Skip Navigation



JXB Advance Access published online on March 3, 2008
Journal of Experimental Botany, doi:10.1093/jxb/ern008
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Supplementary Material
Right arrow All Versions of this Article:
59/7/1875    most recent
ern008v1
Right arrow E-letters: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when E-letters are posted
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Cushman, J. C.
Right arrow Articles by Schlauch, K. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Cushman, J. C.
Right arrow Articles by Schlauch, K. A.
Agricola
Right arrow Articles by Cushman, J. C.
Right arrow Articles by Schlauch, K. A.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author [2008]. Published by Oxford University Press [on behalf of the Society for Experimental Biology]. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org

RESEARCH PAPER

Large-scale mRNA expression profiling in the common ice plant, Mesembryanthemum crystallinum, performing C3 photosynthesis and Crassulacean acid metabolism (CAM)

John C. Cushman1,*, Richard L. Tillett1, Joshua A. Wood2, Joshua M. Branco1 and Karen A. Schlauch1

1Department of Biochemistry, MS200, University of Nevada, Reno, NV 89557-0014, USA
2Department of Animal Biotechnology, MS202, University of Nevada, Reno, NV 89557-0014, USA

* To whom correspondence should be addressed. E-mail: jcushman{at}unr.edu

Received 26 September 2007; Revised 16 December 2007 Accepted 7 January 2008


   Abstract
Top
 Abstract
Introduction
 Materials and methods
 Results and discussion
 Supplementary material
 References
 
The common ice plant (Mesembryanthemum crystallinum L.) has emerged as a useful model for molecular genetic studies of Crassulacean acid metabolism (CAM) because CAM can be induced in this species by water deficit or salinity stress. Non-redundant sequence information from expressed sequence tag data was used to fabricate a custom oligonucleotide microarray to compare large-scale mRNA expression patterns in M. crystallinum plants conducting C3 photosynthesis versus CAM. Samples were collected every 4 h over a 24 h time period at the start of the subjective second day from plants grown under constant light and temperature conditions in order to capture variation in mRNA expression due to salinity stress and circadian clock control. Of 8455 genes, a total of 2343 genes (~28%) showed a significant change as judged by analysis of variance (ANOVA) in steady-state mRNA abundance at one or more time points over the 24 h period. Of these, 858 (10%) and 599 (7%) exhibited a greater than two-fold ratio (TFR) increase or decrease in mRNA abundance, respectively. Functional categorization of these TFR genes revealed that many genes encoding products that function in CAM-related C4 acid carboxylation/decarboxylation, glycolysis/gluconeogenesis, polysaccharide, polyol, and starch biosynthesis/degradation, protein degradation, transcriptional activation, signalling, stress response, and transport facilitation, and novel, unclassified proteins exhibited stress-induced increases in mRNA abundance. In contrast, salt stress resulted in a significant decrease in transcript abundance for genes encoding photosynthetic functions, protein synthesis, and cellular biogenesis functions. Many genes with CAM-related functions exhibited phase shifts in their putative circadian expression patterns following CAM induction. This report establishes an extensive catalogue of gene expression patterns for future investigations aimed at understanding the complex, transcriptional hierarchies that govern CAM-specific expression patterns. A novel graph-theoretic approach called ‘Max Clique Builder’ is introduced that identifies and organizes sets of coordinately regulated genes, such as those encoding subunits of the vacuolar H+-ATPase complex, into tighter functionally related clusters with more similar expression patterns compared with standard hierarchical clustering methods.

Key words: Common ice plant, Crassulacean acid metabolism, Max Clique Builder, Mesembryanthemum crystallinum L., mRNA expression profiling, salinity stress


   Introduction
Top
 Abstract
 Introduction
Materials and methods
 Results and discussion
 Supplementary material
 References
 
Crassulacean acid metabolism (CAM) is a specialized form of photosynthetic carbon assimilation estimated to occur in ~7% of vascular plant species (Crayn et al., 2004; Silvera et al., 2005). CAM is characterized by nocturnal uptake and assimilation of atmospheric and/or respiratory CO2 by the enzyme phosphoenolpyruvate (PEP) carboxylase (PEPC), resulting in the accumulation of C4 acids, such as malate, which is stored in large vacuoles within chloroplast-containing cells. The glycolytic breakdown of carbohydrate formed during the previous day supplies the C3 substrate, PEP, for nocturnal malate formation. By conducting the bulk of atmospheric CO2 uptake at night, when evapotranspiration rates are low, CAM plants achieve water use efficiencies that are 3- to 6-fold higher than C4 and C3 plants, respectively (Nobel, 1996). During the day, the C4 acids are transported to the cytoplasm and are decarboxylated to release PEP or pyruvate, which provide the substrate for daytime CO2 fixation by ribulose 1,5-bisphosphate carboxylase/oxygenase (Rubisco), gluconeogenesis, and carbohydrate formation and storage in the chloroplast as starch or in the vacuole as sucrose/hexose, depending on the plant species (Christopher and Holtum, 1996).

The expression of the CAM pathway is highly plastic in some species, and the extent to which the CAM pathway is employed is governed largely by a range of environmental conditions that impose water deficit stress upon the plant (Cushman and Bohnert, 1999; Cushman and Borland, 2002). For example, the widely studied, halophytic annual and so-called C3–CAM species, Mesembryanthemum crystallinum (common or crystalline ice plant), displays C3 photosynthesis when grown under non-stressed conditions and is capable of completing its life cycle in the C3 mode without ever exhibiting net nocturnal CO2 uptake (Winter and Holtum, 2007). However, when plants are grown under various stress treatment conditions that reduce leaf water potential, such as high salinity, high light, or water deficit, they exhibit all of the physiological features of a CAM plant (Winter and Holtum, 2005). CAM induction is mediated by a calcium-dependent signalling pathway (Taybi and Cushman, 1999), that can be either abscisic acid (ABA) dependent or ABA independent, but which requires protein synthesis (Taybi and Cushman, 2002). The stress-induced shift from C3 photosynthesis to CAM is also accompanied by an increased activity of the antioxidative stress response system (Hurst et al., 2004); however, oxidative stress alone is apparently insufficient to induce CAM (Borland et al., 2006). The inducibility of CAM in M. crystallinum has made it a favourite model for studying the molecular genetics of CAM (Bohnert and Cushman, 2001; Cushman, 2001). More than 20 cDNA libraries exist from M. crystallinum from different tissues, such as meristems, roots, shoots, leaves, epidermal bladder cells, flowers, and seed capsules, and different stress treatments (Bohnert et al., 2001). An ice plant expressed sequence tag (EST) database was created to facilitate gene discovery efforts in this species (Kore-eda et al., 2004).

In addition to being an important ecophysiological adaptation to limiting water supply, CAM is also one of the best characterized physiological rhythms in plants and represents an interesting example of circadian clock specialization (Boxall et al., 2005). Because all the enzymatic machinery for CAM is present within a single cell, competing carboxylation reactions by PEPC and Rubisco must be strictly controlled in time. Carbon flux through PEPC is regulated, in part, by reversible protein phosphorylation catalysed by a circadian-controlled PEPC kinase (Hartwell et al., 1999; Taybi et al., 2000). Rubisco activity is also modulated, with peak activity apparent during the mid to late part of the light period (Maxwell et al., 1999; Griffiths et al., 2002). PEPC kinase transcript abundance is inversely correlated with cytoplasmic malate concentrations, suggesting that malate exerts a negative effect on PEPC kinase gene expression or mRNA stability and appears to over-ride its circadian control (Borland et al., 1999; Nimmo, 2000; Borland and Taybi, 2004). Circadian control of the large, reciprocating pools of carbohydrates (Dodd et al., 2003) and associated transport activities (Häusler et al., 2000; Kore-eda et al., 2005), and the distinction between different classes of carbon pools appear to be critical for the performance of CAM (Borland and Dodd, 2002).

In this study, the large-scale mRNA expression profiles in leaf tissues of M. crystallinum plants performing either C3 photosynthesis or CAM grown under constant light and temperature conditions are reported in order to assess mRNA expression changes under circadian clock control. This work establishes an extensive catalogue of gene expression patterns for future investigations into the complex, transcriptional hierarchies that govern CAM-specific expression patterns. Finally, a novel graph-theoretic clustering approach is introduced that employs both correlation coefficient and Euclidean distance metrics for generating clusters of genes with coordinate expression patterns. This clustering method generates clusters with less variation, allowing biologically relevant expression patterns to be discerned that are useful for recognizing and predicting genes with related functions.


   Materials and methods
Top
 Abstract
 Introduction
 Materials and methods
Results and discussion
 Supplementary material
 References
 
Plant materials
For mRNA expression analyses, M. crystallinum L. plants were grown in soil (Metromix 200; Scotts Sierra Horticultural Products, Marysville, OH, USA) and watered daily with 0.25x Hoagland's solution (Hoagland and Arnon, 1950) in a growth chamber on a 12 h light (28 °C, 450–500 µE m–2 s–1)/12 h dark (18 °C) cycle. Plants performing C3 photosynthesis were irrigated daily with 0.25x Hoagland's solution. Five-week-old plants were stressed by watering with 0.5 M NaCl (salt-stressed) in 0.25x Hoagland's for 14 d to induce CAM. Before tissue collection, all plants were exposed to a 24 h period of constant light and temperature (28±1.5 °C) without humidity control to ensure that changes in gene expression were due to endogenous circadian control as opposed to external cues (Boxall et al., 2005). The third and fourth leaf pairs were collected at subjective dawn of the second day, corresponding to Zeitgeber time (ZT) point ZT24, every 4 h over a 24 h period (Fig. 1), frozen in liquid nitrogen, and stored at –80 °C until RNA extraction. Two biological replicates were collected and analysed for all time points.


