Author for correspondence: Andreas M. Fischer Tel: +1 406 9945908 Fax: +1 406 9947600 Email: firstname.lastname@example.org
• Leaf senescence can be described as the dismantling of cellular components during a specific time interval before cell death. This has the effect of remobilizing N in the form of amino acids that can be relocalized to developing seeds. High levels of carbohydrates have previously been shown to promote the onset of the senescence process.
• Carbohydrate accumulation in barley (Hordeum vulgare) plants was induced experimentally by steam-girdling at the leaf base, occluding the phloem, and gene regulation under these conditions was investigated using the Affymetrix Barley GeneChip array and quantitative real-time reverse transcriptase polymerase chain reaction (qRT-PCR).
• Transcript levels of plastidial (aminopeptidases, cnd41) and vacuolar (thiol and serine) proteases clearly increase in girdled leaves. Of special interest are cnd41, a plastidial aspartyl peptidase that has been implicated in Rubisco degradation in tobacco; and cp-mIII, a highly upregulated carboxypeptidase. SAG12, hexokinases and other senescence-specific genes are also upregulated under these conditions.
• Applying a genomic approach to the innovative experimental system described here significantly enhances our knowledge of leaf proteolysis and whole-plant N recycling.
Leaf senescence is characterized by nutrient remobilization to developing seeds of annual plants, or surviving organs of perennial species. It is a genetically controlled process initiated at a specific point in the life of the leaf that causes an ordered degradation of cell constituents (Masclaux et al., 2000). Of major interest is the remobilization of nitrogen during leaf senescence, considering that the bulk of N for seed storage protein formation in crop plants is exported from leaves (Peoples & Dalling, 1988), and that leaf N is located mainly in chloroplasts as Rubisco (ribulose-1,5-bisphosphate carboxylase/oxygenase; Peoples & Dalling, 1988; Hörtensteiner & Feller, 2002). Because Rubisco and other proteins must be degraded before remobilization, it is important to understand the role proteolysis plays in relation to leaf senescence. Kato et al. (2004, 2005) determined that the DNA-binding protease CND41 is strongly implicated in the degradation of Rubisco in senescing tobacco leaves. However, it was found that CND41 would not act on Rubisco unless its structure was denatured, suggesting that normally active Rubisco in chloroplasts was not affected until leaf senescence was under way (Kato et al., 2004). Other proteases have been associated with proteolysis and N remobilization (Gepstein et al., 2003; Schaller, 2004; Buchanan-Wollaston et al., 2005). A previous study from our laboratory has indicated that one or several carboxypeptidases are involved in leaf N remobilization (Yang et al., 2004). It has been postulated that the various stages of leaf cell component degradation require different signals, and that proteolysis is generally enhanced after induction of senescence (Yoshida, 2003).
We have previously used a model system allowing for the manipulation of leaf carbohydrates in order to investigate the processes important during early phases of senescence, such as plastidial protein degradation. In this study, sugar accumulation was induced in mature barley (Hordeum vulgare) leaves by the interruption of sieve tubes at the base of the leaf lamina by ‘steam-girdling’ (Parrott et al., 2005). As Jongebloed et al. (2004) have shown that naturally occurring sieve tube occlusion and carbohydrate accumulation is associated with chlorophyll degradation, it appears likely that our steam-girdling system enhances a naturally occurring process, thus facilitating its experimental analysis. Girdling has also been shown to increase the levels of carbohydrates in spinach leaves, accompanied by a decrease in Rubisco transcript levels and chlorophyll and protein degradation (Krapp & Stitt, 1995). We noted accelerated chlorophyll degradation and net proteolysis as confirmation of successful senescence induction, while protease activity assays indicated an early and strong induction of both endo- and exopeptidases (Parrott et al., 2005). Here we expand on those findings using the Affymetrix Barley1 genome array and quantitative real-time RT-PCR (qRT-PCR) to determine upregulated protease genes (and other genes potentially involved in senescence regulation and N recycling) in girdled leaves as compared with untreated control leaves. Additionally, we examine leaf carbohydrate accumulation and Rubisco degradation in greater detail, and determine biological processes affected by carbohydrate accumulation in barley leaves. We also compare differential gene regulation in the leaf-girdling system with that in naturally senescing barley leaves. Using this functional genomic approach to studying degradation of major leaf proteins may considerably improve our understanding of whole-plant N recycling.
