Biosynthesis of cellulose-enriched tension wood in Populus: global analysis of transcripts and metabolites identifies biochemical and developmental regulators in secondary wall biosynthesis


  • Sara Andersson-Gunnerås,

    1. Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, 901 83 Umeå, Sweden,
    Search for more papers by this author
  • Ewa J. Mellerowicz,

    1. Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, 901 83 Umeå, Sweden,
    Search for more papers by this author
  • Jonathan Love,

    1. Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, 901 83 Umeå, Sweden,
    Search for more papers by this author
  • Bo Segerman,

    1. Umeå Plant Science Centre, Department of Plant Physiology, Umeå University, 901 83 Umeå, Sweden,
    Search for more papers by this author
  • Yasunori Ohmiya,

    1. Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, 901 83 Umeå, Sweden,
    Search for more papers by this author
  • Pedro M. Coutinho,

    1. Architecture et Fonction des Macromolécules Biologiques, UMR6098, CNRS, Universités Aix-Marseille I & II, case 932, 163 Avenue de Luminy, 13288 Marseille cedex 9, France, and
    Search for more papers by this author
  • Peter Nilsson,

    1. Department of Biotechnology, Royal Institute of Technology, 100 44 Stockholm, Sweden
    Search for more papers by this author
  • Bernard Henrissat,

    1. Architecture et Fonction des Macromolécules Biologiques, UMR6098, CNRS, Universités Aix-Marseille I & II, case 932, 163 Avenue de Luminy, 13288 Marseille cedex 9, France, and
    Search for more papers by this author
  • Thomas Moritz,

    1. Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, 901 83 Umeå, Sweden,
    Search for more papers by this author
  • Björn Sundberg

    Corresponding author
    1. Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, 901 83 Umeå, Sweden,
      *(fax +46 90 786 8165; e-mail
    Search for more papers by this author

*(fax +46 90 786 8165; e-mail


Stems and branches of angiosperm trees form tension wood (TW) when exposed to a gravitational stimulus. One of the main characteristics of TW, which distinguishes it from normal wood, is the formation of fibers with a thick inner gelatinous cell wall layer mainly composed of crystalline cellulose. Hence TW is enriched in cellulose, and deficient in lignin and hemicelluloses. An expressed sequence tag library made from TW-forming tissues in Populus tremula (L.) × tremuloides (Michx.) and data from transcript profiling using microarray and metabolite analysis were obtained during TW formation in Populus tremula (L.) in two growing seasons. The data were examined with the aim of identifying the genes responsible for the change in carbon (C) flow into various cell wall components, and the mechanisms important for the formation of the gelatinous cell wall layer (G-layer). A specific effort was made to identify carbohydrate-active enzymes with a putative function in cell wall biosynthesis. An increased C flux to cellulose was suggested by a higher abundance of sucrose synthase transcripts. However, genes related to the cellulose biosynthetic machinery were not generally affected, although the expression of secondary wall-specific CesA genes was modified in both directions. Other pathways for which the data suggested increased activity included lipid and glucosamine biosynthesis and the pectin degradation machinery. In addition, transcripts encoding fasciclin-like arabinogalactan proteins were particularly increased and found to lack true Arabidopsis orthologs. Major pathways for which the transcriptome and metabolome analysis suggested decreased activity were the pathway for C flux through guanosine 5′-diphosphate (GDP) sugars to mannans, the pentose phosphate pathway, lignin biosynthesis, and biosynthesis of cell wall matrix carbohydrates. Several differentially expressed auxin- and ethylene-related genes and transcription factors were also identified.


