enzyme Activities as Diagnostic Markers
In machine learning, a training data set obtained in a set of defined conditions is used to generate a classification model, which can be used as a diagnostic tool to automatically classify further unknown samples. Their relative stability indicated that enzyme activities might provide a suitable data set for machine learning. A data set made of two experiments for wild-type plants, including 60 samples collected at 12 different times during the normal diurnal cycle and 55 samples collected at different times during a prolonged night were used to construct a decision tree. The decision tree separated the wild-type samples into six classes corresponding to samples harvested at any time during a normal diurnal cycle, samples harvested during the first 12 h of an extended night, and samples harvested 1, 2, 3, and 7 d into a prolonged night (Figure 10A). Four (NIA, AGPase, NAD, and NADP-GAPDH) of the 23 enzymes were excluded because their activity underwent very large changes in the normal diurnal cycle, which made the decision tree unstable (data not shown). It used information with approximately nine (GLDH, PK, shikimate dehydrogenase, G6PDH, GS, glycerokinase, fructose-1,6-bisphosphatase, fructokinase, and invertase) of the remaining 19 enzymes.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
Several of the assays provide large advantages over previous approaches. Glycerokinase and shikimate dehydrogenase, which have rarely been assayed in plants, can now be routinely determined in crude extracts. The development of an assay that allows glutamate to be measured via GLDH allows hundreds of measurements of Fd-GOGAT within a day, compared with previous assays that used HPLC (Suzuki et al., 1994) or ninhydrin (Hecht et al., 1988) to measure glutamate. The new assay for AGPase is a large advance for measurements in photosynthetic tissues, where previously a cumbersome radioactive test was needed. Although a conventional coupled assay exists in which AGPase is assayed by coupling glucose-1-P formation via phosphoglucose isomerase and G6PDH to NADP+ reduction, this assay is unusable in leaves. The other product is ATP, and to obtain maximum activity, the assay must include the allosteric activator glycerate-3-P. Leaf extracts contain high activities of phosphoglycerate kinase and NADP-GAPDH, which convert ATP and glycerate-3-P to triose-phosphate and reoxidize NADPH. In the new assay, ATP reacts with glycerol in the presence of an excess of glycerokinase, and glycerate-3-P can be included in the assay without danger of interference. Previous assays for SPS also had major drawbacks. The widely used anthrone test (Huber et al., 1989) suffers from a low sensitivity and requires a desalting step and the use of concentrated sulphuric acid. In our lab, 40 extracts could be analyzed per day with this method. An alternative assay measures UDP via PK and lactate dehydrogenase functions (Stitt et al., 1988). However, PK has a relatively poor affinity for UDP (http://www.brenda.uni-koeln.de), which makes this assay unsuitable for many plant tissues because they contain high UDPase activity. The new assay detects UDP via UMP-kinase, glycerokinase, and a cycling assay and allows efficient recovery of UDP, especially when it is combined with high (1000-fold) dilution of the extract. It also allows >10-fold higher throughput than the anthrone assay.
Relation between Changes of Transcripts and Changes of Enzyme Activities
We have used measurements of enzyme activities in optimized conditions as a proxy for quantitative determination of proteins. The assumption that the activity in the assay reflects the amount of enzyme protein was checked by optimizing the assays with respect to all available information about the kinetic properties of the enzymes. It remains possible that some enzymes have unknown effectors or are subject to posttranslational regulation that affects their Vmax rather than kinetic properties and is not overridden by including high levels of substrates and effectors and optimizing the pH and ionic conditions. However, this would affect only the results for individual enzymes and not the general conclusions. A second limitation is that the enzyme assays do not distinguish among isoforms. For this reason, we have been cautious in using data for five enzymes, where the gene family members showed contrasting changes of their transcript levels.
The results from our study show that quantification of enzyme activities or proteins will be crucial in system biology. First, changes in transcript levels do not lead to quantitatively similar changes of enzyme activity. The diurnal changes of transcript levels were usually larger than the changes of the corresponding enzyme activities, the increased amplitude of the diurnal changes of transcripts in pgm did not lead to increased diurnal changes of enzymes, and the large and sustained changes of transcript levels in a prolonged night did not lead to correspondingly large changes in enzyme activities. In general, our results indicate that the response of enzyme activities is often damped by a factor of two or more. Second, enzyme activities typically change with a time delay. This is clearly seen in diurnal cycle in Col-0 and pgm, where the changes of activity were delayed compared with the change of transcript levels in a manner that varied from enzyme to enzyme. Most enzyme activities were also maintained for a very long time in an extended night, even though transcript levels decreased rapidly. In general, enzymes turn over more slowly than transcripts, but the relation varies on an enzyme-to-enzyme basis. Third, whereas a global analysis may provide rough guidelines about what constitutes a typical response, in some cases enzymes show very particular responses. Three examples are the large diurnal changes in AGPase activity and to a lesser extent NIA activity relative to the changes of transcript, the striking and rapid decrease of NIA activity during a prolonged night, and the unusually close correlation between the momentary changes of transcript levels and activity for fructokinase.
