Transcriptional profiling by cDNA-AFLP and microarray analysis reveals novel insights into the early response to ethylene in Arabidopsis


  • Annelies De Paepe,

    1. Unit Plant Hormone Signalling and Bio-imaging, Department of Molecular Genetics, Ghent University, K.L. Ledeganckstraat 35, B-9000 Gent, Belgium
    Search for more papers by this author
  • Marnik Vuylsteke,

    1. Department of Plant Systems Biology, Ghent University, Flanders Interuniversity Institute for Biotechnology (VIB), Technology Park 927, B-9052 Gent, Belgium
    Search for more papers by this author
  • Paul Van Hummelen,

    1. VIB Microarray Facility, Gasthuisberg, Onderzoek en Navorsing, B-3000 Leuven, Belgium
    Search for more papers by this author
  • Marc Zabeau,

    1. Department of Plant Systems Biology, Ghent University, Flanders Interuniversity Institute for Biotechnology (VIB), Technology Park 927, B-9052 Gent, Belgium
    Search for more papers by this author
  • Dominique Van Der Straeten

    Corresponding author
    1. Unit Plant Hormone Signalling and Bio-imaging, Department of Molecular Genetics, Ghent University, K.L. Ledeganckstraat 35, B-9000 Gent, Belgium
      (fax + 32 (0) 9 264 53 33; e-mail
    Search for more papers by this author


This article is corrected by:

  1. Errata: Correction Volume 56, Issue 1, 180, Article first published online: 25 September 2008

(fax + 32 (0) 9 264 53 33; e-mail


A comprehensive transcriptome analysis by means of cDNA-amplified fragment length polymorphism (AFLP) and cDNA-microarray technology was performed in order to gain further understanding of the molecular mechanisms of immediate transcriptional response to ethylene. Col-0 plants were treated with exogenous ethylene and sampled at six different time-points ranging from 10 min until 6 h. In order to isolate truly ethylene-responsive genes, both the ethylene-insensitive mutant ein2-1 and the constitutive mutant (ctr1-1) were analysed in parallel by cDNA-AFLP while ein2-1 was included for the microarray experiment. Out of the cDNA-transcript profiling covering about 5% of the Arabidopsis transcriptome, 46 ethylene-responsive genes were isolated, falling in different classes of expression pattern and including a number of novel genes. Out of the 6008 genes present on the chip, 214 genes were significantly (α = 0.001) differentially expressed between Col-0 and ein2-1 over time. Cluster analysis and functional grouping of co-regulated genes allowed to determine the major ethylene-regulated classes of genes. In particular, a large number of genes involved in cell rescue, disease and defence mechanisms were identified as early ethylene-regulated genes. Furthermore, the data provide insight into the role of protein degradation in ethylene signalling and ethylene-regulated transcription and protein fate. Novel interactions between ethylene response and responses to several other signals have been identified by this study. Of particular interest is the overlap between ethylene response and responses to abscisic acid, sugar and auxin. In conclusion, the data provide unique insight into early regulatory steps of ethylene response.


Ethylene is a simple two-carbon gaseous molecule with profound effects on plant growth and development, including promotion of phenomena such as the seedling triple response, responses to environmental stresses such as water deficit, mechanical wounding and pathogen attack, fruit ripening and tissue senescence (Abeles et al., 1992; Johnson and Ecker, 1998). The current knowledge of the regulation of ethylene signalling in Arabidopsis has emerged from genetic studies on mutants either fully or partially defective in ethylene responses (etr, ein) or with constitutive characteristics (ctr), providing a view on the mechanism underlying ethylene signalling (Bleecker et al., 1988; Guzman and Ecker, 1990; Kieber et al., 1993; Roman et al., 1995; Van Der Straeten et al., 1993). A linear pathway was proposed that initiates with ethylene binding at a family of ethylene receptors and terminates in a transcriptional cascade (Chang, 2003). In addition to the identification of receptors and downstream signalling components, these studies also have indicated the importance of transcriptional regulation in ethylene responses (Chao et al., 1997; Fujimoto et al., 2000; Solano et al., 1998). Although genetic screens were originally designed to identify specific components in ethylene signalling, mutations in these genes often confer changes in sensitivity to other hormones as well. For instance, mutations in ethylene signalling components have been recovered not only in screens using auxin transport inhibitors or cytokinins, but also in screens for suppressor and enhancer mutations of abscisic acid (ABA) mutants and for regulators of sugar metabolism (Beaudoin et al., 2000; Ghassemian et al., 2000; Vogel et al., 1998; Zhou et al., 1998). Together, these facts indicate that the linear representation of the hormone signalling pathways controlling a specific aspect of plant growth and development is oversimplified, and that hormones interact with each other and with a plethora of developmental and metabolic signals. Therefore, dissecting crosstalk between ethylene and other pathways is critical to the understanding of a plant's response to ethylene. To this end, comparing genetic interaction maps with patterns based on transcript profiling and other genomic technologies may create a more representative view of hormone interactions within the cell. Various targeted gene expression studies have identified ethylene-regulated genes in different processes and in different tissue types. Ethylene-response genes have been isolated using differential display techniques in etiolated seedlings and during tomato ripening (Trentmann, 2000; Zegzouti et al., 1999). Furthermore, previous research using cDNA microarrays revealed an important overlap of genes regulated by jasmonic acid, ethylene, and upon infection with an avirulent pathogen, indicating a network of regulatory interactions and coordination during pathogen and wounding responses (Schenk et al., 2000). Another study focused on transcriptional profiling of genes in response to wounding and a number of genes involved in ethylene biosynthesis and signalling were identified (Cheong et al., 2002). Very recently, Van Zhong and Burns (2003) described the profiling of ethylene-regulated gene expression in Arabidopsis by microarray analysis. In this analysis only one time-point (24 h) was investigated; therefore restricted to later ethylene-response genes. In addition, their data was compared with the insensitive mutant etr1-1 and the constitutive mutant ctr1-1, but no treatment was performed on the mutants. Little is known about early ethylene-regulated genes in Arabidopsis, modulated shortly after ethylene exposure. An extensive kinetic analysis of the transcriptional cascade in response to ethylene would allow mapping of the interaction between ethylene and other pathways in a comprehensive manner at the genomic level, besides identification of novel genes involved in ethylene signalling. Furthermore, the temporal cascade of gene expression in response to ethylene has not been analysed in detail. To this end, we performed a pilot study using the cDNA-amplified fragment length polymorphism (AFLP) transcript profiling approach (Breyne and Zabeau 2001). In this study, 19 days old Arabidopsis wild type (Col-0) plants were treated with two different concentrations of ethylene, for a number of time-points ranging from 10 min until 6 h, allowing a clear distinction between immediate early, early and late ethylene response genes. In order to isolate truly ethylene-responsive genes, the ethylene-insensitive mutant ein2-1 (Alonso et al., 1999) and the constitutive ethylene-response mutant ctr1-1 (Kieber et al., 1993) were included for analysis. At this stage of development ethylene plays a role in senescence (Park et al., 1998; Weaver et al., 1998). In addition, it is known that wounding, pathogen attack and different abiotic stresses induce ethylene production (Johnson and Ecker, 1998; Kende, 1993). Therefore, genes involved in ethylene-regulated stress responses are expected to come across our analysis. Moreover, this set-up allows exploring which vegetative processes are affected by ethylene.

In order to gain further understanding of the molecular mechanisms of immediate early ethylene action, we made use of cDNA arrays spotted with 6008 unique cDNAs from Arabidopsis ( This was performed both in wild type Arabidopsis plants and in the ethylene-insensitive mutant ein2-1.

Altogether 46 transcript tags were isolated from the cDNA-AFLP experiment and 214 genes were significantly modulated between the two genotypes Col-0 and ein2-1 over time in the microarray analysis. In particular, a large number of genes involved in cell rescue, disease and defence mechanisms were identified as early ethylene-regulated genes. Furthermore, the data provide insight into the role of protein degradation in ethylene signalling and ethylene-regulated transcription and protein fate. Novel players in the interaction between ethylene and several other signals have been identified.


Isolation and identification of ethylene-modulated genes by cDNA-AFLP analysis

As cDNA-AFLP allows the simultaneous analysis of multiple samples, this approach was chosen for a pilot experiment. To perform a detailed study of the very early responses to ethylene, we set up a kinetic analysis over six time-points. Plants were treated for 10, 20, 30 min, 1, 2 and 6 h with ethylene. The samples harvested after 6 h of treatment allowed us to investigate whether mRNA levels of the early genes remained constant during longer periods of ethylene exposure. A kinetic analysis with multiple sequential time-points increases the temporal resolution and thus allows discrimination between immediate early, early and late ethylene response genes. In order to confirm that the genes are modulated by ethylene, the ethylene-insensitive mutant ein2-1 and the constitutive ethylene-response mutant ctr1-1 were treated in parallel. The ein2-1 mutant was chosen because EIN2 is the only known gene for which a loss-of-function mutation leads to complete ethylene insensitivity (Roman et al., 1995). Consequently, it could be expected that truly ethylene-regulated genes would not be affected in the ein2-1 mutant and would display constitutive expression in ctr1-1. The experiment was performed twice, on independent batches of plants, once with 100 ppb and once with 10 ppm ethylene. Because the first (10 min) and the last (6 h) time-points of treatment were 5 h 50 min apart, untreated controls at both time-points were included for wild type and mutants. A substantial number of genes are known to be regulated by the circadian clock (Harmer et al., 2000). In this way, most genes that are under circadian control could be eliminated. In this analysis 10 primer combinations with selectivity +2/+1 and seven primer combinations with selectivity +2/+2 were used, covering on estimation 5% of the Arabidopsis transcriptome. A total of 1200 gene tags were monitored. After normalization of the AFLP-QuantarPro expression data and selection of differentially expressed genes based on the coefficient of variation (CV) >0.5 criterion, 46 genes were found to be strongly affected in their expression upon ethylene exposure. The reproducibility between the two experiments was as high as 96%. Only those genes that displayed a similar trend in expression pattern in both independent experiments were taken for further analysis, resulting in a total of 46 ethylene-regulated genes (Table 1). Differentially expressed transcript tags were excised from the gels, re-amplified with the selective primers and sequenced. Significant homology with Arabidopsis genes with known or putative functions was found for 36 of the 46 transcript fragments, whereas eight tags were from genes of unknown function. Only for two tags, no reliable sequence was obtained (displayed in Table 1 as unidentified tags).

