Molecular phenotyping of lignin-modified tobacco reveals associated changes in cell-wall metabolism, primary metabolism, stress metabolism and photorespiration


  • Rebecca Dauwe,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology, Technologiepark 927, 9052 Gent, Belgium,
    2. Department of Molecular Genetics, Ghent University, Technologiepark 927, 9052 Gent, Belgium,
    Search for more papers by this author
    • Present address: Wood Science Department, The University of British Columbia, Vancouver, BC V6T 1Z4, Canada.

  • Kris Morreel,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology, Technologiepark 927, 9052 Gent, Belgium,
    2. Department of Molecular Genetics, Ghent University, Technologiepark 927, 9052 Gent, Belgium,
    Search for more papers by this author
  • Geert Goeminne,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology, Technologiepark 927, 9052 Gent, Belgium,
    2. Department of Molecular Genetics, Ghent University, Technologiepark 927, 9052 Gent, Belgium,
    Search for more papers by this author
  • Birgit Gielen,

    1. Department of Biology, University of Antwerp, Universiteitsplein 1, 2610 Wilrijk, Belgium,
    Search for more papers by this author
  • Antje Rohde,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology, Technologiepark 927, 9052 Gent, Belgium,
    2. Department of Molecular Genetics, Ghent University, Technologiepark 927, 9052 Gent, Belgium,
    Search for more papers by this author
  • Jos Van Beeumen,

    1. Department of Biochemistry, Physiology and Microbiology, Ghent University, K.L. Ledeganckstraat 35, 9000 Gent, Belgium,
    Search for more papers by this author
  • John Ralph,

    1. US Dairy Forage Research Center, USDA–Agricultural Research Service and Department of Biological Systems Engineering, University of Wisconsin-Madison, Madison, Wisconsin 53706-1108, USA,
    Search for more papers by this author
  • Alain-Michel Boudet,

    1. Surfaces Cellulaires et Signalisation chez les Végétaux, Centre National de la Recherche Scientifique-Université Paul Sabatier, Unité Mixte de Recherche 5546, Pôle de Biotechnologie Végétale, 24 chemin de Borderouge, BP 42617, 31326 Castanet, France,
    Search for more papers by this author
  • Joachim Kopka,

    1. Max-Planck Institute of Molecular Plant Physiology, Cooperative Research Group, Am Mühlenberg 1, 14476 Golm, Germany
    Search for more papers by this author
  • Soizic F. Rochange,

    1. Surfaces Cellulaires et Signalisation chez les Végétaux, Centre National de la Recherche Scientifique-Université Paul Sabatier, Unité Mixte de Recherche 5546, Pôle de Biotechnologie Végétale, 24 chemin de Borderouge, BP 42617, 31326 Castanet, France,
    Search for more papers by this author
  • Claire Halpin,

    1. University of Dundee, Plant Research Unit at the Scottish Research Institute, Invergowrie, Dundee DD2 5DA, UK
    Search for more papers by this author
  • Eric Messens,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology, Technologiepark 927, 9052 Gent, Belgium,
    2. Department of Molecular Genetics, Ghent University, Technologiepark 927, 9052 Gent, Belgium,
    Search for more papers by this author
  • Wout Boerjan

    Corresponding author
    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology, Technologiepark 927, 9052 Gent, Belgium,
    2. Department of Molecular Genetics, Ghent University, Technologiepark 927, 9052 Gent, Belgium,
      (fax +32 9 3313809; e-mail
    Search for more papers by this author

(fax +32 9 3313809; e-mail


Lignin is an important component of secondarily thickened cell walls. Cinnamoyl CoA reductase (CCR) and cinnamyl alcohol dehydrogenase (CAD) are two key enzymes that catalyse the penultimate and last steps in the biosynthesis of the monolignols. Downregulation of CCR in tobacco (Nicotiana tabacum) has been shown to reduce lignin content, whereas lignin in tobacco downregulated for CAD incorporates more aldehydes. We show that altering the expression of either or both genes in tobacco has far-reaching consequences on the transcriptome and metabolome. cDNA-amplified fragment length polymorphism-based transcript profiling, combined with HPLC and GC–MS-based metabolite profiling, revealed differential transcripts and metabolites within monolignol biosynthesis, as well as a substantial network of interactions between monolignol and other metabolic pathways. In general, in all transgenic lines, the phenylpropanoid biosynthetic pathway was downregulated, whereas starch mobilization was upregulated. CCR-downregulated lines were characterized by changes at the level of detoxification and carbohydrate metabolism, whereas the molecular phenotype of CAD-downregulated tobacco was enriched in transcript of light- and cell-wall-related genes. In addition, the transcript and metabolite data suggested photo-oxidative stress and increased photorespiration, mainly in the CCR-downregulated lines. These predicted effects on the photosynthetic apparatus were subsequently confirmed physiologically by fluorescence and gas-exchange measurements. Our data provide a molecular picture of a plant’s response to altered monolignol biosynthesis.


Lignin is an aromatic heteropolymer that is mainly present in the walls of secondarily thickened cells (Boerjan et al., 2003; Ralph et al., 2004). It is generated from the oxidative polymerization of monolignols, and makes the wall rigid and hydrophobic, allowing transport of water and solutes through the vascular system. In addition to its structural role in the vascular tissues, lignin also protects plants against invading pathogens. The lignin biosynthetic pathway has attracted significant research attention because lignin is a limiting factor in a number of agro-industrial processes, such as chemical pulping, forage digestibility, and the processing of lignocellulosic plant biomass to bioethanol (Baucher et al., 2003; US Department of Energy, 2006). Consequently, most of the genes involved in the monolignol biosynthetic pathway have been cloned and their function assigned through genetic modification and the characterization of mutants (Boerjan et al., 2003). Although analyses of the consequences of altering the expression of these genes have mostly been restricted to the study of lignin amount and composition, more recent transcript profiling of Arabidopsis thaliana pal, cad and c3h mutants, maize (Zea mays) bm3 mutants, and antisense 4CL1 and F5H-overexpressing aspen (Populus sp.) indicates far-reaching consequences of the mutations on various pathways (Abdulrazzak et al., 2006; Ranjan et al., 2004; Robinson et al., 2005; Rohde et al., 2004; Shi et al., 2006; Sibout et al., 2005). Hence, it is conceivable that the observed effects on the cell wall in transgenic plants with altered monolignol biosynthesis are not merely due to the reduced flux through the altered biosynthetic step, but are the combined outcome of a much more complex interplay of various metabolic pathways. In addition, adverse effects on plant growth in lignin-modified plants may be caused by secondary effects, rather than defects in lignin biosynthesis per se. Indeed, some transgenic tobacco (Nicotiana tabacum) lines with reduction in lignin amount have growth defects, whereas other lines with comparable reductions develop normally (Chabannes et al., 2001a,b). In Arabidopsis, it has been shown that growth defects in hct mutants are attributable to elevated flavonoid levels (Besseau et al., 2007). Understanding the molecular mechanisms that plants use to respond to cell-wall defects is essential in order to design plant cell walls with improved properties without compromising plant health and viability (Möller, 2006).

The last two steps in the biosynthetic pathway of the monolignols are catalysed by cinnamoyl CoA reductase (CCR) and cinnamyl alcohol dehydrogenase (CAD). In tobacco, only one cDNA has been identified so far for CCR (O’Connell et al., 2002; Piquemal et al., 1998) and two for CAD (Knight et al., 1992). The two CAD cDNA sequences show strong homology both in the coding region (94%) and in the 5′ and 3′ untranslated regions (88% and 82%, respectively), supporting the hypothesis that these genes are derived from the two ancestral species, N. sylvestris and N. tomentosiformis, rather than representing two distinct CAD genes. Transgenic tobacco plants downregulated for the corresponding genes have already been thoroughly characterized for alterations in morphology, lignin amount and composition, and cell-wall structure (Chabannes et al., 2001a,b; Halpin et al., 1994; O’Connell et al., 2002; Piquemal et al., 1998; Ralph et al., 1998). Downregulation of CCR in tobacco results in reduced lignin content. Lignin in these plants is characterized by a reduced incorporation of cinnamaldehydes and elevated levels of ferulate and tyramine ferulate. Xylem vessels are collapsed, leaves are dark-green and wrinkled, and, in addition, strongly downregulated lines are smaller in size with yellowish areas between the leaf veins. Tobacco plants downregulated for CAD have higher cinnamaldehyde levels in lignin, but normal lignin content and overall plant appearance. Double transformants with reductions in the expression of both CCR and CAD have less lignin but no collapsed vessels. All three types of transgenic lines have a reddish-colored xylem, attributable to the presence of unusual phenolics during lignin polymerization. Because these lines are so well characterized, they are ideal models to study how the various phenotypes are elaborated at the molecular level and how plants respond to cell-wall defects. Here, we analysed the transcriptome and metabolome of these transgenic lines. Our data show that, depending on the modified step, alteration of monolignol biosynthesis can affect the transcript levels of other monolignol biosynthesis genes, and additionally those of starch and hemicellulose metabolism, respiration, photorespiration and stress pathways. The suggested effects of lignin modification on photorespiration and oxidative stress were further corroborated by physiological measurements. Together, our combined transcript and metabolite data reveal how monolignol and cell-wall biosynthesis are integrated into general metabolism, and show that perturbations of such processes are sensed and partially compensated for by plants.


Comparative cDNA-AFLP profiling of xylem from CCR, CAD and double downregulated tobacco

cDNA-amplified fragment length polymorphism (AFLP)-based transcript profiling was performed on two pools of scraped xylem tissue, derived from wild-type tobacco (WT) and transgenic tobacco deficient in CCR (asCCR), CAD (asCAD) or both (DT) (Figure 1). Both single and double transformants developed normally and were hemizygous for the antisense constructs. Gene expression was surveyed with 189 primer combinations, yielding a total of approximately 11 000 transcript-derived fragments (TDFs) ranging from 50 to 500 bp. To determine the TDFs that accumulated differentially between the lines (dTDFs), the cDNA-AFLP gels were analysed by a combination of visual scoring and statistical analyses; 365 non-redundant bands had a significantly altered intensity in one or more transgenic lines, 186 of which were proven, after excision, re-amplification and sequencing, to represent a single transcript (Table 1; Tables S1 and S2). All further analyses were based on this set of 186 dTDFs.

Figure 1.

 cDNA-AFLP transcript profiling of WT, asCCR, asCAD and DT tobacco xylem.
A typical cDNA-AFLP gel is shown. Amplification was performed with BstYIT + 2 and MseI + 1 primers. Gel sections containing dTDFs (a, b, c, d) that have been excised and re-amplified are enlarged. Two pools (lanes 1–8) are presented for each profile.

Table 1.   Identities, expression patterns and functional classification of the dTDFs isolated by cDNA-AFLP transcript profiling of WT, asCCR, asCAD and DT tobacco
dTDFBest hitDescriptionMean band intensityArabidopsis closest homologue
WTasCCRasCADDTAGIE-value (blastp)
  1. For each dTDF, the nearest related gene as obtained by homology search in public databases (see Appendix S1) is given. The identification code and description of the genes was retrieved from the Swiss-Prot database (sp) or from the computer-annotated supplement to Swiss-Prot that contains all the translations of EMBL nucleotide sequence entries not yet integrated in Swiss-Prot, TrEMBL (spt). Transcript levels were scored in the WT and asCCR, asCAD and DT transformants, and the mean band intensity values are given per genotype. One-way anova and post hoc LSD tests were carried out. Values significantly different from the WT value at < 0.05 are indicated by asterisks, with italic numbers indicating lower transcript levels, and bold numbers higher transcript levels. The closest Arabidopsis homologue of these genes was searched by sequence comparison at the amino acid level (BLASTP). AGI, Arabidopsis Genome Initiative. dTDFs belonging to the classes of ‘expressed protein’, ‘hypothetical protein’, ‘unknown protein’ and ‘no hit’ are presented in Table S2.