Wood, which is composed chiefly of cell walls, is one of the major carbon (C) reservoirs on Earth. The wall of a wood cell consists of compound middle lamella and secondary wall layers (S1, S2 and S3) composed of a cellulose/hemicellulose network impregnated with lignin. In the wood of Populus spp., these polymers occur in approximate proportions of 45% cellulose, 25% hemicellulose and 20% lignin (McDougall et al., 1993; Timell, 1969). In many angiosperm trees, including Populus spp., the normal development of secondary wall biosynthesis is greatly modified when tension wood (TW) is formed as a gravitational response to stem movements caused, for example, by wind or load (Hellgren et al., 2004; Mellerowicz et al., 2001; Pilate et al., 2004). TW is asymmetrically induced on one side of the stem and characterized by fibers with a gelatinous cell wall layer (G-layer) that normally replaces most of the S2 and the entire S3 layer. The G-layer in Populus tremula (L.) consists mainly of highly crystalline cellulose with microfibrils oriented parallel to the fiber axis, and some matrix carbohydrates (Fujita et al., 1974; Norberg and Meier, 1966). Whether lignin is present in the G-layer is still a matter of debate (Joseleau et al., 2004), but if present at all its content is much reduced compared to that of the S-layers (Timell, 1969). In addition to having fibers with a G-layer (G-fibers), TW is characterized by a reduced number of vessel elements, and it is often formed at an accelerated rate compared with normal wood. Formation of G-fibers is unique to woody species and demonstrates that the flow of C into the various wood polymers can be dramatically modified. The resulting TW has an overall increase in cellulose content of about 10–20% (depending on species) compared with normal wood, and a corresponding decrease in lignin and hemicelluloses (Timell, 1969).

TW can be induced experimentally, and the modified structure and chemistry of the G-fiber makes TW formation an attractive experimental system to explore the development and biochemical pathways of secondary cell wall formation, and the control of C flow into lignin, cellulose and hemicellulose. Limited information is currently available on the mechanisms underlying TW formation. Plant hormones, such as auxins, ethylene and gibberellins, have been suggested to mediate the gravitational response resulting in TW formation, but their role and mechanisms of action are not conclusively established (Andersson-Gunnerås et al., 2003; Hellgren et al., 2004; Mellerowicz et al., 2001). Several unidentified proteins were found to be induced in TW-forming tissues (Baba et al., 2000), and recently expressed sequence tag (EST) libraries from TW-forming tissues have been constructed from Populus tremula × alba (L.) (Dejardin et al., 2004) and Populus tremula (L.) × tremuloides (Michx.) (Sterky et al., 2004). These libraries are rich in cDNAs encoding several wall-associated proteins, most notably several arabinogalactan proteins (AGPs) (Lafarguette et al., 2004). As a further step in understanding the molecular mechanisms resulting in G-layer formation, we have used a 13 000 poplar microarray (Andersson et al., 2004) together with a metabolomic approach to search for developmental regulators in G-fiber formation, and important components in the biosynthesis of wood cell walls. We identified key steps for the divergence of the C flow from lignin and hemicellulose to cellulose biosynthesis, and genes encoding components of hormone signaling pathways and transcription factors differentially expressed between TW and normal wood. By using the recently identified gene models from the poplar genome-sequencing project (, annotation was much improved and identification and separation of gene members in large gene families were facilitated.

Results and discussion

Formation of TW differs in three major aspects from formation of normal wood. The rate of wood production is increased, the number of vessel elements that differentiate from the vascular cambium is decreased, and the cell wall biosynthesis in wood fibers is modified by the production of the G-layer. Accelerated growth and developmental decisions regarding the cell type formed are determined in the meristematic cambial zone, whereas the formation of G-layers takes place late during xylem differentiation (Timell, 1986). This study focused on G-layer formation, and the analyzed tissues consisted of developing xylem in the stage of secondary wall formation (excluding the cambial zone and early stages of xylem cell expansion). The data should therefore be interpreted in the light of secondary wall biosynthesis, i.e. the replacement of most of the S2 and the entire S3 layer by a G-layer. In analogy to similarly induced compression-wood tracheids in conifers (Timell, 1986) and our own observations in Populus (E. J. Mellerowicz and B. Sundberg, unpublished data), it can be assumed that differentiating xylem cells respond to the gravitational signal during the developmental stage of the secondary wall. Therefore the analyzed tissue should reveal genes involved both in the induction and in the biosynthesis of the G-layer.

Highly abundant transcripts in TW-forming tissues

To complement the poplar microarray with genes expressed during TW formation, an EST library was constructed from secondary wall-forming TW tissues, and 5723 cDNA clones were sequenced, clustered and annotated together with other poplar sequences (Sterky et al., 2004; The availability of the genome sequence now offered the opportunity to use gene models for establishing the identity of ESTs. Similarity among P. tremula, P. tremuloides and P. trichocarpa was found to be as high as 98% in the coding sequences (J. Geisler-Lee and E. J. Mellerowicz, unpublished results) and therefore it was possible to map P. tremula × tremuloides ESTs onto P. trichocarpa gene models.