Materials and Methods
Plant material, growth conditions and treatments
Barley plants (Hordeum vulgare L. cv. Harrington) were grown and treated as described by Parrott et al. (2005). Briefly, mature second leaves of 14-d-old seedlings were treated with a steam-heated (95°C) hypodermic needle for 5 s at the base of the leaf lamina to interrupt the phloem (girdled leaves). Untreated leaves (control leaves) were compared with girdled leaves or leaves that were partially girdled (shift-girdled: two opposing phloem interruptions spanning half the lamina separated by approx. 1 cm; Feller & Fischer, 1994). For naturally senescing leaf analysis, barley seeds (H. vulgare cv. Lewis) were grown three to a 1-gallon pot in potting soil in a glasshouse of the Montana State University Plant Growth Center (Bozeman, MT, USA) between February and April 2006, with a 22 : 18°C day : night temperature cycle. When necessary, days were extended to a 16-h photoperiod, using Son-Agro 430 W high-pressure sodium lamps (Philips, Somerset, NJ, USA). Seedlings were fertilized once per week until anthesis with Peter's Professional General Purpose fertilizer (250 ml per pot; 4 g l−1; Scotts-Sierra Horticultural Products Company, Marysville, OH, USA). Plants were tagged at anthesis, and second leaves were harvested at 7, 14 and 21 d past anthesis (dpa). Leaves were immediately frozen in liquid N and stored at –80°C for RNA extraction and analysis.
Leaf samples for fructose, glucose, sucrose and fructan analysis were ground to a fine powder in liquid N, using a mortar and pestle. Soluble sugars were extracted from 50 mg liquid N powder in 5 ml H2O heated to 60°C. Samples were equilibrated to 60°C in a water bath for 20 min, vortexed several times, and cooled to 25°C. Fructose and glucose were assayed using a Sigma FA-20 assay kit (Sigma, St Louis, MO, USA) according to the supplied procedure. For each sample, 25 µl extract was used and absorbance at 340 nm was compared with standard curve values (0–25 µg fructose or glucose). Sucrose was assayed using a sucrose assay kit (K-SUCGL, Megazyme Wicklow, Ireland) following the supplied procedure and using 100 µl of both extract and reagent. Absorbance was read at 510 nm against a glucose standard, and sucrose concentrations were calculated based on the difference in glucose concentrations before and after enzymatic inversion (using β-fructosidase) of sucrose to glucose and fructose. Fructans were assayed using a fructan assay kit (K-Fruc, Megazyme) following procedures and calculations provided by the manufacturer. For each sample, 200 µl extract was used and fructan concentrations were determined by comparing absorbance at 410 nm for samples and standards. All assays were performed on three biological replications for 0-, 4- and 8-d control, girdled and shift-girdled treatments.
Two antipeptide antibodies, directed against sequences close to the N- and C-terminus of the large subunit of wheat (Triticum aestivum L.) Rubisco, were used for this study. The N-terminal sequence used was VGFKAGVKDYKLTYYTPEYE (starting with position 11 of the wheat and barley Rubisco large chain precursor). The barley sequence (compare GenBank accessions AY137456 and AY328025) is identical, with the exception of peptide position four (Q instead of K in barley). This peptide was commercially synthesized, conjugated to keyhole limpet hemocyanin, and used for the immunization of two rabbits using standard procedures (BioWORLD, Dublin, OH, USA). The C-terminal sequence used was CVQARNEGRDLAREG (starting with position 427 of both the wheat and barley Rubisco large chain precursor). This peptide sequence is identical in both species (see GenBank accessions AY137456 and AY328025). This peptide was again commercially synthesized, conjugated to keyhole limpet hemocyanin, and used for the immunization of two rabbits (Genosphere Biotechnologies, Paris, France). The reactivity of crude serum against barley Rubisco was confirmed on immunoblots, using different concentrations of both protein and antibodies.
Denaturing electrophoresis and immunoblotting
Proteins were extracted from liquid N-ground barley leaves as described by Parrott et al. (2005), using 25 mm Tris–HCl pH 7.5 containing 1% (w/v) insoluble polyvinylpolypyrrolidone and 0.1% (v/v) β-mercaptoethanol as an extraction buffer. Crude samples were centrifuged (10 min, 20 000 g, 4°C), and supernatants were used for the analysis of soluble proteins. Pellets were washed twice in extraction buffer without polyvinylpolypyrrolidone (PVPP), resuspended in extraction buffer without PVPP, and utilized for the analysis of membrane proteins. SDS-polyacrylamide gel electrophoresis was performed (Laemmli, 1970) with 10 µl extracted soluble and membrane proteins (corresponding to approx. 2% of a leaf or approx. 20 µg soluble protein at day 0) loaded in each lane, and visualized by staining with Coomassie Brilliant Blue R-250. Immunoblotting was performed as described by Fischer et al. (1999) using samples from SDS-PAGE. Each lane was loaded with 15 µl extracted soluble protein equating to approx. 3% of a leaf (approx. 30 µg protein at day 0). Proteins were electrotransferred to nitrocellulose, and membranes were blocked and probed with 1 : 5000 diluted antibodies specific for the N- and C-terminus of the large subunit of Rubisco. Secondary antibodies were a goat antirabbit IgG, horseradish peroxidase conjugate (Bio-Rad, Hercules, CA, USA) diluted 1 : 5000 before use. Blots were incubated with chemiluminescent substrate (SuperSignal West Pico Chemiluminescent Substrate, Pierce, Rockford, IL, USA) for 10 min before exposure to X-ray film.