1 Metabolism, 18 entries (9.7%)
1.1 Amino acid metabolism, six entries (3.2%)
 C32M4_019spt:Q94BP8Putative alanine aminotransferase (fragment), ACC synthase435441 77*411 At1g17290.10
 C34M4_038spt:Q83JW4Hypothetical protein csdA, Aminotransferase, class V6019976*5969 1085*At1g08490.1E-80
 C33M2_033sp:P38561Glutamine synthetase root isozyme 3 (GS1-3)13125156*1340 1247 At5g37600.10
 T31M24_034sp:P54767Glutamate decarboxylase60761707*2111*1383*At5g17330.10
 C13M3_009spt:Q9ZUY3Putative chorismate mutase/ prephenate dehydratase418168*121*132*At2g27820.10
 T41M13_010spt:Q6FA02Phospho-2-dehydro-3-deoxyheptonate aldolase/ 3-deoxy-d-arabino-heptulosonate-7-phosphate synthase70246*300*307*No hit 
1.2 C-compound and carbohydrate metabolism, eight entries (4.3%)
 T34M13_024sp:Q42962Phosphoglycerate kinase, cytosolic2571681*254 1671*At1g79550.20
 T44M3_100spt:Q9AXL6β-glucosidase (fragment)2526749*2528 761*At1g02850.2E-147
 T24M3_013spt:Q9XH69β-amylase (fragment)2531660*1057*3281*At5g18670.1E-170
 T23M2_052spt:Q9FKY9Pectinesterase like protein1319620*2960*577*At5g66920.10
 T44M4_020spt:Q40161Polygalacturonase isoenzyme 1 β subunit (precursor)213   137644*151 At1g70370.10
 C32M21_045spt:Q6IVK7Putative UDP-glucose dehydrogenase 112 5277486*16 798*9135*At5g15490.10
 C42M33_004spt:Q6IVK6Putative UDP-glucose dehydrogenase 21122486*1264 462*At5g15490.10
 C22M4_007spt:Q9STB3ADP-glucose pyrophosphorylase (fragment)2241306*545*107*At2g21590.10
1.3 Lipid, fatty acid and isoprenoid metabolism, three entries (1.6%)
 C32M3_007spt:Q89R91Epoxide hydrolase347129*498*161*At4g02340.1E-44
 C12M4_227spt:Q8VY26Hypothetical protein At4g32810, related to neoxanthin cleavage enzyme259113 964*3237 7883*At4g32810.10
 T22M1_092spt:Q9SP651-deoxy-d-xylulose 5-phosphate synthase 8459  63346627 2671*At4g15560.10
1.4 Secondary metabolism, nine entries (4.8%)
 C12M12_025spt:Q9SDZ22′-hydroxy isoflavone/dihydroflavonol reductase homologue (fragment) DFR11933205*75*224*At2g45400.17.00E-83
 T34M12_050spt:Q42958Caffeic acid O-methyltransferase OMT I4817783*1180*1463*At5g54160.1E-160
 T34M11_054spt:Q42958Caffeic acid O-methyltransferase OMT I10 7525020*4218*4177*At5g54160.1E-160
 T11M1_040spt:Q9SDU04-coumarate:CoA ligase (fragment)86 55643 484*18 239*54 693*At4g19010.12.00E-45
 T33M4_015spt:Q9FSC7Cinnamyl alcohol dehydrogenase326117*80*75*At3g19450.1E-168
 T12M3_027spt:Q8H6 W0Phenylalanine ammonia lyase 11401393*502*2000*At3g53260.10
 T43M1_043spt:Q8H6 V6Phenylalanine ammonia lyase340162*86*80*At3g53260.10
 T21M21_028spt:Q925P9Adenosine kinase (fragment)5098  35468920*4209 At3g09820.22.00E-11
 C42M3_003sp:Q9SF85Adenosine kinase 148302659*6371*2905*At3g09820.10
2 Transcription, eight entries (4.3%)
2.1 mRNA transcription, eight entries (4.3%)
 T34M12_054sp:P06269RNA polymerase α subunit139318 329*2110 1876 No hit 
 C32M23_005spt:Q9FZ15Tuber-specific and sucrose-responsive element binding factor, Myb-like transcription factor26701314*3750 233*At5g67300.17.00E-79
 C34M4_005spt:O22230Putative heat shock transcription factor1832562*586*436*At2g41690.1E-144
 C42M2_031spt:Q9FEL0AP2 domain-containing transcription factor (fragment)2095767*1145*862*At3g16770.14.00E-37
 T21M14_041spt:Q84UB0Transcription factor Myb117591449 470*560*At1g19000.24.00E-57
 T21M2_040spt:Q8S9H7MYB-like transcription factor DIVARICATA101444*82 224*At5g58900.14.00E-86
 T41M3_003spt:Q93YF1Nucleic acid binding protein149224*126 332*At1g47490.1E-142
 T34M12_027spt:Q8S249Putative pumilio/Mpt5 family RNA-binding protein20141897 901*2101 At2g29190.10
3 Protein synthesis,11 entries (5.9 %)
3.1 Ribosome biogenesis, seven entries (3.8 %)
 T24M33_016spt:Q9MAV9Cytoplasmic ribosomal protein S13530624 1704*894*At3g60770.17.00E-80
 T21M44_019spt:Q8L6A3Putative ribosomal protein L2 (fragment)2432178*229 189 At2g44065.28.00E-28
 T21M23_022spt:Q9ATF6Ribosomal protein L1715193052*227*1370 At3g04400.18.00E-77
 T31M44_010sp:P06379Chloroplast 50S ribosomal protein L2531012*73 59 At2g44065.22.00E-30
 T11M4_025spt:Q8S8 V3Chloroplast 50S ribosomal protein L21808173 179*1569 1594 At2g44065.22.00E-30
 T11M1_026spt:Q8S8 V3Chloroplast 50S ribosomal protein L256 8254961*55 897 68 607 At2g44065.22.00E-30
 T23M1_023sp:Q0806940S ribosomal protein S82463498*1181*3227*At5g20290.1E-106
3.2 Translation, three entries (1.6%)
 T23M22_009spt:O82278Putative GTP-binding protein (this protein promotes the GTP-dependent translocation of the nascent protein chain from the A-site to the P-site of the ribosome)646266*631 363*At2g31060.10
 C13M1_044spt:Q9SQV1Putative RNA helicase4591221*378 1223*At3g06480.10
 C23M2_222spt:Q9M9P8T17B22.1 protein, helicase C, Rnase III76115 211*3163 13 946*At3g03300.10
3.3 Aminoacyl-tRNA synthetases, one entry (0.5 %)
 T31M2_015sp:Q9KTX0Histidyl-tRNA synthetase166851*550*475*At3g46100.18.00E-29
4 Protein fate (folding, modification, destination), eight entries (4.3%)
4.1 Protein modification, one entry (0.5%)
 T21M24_045sp:Q42539Protein-l-isoaspartate O-methyltransferase45718016*8183*9368*At3g48330.2E-129
4.2 Proteolytic degradation, seven entries (3.8%)
 C41M1_050spt:Q9SD67FtsH metalloproteinase-like protein206449*70*369*At3g470600
 C43M3_067spt:Q9LM69Hypothetical protein At1g20823, RING-type zinc finger21 0049992*18 520 15 460*At1g20823.1E-111
 T31M2_094spt:Q9M7 K7Ubiquitin protein ligase 210132570*905 2420*At1g55860.10
 C12M2_018spt:Q9M7 K6E3 ubiquitin protein ligase UPL2269678*604*767*At1g70320.10
 T24M32_013spt:Q6K3D2Putative F-box protein FBL26331098*122*719 At3g58530.1E-100
 T41M11_031spt:Q9SB64Hypothetical protein At4g24690, ubiquitin-associated domain (UBA)-containing protein19345404*3179*5142*At4g24690.10
 C12M4_235spt:O65493Papain-type cysteine endopeptidase XCP153204085 12 128*4037At4g353500
5 Cellular communication/signal transduction mechanism, 11 entries (6.0%)
5.1 Transport facilities, three entries (1.6%)
 T41M2_008sp:Q9SU64Cyclic nucleotide- and calmodulin-regulated ion channel 161588285*1556 274*At3g48010.10
 T21M4_036sp:O04289Sulfate transporter 3.2737727 705 185*At4g02700.10
 T24M12_002spt:Q949 J9Hypothetical protein, mechanosensitive ion channel34259*47 27 At1g58200.2E-93
5.2 Intracellular signalling, four entries (2.2%)
 T13M1_078spt:Q9SFE3Putative phosphatidylinositol-4-phosphate 5-kinase423309 1143*536 At1g21920.10
 T33M2_005spt:Q942X2Putative serine/threonine kinase PBS1 protein164261*129 274*At5g18610.10
 T41M3_018spt:Q8RWN3Protein kinase-like protein4556*25*65*At3g25840.10
 C43M2_033spt:Q84TS1Hypothetical protein OSJNBb0097 F01.10, PP2C-like11613958*1836*4093*At2g46920.2E-150
5.3 Transmembrane signal transduction, four entries (2.2%)
 C32M22_018spt:O65238T26D22.12 protein, B_lectin, serine/threonine protein kinase314328 230 1093*At5g35370.10
 C32M21_016spt:Q9FZP3S-receptor kinase185876*71*148*At5g35370.10
 C34M1_029spt:Q9SSL9Highly similar to receptor-like protein kinase474263 3415*387 At1g73080.10
 C34M1_027spt:Q9SSL9Highly similar to receptor-like protein kinase236129 1634*242 At1g73080.10
6 Cell rescue, defence and virulence, 12 entries (6.5%)
6.1 Detoxification, 12 entries (6.5%)
 T34M12_036sp:P20238Metallothionein-like protein 11123249*699*333*At3g09390.14.00E-18
 T32M4_014spt:Q9FNU2Putative ABC transporter transmembrane protein18314594*801*1003*At5g39040.10
 T14M1_023spt:Q9FNU2Putative ABC transporter transmembrane protein194762*224 651*At5g39040.10
 T21M12_018spt:Q9FQE6Glutathione S-transferase GST 129013734*840 3406*At5g62480.12.00E-65
 T24M13_029sp:P25317Probable glutathione S-transferase PARA/auxin-regulated protein PARA32643076 2948 1658*At1g17180.14.00E-83
 T24M13_025sp:P25317Probable glutathione S-transferase PARA/auxin-regulated protein PARA1671491*123 1442*At1g17180.14.00E-83
 C13M2_027spt:Q9FQE6Glutathione S-transferase GST 12154213 115*1363 10 337*At5g62480.12.00E-65
 T44M1_020sp:P25317Probable glutathione S-transferase PARA/auxin-regulated protein PARA95729*59 742*At1g17180.14.00E-83
 T32M1_022sp:P25317Probable glutathione S-transferase PARA/auxin-regulated protein PARA3658649*988 10 001*At1g17180.14.00E-83
 T14M2_036sp:P25317Probable glutathione S-transferase PARA/auxin-regulated protein PARA3331315*463 1246*At1g17180.14.00E-83
 T12M1_020sp:P25317Probable glutathione S-transferase PARA/auxin-regulated protein PARA6842981*636 3852*At1g17180.14.00E-83
 T14M1_030sp:P25317Probable glutathione S-transferase PARA/auxin-regulated protein PARA6857592*1273 6074*At1g17180.14.00E-83
7 Control of cellular organization, five entries (2.7%)
7.1 Cell wall, five entries (2.7%)
 T33M2_020spt:Q9ZP31Expansin precursor142121 744*153 At1g26770.1E-114
 C43M3_079spt:Q8W3 M4Extensin-like protein35513760 10 272*3741 At4g06744.10
 T31M4_058spt:O82337Arabinogalactan protein718754 766*242*At2g46330.19.00E-35
 T13M3_018spt:Q6Z387Lipid transfer protein-like273173*875*207 At2g44300.12.00E-35
 T21M2_038sp:P43643Elongation factor 1-α/vitronectin-like adhesion protein 1 PVN16602152*719 1227 At5g60390.10
8 Light-related, 11 entries (5.9%)
8.1 Photosynthesis, six entries (3.2%)
 C43M34_042spt:Q41003Chlorophyll a/b binding protein (fragment)23112159*4126*2224 At2g34430.13.00E-62
 T21M24_007sp:P27492Chlorophyll a/b binding protein 16, chloroplast (precursor)2341684*5053*1468*At1g29930.1E-137
 T21M44_005sp:P10707Chlorophyll a/b binding protein 1D (fragment)325121*1062*206 At2g34430.12.00E-57
 T12M2_101spt:Q84TM7Chlorophyll a/b binding protein.54931543*10 817* 2889*At2g05100.1E-138
 T13M1_017spt:O64444Light-harvesting chlorophyll a/b–binding protein (precursor)850   973 526* 1229 At1g29930.1E-139
 C32M2_023sp:P13869Chlorophyll a/b binding protein, chloroplast (precursor)350   259 976*  344 At3g61470.1E-130
8.2 Photorespiration, one entry (0.5%)
 C24M1_008spt:Q43775Glycolate oxidase1881  718*2699*  3229*At3g14420.3E-147
8.3 Transcriptional control, two entries (1.1%)
 T21M22_033spt:Q8L5P7LHY protein767 1374 7322*  427 At1g01060.2E-151
 T21M42_046spt:Q8L5P7LHY protein77    63 394*   41 At1g01060.2E-151
8.4 Cellular sensing and response, two entries (1.1%)
 T33M3_035spt:Q9LIE5Phytochrome A signalling protein327  379 5295*  338 At3g22170.10
 T33M3_034spt:Q9LIE5Phytochrome A signalling protein17105270*   872*  2706*At3g22170.10
9. Unclassified proteins, six entries (3.2%)
 T14M3_024sp:P24422Probable aquaporin TIP-type RB7-18C736   86*163*   66*At4g17340.1E-120
 T21M3_077spt:Q93Y38Transmembrane protein FT27/PFT27-like65872513*9765* 8573 At5g36290.2E-161
 T21M3_072spt:Q93Y38Transmembrane protein FT27/PFT27-like5582481*   442   226 At5g36290.2E-161
 T13M2_015spt:Q9A636Pentapeptide repeat family protein291  326427*   86 At5g55000.28.00E-15
 C42M3_056spt:Q6YAT7Actin (fragment)1911062*2317*  3217*At3g12110.13.90E-157
 C43M34_029spt:Q6Z8 N9Putative AT-hook DNA-binding protein182  578*520*  686*At2g33620.34.00E-86