Genes with the most abundant ESTs in the TW library are presented in Figure 1. In some cases these genes represent large gene families whose members, although similar in general function, may be specialized in different developmental processes or biochemical pathways. The sampled TW tissues were enriched not only in transcripts involved in the G-layer formation, but also in transcripts functioning in the formation of the S1 and early S2 layers. Thus, many of the abundant transcripts were also highly represented in other libraries from wood-forming tissues, and according to subsequent microarray analysis they were not affected, or even decreased, during TW formation, as in the case of the methionine metabolism genes (Figure 1). Many of the well-represented genes were related to cell wall formation, especially to carbohydrate metabolism and cytoskeleton. Some housekeeping genes, the fasciclin-like arabinogalactan proteins (FLAs), and two genes with unknown function were also found to be highly expressed in TW.

Figure 1.

Most abundant transcripts in the tension wood (TW) EST library. Genes with more than 14 ESTs in the TW library (black bar, total 5,723 ESTs) are shown. The number of ESTs in the cambial region libraries (light gray bar; total 6326 ESTs), in the wood cell death library (dark gray bar; total 4867 ESTs) and in the remaining libraries from non-woody tissues (white bar; total 85 103 ESTs) in the Populus database is also given. An arrow indicates increase or decrease of a transcript level in the microarray experiment. Significance level: ***, P ≤ 0.001; **, P ≤ 0.01; *, P ≤ 0.05; n.s., not significantly regulated; no poplar unigene ID (PU ID), no EST representing this gene was spotted on the microarray. 1, estExt_fgenesh4_pg.C_280066; 2, grail3.0066005802; 3, estExt_Genewise1_v1.C_LG_XIV0850; 4, eugene3.00180798; 5, estExt_fgenesh4_pm.C_LG_XVIII0009; 6, estExt_fgenesh4_pg.C_LG_II1020; 7, eugene3.00141104; 8, estExt_Genewise1_v1.C_LG_XIX1125; 9, estExt_fgenesh4_kg.C_LG_IV0063; 10, grail3.0028013201; 11, gw1.IX.2621.1; 12, estExt_Genewise1_v1.C_LG_I3343; 13, estExt_fgenesh4_pg.C_LG_X1860; 14, estExt_fgenesh4_pg.C_1700003; 15, grail3.0124006001; 16, estExt_fgenesh4_pm.C_LG_XIV0257; 17, eugene3.00040471; 18, estExt_fgenesh4_pm.C_LG_I0969; 19, grail3.0263001401; 20, eugene3.00151077; 21, estExt_fgenesh4_pg.C_LG_II0928; 22, grail3.0049008501, 23, eugene3.00091518; 24, estExt_Genewise1_v1.C_LG_X0543; 25, grail3.0050012801; 26, estExt_Genewise1_v1.C_640357.

Carbohydrate-active enzyme families of TW and their significance in cell wall formation

The TW EST library was specifically mined for carbohydrate-active enzyme (CAZyme)-encoding genes by searches against annotated entries found in the CAZy database (Coutinho and Henrissat, 1999) ( CAZyme genes are likely to encode most important known but not yet characterized cell wall biosynthetic enzymes. We could identify 119 gene models, representing 37 CAZyme families, including glycosyl transferases (GTs), glycoside hydrolazes (GHs), polysaccharide lyases (PLs) and carbohydrate esterases (CEs) (Figure 2). The most abundant families (GT4, GH9, GT2 and GH19) include enzymes directly or indirectly involved in cellulose biosynthesis. Other abundant families (GH17, GT47, expansins, PL1, GT8 and GH16) include wall matrix biosynthetic enzymes and cell wall degrading and modifying enzymes. These enzymes are described in subsequent sections.

Figure 2.

Number of EST clones corresponding to carbohydrate active enzymes (CAZymes) and expansins in the tension wood (TW) library. EST numbers in other libraries from woody and non-woody tissues are given for comparison as in Figure 1.