RNA extraction for microarray and quantitative real-time RT-PCR analysis
Leaf samples frozen at –80°C were placed in a liquid N-cooled mortar and pestle, and ground to a fine powder. RNA was extracted from 100 mg liquid N powder utilizing the RNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA) following the supplied procedure and using buffer RLC. Three individual extractions for each sample were pooled and EtOH-precipitated overnight at –80°C using glycoblue (Ambion, Austin, TX, USA) as a coprecipitant. Samples were pelleted at 15 000 g for 20 min at 4°C, washed twice with 70% EtOH, dried and resuspended in 15 µl sterile H2O. Samples were incubated with recombinant RNAsin (Promega, Madison, WI, USA) at 50°C for 15 min, then RNA concentrations were determined at 260 nm using a spectrophotometer (Spectra Max Plus, Molecular Devices, Sunnyvale, CA, USA). RNA quality was tested using an RNA 6000 nano assay (Bioanalyser 2100, Agilent Technologies, Palo Alto, CA, USA). RNA used for qRT-PCR was extracted as above, with the addition of an on-column DNase digestion (RNase-free DNase kit, Qiagen) during the RNeasy extraction procedure.
Growth chamber-grown barley leaf samples were used for microarray analysis. Three biological replications were tested for each of the treatments (control, girdled and shift-girdled) at each time point (0, 4 and 8 d), resulting in 21 samples (three replications of seven samples) each hybridized to a corresponding chip (GeneChip Barley1 22 k genome array, Affymetrix, Santa Clara, CA USA). Chip preparation, hybridization and analysis were performed at the Montana State University Functional Genomics Core Facility (Bozeman, MT, USA). Procedures for reverse transcription, cRNA synthesis and labelling, chip hybridization, washing, staining and scanning were performed according to manufacturer's instructions (Affymetrix). For the Barley1 genome array, 10 µg cRNA was hybridized to each chip. Raw data were analysed using genespring GX ver. 7.3 (Agilent Technologies). Robust multichip averaging (Irizarry et al., 2003), with per chip and per gene median polishing, was used to normalize CEL file data generated from the 21 chips. Data were filtered first on a baseline raw signal intensity threshold of 50 in at least one replicate set equivalent (RSE; one set of three replications for one treatment), then filtered on present or marginal calls in at least one RSE, leading to a reduction in the number of genes analysed from 22 840 to 13 654. Data were filtered on a fold change of 1.5 in any comparison (treatment vs time point) and analysed using a two-way anova with Student's t-test and a Benjamini and Hochberg false discovery rate of 0.05 between time point (0, 4 or 8 d) and treatment (control, girdled or shift-girdled leaves). Comparisons were run for each time/treatment, and gene lists were generated. Hierarchical trees and line diagrams were produced from these lists, as were candidate genes for qRT-PCR microarray validation. Gene ontology enrichment for biological processes was determined using the genespring software, and lists of genes up- and downregulated in 4- and 8-d-girdled leaves were compared with 4-d control leaves.
Quantitative real-time reverse transcriptase-PCR (qRT-PCR) was used to validate microarray data and to determine differential gene expression in naturally senescing leaves. First-strand cDNA synthesis and subsequent PCR were performed using the Superscript III Platinum Two-Step qRT-PCR Kit with SYBR Green (Invitrogen, Carlsbad, CA, USA) following the manufacturer's procedure with 1 µg sample RNA. Primers were designed using the oligoperfect primer design software (Invitrogen) to have a melting temperature between 58 and 60°C, with no more than a 2°C difference between forward and reverse primers. PCR products were between 75 and 150 bp and had a 40–80% GC content. Standards for high and low gene expression were designed by screening the raw Barley1 GeneChip data and identifying genes with stable expression levels between treatments and time points. Two additional genes with stable expression levels were chosen as controls for direct inclusion in the resulting figure. The primer sequences for standards and candidate genes are as follows: high standard, Contig97_s_at, translationally controlled tumour protein homologue (forward 5′-cttggcagtgaggttcttga-3′; reverse 5′-ttgacaccttccgtcttcag-3′); low standard, contig13298_at, unknown function (forward 5′-atctcagcctcgcagaagag-3′; reverse 5′-aagctcgatccaagaaggaa-3′); contig10941_at, SAG12 cysteine protease (forward 5′-acgaggagcgagctatcatt-3′; reverse 5′-gaccattgtacacgccattc-3′); 12029_s_at, subtilase family protein (forward 5′-cagaaaacgcaggagtacgt-3′; reverse 5′-atcgacccgaacgtgtactt-3′); contig6013_at, CND41-like chloroplast nucleoid DNA-binding protein (forward 