Table 1.  Overview of ethylene-regulated genes from the cDNA-AFLP experiment and the microarray analysis
Description (MIPS)Description (TIGR)AccessionCluster(1)(2)(3)
cDNA-AFLP experiment
Polyubiquitinpolyubiquitin UBQ14At4g02890A21.124.24−5.64
Unidentified tagunidentified tag A9.872.68−7.02
Phosphoenolphyruvate carboxylase (PPC)posphoenolypyruvate carbozylase (PPC)At3g14940A35.877.29−9.96
Unknown proteinputative guanine nucleotide-binding proteinAt1g48630A7.033.11−1.83
Photosystem II oxygen-evolving complex protein 3-likeoxygen-evolving complex protein 16 (OEC16)At4g21280B−1.401.063.06
Unidentified tagunidentified tag B1.381.407.94
Unknown proteinunknown proteinAt5g05250B1.481.254.48
Pectinesterase-like protein(***)pectinesterase family (***)At3g49220B1.341.134.33
ENOD8-like proteinGDSL-motif lipase/hydrolase protein familyAt1g67380B1.271.057.37
Amino acid permease AAP3putative amino acid carrierAt1g77380B2.182.084.86
Unknown protein (*)putative 1-aminocyclopropane- 1-carboxylate oxidase (*)At1g62380B2.512.386.35
Putative bHLH transcription factor (bHLH037)bHLH protein family proteinAt3g50330B1.191.1710.87
Peroxidase ATP3a homologperoxidase, putativeAt5g64110B−1.111.0712.98
Male sterility2-like proteinacyl CoA reductase, putative/ male-sterility proteinAt5g22500B1.861.268.34
Unknown proteinunknown proteinAt1g06510B2.642.2010.05
Putative auxin-regulated protein (***)auxin-induced protein-related (***)At2g45210B3.942.155.95
Germin-like protein (GLP2a) copy2 (**)germin-like protein (AtGER2)(**)At5g39190B2.231.784.37
Putative carboxyphosphonoenol- pyruvate mutaseunknown proteinAt2g43180B1.161.022.12
Unknown proteinunknown proteinAt2g25510B1.671.812.48
Unknown proteinunknown proteinAt1g68600B5.703.441.09
Unknown proteinSOUL heme-binding protein familyAt3g10130C13.422.53−1.27
Unknown protein(**)acireductone dioxygenase (ARD) family (**)At5g43850C15.582.234.64
Ubiquitin-conjugating enzyme E2-17 kD 10E2 ubiquitin-conjugating enzyme 10 (UBC10)At5g53300C13.822.455.08
Unknown proteinputative Eukaryotic translation initiation factorAt1g36730C19.583.021.37
Putative RAV-like B3 domain DNA binding proteinDNA-binding protein, putativeAt2g46870C15.473.255.29
Plasma membrane proton ATPase (PMA)ATPase 1, plasma membrane-type, putativeAt2g18960C15.103.061.42
Putative pre-mRNA splicing factor RNA helicaseRNA helicase, putativeAt2g35340C15.315.031.38
Arabinogalactan protein (AGP7)arabinogalactan protein (AGP7)At5g65390C19.686.841.27
Putative receptor-like protein kinase (***)S-locus lectin protein kinase family (***)At1g11330C114.734.10−3.00
Unknown proteinmechanosensitive (MS) ion channel domain proteinAt3g14810C132.468.39−2.39
RNA-binding protein cp29RNA-binding protein cp29At3g53460C123.346.441.31
Putative transcription factor (MYB3)myb family transcription factor (MYB4)At1g22640C29.395.20−1.27
Senescence-associated gene 20 (*)expressed protein (*)At3g10980C25.312.802.95
Putative tropinone reductaseputative tropinone reductaseAt2g29310C28.084.627.89
Cu/Zn-superoxide dismutase copper chaperone (**)Cu/Zn-superoxide dismutase copper chaperoneAt1g12520C25.645.45−1.04
Unknown proteinsenescence-associated protein-relatedAt5g47060C213.874.82−5.07
Heat shock protein 1717.4 kDa class I heat shock protein (hsp 17.4-CI)At3g46230C243.387.731.66
Cysteine proteinase RD21ACysteine proteinase RD21AAt1g47128C238.097.111.31
Putative ribosomal protein60S ribosomal protein L37 (RPL37C)At3g16080D14.391.09−6.83
Unknown proteinunknown proteinAt1g54380D12.561.08−7.68
Putative ribosomal protein60S ribosomal protein L37 (RPL37C)At3g16080D14.391.09−6.83
Unknown proteinunknown proteinAt1g54380D12.561.08−7.68
Ferritin 1 precursor (***)ferritin 1 (FER1) (***)At5g01600D15.16−1.29−2.43
IRE-like protein kinaseprotein kinase, putativeAt1g48490D1−1.58−6.16−4.75
Putative protein (**)putative protease inhibitor Dr4 (**)At1g73330D21.04−3.89−6.67
Putative MADS Box / AGL proteinMADS-box protein AGL 19At4g22950D2−1.60−5.21−1.22
BHLH transcription factor (bHLH027)-like proteinbHLH transcription factor familyAt4g29930D2−1.01−6.76−1.54
Putative xyloglucan endo-transglycosylaseputative xyloglucan endo-transglycosylaseAt2g18800D21.25−2.43−1.05
Acyl-CoA synthetase like proteinAMP-dependent synthetase and ligase familyAt2g17650D2−1.57−4.97−2.14
 Microarray analysisAccessionCluster(1)(2)
  1. Five previously identified ethylene-regulated genes are indicated by (*), four transcripts in common between both analyses and present in both final data-sets are indicated by (**), four transcripts in common between both analyses, but not in both final data-sets are indicated by (***). (1) maximal induction in wild type compared to wild type control (2) ratio average values of wild type over average values of the ein2-1 mutant (3) ratio average values of the ctr1-1 mutant over average values of the wild type.