Figure 1
View larger version (25K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 1. Experimental strategy for comparing mRNA expression patterns in M. crystallinum plants performing C3 photosynthesis versus those performing CAM. L, light; D, dark; SL, subjective light; SD, subjective dark. Samples were collected every 4 h over a 24 h time period (arrows) from plants grown under constant light and temperature conditions for the preceding 24 h period.

 
RNA isolation for microarray analysis
The frozen leaf tissue was ground in liquid nitrogen with a mortar and pestle, and total RNA was extracted from the frozen powder using a cetyltrimethyl ammonium bromide-based RNA extraction method (Hartwell et al., 1996). The total RNA was further purified using a Qiagen RNeasy plant mini kit (Qiagen Inc., Valencia, CA, USA) according to the manufacturer's instructions. RNA integrity was confirmed by electrophoresis on 1.5% agarose gels containing formaldehyde (Sambrook et al., 1989).

NimbleGen microarray analysis
Custom oligonucleotide microarrays were fabricated by NimbleGen, Inc. (Madison, WI, USA) using photolithography directed by the Maskless Array Synthesizer (MAS) (Singh-Gasson et al., 1999). The custom M. crystallinum oligonucleotide microarray contained probe sets for 8455 unique genes as defined by The Dana Farber Cancer Institute (DFCI) Ice Plant Gene Index Release 4.0 (April 23, 2003; http://compbio.dfci.harvard.edu/tgi/cgi-bin/tgi/gimain.pl?gudb=ice_plant). Each probe set (excluding the two controls Bluescript and polyubiquitin, which consisted of 1507 and 1595 probe pairs, respectively) consisted of 11 perfect match (PM) and 11 mismatch (MM) 24mer oligonucleotides per gene for a total of 192 546 features. Total RNA (15 µg) of each biological replicate sample was converted to biotin-labelled cRNA and hybridized and washed as described (Nuwaysir et al., 2002). Original probe annotation from TIGR was updated by nucleotide sequence query against the UniProt Knowledgebase Release 9.2 (UniProtKB/Swiss-Prot Release 51.2 and UniProtKB/TrEMBL Release 34.2) (Bairoch et al., 2005) using Tera-BLASTP (double-banded Smith–Waterman), word size 3, gap open penalty=11, extension penalty=1, and a minimum significance of 1e–5 using DeCypherTM programmable logic hardware (TimeLogic, Inc., Carlsbad, CA, USA) at the Nevada Center for Bioinformatics. Functional categories were assigned automatically by amino acid homology to similar proteins categorized according to the Munich Information Center for Protein Sequences (MIPS; http://mips.gsf.de) Funcat 2.1 classification scheme (Ruepp et al., 2004). Manual annotation and bibliographic searches were also performed to correct annotation errors as necessary.

Microarray data processing and analysis
All NimbleGen custom oligonucleotide arrays were processed first by RMA (Robust Multi-Array Average) (Irizarry et al., 2003) using the R package affy (Gautier et al., 2004). Specifically, expression values were computed from the pair files by first applying the RMA model of probe-specific correction of PM probes. A visual inspection showed that the distributions of raw PM probe values of all 28 arrays in this study were very similar, with no apparent outlying arrays. Digestion curves describing trends in RNA degradation between the 5' end and the 3' end in each probe set were generated, and all 28 proved very similar (data not shown). The corrected probe values were then normalized via quantile normalization, and a median polish was applied to compute one expression measure from all probe values. To adjust for the batch effect between replicates in this experiment, a method was used that employs an Empirical Bayes framework effective in small sample sizes (Johnson et al., 2007). This adjustment resulted in a mean correlation of replicate arrays of 92.4%. Resulting RMA expression values were averaged across replicates, and ratios between these stress and control averages were log2-transformed.

ANOVA
The following model was used for the analysis of variance (ANOVA): yijk=Ei+Tj+(ET)ij+{epsilon}ijk, where yijk denotes the log2 signal measured for experimental state i, time j, and biological replicate k, with 1≤i≤2, 1≤j≤7, and 1≤k≤2. The terms Ei and Tj measure the effect of the experimental condition and time point, respectively, and the interaction term (ET)ij accounts for the interaction between environmental condition and time. After the application of a multiple testing correction [false discovery rate (FDR); Benjamini and Hochberg, 1995], the ANOVA indicated that no genes had a significant interaction effect (i.e. all interaction P-values were >0.05). This implies that the differences in expression measures between stress and control were consistent across the time points, or that two replications were simply not enough to detect differences of significance at an FDR of 0.05.

Hierarchical clustering
Hierarchical clustering was performed using Pearson's correlation coefficient as distance metric and the average agglomeration method. Clustering dendrograms were examined below the 0.15 height threshold, allowing a close inspection of genes clustered at or above a cluster-average correlation coefficient of 0.85.

Maximal clique builder
As a complement to standard hierarchical clustering, a novel graph-theoretic approach to constructing putative functional gene clusters was used in this study. This method generates clusters with apparently less variation (i.e. ‘tighter’ clusters), more biologically meaningful results, and improved hypotheses about functions of unknown genes. This method is based on two components: a user-defined weighted function designating a measure of similarity (or distance) between two gene expression profiles and structural properties of a bi-directional graph.

The distance metric associates with each pair of genes a measure of similarity, where the notion of similarity may be experiment dependent, and is determined by the researcher as a combination of several distance metrics including standard metrics such as Pearson's or Spearman's correlation coefficient, the cosine of the angle between the expression vectors, the Euclidean distance, and less commonly applied distance metrics such as the Manhattan (city block) metric and the Munneke penalized metric (Munneke et al., 2005). Combining, for example, the correlation and Euclidean distance allows the user to define two expression profiles as similar if they are both alike in pattern and close in magnitude across the time series. More specifically, in the example above, the researcher can define two genes as similar if the Euclidean distance between their expression vectors is less than a specified threshold, and their correlation coefficient is greater than a second threshold. From such relationships, a bi-directional graph is drawn, in which two genes are connected by an edge if they are deemed similar with respect to the rules indicated above. Clusters are detected within the structure of the graph as maximal cliques. Maximal cliques represent gene clusters in which connectivity satisfies the transitive property of similarity, generating clusters in which variation of profiles across group members is typically less than those generated under standard hierarchical clusters using the correlation coefficient as distance measure. For this analysis, Max Clique Builder (MCB) parameters were used as follows. Let µeuc and {sigma}euc denote the mean and standard deviation of the set of pair-wise Euclidean distances across all gene pairs in group G. Let cor(g1,g2) represent the Pearson correlation coefficient between the two genes g1, g2 of group G. Similarly, let euc(g1,g2) denote the Euclidean distance between the two gene profile vectors. Then g1 and g2 were considered similar if one of the four conditions below was satisfied.

Quantitative real-time RT-PCR
The frozen leaf tissue was ground in liquid nitrogen with a mortar and pestle, and total RNA was extracted from the frozen powder using a Qiagen RNeasy plant mini kit (Qiagen) with on-column DNase treatment according to the manufacturer's instructions. RNA integrity was confirmed by electrophoresis on 1.5% agarose gels containing formaldehyde. cDNA was synthesized using an iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer's instructions with a uniform 1 µg of RNA per reaction volume reverse-transcribed. Primers for genes assayed by real-time PCR were selected using Primer3 software (Rozen and Skaletsky, 2000). Quantitative real-time PCRs were prepared using an iTaq SYBR Green Supermix with ROX (Bio-Rad) and performed using the ABI PRISM® 7000 Sequence Detection System (Applied Biosystems, Foster City, CA, USA). Expression was determined for triplicate biological replicates by use of serial dilution cDNA standard curves per gene and normalized by comparison with a polyubiquitin control gene (TC4886). Gene-specific primers used included: polyubiquitin (TC4886), forward primer (FP) 5'-CGCACCTTGGCTGACTACA-3', reverse primer (RP) 5'-AACAACCAGACCATGCAACA-3', product size=103 bp; NADP-dependent malate dehydrogenase (NADP-MDH; TC4960), FP 5'-TCCTCAACCAGCCGATCTT-3', RP 5'-TCTTCACAGTGGAAGCACAGA-3’; product size=127 bp; {alpha}-glucan water dikinase, chloroplast precursor (TC4965), FP 5'-AAGATATTGAAGGAGTTGTCAAGGA-3', RP 5'-GTGCTGAGACACGACTCTATGATG-3', product size=117 bp; pyrophosphate-dependent phosphofructo-1-kinase (TC5411), FP 5'-GCAATGGATTTGCCAAGAGT-3', RP 5'-ATATGCGCCTCCAAGACATC-3', product size=85 bp; {alpha}-glucan (starch) phosphorylase H isozyme (TC5982), FP 5'-TTAGCAGCGACCGGACAAT-3', RP 5'-ACATCAACAGCACAAGGCATA-3', product size=102 ; and PEPC 1 (TC6284), FP 5'-TGCTCCTGGACTTGAAGACA-3', RP 5'-CCCACAAACCGTAGCCATTA-3', product size=155 bp. The Pearson correlation coefficient of the linear regression of 10 pairs of microarray/qRT-PCR log2 (CAM/C3) expression ratios was computed as 0.942. Using Fisher's Z-transform, the confidence interval of the coefficient was computed, with a P-value of P <<0.001.