5′-tcttcgggttcaagtccaa-3′; reverse 5′-cctagctagccacccttcat-3′); contig6202_at, proline iminopeptidase (forward 5′-ttaaggtggttccagatgca-3′; reverse 5′-taggctgctcatttcctcaa-3′); contig4312_s_at neutral leucine aminopeptidase preprotein (forward 5′-taccagcgtcgctattgttc-3′; reverse 5′-tcgactcagccttgaatctg-3′); contig2779_at, aspartic endopeptidase (forward 5′-ggtgagcctcttgttgatga-3′; reverse 5′-tgtgcttcatcccatactgc-3′); contig9006_at, papain-like cysteine protease (forward 5′ -aaaggtggcaaggattattgg-3′; reverse 5′-tgccacagatacctgacgat-3′); contig600_at, serine carboxypeptidase III precursor (CP-MIII) (forward 5′-ccaagcaatgctcacagact-3′; reverse 5′-gagtggacccaccttgagtt-3′); control with stable expression, contig8733_at, 15.9 kDa subunit RNA polymerase II (forward 5′-ctcgtcgagttcgagctatg-3′; reverse 5′-atgagggagaggtcgttgag-3′); second control with stable expression, HW03O14 µ_s_at, Sec61p alpha subunit (forward 5′-ctggccattctatccctgtt-3′; reverse 5′-aacggattcgccagaatatc-3′). PCR was carried out in a RotorGene 3000 thermal cycler (Corbett Life Science, Mortlake, NSW, Australia). Each 20-µl PCR reaction contained 2 µl cDNA and 0.5 µm of each primer. The cycling profile consisted of 50°C for 2 min, 95°C for 2 min, 40 cycles of 94°C for 5 s, 55°C for 15 s and 72°C for 15 s. A melt curve from 72 to 95°C was run following the PCR cycling. For all qRT-PCR assays, the expression levels of target genes were normalized to the levels of the two genes chosen as standards (high, Contig97_s_at; low, contig13298_at) utilizing a standard curve (rotor-gene 6 software (Build 27), Corbett Life Science). The standard curves were then used to determine relative expression values for each gene after qRT-PCR analysis of each sample. Correct amplicon size for each primer set was verified using gel electrophoresis.
Leaf girdling leads to carbohydrate accumulation and induces leaf senescence
Fructose, glucose, sucrose and fructan levels were analysed in leaves of barley plants grown in a growth chamber in order to determine the effect of girdling and shift-girdling on carbohydrate accumulation, as compared with untreated (control) leaves. Fructose and glucose accumulated two- to fourfold higher in girdled leaves after 4 d, and four- to eightfold after 8 d, when compared with control leaves at day 0, while levels in control and shift-girdled leaves decreased from 0 to 8 d (Fig. 1a,b). Sucrose accumulated 17- to 21-fold higher in girdled leaves than in control and shift-girdled leaves from 0 to 8 d, while only modest increases were seen in control and shift-girdled leaves (Fig. 1c). Accumulation of fructan followed a similar pattern, but increased in girdled leaves 15- to 20-fold higher than either control or shift-girdled leaves, both of which had negligible fructan levels (Fig. 1d).
Increased carbohydrate levels were associated with a profound visible effect, clearly showing senescence in girdled leaves when compared with control and shift-girdled leaves. Leaves that were sampled and photographed immediately after girdling or shift-girdling (day 0) were indistinguishable from untreated control leaves (Fig. 2a). After 4 d, girdled leaves showed a decrease in chlorophyll, as noted by the lightening of the lamina colour, compared with shift-girdled and control leaves (Fig. 2b). At 8 d, girdled leaves were pale green (Fig. 2c) and after 12 d, girdled leaves were yellow and becoming necrotic at the leaf tip (Fig. 2d). Shift-girdled leaves showed no visible difference when compared with untreated control leaves at 8 and 12 d.
To further correlate carbohydrate accumulation with leaf senescence, both membrane (Fig. 3a) and soluble (Fig. 3b) proteins were analysed by SDS-PAGE. Soluble proteins remained quite stable in all samples between 0 and 8 d, but some decrease of protein bands analysed with the membrane fraction was visible over the same time span in most lanes. A prominent double band, probably derived from the Rubisco large subunit, was visible in the membrane fraction, but the significance of this observation, which was seen in all control and treated samples, has not been investigated further. At 12 d, a decrease of the Rubisco large subunit was evident in the soluble protein fraction of girdled leaves; these data correlate well with findings from previously published, similar gels, in which only approx. 34% (based on densitometric analysis, compared with day 0) of the Rubisco large subunit was left after 12 d in samples from girdles leaves, compared with 60–80% in control and shift-girdled leaves (Parrott et al., 2005). Data shown in Fig. 3a suggest an even more substantial degradation of membrane, including light-harvesting chlorophyll-binding proteins after 12 d in girdled leaves, confirming that visual senescence symptoms are accompanied by accelerated/enhanced net proteolysis in our experimental system.