Overall, a relatively stronger effect on the transcriptome was seen in asCCR (145 dTDFs) and in DT (131 dTDFs) compared to asCAD (110 dTDFs; Figure 2a). All genes for which the transcript level was either upregulated in both asCCR and asCAD, or downregulated in both, also displayed the same change in transcript level in DT. For a set of 20 dTDFs, the accumulation was affected in the opposite way in asCCR and DT compared with asCAD. Only six dTDFs accumulated differentially in DT alone.

Figure 2.

 Distribution of dTDFs and metabolites in asCCR, asCAD and DT.
(a) Differential transcripts. Venn diagram indicating the 186 non-redundant dTDFs in the tobacco xylem of asCCR, asCAD or DT based on visual scoring of the cDNA-AFLP gels, anova (< 0.05) and a post hoc LSD test of the quantified values and sequencing. Bold, number of dTDFs with higher transcript level as compared to WT; underlined, number of dTDFs with lower transcript level; italics, number of dTDFs that are oppositely regulated in different lines.
(b) Differential metabolites. Venn diagram indicating the number of metabolites that are differential in asCCR, asCAD and DT, as identified by GC-MS (bold) and HPLC (italics). The numbers for GC-MS include only the metabolites with known identity; those for HPLC include metabolites with known and unknown identity.

Based on sequence homology, 53% (98 dTDFs) of the 186 dTDFs were assigned a putative function, 14% (26 dTDFs) were similar to genes without known function, and for 33% (62 dTDFs), no hit with any sequence was found. Functional classification of the dTDFs (Table 1; Tables S1 and S2) indicated that transcript levels of genes involved in many aspects of both primary and secondary metabolism were affected, and hence that the effect of deficiency in one or both monolignol biosynthesis genes was not restricted to the phenylpropanoid biosynthetic pathway.

Correlation between expression patterns and function of the differentially expressed genes

Using hierarchical clustering, the dTDFs were divided into groups that had similar accumulation patterns across the various transgenic lines, and functional relationships between members of each group were examined (Figure S1). Of all 186 dTDFs, 59% could be divided into five main groups, each consisting of at least 18 dTDFs with highly similar accumulation patterns; the remaining 41% had no such patterns.

Group 1 (30 dTDFs) is the largest, and contains genes that were upregulated similarly in asCCR and in DT but were not or only modestly differentially regulated in asCAD. In this group, nine out of 13 dTDFs with known function belong to the detoxification class, eight of which represent glutathione S-transferases (GSTs) of at least two different types (including six dTDFs for the auxin-regulated GST, parA) and one dTDF that originates from an ABC transporter gene. Group 1 also contains a dTDF that corresponds to 3-phosphoglycerate (PGA) kinase, which is active centrally in carbon metabolism.

Group 2 (20 dTDFs) consists of genes that were mainly strongly upregulated in asCAD, including several light-related dTDFs [three dTDFs for chlorophyll a/b-binding protein (CAB), one for the phytochrome gene member PHYA, and one for late elongated hypocotyl (LHY)]. Several cell-wall-related dTDFs, corresponding to proteins involved in cell-wall expansion [expansin, lipid transfer protein (LTP), polygalacturonase (PGase), extensin] are found in this group, as well as two dTDFs corresponding to the same leucine-rich repeat receptor-like protein kinase (LRR-RLK) of unknown function.

Group 3 (22 dTDFs) comprises genes that were similarly downregulated in all transgenic lines. Half of the dTDFs to which a function could be assigned are involved in secondary metabolism [phenylalanine ammonia lyase (PAL), 4-coumarate:CoA ligase (4CL), caffeic acid O-methyltransferase (COMT) (two fragments), CAD and dihydroflavonol 4-reductase (DFR)]. The PAL dTDF is identical to a completely conserved region present in the three known tobacco PAL sequences (Fukasawa-Akada et al., 1996; Nagai et al., 1994; Pellegrini et al., 1994). For the Arabidopsis homologues that are most closely related to these tobacco PAL sequences, AtPAL1 and AtPAL2, a role in lignification has been demonstrated (Rohde et al., 2004). The 4CL dTDF aligns with a conserved region in the three available 4CL-coding sequences from tobacco (Lee and Douglas, 1996) but the sequence is not identical (78.6% identity at the amino acid level). Compared to the Arabidopsis 4CL genes, the dTDF shows equal homology to At4CL1 and At4CL2, which have been proposed as the best candidates for a function in monolignol biosynthesis (Ehlting et al., 1999; Raes et al., 2003), and to the Arabidopsis 4CL-like genes 5, 6 and 7, for which no role in monolignol biosynthesis has yet been shown. The two COMT dTDFs overlap and correspond to a tobacco COMT of class I, which is implicated in the biosynthesis of the syringyl units in lignin (Atanassova et al., 1995; Jaeck et al., 1996). The identified CAD dTDF is distinct from the CAD gene used in the antisense construct to suppress CAD expression in asCAD; it is 145 bp long and shares 85% nucleotide identity with the two known tobacco CAD cDNAs. The CAD dTDF clusters in the class of ‘true’CAD genes, characterized for their involvement in lignification (Raes et al., 2003).

Also within group 3 are dTDFs corresponding to dihydroflavonol 4-reductase (DFR, a key enzyme in anthocyanin and proanthocyanidin biosynthesis), chorismate mutase (CM, the rate-limiting enzyme in phenylalanine biosynthesis), ADP-glucose pyrophosphorylase (AGPase, the rate-limiting enzyme in starch biosynthesis), glutamate decarboxylase (GAD), and two stress-related transcription factors (a putative heat shock transcription factor and an AP2 domain-containing transcription factor). Finally, dTDFs corresponding to an unknown auxin-responsive gene and an aquaporin were also grouped with the phenylpropanoid biosynthesis genes.

Group 4 (20 dTDFs) consists of genes that were similarly downregulated in asCCR and in DT, and comprised four dTDFs involved in (cell wall) carbohydrate metabolism [UDP-glucose dehydrogenase (UGDH) (two dTDFs), pectin methylesterase (PME) and β-glucosidase]. Additionally, one dTDF of this group corresponded to a sucrose-responsive element binding MYB factor, TSF (Q9FZ15). Other dTDFs of group 4 corresponded to CAB proteins, an adenosine kinase (ADK) and an aminotransferase.

Group 5 (17 dTDFs) represents genes that are mainly strongly upregulated in asCCR but not in DT in which CAD is also deficient. No prevalence of a functional category was apparent. The only dTDF with a known function in metabolism corresponds to glutamine synthetase (GS).

Among the ungrouped dTDFs, all functional categories were represented and some are noteworthy: a dTDF corresponding to an uncharacterized tobacco PAL gene was induced in all transformants and most strongly in the CCR-deficient lines. When compared to Arabidopsis, this dTDF is most closely related to AtPAL1 and AtPAL2 (Raes et al., 2003), hinting at a role for this induced PAL gene in lignification. A similar accumulation pattern was seen for the dTDF corresponding to 3-deoxy-d-arabino-heptulosonate-7-phosphate (DAHP) synthase. DAHP synthase regulates the carbon flux into the shikimate pathway (Figure 3). A β-amylase dTDF was induced and an AGPase dTDF was repressed in all transformants, suggesting a general mobilization of starch in CCR- and CAD-deficient plants. Furthermore, dTDFs that might be involved in the transduction of signals from the cell wall to the cytoplasm were mainly found in the CCR-deficient lines. A dTDF corresponding to an arabinogalactan protein (AGP) was strongly repressed specifically in DT. One dTDF corresponded to EF-1α/PVN1, a protein that has been suggested to connect the plasma membrane with the cell wall (Zhu et al., 1994). This dTDF was induced most strongly in the single asCCR transformant. Finally, a dTDF corresponding to a lectin domain-containing RLK was induced specifically in DT.

Figure 3.