Identifying genes differentially expressed in TW

The poplar microarray (Andersson et al., 2004) was used to identify transcripts differentially regulated during TW formation. TW was induced by bending field-grown trees which were harvested when typical G-layers were forming in all differentiating xylem cells. The experiment was repeated in two growing seasons to obtain independent biological replicates. The microarray data were analyzed using a two-stage analysis of variance (anova) mixed model (Wolfinger et al., 2001) to account for sources of systematic errors, such as position on the microarray, dye and hybridization conditions. The poplar microarray has been used in a number of earlier published experiments ( and the reliability of the results was verified by adequate quality control experiments. Independent experiments by us and others using real-time reverse transcrpitase–polymerase chain reaction (RT-PCR) and Northern blot for, for example, genes encoding cellulose synthase catalytic subunit (CesAs), AGPs and ACC oxidase, as discussed below, served as positive controls for our results. In addition, we performed a small quality test for some genes with decreased transcript abundance in TW with an acceptable outcome, as follows: PU08184-0.17/0.38., 0.32/0.39 (Northern ratio/microarray ratio, for both years), PU02436-0.42/0.49, PU00110 0.31/0.33, 0.45/0.30.

In the first bulk analysis to identify differentially regulated genes, we used a stringent significance level of P ≤ 0.001 and a cut-off filter of a minimum of ±30% in transcript abundance. In cases where several clones representing the same gene were present on the microarray, the data for the most affected clone are presented. After removing clones representing the same gene model, 444 genes were found to be differentially expressed (235 up- and 209 down-regulated) compared with control developing wood tissues in the upright trees. Among these 444 genes, c. 75% could be assigned a putative function. A selection of the most affected transcripts is presented in Table 1 and the full list of differentially expressed genes is presented in Table S1. In the subsequent analysis of specific functions and/or metabolic pathways discussed below, we have also considered differentially expressed genes with a lower level of significance and/or up- and down-regulation. The use of a less stringent significance level can be justified when a limited set of genes is analyzed. For closely related gene family members, it was sometimes observed that highly abundant wood-specific transcripts cross-reacted with, for example, leaf-specific cDNAs as determined from EST library frequencies. Therefore we have considered signal strength as well as relative EST frequency in wood-related libraries when identifying key genes.

Table 1.  Differentially regulated genes in TW compared with normal wood
PU IDDescriptionAverage ratio TW/CAGI IDScore
  1. (A) Genes functionally classified to carbohydrate metabolism, cell walls, and transcription factors. Only the genes that differed more than ±30% in expression with a P ≤ 0.001 (ANOVA) are shown.

  2. (B) Genes related to cell cycle, cell defense, cellular communication, cellular transport, energy, metabolism, protein fate and protein synthesis and unclassified. Only transcripts modified more than ±100% with a P ≤ 0.001 (ANOVA) are shown.