To investigate further the pattern of Rubisco degradation, the soluble protein fraction was investigated using antibodies specific to the N- (Fig. 3c) and C-terminus (Fig. 3d) of its large subunit. In addition to strongly visualizing the large subunit, both antibodies reacted with several additional protein bands. Among these, a band of approx. 70 kDa (above large subunit in Fig. 3c,d) is conspicuous. The significance of the this band is not clear but, based on the fact that it was detected by both antibodies, it is unlikely to be caused by a lack of antibody specificity. The N-terminal-specific antibody showed degradation of the large Rubisco subunit at 12 d in the girdled sample, and also showed specificity for two fragments of 31.2 and 21.6 kDa, which were especially prominent in girdled samples (Fig. 3c). The C-terminal-specific antibody yielded similar results to the N-terminal antibody (Fig. 3d), but with fragments visible at 46.0, 40.0 and 34.7 kDa. The latter band was again more conspicuous in (4- and 8-d) girdled samples. It should be noted that the sum of the weights of the N-terminal 21.6-kDa and C-terminal 34.7-kDa fragments is close to the weight of the intact Rubisco large subunit, suggesting that these fragments arise from a single cleavage event. No corresponding C-terminal fragment to the N-terminal 31.2 kDa band was detected in our experiment. This finding could be explained either by a lack of sensitivity, or by the rapid further degradation of such a fragment with a predicted weight of approx. 25 kDa. It should be noted that our experimental approach does not exclude the possibility that detected fragments were formed in vitro during sample preparation (as opposed to being derived from – more interesting –in vivo proteolysis). However, together, SDS-PAGE and immunoblot data clearly confirm that carbohydrate accumulation is correlated with enhanced degradation of both soluble proteins (Rubisco) and membrane proteins (e.g. light-harvesting chlorophyll-binding proteins, LHCP).
Differences in gene expression in girdled barley leaves
Major differences in expression levels of numerous genes in girdled leaves when compared with control and shift-girdled leaves (Fig. 4) were determined by utilizing Affymetrix Barley1 microarray gene chips and qRT-PCR. For simplicity, we use the term ‘genes’ when referring to gene chip probe sets and contigs, as well as qRT-PCR probe sets. Three biological replicates of control, girdled and shift-girdled leaf samples were harvested 0, 4 and 8 d after treatment. Labelled RNA from the resulting 21 samples was hybridized to gene chips, stained and analysed as described under Materials and Methods. Two-way anova found 6582 genes to be significant for either harvest day or treatment. Comparisons between girdled leaves and control and shift-girdled leaves assayed at 4 d revealed 5613 genes that were 1.5-fold or more up- or downregulated, while the same comparison at 8 d yielded 6111 genes 1.5-fold or more up- or downregulated (Fig. 4). We have previously shown that the shift-girdling control indicates that observed effects in girdled leaves are caused by the girdling treatment per se (carbohydrate accumulation), and not to wounding (Feller & Fischer, 1994; Parrott et al., 2005). Microarray analysis confirms this interpretation, as gene-expression patterns in control and shift-girdled leaves are very similar (Fig. 4). Complete gene lists are shown in Tables S1 and S2 in Supplementary material.
Biological processes that were influenced by upregulated genes in girdled leaves at 4 and 8 d are shown in Fig. 5. Gene ontologies were determined using genespring software, and the lists of biological processes were created based on the comparisons made between girdled leaf samples at 4 and 8 d vs control samples at 4 and 8 d. The biological processes listed in Fig. 5 (carbohydrate metabolism, other metabolism, electron transport, energy pathways, cell organization and biogenesis, cell growth and/or maintenance, cell death, protein biosynthesis protein transport, development, regulation of gene expression) represent the genes overlapping with gene ontology biological process (GO BP) classifications with a random overlap P value of 0.01 or better. Detailed lists for processes influenced by both up- and downregulated genes in 4- and 8-d girdled leaves are shown in Table S3.
Some differences in GO BP between 4- and 8-d girdled leaf samples (upregulated genes) were seen in many of the GO BP categories (Fig. 5; Table S3). In many of the GO BP classifications, and elsewhere in our data, there were genes that may have implications in leaf senescence, and in most instances there were increases in the number of these genes active at 8 vs 4 d (Tables S1, S2, S3). Enhanced carbohydrate metabolism was correlated with increases in senescence-associated genes, with increased expression of sugar-related epimerases, decarboxylases and kinases noted. Some transcription factors were seen in the electron transport category, and amino acid transporters and channel proteins were seen in the energy pathways and cell organization categories. Many more transcription factor classes were observed in the cell growth and maintenance category, while numerous additional protein transporters were present in the protein transport category. Finally, many proteases, senescence-associated proteins and transcription factors (MYB, bZIP, HIN, WRKY) were seen in the regulation of gene expression category.
Of the 5613 genes 1.5 fold or more up- or down regulated in girdled leaves at 4 d, 109 protease genes were upregulated while 52 were downregulated as compared with 4-d control leaves (Table S4). Similarly, of the 6111 genes 1.5 fold or more up- or down regulated in girdled leaves at 8 d, 103 protease genes were upregulated and 74 were downregulated as compared with 8-d control leaves (Table S5). While the numbers of upregulated protease genes (products of which may be functionally important for N remobilization) are higher at both time points, our data suggest that leaf senescence, at least in our system, is associated with a shift in the protease complement, with the expression of a considerable number of genes decreasing. Closer examination of the 50 protease genes with twofold or more upregulation, using hierarchical clustering of treatment conditions, showed that shift girdling produced little change in gene expression in either 4- or 8-d samples as compared with 0- and 4- or 8-d controls, whereas the impact of girdling was profound (Fig. 6). A similar pattern was evident when all differentially expressed genes were included in the analysis (data not shown). Girdled treatments at 4 and 8 d were clustered separately from all control and shift-girdled treatments, resulting in a progression from mature to senesced leaves. Clustering on the basis of individual gene-expression patterns across all conditions revealed three distinct clusters. The first cluster had low expression levels in 0-d and high expression levels in 4- and 8-d girdled samples; the second had higher expression levels in 0-d (as compared with the first cluster) and high levels in 4- and 8-d girdled samples; and the third had high expression levels in 0-d, decreasing expression in 4- and 8-d control and shift-girdled samples, and high expression in 4- and 8-d girdled samples. Within each of these clusters, genes were grouped according to similar expression patterns, resulting in groupings of either genes of (potentially) similar function, or groups of probe sets each describing the same gene on the gene chip array (e.g. for serine carboxypeptidase III).