 Hypothetical relationship between lignification, respiration and detoxification of accumulating phenylpropanoids.
Several lines of evidence from both the transcript and the metabolite profiles from asCCR, asCAD and DT tobacco reflect an elevated respiration, involving breakdown of starch and elevated glycolytic flux. Increased respiration may be associated with an increased release of glucose and an increased biosynthesis of quinate, which is necessary to store/detoxify the accumulating phenylpropanoid intermediates. The release of glucose from starch breakdown may be destined directly for conjugation with accumulating phenylpropanoids. On the other hand, the induced transcript levels of DAHP synthase in asCCR, asCAD and DT indicate that the elevated carbon flux is extended towards the shikimate pathway, and possibly includes the biosynthesis of quinate. For the genes that displayed differential transcript levels in asCCR, asCAD and/or DT tobacco, transcript profiles of WT, asCCR, asCAD and DT (from left to right) are given. Compounds that display differential accumulation in a given transgenic line are followed by asCCR, asCAD or DT. Red, higher; green, lower. Abbreviations: AGPase, ADP-glucose pyrophosphorylase; G-1-P, glucose-1-phosphate; G-6-P, glucose-6-phosphate; F-6-P, fructose-6-phosphate; M-6-P, mannose-6-phosphate; PGA kinase, 3-phosphoglycerate kinase; E4P, erythrose-4-phosphate; PEP, phospho-enol-pyruvate; α-KG, α-ketoglutarate; Gln, glutamine; Glu, glutamate; GS, glutamine synthetase; GOGAT, glutamate synthase; DAHP, 3-deoxy-d-arabino-heptulosonate 7-phosphate; CM, chorismate mutase; PAL, phenylalanine ammonia lyase; asCCR, CCR-downregulated tobacco; asCAD, CAD-downregulated tobacco; DT, tobacco downregulated for both CAD and CCR.

Differential accumulation of phenylpropanoid pathway intermediates and products

Because the altered transcript levels of genes involved in phenylpropanoid metabolism do not necessarily mean that metabolism is altered, it is essential to perform biochemical analyses. Therefore, we first carried out a targeted HPLC-based metabolite profiling of xylem methanol extracts to identify differentially accumulating phenylpropanoid pathway intermediates and products (oligolignols). For this metabolite profiling, a new batch of plants was grown and harvested under the same conditions as for the transcriptome analysis. The set included two lines each for asCAD, asCCR and DT to increase the statistical robustness of the data set. Moreover, all plants were analysed individually (see Experimental procedures). Out of 95 compounds detected, the abundance of 66 compounds was statistically different in at least one of the transgenic lines relative to WT (Table 2; Table S3), of which 18 were below the detection limit in WT and 30 below the detection limit in at least one transgenic line. As was the case for the differential transcriptome, the pattern of differential compounds in DT largely confirmed the pattern in asCCR (Figure 2b), in agreement with the upstream position of CCR relative to CAD in the monolignol biosynthetic pathway. Only three compounds differentially accumulated specifically in DT. The structures of the compounds that were sufficiently pure were resolved by LC–MS analysis (Appendix S1; Figure S2).

Table 2.   Identified differentially accumulating metabolites in asCCR, asCAD and DT, as revealed by HPLC
  1. Values are expressed as μV mg−1 dry weight; values significantly different from WT at < 0.05 are indicated in bold or italics, respectively, when they are higher or lower in abundance. n.d., not detected.

  2. aCompound detected by MS, but co-elution hampered its quantification by UV/Vis; asterisks indicate compounds that are stereoisomers. Caffeoyl quinates were detected, but could not be quantified because of co-elution. Nomenclature is as described by Morreel et al. (2004a). Transgenic lines and number of plants analysed are described in Experimental procedures. Values are means ± SE. Only the differentially accumulating metabolites with known identities are shown; those with unknown identities are presented in Table S3.

 5.64Vanillic acid glucoside132 ± 10602 ± 90485 ± 60985 ± 60
 5.91Syringic acid hexoside12.6 ± 2357 ± 4052.8 ± 4424 ± 50
 6.59Ferulic acid hexosiden.d.2420 ± 300n.d.7670 ± 500
 6.99Sinapic acid glucosiden.d.4250 ± 500165 ± 305260 ± 200
 7.13Scopolin248 ± 203530 ± 3002700 ± 4005700 ± 300
 7.81Sinapoyl hexosen.d.72.0 ± 14n.d.319 ± 30
 7.89Feruloyl hexosen.d.72.2 ± 10n.d.166 ± 20
 9.26Feruloyl quinic acidn.d.343 ± 50n.d.661 ± 110
10.08Sinapyl alcohol87.5 ± 1022.6 ± 5115 ± 1016.5 ± 3
10.17Feruloyl quinic acidn.d.48.6 ± 10n.d.92.7 ± 11
10.29Coniferyl alcohol36.5 ± 3n.d.40.0 ± 8n.d.
10.65G(t8–O–4)G39.6 ± 2n.d.110 ± 10n.d.
10.91G(t8–O–4)G(t8–O–4)G67.3 ± 5n.d.n.d.n.d.
11.26Sinapic acidn.d.32.8 ± 3n.d.30.3 ± 2
11.47Ferulic acidn.d.117 ± 10n.d.79.0 ± 5
13.06Sinapaldehyden.d.n.d.52.9 ± 9n.d.
13.39Coniferaldehyden.d.n.d.326 ± 40n.d.
14.32Feruloyl caffeoyl quinic acidn.d.53.7 ± 7n.d.71.2 ± 12
14.62G(8–5)G544 ± 50n.d.27.4 ± 4n.d.
15.08GMe(t8–O–4)G42.2 ± 5n.d.9.12 ± 1.4n.d.
15.33Feruloyl tyramine36.1 ± 638.3 ± 745.8 ± 683.3 ± 17
15.56G(t8–O–4)S(8–5)G186 ± 20n.d.5.96 ± 1.7n.d.
16.03G(t8–O–4)G(8–8)G8.64 ± 0.7n.d.n.d.n.d.
16.22G(t8–O–4)G(8–5)G′14.9 ± 1n.d.n.d.n.d.
16.38G(e8–O–4)G(8–5)G′66.3 ± 6n.d.4.09 ± 0.7n.d.
16.66G(8–8)G12.5 ± 6n.d.1.03 ± 0.2n.d.
17.10G(8–O–4)G(8–8)S(8–O–4)G40.2 ± 5n.d.11.5 ± 1n.d.
17.28G(8–5)G′93.2 ± 12n.d.29.3 ± 4n.d.
17.48*G(t8–O–4)S(8–8)G49.5 ± 5n.d.8.11 ± 1.2n.d.
17.75G(t8–O–4)S(8–5)G′30.9 ± 4n.d.9.29 ± 1.8n.d.
17.94*G(t8–O–4)S(8–8)G25.4 ± 212.5 ± 31.65 ± 0.418.3 ± 3
18.41G(t8–O–4)S(8–8)S36.7 ± 3n.d.3.25 ± 0.7n.d.
18.64G(e8–O–4)S(8–8)G18.0 ± 3n.d.3.86 ± 0.8n.d.
18.98GMe(t8–O–4)S(8–5)G23.6 ± 3n.d.18.2 ± 2n.d.
19.16G(8–O–4)S(8–8)S(8–O–4)G21.8 ± 3n.d.2.60 ± 0.8n.d.

Generally, the levels of compounds downstream of CAD and CCR decreased: coniferyl and sinapyl alcohol levels were strongly reduced in asCCR and DT. The levels of 16 oligolignols (oligomers of monolignols and their corresponding aldehydes) were strongly reduced in all transgenic lines, and the abundance of three additional oligolignols was strongly reduced specifically in asCCR and DT. In contrast, the substrates of CAD or CCR, and derivatives thereof, strongly accumulated in the transgenic lines: in asCAD, coniferaldehyde and sinapaldehyde increased, and ferulic acid and sinapic acid increased in asCCR and DT, whereas these four compounds were below the detection limit in WT xylem. In addition, in asCCR and DT, levels of derivatives of ferulic acid and sinapic acid, i.e. ferulic acid hexoside, a feruloyl hexose, feruloyl quinic acid, feruloyl caffeoyl quinic acid, sinapoyl hexose, scopoline, vanillic acid glucoside, sinapic acid glucoside and syringic acid hexoside, were elevated. The latter four compounds also accumulated in asCAD. An increased accumulation of feruloyl tyramine was specifically observed in DT.

Non-targeted metabolite profiling

An unbiased analysis of the metabolite profiles of xylem of WT, asCCR, asCAD and DT was conducted on the same plants as for the targeted phenylpropanoid analysis, by GC-MS profiling and subsequent principal component analysis and univariate statistical tests (Figure S3). Of the compounds with known identity, 28 were differentially present between at least one of the transgenic lines and WT (Table 3; Figure 2b). In addition to GC-MS analysis, the levels of all amino acids were determined with an amino acid analyser (Table 4). Again, most differentially present compounds were found in asCCR and DT. The differential metabolite profiles indicated effects on sugar and hemicellulose metabolism, respiration and photorespiration, as well as the involvement of stress responses. Differentially accumulating metabolites involved in these various metabolic aspects and supporting transcript profiling data are presented in more detail below.

Table 3.   Differentially accumulating metabolites in asCCR, asCAD and DT, as identified by GC-MS
PC1PC3PC4P-valueFold changeP-valueFold changeP-valueFold change
  1. ID, mass spectral identifier (see Experimental procedures); PCA, principal component analysis; PC, principal component; P, significance. PC1 separated asCAD from the other tobacco lines, whereas asCCR and DT were distinguished from the other tobacco lines by PC3. PC4 was involved in the distinction between asCCR and DT. The loading factors of those compounds that contributed prominently (see Experimental procedures) to PC1, PC3 and/or PC4 are shown. Italics, lower abundance; bold, higher abundance compared with WT. Transgenic lines and the number of plants analysed are described in Experimental procedures.

Sugar metabolism
Trehalose274 0020.87−0.25   <0.052.2  
Raffinose337 0020.9−0.26   <0.012.8<0.050.7
Melezitose346 0010.88  <0.051.6<0.012.6  
Galactinol299 002 0.53 <0.0011.8    
myo-Inositol209 002 −0.59     <0.0010.6
Isomaltose287 0010.8−0.45     <0.0010.3
Maltose274 0010.91    0.060.6  
Fructose-6-phosphate232 002 0.56 <0.011.9  <0.00012.1
Glucose-6-phosphate235 002 0.64 <0.00012.5  <0.00012.6
Mannose-6-phosphate231 001 0.58 <0.011.8  <0.00011.9
Galactose-6-phosphate232 001 0.56 <0.011.7  <0.0012
Citric acid182 0040.85  <0.0012.1<0.013.1<0.051.3
Hemicellulose metabolism
Xylose166 001 0.470.28<0.00012.4  <0.0011.9
Arabinose167 002 −0.220.34<0.051.3    
Amino acids and polyamines
Putrescine175 0020.84  <0.012.1<0.052.2<0.051.6
γ-Amino butyric acid153 003 0.450.31<0.00012.3  <0.051.7
β-Alanine144 001 0.41 <0.051.5    
Serine138 001 0.400.26<0.011.8    
Threonine140 001 0.540.41<0.011.9    
Glutamic acid163 001 0.550.52<0.00012.7  <0.0011.8
Tyrosine194 002 0.64−0.24<0.0013.7  <0.00016
Methionine152 001  0.3<0.00011.9  <0.011.5
D-Quinic acid185 0010.72−0.56 <0.0010.5<0.051.9<0.00010.4
4-Caffeoyl quinic acid317 001−0.290.39     <0.0012.5
3-Caffeoyl quinic acid311 001       <0.00013.6
1-Caffeoyl quinic acid340 0010.84−0.28   <0.052.1  
Pyroglutamic acid153 002 0.400.53<0.00012.3  <0.0011.6
Threonic acid156 001 −0.340.28<0.050.5  <0.010.4
Glycerol129 003 −0.5 <0.0010.5<0.012.7<0.00010.5
Table 4.   Amino acid concentration in WT, asCCR, asCAD and DT, as determined by HPLC
Amino acidPeak height (μAU μg−1 dry weight; mean ± SE)
  1. *Amino acids of either asCCR, asCAD or DT that accumulate differentially in both transgenic lines compared with WT at < 0.05. Bold, higher abundance compared to WT. n.d., not detected. Transgenic lines and the number of plants analysed are described in Experimental procedures.