In cases where enzyme activity showed a large change relative to that of the encoding transcripts, closer examination of the data and comparison with the literature indicates that the change is at least partly because of posttranscriptional regulation. For example, the marked changes of AGPase activity in the diurnal cycle were accompanied by reciprocal changes of transcripts for the two subunits. Even more strikingly, whereas these transcripts rise and AGPase activity falls after illumination at the start of the day, AGPase activity stays high and the transcript levels fall if the night is prolonged. This indicates that the decrease in the first part of the normal day is because of posttranscriptional events that are triggered by a signal that is generated in the light. Jeannette and Prioul (1994) reported a similar diurnal activity change in maize (Zea mays) leaves. Further studies are needed to elucidate the biochemical mechanisms and understand their physiological function.
Comparison with published studies in tobacco indicates that the marked diurnal changes of NIA activity are also partly because of posttranscriptional regulation. Tobacco exhibits similar diurnal changes of NIA transcript and activity (Scheible et al., 1997b, 2000). In this species, overexpression of NIA under the control of a constitutive promoter abolished the diurnal changes of NIA transcript but did not alter the diurnal changes of NIA activity (for references, see Meyer and Stitt, 2001). Weiner and Kaiser (1999) have shown that synthesis of NIA protein is inhibited in the dark and stimulated in the light, whereas degradation is effected in a reciprocal manner.
The slow and damped response of enzyme activities has important implications for the analysis of multilevel genomics data sets. There is, for example, obvious interest in combining data sets from expression arrays and metabolite profiling. In the conditions we have studied, any correlations found among the levels of individual or groups of transcripts and metabolites will provide information about the role of metabolites as regulators of transcription. They cannot be because of the impact of transcription and the resulting changes of enzyme activities on metabolism because enzyme activities change too slowly and are too strongly damped to exert an effect. In other sets of biological samples, the changes in enzyme activities may be more marked, for example, when samples are compared in which a long time period is allowed for the system to stabilize or when different genotypes are compared. However, it will clearly always be important to obtain quantitative data about the response at the enzyme activity/protein level to provide an empirical base for the interpretation of the direction in which causality can operate in comparison of expression profiles and downstream phenotypic changes.
Functional Consequences of Different Dynamics for Change of Transcript Levels and Enzyme Activities
Partial uncoupling of transcription and posttranscriptional regulation generates a regulatory network that combines flexibility with robustness. Whereas transcription responds rapidly to short-term changes in the physiological status, the damped and delayed response of enzyme activities means that the changes of transcription will only affect metabolism if they are sustained for a long time. This has obvious advantages in a leaf, which is subjected every day to recurring changes of the light intensity, leaf water potential, and the levels of many metabolites every day. On the one hand, it is important to regulate transcription in response to changes in these parameters, otherwise the plant will be unable to respond to changes in its environment. On the other hand, alterations in protein levels every day would be energetically wasteful because the levels of proteins are far higher than of individual transcript. Even more crucially, changes of enzyme levels would impair metabolism and growth on the next day, until they are reversed (Gibon et al., 2004). The results in this article show that for most enzymes, changes in transcript levels have only a small impact on enzyme activities within a single day but lead to marked changes if the diurnal pattern is modified for several days. For example, the period of carbon starvation at the end of every night in pgm (Gibon et al., 2004; Thimm et al., 2004) leads to recurring changes in the levels of transcripts that generate stable changes in the levels of several enzymes, including increased activities of GLDH and invertase and decreased activities of NIA and GOGAT. In a more extreme treatment, whereas transcript levels respond to a prolonged night within 4 to 8 h, enzyme activities, with one exception, do not change markedly for 48 h.