genotype × time-effect list:
Unknown proteinexpressed proteinAt3g14190A13.1−1.0
Glycine-rich RNA binding protein,putativeRNA-binding protein, putativeAt3g23830A13.1−1.0
Unknown proteinexpressed proteinAt5g43970A13.0−0.9
Unknown proteinexpressed proteinAt3g22105A12.7−1.1
Possible aposory-associated like proteinaldose-1 epimerase familyAt4g25900A12.9−1.4
Unknown proteindehydrin (ERD 14)At1g76180A13.31.0
Unknown proteinlate embryogenesis abundant (LEA) protein familyAt4g02380A12.11.6
Lipid-transfer protein-like, predicted GPI-anchored proteinprotease inhibitor/seed storage/lipid transfer protein (LTP)At3g43720A13.11.1
Putative fibrillinplastid-lipid associated protein PAP/fibrillin, putativeAt4g04020A12.11.1
Cinnamoyl-CoA reductase-like proteincinnamoyl-CoA reductase-relatedAt4g30470A17.01.5
Integral membrane protein, putativeMATE efflux protein familyAt3g21690A12.01.5
ABC transporter like proteinABC transporter family proteinAt5g06530A14.11.8
Unknown proteinexpressed proteinAt5g56980A12.31.4
GDSL-motif lipase/hydrolase-like proteinGDSL-motif lipase/hydrolase protein familyAt5g45950A19.91.2
4-coumarate-CoA ligase-like proteinAMP-dependent synthetase and ligase familyAt3g48990A1−2.01.4
1-aminocyclopropane-1-carboxylate oxidase1-aminocyclopropane-1-carboxylate oxidaseAt1g05010A13.51.9
Unknown proteinexpressed proteinAt5g18460A12.11.3
ATP-dependent transmembrane transporter, putativeABC transporter family proteinAt1g51460A12.01.6
Unknown proteinexpressed proteinAt5g44580A13.03.4
Formate dehydrogenase (FDH)formate dehydrogenase (FDH)At5g14780A13.63.5
Putative ethylene receptor ERS2 (*)ethylene receptor-related (*)At1g04310A14.25.0
UDP-glucose:indole-3-acetate beta-D-glucosyltransferase (iaglu)UDP-glucose:indole-3-acetate beta-D-glucosyltransferaseAt2g40940A12.11.9
Anthranilate N-benzoyltransferase-like proteintransferase familyAt4g15550A15.55.3
Putative proteinexpressed proteinAt5g01210A14.44.2
RSH3 (*)RelA/SpoT protein, putative (RSH3) (*)At3g59900A111.35.1
Unknown proteinexpressed proteinAt1g54130A12.21.8
Unknown proteinexpressed proteinAt5g27320A12.71.3
Hypothetical proteinRNA binding protein 47 (RBP47), putativeAt1g47500A12.01.8
Putative mRNA capping enzyme, RNA guanylyltransferasemRNA capping enzyme-relatedAt3g09100A12.21.4
1-aminocyclopropane-1-carboxylate oxidase1-aminocyclopropane-1-carboxylate oxidaseAt5g43450A12.31.6
Abscisic acid responsive elements-binding factor (ABRE/ABF3)abscisic acid responsive elements-binding factor (ABF3)At4g34000A12.01.4
Predicted protein of unknown functionTolB protein-relatedAt4g01870A14.63.6
Unknown proteinexpressed proteinAt1g19180A12.82.4
Unknown proteinexpressed proteinAt4g27900A17.12.0
Unknown proteintransport protein Gsa12p-relatedAt3g62770A12.91.6
Putative proteinhydrolase familyAt4g33540A13.61.4
Hypothetical proteincalcium-binding EF-hand protein familyAt1g54450A12.91.3
Ids4-like proteinSPX (SYG1/Pho81/XPR1) domain proteinAt5g20150A13.81.4
60S ribosomal protein L3060S ribosomal protein L24,putativeAt2g44860A12.11.2
Ribosomal protein L36-likeribosomal protein familyAt5g20180A12.61.2
Ethylene responsive element binding factor 2 (ATERF2) (*)ethylene responsive element binding factor 2 (EREBP-2) (*)At5g47220A12.02.6
Unknownexpressed proteinAt1g53890A12.21.9
Nitrilase 4nitrilase 4(NIT4)At5g22300A12.01.7
Cinnamyl-alcohol dehydrogenase ELI3-1mannitol dehydrogenase (ELI3-1), putativeAt4g37980A12.51.9
Putative mitogen-activated protein kinase MMK2mitogen-activated protein kinase (MAPK), putative (MPK2)At1g59580A12.01.3
S-adenosylmethionine decarboxylase (adoMetDC2)adenosylmethionine decarboxylase familyAt5g15950A17.62.8
Unknown proteinexpressed proteinAt4g25670A17.31.4
Protein phosphatase 2C AB12 (PP2C)protein phosphatase 2C, AB12At5g57050A14.51.7
Abscisic acid insensitive protein (ABI1)protein phosphatase ABI1At4g26080A119.03.3
14-3-3 protein homolog RCI1 (pir‖S47969)14-3-3 protein GF14 psi (grf3/RCI1)At5g38480A13.41.5
Receptor kinase (CLV1)CLAVATA1 receptor kinase (CLV1)At1g75820A12.81.6
Putative proteinaldehyde dehydrogenase, putative (ALDH)At1g54100A1−3.41.1
Putative proteinKelch repeats protein familyAt5g30310B12.01.5
Putative protein (**)acireductone dioxygenase (ARD) family (**)At5g43850B12.11.6
Expansin S2 presursor like proteinexpansin, putative (EXP11)At1g20190B12.71.0
Zinc finger like proteinzinc finger (CCCH type) protein familyAt5g44260B1−2.41.6
Nucleoid DNA-binding - like proteinnucleoid DNA-binding - like proteinAt3g54400B12.31.7
Putative SF16 protein {Helianthus annuus}calmodulin-binding protein familyAt2g43680B12.31.5
Plasma membrane proton ATPase-likeATPase, plasma membrane-type (proton pump), putativeAt5g62670B14.11.4
Putative proteinexpressed proteinAt1g80865B12.62.0
Subtilisin proteinase like proteinsubtilisin-like serine proteaseAt4g21650B12.11.8
Arginine decarboxylasearginine decarboxylase 1 (SPE1)At2g16500B12.21.6
Elongation factor G, putativeelongation factor Tu family proteinAt1g62750B12.11.6
GTP-binding protein typA (tyrosine phosphorylated protein A)GTP-binding protein typA (tyrosine phosphorylated protein A)At5g13650B12.01.5
Putative translation initiation factor IF2translation initiation factor IF-2 [chloroplast], putativeAt1g17220B12.01.5
Putative proteinWD-40 repeat protein familyAt1g15750B12.11.3
Unknown proteinelongation factor Ts familyAt4g29060B12.31.3
Hsp 70-like proteinheat shock protein cpHsc70-1At4g24280B1−1.42.0
Unknown proteinhypothetical proteinAt1g33860B1−3.41.9
Calcium-dependent protein kinasecalcium-dependent protein kinase family (CDPK)At5g66210B12.81.8
Arabinogalactan-protein AGP1 (gb|AAC77823.1)arabinogalactan-protein (AGP1)At5g64310B12.62.2
Xyloglucan endo-1,4-beta-D-glucanase-like proteinxyloglucan endotransglycosylase, putativeAt4g30280B12.42.5
Nitrate reductase 1 (NR1)nitrate reductase 1 (NR1)At1g77760B1−3.81.4
Unknown proteinsurp module familyAt5g23080B13.72.0
Acyl-(acyl carrier protein) thioesterase, putativeacyl-(acyl carrier protein) thioesterase (acyl-ACP thioesterase)At1g08510B12.01.4
Ethylene-insensitive3-like1 (EIL1)ethylene-insensitive3-like1 (EIL1)At2g27050B1−2.41.8
Receptor-like kinase, putativeleucine rich repeat protein kinase familyAt3g23750B1−3.71.9
DNA-binding protein, putativemyb family transcription factorAt1g01060B1−3.72.2
Unknown proteincell wall protein precursor, putativeAt2g20870B12.21.7
Unknown proteinpectinesterase familyAt1g14890B12.22.1
AP2 domain containing protein RAP2.3AP2 domain transcription factor RAP2.3At3g16770B1−2.212.8
Unknown proteinglycosyl hydrolase family 5At5g01930B1−2.01.4
PSI 9KDa proteinpsaCAt4g14160B12.51.8
Putative receptor-like protein kinase, ERECTAleucine rich repeat protein kinase family (ERECTA)At2g26330B12.01.3
PsalPSI I proteinAt5g51790B13.61.7
Ribosomal protein S7hypothetical proteinAt5g35130B13.62.0
Putative proteinexpressed proteinAt5g16730B12.01.1
Eukaryotic translation initiation factor 4E, putativeeukaryotic translation initiation factor 4E (eIF-4E)At1g29550C12.8−1.4
Hypothetical proteinisochorismate synthase (isochorismate mutage), putativeAt1g18870C12.3−1.6
Unknown proteinunknown proteinAt3g15100C1−2.7−1.5
Putative auxin-induced protein AUX2-11auxin-responsive protein (IAA3 3)At1g04240C1−6.81.0
Nodulin-like proteinnodulin MtN21 family proteinAt1g75500C1−3.3−1.7
Putative proteinpolygalacturonase (pectinase) familyAt1g70370C1−3.7−1.5
Two-pore calcium channel (AtTPC1)two-pore calcium channel (TPC1)At4g03560C1−2.0−1.2
Peroxidase (Unknown protein)peroxidase, putativeAt5g42180C1−2.8−1.8
Hypothetical proteinSWIB complex BAF60b domain-containing proteinAt2g14880C1−2.0−1.3
Putative proteinexpressed proteinAt5g65480C1−3.8−1.7
Hypothetical proteinaminotransferase-relatedAt4g28410C1−6.1−5.0
Putative DNA-binding protein, GT-1DNA binding protein GT-1-relatedAt3g25990C1−2.6−3.2
Unknown proteinFAD-linked oxidoreductase familyAt2g46750C1−2.3−2.1
Fe (II) transport proteiniron-regulated transporter (IRT1)At4g19690C1−2.1−6.1
Unknown proteinpectin methylesterase, putativeAt1g11580C1−2.8−3.3
Peroxidase like proteinperoxidaseAt2g38380C1−2.2−2.7
Unknown proteinearly nodulin ENOD18 familyAt3g03270C12.0−2.8
Putative peroxidase ATP12aperoxidase, putativeAt1g05240C1−2.4−3.1
Putative peroxidaseperoxidase, putativeAt3g21770C1−2.0−2.7
Putative proteinamino acid permease familyAt3g13620C1−1.8−2.4
Photoassimilate-responsive protein PAR-1b-like proteinphotoassimilate-responsive protein-relatedAt3g54040C1−2.0−1.6
Peroxidase ATP3aperoxidase, putativeAt5g64100C1−2.0−2.0
Peroxidase ATP13aperoxidase, putativeAt5g17820C1−2.1−1.9
Unknown proteinmajor latex protein (MLP)-relatedAt2g01520C1−2.8−7.4
UTP-glucose glucosyltransferaseglycosyltransferase familyAt5g66690C1−2.3−3.6
Thioglucosidase 3D precursorglycosyl hydrolase familyAt3g09260C1−2.9−5.7
Putative lectinmyrosinase binding protein, putativeAt3g16420C1−2.7−5.0
Phytocyanin, blue copper-binding protein IIuclacyanin IIAt2g44790C1−2.4−5.8
Receptor-protein kinase-like proteinprotein kinase familyAt5g24010C1−2.2−4.9
Unknown proteinmeprin and TRAF homology (MATH) domain proteinAt5g26280C1−2.6−5.5
Putative jasmonate inducible proteinjacalin lectin familyAt3g16400C1−2.0−3.3
1-aminocyclopropane-1-carboxylic acid oxidase-like proteinflavonol synthase, putativeAt5g63600C1−2.3−2.8
Cytochrome P450-like proteincytochrome p450 familyAt5g48000C1−2.5−3.3
Unknown proteinmeprin and TRAF homology (MATH) domain proteinAt5g26260C1−2.1−4.0
Unknown proteinprotease inhibitor/seed storage/lipid transfer protein (LTP)/familyAt2g48140C1−2.0−2.5
Cytochrome P450 monooxygenase like proteincytochrome P450, putativeAt4g37410C1−2.5−3.2
ABC transporter-like proteinABC transporter family proteinAt3g53480C1−2.1−2.4
Phosphate transporterinorganic phosphate transporter (PHT1)At5g43350C1−2.5−3.9
Patatin-like proteinpatatin, putativeAt4g37070C1−2.8−2.1
Feebly like proteinpathogen-inducible alpha-dioxygenase, putativeAt3g01420C1−5.0−3.3
HSR201 like proteinacyltransferase familyAt4g15390C1−3.0−4.8
Putative protein (**)protease inhibitor, putative (Dr4) (**)At1g73330C1−2.6−4.7
Periaxin-like proteinproline-rich protein familyAt5g09530C1−2.2−2.5
Unknown proteinintegral membrane protein familyAt4g27860C1−2.6−1.9
Putative beta-fructosidase (At1g62660)glycosyl hydrolase family 32At1g62660C1−3.3−2.0
Putative proteinmethionine sulfoxide reductase (SelR) domain proteinAt4g21850C1−3.1−2.3
Unknown proteinauxin efflux carrier protein familyAt2g17500C1−2.4−1.7
Beta-1,3-glucanase class I precursorglycosyl hydrolase family 17At4g16260C1−2.8−1.6
Putative trypsin inhibitor (At1g73260)trypsin and protease inhibitor (Kunitz) familyAt1g73260C1−6.4−1.9
Hypothetical proteinexpressed proteinAt3g21710C1−2.5−2.2
Osmotin precursorosmotin-like protein (OSM34)At4g11650C1−2.7−1.9
Putative trypsin inhibitortrypsin inhibitor, putativeAt2g43510C1−3.1−1.9
VIB02-016putative trypsin inhibitor C1−3.2−1.9
Superoxide dismutase (EC (Fe)iron superoxide dismutase (FSD1)At4g25100C1−2.4−3.8
Peroxidase, prxr2peroxidise, putativeAt4g37520C1−8.1−1.5
Peroxidase, ATP8aperoxidise, putativeAt4g30170C1−5.3−2.2
Unknown proteinglycine-rich proteinAt2g36120C1−2.3−1.8
Phosphoenolpyruvate carboxykinase (ATP)-like proteinPhosphoenolpyruvate carboxykinase [ATP]At1g37870C1−2.0−1.6
Amino acid transport protein AAP2amino acid permease 2 (AAP2)At5g09220C1−2.3−1.9
Putative Iron/Ascorbate oxidoreductase family protein2-oxoglutarate-dependent dioxygenase, putativeAt1g06640C1−2.3−1.6
Unknown proteinexpressed proteinAt5g62130C1−2.2−5.3
Cytochrome C oxidoreductase like proteinnodulin MtN3 family proteinAt4g15920C1−2.2−1.8
Germin-like proteingermin-like protein (GLP4) (GLP5)At1g09560C1−2.0−1.8
Sugar transporter like proteinmannitol transporter, putativeAt4g36670C1−5.4−2.4
Heat shock protein 101heat shock protein 101 (HSP 101)At1g74310C1−3.6−1.0
Putative cysteine proteinasecysteine proteinase XCP2At1g20850C12.91.0
genotype difference-list:
Unknown proteinbrassinosteroid signalling positive regulator, putativeAt1g19350A2−1.82.8
Ethylene response sensor (ERS) (*)ethylene response sensor (ERS) (*)At2g40940A2−1.82.0
Putative Cu/Zn superoxide dismutase copper chaperone (**)copper/zinc superoxide dismutase copper chaperone, putative (**)At1g12520A2−1.72.0
Putative proteinexpressed proteinAt1g70230A21.72.0
Unknown proteinsporulation protein AA-relatedAt3g10420A21.84.2
Unknown proteinactin-related protein 4 (ARP4)At1g18450A2−1.42.0
Lycopene epsilon cyclaselycopene epsilon cyclaseAt5g57030A21.62.0
Unknown proteinVQ motif-containing protein familyAt3g22160A21.42.0
Putative proteinexpressed proteinAt3g61700A21.52.3
Putative protease SppA (SppA)SppA protease IV (SppA)At1g73990A21.22.0
Putative copper/zinc superoxide dismutasecopper/zinc supperoxide dismutase (CSD2)At2g28190A2−1.42.1
Unknown proteinexpressed proteinAt2g47270A21.22.0
Germin-like protein (GLP2a) copy2 (**)germin-like protein (AtGER2) (**)At5g39190A21.92.6
Expansin At-EXP5expansin, putative (EXP5)At3g29030B2−1.3−2.0
1-aminocyclopropane-1-carboxylate oxidase1-aminocyclopropane-1-carboxylate oxidase (ACC oxidase), putativeAt2g19590B2−1.7−2.0
Nicotianamine synthase (dbj|BAA74589.1)nicotianamine synthase, putativeAt5g04950B2−1.3−2.7
Unknown proteinmeprin and TRAF homology (MATH) domain proteinAt5g26290B2−1.5−2.0
Unknown proteincatalase 3At1g20620B2−1.9−3.2
Disease resistance response/dirigent-like proteindisease resistance response protein-relatedAt3g55230B2−1.0−2.1
Unknown proteinmeprin and TRAF homology (MATH) domain proteinAt5g26260B2−1.6−3.8
Unknown proteinFAD-linked oxidoreductase familyAt2g46750B2−1.6−2.3
Glutamate decarboxylase 1 (GAD 1)glutamate decarboxylase 1 (GAD 1)At5g17330B2−1.7−2.9
Pyruvate decarboxylase-1 (Pdc1)pyruvate decarboxylase, putativeAt4g33070B2−1.5−5.4
Nitrate/chlorate transporter CHL1nitrate/chlorate transporter (NRT1.1/CHL1)At1g12110B2−1.2−2.4
Putative fructokinasepfkB type carbohydrate kinase protein familyAt2g31390B2−1.2−2.3
Putative proteinglyceraldehyde 3-phosphate dehydrogenaseAt1g79530B2−1.2−2.4
Phosphoenolpyruvate carboxylase (PPC)phosphoenolpyruvate carboxylase (PEP carboxylase/PEPCase)At3g14940B2−1.6−2.0
Unknown proteinmyrosinase-associated protein, putativeAt1g54010B2−1.6−2.7
Phospholipase like proteinesterase/lipase/thioesterase familyAt2g39420B2−1.6−2.0
Putative proteindisease resistance response protein-relatedAt4g13580B2−1.7−2.3
Unknown proteinexpressed proteinAt4g22212B2−1.8−3.7
Ferrodoxin NADP oxidoreductase-like proteinferrodoxin-NADP(+) reductase (adrenodoxin reductase), putativeAt1g30510B2−1.1−2.0
Beta-glucosidase, putativeglycosyl hydrolase family 1At1g66280B2−1.8−4.9
Putative myrosinase-binding proteinjacalin lectin familyAt2g39310B2−1.9−4.1
Putative lectinjacalin lectin familyAt3g16390B2−1.5−2.7
Putative alcohol dehydrogenaseshort-chain dehydrogenase/reductase family proteinAt2g47130B2−1.5−3.4
Putative endochitinaseglycosyl hydrolase family 19 (chitinase)At2g43610B2−1.4−2.8
Putative proteindisease resistance protein (TIR class), putativeAt5g44920B2−1.3−2.0
Putative peroxidase ATP2aperoxidase, putative (ATP2a)At2g37130B2−1.8−2.5
Cinnamoyl CoA reductase like proteincinnamoyl-CoA reductase, putativeAt1g80820B2−1.9−2.8
WRKY transcription factor 11 (WRKY 11)WRKY family transcription factorAt4g31550B2−1.4−4.0
Unknown proteinmyrosinase-associated protein, putativeAt1g54000B2−1.4−2.6
Unknown proteinexpressed proteinAt1g47480B2−1.1−2.0
Cytochrome p450 (ROTUNDIFOLIA3)cytochrome P450 90C1/rotundifolia3 (rot3)At4g36380B2−1.7−2.0
Putative Fe (II) transporterzinc transporter (ZIP3)At2g32270B2−1.3−2.0
Putative protein kinaseprotein kinase familyAt3g56050B2−1.3−2.0
Unknown proteinMuDR (mutator) transposase familyAt5g50315B2−1.2−3.5
Putative laccaselaccase) diphenol oxidase), putativeAt2g30210B2−1.8−2.1
Unknown proteinlateral organ boundaries (LOB) domain protein 25 (LBD25)At3g27650B2−1.7−2.1
Putative hexose transportersugar transporter, putativeAt4g02050B2−2.0−1.5
Thioredoxin (clone GIF1) (pir||S58118)thioredoxin H-type 3 (TRX-H-3)At5g42980B2−1.4−2.1
Putative proteinproline-rich protein familyAt4g38080B2−1.9−2.0