Asp*107 ± 10268 ± 30133 ± 10239 ± 20
Glu218 ± 20506 ± 60216 ± 20451 ± 30
Asn14.0 ± 235.3 ± 417.3 ± 131.1 ± 2
Ser*21.8 ± 339.6 ± 425.9 ± 534.0 ± 3
Gln*107 ± 10179 ± 2090.3 ± 6143 ± 11
Gly*8.49 ± 1.112.2 ± 18.07 ± 1.112.2 ± 1
His*4.61 ± 0.46.21 ± 0.85.45 ± 0.68.37 ± 1.1
Arg*4.72 ± 0.36.89 ± 0.85.08 ± 0.36.69 ± 0.6
Thr31.8 ± 457.9 ± 529.0 ± 349.3 ± 3
Ala*23.1 ± 377.3 ± 1049.7 ± 798.4 ± 7
Pro10.7 ± 142.0 ± 89.13 ± 1.116.2 ± 1
Tyr*10.4 ± 122.9 ± 211.1 ± 124.2 ± 1
Val14.6 ± 223.3 ± 116.2 ± 251.6 ± 32
Met4.18 ± 0.44.72 ± 0.622.3 ± 55.99 ± 0.8
Cystine*n.d.0.551 ± 0.5516.1 ± 4n.d.
Ile7.08 ± 1.09.71 ± 0.86.67 ± 1.211.0 ± 1
Leu26.6 ± 241.7 ± 321.8 ± 238.6 ± 1
Phe11.8 ± 111.7 ± 19.76 ± 1.012.1 ± 1
Trp7.05 ± 1.18.00 ± 0.75.64 ± 0.98.96 ± 0.9
Lys10.0 ± 18.67 ± 0.65.75 ± 0.89.15 ± 0.6

Starch and hemicellulose metabolism and respiration.  The differential levels of maltose and isomaltose, which are major products of starch degradation, probably indicate an altered sugar–starch conversion. This observation is in accordance with the general mobilization of starch in CCR- and CAD-deficient plants, as suggested by the lower transcript levels of AGPase and the higher transcript levels of β-amylase; Table 1).

Glucose is used as a substrate for respiration, entering glycolysis via conversion into glucose-6-phosphate (G-6-P) and fructose-6-phosphate (F-6-P). Increased concentrations of the interconvertible G-6-P, F-6-P and mannose-6-phosphate (M-6-P) were detected in asCCR and DT, suggesting elevated respiration (Roessner et al., 2001; Figure 3).

The major hemicellulose present in cell walls of tobacco is arabinoxyloglucan, whose breakdown products are arabinose, xylose and glucose. Increased levels of arabinose in asCCR and xylose in asCCR and DT result from an increased breakdown of hemicellulose compared to WT. The hemicellulose breakdown products, arabinose and xylose can, via conversion to xylulose-5-P, enter the pentose phosphate pathway and function as substrates for respiration (Michal, 1999), or they can be salvaged by converting them back to their UDP conjugates that may re-enter hemicellulose biosynthesis, supporting hemicellulose remodeling (Kotake et al., 2004). Together, these data suggest that the transgenic lines, and mainly asCCR and DT, break down starch and hemicellulose and that the breakdown products enter respiration.

Photorespiration.  Citrate levels were increased in all transformants, but no differential accumulation was observed for any other Krebs cycle intermediate. Such specific accumulation of citrate has been associated with elevated photorespiration (Bykova et al., 2005). The levels of the amino acids Gly, Ser, Gln, Glu, Asp and Ala were increased in asCCR and DT (Tables 3 and 4). Elevated levels of the photorespiratory metabolites Gly and Ser, and increased levels of Gln and Glu, have previously been used as indicators for elevated photorespiration (Jeong et al., 2004; Kozaki and Takeba, 1996; Oliveira et al., 2002). Gln and Glu are intermediates in the re-assimilation of photorespiratory ammonium via the glutamine synthetase/glutamate synthase (GS/GOGAT) cycle. Asp and Ala are derived from Glu. Together with the dTDFs corresponding to the photorespiratory enzymes glycolate oxidase and GS, these differential metabolites indicate altered photorespiration in asCCR and DT.

Indicators of a stress response.  Other metabolic changes, such as the altered levels of trehalose, raffinose and galactinol, the phenolic compound caffeoyl quinic acid, and the metabolites involved in polyamine biosynthesis, i.e. Arg, putrescine, β-alanine and γ-amino butyric acid (GABA), point towards a potential stress response in the transgenic plants, mostly in asCCR and DT [trehalose (Wingler, 2002); raffinose and galactinol (Panikulangara et al., 2004); polyamines (Kaplan et al., 2004; Nikiforova et al., 2005)].

Taken together with the transcriptome data that indicate altered transcript levels for metallothioneins, GSTs and heat shock transcription factors, which are typical defence genes induced in response to oxidative damage, the differential accumulation of stress-related metabolites revealed by GC-MS indicates that CCR deficiency may result in oxidative stress in tobacco.

Photo-oxidative stress.  The altered threonic acid and pyroglutamic acid concentrations in asCCR and DT may indicate altered redox status in the chloroplast, reflecting photo-oxidative stress. Threonic acid and pyroglutamic acid are derived from l-ascorbate and glutathione (GSH), respectively, which constitute, via the ‘ascorbate–glutathione cycle’, two major anti-oxidants of the chloroplast’s photo-oxidative stress defence system (Bray et al., 2000). Photo-oxidative stress was also suggested by the transcriptome data. One dTDF was most homologous to a thylakoid-located FtsH protease that is probably involved in the degradation of photo-oxidatively damaged photosystem II (PSII) proteins (Lindahl et al., 1996), and three dTDFs corresponded with chloroplastic ribosomal proteins. Furthermore, one dTDF corresponded to an epoxide hydrolase that is possibly functional in the xanthophyll cycle involved in photoprotective energy dissipation (Loggini et al., 1999). Taken together, both transcriptome and metabolome data indicate photo-oxidative stress in the lignin-modified lines, again predominantly in asCCR and DT.

Validation of hypotheses generated by molecular phenotyping

Pigments.  Various methods were used to validate the hypotheses generated on the basis of the molecular phenotyping. Transcript profiling data suggested there were effects on pigment biosynthesis in the transformants. A general repression of chloroplast-localized steroid biosynthesis was prompted by the reduced transcript levels, particularly in DT, of 1-deoxy-d-xylulose-5-phosphate synthase, a key enzyme in the biosynthesis of chloroplast-bound isoprenoids, such as carotenoids (carotenes and xanthophylls), prenyl chains of chlorophylls and plastoquinone-9 (important in plant respiration), and ABA. On the other hand, reduced DFR transcript levels (in asCCR, asCAD and DT) suggested reduced levels of anthocyanins.

To check for altered pigment accumulation, chlorophyll a (ChlA), chlorophyll b (ChlB), carotenoids and anthocyanins were quantified spectrophotometrically in extracts from xylem, cortex and leaves (Table 5). ChlA levels (and to a lesser extent ChlB levels) were markedly increased in leaf, xylem and cortex of all transgenic lines. Anthocyanin levels were significantly reduced in leaves of all transgenic lines, whereas, in xylem, a significant decrease in anthocyanin and carotenoid content was only obvious in asCAD and DT.

Table 5.   Pigment concentrations
  1. Chlorophyll a and b and carotenoid data are mean concentrations (±SE) in μg mg−1 dry weight, anthocyanin data are absorbance values (±SE). Values significantly different from WT at < 0.05 have a different font: bold, increased concentration; italics, decreased concentration. Transgenic lines and the number of plants analysed are described in Experimental procedures.

ChlA74.7 ± 4146 ± 10135 ± 10171 ± 10
ChlB25.1 ± 341.5 ± 353.0 ± 655.6 ± 6
Carotenoid36.0 ± 377.4 ± 377.3 ± 384.6 ± 5
Anthocyanin1.73 ± 0.31.32 ± 0.30.872 ± 0.160.727 ± 0.21
ChlA527 ± 201540 ± 100995 ± 601450 ± 100
ChlB177 ± 10449 ± 30290 ± 20498 ± 80
Carotenoid175 ± 10445 ± 30317 ± 20399 ± 40
Anthocyanin0.197 ± 0.29−3.10 ± 0.40.442 ± 0.68−3.42 ± 0.6
ChlA563 ± 40952 ± 601730 ± 7001330 ± 100
ChlB170 ± 10236 ± 20423 ± 180348 ± 30
Carotenoid164 ± 10249 ± 20463 ± 200352 ± 30
Anthocyanin−0.896 ± 0.213.95 ± 0.33.56 ± 0.44.04 ± 0.2

Chlorophyll fluorescence and efficiency of photochemistry.  The differential transcriptome and metabolome, together with the elevated chlorophyll content, suggested over-excitation of the photosynthetic apparatus, photo-oxidative damage, and alterations at the levels of photosynthesis and photorespiration in the transformants. To test whether the profiling data truly reflected changes in these physiological processes in planta, chlorophyll fluorescence and gas exchange were measured.

Fluorescence from leaves of WT, asCCR, asCAD and DT in the dark-acclimatized state upon application of a saturating flash gives an indication of the maximal amount of absorbed light that can be used to drive the electron transport chain, allowing determination of the maximum photochemical quantum yield of PSII (Fv/Fm) in the absence of non-photochemical quenching. The Fv/Fm ratio in the yellowish tissue between the leaf veins of asCCR and DT was significantly decreased when compared to WT (Table 6). In contrast, when measurements were performed on the dark-green tissue surrounding the lesions, a significant increase in Fv/Fm was observed in asCCR and DT. Chlorophyll fluorescence parameters measured in the light-acclimatized state give an indication of the actual proportion of photochemical and non-photochemical quenching. The actual proportion of absorbed light that is used to drive the electron transport chain is represented by the photochemical quantum yield ΦPSII, and was significantly increased in asCCR and DT. In the other lines that were tested, no significant changes in fluorescence parameters were apparent.

Table 6.   Fluorescence
  1. Chlorophyll fluorescence variables of the youngest mature leaf of tobacco plants. Measurements were obtained in the dark-acclimatized (Fv/Fm) and light-acclimatized states (ΦPSII). Means  ± SE are indicated. Fv/Fm measurements for asCCR and DT were made on leaf sections avoiding and focusing on the lesions. Asterisks indicate values that are significantly different from the WT (< 0.10). Fv/Fm, maximum photochemical quantum yield of PSII; ΦPSII, photochemical quantum yield of PSII. Transgenic lines and the number of plants analysed are described in Experimental procedures.

Fv/Fm0.779 ±  0.010.796 ±  0.01*0.790 ±  0.01  0.0467
Fv/Fm (lesions) 0.758 ±  0.01*0.747 ±  0.01*   
ΦPSII0.132 ±  0.010.183 ±  0.02*0.156 ±  0.01*0.3130.0170.0532

Gas exchange.  The increase in the photochemical quantum yield of PSII may result in altered photosynthesis and/or photorespiration. Because photorespiration measurements are time-consuming, and, for optimal comparability, should be performed in the shortest possible time interval, we first assayed a limited set of asCAD, asCCR and DT lines. Differential photorespiration was most obvious in the homozygous CCR-deficient transformants asCCR (B3.1) and DT (Dt52.6) (data not shown). These lines were further examined by gas exchange tests (Table 7). Photorespiration was significantly increased in both asCCR (B3.1) and DT (Dt52.6), whereas no significant effects on photosynthesis were observed.

Table 7.   Gas exchange analyses
  1. CO2 assimilation rate at saturating light intensity (Asat) in air containing 21% O2 and rates of photorespiration of the youngest mature leaf of WT, asCCR and DT. Means ± SE are given. Asterisks indicate values that are significantly different (< 0.10) from WT. Transgenic lines and the number of plants analysed are described in Experimental procedures.