   Results and discussion
Top
 Abstract
 Introduction
 Materials and methods
 Results and discussion
Supplementary material
 References
 
Identification of salinity stress-induced or repressed genes sampled over 24 h
The mRNA expression profiles of leaf tissue of 5-week-old unstressed M. crystallinum plants conducting C3 photosynthesis were compared with those of 5-week-old plants that were salinity stressed every day for 14 d to induce CAM. High salinity stress was used because it results in a strong and predictable degree of CAM expression in mature leaves (Winter and Holtum, 2007). All plants were exposed to constant light and temperature conditions for 24 h prior to sampling every 4 h over a 24 h period for a total of seven samples per condition (Fig. 1). Multiple time points were collected per treatment, to survey not only for stress-induced changes, but also for circadian clock-controlled changes in mRNA expression, as the majority of stress-induced genes are known to be under circadian clock control (Kreps et al., 2002). Duplicate cDNA samples were hybridized to a custom M. crystallinum NimbleGen oligonucleotide-based microarray representing 8455 genes. The gene content for microarray fabrication was derived from the DFCI Ice Plant Gene Index database (Rel. 4.0) that consisted of 25 640 ESTs and 200 ETs, assembled into 2851 clusters, and 5557 singletons representing 8455 unique sequences from M. crystallinum. It is estimated that more than one-third of the M. crystallinum transcriptome was surveyed with this microarray, with a bias towards leaf-expressed genes as most ESTs were sampled from cDNA libraries prepared from leaf tissue.

After performing an ANOVA and a multiple testing correction (FDR) (Benjamini and Hochberg, 1995), 2343 probe sets (28%) were found to be differentially expressed (P ≤0.05) between experimental states at one or more of the seven time points (Supplementary Table S1 available at JXB online). These probe sets are referred to as those passing the ANOVA filter. From this set of genes, a subset of 1458 probe sets was extracted that displayed a 2-fold (or greater) ratio change in steady-state transcript abundance at one or more of the seven time points sampled (Table S2 at JXB online). This subset of genes is referred to as the two-fold ratio (TFR) set. Of these, 858 (10%) and 599 (7%) exhibited a TFR increase or decrease in mRNA abundance, respectively. One probe set for NADP-dependent malic enzyme (TC4960) exhibited both a significant increase (at ZT36) and a decrease (at ZT24) in transcript abundance. It was decided to examine genes with at least a 2-fold statistically significant differential regulation more closely; the 2-fold threshold was chosen in this study as it encompasses roughly a spread of 2 SDs from the mean of all ratios exhibited in the set of 2343 probe sets (see Fig. S1 at JXB online). Additionally, of all genes represented on the microarray, 10.4% of all 59 248=8464x7 ratios are captured by this threshold, which we believe is a reasonable percentage (see Fig. S2 at JXB online).


View this table:
[in this window]
[in a new window]

  		Table 2. Identification and functional descriptions of the 30 genes exhibiting the most highly increased steady-state transcript abundance in CAM-performing M. crystallinum leaves

 

Figure 2
View larger version (8K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 2. Principal component analysis (PCA) of mRNA expression profiles in M. crystallinum plants performing C3 photosynthesis (squares) and CAM (circles) during a 24 h time period under constant light and temperature conditions. Samples were collected every 4 h corresponding to ZT24, ZT28, ZT32, ZT36, ZT40, ZT44, and ZT48. Each symbol represents the average of two biological replicates for each indicated sampling time. The x-axis (first principal component) is associated with 76.4% and the y-axis (second principal component) is associated with 20.4% of the total variation, clearly indicating differences between stress conditions and among sampling times.

 
Principal component analysis (PCA) was used to simplify and visualize the greatest variances among different time points within the large-scale transcriptomic data set (Fig. 2). The two first principal components explaining 96.8% of the overall variance of transcription profiles (76.4% and 20.4% for axes 1 and 2, respectively) allowed the clear differentiation of samples from plants performing C3 photosynthesis and CAM (Fig. 2). Samples from the subjective morning time points clustered together, as did samples from the subjective evening, indicating that pronounced differences in expression patterns were also occurring over time, most probably due to circadian clock control.

Functional categorization of salinity stress-repressed or -induced genes
To determine the overall functional categorization of stress-induced/repressed genes with temporally fluctuating expression patterns, functions were assigned to each gene according to the MIPS Funcat 2.1 classification scheme (Fig. 3). Manual curation using this functional categorization scheme was used rather than an automated Gene Ontology scheme (Al-Shahrour et al., 2005) because more complete and accurate functional assignments could be achieved. Comparison of the relative number of genes within each functional category revealed several distinct differences in mRNA abundance patterns between unstressed plants performing C3 photosynthesis and salinity-stressed plants performing CAM. Gene categories showing significant reductions in transcript abundance following salinity stress included photosynthesis (i.e. 02 Energy), which included genes encoding selected isozymes of carbonic anhydrase (CA), Rubisco small subunit precursors, subunits of the photosystem (PS) I and II reaction centre complexes, associated light-harvesting chlorophyll a/b-binding proteins of the light-harvesting antenna complexes, and the PSII-associated oxygen-evolving complex. Transcript abundance for genes encoding protein synthesis machinery was also reduced (i.e. 12 Protein synthesis) including genes encoding ribosomal protein subunits, translational apparatus components, and proteins that mediate protein folding (e.g. chaperonins). Genes encoding cell wall components such as pectin and polygalacturonidase (i.e. 42 Biogenesis of cellular components) also showed reduced transcript abundance (Fig. 3; Table S2 at JXB online). A list of the 30 genes exhibiting the greatest declines in relative steady-state transcript abundance is shown in Table 1. These results indicate that chloroplast functions related to C3 photosynthesis and photosynthetic electron transport activities are sharply reduced following salinity stress.


Figure 3
View larger version (16K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 3. Functional categories of genes with increased (red bars) or decreased (blue bars) mRNA abundance with temporally dynamic expression patterns. Functional categories were assigned by homology search comparison using the UniProt Knowledgebase annotations and the Munich Information Center for Protein Sequences (MIPS) Funcat 2.1 classification scheme. Gene functional assignments were not made for genes that reported a significant BLASTP match to a hypothetical or unknown gene product (i.e. 98 Classification not yet clear-cut). Mesembryanthemum crystallinum genes with no significant BLASTP hit e-values <1e–5 were categorized as unclassifiable (i.e. 99 Unclassified proteins).

 

View this table:
[in this window]
[in a new window]

  		Table 1. Identification and functional descriptions of the 30 genes exhibiting the most highly decreased steady-state transcript abundance in CAM-performing M. crystallinum leaves

 
Functional categories of genes showing significant increases in steady-state transcript abundance following salinity stress included those encoding products involved in general metabolic processes (i.e. 01 Metabolism), which included glycolysis/gluconeogenesis, starch degradation and synthesis, polyol and raffinose series sugar biosynthesis, and chlorophyll degradation (Fig. 3). Transcript abundance for genes encoding transcription factors and transcript processing were also increased (i.e. 11 Transcription) including many homeobox-leucine zipper and basic-leucine zipper (BZip) transcription factors. Dramatic increases in transcript abundance for genes encoding functions associated with protein folding, modification, and destination (i.e. 14 Protein fate) are evident, including multiple heat shock, DnaJ proteins, and peptidyl-prolyl cis-trans isomerases, a large number of cysteine and aspartate proteinases, and ubiquitin-activating and -conjugating enzyme subunits (Fig. 3). Transcript abundance for genes encoding RNA-, lipid-, and metal-binding proteins was also increased (i.e. 16 Proteins with binding functions). Genes encoding transport-related functions (i.e. 42 Cellular transport, transport facilitation, and transport routes), including mitochondrial dicarboxylate carrier proteins, Na+/H+ antiporters, and vacuolar ATPase subunits, also showed increased transcript abundance. Genes encoding signalling functions (i.e. 30 Cellular communication/signal transduction mechanism) also showed increased transcript abundance including protein 2C phosphatases, a calcium-dependent protein kinase, a calcium sensor calcineurin B-related protein, and multiple phospholipases involved in phosphoinositol signalling. Many gene products with stress response functions (i.e. 32 Cell rescue, defence, and virulence; 36 Systemic interaction with the environment), such as salt-induced proteins, late embryogenesis abundant (LEA) proteins, oxidative stress response, oxygen radical detoxification enzymes, elicitor- and wound-induced proteins, and jasmonic acid/ethylene-dependent response proteins, also exhibited large increases in steady-state transcript abundance (Fig. 3; Table 2S at JXB online). A list of the top 30 genes that exhibited the greatest increases in relative steady-state transcript abundance is shown in Table 2. Several of these most strongly induced genes encoded products with CAM-related functions, including glucose-6 phosphate/Pi translocator isogene 2 (Kore-eda et al., 2005), PEPC 1 (Cushman et al., 1989), and enolase (Forsthoefel et al., 1995a). Finally, genes involved in the production of osmoprotectant such as inositol 4-methyltransferase (Vernon and Bohnert, 1992), which catalyses a key step in the production of pinitol, and stachyose synthase, which catalyses a range of different galactosyl transfer reactions leading to the formation of stachyose, a tetrasaccharide of the raffinose series of sugars (raffinose+galactinol->stachyose+myo-inositol) and galactopinitol A (D-pinitol+galactinol β{leftrightarrow}galactopinitol A+myo-inositol) (Hoch et al., 1999), are among the most strongly induced in the ice plant following salinity stress. Overall, these results indicate that a set of genes encoding a wide variety of stress-adaptive functions are sharply increased following salinity stress.