Carbohydrate metabolism (ratio > +/30%)
PU05676Ribokinase4.07At1g 196001405
PU07211UDP-glucose pyrophosphorylase, PHUGP22.31At5g173102079
PU07012Glucose-6-phosphate 1-dehydrogenase1.50At5g131102441
PU08630Glucose-6-phosphate 1-denydrogenase0.76At5g407602356
PU02384D-3-phosphoglycerate dehydrogenase0.73At4g342002432
PU01407DAHP synthetase0.72At1g224102212
PU06153UDP-D-xylose 4-epimerase, MUR40.67At1g306201826
PU02218DAHP synthetase0.61At1g224102216
PU06682GDP-mannose pyrophosphorylase0.58At2g397701691
PU11304Phosphoenolpyruvate carboxylase kinase0.58At3g04530962
PU11655Dehydroquinate dehydratase/shikimate: NADP oxidoreductase0.47At3g063502001
PU02562GDP-mannose pyrophosphorylase0.44At2g397701700
PU08268D-3-phosphoglycerate dehydrogenase, 3-PGDH0.41At1g177452324
PU00110Malate dehydrogenase, cytosolic0.35At1g044101463
PU03052GDP-mannose pyrophosphorylase0.34At2g397701732
Cell wall (ratio > +/30%)
PU02114Fasciclin-like arabinogalactan-protein, PtFLA12K18.90At5g60490547
PU06339Fasciclin-like arabinogalactan-protein, PtFLA12E18.26At5g60490527
PU06797Fasciclin-like arabinogalactan-protein, PtFLA12V18.26At5g60490576
PU06458Fasciclin-like arabinorjalactan-protein, PtFLA12D17.75At5g60490517
PU07802Fasciclin-like arabinogalactan-protein, PtFLA12S13.80At5g60490547
PU07326Fasciclin-like arabinogalactan-protein, PtFLA12Q13.35At5g60490537
PU07173Fasciclin-like arabinogalactan-protein, PtFLA12F10.75At5g60490519
PU07213Fasciclin-like arabinogalactan-protein, PtFLA12J9.56At5g60490550
PU07966Fasciclin-like arabinogalactan-protein, PtFLA12P8.86At5g60490537
PU07280Fasciclin-like arabinogalactan-protein, PtFLA12G6.79At5g60490501
PU06434COBRA protein, COBL44.54At5g156301843
PU08742Proline-rich family protein, PRP42.45At4g38770678
PU06569Microtubule-associated protein EB1-like protein2.42At3g476901036
PU08244Peroxidase 12, PER122.11At1g716951145
PU06848Kinesin motor protein-related1.87At3g101803643
PU01104Cupin family protein1.74At2g18540155
PU12639Fasciclin-like arabinogalactan-protein, FLA101.69At3g609001525
PU08631Proline-rich family protein.1.67At4g38770695
PU06488Peroxidase 12, PER121.59At1g716951131
PU07125Proline-rich family protein1.54At3g22o70270
PU08476Proline-rich family protein, PRP41.39At4g38770381
PU02638Caffeic acid O-methyltransferases, COMT10.74AI5g541601475
PU03723Hydroxyproline-rich glycoprotein family protein0.74AI2g2824083
PU03517Cinnamoyl-CoA reductase, CCR0.71At1g159501383
PU00591Cinnamoyl-CoA reductase, CCR10.71At1g159501393
PU03752Myosin heavy chain-related0.69At1g77580743
PU02804Laccase, LAC30.67At2g380802058
PU01735Germin-fike protein, GLP100.67At3g62020834
PU11256Hydfoxyproline-rich giycoprotein family protein0.65At1g083701184
PUQ2719Phenylalanine ammonia-lyase, PAL10.65At2g370403062
PU12018Microtubule-associated protein0.64At4g299502187
PU06275Kinesin motor protein-related0.63At5g275501767
PU08465Myosin heavy chain-related0.63At2g14680277
PU01728Arabinogalactan peplide, AtAGPl4-like0.62AI5g56540188
PU07065Germin-lika protsin, GLP100.60At3g62020845
PU01147Kinesin light chain gene-like0.60At1g275002098
PU02461Caifeoyi-cosnzyms A O-methyltransferase, CCoAOMT20.59At4g340501151
PU01148Ferulaie-5-hydroxylase, F5H0.58At4g362202070
PU00666Myosin class 11 heavy chain0.55At1g03080334
PU02041Kinesin motof protein-related0.44At1g555501655
PU00572Kinesin motor protein-related0.41At1g555501707
Transcription (ratio > +/30%)
PU03049CND-binding protein, putative13.19At5g10770713
PU09933Zinc finger family protein6.03At3g21890274
PU06749Zinc finger (GATA type) family protein5.31At3g54Blo507
PU04164Zinc finger (C3HC4-type RING finger) family protein4.70Atg273301171
PU07328WRKY transcription factor, WRKY203.47At4g266401520
PU04307TCP family transcription factor, putative2.22At3g27010771
PU04669Linker histone protein-putative2.16At1g 14900519
PU06842Zinc finger (C3HC4-type RING finger) family protein2.00At2g01275640
PU06790Zinc finger (C3HC4-type RING finger) family protein1.86At5g019601448
PU01258Transcription factor LIM1.85At1g10200841
PU00292Transcription factor LIM1.82At1g 10200847
PU07724Homeodomain transcription factor,1.63At1g629901196