Tobacco lineAsat (μmol m−2 sec−1)Photorespiration (μmol m−2 sec−1)
WT14.01 ± 0.66.57 ± 0.2
B3.114.62 ± 0.67.87 ± 0.3*
Dt52.612.85 ± 0.77.61 ± 0.2*
Pmodel 0.14220.0020


Characterization of the molecular phenotype by transcriptome and metabolome profiling

In an attempt to map the metabolic sphere of influence of the perturbation of an essentially structural process, lignification, alterations in transcript and metabolite levels were studied in the xylem of tobacco plants in which the expression of CCR, CAD or both were downregulated. We hypothesized that interpretation of the snapshots of the differential transcriptomes and metabolomes of the transformants would help reveal the molecular and metabolic processes that are associated with the morphological phenotypes that have been described for these transformants by Chabannes et al. (2001a). Supporting this presumption, the molecular phenotypes were proportional to the morphological phenotypes (Chabannes et al., 2001a; Halpin et al., 1994; Piquemal et al., 1998); a more restricted portion of the transcriptome was affected in asCAD than in asCCR, and this tendency was also evident at the metabolite level. Furthermore, when both CCR and CAD were deficient (in the DT), the effects of downregulation of the most upstream enzyme, CCR, prevailed both at the transcriptome and metabolome levels. Phenylpropanoid biosynthesis was affected in asCAD as well as in asCCR and DT. However, a clear distinction was apparent between CCR and CAD downregulated plants with regard to the functions of the other affected metabolic pathways. The molecular phenotype of asCCR and DT was characterized mainly by changes at the level of detoxification and carbohydrate metabolism, whereas the molecular phenotype of asCAD was enriched by differential transcript levels of light- and cell-wall-related genes. Interpretation of the differences at the level of the transcriptome and metabolome, although obtained from distinct plant batches because of practical restrictions, led to a coherent image of the molecular adaptations in these transgenic plants.

Candidate genes for a function in the control of lignin biosynthesis-associated processes

Transcript profiling revealed differential expression in the three transgenic lines (asCCR, asCAD and DT) of genes involved in monolignol biosynthesis (CAD and COMT), in the general phenylpropanoid pathway (PAL, 4CL) and in the substrate-supplying shikimate-aromatic amino acid biosynthesis pathway (CM). With the exception of one of the two differential PAL genes, all these monolignol biosynthesis-linked genes were similarly downregulated in asCCR, asCAD and DT. These data prove that our approach is valid to identify genes that are metabolically closely linked to phenylpropanoid metabolism. Hence, the set of known and unknown genes whose transcript levels were similarly downregulated in asCCR, asCAD and DT (Figure S1) are good additional candidates to be closely associated with the pathway.

Comparison of the transcript profiling data with those obtained for other mutants in the monolignol biosynthetic pathway also pointed to a set of interesting candidate genes that are likely to play a role in flux control through primary and secondary metabolism in response to changes in monolignol biosynthesis. For example, for many genes that displayed altered transcript levels in asCCR, asCAD or DT, i.e. those encoding PAL, 4CL, COMT, CM, PGA Kinase, β-glucosidase, glycolate oxidase, AGPase, DAHP, UGDH, GAD, β-amylase and GS, the closest Arabidopsis homologues were also differentially expressed in Arabidopsis cad-c cad-d, pal1, pal2 and/or c3h mutants (Abdulrazzak et al., 2006; Rohde et al., 2004; Sibout et al., 2005). Therefore, the corresponding enzymes are good candidates for a direct or indirect flux-controlling role in lignin-associated processes.

Two transcription factor genes displayed a similarly strong downregulated transcript level in asCAD and asCCR lines, as was observed for the phenylpropanoid biosynthesis genes. These genes, encoding a putative heat shock transcription factor and an AP2 domain-containing transcription factor, might have a putative function in the coordinated transcriptional regulation of phenylpropanoid metabolism. This hypothesis is further supported by the fact that the transcript level of the closest Arabidopsis homologue of the AP2 transcription factor (At3g16770) was reduced in the Arabidopsis cad-c cad-d mutant (Sibout et al., 2005). In contrast, the Myb-like transcription factor DIVARICATA was induced in the asCCR and DT lines, and may be induced in response to low lignin contents. Its closest Arabidopsis homologue (At5g5890) is similarly induced in an Arabidopsis c3h insertion mutant (Abdulrazzak et al., 2006).

Based on their differential transcript levels in various transgenic lines/mutants with defects in monolignol biosynthesis, several signal transduction genes are good candidates for a regulatory function associated with lignification: an RLK showing differential transcript levels in asCAD tobacco and whose closest Arabidopsis homologue (At1g73080) is differentially expressed in the cad-c cad-d and pal1 Arabidopsis mutants (Rohde et al., 2004; Sibout et al., 2005), a gene encoding a PHYA signalling protein that has differential transcript levels in asCCR, asCAD and DT and whose closest Arabidopsis homologue (At3g22170) is differentially expressed in the pal2 Arabidopsis mutant (Rohde et al., 2004), and a phosphatidylinositol-4-phosphate 5-kinase with differential transcript levels in asCAD tobacco and whose closest Arabidopsis homologue (At1g21920) is differentially expressed in the pal1 and pal1 pal2 mutants (Rohde et al., 2004).

Because auxin is a key signal for secondary xylem formation (Sundberg et al., 2000), an unknown auxin-responsive gene, which was identified here, with reduced transcript levels in asCCR, asCAD and DT, and six tags corresponding to the auxin-related GST parA, upregulated in asCCR and DT, might play a role in integration of monolignol biosynthesis into the developmental process of secondary xylem formation. Additionally, an aquaporin with a highly similar transcript profile as this auxin-responsive protein across the transgenic tobacco lines might be involved in the coordinated cyclic regulation of cell expansion and lignification during secondary xylem formation (Harmer et al., 2000). Aquaporins mediate the water influx associated with cell expansion. Because expression of the closest Arabidopsis homologue of both the auxin-responsive unknown (At3g25290) and aquaporin (At4g17340) genes were reduced as well in the cad-c cad-d mutant (Sibout et al., 2005) we believe that these genes are worthy targets for further functional studies.

Lignin biosynthesis and starch metabolism are closely linked

In xylem cells, a major portion of the carbon that is channelled into phenylpropanoid metabolism is diverted to lignin biosynthesis (van Heerden et al., 1996). The flux into secondary metabolism can be limited by the precursor supply in primary metabolism (Henkes et al., 2001; Rogers et al., 2005). In plants, starch is the major carbon storage form and sucrose the major carbon transport form, and the latter may be considered as the main substrate for lignin biosynthesis (Amthor, 2003). A dependency of the transcript levels of lignin biosynthetic genes on starch turnover and availability of sucrose has been demonstrated in the Arabidopsis sex1 mutant. This mutant is impaired in its ability to degrade starch granules, has low transcript levels of most monolignol biosynthetic genes, and accumulates less lignin (Rogers et al., 2005). In the transgenic tobacco lines analysed, the opposite effect was apparent; in concert with a general downregulation of phenylpropanoid biosynthesis genes in asCCR, asCAD and DT, the transcript levels of a rate-determining starch biosynthetic enzyme, AGPase, were reduced, and those of a starch-degrading exo-enzyme, β-amylase, were induced. These observations indicate a general induction of starch mobilization that is linked with monolignol biosynthesis. Alterations in starch and sucrose metabolism were further supported by the altered levels of maltose and isomaltose.

Metabolic link between the detoxification of accumulating phenylpropanoid intermediates and carbon supply

In asCAD, the substrates of the suppressed enzyme, coniferaldehyde and sinapaldehyde, accumulated strongly, whereas in asCCR and DT, the abundances of glycosylated and quinylated derivatives of the substrate of CCR, feruloyl CoA, and, additionally, ferulic acid and sinapic acid were strongly elevated. Similar to what was proposed for CCoAOMT-deficient plants, feruloyl CoA, the substrate of CCR, might be quinylated or re-directed to ferulate by a thioesterase. Ferulic acid would then be partially converted to sinapic acid, and both acids detoxified by glycosylation (Guo et al., 2001; Meyermans et al., 2000).

The accumulation of glycosylated and quinylated phenylpropanoid derivatives, which is particularly strong in asCCR and DT, indicates that the accumulation of monolignol biosynthesis intermediates in CCR- and/or CAD-deficient plants provokes a concomitant induction of the detoxification and storage mechanisms, requiring, in turn, an increased carbon supply in the form of glucose and quinate. Where do the glucose and quinate come from? In asCAD, asCCR and DT, the lower transcript levels for AGPase and the higher transcript levels for β-amylase than in WT indicated a reduced biosynthesis and increased degradation of starch to glucose. In asCCR and DT, several lines of evidence from both the transcript and the metabolite profiles additionally reflected an elevated respiration, involving the breakdown of carbohydrate polymers or oligomers to glucose, followed by the utilization of glucose by the glycolytic pathway. The release of glucose from starch or hemicellulose catabolism may be destined directly for conjugation with accumulating phenylpropanoids. On the other hand, the induced transcript levels of DAHP synthase in asCCR, asCAD and DT indicate that the elevated carbon flux is at least partly extended towards the shikimate pathway, and, because CM transcript levels are reduced in these lines, an increased flux towards the biosynthesis of quinate is expected. The level of free quinate then depends on the balance between quinate biosynthesis and quinate acylation. Free quinate levels are increased in the asCAD plants but are decreased in the asCCR and DT plants. A stronger depletion of quinate in the asCCR plants is in accordance with the stronger accumulation of phenylpropanoid quinic esters in these plants, and does not contradict a concomitant increase of quinate biosynthesis (Figure 3).

A link between cell-wall expansion and light-related genes is apparent in asCAD

Specifically, in asCAD tobacco, the transcript levels of several genes related to cell-wall expansion (expansin, LTP, PGase, PME), and of genes involved in the light-harvesting complex/circadian clock (CAB,LHY, and PHYA), and light signalling (PhyA), were elevated. Similarly, microarray analyses of the xylem of transgenic CAD-downregulated poplar (Populus tremula x alba) revealed differential transcript levels of predominantly light/circadian clock-related genes, including CAB and LHY (Andersson Gunnerås, 2005), and an important fraction of differentially expressed genes in Arabidopsis cad mutants encoded cell-wall (expansion) proteins (Sibout et al., 2005). These data suggest unanticipated alterations of cell-wall structure in asCAD that have so far remained unnoticed. Further detailed experiments are required to reveal these effects at the structural level.

Oxidative stress

Differential transcript levels of stress-related genes and the accumulation of low-molecular-weight anti-oxidants and other stress-related compounds, particularly in the CCR-deficient plants, indicate that the transgenic plants experience oxidative stress. What could trigger oxidative stress in plants with perturbed monolignol biosynthesis? One reason could be that the anti-oxidant defence system is partly deficient. In some plant species, hydroxycinnamic acids play an important role as anti-oxidants (Neill et al., 2002). In tobacco plants with inhibited phenylpropanoid metabolism caused by overexpression of the transcription factor AmMYB308, or with inhibited PAL expression, premature cell death in the leaves has been reported, a phenotype similar to that in the leaves of CCR-deficient tobacco (Elkind et al., 1990; Tamagnone et al., 1998a), and the decreased levels of phenolic acid derivatives, and particularly caffeoyl quinic acid, have been suggested to be the cause of the oxidative stress (Tamagnone et al., 1998b). However, in CCR- and CAD-deficient tobacco, the levels of hydroxycinnamic acids and their derivatives, including caffeoyl quinic acid, were increased, making this hypothesis not very likely.