Gene functional assignments were not designated (i.e. 98 Classification not yet clear-cut) for 10.6% and 13.8% of genes with decreased or increased transcript abundance, respectively, which had significant BLASTP hits. Another 3.0% and 10.8% of genes with increased or decreased transcript abundance, respectively, had no significant BLASTP hit (i.e. 99 Unclassified proteins) (Fig. 3; Table S2 at JXB online). The much larger number of unknown or novel genes in salinity-stressed plants indicates that such salinity stress-induced genes are less likely to be functionally characterized than genes from plants grown under unstressed conditions (Kore-eda et al., 2004).

CAM-related C4 metabolism genes
From within the 2343 probe sets passing the ANOVA filter (Table S1), genes with expected CAM-related functions were extracted and subjected to hierarchical cluster analysis. The vast majority of these genes exhibited circadian clock-regulated expression patterns, with peak expression evident in the early morning, evening, night, and late morning/mid-day on scanning the heat map from top to bottom (Fig. 4). These genes were identified and organized by functional category in Table 3. Additional putative CAM-related isogenes that did not show a significant change in mRNA expression are listed in the footnotes of Table 3.


Figure 4
View larger version (52K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 4. A heat map displaying simple hierarchical clustering of 56 genes with CAM-related functions and associated isogenes that exhibited ANOVA-significant changes in expression at one or more time points. Clustering was performed using the Pearson correlation coefficient as distance metric and the average agglomerative method. Corresponding gene descriptions organized by functional category are listed in Table 3.

 

View this table:
[in this window]
[in a new window]

  		Table 3. Functional categorization of CAM-related genes and non-CAM isogenes in CAM-performing M. crystallinum leaves

 
C4 metabolism enzymes responsible for the conversion of PEP and CO2 into malate include plastidic CA, cytoplasmic PEPC, cytoplasmic NADP-MDH, and chloroplastic, mitochondrial, and glyoxysomal isogenes encoding NAD-malate dehydrogenase (NAD-MDH) (Table 3). Cah1 and Cah2 isogenes encoding CA, which catalyse the conversion of CO2 to HCO3–, the substrate for PEPC, showed a >8-fold increase in mRNA abundance, with peak putative circadian expression at subjective evening (Fig. 4). Transcript abundance for two additional CA isogenes (Cah3 and Cah4) was repressed to a similar magnitude following salinity stress. The Ppc1 isogene showed a >30-fold increase in transcript abundance following stress induction, in agreement with previous observations (Cushman et al., 1989), with a putative circadian expression pattern that peaks at subjective evening. In contrast, the Ppc2 isogene was repressed by salinity stress. Mdh1, which encodes plastidic NADP-MDH (Cushman, 1993), and Mdh2, which encodes cytosolic NAD-MDH, showed a 2- and 8-fold increase in transcript abundance, respectively, with peak transcript accumulation at subjective night and evening, respectively (Fig. 4). In contrast, transcript accumulation for the glyoxysomal isoforms of NAD-MDH were repressed by salinity stress (Table 3). Also included in this group was PEPC kinase (Ppck1), the minimal protein kinase responsible for nocturnal phosphorylation and activation of PEPC (Hartwell et al., 1999; Taybi et al., 2000). Transcripts for Ppck1 peaked during the subjective late night/early morning, consistent with previous observations in ice plant (Taybi et al., 2000).

Isogenes encoding cytoplasmic NADP-malic enzyme (Mod1, Mod2), which catalyses the carboxylation of pyruvate to malate or the reverse reaction involving decarboxylation of malate to pyruvate in the cytosol, also exhibited stress-induced/putative circadian expression patterns. Mod1 showed a >8-fold induction and peak putative circadian expression at subjective evening, indicating a role in carboxylation (Cushman, 1992). The second isogene (Mod2) shows a 4-fold increase in mRNA abundance following salinity stress, with peak expression at subjective morning (Fig. 4), indicating that this isoenzyme may be responsible for daytime decarboxylation of malate to pyruvate with the release of CO2 during the day for subsequent refixation by Rubisco. However, it should be noted that the steady-state transcript abundance changes surveyed here serve as only a partial indicator of the possible relative importance of a particular enzyme due to possible post-translational modification and allosteric regulatory events that can influence enzyme activity. The chloroplast-localized enzyme pyruvate, orthophosphate dikinase (Ppdk1) catalyses the conversion of pyruvate to PEP, which then enters the gluconeogenic pathway. Ppdk1 transcript abundance, like that of Mod1, is induced >8-fold (Fißlthaler et al., 1995) and exhibits peak putative circadian expression at subjective dusk. Whereas the increased transcript abundance for several C4 metabolism enzymes including PEPC (Ppc1) (Cushman et al., 1989), chloroplast NADP-MDH (Mdh1) (Cushman, 1993), cytosolic NADP-malic enzyme (Mod1) (Cushman, 1992), and pyruvate, orthophosphate dikinase (Ppdk1) (Fißlthaler et al., 1995) was known to occur following CAM induction by salinity stress, none of these genes was previously recognized to be circadianly regulated. However, the activity of these enzymes was reported previously to be induced by salinity stress, with small peaks in activity evident during the night period (Holtum and Winter, 1982). The salinity-induced increase and apparent putative circadian fluctuation in mRNA abundance of Mdh2 encoding NAD-MDH, which catalyses the conversion of oxaloacetate to L-malate in the cytosol, had not been described previously; however, the activity of this enzyme was shown to increase during the C3 photosynthesis to CAM transition (Holtum and Winter, 1982), consistent with the observed increases in mRNA abundance following CAM induction. Likewise, the salinity stress-induced and apparent putative circadian expression pattern of Mod2, which encodes a cytosolic NADP-malic enzyme, had not been reported previously.

CAM-related glycolysis/gluconeogenesis genes
A prerequisite for CO2 pumping during CAM is the supply of C3 carbon skeletons by glycolysis to fuel nocturnal CO2 fixation by PEPC and the reciprocal flow of carbon through gluconeogenesis leading to daytime starch biosynthesis in the chloroplast. A suite of glycolytic/gluconeogenic enzymes is responsible for the interconversion of PEP to glucose-1-phosphate, which is the starting point for starch biosynthesis in the chloroplast (Table 3). The genes encoding the cytoplasmic isozymes of enolase (Pgh1), phosphoglycerate mutase (Pgm1), phosphoglycerate kinase (Pgk1), NAD-dependent glyceraldehyde 3-phosphate dehydrogenase (GapC), transaldolase (Tal1), cytoplasmic fructose-bisphosphate aldolase (Fba1), pyrophosphate-dependent phosphofructo-1-kinase (Pfk1), and cytoplasmic glucose-6-phosphate isomerase (Gpi1) exhibited coordinate patterns of salinity stress-induced and putative circadian expression, with peak expression occurring at subjective evening (Fig. 4). In contrast, plastidic triose phosphate isomerase (Tpi1) exhibited peak transcript abundance at ZT40, and the cytoplasmic isozyme of glucose-6-phosphate isomerase (Gpi1) exhibited peak expression at 10.00. The majority of these enzymes also had one or more, mainly chloroplastic, isogenes, which displayed reduced steady-state transcript abundance following salinity stress (Table 3).

Several glycolytic/gluconeogenic enzymes including enolase (Pgh1) (Forsthoefel et al., 1995a), phosphoglycerate mutase (Pgm1) (Forsthoefel et al., 1995b), and NAD-glyceraldehyde 3-phosphate dehydrogenase (GapC) (Ostrem et al., 1990) have been reported previously to exhibit increased transcript abundance following CAM induction by salinity stress. However, none of these genes was recognized previously to be under the control of the circadian clock. It is reported that several additional enzymes including the cytoplasmic enzymes for phosphoglycerate kinase (Pgk1), triose phosphate isomerase (Tpi1), transaldolase (Tal1), fructose-bisphosphate aldolase (Fba1), pyrophosphate-dependent phosphofructo-1-kinase (Pfk1), and the cytosolic isozyme of glucose-6-phosphate isomerase (Gpi1) also show coordinate salinity stress-induced (between 1.5- and 16-fold) and circadian clock-controlled fluctuations in mRNA abundance (Fig. 4). Transaldolase catalyses the reversible transfer of a three-carbon ketol unit from sedoheptulose 7-phosphate to glyceraldehyde 3-phosphate to form erythrose 4-phosphate and fructose 6-phosphate. This enzyme, together with transketolase (TC6417), which undergoes a significant reduction in mRNA abundance following salinity stress (see Table S1 at JXB online), provides a link between the glycolytic and pentose-phosphate pathways.