In Figure 1(a), the hierarchical average linkage clustering (Eisen et al., 1998) of the 46 differentially expressed genes for both treatments is presented (Eisen et al., 1998). Based on this clustering, we were able to distinguish four main expression profiles, designated A, B, C and D. In the first group (Figure 1a, A) expression was rapidly modulated and transient, with maximum mRNA levels after 10 and 20 min of treatment. The common feature for the second group (Figure 1a, B) is the high expression in ctr1-1. A third profile (Figure 1a, C) revealed an early temporal increase in mRNA levels during the treatment. This cluster can be further divided in two subclusters. In the first subcluster (Figure 1a, C-1) the peak of expression occurred after 1–2 h of treatment. In the second subcluster (Figure 1a, C-2), upregulation by ethylene was observed after 10 min of treatment and remained high until 2 h. Finally, a fourth expression profile is seen in cluster D, with a low expression in ctr1-1. Here again, it was possible to classify the expression profiles in two subclusters. In the first subcluster (Figure 1a, D1), mRNA levels were specifically low in the ctr1-1 mutant, whereas comparable mRNA levels were seen in the wild type and the ein2-1 mutant. This was not the case for the second subcluster (Figure 1a, D2), where a higher expression level could be observed in the ein2-1 mutant, indicating that these genes are possibly downregulated by ethylene.

Figure 1.

Hierarchical average linkage clustering of the isolated ethylene-regulated transcripts from the cDNA-AFLP experiment and the ethylene-modulated genes from the microarray analysis.
(a) cDNA-AFLP experiment: each column represents a time-point of sampling during the treatments for wild type, ein2-1 and ctr1-1 and each row represents the expression profile of a tag over the seven time-points for Col-0 and ein2-1. For the ctr1-1 samples the results are displayed for the three time-points. Red and green reflect transcriptional activation and repression respectively. The main expression patterns are divided in four groups (A, B, C, D). Clusters C and D were divided in two subclusters (1, 2). On the left side, the data from the 100ppb treatment are presented, on the right the data from the 10ppm treatment.
(b) Significantly ethylene-modulated genes out of the genotype × time analysis: each column represents the time-point of sampling during the treatment with 10 ppm ethylene in both wild type and the ein2-1 mutant. Three main clusters are indicated on the left (A1, B1, C1).
(c) Genes displaying significant difference in expression between the two genotypes. Two main clusters are indicated on the left (A2, B2).

Gene identity, expression values and extra information are provided as supplementary data (list cDNA-AFLP).

6K-Arabidopsis microarray: experimental set-up, validation of data and data analysis

A microarray containing 6008 unique Arabidopsis ESTs spotted in duplicate, representing about a fourth of the Arabidopsis transcriptome, was used to perform a broader monitoring of early transcriptome changes in wild type (Col-0) and mutant plants (ein2-1) upon treatment with 10 ppm ethylene. Here again, the analysis of the ethylene-insensitive mutant ein2-1, allowed us to distinguish between truly ethylene responsive genes from genes influenced by circadian rhythm during the treatment. A reference design was used, consisting of 14 dye-swap experiments wherein the two test samples (Col-0 and ein2-1 sampled across the seven time-points each) were directly compared with the ‘reference’ sample ctr1-1. In a first experiment, each of the genotype samples harvested at the various time-points were assigned to the green dye and the reference sample to the red dye. In a second experiment, the assignment of dyes was swapped, resulting in a total of 28 cDNA microarrays.

A total of 747 (12.5%) clones for which no ‘consistent‘ (see Experimental procedures) positive signal for all Col-0 and ein2-1 samples was detected, were excluded from the analysis. Subsequently, we applied two sequential mixed model analyses of variance (anovas) (Wolfinger et al., 2001) to the base-2 logarithm of all the ‘LOWESS’-transformed spot measurements, using the REML procedure as implemented in Genstat (Genstat Release 6.1 for Windows; VSN International, Hemel Hempstead, UK) (see Experimental procedures). To test differences between the two genotypes Col-0 and ein2-1, the seven time-points and the 14 genotype × time effects, we used Wald statistics. Expression differences with P-values ≤0.001 were called significant resulting in 476, 1368 and 231 genes showing significant genotype, time and genotype × time effects on their expression, respectively (see Figure S1a). Genes affected in their expression across the seven time-points in a non-genotype-specific manner were considered as affected by circadian rhythm and were therefore ruled out for further analysis. An additional selection of differentially expressed genes was based on a two-fold-change criterion resulting in a set of 57 genes only affected in a genotype specific manner and 157 significantly (P < 0.001) ethylene-regulated genes showing a two-fold-change in at least one time-point in Col-0 but not in ein2-1 (Table 1). The high reproducibility of the hybridization signal was supported by 12 significantly differentially expressed genes for which two independent cDNA clones were spotted on the array. The expression profiles obtained by the clones were comparable in all cases (Figure S1b).

Cluster analysis of the microarray data

To group genes with similar expression patterns, the profiles of the 57 genes only displaying a significant genotype-effect (Figure 1c) and the 157 significantly differentially expressed genes (genotype × time effect) (Figure 1b) were analysed using the hierarchical average linkage clustering (Eisen et al., 1998).

The 157 significantly differentially expressed genes could be divided in three subclusters. For the genes belonging to cluster A1 (Figure 1b) the expression level in wild type was significantly upregulated by ethylene in wild type at later time-points of treatment (comparable with cluster C1 from the cDNA-AFLP). From the top to the bottom of this subcluster, a shift from the latest time-point of treatment (6 h) to intermediate time-points (1 h–6 h) and finally to the earlier (30 min–6 h) ethylene regulation is visible. Cluster B1 (Figure 1b) contained genes upregulated by ethylene during the earliest time periods of treatment (comparable to cluster C2 from the cDNA-AFLP study). Within cluster B1, genes that display an immediate early temporal increase in expression (much comparable to cluster A of the cDNA-AFLP), are present. Finally, the largest cluster, cluster C1 (Figure 1b), contained downregulated genes. This cluster corresponds to cluster D2 observed in the cDNA-AFLP. Gene identity, expression values and extra information are provided as supplementary data (list average linkage clustering-genotype difference and list average linkage clustering-genotype × time-effect).

For the genes only affected by genotype difference, two main expression profiles could be identified: higher expression in the ein2-1 mutant (Figure 1c, A2) and higher expression in wild type (Figure 1c, B2).