On the other hand, anthocyanins are excellent scavengers of free radicals. High light induces anthocyanin accumulation that might protect plants against photo-oxidative stress and cell death (Vandenabeele et al., 2004; Vanderauwera et al., 2005). The decreased anthocyanin levels in the xylem of asCAD and DT and in the leaves of asCCR, asCAD and DT might thus contribute to oxidative stress. Anthocyanin levels were indeed most severely decreased in the lines that developed lesions in the leaves. However, the possibility cannot be excluded that the observed decrease in anthocyanin levels are a secondary effect of oxidative stress rather than its cause. Indeed, transcriptome analysis of catalase-deficient Arabidopsis indicated that accumulating H2O2 impairs the high-light induction of genes involved in the regulation, biosynthesis and sequestration of anthocyanins (Vandenabeele et al., 2004; Vanderauwera et al., 2005).

The altered cell wall as a stress generator

Alternatively, the molecular stress phenotype may not be primarily due to a defective anti-oxidant defense system, but may be actively induced. One possibility is that the altered cell wall mimics wounding or the early stages of pathogenesis when hydrolysing enzymes secreted by pathogens depolymerize the plant cell wall. An induction of the physiological response characteristic of wounded and infected plants has been described previously for the cellulose synthase mutant cev1, whose altered cell wall has been proposed to allow increased release of elicitors (Ellis et al., 2002). In agreement with a wound response, feruloyl tyramine levels were higher in CCR-downregulated lines, both as a soluble compound (this study) and incorporated into the wall (Chabannes et al., 2001a; Ralph et al., 1998). Feruloyl tyramine is an amide formed by the conjugation of tyramine with feruloyl CoA, and is typically elicited in solanaceous plants upon wounding or pathogen attack to reinforce the cell wall by cross-linking with lignin (Guillet and De Luca, 2005).

A second possibility is that the altered structure of the cell wall itself may signal a stress response. Lignin content is strongly reduced in asCCR and DT, and structural disorders are clearly visible in the xylem cell walls (Chabannes et al., 2001b). Furthermore, alterations in the metabolism of hemicellulose and pectin were apparent from the differential transcriptome and metabolome of the CCR-deficient transformants. Similar alterations in cell-wall biogenesis were observed in cellulose synthase mutants and in the lignin-deficient Arabidopsis c3h mutant (Abdulrazzak et al., 2006; Caño-Delgado et al., 2003). Because hemicellulose cross-links the cellulose microfibrils, reduced hemicellulose content might, in turn, contribute to the loss of cohesion of the cell walls observed in asCCR and DT. Ultimately, alterations in cell-wall structure need to be signalled to the nucleus to regulate gene expression. Integrin-like linkages between the cell wall and the plasma membrane are required to mediate some of the signals that can activate defence responses (Mellersh and Heath, 2001). The differentially expressed genes encoding potential linking proteins in the CCR-deficient transformants (AGP, RLKs and EF-1α/PVN1) are thus candidate signal transducers.

Photorespiration dissipates increased photosynthetic efficiency

Another possibility is that the oxidative stress is caused by photorespiratory H2O2. An increase in photorespiration was suggested by the differential transcript and metabolite profiles, and was confirmed in asCCR and DT by gas-exchange experiments. Photorespiration is a photoprotective mechanism that dissipates excess electron flow through the photosynthetic apparatus when excess light is absorbed, and is an important source of H2O2. In accordance with an induction of photorespiration in response to excess light absorption in the CCR-deficient plants, an increased turnover of PSII proteins was apparent from the transcript profiling, and strong accumulation of H2O2 could be visualized in the leaves of the CCR-deficient lines (Figure S4). Moreover, the strongly CCR-downregulated lines had lesions in the leaves that were highly similar to those in catalase-deficient tobacco when grown under high light conditions (Dat et al., 2003), and may thus result from the photo-oxidative stress and subsequent photo-inhibition.

The elevated photorespiration itself is probably caused by the increased efficiency of PSII. Indeed, the electron flow in the photosynthetic electron transport chain depends on the efficiency of the absorption of light energy and the efficiency of the photosynthetic assimilation of CO2. An increased efficiency of PSII was suggested by increased chlorophyll levels in leaves, cortex and xylem of CCR- and CAD-downregulated transformants. Chlorophyll fluorescence parameters confirmed an increased maximum photochemical quantum yield and an increased PSII efficiency in the leaves of the strongly CCR-downregulated transformants. However, gas-exchange analyses showed that photosynthetic CO2 assimilation was not significantly altered in these lines. The absorption of more light energy than can be utilized for photosynthesis may increase the level of photorespiration to protect the photosynthetic apparatus from light-induced damage (Niyogi, 2000). The question remains as to which mechanism induces the efficiency of the photosynthetic apparatus in these plants.


Taken together, the transcriptome and metabolome data suggest the following hypothetical scenario (Figure 3). Downregulation of CAD and CCR reduces the flux of carbon towards lignin. This reduced flux is most pronounced for CCR-downregulated plants, which make significantly less lignin. The primary response of the partial block in these pathway steps is the accumulation of pathway intermediates and derivatives thereof, the disappearance of monolignol coupling products, and the alteration of cell-wall structure. The pathway intermediates repress gene expression in the phenylpropanoid pathway, possibly by feedback mechanisms. As a secondary response, the cell wall senses its altered structure and the need for further reinforcement of the wall with lignin. Therefore, starch is broken down to supply new monolignols for lignin, but because CAD and CCR are defective, the result is a further accumulation of pathway intermediates and a stronger repression of phenylpropanoid biosynthesis. The breakdown of starch and hemicellulose serves as the initial source of the glucose and quinate needed to detoxify and store the accumulating phenylpropanoid pathway intermediates (Figure 3).

A third event is the stress response that is most apparent in CCR-deficient lines. Possibly, plasma membrane-localized proteins sense the altered cell-wall structure and signal it to the nucleus, or the altered cell wall releases cell-wall fragments that act as elicitors to switch on gene expression. The increased photorespiration in CCR-deficient lines may as well be an important contributor to the oxidative stress.

Our data show that altering the expression of the late monolignol biosynthetic pathway genes CCR and CAD affects the transcript level of a high number of genes, revealing a substantial network of regulatory interactions both within monolignol biosynthesis, and among monolignol and other metabolic pathways. In the past, such effects were classified as pleiotropic and considered uninteresting because they could not be properly explained and linked with the gene under study. However, with the advent of transcript and metabolite profiling tools, the molecular details of these pleiotropic phenotypes can be revealed at a much larger scale than before, and scenarios for how they relate to the modified gene can be proposed. Understanding these pleiotropic effects is important for several reasons. It provides new candidate genes that are closely related to the modified pathway. In addition, the genes of a given pathway for which transcript levels are affected might be those that regulate the flux through that pathway, and therefore correspond to interesting targets for pathway engineering.

Furthermore, knowing how the biosynthesis of the various cell-wall polymers is integrated, and how plants compensate for or respond to alterations in the biosynthesis of one of the wall polymers, is essential for designing cell walls for various end uses. Importantly, dissecting the complex pleiotropic phenotype at the molecular level might help in predicting the negative and unforeseen consequences of a mutation on plant health and quality. This knowledge opens up possibilities for mitigating these effects, for example by blocking adverse metabolic routes through genetic engineering or breeding. Our data also put the concept of substantial equivalence into perspective. Apparently, alterations in the expression of a single endogenous structural gene affect the expression of a plethora of other genes in various unanticipated pathways, putting into perspective concerns about unanticipated effects on metabolism caused by transgenes in genetically modified organisms.

Finally, our data reveal the amazing plasticity of the plant’s metabolism in response to genetic variation. This metabolic plasticity, enabling plants to adapt to genetic variation, is probably important for plant populations to adapt to environmental conditions and must be considered essential for the long-term survival of populations (Booy et al., 2000).

Experimental procedures

Tobacco lines

All tobacco plants used were Nicotiana tabacum L. cv. Samsun; wild-type plants are referred as WT. B3.1 and B6.8 are cinnamoyl CoA reductase (CCR) downregulated lines, homozygous for a construct containing a 1.3-kb fragment, corresponding to a full-length tobacco CCR cDNA, in the antisense orientation, coupled to the 35S CaMV promoter and the 3′ terminator of the nopaline synthase gene (Piquemal et al., 1998). Seeds resulting from self-pollination of the homozygous line were used.

J48.7 and J40.31 are cinnamyl alcohol dehydrogenase (CAD) downregulated lines, homozygous for a construct containing a 1 kb fragment of the tobacco CAD2 cDNA in the antisense orientation, coupled to the 35S CaMV promoter and the 3′ terminator of the nopaline synthase gene (Halpin et al., 1994). Seeds resulting from self-pollination of the homozygous lines were used.

T16-1 T2F is a CAD-downregulated line, homozygous for an RNAi construct containing a 550 bp fragment of the 5′ region of tobacco CAD2 in the sense and the antisense orientations flanking a 220 bp intron, coupled to the CaMV 35S promoter and 3′ terminator. The hemizygous lines resulting from the backcross of B3.1 and J48.7 with WT are B3 h and J48 h, respectively (Chabannes et al., 2001a).

The double transformant Dt1a, obtained by crossing of the homozygous lines J48.7 x B3.1, is hemizygous for the two transgenes (Chabannes et al., 2001a). Dt52.6 is a CCR- and CAD-downregulated line, homozygous for a chimeric construct with partial sense sequences of both CAD and CCR (Abbott et al., 2002). All transgenic lines downregulated for CCR had slightly curled leaves that were a darker green than those of the WT. In addition, B3.1 and Dt52.6 developed yellowish areas between the leaf veins and were smaller than the other lines.

For convenience, throughout the text, ‘asCCR’ refers collectively to all CCR single-transformed lines described above, ‘asCAD’ to all CAD single-transformed lines described above, and ‘DT’ to the double transformants Dt1a and Dt52.6.

Growth and sampling conditions

Seeds of the various tobacco lines were surface-sterilized and germinated on solid Murashige and Skoog medium, grown under a light/dark regime of 16 h/8 h, 20–30 E m−2 sec−1, at 27°C, and transferred after 4 weeks to soil and cultivated in a greenhouse (at 16 h/8 h and 25°C/22°C day/night).