An earlier study reported that salinity stress increased the activities of enolase, pyruvate, orthophosphate dikinase, phosphoglycerate mutase, phosphoglycerate kinase, NAD/NADP glyceraldehyde 3-phosphate dehydrogenase, fructose-1,6-bisphosphatase, phosphofructo-1-kinase, and glucose-6-phosphate isomerase, and suggested their participation in CAM (Holtum and Winter, 1982).

CAM-related starch synthesis/degradation genes
The diel breakdown and resynthesis of a transient starch reserve is a central requirement of CAM. Up to 20% of leaf dry weight may be committed to this daily cycle of carbohydrate (Borland and Dodd, 2002). In M. crystallinum plants performing CAM, an estimated 40–60% of the starch degraded at night is used to supply PEP for CAM, with the remainder being used for export to soluble sugars and consumption through respiration (Borland and Dodd, 2002). Many genes associated with starch biosynthesis and breakdown that exhibited stress-induced and putative circadian expression patterns were identified (Fig. 4 and Table 3). Two genes encoding the chloroplast precursor of ADP glucose pyrophosphorylase small (Agp1 and Agp2) and large subunit (Agp3), which catalyses the committed step in starch biosynthesis, showed a 3- to 8-fold increase in mRNA expression following CAM induction and exhibited peak expression during the subjective evening (Agp1 and Agp2) or afternoon (Agp3) (Fig. 4). Two different genes (Sss1 and Sss2) were identified, which encode soluble starch synthase isoenzymes I and IV, respectively. Starch synthase I catalyses the elongation of glucans by the addition of glucose residues from ADP-glucose through the formation of {alpha}-1,4 linkages and is a major determinant for the synthesis of transient starch reserves in plants (Delvalle et al., 2005). Both genes exhibited stress-inducible increases in steady-state transcript abundance, with Sss2 showing preferential expression in the subjective night. Three isogenes encoding 1,4-{alpha}-glucan-branching enzyme (Gbe1-3), which catalyses the transfer of a segment of a 1,4-{alpha}-D-glucan chain to a primary hydroxyl group in a similar glucan chain (Tetlow et al., 2004), were identified that exhibited increased mRNA abundance. Each isogene displayed a different time of peak expression. Gbe1, Gbe2, and Gbe3 showed peak expression during the subjective evening, night, and morning, respectively. The observed increases in the expression of genes encoding starch biosynthetic enzymes is consistent with the increased transient starch turnover required for reciprocal C4 acid production during CAM (Borland and Dodd, 2002).

Increased mRNA expression of many genes encoding starch-degrading enzymes was also observed (Fig. 4 and Table 3). Plastidic {alpha}-amylase catalyses the endohydrolysis of 1,4-{alpha}-D-glucosidic linkages, in polysaccharides containing three or more 1,4-{alpha}-linked D-glucose units. The {alpha}-amylase isogene (AmyA1) showed an 8-fold increased mRNA abundance, with peak expression during the subjective evening and night reflecting its likely role in nocturnal starch degradation to fuel C4 acid production (Fig. 4). A second isogene (AmyA2) was also identified, whose expression did not change significantly, but nonetheless showed peak evening expression with declining expression during the subjective night (see Table 3 footnote e). Plastidic β-amylase catalyses the exohydrolysis of 1,4-{alpha}-D-glucosidic linkages, removing β-maltose units from non-reducing ends of 1,4-{alpha}-linked D-glucose units. Two isogenes (AmyB1 and AmyB2) showed 5- to 9-fold stress-induced increases in transcript abundance, with putative circadian expression that peaked during the subjective night. A third isogene (AmyB3) showed less induction, but exhibited peak transcript abundance during the subjective morning (Fig. 4). One pullanase isogene (Sde1) was identified that showed a >4-fold induction in mRNA abundance following salinity stress, with peak subjective evening expression. Pullanase, a member of the isoamylase type of starch-debranching enzymes that hydrolyses {alpha}-1,6 glycosidic linkages, has been shown to play a major role in starch mobilization (Wattebled et al., 2005). The evening expression pattern of this isogene indicates that it may contribute to evening starch degradation to fuel nocturnal malate production.

Stp1, which encodes {alpha}-glucan phosphorylase H, a starch mobilization enzyme that phosphorylates amylopectin to catalyse the release of glucose-1-phosphate, also showed a pronounced (>8-fold) increase in transcript abundance, with peak expression during the subjective evening. A second isogene was also identified (Stp2) with peak expression during the subjective evening; however, its expression was not found to be significantly induced during CAM induction (see Table 3 footnote e). Gtf1, which encodes 4-{alpha}-glucanotransferase or disproportionating enzyme (D-enzyme) that transfers small chain 1,4-{alpha}-D-glucan units to a new position in an acceptor, which may be glucose or a 1,4-{alpha}-D-glucan for further mobilization by starch phosphorylase, exhibited a >5-fold increase in transcript abundance that peaked during the subjective evening and night. Gwd1, which encodes {alpha}-glucan, water dikinase, an enzyme that catalyses the transfer of β-phosphate of ATP to either the C-3 or C-6 position of the glucosyl residue of {alpha}-glucan, and is essential for the nocturnal degradation of transitory leaf starch granules (Yu et al., 2001; Mikkelsen et al., 2005), exhibited a stress-induced putative circadian expression pattern with peak expression at subjective evening and night (Fig. 4). These results are consistent with previous observations that starch-degrading enzymes increase substantially following CAM induction (Paul et al., 1993), with the expression of β-amylase and starch phosphorylase genes exhibiting enhanced mRNA accumulation following CAM induction (Dodd et al., 2003). Thus, it is perhaps not surprising that enhanced expression and putative circadian clock control of several genes associated with nocturnal starch breakdown were observed (Borland and Taybi, 2004). The coordinate, putative circadian regulation of such a large number of genes with functional roles in glycolysis/gluconeogenesis and starch synthesis/degradation suggests that the overt, self-sustaining rhythms of nocturnal CO2 fixation and associated carbon flux through glycolysis and gluconeogenesis is dependent upon the diel breakdown and resynthesis of transient starch reserves (Borland and Taybi, 2004) and is not controlled simply by the carboxylation through PEPC, which apparently becomes dysfunctional within a few subjective day/night cycles under constant light and temperature conditions (Wyka and Lüttge, 2003). Long-term microarray-based mRNA expression profiling over 2–3 subjective day/night cycles coupled with detailed gas exchange measurements and metabolite profiling is needed to elucidate further the relative importance of individual components, such as malate production, that might be required to support the sustained CAM rhythm in the absence of environmental cues.

CAM-related transport genes
Genes encoding CAM-related metabolite transporters are also subjected to salinity stress induction and putative circadian clock control. For example, mRNA abundance increases for several plastidic glucose-6 phosphate/Pi translocators (Gpt2–4), all of which exhibited putative circadian expression patterns that peaked during the subjective evening, consistent with their likely role in transporting glucose-6 phosphate across the plastid envelop to fuel nocturnal glycolysis and carboxylation reactions. One of these genes (Gpt2) showed a 64-fold increase in transcript abundance following salinity stress (Fig. 4 and Table 3), consistent with previous results (Kore-eda et al., 2005). The other two isogenes (Gpt3 and Gpt4) had not been previously characterized. However, a related plastidic glucose-6 phosphate/Pi translocator isogene (Gpt1) (Häusler et al., 2000) did not show ANOVA significant changes in mRNA abundance (Table 3), in contrast to an earlier report (Kore-eda et al., 2005). This discrepancy is likely to be explained by variations in plant growth conditions and/or duration of stress treatments. A plastidic adenylate transporter (Ant1) with ATP:ADP antiport activity, which catalyses the reaction ATP(out)+ADP(in)=ATP(in)+ADP(out), exhibited a significant increase in mRNA abundance following salinity stress, with peak evening expression. This pattern of expression is similar to that observed for this isogene in salinity-stressed plants grown under a diel day/night cycle (Kore-eda et al., 2005). Transcript abundance for a triose phosphate/phosphate transporter (Tpt1) isogene was found to be depressed by salinity stress, with peak expression occurring during the subjective day, consistent with previous results (Kore-eda et al., 2005). A previously characterized Dct1 isogene encoding a plastidic 2-oxoglutarate/malate carrier protein showed reduced transcript abundance following salinity stress and the lowest expression during the evening (Fig. 4 and Table 3), consistent with previous analyses (Kore-eda et al., 2005). However, two additional putative mitochondrial 2-oxoglutarate/malate carrier protein isogenes (Dct3 and Dct4) not previously characterized were found to be strongly induced (>16-fold) following salinity stress, with peak expression during the day and night, respectively (Fig. 4 and Table 3). Although the exact substrates for these carriers remain unknown, their mRNA expression patterns indicate possible functions as either di- or tricarboxylate transporters.