From the list of 50 upregulated protease genes from the gene/condition tree (Fig. 6), eight were chosen as candidates for further analysis and gene chip validation using qRT-PCR. Table 1 shows the eight candidate genes that were chosen based on their representation of different catalytic classes of proteases (cysteine, aspartic, serine and metalloproteases), their high upregulation in 4- and 8-d girdled treatments relative to 0-, 4- and 8-d control and shift-girdled treatments, and their possible implication in senescence-related protein degradation (SAG12, CND41).
Table 1. Relative expression data for candidate genes in 4- and 8-d girdled barley (Hordeum vulgare) leaves
Affymetrix probe set identifier, description and fold-change data are shown. Selected genes represent various catalytic classes of proteases upregulated in girdled leaves at 4 or 8 d vs control and/or shift-girdled leaves at 0, 4 or 8 d.
c, control; g, girdled; s, shift-girdled.
subtilase family protein
cnd41-like chloroplast nucleoid DNA-binding protein
neutral leucine aminopeptidase preprotein
papain-like cysteine peptidase
serine carboxypeptidase III precursor (CP-MIII)
Fold-change data from the Barley1 genome array (Table 1) and relative expression levels from qRT-PCR (Fig. 7) of the eight protease genes show similar expression patterns. Data from the gene chips for each of the eight genes show fold changes increased in 4- or 8-d girdled leaves when compared with 4- or 8-d control or shift-girdled leaves and 0-d controls. The gene chip data were validated by qRT-PCR with primers specific for each of the eight protease genes, and while absolute values differ between gene chip fold-change values and qRT-PCR relative expression values, the pattern of gene expression was comparable. However, while there were similarities in gene regulation between the two methods within either 4- or 8-d time points, there was some variation of expression levels seen between 0-, 4- and 8-d time points. For instance, gene chip fold-change values for contig10941_at (SAG12 protein) at 4 d in girdled leaves were slightly lower than the values for 8 d when compared with 0-, 4- and 8-d controls and shift-girdled leaves, but qRT-PCR values show an increase in relative expression from 4 to 8 d in girdled leaves. Contig6013_at (CND41-like chloroplast nucleoid DNA-binding protein) showed higher expression at 0-d expression levels using qRT-PCR than at 4 d in girdled leaves.
Protease genes are upregulated in naturally senescing barley leaves
In order to validate our model system against natural leaf senescence, gene expression of the eight candidate genes was determined in naturally senescing barley leaves. Second leaves (below the head) of naturally senescing barley plants showed increased protease gene-expression levels as determined by qRT-PCR analysis. The eight candidate genes selected for microarray validation (Table 1) were used to determine expression levels of these genes in naturally senescing second leaves ranging in age from 7 to 21 d postanthesis (dpa). Expression levels of these genes in naturally senescing leaves (Fig. 8) increase with the progression of natural leaf senescence, indicating that our experimental approach correctly identifies genes important for this process (Table 1; Fig. 7). However, one notable exception was contig6013_at (CND41-like chloroplast nucleoid DNA-binding protein; Fig. 8c), which decreased considerably at 21 d. Two controls with (based on our gene chip data) stable expression levels were included in Fig. 8 to demonstrate that observed upregulation of protease genes is specific.