For analyses of stem tissue (cDNA-AFLP, soluble phenolics, GC-MS, amino acid and pigment analyses), 10-week-old plants were harvested at 10 weeks old, just before flowering, and whole stems of WT and transgenic lines were frozen immediately in liquid nitrogen. All experiments were carried out on the basal half of the stem of the various tobacco lines. The cortex was removed, and the outer part of the woody ring was scraped off with a scalpel and ground for further analysis. For pigment analyses, the cortex was ground separately, and each sample of leaf tissue consisted of a pool of three 12 mm leaf discs, cut out of the 4th, 5th and 6th leaves, counting from the top. Samples were frozen in liquid nitrogen immediately upon harvest and were ground with metal balls in an MM300 high-frequency shaking instrument (Retsch); The following lines were used: in the cDNA-AFLP experiment, WT (n = 7) and the hemizygous downregulated lines B3 h (n = 7), J48 h (n = 7) and Dt1a (n = 7) for amino acid, soluble phenolics and GC-MS analyses, the lines B3.1 (n = 6), B6.8 (n = 6), J48.7 (n = 3), J40.31 (n = 3), Dt52.6 (n = 6), Dt1a (= 6), T16-1 T2 F (= 6) and WT (= 12) for pigment analyses, the lines B3.1 (= 5), B6.8 (= 5), J48.7 (= 3), J40.31 (= 3), Dt52.6 (= 5), Dt1a (= 5) and WT (= 5) for gas exchange analysis, WT (= 16), and the lines B3.1 (= 12) and Dt52.6 (= 12) for chlorophyll fluorescence measurements in the light-acclimatized state, WT (= 6), B3.1 (= 6) and Dt52.6 (= 6) and for chlorophyll fluorescence measurements in the dark-acclimatized state, WT (= 29), B3.1 (= 13), B6.8 (= 13), J48.7 (= 13), Dt52.6 (= 13) and Dt1a (= 13). Independently grown series of plants were utilized for (i) cDNA-AFLP, (ii) the amino acid, phenolic and GC-MS profiling, (iii) the pigment analysis and (iv) gas exchange and chlorophyll fluorescence measurements. Multiple lines were used to increase the robustness of the data. Statistical analysis revealed only those metabolites that were differential in all lines.

cDNA-AFLP analysis and data processing

For each genotype, two pools of xylem tissue, each from seven plants, were prepared for RNA extraction. Total RNA was prepared by the LiCl extraction method (Goormachtig et al., 1995), from approximately 300 mg of pooled ground xylem. Using 5 μg of total RNA, cDNA-AFLP-based transcript profiling was performed essentially as described previously (Breyne et al., 2002). The cDNAs were screened using all possible 128 MseI + 1/BstYIT/C + 2, 55 MseI + 2/BstYIT/C + 2 and six MseI+(2 or 3)/BstYIT/C+(2 or 3) primer combinations. Quantitative data were analysed using AFLP-QuantarPro software (Keygene; The exported normalized band intensity values were corrected for the lane intensity according to the method described by Breyne and Zabeau (2001). Bands with similar intensity in the two pools within each line, but different intensity in one or more transgenic lines, compared to WT, were selected first by visual scoring. From this selection, significant (< 0.05) differential expression between the genotypes was determined for each transcript using one-way analysis of variance (anova) with one fixed factor (genotypes) and the least significant difference (LSD) post hoc test (version 8; SAS Institute; TDFs that, according to visual scoring and both statistical analyses, accumulated differentially between the WT and one or several other genotypes, were isolated from the gel, re-amplified and sequenced. Only when a sequence could be obtained directly from the re-amplified PCR product, or when, after cloning of this PCR product into the pGEM-T easy vector (Promega,, individual clones gave one unambiguous sequence, was the corresponding intensity pattern on the gel assumed to be the result of a single transcript and was withheld as ‘differential TDF’ or dTDF. Sequences shorter than 21 nucleotides after removal of the primer sequence were not further considered. After log transformation and normalization of the mean expression data (with two replicates per genotype), the complete linkage hierarchical clustering algorithm and the treeview algorithm (Eisen et al., 1998) were used to group the dTDFs with similar behaviour in terms of transcript levels in the various transformants.

Homology search and identification of the transcript tags

The sequences obtained either from the re-amplified PCR product or from individual clones after cloning into pGEM-T easy (Promega) were compared by blastx (Altschul et al., 1997) to a database containing all plant amino acid sequences from the Swiss-Prot and TrEMBL databases. If at least 10 amino acids were aligned in the best blastx hit, the alignment covered at least 80% of the overlap between the tag (query) and the sequence in the database (subject), and at least 50% of the amino acids in the alignment were identical between the query and the subject, the best blastx hit was accepted as homologue of the dTDF and its function assigned to the dTDF. If one or more of these parameters were not reached in the blastx output, blastn was performed on the transcript tag (Altschul et al., 1997) against a database containing all EMBL and GenBank non-redundant plant nucleotide sequences, including expressed sequence tag and genome survey sequences. A blastn hit was accepted if an alignment of 30 bp or longer was obtained that covered at least 60% of the overlap, and if at least 80% of the base pairs within the alignment were identical between query and subject. If the alignment did not cover 60% of the overlap, a blastn hit was still accepted when at least 50 bp were aligned with at least 85% identity, or when at least 21 bp had a completely identical alignment, covering the complete overlap between the dTDF and the blastn hit. In this manner, TDFs that tend to be short and/or situated in the 3′ untranslated region could be replaced by longer EST or isolated cDNA sequences, increasing the chance of finding significant homology in a subsequent blastx.

The best blastx hit for each transcript was submitted to blastp against the protein sequences of the Arabidopsis thaliana database of the Institute for Genome Research. The Arabidopsis homologues of the differentially expressed transcripts were compared, at the amino acid level, to the differentially expressed genes in published Arabidopsis transcriptome analyses.

Metabolite profiling

Xylem was harvested and extracted as described previously (Damiani et al., 2005; Morreel et al., 2004a,b). Amino acid abundances were determined according to the method described by Rohde et al. (2004). Methanol-soluble phenolics were analysed by HPLC (Damiani et al., 2005). Commercially available monolignol biosynthesis intermediates were spiked to identify their position on the chromatogram. Other HPLC peaks were identified using procedures described in Appendix S1. In addition, 802 derivatized polar metabolites, of which 159 are known metabolites, were revealed by GC-MS. Information on the polar metabolites, using the corresponding mass spectral identifiers (ID), may be found at (Kopka et al., 2005; Schauer et al., 2005).


Sinapic acid and syringic acid were purchased from Janssen Chimica;, and quinic acid, ferulic acid, coniferaldehyde, sinapaldehyde, coniferyl alcohol and sinapyl alcohol were purchased from Sigma-Aldrich ( The synthesis of feruloyl tyramine is an unpublished method available from J. Ralph and F. Lu (USDA, University of Wisconsin).

Quantitative analysis of pigments (chlorophylls a and b, carotenoids and anthocyanins)

Chlorophyll a (ChlA), chlorophyll b (ChlB), carotenoids and anthocyanins were quantified in xylem, cortex and leaf tissue. ChlA, ChlB and carotenoids were extracted by adding 500 μl of 80% acetone to the samples, centrifugation at 5000 g, washing twice with 80% acetone, and centrifugation for 10 min at 5000 g. Their absorption was measured using a SPECTRAmax™ 250 microplate spectrophotometer (Molecular Devices Corporation; at 663, 646 and 470 nm. The concentrations of ChlA, ChlB and carotenoids were quantified according to the method described by Lichtenthaler and Wellburn (1983), as follows: ChlA (μg ml−1) = 12.21 A663– 2.81 A646; ChlB (μg ml−1) = 20.31 A646– 5.03 A663; carotenoids (μg ml−1) = [1000 A470– 1.82 ChlA (μg ml−1) – 85.02 ChlB (μg ml−1)]/198.

Anthocyanins were extracted by adding 2 ml of 0.1% HCl in methanol to the sample, followed by 48 h of incubation at 4°C and centrifugation for 10 min at 5000 g. By measuring the absorption at 530 and 675 nm, anthocyanins were quantified according to the method described by Rabino and Mancinelli (1986): anthocyanins = A530– (0.3 × A657).

Statistical analysis of amino acids, pigments and soluble phenolics

The concentrations of the amino acids, pigments and the soluble phenolics were calculated per mg dry weight, and the data analysed statistically as described previously (Morreel et al., 2004a,b). Briefly, significant differences (αmodel = 0.05; αline = 0.01; αenz = 0.05) were obtained by nested anova model (1):


where yijk is the value of the trait in individual k belonging to the jth tobacco line of the ith enzyme; μ is the overall mean trait value; enzi and linej(i) are the deviations from the overall mean due to the effects of the downregulation of the ith enzyme and the jth tobacco line for the ith downregulated enzyme, respectively; and εijk is the residual deviation of the trait value in individual k of the jth tobacco line for the ith downregulated enzyme. Box–Cox transformations were performed whenever necessary. If no homoscedasticity could be obtained for model (1), a one-way anova was applied (αmodel= 0.05). When appropriate, anova was followed by an LSD post hoc test (αLSD = 0.05).

Statistical analysis of GC-MS data

Data were analysed according to the method described by Damiani et al. (2005). Principal component analysis (PCA) of the GC-MS data revealed four main principal components (PC) that were able to separate the transgenic lines, explaining cumulatively 56.4% of the variance present in the dataset. Components that mainly contributed to these PC were determined when the absolute value of their loading factor was larger than 0.20 and larger than one standard deviation from the mean loading factor in that particular PC.

GC-MS profiles of WT, asCCR, asCAD and DT could be distinguished by PC1, PC3 and PC4. PC1 clustered asCAD partly independently from the other lines (see Figure S3). Nevertheless, values for asCAD were much more spread by PC1 than for the other lines because downregulation was obtained with either RNAi or antisense T-DNA constructs. Those individual PC1 values that afforded the partial separation all belonged to the more strongly RNAi downregulated CAD lines. PC3 distinguished asCCR and DT lines from WT plants, whereas PC4 separated asCCR lines from DT. Known metabolites that contributed predominantly to these PC and that were significantly differential based on a t test (α = 0.05), as well as those known metabolites that were only detected by a t test with a more stringent threshold (α = 0.001), are listed in Table 3.

Gas exchange and chlorophyll fluorescence analysis

Levels of photorespiration were determined from measurements of photosynthetic rates in air containing 21% O2 and 2% O2 using an infra-red CO2 gas analyser LI-6400 (LI-COR; Pre-mixed gases were supplied from pressurized tanks (Air America Liquide; Measurements were made at saturating photon flux density (1200 μmol m−2 sec−1), a leaf temperature of 25.5°C (SD = 0.7) and a vapour pressure deficit of 0.80 kPa (SD = 0.17).

A fluorescence chamber head (LI-6400-40; LI-COR) was integrated with the open gas exchange system LI-6400 to measure steady-state fluorescence (Fs), maximum fluorescence during a light saturating pulse (Fm′), and minimum fluorescence during far-red light illumination (Fo′). The photochemical quantum yield of electron transport through PSII was calculated as inline image (Genty et al., 1989). To account for the fact that PSII units are connected rather than independent, we used inline image as an estimate of the fraction of open PSII centers as proposed by Kramer et al. (2004). Because of the size of the LI-6400-40 leaf chamber fluorometer, it was not possible to avoid the lesions on the leaves of the CCR-downregulated tobacco plants.

Fv/Fm, i.e. the maximum quantum yield of primary photochemistry of PSII, was measured in the morning on dark-acclimatized (1 h) leaves using a ‘plant efficiency analyser’ (Hansatech; as described previously (Gielen et al., 2005). With this device, focus on the dark-green tissue was possible, thus avoiding the lesions. Photorespiration data and fluorescence measurements in the dark-acclimatized state were analysed by one-way anova followed by an LSD post hoc test using SAS version 8.2 with the mixed procedure (Littell et al., 1996). Because all data for asCCR and DT revealed similar differences and the fluorescence measurements obtained under light-acclimatized conditions were highly variable, the nested anova model (2) was applied using R, version 2.4.1


where yijk is the value of the trait in individual k belonging to the jth line of the ith tobacco group (I = 0 or 1 for WT or transgenic tobacco), μ is the overall mean trait value, tri is the deviation from the overall mean due to the effects of CCR downregulation, enzj(i) expresses the particular deviation of either CCR or both CCR and CAD, and εijk is the residual deviation of the trait value in individual k. A threshold level of 0.10 was used for all model parameters.


The authors thank Fachuang Lu for synthesizing the tyramine ferulate, Paul Schatz for providing the caffeoyl and feruloyl shikimic acids, Hoon Kim for oligolignol synthesis, Alexander Erban and Ines Fehrle for technical assistance with GC-MS metabolite profiling, and Martine De Cock for preparing the manuscript. This work was supported by the European Commission program (QLK5-CT-2000-01493; COPOL) and Research Foundation-Flanders (grant G.0352.05). R.D. is indebted to the Institute for the Promotion of Innovation by Science and Technology in Flanders for a pre-doctoral fellowship. B.G. and A.R. are Postdoctoral Researchers of the Research Foundation-Flanders.