CAM-related genes exhibit stress-induced phase-shifted putative circadian clock-regulated patterns of mRNA expression
The molecular characterization and expression of central circadian clock components in M. crystallinum showed that the circadian clock is largely compensated against development and salinity stress (Boxall et al., 2005). The expression profiles of a subset of CAM-related genes, which encode enzymes with function in C4 metabolism and glycolysis/gluconeogenesis including Ppck1, Ppdk1, Ppc1, Pgh1, Mdh2, Pgm1, Mod1, and Fba1 (see Table 3), were dramatically changed in response to salinity stress (Fig. 5). This graphical presentation also illustrates the relative magnitude of change in mRNA abundance between the C3 photosynthesis and CAM states. When performing C3 photosynthesis, most of these genes exhibited putative circadian expression patterns, with the lowest expression occurring during the subjective evening and the highest expression during the subjective night. In contrast, all of the genes exhibit stress-induced expression and peak transcript abundance at subjective afternoon or evening, with the exception of Ppck1 and Ppdk1, which exhibited peak expression during the subjective night (Fig. 5A, B). These results indicate that the putative circadian clock output patterns of gene expression for genes with CAM-related functions is fundamentally altered following CAM induction. The observed phase shift of 4–8 h in the peak expression of these genes may be modulated by the relative expression of clock-regulated activators or repressors that regulate the expression of evening phased genes (Harmer and Kay, 2005).


Figure 5
View larger version (30K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 5. mRNA expression of selected genes with CAM-related functions that exhibited phase-shifted circadian clock-controlled expression patterns following salinity stress. Normalized expression values were plotted on a linear scale from plants expressing C3 photosynthesis (blue) versus CAM (red). (A) Ppck1, (B) Ppdk1, (C) Ppc1, (D) Pgh1, (E) Mdh2, (F) Pgm1, (G) Mod1, (H) Fba1. Corresponding gene descriptions organized by functional category are listed in Table 3. The shaded bar indicates subjective dark. The open bar indicates subjective light.

 
Hierarchical and graph-theoretical clustering of coordinately regulated vacuolar ATPase genes
We used hierarchical clustering with the Pearson correlation coefficient as the distance metric and average agglomeration method to identify genes belonging to the vacuolar ATPase complex and to analyse their coordinate expression patterns (Fig. 6). The vacuolar H+-ATPase (V-ATPase) complex plays an essential role in maintaining ionic and metabolic gradients across endomembranes and, under conditions of salinity stress, drives vectoral proton pumping from the cytoplasm to the tonoplast interior, which energizes secondary transport functions such as Na+/H+ antiporters, which sequester Na+ ions into the vacuole (Kluge et al., 2003a; Sondergaard et al., 2004). In CAM plants, the V-ATPase also provides the proton motive force to facilitate nocturnal malate transport into the vacuole. In vascular plants, the V-ATPase complex is composed of 12 different subunits. Using the hierarchical clustering approach, it was possible to identify eight V-ATPase subunits in a cluster containing 31 genes (25.8% discovery rate), having an average pair-wise correlation coefficient of 0.931 (Fig. 6A and Table 4). The utility of MCB, a novel graph-theoretic clustering tool, to define groups of genes using cliques of a graph was tested next. To compare the above hierarchical cluster with a maximal clique properly, the maximal clique containing the same eight V-ATPase units was identified that had an average pair-wise correlation coefficient of 0.95. This clique identified 10 out of 18 members (55% discovery rate) of the V-ATPase genes (Table 4). Thus, MCB had a V-ATPase subunit discovery rate of 55% versus only 25.8% with standard hierarchical clustering (using an average pair-wise correlation threshold of 0.85). In addition, the MCB grouping exhibits a much lower degree of variation among the profile magnitudes than obtained with hierarchical cluster, with the mean and standard deviation of pair-wise differences in magnitude 1.29 and 0.38, and 2.47 and 2.7, respectively (Fig. 6B). Genes within this clique showed tightly coordinated expression profiles peaking during the subjective evening, with all V-ATPase genes exhibiting between a 4- and 6-fold increase in transcript abundance following 2 weeks of salinity stress (Fig. 6B). These results are consistent with a similar study in which the mRNA abundance for five of 12 subunits showed increased mRNA abundance from 2- to 5-fold depending on the subunit gene following short-term (6 h) salinity stress (Kluge et al., 2003b). Only the V-ATPase C subunit was recognized previously as being under circadian clock control (Rockel et al., 1997). The relatedness of coordinately regulated genes within the V-ATPase clique is illustrated by the connectivity graph (Fig. 6C). MCB was also able to identify more genes (8 versus 10) encoding subunits of the V-ATPase complex from among a much smaller set of coordinately regulated genes. In general, MCB cliques consisted of genes with coordinate expression patterns having less variation than the hierarchical clusters, allowing biologically relevant expression patterns to be recognized, which may prove useful in predicting genes with related functions.


Figure 6
View larger version (37K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 6. Vacuolar ATPase and genes with similar expression patterns peak near the subjective evening. Expression values were plotted on a log2 scale as the ratio of relative expression between plants expressing CAM versus C3 photosynthesis. (A) Cluster of 31 genes defined by hierarchical clustering using an average pair-wise correlation coefficient cut-off of 0.85, resulting in the average pair-wise correlation of this cluster being 0.93. The V-ATPase genes are indicated by red lines; other genes are indicated by grey lines. (B) Maximal clique of dimension 18 for V-ATPase components. Line colours are the same as in A. Gene index or accession numbers and gene descriptions are listed in Table 4. (C) Connectivity graph of the 18-member maximal clique illustrates an identical degree of connectivity among all 18 gene expression profiles.

 

View this table:
[in this window]
[in a new window]

  		Table 4. Coordinate mRNA expression of V-ATPase genes and genes with similar expression patterns identified using either hierarchical clustering or Max Clique Builder

 
Quantitative real-time RT-PCR validation of microarray expression values
In order to validate the accuracy of the expression patterns determined by microarray analysis, the mRNA abundance estimates obtained were compared using the custom oligonucleotide array with results obtained by quantitative real-time RT-PCR (qRT-PCR). Five genes with CAM-related functions (Ppc1, Mod1, Pfk1, Stp1, and Gwd1; see Table 3) were selected for analysis by qRT-PCR and microarray at two time points (ZT24 and ZT36) that corresponded to subjective dawn and dusk, respectively. The relative fold inductions for both technologies are directly comparable, as shown in Fig. 7. Overall, statistically significant correlations (P <<0.01) between qRT-PCR and RMA normalized data were observed for these 10 observation pairs, having a correlation of 0.942 (Fig. 7). All five genes showed consistent trends in expression at ZT36; however, all but two genes (Stp1 and Gwd1) show consistent trends at ZT24 (fourth quadrant, Fig. 7), probably due to differences in RNA extraction or handling and, of course, detection methodology.


Figure 7
View larger version (7K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 7. Comparison of gene expression ratios reported by the custom NimbleGen M. cystallinum oligonucleotide microarray and by quantitative real-time RT-PCR. Data of five genes and corresponding probe sets (Mod1, Pfk1, Stp1, Ppc1, and Gwd1; see Table 3) from leaf samples at both ZT24 and ZT36 from C3 photosynthesis- and CAM-performing plants. Microarray log-transformed expression ratios (CAM/C3) of photosynthesis (x-axis) are plotted against the same ratios obtained by quantitative real-time RT-PCR (y-axis). Linear regression showed an overall correlation coefficient of 0.9417.

 
Concluding remarks
The first large-scale mRNA expression profiling of any CAM plant is reported. Taking advantage of the stress-inducible response of M. crystallinum to shift to CAM following 2 weeks of salinity stress, a direct comparison of the mRNA expression patterns in plants performing C3 photosynthesis and CAM was possible. These results demonstrated that ~28% of the >8400 genes surveyed exhibited a significant change in gene expression, with ~17% displaying a 2-fold (or greater) ratio change in mRNA abundance. This study revealed isogenes with putative CAM-related functions by their increased steady-state transcript abundance. Most CAM-related isogenes displayed putative circadian clock-controlled expression patterns, with a majority exhibiting peak expression during the subjective evening. Salinity stress resulted in a pronounced phase shift for many of these genes. In addition, the utility of a novel graph-theoretical clustering algorithm for the identification of coordinately regulated genes within the vacuolar ATPase complex was demonstrated. Future studies employing more long-term mRNA expression profiling over several subjective day/night cycles integrated with gas exchange and metabolite profiling are needed to dissect the complex interplay between circadian clock-controlled transcriptional networks and metabolite control that support sustained CAM rhythms in the absence of environmental clues.


   Supplementary material
Top
 Abstract
 Introduction
 Materials and methods
 Results and discussion
 Supplementary material
References
 
Supplementary material is available at JXB online.