In parallel to the hierarchical average linkage clustering, the adaptive quality-based clustering (AQBC) algorithm was applied (De Smet et al., 2002). Application of several clustering methods is not only useful as an independent confirmation of the grouping of expression profiles, but also allows discovering additional interesting features. For this analysis, the minimal number of genes in a cluster was set to 5. The algorithm allowed clustering 123 out of the 157 genes from the genotype × time analysis into five clusters (Figure 2a). Cluster 1a (55 genes) and cluster 2a (35 genes) were predominant and include transcripts repressed and induced by ethylene, respectively (covered by clusters C1 and A1 in the hierarchical clustering analysis). Cluster 3a (21 genes) corresponded largely to cluster B1 in the hierarchical average linkage clustering, containing early induced genes. Cluster 4a (28 genes) contained transcripts that are specifically upregulated after 1–2 h of treatment, also covered by cluster B1 from the hierachical clustering. The genes belonging to cluster 5a (five genes) are strongly repressed by ethylene, especially after 2–6 h (grouped in cluster C1 from the hierarchical clustering).

Figure 2.

Gene expression profiles obtained by quality-based clustering.
The 157 differentially expressed genes from the genotype × time analysis (a) and the 57 genes displaying significant genotype difference (b) were clustered using specific software (De Smet et al., 2002). Dashed lines represent the normalized expression profiles of individual genes. The full line graphs represent the mean expression profile of all transcripts belonging to the respective cluster. Values on the y-axis are the normalized expression data. For each cluster the number of genes is indicated.

Similar to the hierarchical average linkage clustering, the genes only affected in genotype difference could be subdivided in genes with higher expression in the ein2-1 mutant (cluster 1b) and wild type (cluster 2b), respectively (Figure 2b).

Complete lists of these clusters are provided as supplementary data (list AQBC-clusters genotype difference and list AQBC-clusters genotype × time-effect).

Assessment of reliability of the microarray data

In the microarray analysis, four known ethylene-regulated genes (see Table 1) were identified displaying an expression profile similar to that previously described. In addition to these internal controls, an independent confirmation was obtained from eight genes that displayed similar expression profiles in both the cDNA-AFLP experiment (see Table 1) and the microarray experiment. To further test the reliability of the microarray results, and to prove repeatability on independent biological samples, the expression patterns of eight candidate genes (belonging to different clusters) were examined by reverse transcription-polymerase chain reaction using the cDNA samples from the 100 ppb treatment. Analysis of the expression patterns of these genes by RT-PCR revealed very consistent profiles compared with the results of the microarray experiment (Figure 3). In total, two induced genes (At3g59900 and At1g15550), one gene with a higher expression in wild type compared with the mutant (At3g16770), two downregulated genes (At5g62130 and At1g62660), a constitutive gene (At5g53180) and two genes with a transient expression profile (At4g01870 and At5g62670) were tested.

Figure 3.

Confirmation of expression profiles obtained by RT-PCR and microarray analysis on independent biological repeats.
For the RT-PCR analysis (left panel), amplifications were performed on the cDNA-AFLP preamplification products of wild type and mutant samples treated with 100 ppb of ethylene. The following genes were analysed: At3g59900, putative gene; At4g15550, UDP-glucose:indole-3-acetate β-d-glucosyltransferase (iaglu); At3g16770, AP2 domain containing protein RAP2.3; At5g62130, putative gene; At1g62660; putative β-fructosidase; At4g01870, putative gene; At5g53180, RNA-transport protein-like gene; At5g62670; plasma membrane proton ATPase-like gene. Specific oligonucleotide primers were used to detect the transcript levels of the genes listed in Experimental procedures. Primer sequences are given in Table S1. The microarray analysis (right panel) was performed on samples of wild type and the ein2-1 mutant treated with 10 ppm of ethylene. The same kinetic analysis was performed in both cases.

Taking together the internal controls, the experiments using different technologies and the independent biological repeat, support the reliability of the ethylene-regulated gene expression patterns detected using the microarray approach.

Functional analysis

All 214 differentially expressed genes (combined from genotype and genotype × time analysis) were classified according to their functional categories derived from MatDB ( The global analysis indicated that an overrepresentation of differentially ethylene-regulated genes could be observed in the following functional classes: metabolism; cell rescue, defence and virulence; subcellular localization; protein binding and cofactor requirement (data not shown). But the degree of representation of functional classes differed from cluster to cluster. The representation of functional classes of the three largest clusters derived from AQBC-clustering (wherein the up- and downregulated cluster from the genotype and the genotype × time analysis were combined, that is, clusters 1a and 1b, 2a and 2b were taken together) are shown in Figure 4. To evaluate the importance of the predominant functional classes in a given cluster, the percentage of the genes belonging to the defined functional group was compared with that of all genes present on the array. In cluster 1 (corresponding to the combination of cluster 1a and 1b) (Figure 4), representing the main group of downregulated genes, genes involved in cell rescue, defence and virulence, as well as metabolic genes and genes encoding proteins that facilitate transport are clearly overrepresented. In cluster 2 (upregulated genes, corresponding to the combination of cluster 2a and 2b) there is again a clear presence of metabolic and defence genes, albeit to a much lesser extent as in cluster 1 for the defence genes. These observations indicate that ethylene negatively regulates some metabolic processes while other metabolic processes are induced in the presence of ethylene. In addition, ethylene mainly not only decreases the transcription of defence genes but also induces some of these genes. In contrast, in cluster 3; containing genes that are early modulated by ethylene (corresponding to cluster 3a), other functional classes are overrepresented besides the metabolic genes. Early ethylene-regulated genes particularly include genes needed for protein synthesis and protein activity. Moreover, genes involved in cellular organization are also activated very early by ethylene. Altogether, it is clear that different classes of genes are regulated by ethylene at early compared with later time-points.

Figure 4.

Graphical representation of the percentage of genes belonging to a given functional class for the major AQBC clusters.
The three largest clusters are presented here (a, b, c). The percentage within a cluster (white bars) was compared to the percentage of all genes present on the array belonging to the corresponding functional class (black bars). Y-axis: percentage; MET, metabolism; CELL RESC, cell rescue and defence and virulence; SUBC LOC, subcellular localization; PROT BIND, protein with binding function or cofactor requirement, TRANSP, transport facilitation; UNCLASS, unclassified proteins/classification not yet clear-cut; PROT F, protein fate (folding, modification, destination); PROT S, protein synthesis; CONT CELL ORG, control of cellular organization.

In addition to the functional analysis, we made use of the expression data that became publicly available from the GARNET facility at Nottingham University, UK (, based on hybridization of Affymetrix chips covering 22.746 Arabidopsis genes. Expression data for different tissues were automatically analysed using a specially designed algorithm (D. Zadik and M. Bennett, unpublished data). It was surprising that about 25% of the analysed ethylene-modulated genes are highly expressed in roots (Figure S2). The second main group of genes (about 25%) is ubiquitously expressed whereas smaller groups of genes are predominant in specific tissues.

Ethylene regulation of genes involved in the ubiquitin-mediated degradation pathway

Recent analyses have connected individual components of the ubiquitin/26S proteasome pathway to almost all aspects of plant development, including hormone signalling (Kepinski and Leyser, 2002; McGinnis et al., 2003; McGinnis et al., 2003; Smalle et al., 2002, 2003). Two genes involved in the ubiquitin degradation pathway in plants, UBC10 and UBQ14, were early induced by ethylene in our cDNA-AFLP experiment. This observation led to the hypothesis that ubiquitin-mediated degradation is possibly involved in ethylene signalling. To find support for this hypothesis, we investigated the response to ethylene of genes belonging to different families of components involved in ubiquitin-mediated proteolysis (Table S1) with RT-PCR. PCR amplifications were performed on the preamplification products of wild type, ein2-1 and ctr1-1 treated with 10 ppm and 100 ppb of ethylene, respectively (see Experimental procedures). In total, 22 transcripts were investigated in samples from both independent treatments and those for which an ethylene-regulated pattern was observed are shown in Figure 5. Three transcripts belonging to the family of ubiquitin-specific proteases (UBPs) exhibited ethylene-inducible expression. UBP4 displayed a higher, albeit transient expression after 10 and 30 min of treatment in the wild type. UBP23 was upregulated by ethylene in time. The mRNA level in the ctr1-1 mutant was higher compared with the baseline expression level in the ein2-1 mutant. For UBP27 higher expression was observed in the presence of ethylene, while expression remained unaffected in ein2-1 and expression was elevated in ctr1-1. Furthermore, one gene encoding a putative ubiquitin-conjugating enzyme (UBC), possibly functioning as an E2 enzyme in the process of ubiquitylation, was found to be ethylene-regulated. This putative UBC (At2g16920) exhibited a transient peak in expression level within 20 min of exposure to the hormone. Moreover, this gene displayed a higher mRNA level in the ctr1-1 mutant, comparable with that seen for the peaks of expression in the wild type.

Figure 5.

Analysis of the steady state messenger levels of genes encoding components of the ubiquitin-mediated protein degradation pathway after treatment with 10 ppm and 100 ppb of ethylene.
Amplifications were performed on the preamplification products for cDNA-AFLP analysis. Specific oligonucleotide primers were used to detect the transcript levels of the genes listed in Table 1.
(a) An EtBr stained gel of the amplification products for UBP4 (20 cycles), UBP23 (20 cycles), UBP27 (27 cycles) and a putative UBC At2g16920 (20 cycles) from the 10 ppm treatment is shown. CAD was used as a control (30 cycles) (Moshkov et al., 2003).
(b) An EtBr stained gel of the amplification products for UBP4 (20 cycles), UBP23 (25 cycles), UBP27 (27 cycles) and a putative UBC At2g16920 (25 cycles) from the 100 ppb treatment is shown. CAD was used as a control (30 cycles).


In this study, the transcriptional changes upon short ethylene treatment were studied using a pilot cDNA-AFLP experiment and an Arabidopsis microarray. An estimated 5% of the transcriptome corresponding to 1200 transcript tags was scanned by cDNA-AFLP. About 4% of these were ethylene responsive. However, with the microarray technique about one-fourth of the Arabidopsis transcriptome was investigated. This analysis yielded 214 (out of the genotype-effects and the genotype × time-effects) ethylene-modulated genes, representing 3.3% of all genes present on the chip. Thus, despite the fact that two different transcript profiling techniques were used, both analyses resulted in a similar percentage of differentially ethylene-regulated genes. Therefore, the threshold of a CV >0.5 used in our cDNA-AFLP analysis corresponds roughly to a twofold difference in expression as applied in the microarray analysis. cDNA-AFLP is a PCR-based, transcript profiling technology that combines both the feature of high specificity and the capability of detection of rare transcript tags; therefore, its sensitivity is higher that of hybridization-based techniques. Conversely, the strength of microarrays lies in the massive parallel nature of the analysis, allowing the simultaneous analysis of up to tens of thousands of genes. Recently, Reijans et al. (2003) reported a good correlation between cDNA-AFLP and microarray results.