PU06184bHLH transcription factor, putative1.57At2g43060214
PU07558Zinc finger (CCCH-type) family protein1.50At2g419002195
PU01433Heterogeneous nuclear ribonucleoprotein (hnRNP), putative1.49At3g078101427
PU06356Transcription factor NtWRKY4-like1.44At5g562701272
PU06901Transcription factor NtWRKY4-like1.41At5g562701325
PU09337WRKY DMA-binding protein1.35At1g139601240
PU06550RNA recognition motif (RRM)-containing protein1.31At2g16940930
PU01598Myb family transcription factor, PttMYB21a1.31At1g17950567
PU03175Scarecrow-like transcription factor, SCL10.76At1g214501848
PU07028NAM like protein0.70At4g285001008
PU08786Basic helix-loop-helix (bHLH) family protein0.70At4g34530541
PU03311Histone H2B, putative0.69At5g2288074
PU00263NAM like protein0.67At4g28500973
PU01538Homeodomain-containing protein0.66At2g16400994
PU07240Zinc finger (C3HC4-type RING finger) family protein0.66At3g19910544
PU02947Remorin family protein0.65At5g23750660
PU05553Transcription initiation factor TFIID-20.62At1g55520938
PU05034Transcription elongation factor-related0.61At5g05140688
PU04208Zinc finger (C3HC4-type RING finger) family protein0.59At3g02290740
PU02436COP1 -interacting protein, CIP70.59At4g27430844
PU09430WRKY family transcription factor0.55At2g4400362
PU11043Zinc finger (C3HC4-type RING finger) family protein0.47At1g15100394
PU00374NAM-like protein0.34At2g180601169
Cell cycle ( ratio > +/100%)
PU06293Cell division control protein 2 homolog A4.22At3g487501369
Cell defence (ratio > +/100%)
PU12810Peptide methionine sulfoxide reductase, putative0.39At5g61640658
Cellular communication (ratio > +/100%)
PU05580LLR family protein2.42At2g20850535
PU11285Protein kinase, putative0.49At1g090201774
PU11549LLR family protein0.47At3g43740705
PU01413LLR family protein0.44At5g634101730
Cellular transport (ratio > +/100%)
PU06614Clathrin assembly protein AP180homolog14.89At5g572001918
PU06101Phosphate/phosphoenol pyruvate translocator3.91At5g058201252
PU01790Transporter-related, GDP-Mannose transporter2.85At1g210701457
PU12026Glucose-6-phosphate/phosphate translocator-related2.62At1g776101593
PU02146Cation efflux family protein2.40At2g04620276
PU06175Intracellular protein transport protein, putative2.06At3g550601957
PU07552Mitochondrial substrate carrier family protein2.01At1g253801172
PU11237Mitochondrial phosphate transporter0.46At5g140401404
PU10105Autophagy 8i (APGSi)0.39At3g15580471
Energy (ratio > +/100%)
PU06238Lipin family protein7.26At5g42870923
PU06751NADP-dependent malic enzyme3.16At1g797502505
Metabolism (ratio > +/100%)
PU06465Cytochrome p450 family4.64At3g52970961
PU13121Decarboxylase family protein, putative3.43At4g35190882
PU011832-oxoglutarate-dependent dioxvaenase3.27At1g06620870
PU11130Cytochrome p450,2.17At1g649501664
PU00211Aldehyde dehydrogenase homolog, putative2.11At1g54l002248
PU01618Glutamine synthetase, putative0.47At5g376001759
No class (ratio > +/100%)
PU06799Expressed protein12.49At3g15351142
PU06337Expressed protein9.46At2g442601894
PU06931Expressed protein8.05At4g34215538
PU08796Expressed protein6.55At5g4305050
PU06489Expressed protein5.34At5g157402108
PU06389Expressed protein4.27At4g33140913
PU04375Expressed protein3.27At1g27930960
PU07149Expressed protein2.83At2g35880541
PU09680Expressed protein2.80At4g201701638
PU06840VQ motif-containing2.71At1g28280556
PU07094Expressed protein2.39At4g201701407
PU04784Expressed protein2.17At3g19310998
PU06435Expressed protein2.08At2g240701082
PU06800Expressed protein2.08At2g35880533
PU06018Expressed protein2.05At5g16720463
PU04754Expressed protein2.05At5g426901049
PU08533Expressed protein0.47At3g21570151
PU08184Tetraspanin-like protein0.40At4g28050942
PU03150Expressed protein0.35At4g39235315
Protein fate (ratio > +/100%)
PU06639F-box family protein5.28At3g13570542
PU02345Cysteine proteinase XCP20.42At1g208501404
PU00170DNAJ heat shock N-terminal domain-containing protein0.40At3g14200494
PU07286Cysteine proteinase XCP20.38At1g208501381
PU02394Cysteine proteinase precursor0.21At1o471281133
Protein synthesis (ratio > +/100%)
PU06768Ribosomal protein L5 - like6.69At5g397401295
PU00399Translation initiation factor 3 (IF-3) family protein2.51At2g24060818