Leaf girdling leads to carbohydrate accumulation and senescence induction in barley leaves
Sugar accumulation or carbon feast has been suggested to induce leaf senescence in a variety of plant species (Feller & Fischer, 1994; Koch, 1996; Wingler et al., 1998, 2006; Masclaux et al., 2000; Ono et al., 2001; Pourtau et al., 2004, 2006; Parrott et al., 2005). We have demonstrated previously that leaf senescence can be induced experimentally by girdling (Parrott et al., 2005). For this study, carbohydrate accumulation was induced in second (mature) leaves of 2-wk-old barley seedlings by stem-girdling the base of the leaf blade, leading to selective phloem interruption. Leaves with incompletely interrupted phloem (shift-girdled) were used as controls, imposing the same wounding stress, but not leading to carbon feast conditions. For both girdled and shift-girdled treatments, the xylem is left intact with no wilting of the leaves observed. As expected, substantial carbohydrate accumulation is seen in girdled as compared with control and shift-girdled leaves. Fructose, glucose, sucrose and fructan increased from 0 to 8 d in the girdled samples, while fructose and glucose decreased in the controls and shift-girdled leaves (Fig. 1). At the same time, chlorophyll was degraded as the leaves became visibly more yellow between 0 and 8 d (Fig. 2). These data extend our previous findings (Parrott et al., 2005) and, while they do not demonstrate a causal relationship between carbohydrate accumulation and senescence induction after leaf girdling, such a connection appears likely based on recent publications in the field (cited above). Furthermore, SDS-PAGE and immunoblot analysis indicated enhanced protein degradation in girdled leaf samples. While soluble and, even more, membrane protein (such as LHCP) levels detected with Coomassie Blue decreased by 12 d posttreatment (Fig. 3a,b), Rubisco antibodies demonstrated enhanced degradation of this major plastidial protein at 4 and 8 d (Fig. 3c,d). Specifically, the combined use of N- and C-terminal antibodies allowed the detection of one N- (21.6 kDa) and one C-terminal fragment (34.7 kDa), both of which are quantitatively more important in 4- and 8-d girdled than control samples, and the combined molecular weight of which is close to the weight of the Rubisco large subunit. These two fragments could be derived from a single cleavage event, representing a first step towards the biochemical characterization of proteolytic or other events involved in the initiation of Rubisco degradation. The fact that these (and other) fragments are weaker (N-terminal antibody) or undetectable (C-terminal antibody) after 12 d may be caused by increased protease activity with progressing senescence (Parrott et al., 2005). Together, data presented in Figs 1–3 indicate that leaf girdling is a powerful model system for the study of early senescence events. As partial sieve tube occlusion has been shown to occur during natural leaf senescence (Jongebloed et al., 2004), it appears likely that our system enhances corresponding signals and processes, thus making them more accessible to experimental (e.g. genomic) analysis.
Carbohydrate accumulation is correlated with differential expression of numerous senescence-associated genes
The Affymetrix Barley1 22 840-element microarray was used specifically to identify genes upregulated in girdled leaves (compared with untreated and shift-girdled controls) at 0, 4 and 8 d after treatment. Results show that expression of several thousand genes 1.5-fold or more up- and downregulated is influenced by this treatment (Fig. 4; Tables S1, S2). These up- and downregulated genes may have an impact on many biological processes in girdled leaves, including carbohydrate metabolism, electron transport, energy pathways, cell organization and biogenesis, cell growth and/or maintenance, cell death, protein biosynthesis protein transport, development, and the regulation of gene expression (Fig. 5; Table S3). Of most interest with respect to this study are genes that might be involved in carbohydrate metabolism, senescence, and N transport and metabolism. Genes associated with N metabolism (Guo et al., 2004), including glutamine synthetase, glutamate synthase, asparagine synthetase, glutamate dehydrogenase and several aminotransferase genes, were upregulated at both 4 and 8 d in girdled leaves (Tables S1, S2). Several hexokinases, implicated in leaf senescence (Moore et al., 2003), were upregulated in 4- and 8-d girdled samples (Tables S1, S2); among these, a (putative) fructokinase (contig101_at) is almost 20-fold upregulated in 4-d girdled leaves (Table S1). Nine amino acid transporters were also upregulated, suggesting their involvement in the transport of amino acids derived from rapid net protein degradation (Tables S1, S2).We found that SAG12 was upregulated in 4- and 8-d girdled barley leaves and in naturally senescing leaves (Table 1; Figs 7a, 8a). It is interesting to note that SAG12 has been shown to increase over time (Wingler et al., 2006), and that SAG12 transcript levels increase from 0 to 8 d in girdled leaves and from 7 to 21 d in naturally senescing leaves. We also found numerous other senescence-related proteins, a glucose-6-phosphate/phosphate translocator and many MYB transcription factors upregulated in the GO BP lists (Table S3). Two interesting transcription factors implicated in leaf senescence in tobacco and Arabidopsis, HIN1 (Takahashi et al., 2004) and WRKY53 (Miao et al., 2004), were also found to be upregulated in 4- and 8-d girdled leaves. Wingler et al. (2006) have studied Arabidopsis gene-expression patterns as they pertain to sugar-induced gene expression, and have found several upregulated genes related to senescence and N remobilization (SAG 12, MYB75, MYB90, GLN1;4) which were induced when Arabidopsis was grown on a medium high in glucose.