Table 1S. Table of mRNA expression data for 2343 probe sets for which steady-state transcript abundance was found to be differentially expressed (P ≤0.05) among experimental states at one or more of the seven time points as determined by ANOVA and a multiple testing correction. A blank or dot (.) indicates no annotation was available for this gene. Data are presented as the log2 ratio of CAM/C3 of expression at each time point indicated.

Table 2S. Table of mRNA expression data for 1458 probe sets that displayed a 2-fold (or greater) ratio (TFR) change in steady-state transcript abundance at one or more of the seven time points sampled as determined by ANOVA and a multiple testing correction. A blank or dot (.) indicates no annotation was available for this gene. Data are presented as the log2 ratio of CAM/C3 of expression at each time point indicated.

Figure 1S. The distribution of log2-transformed ratios of ANOVA-significant (n=2343) genes at all temporal states. The distribution has mean m=0.13 and standard deviation s=1.17. Note that 21.5% of the ratios (16 401=2343x7 total ratios) in this subset are up-regulated beyond the 2-fold threshold [log2(ratio)>1], and 12.1% of the ratios are down-regulated beyond the 2-fold threshold [log2(ratio)< –1].

Figure 2S. The distribution of log2-transformed ratios of all genes (n=8464) at all temporal states. Note that 6.6% of all ratios (59 248=8464x7 total ratios) are up-regulated beyond the 2-fold threshold [log2(ratio)>1], and 3.8% of the ratios are down-regulated beyond the 2-fold threshold [log2(ratio)<–1].


View this table:
[in this window]
[in a new window]

 
 


   Acknowledgements
 
This research was supported, in part, by funding from the National Science Foundation (DBI-0136561) to KAS, the National Science Foundation (IBN-0196070 and DBI-9813360) to JCC, and the Nevada Agricultural Experiment Station, and is published as publication #03077060. The authors thank Mary Ann Cushman and Rebecca Albion for technical support. The Nevada Genomics, Proteomics and Bioinformatics Centers are supported by grants from the NIH Biomedical Research Infrastructure Network (NIH-NCRR, P20 RR16464) and NIH IDeA Network of Biomedical Research Excellence (INBRE, RR-03-008).


   References
Top
 Abstract
 Introduction
 Materials and methods
 Results and discussion
 Supplementary material
 References
 
Al-Shahrour F, Minguez P, Vaquerizas J, Conde L, Dopazo J. BABELOMICS: a suite of web tools for functional annotation and analysis of groups of genes in high-throughput experiments. Nucleic Acids Research (2005) 33:W460–W464.[Abstract/Free Full Text]

Bairoch A, Apweiler R, Wu C, et al. The Universal Protein Resource (UniProt). Nucleic Acids Research (2005) 33:D154–D159.[Abstract/Free Full Text]

Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Statistical Methodology) (1995) 57:289–300.

Bohnert H, Ayoubi P, Borchert C, et al. A genomics approach towards salt stress tolerance. Plant Physiology and Biochemistry (2001) 39:95–311.

Bohnert H, Cushman J. The ice plant cometh: lessons in abiotic stress tolerance. Journal of Plant Growth Regulation (2001) 19:334–346.

Borland A, Dodd A. Carbohydrate partitioning in crassulacean acid metabolism plants: reconciling potential conflicts of interest. Journal of Experimental Botany (2002) 29:707–716.

Borland A, Elliott S, Patterson S, Taybi T, Cushman J, Pater B, Barnes J. Are the metabolic components of crassulacean acid metabolism up-regulated in response to an increase in oxidative burden? Journal of Experimental Botany (2006) 57:319–328.[Abstract/Free Full Text]

Borland A, Hartwell J, Jenkins G, Wilkins M, Nimmo H. Metabolite control overrides circadian regulation of phosphoenolpyruvate carboxylase kinase and CO2 fixation in Crassulacean acid metabolism. Plant Physiology (1999) 121:889–896.[Abstract/Free Full Text]

Borland A, Taybi T. Synchronization of metabolic processes in plants with Crassulacean acid metabolism. Journal of Experimental Botany (2004) 55:1255–1265.[Abstract/Free Full Text]

Boxall S, Foster J, Bohnert H, Cushman J, Nimmo H, Hartwell J. Conservation and divergence of circadian clock operation in a stress-inducible crassulacean acid metabolism species reveals clock compensation against stress. Plant Physiology (2005) 137:969–982.[Abstract/Free Full Text]

Christopher J, Holtum J. Patterns of carbohydrate partitioning in the leaves of Crassulacean acid metabolism species during deacidification. Plant Physiology (1996) 112:393–399.[Abstract]

Crayn D, Smith J, Winter K. Multiple origins of Crassulacan acid metabolism and the epiphytic habit in the neotropical family Bromeliaceae. Proceedings of the National Academy Sciences, USA (2004) 101:3703–3708.[Abstract/Free Full Text]

Cushman J. Characterization and expression of a NADP-malic enzyme cDNA induced by salt stress from the facultative crassulacean acid metabolism plant, Mesembryanthemum crystallinum. European Journal of Biochemistry (1992) 208:259–266.[Web of Science][Medline]

Cushman J. Molecular cloning and expression of chloroplast NADP-malate dehydrogenase during Crassulacean acid metabolism induction by salt stress. Photosynthesis Research (1993) 35:15–27.[Medline]

Cushman J. Crassulacean acid metabolism. A plastic photosynthetic adaptation to arid environments. Plant Physiology (2001) 127:1439–1448.[Free Full Text]

Cushman J, Bohnert H. Crassulacean acid metabolism: molecular genetics. Annual Review of Plant Physiology and Plant Molecular Biology (1999) 50:305–322.[CrossRef][Web of Science]

Cushman J, Borland A. Induction of Crassulacean acid metabolism by water limitation. Plant, Cell and Environment (2002) 25:295–310.[CrossRef][Medline]

Cushman J, Meyer G, Michalowski C, Schmitt J, Bohnert H. Salt stress leads to differential expression of two isogenes of phosphoenolpyruvate carboxylase during Crassulacean acid metabolism induction in the common ice plant. The Plant Cell (1989) 1:715–725.[Abstract/Free Full Text]

Delvalle D, Dumez S, Wattebled F, et al. Soluble starch synthase I: a major determinant for the synthesis of amylopectin in Arabidopsis thaliana leaves. The Plant Journal (2005) 43:398–412.[CrossRef][Web of Science][Medline]

Dodd A, Griffiths H, Taybi T, Cushman J, Borland A. Integrating diel starch metabolism with the circadian and environmental regulation of Crassulacean acid metabolism in Mesembryanthemum crystallinum. Planta (2003) 6:789–797.

Fißlthaler B, Meyer G, Bohnert H, Schmitt J. Age-dependent induction of pyruvate, orthophosphate dikinase in Mesembryanthemum crystallinum L. Planta (1995) 196:492–500.[Web of Science][Medline]

Forsthoefel N, Cushman M, Cushman J. Posttranscriptional and posttranslational control of enolase expression in the facultative Crassulacean acid metabolism plant Mesembryanthemum crystallinum L. Plant Physiology (1995a) 108:1185–1195.[Abstract]

Forsthoefel N, Vernon D, Cushman J. A salinity-induced gene from the halophyte M. crystallinum encodes a glycolytic enzyme, cofactor-independent phosphoglyceromutase. Plant Molecular Biology (1995b) 29:213–226.[CrossRef][Web of Science][Medline]

Gautier L, Cope L, Bolstad M, Irizarry R. affy—analysis of Affymetrix GeneChip data at the probe level. Bioinformatics (2004) 20:307–315.[Abstract/Free Full Text]

Griffiths H, Helliker B, Roberts A, Haslam R, Robe W, Borland A, Maxwell K. Regulation of Rubisco activity in crassulacean acid metabolism plants: better late than never. Functional Plant Biology (2002) 29:689–696.[CrossRef][Web of Science]

Harmer S, Kay S. Positive and negative factors confer phase-specific circadian regulation of transcription in Arabidopsis. The Plant Cell (2005) 17:1926–1940.[Abstract/Free Full Text]

Hartwell J, Nimmo G, Wilkins M, Jenkins G, Nimmo H. Phosphoenolpyruvate carboxylase kinase is a novel protein kinase regulated at the level of gene expression. The Plant Journal (1999) 20:333–342.[Web of Science][Medline]

Hartwell J, Smith L, Wilkins M, Jenkins G, Nimmo H. Higher plant phosphoenolpyruvate carboxylase kinase is regulated at the level of translatable mRNA in response to light or a circadian rhythm. The Plant Journal (1996) 10:1071–1078.[CrossRef][Web of Science]

Häusler R, Baur B, Scharte J, Teichmann T, Eicks M, Fischer K, Flügge U-I, Schubert S, Weber A, Fischer K. Plastidic metabolite transporters and their physiological functions in the inducible crassulacean acid metabolism plant Mesembryanthemum crystallinum. The Plant Journal (2000) 24:285–296.[CrossRef][Web of Science][Medline]

Hoagland D, Arnon D. The water culture method of growing plants without soil. California Agricultural Experiment Station Circular No. 347 (1950) Berkeley, CA: College of Agriculture, University of California.