Identifying metabolites differentially represented in developing TW

The same tissues that were used for the microarray analysis were also used for global analysis of metabolites by gas chromatography/time-of-flight mass spectrometry (GC/TOF-MS) after extraction and derivatization of extracts according to a method by Gullberg et al. (2004). More than 350 peaks were detected, and about 25% of them were identified by comparing their retention index and mass spectra with mass spectra libraries. Multivariate analysis by partial least squares-discriminate analysis (PLS-DA) revealed that 26 metabolites were significantly changed at P ≤ 0.01. Sixteen of these compounds could be identified (Table 2). Among these there were several carbohydrate-related metabolites such as sucrose, arabinose and inositol, which were all decreased in TW, whereas xylose and xylitol were more abundant in TW. Two fatty acid metabolites were increased in TW. Other metabolites with decreased abundance were shikimate, a precursor of secondary metabolites including monolignols, and gamma-butyric acid (GABA), suggested to be involved in the regulation of carbon/nitrogen balance and/or to act as a signaling molecule (Bouche and Fromm, 2004).

Table 2.  Differentially abundant metabolites in developing TW compared with normal wood
CompoundRatio (TW/C)Retention Index LibraryaBase Peakb (m/z)
  1. The metabolites were identified using PLS-DA analysis (P ≤ 0.01).

  2. aMass spectra libraries used for the identification of metabolites: U = user library; M =; N = NIST 98.

  3. bBase peaks in the obtained mass spectra, m/z 73 is not included as base peak.

  4. cEITTMS_N12C_ATHF_2021.546.

Amino acid
Fatty acid
 Oleic acid1.192215U96
 Linoleic acid1.112209U81
Organic acid
 Shikimic acid0.871809U204
 Pentonic acid0.841748M147
 Citric acid0.801866U147
 Galaciaric acid0.662037M147
 2-Hydroxyhexane- dioic acid0.491668M129
 Uracif0.851338M. N99
Sugar alcohol
No class
 PtXyfTWQORI324Q2.643240 108
 PtXylTWOORI17851.391785 89
 PtXylTWOORI16731.271673 117
 PtXylTWCX)RI23501.122350 147
 PtXylTWOORI17760.731776 217
 PtXylTWOORI20520.692052 147
 PtXylTWOOR]23620.632362 204
 PtXylTWOQRI17990.581799 117
 PtXylTWOORI14700.481470 155

Fasciclin-like AGP transcripts are highly abundant during TW formation

Among the most abundant and most increased transcripts in TW there were several fasciclin-like AGPs (FLAs) (Figure 1, Table 1). Arabidopsis has 21 FLAs divided into four subgroups (A–D) based on the presence of one or two fasciclin domains, one or two AGP domains and a glycosylphosphatidylinositol (GPI) anchor (Johnson et al., 2003). The FLAs that were greatly increased in TW were similar to subgroup A FLAs. We identified 24 genes representing this subgroup by searching the poplar genome for the fasciclin domain. The phylogenetic analysis (Figure 3) showed that AtFLA7, 11 and 13 had two putative poplar orthologs, whereas AtFLA12 had as many as 22 similar genes in poplar, giving rise to a large difference in the genetic diversity within subgroup A between the two species. Several hundred FLA-like ESTs were identified in the Populus database (, more than 200 of these belonging to subgroup A. The majority of them were present in wood-forming tissues (Table S2), indicating the importance of subgroup A FLAs in the formation of secondary walls. This is further supported by a recent study demonstrating the co-regulation of AtFLA11 and AtFLA12 with secondary wall-specific CesA genes (Persson et al., 2005). Interestingly, the FLA genes with increased transcript abundance in TW were the ones most divergent from AtFLA12 (Figure 3a). Our findings confirm a recent report by Lafarguette et al. (2004) who used quantitative RT-PCR to find 10 members of subgroup A up-regulated in TW. In addition, transcript levels for a subgroup C AtFLA10-like gene and a lysine-rich AGP similar to AtAGP19 were increased in TW (Table 1a). The poplar homolog of AtAGP19 was also similar to the PtAGP6 that was recently reported to have a role in the wood-forming tissues of loblolly pine (Yang et al., 2005). The most abundant AGP in wood EST libraries was an AG peptide similar to AtAGP14 (data not shown). This AGP was the only AGP transcript found to be decreased in TW (Table 1a).