Use of Enzyme as Diagnostic Markers
The relative stability of enzyme activities and the likelihood that they reflect long-term acclimation rather than a rapid response to a transient situation makes them ideal diagnostic markers. To illustrate this, we have used a machine-learning approach to diagnose the physiological status of the starchless pgm mutant. After several hours of acute sugar depletion during the second part of the night (Gibon et al., 2004), the transcript profile in pgm at the end of the night resembles a wild-type plant that has experienced a 4 to 6 h extension of the night (Thimm et al., 2004). The decision tree that we constructed revealed that the enzyme activity profile in pgm resembles that in a wild-type plant that has experienced 3 d of continuous darkness. This result is obtained irrespective of the time of day when pgm is harvested. This illustrates that the rapid changes of transcripts generated by the daily recurrence of a period of sugar starvation are integrated to generate a quasi-stable enzyme profile that resembles a wild-type plant that has been exposed to several days of carbon starvation. We are presently collecting a portfolio of potential diagnostic enzyme activities from plants based on data obtained across a range of growth conditions. The data will be used as training sets for machine-learning systems. We expect this approach to become useful for the characterization of ecotypes, mutants, and other genetically altered plants and for the evaluation of plant performance in different conditions.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
) were grown in an 8-h day phytotron. At least 3 weeks before their use, the plants were transferred into a small growth cabinet with a 12-h d of 160 µE and 20°C throughout the day/night cycle. Harvests of 15 plant rosettes at a time point were performed sequentially every 2 h within a day/night cycle or 0, 2, 4, 8, 24, 48, 72, and 148 h in total darkness. Each sample typically contained three rosettes, equivalent to 500 mg FW. The entire sample was powdered under liquid nitrogen and stored at –80°C until its use.
Reagents
Chemicals were purchased from Sigma (Taufkirchen, Germany), except NAD+, NADH, NADP+, NADPH, and phosphoenolpyruvate (Roche, Mannheim, Germany) and glucose, fructose, sucrose (Merck, Darmstadt, Germany). Xylulose-5-phosphate was purchased from W.D. Fessner (University of Darmstadt, Germany). Dihydroxyphenoxazine (Amplex red) was purchased from Molecular Probes Europe (Leiden, The Netherlands). Enzymes for analysis were purchased from Roche except phosphoglycerokinase and glycerokinase (Sigma-Aldrich). UMP-kinase was overexpressed and purified as in Serina et al. (1995). The clone encoding UMP-kinase is a generous gift from Octavian Barzu (Institut Pasteur, Paris, France).
Preparation of RNA, cDNA, and Labeled cRNA, Array Hybridization, and Data Evaluation
Extraction of RNA, cDNA synthesis and labeling, and hybridization on Affymetrix ATH1 arrays were performed as in Thimm et al. (2004) and data evaluation as in Gibon et al. (2004).
Extraction of Enzymes
For routine measurements, samples corresponding to 500 mg FW were collected, ground to a powder in liquid nitrogen, and stored at –80°C. Aliquots of 10 to 20 mg FW were subaliquoted at –180°C and extracted by vigorous vortexing with extraction buffer. The composition of the extraction buffer was 10% (v/v) glycerol, 0.25% (w/v) BSA, 0.1% (v/v) Triton X-100, 50 mM Hepes/KOH, pH 7.5, 10 mM MgCl2, 1 mM EDTA, 1 mM EGTA, 1 mM benzamidine, 1 mM 
-aminocapronic acid, 1 mM phenylmethylsulfonyl fluoride, 10 µM leupeptin, and 1 mM DTT. Phenylmethylsulfonyl fluoride was added just before extraction. DTT was omitted when using peroxidase-based or MTT-based indicator reactions. In all cases, 500 to 1000 µL of extraction buffer was used, leading to an initial 50-fold (w/v) dilution. Subaliquots were diluted further to generate appropriate dilutions for the different enzymes.
Pipetting Scheme for Stopped Assays
Ninety-six–well microplates were prepared using a Multiprobe II pipetting robot equipped with a cooling block, an incubation block, a shaker, and a gripper (Perkin-Elmer, Zaventem, Belgium). When necessary, the extracts were further diluted in the extraction buffer using an automated program on the robot. Then, aliquots of 5 µL of extracts as well as standards were dispensed into a 96-well microplate held at 4°C, then 40 µL of reaction mix were added. The microplate was transferred to 25°C, gently shaken, and preincubated for 5 min. Then, 5 µL of a starter solution were added, and the microplate was shaken again. After a defined time (0, 10, or 20 min, depending on the assay), 20 µL of a stop solution (NaOH, HCl, or N-ethylmaleimide, depending on the assay) were added, and the plates were incubated at room temperature for a further 10 min. T
0.75; 0.50 
R2 < 0.75; 0.25 