As expected, known ethylene-regulated genes were picked up in both analyses (see Table 1) displaying similar expression patterns as previously described. Furthermore, several genes identified by cDNA-AFLP were also present on the 6K-chip and displayed similar expression profiles (see Table 1). Clustering in both analyses yielded comparable groups of co-regulated genes consisting of very early, early, later up- and downregulated clusters. From the functional analysis of these differentially expressed genes, some major functional groups appeared which provided novel insights in the early role of ethylene signalling. The most interesting observations are discussed below.

Ethylene in plant disease resistance and abiotic stress

Besides its physiological role at different developmental stages, ethylene is also a stress hormone. Its synthesis is induced by a variety of stress signals, such as mechanical wounding, chemicals and metals, drought, extreme temperatures and pathogen infection (Johnson and Ecker, 1998). For the latter, depending on the plant species, and the type of pathogen and its offensive strategies, the role of ethylene can be very different. Ethylene-insensitive signalling mutants may show either increased susceptibility or increased resistance (Asai et al., 2000; Bent et al., 1992; Berrocal-Lobo et al., 2002; Greenberg et al., 2000; Thomma et al., 1999). This apparent discrepancy in different plant–pathogen interactions may be reconciled by the different infection mechanisms of pathogens, and by the fact that ethylene is not only involved in pathogen response, but is also in many aspects of plant development including senescence, cell death and ripening (Abeles et al., 1992). Besides its involvement in pathogen infection, ethylene is also implicated in response to abiotic stress. An enhanced ethylene emanation is one of the earliest responses to ozone stress (Moeder et al., 2002; Vahala et al., 2003). It is suggested that ethylene is involved in the regulation of cell death by amplifying ROS production, which is responsible for the execution of spreading cell death (Overmyer et al., 2000). Compared with the two other hormones involved in responses to abiotic stress and pathogen defence, JA and SA, ethylene is involved in the early responses whereas JA and SA may control more prolonged effects. In agreement with this hypothesis, a large group of genes involved in defence and disease were rapidly affected by ethylene in our analysis. Interestingly, these defence genes were mainly downregulated by ethylene (see Figure S3) which is contrary to what is described for the defence genes PDF1-2, and the PR-genes PR-3 and PR-4 in Arabidopsis (Penninckx et al., 1996). This discrepancy is probably results from the emphasis on early time-points whereas other studies focused on longer treatments with ethylene. In accordance with our findings, the defence gene catalase was downregulated in tomato after 15 min of ethylene treatment (Zegzouti et al., 1999).

The heat shock protein (HSP) 101, a superoxide dismutase, a catalase 3-homologue, an HSR201-like, two putative disease resistance response genes, an osmotin precursor, a β-glucanase precursor and a putative endochitinase were downregulated by ethylene or displayed a lower level in the wild type compared to ein2-1. The HSR201 (hypersensitivity-related)-like protein was isolated during an incompatible interaction between tobacco and the bacterial pathogen Pseudomonas solanacearum (Czernic et al., 1996). Interestingly, HSR201 is homologous to an ethylene-inducible tomato gene involved in fruit ripening and senescence. All major HSPs have related functions, protecting proteins against misfolding and aggregation. The expression of HSP101 was downregulated very early on and the amount of cpHsc70-1mRNA was higher in wild type compared with ein2-1. In addition, the small HSP 17.4 gene, isolated from the cDNA-AFLP, displayed an early ethylene-regulated expression profile (see supplemental list cDNA-AFLP).

One striking result was that a fair number of peroxidase genes were downregulated by ethylene or displaying a significantly higher expression level in ein2-1. All these peroxidases belong to the classical class III family of peroxidases which are targeted via the endoplasmic reticulum (ER) either extracellullarly or to the vacuole. They are ascribed a variety of functional roles, including lignification, suberization, auxin metabolism, defence, stress and developmentally related processes (Welinder et al., 2002). Together with the other downregulated defence genes, these peroxidases can be involved in the amplification of the cell death signal by loosening cell walls and inhibiting H2O2 scavenging, which leads to local resistance responses. Moreover, the mRNA for the phytocyanin blue copper-binding protein was also downregulated by ethylene and could have an analogous role as some phytocyanins are involved in redox reactions occurring during primary defence responses in plants and/or in lignin formation (Nersissian et al., 1998). A similar pattern was observed for the mRNA of thioredoxin H-type 3 (TRX-H-3). It was demonstrated that TRX-H-3 induces H2O2 tolerance in Saccharomyces cerevisiae, suggesting that these isoforms could act as antioxidants possibly by serving as hydrogen donors for a thioredoxin-dependent peroxidases (Brehelin et al., 2000; Verdoucq et al., 1999). Recently, a new antioxidant route is described for the action of the enzyme methionine sulfoxide reductase (Nakao et al., 2003). In our study, the methionine sulphoxide reductase (SelR) domain protein also displays a downregulated pattern. Finally, a superoxide dismutase, known as an antioxidant, was downregulated, whereas a Cu/Zn-superoxide dismutase copper chaperone precursor (CCS) (picked up in both analyses) and a Cu/Zn superoxide dismutase was early induced by ethylene.

Another group of ethylene-regulated genes encode enzymes in intermediate steps in lignin biosynthesis. These include two cinnamoyl-coA reductase-like genes and the cinnamyl-alcohol dehydrogenase ELI3-1. For both enzymes other isozymes have been reported to function in defence and wounding response (Cheong et al., 2002; Kiedrowski et al., 1992). One of the two cinnamyl-coA reductase-like genes and the ELI3-1 gene are both upregulated by ethylene at the latest time-points of treatment (2–6 h), suggesting a more protective role for ethylene at later stages of defence. Two other genes involved in lignin synthesis in plants are a putative laccase, with an action similar to that of the plastocyanin blue copper-containing protein (Claus, 2003) and a 4-coumarate-CoA ligase-like protein, similar to a key enzyme of phenylpropanoid metabolism (Lee et al., 1995).

A number of transcription factors – amongst which the WRKY DNA-binding proteins – are associated with plant defence responses. Dong et al. (2003) demonstrated that WRKY transcription factor 11 is induced within the first hours after treatment with an avirulent pathogen or with SA. WRKY11 showed little change in transcript level after treatment with ethylene, but the amount of mRNA was about 2.5-fold higher in ein2-1 compared with the wild type, indicating that the involvement of ethylene cannot be ruled out.

Another group of early downregulated defence-related genes are three putative myrosinase-binding proteins (MBPs), two putative lectins, and a putative jasmonate-induced protein of the jacalin lectin family. Jacalin-related proteins have been suggested to be involved in resistance against bacteria, insects and fungi (Chisholm et al., 2000). Myrosinase is an enzyme capable of hydrolysing glucosinolates into various compounds that function in defence (Taipalensuu et al., 1996). In Brassica napus MBPs are upregulated systemically in response to JA during insect or fungal attack.

Downregulation by ethylene is also found for the pleiotropic drug resistance gene PDR9 belonging to the ATP-binding cassette (ABC) transporters. These proteins have been implicated in the transport of antifungal agents (van den Brule and Smart, 2002).

The mRNA that codes for the anthranilate N-benzoyltransferase-like protein was highly upregulated after 1 h of treatment. This enzyme plays a role in biosynthesis of phytoalexins which are important in plant defence against microorganisms (Yang et al., 1997).

Previous transcript profiling studies indicated that cell wall components and enzymes were affected in defence responses (Cheong et al., 2002; Schenk et al., 2000; Van Zhong and Burns, 2003). In our microarray analysis two α-expansins (EXP) were found to be regulated by ethylene. Expansins are cell-wall-loosening proteins that induce stress relaxation and extension of plant cell walls. For EXP5 downregulation was observed after 2 h of treatment. In contrast, the putative EXP11 displayed a rapid twofold transient peak in expression (10 min). This observation strengthens the hypothesis that at early time-points ethylene could play a role in cell-wall loosening while at later time-points it appears to have a protective role enforcing the cell wall (cf. genes involved in lignin biosynthesis). Cosgrove et al. (2002) suggested a wall structural model in which cellulose microfibrils are linked together by a relatively large and inaccessible xyloglucan complex that is opened by EXP, which allows immediate wall extension and access to enzymatic attack by endoglucanases. Thus, the xyloglucan endo-1,4-β-d-glucanase-like protein could be involved in a cooperation with EXP11 as the former is upregulated after 20 min of treatment and reaches its highest level after 1 h, subsequently dropping below control level. In our cDNA-AFLP, another putative xyloglucan endo-transglycosylase (XET) was highly upregulated in the ein2-1 mutant. The arabinogalactan-proteins AGP1 (microarray) and AGP7 (cDNA-AFLP) displayed both an early temporal increase. AGPs are hydroxyproline-rich cell-wall proteins implicated in plant growth and development. In previous research AGP1 was shown to be affected by different forms of stress (Schultz et al., 2002). Finally, a putative pectin methylesterase was downregulated by ethylene and a pectinesterase was highly expressed in the ctr1-1 mutant.

Out of these observations, we can conclude that ethylene is an early active component controlling the damage process at different levels. We hypothesize that ethylene primes the plant cells for amplifying the danger signal in order to contain the damager. Ethylene mainly inhibited several groups of enzymes involved in the early defence response, although the expression of a smaller group of defence genes was stimulated (see Figure S3). The response to ethylene for the latter group of genes was mostly more specific to the later time-points. This reaction can possibly switch on the JA and/or the SA signalling pathways which are involved in more prolonged effects in defence and resistance.

Autoregulation of ethylene biosynthesis and novel interactions between responses to ethylene, ABA, sugars and auxins

In Arabidopsis, the enzyme ACC-oxidase (ACO), which catalyses the final step in ethylene biosynthesis, is present as a multigene family, but little information about these genes has been reported (Gomez-Lim et al., 1993). Four ACC-oxidases (ACO) and one ACC-oxidase-like gene were differentially affected by ethylene in our TP-analysis. In the cDNA-AFLP the upregulated putative ACC-oxidase gene corresponds to ACO2, corroborating previously published data (Gomez-Lim et al., 1993). The expression of two of the ACC-oxidases from the microarray analysis (At1g05010 and At5g43450) peaked after 2 h of treatment, whereas the gene encoding an ACC-oxidase like protein was downregulated in WT, and the fourth ACO (At2g19590) only displayed a higher level in the ein2-1 mutant. Ethylene regulation of its own biosynthesis implicates the hormone as a messenger for the induction of later ethylene-responsive genes or amplification of its signal. Moreover, the differential expression of multiple ACOs proves a more complex autoregulation of ethylene biosynthesis. Furthermore, nitrilase 4 (NIT4) expression was upregulated by ethylene. This gene encodes a β-cyano-1-alanine hydratase/nitrilase and is proposed to take part in cyanide detoxification (Piotrowski et al., 2001). Therefore, NIT4 may play a role in cyanide detoxification during ethylene biosynthesis.