Carbohydrate accumulation is correlated with drastically increased protease gene regulation in senescing barley leaves
Of significant interest in our data were 50 protease genes that were specifically upregulated at least twofold in girdled leaves. It is noteworthy that genes for both endo- and exoproteolytic enzymes, belonging to several catalytic classes (cysteine, aspartic, serine and probably metalloproteases), were induced. This is a considerably greater number of proteases upregulated in senescing leaves than have been reported in other studies (Gepstein et al., 2003; Wingler et al., 2006), indicating that our experimental approach is ideal to detect genes differentially regulated during early phases of the senescence process. The hierarchical tree cluster analysis (Fig. 6) shows three clusters of genes, each demonstrating relatively high to high expression in the girdled leaf samples: the first with low expression levels at 0 d; the second a mix of moderately low and moderate gene expression in controls at 0–8 d; and the third with higher levels of expression at 0 d. There appears to be no clear division between the three clusters, as proteases of all catalytic classes were represented in each cluster. However, probe sets for genes or gene families were clustered together in subclusters, and each of these is contained in one of the three main clusters. It is interesting to see that the hierarchical tree was also organized based on the ‘age’ of the leaf samples, where the ‘youngest’ leaves, 0-d and 4- and 8-d control and shift-girdled leaf samples, clustered followed by the ‘senescing’ 4-d girdled samples and finally the ‘senesced’, the 8-d girdled leaf samples. Our choice of gene candidates for microarray validation and gene-expression analysis in naturally senescing leaves comes from the second cluster, where gene expression is generally moderate in controls and shift-girdled leaves and highly upregulated in girdled leaves. This gene list suggested eight candidate genes for the amino-, carboxy- and possibly for the endopeptidase activities detected with standard biochemical assays (Parrott et al., 2005). Transcript levels of plastidial (aminopeptidases, CND41) and vacuolar (thiol and serine) proteases clearly increase in girdled leaves accumulating high carbohydrate levels when compared with control and shift-girdled leaves. In most instances, increases were also seen from 4 to 8 d. However, both CND41 and a proline iminopeptidase decreased in transcript level from 4 to 8 d (Table 1). Validation of the microarray data using qRT-PCR mostly fits with the increases and decreases seen in the chip data. Some discrepancies exist in the data, possibly caused by variations in sample preparation methods used for gene chip target preparation and qRT-PCR.
Only three of the eight upregulated proteases had detailed annotations from the Barley1 chip, making further analysis of these genes more challenging. However, in choosing members of the candidate list, we included two genes of special interest: cp-mIII, a highly expressed carboxypeptidase (Parrott et al., 2005); and CND41, a plastidial aspartic protease that has been implicated in Rubisco degradation in tobacco (Kato et al., 2004, 2005). It has been shown previously in our laboratory that there is a strong correlation between one or several carboxypeptidases and leaf N remobilization (Yang et al., 2004). Here we see very strong upregulation of cp-mIII in 4- and 8-d girdled and naturally senescing leaves (Table 1; Figs 7h, 8h), expanding on and reinforcing our notion that proteases are induced in senescing leaf tissues. The upregulation of CND41 in girdled barley leaves is intriguing. Kato et al. (2004) have shown increases in the level of CND41 in senescing leaves and corresponding decreases in Rubisco levels. In antisense CND41 tobacco, however, Rubisco degradation is not seen. Here we see similar increases in CND41 transcript levels (Table 1; Fig. 7c) and corresponding decreases in Rubisco levels (Fig. 3b–d). It has long been debated if major plastidial proteins are (partially) degraded inside intact chloroplasts during the early phases of leaf senescence, or whether extraplastidial (such as vacuolar) proteases are important as well (Hörtensteiner & Feller, 2002). Although, at present, our data are limited to the quantification of mRNA levels, they implicate a CND41 protein in plastidial proteolysis in a monocotyledonous species, supporting the notion that photosynthetic proteins are (partially?) degraded in the intact organelles during early phases of leaf senescence. In naturally senescing leaves, CND41 transcript levels at 7 and 14 d were high, but decreased drastically by 21 d. Second leaves (below the ear) at 21 dpa are almost completely senesced; presumably, most or all plastidial proteins are degraded by this time, explaining the observed decrease in CND41 transcript levels.
In conclusion, we have demonstrated that leaf girdling in barley leads to carbohydrate accumulation and induces leaf senescence, facilitating the analysis of many senescence-associated genes. Unique to our study is the discovery of many upregulated proteases, some of which may be directly or indirectly implicated in bulk protein degradation and N remobilization. While a previous study (Gepstein et al., 2003) has suggested that plastidial proteases (Clp, Fts, Spp etc.) play a role in leaf senescence, we find transcript levels of several of the corresponding genes to be downregulated (Tables S1, S2, S4, S5). Although it is well known that gene function is regulated at other than transcriptional levels, data presented in this manuscript do not suggest a major role for these particular genes in senescence-associated N recycling. Our findings correlate well with studies in both tobacco, where senescing leaves have been shown to upregulate genes responsible for Rubisco degradation (Kato et al., 2004, 2005), and Arabidopsis, where sugar accumulation has induced leaf senescence and senescence-associated genes, including N remobilization-associated genes (Pourtau et al., 2004, 2006; Wingler et al., 2006). It appears likely that the further characterization of many of our candidate genes will shed new light on the mechanism of bulk protein degradation during the early phases of leaf senescence.
We thank Regina Hölzer for the careful analysis of the Rubisco antibody preparations. Funding for this research was provided by grants from the USDA-NRI (project 2005-02022) and the Swiss National Science Foundation (project 31-55289.98) to A.M.F. and U.F., respectively. Additional support from the Montana Agricultural Experiment station to A.M.F. is also gratefully acknowledged.