Hoch G, Peterbauer T, Richter A. Purification and characterization of stachyose synthase from lentil (Lens culinaris) seeds: galactopinitol and stachyose synthesis. Archives of Biochemistry and Biophysics (1999) 366:75–81.[CrossRef][Web of Science][Medline]

Holtum J, Winter K. Activities of enzymes of carbon metabolism during the induction of crassulacean acid metabolism in Mesembryanthemum crystallinum L. Planta (1982) 155:8–16.[CrossRef][Web of Science]

Hurst A, Grams T, Ratajczak R. Effects of salinity, high irradiance, ozone, and ethylene on mode of photosynthesis, oxidative stress and oxidative damage in the C3/CAM intermediate plant Mesembryanthemum crystallinum. Plant, Cell and Environment (2004) 27:187–197.[CrossRef]

Irizarry R, Hobbs B, Collin F, Beazer-Barclay Y, Antonellis K, Scherf U, Speed T. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics (2003) 4:249–264.[Abstract]

Johnson W, Rabinovic A, Li C. Adjusting batch effects in microarray expression data using Empirical Bayes methods. Biostatistics (2007) 8:118–127.[Abstract/Free Full Text]

Kluge C, Lahr J, Hanitzsch M, Bolte S, Golldack D, Dietz K. New insight into the structure and regulation of the plant vacuolar H+-ATPase. Journal of Bioenergetics and Biomembranes (2003a) 35:377–388.[CrossRef][Web of Science][Medline]

Kluge C, Lamkemeyer P, Tavakoli N, Golldack D, Kandlbinder A, Dietz K. cDNA cloning of 12 subunits of the V-type ATPase from Mesembryanthemum crystallinum and their expression under stress. Molecular Membrane Biology (2003b) 20:171–183.[CrossRef][Web of Science][Medline]

Kore-eda S, Cushman M, Akselrod I, Bufford D, Fredrickson M, Clark E, Cushman J. Transcript profiling of salinity stress responses by large-scale expressed sequence tag analysis in Mesembryanthemum crystallinum. Gene (2004) 341:83–92.[CrossRef][Web of Science][Medline]

Kore-eda S, Noake C, Ohishi M, Ohnishi J, Cushman J. Transcriptional profiles of organellar metabolite transporters during induction of Crassulacean acid metabolism in Mesembryanthemum crystallinum. Functional Plant Biology (2005) 32:451–466.[CrossRef][Web of Science]

Kreps J, Wu Y, Chang H-S, Zhu T, Wang X, Harper J. Transcriptome changes for Arabidopsis in response to salt, osmotic and cold stress. Plant Physiology (2002) 130:2129–2141.[Abstract/Free Full Text]

Maxwell K, Borland A, Haslam R, Helliker B, Roberts A, Griffiths H. Modulation of ribulose-1,5-bisphosphate carboxylase activity during the diurnal phases of the crassulacean acid metabolism plant Kalanchoë daigremontiana. Plant Physiology (1999) 121:849–856.[Abstract/Free Full Text]

Mikkelsen R, Mutenda K, Mant A, Schurmann P, Blennow A. Alpha-glucan, water dikinase (GWD): a plastidic enzyme with redox-regulated and coordinated catalytic activity and binding affinity. In: Proceedings of the National Academy of Sciences, USA (2005) 102:1785–1790.[Abstract/Free Full Text]

Munneke B, Schlauch K, Simonsen K, Beavis W, Doerge R. Adding confidence to gene expression clustering. Genetics (2005) 170:2003–2011.[Abstract/Free Full Text]

Nimmo H. The regulation of phosphoenolpyruvate carboxylase in CAM plants. Trends in Plant Science (2000) 5:75–80.[CrossRef][Web of Science][Medline]

Nobel P. High productivities of certain agronomic CAM species. In: Crassulacean acid metabolism. Biochemistry, ecophysiology and evolution—Winter K, Smith JAC, eds. (1996) Berlin: Springer-Verlag. 255–265.

Nuwaysir E, Huang W, Albert T, et al. Gene expression analysis using oligonucleotide arrays produced by maskless photolithography. Genome Research (2002) 12:1749–1755.[Abstract/Free Full Text]

Ostrem J, Vernon D, Bohnert H. Increased expression of a gene coding for NAD-glyceraldehyde-3-phosphate dehydrogenase during the transition from C3 photosynthesis to crassulacean acid metabolism in Mesembryanthemum crystallinum. Journal of Biological Chemistry (1990) 265:3497–3502.[Abstract/Free Full Text]

Paul M, Loos K, Stitt M, Ziegler P. Starch-degrading enzymes during the induction of CAM in Mesembryanthemum crystallinum. Plant, Cell and Environment (1993) 16:531–538.[CrossRef]

Rockel B, Blasius B, Beck F, Ratajcek R, Lüttge U. Endogenous oscillations of the transcript amounts of subunit C of the V-ATPase of Mesembryanthemum crystallinum with harmonic frequency resonances under continuous illumination. Cell and Molecular Biology Letters (1997) 2:69–76.

Rozen S, Skaletsky H. Primer3 on the WWW for general users and for biologist programmers. Methods in Molecular Biology (2000) 132:365–386.[Medline]

Ruepp A, Zollner A, Maier D, et al. The FunCat, a functional annotation scheme for systematic classification of proteins from whole genomes. Nucleic Acids Research (2004) 32:5539–5545.[Abstract/Free Full Text]

Sambrook J, Fritsch E, Maniatis T. Molecular cloning: a laboratory manual (1989) Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press.

Silvera K, Santiago L, Winter K. Distribution of Crassulacean acid metabolism in orchids of Panama: evidence of selection of weak and strong modes. Functional Plant Biology (2005) 32:397–407.[CrossRef][Web of Science]

Singh-Gasson S, Green R, Yue Y, Nelson C, Blattner F, Sussman M, Cerrina F. Maskless fabrication of light-directed oligonucleotide microarrays using a digital micromirror array. Nature Biotechnology (1999) 17:974–978.[CrossRef][Web of Science][Medline]

Sondergaard T, Schulz A, Palmgren M. Energization of transport processes in plants. Roles of the plasma membrane H+-ATPase. Plant Physiology (2004) 136:2475–2482.[Free Full Text]

Taybi T, Cushman J. Signaling events leading to Crassulacean acid metabolism (CAM) induction in the common ice plant, Mesembryanthemum crystallinum. Plant Physiology (1999) 121:545–555.[Abstract/Free Full Text]

Taybi T, Cushman J. Abscisic acid signaling and protein synthesis requirements for CAM induction in the common ice plant. Journal of Plant Physiology (2002) 159:1235–1243.[CrossRef][Web of Science]

Taybi T, Patil S, Chollet R, Cushman J. A minimal Ser/Thr protein kinase circadianly regulates phosphoenolpyruvate carboxylase activity in CAM-induced leaves of Mesembryanthemum crystallinum. Plant Physiology (2000) 123:1471–1482.[Abstract/Free Full Text]

Tetlow I, Wait R, Lu Z, Akkasaeng R, Bowsher C, Esposito S, Kosar-Hashemi B, Morell M, Emes M. Protein phosphorylation in amyloplasts regulates starch branching enzyme activity and protein–protein interactions. The Plant Cell (2004) 16:694–708.[Abstract/Free Full Text]

Vernon D, Bohnert H. Increased expression of a myo-inositol methyl transferase in Mesembryanthemum crystallinum is part of a stress response distinct from crassulacean acid metabolism induction. Plant Physiology (1992) 99:1695–1698.[Abstract/Free Full Text]

Wattebled F, Dong Y, Dumez S, et al. Mutants of Arabidopsis lacking a chloroplastic isoamylase accumulate phytoglycogen and an abnormal form of amylopectin. Plant Physiology (2005) 138:184–195.[Abstract/Free Full Text]

Winter K, Holtum J. The effects of salinity, crassulacean acid metabolism and plant age on the carbon isotope composition of Mesembryanthemum crystallinum L. a halophytic C3–CAM species. Planta (2005) 222:201–209.[CrossRef][Web of Science][Medline]

Winter K, Holtum J. Environment or development? Lifetime net CO2 exchange and control of the expression of crassulacean acid metabolism in Mesembryanthemum crystallinum. Plant Physiology (2007) 143:98–107.[Abstract/Free Full Text]

Wyka T, Lüttge U. Contribution of C3 carboxylation to the circadian rhythm of carbon dioxide uptake in a Crassulacean acid metabolism plant Kalanchoë daigremontiana. Journal of Experimental Botany (2003) 54:1471–1479.[Abstract/Free Full Text]

Yu T, Kofler H, Hausler R, et al. The Arabidopsis sex1 mutant is defective in the R1 protein, a general regulator of starch degradation in plants, and not in the chloroplast hexose transporter. The Plant Cell (2001) 13:1907–1918.[Abstract/Free Full Text]


Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Supplementary Material
Right arrow All Versions of this Article:
59/7/1875    most recent
ern008v1
Right arrow E-letters: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when E-letters are posted
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Cushman, J. C.
Right arrow Articles by Schlauch, K. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Cushman, J. C.
Right arrow Articles by Schlauch, K. A.
Agricola
Right arrow Articles by Cushman, J. C.
Right arrow Articles by Schlauch, K. A.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?