Genetic analysis of ethylene and ABA suggested that these hormones antagonize each other at the level of germination (Beaudoin et al., 2000; Ghassemian et al., 2000). After germination, ABA and ethylene signalling display complex interactions. As both hormones inhibit root growth, they may act in the same or in parallel pathways. However, ethylene-overproducing mutants have decreased ABA sensitivity, implying another antagonistic interaction. One suggested explanation for this apparent inconsistency is that ABA inhibits root growth by signalling through the ethylene response pathway, but is unable to use this pathway in the presence of ethylene (Ghassemian et al., 2000). Furthermore, ethylene does not appear to interfere in ABA-regulated processes such as stomatal closure or induction of some ABA-response genes. In our analysis, both ABI1 and ABI2 were highly upregulated by ethylene with a peak of expression after 1–2 h of treatment. Both genes encode homologous type 2C protein phosphatases acting as negative regulators in ABA signalling (Gosti et al., 1999). As it is known that ABI1 is upregulated by ABA (Leung et al., 1997), we can hypothesize that both hormones work additively in the regulation of transcription of these genes. Upregulation by ethylene was also observed for the ABA responsive element-binding factor (ABRE/ABF3), which is an ABA-inducible bZIP transcription factor and for dehydrin ERD14 (Nylander et al., 2001). The pattern of ABA-inducibility of ABF3 (Choi et al., 2000) corresponds to that seen by ethylene treatment. In addition, nitrate reductase 1 (NR1) displayed a peak in expression after 1 h of treatment. NR is the first enzyme of the nitrate assimilation in plants and hormonal control of the transcription and activity of NR1 has been described. Moreover, NR-mediated NO synthesis in guard cells is required for ABA-induced stomatal closure (Desikan et al., 2002). Possibly the ABA and ethylene pathways are integrated in this response. Although the interaction at this stage of development is not yet clear, it can be concluded that for some responses both hormones work in parallel on the expression of genes involved in ABA signalling.

Characterization of sugar-signalling mutants in Arabidopsis has unravelled a complex signalling network that links sugar responses to ABA and ethylene (Leon and Sheen, 2003; Zhou et al., 1998). Glucose activates ABA biosynthesis, and ABA and glucose antagonize ethylene signalling (Leon and Sheen, 2003). In this perspective the downregulation of expression of a putative hexose transporter and a sugar transporter-like gene, further supports the antagonistic relationship between sugar and ethylene signalling. Alternatively, ethylene regulation of sugar transporters could also result in delayed senescence as sugars are known to repress photosynthetic gene expression (Leon and Sheen, 2003; Rolland et al., 2002). A putative fructokinase displayed a lower level in expression in the wild type compared with ein2-1, indicating that ethylene is also negatively involved in sucrose metabolism.

Interactions between ethylene and auxin pathways were also observed in this study. One striking result is the induction of UDP-glucose:indole-3-acetate β-d-glucosyltransferase (iaglu) upon prolonged ethylene exposure (1–6 h). IAA has been shown to form conjugates with sugars, amino acids and small peptides, which are believed to be involved in IAA transport, in IAA storage, in the homeostatic control of the pool of free hormone, and as a first step in catabolic pathways (Jackson et al., 2001). To our knowledge, this is the first report showing the involvement of ethylene in IAA conjugation. In addition, one of the primary auxin response genes, IAA3 (MIPS description: putative auxin-induced protein AUX2-11), is significantly downregulated by ethylene from 1 h treatment on. IAA3 plays a central role in auxin signalling as a repressor of auxin-regulated gene expression and may also regulate light responses (Tian et al., 2002). Furthermore, in our cDNA-AFLP a putative auxin-regulated protein displayed a high expression level in the ctr1-1 mutant.

Early ethylene response genes playing a role in protein degradation

An interesting aspect that came across in this analysis is the link between ethylene and protein degradation. Ubiquitin-mediated proteolysis has emerged as being fundamentally important in many aspects of development such as hormone signalling, light perception and circadian rhythm, as well as in plant defence signalling (Callis and Vierstra, 2000; Kepinski and Leyser, 2002; Kim and Delaney, 2002; Liu et al., 2002; Xu et al., 2002). In this proteolysis pathway, ubiquitin becomes covalently attached to cellular proteins by an ATP-dependent reaction cascade, which requires three distinct enzymes: ubiquitin-activating enzyme (UBA) (E1), UBC (E2) and ubiquitin-protein ligase (UBL) (E3). UBPs help regulate the ubiquitin/26S proteolytic pathway by generating free ubiquitin monomers from their initial translational products, recycling ubiquitins and/or by removing ubiquitin from specific targets and thus presumably preventing target degradation (Yan et al., 2000). Until now, a role for protein degradation is demonstrated in auxin, jasmonate, cytokinin, absiscic acid and GA signalling (Kepinski and Leyser, 2002; McGinnis et al., 2003; Smalle et al., 2002; Smalle et al., 2003). Very recently, two reports have illustrated the importance of ubiquitin-mediated protein degradation in ethylene signalling. In the absence of ethylene, EIN3, a key transcription factor in ethylene signalling, is quickly degraded through a ubiquitin/proteasome pathway mediated by two F box proteins, EBF1 and EBF2 (for EIN3-binding F box protein 1 and 2) (Guo and Ecker, 2003; Potuschak et al., 2003). These results revealed that a ubiquitin/proteasome pathway negatively regulates ethylene responses by targeting EIN3 for degradation. In our cDNA-AFLP-analysis, the UBC10 (cluster C), coding for an E2 protein, was isolated as an early ethylene response gene, peaking at 1–2 h of treatment. UBQ14, a polyubiquitin gene, was identified in the class of very early ethylene-regulated genes with maximal expression after 10–20 min of treatment (cluster A). Polyubiquitin (UBQ) genes encode precursor proteins which need proteolytic processing by previously described UBPs to release mature ubiquitin which on its turn is covalently attached to substrate proteins usually targeting them for degradation (Bachmair et al., 2001). The response to ethylene for a number of components involved in the process of ubiquitin-mediated proteolysis was further investigated by analysing their steady-state messenger RNA levels (listed in Table S1). Our RT-PCR analysis did not reveal ethylene inducibility of EBF1; rather was the pattern of increasing transcript abundance similar in the wild type and in the insensitive control mutant (Figure S4, panel a). Comparable results were observed in our microarray analysis (Figure S4, panel b). Such expression profile probably reflects circadian control rather than a bona fide regulation by ethylene. This conclusion is in contrast to the interpretation of results published by Guo and Ecker (2003) and Potuschak et al. (2003). One reason could be the difference in tissue and age of the plants used in these analyses. However, the data of Potuschak et al. actually also revealed a clear difference in expression of EBF1 in ein3-1 between the zero and 1 h time-points of ACC treatment. In contrast to EBF1, the EBF2 mRNA displayed an ethylene-regulated pattern in our analysis, corroborating the results of Guo and Ecker (2003), Potuschak et al. (2003) and Gagne et al. (2004).

In addition, three UBPs and one putative UBC exhibited a bona fide ethylene-regulated expression profile in the RT-PCR experiment (Figure 5). At first sight, the involvement of ethylene in the ubiquitination pathway was not strictly confirmed by the microarray results; only one unknown gene similar to the ubiquitin-specific protease 12 ended up in the final list of significantly altered expression patterns. However, a detailed analysis of the data confirmed that several components of the ubiquitination pathway (UBC-like, ubiquitin carboxyl-terminal hydrolase) were responsive to ethylene but did not end up in the final data set because the induction-fold was just below 2. Furthermore, the role of ethylene in protein degradation was confirmed in the microarray by the regulation of a putative cysteine proteinase, a subtilisin proteinase, a putative protease SppA and three putative trypsin inhibitors. The profile of the putative cysteine proteinase was similar to that of the cysteine proteinase RD21A isolated from the cDNA-AFLP. All together, we can conclude that early responses to ethylene include protein degradation.

Early ethylene response genes involved in transcriptional events

Five transcription factors were identified in the cDNA-AFLP and at least 11 genes encoding DNA-binding/transcription factors showed significant differences in expression by ethylene treatment in the microarray analysis. Amongst them are AP2 (2), WRKY (1), Myb (1), Zn-finger (2), bHLH (2), MADS-box containing proteins (1), AtERF2, EIL1, ABF3, IAA3 and the DNA-binding protein GT-1. Of particular interest is the putative RAV-like B3 domain DNA binding protein. The implication of this family of proteins in ethylene signalling has been reported very recently by Alonso et al. (2003).

General conclusion

Our study confirmes the importance of ethylene in the response to environmental stresses and provides new insights into the role of ethylene in these processes. Furthermore, the results reveal novel understanding of the integration of ethylene signalling and two other hormones, ABA and auxin, or sugar signalling; and the role of protein degradation in ethylene signalling. Future characterization of these genes will help to better understand the early response to ethylene and its implication on the later events in the cascade.

Experimental procedures

Plant materials and growth conditions

Arabidopsis thaliana (L.) Heynh. (ecotype Columbia-0) was purchased from Lehle seeds (Round Rock, TX, USA). Ethylene response mutants ctr1-1 and ein2-1, both in Col-0 background, were obtained from the Arabidopsis Biological Resource Center (ABRC) at Ohio State University. Seeds were sown under sterile conditions as described previously (Smalle et al., 1997). The growth medium used was MS/2 (half-strength Murashige and Skoog; Sigma supplemented with 1% sucrose). After sowing, plates were stored at 4°C in the dark for 1 day and then put in a growth chamber at 22°C and 60% relative humidity under white fluorescent light [photosynthetic photon flux density (PPFD): 75 μmol m−2 sec−1] and long day conditions (16 h light/8 h dark). Plants were 19 days old at the time of ethylene exposure.

Ethylene treatments

Plants were placed inside growth chambers dedicated to gas exposures. Control plants and the plants to be treated with ethylene were placed in two adjacent identical growth cabinets. Ambient conditions were 22°C, 60% humidity and white fluorescence light (75 μmol m−2 sec−1) under long day conditions (16 h light/8 h dark). Ethylene (100 ppb or 10 ppm) in air (organic carbon free; Air Liquide Belge N.V., Aalter, Belgium) were flushed through at a flux rate of four refreshments per hour. In order to allow identification of very early response genes, harvesting was performed after 10, 20, 30, 1, 2 and 6 h of treatment. Control samples treated with air were harvested after 10 min and 6 h.

RNA preparation and cDNA-AFLP analysis

Total RNA was extracted from whole plants (19 days old) using TrizolR reagent (GIBCO/BRL, Gaithersburg, MD, USA) according to the manufacturer's instructions. cDNA synthesis and cDNA-AFLP analysis were performed as described by Breyne et al. (2002). For detailed information, visit The first step was the conversion of mRNA into ds cDNA using a biotinylated oligo-dT primer starting from about 10 μg total RNA. The cDNAs were digested with two restriction enzymes in a two-step reaction. After digestion with the first enzyme (BstYI; New England Biolabs, Beverly, MA, USA), the 3′ regions were captured on Dynabeads carrying streptavidin. Digestion with the second enzyme (MseI; New England Biolabs), releases the restriction fragments or transcript tags. The preamplifications were performed with the MseI and BstYI primers – with either T or C as 3′ nucleotide – without selective nucleotides. From a 600-fold dilution of the pre-amplified samples, 5 μl was used for the final selective amplifications using a BstYI primer with two selective nucleotides and an MseI primer with one or two selective nucleotides. Amplification products were separated on 5% polyacrylamide gels using the Sequigel system (Bio-Rad, Hercules, CA, USA). A total of three selective nucleotides resulted in profiles that were not too dense (the average number of tags was about 60), while the sensitivity was already high enough to detect low abundant messengers. Gels were dried on 3MM Whatman paper, exposed to Kodak Biomax Films, and scanned in a PhosphorImager 445 SI (Amersham Biosciences, Little Chalfont, UK).

Quantitative measurement of the expression profiles and data analysis

Gel images were quantitatively analysed with the AFLP-QuantarPro image analysis software (Keygene N.V., Wageningen, the Netherlands) by which all visible AFLP fragments were scored and individual band intensities were measured in each lane. The raw data were corrected for differences by using a total lane intensity correction. To that end, the intensity values were summed per lane for each primer combination and each of the sums was divided by the maximal value to yield the correction factors. Finally, all raw data were divided by these correction factors. For the corrected data, the CV value (=SD/mean) for each tag was calculated and a threshold of 0.5 was chosen for significant differentially expressed genes. The higher the CV value, the higher the differences in expression levels. Forty-nine genes were found to display a significantly (with a coefficient of variance >0.5) altered expression pattern upon ethylene exposure for the 10 ppm treatment, 48 genes for the 100 ppb treatment, respectively. Only those genes that displayed a similar pattern in both independent experiments were taken for further analysis, resulting in a total of 46 ethylene-regulated genes. Hierarchical clustering was performed using the Cluster and Treeview program (Eisen et al., 1998).

Characterization of AFLP fragments

Bands corresponding to differentially expressed genes were cut out from the gel and eluted DNA was re-amplified under the same conditions as for the selective amplification (selectivity +2/+2 or +2/+1). Sequence information was obtained either by direct sequencing of the re-amplified product with the BstYI or MseI primer or after cloning the fragments in pGEM-T easy (Promega, Madison, CA, USA) and sequencing of three individual clones. Only when the three sequences were identical, they were included for further analysis. The obtained sequences were compared to nucleotide and protein sequences in the publicly available databases by blast sequence alignments (nblast and tblastx).

Microarray preparation

The microarray consisted of 6008 Arabidopsis genes composed from the unigene clone collection from Incyte (Arabidopsis Gem I, Incyte, Wilmington, DE, USA) and 520 positive and negative controls (for details see All 6528 clones were spotted in duplicate on a single array. The cDNA inserts were PCR amplified using M13 primers, purified with Multiscreen-PCR plate (cat: MANU03050; Millipore, Brussels, Belgium) and arrayed in 50% DMSO on Type VII silane coated slides (cat# RPK2331; Amersham Biosciences, Buckinghame, UK) using a Molecular Dynamics Generation III printer (Amersham BioSciences). Slides were blocked in 2× SSPE, 0.2% SDS for 30 min at 25°C.

Microarray hybridization

Total RNA was extracted from the samples using TrizolR reagent (GIBCO/BRL, Gaithersburg, MD, USA) according to the manufacturer's instructions.

A minimum of 5 μg total RNA was linearly amplified using in vitro transcription as described in detail in Puskas et al. (2002). The probes were resuspended in 210 μl hybridization solution containing 50% formamide, 1× HybridizationBuffer (cat# RPK0325; Amersham BioSciences), 0.1% SDS and 60 μg ml−1 poly-dT. Hybridization and post-hybridization washing was performed at 45°C using an Automated Slide Processor (ASP) (Amersham BioSciences). Post-hybridization washing was performed in 1× SSC, 0.1% SDS, followed by 0.1× SSC, 0.1% SDS and 0.1× SSC. The complete ASP program can be downloaded from

Data analysis

Arrays were scanned at 532 nm and 635 nm using a Generation III scanner (Amersham BioSciences, UK). Image analysis was performed with ArrayVision (Imaging Research Inc, Ontario, Canada). A reference design was applied, including reciprocal labelling (dye-swap) of all samples. As a reference, we used a pool of ctr1-1 samples (0–10 min to 6-h treated samples). Altogether, this accounted for a total of 28 DNA microarrays. Spot intensities were measured as artefact-removed total intensities (ARVol) without correction for background. For 24 negative control spots containing a Bacillus subtilis-specific cDNA and 6008 Arabidopsis spots, we first addressed within-slide normalization by plotting for each single slide a ‘MA-plot’ (Yang et al., 2002) where M = log2 (R/G) (where R = red fluorescent dye Cy5 and G = green fluorescent dye Cy3) and A = log2 √R × G for each spot. By means of data preparation, ‘LOWESS’ normalization was applied, with a value of 0.2 for the smoothing parameter to correct for dye intensity differences. Based on the M′ (adjusted M) and A values for each gene, adjusted log2 R and log2 G signal intensities were obtained.

For the 96 adjusted log2 R and log2 G signal intensities of the negative control spots, the median and the 95 percentile were calculated. The 95 percentile was arbitrarily defined as the signal threshold. For each gene, the adjusted log2 R and log2 G signal intensities were compared with the signal threshold; 747 genes were below the signal threshold in at least 25% of the number of observations in the 14 genotype-time samples (n = 56) per gene and were subsequently removed from the data set. All values of the remaining 5259 genes below the 95 percentile threshold were reset to the median value of the negative control intensities. Mixed anova models, in which some effects are considered fixed and others are considered random, have been used as described previously by Wolfinger et al. (2001): let yiklm be the base-2 logarithm of the ‘LOWESS’-transformed spot measurement from gene i (i = 1, …, 5259); we first applied a linear normalization anova model of the form yiklm = μ + Ak + (ADR)klm + εiklm to estimate global variation of the collection of i selected cDNA fragments in the form of random array effects (Ak; k = 1, …, 28), random replicates within array × dye combinations, or channel-replicate effects [(ADR)klm; k = 1, …, 28, l = 1, …, 2, m = 1, …, 2] and random error εiklm. Residuals, computed by subtracting the fitted values for the effects from the yiklm, were then subjected to 5259 gene-specific models of the form rjgklmn = μ + Dl + Sj + (STjn + (SGjg + (SGTjgn + Ak + γjgklmn partitioning gene-specific variation into fixed gene-specific dye effects (Dl), fixed sample effects [Sj, j = 1, …, 2; control (ctr1-1) and test samples (Col-0 and ein2-1)], fixed time effects [(STjn, n = 1…7), fixed genotype effects ((SG)jg, g = 1…2; Col-0 and ein2-1], fixed genotype × time effects (effect of interest) [(SGT)jgn], random spot effects (Ak) and random error γjgklmn.

To test differences between the two genotypes Col-0 and ein2-1, the seven time-points and the 14 genotype × time effects, we used Wald statistics, which should follow approximate χ2-distributions under the null hypothesis with degrees of freedom (d.f.) equal to 1, 6 and 6, respectively. Expression differences with P-values ≤0.001 were called significant resulting in 476, 1368 and 231 genes showing significant genotype, time and genotype × time effects on their expression, respectively. Genes affected in their expression across the seven time-points in a non-genotype-specific manner were considered as affected by circadian rhythm and were therefore ruled out for further analysis. An additional selection of differentially expressed genes was based on a twofold-change criterion resulting in a set of 57 genes only affected in a genotype-specific manner (average values of wild type samples and ein2-1 samples were compared) and 157 significantly (P < 0.001) ethylene-regulated genes showing a twofold-change in at least one time-point in Col-0 but not in ein2-1.

All standard calculations and statistics, including the LOWESS fit were carried out using Genstat (Genstat Release 6.1 for Windows; VSN International, Hemel Hempstead, UK (Payne and Arnold, 2000)). We used REML procedure as implemented in Genstat to perform both the normalization and gene model fits. The Genstat ‘VWALD’ procedure (Goedhard and Thissen, was used to save the non-hierarchical Wald test for the fixed terms in the REML analysis.

Hierarchical clustering was performed using the Cluster and Treeview program (Eisen et al., 1998). Quality-based clustering was performed with a recently developed software program (De Smet et al., 2002). This program is similar to K-means clustering, except that the number of clusters need not be defined in advance and the expression profiles that do not fit in any cluster are rejected. The minimal number of tags in a cluster and the required probability of genes belonging to a cluster were set to 5 and 0.95, respectively.

RT-PCR analysis

A 200-fold dilution of the cDNA-AFLP preamplifications of the samples treated with 100 ppb and 10 ppm was used for each RT-PCR reaction (Vandenbussche et al., 2003). Five microlitres of the diluted preamplification product was used as a template. The PCR amplification cycle was as follows: 95°C for 30 sec, 56°C for 30s, 72°C for 30 sec. Samples were taken after 20, 25 and 27 cycles and 15 μl of the PCR product was visualized on an 1% agarose gel. All PCRs were carried out in a Mastercycler (Eppendorf, Hamburg, Germany). DNA was stained with ethidium bromide in the gel. A list of the gene-specific primers is given in Table S1.

A series of oligonucleotide primers were designed for specific amplification of genes involved in the ubiquitin-mediated degradation pathway. The selection was primarily based on the presence of an ethylene-responsive element in the 500 bp-region upstream of the start of the respective gene [list of genes presented in Yan et al. (2000) and Bachmair et al. (2001)]. This was the case for 38 out of the 86 genes analysed. Secondly, only those transcripts containing a BstYI and a MseI-cutting site could be analysed, resulting in the analysis of 22 transcripts (see Table S1). CAD transcript accumulation was used as an internal control (Moshkov et al., 2003) for the detection of early ethylene regulation (forward primer: CATGGGAGTTATCAACAATCCA, reverse primer: CATAATCCATCTTCACAACTTCG). For the internal control CAD, 30 cycles were run, according to Moshkov et al. (2003).


We thank Wilson Ardiles-Diaz, Caroline Buysschaert, Rebecca De Clercq, Jan Gielen and Raimundo Villaroel for sequencing the transcript tags, Els Fostier for technical assistance, Steven Vercruysse for the bioinformatics support, Daniel Zadik and Malcolm Bennett for providing us the algorithm for expression data in different tissues and Wim Vriezen for critical reading and helpful discussions. This work was supported by a PhD fellowship to Annelies De Paepe and research grants (G.0281.98 and G.0345.02) to Dominique Van Der Straeten from the Fund for Scientific Research (Flanders).

Supplementary material

The following material is available from

Figure S1. Data analysis and reproducibility of microarray data.

Figure S2. Tissue-based distribution of ethylene-modulated genes.

Figure S3. Clustering of differentially expressed genes involved in defence responses.

Figure S4. Analysis of the steady state messenger levels of EBF1 and EBF2.

Table S1 Primer sequences used in RT-PCR analysis.

Table S2 List cDNA-AFLP results

Table S3 List AQBC-clusters-genotype difference

Table S4 List AQBC-clusters-genotype x time-effect

Table S5 List average linkage clustering-genotype difference

Table S6 List average linkage clustering-genotype x time-effect