Figure 3.

Phylogenetic analysis and expression of subgroup A fasciclin-like arabinogalactan proteins (FLAs).
(a) Phylogenetic tree of subgroup A FLAs from Arabidopsis and Populus spp. Populus genes present on the microarray are indicated with numbers followed by capital letters. The transcript levels of genes shaded in gray were greatly increased in tension wood (TW) compared with normal wood (P ≤ 0.001). Names of published Populus tremula × alba and Populus × canescens Ait. genes are also indicated. clustalw was used to align protein sequences derived from gene models and the unrooted dendrogram was obtained using treeview (version 1.6.6). Arabidopsis Genome Initiative (AGI) codes, GenBank accession numbers and PU IDs or gene models (when PU numbers were missing) for the genes used were as follows: Arabidopsis thaliana, AtFLA6 (At2g20520), AtFLA7 (At5g65390), AtFLA9 (At1g03870), AtFLA11 (At5g03170), AtFLA12 (At5g60490), AtFLA13 (At5g44130); Populus × canescens, Pop14A9 (AAD56235); Populus tremula × alba, AGP1-AGP14 (AY607753-AY607766); Populus trichocarpa, 7a (estExt_Genewise1_v1.C_LG_II3738), 7B (PU02553), 11A (PU00903), 11B (PU07340), 12A (eugene3.00012224), 12B (PU06862), 12C (PU01738), 12D (PU06458), 12E (PU06339), 12F (PU07173), 12G (PU07280), 12h (eugene3.10810001), 12i (eugene3.00130132), 12J (PU02214), 12K (PU02214), 12l (fgenesh1_pg.C_LG_XIX001005), 12m (eugene3.06840001), 12n (fgenesh1_pg.C_scaffold_858000001), 12o (eugene3.00191024), 12P (PU07966), 12Q (PU07326), 12r (fgenesh1_pg.C_scaffold_1081000003), 12S (PU07802), 12t (eugene3.00191022), 12u (eugene3.00660250), 12V (PU06797), 13a (grail3.0048017501), 13B (PU09032).
(b) Expression profiles of P. tremula×tremuloides FLA genes similar to AtFLA11 and AtFLA12 in developing secondary xylem in upright trees. The figure shows that transcripts greatly increased during tension wood (TW) formation (solid line corresponding to 12D, 12E, 12F, 12G, 12V, 12J, 12K, 12P, 12Q and 12S) are expressed later in xylem development as compared with those that were not increased (dotted line corresponding to 11A, 11B, 12B and 12C). A, cambial zone; B, early radial expansion; C, late radial expansion; D, secondary wall formation; E, late maturation. Data from Schrader et al. (2004) are presented as log2 ratios.

The function of FLAs in wood formation is unknown. However, as Arabidopsis does not form TW, it is tempting to speculate that the TW-induced FLAs that lack true Arabidopsis orthologs acquired a specific function during TW formation, such as the biosynthesis of the G-layer. The presence of AGPs in the G-layer and at the plasma membrane in G-fibers was demonstrated by Lafarguette et al. (2004). However, the FLA transcripts that increased during TW formation were also expressed in control trees (according to the signal intensity on the microarray) and present in other wood-related libraries (Table S2), suggesting their general role in secondary wall formation. It is notable that TW-induced subgroup A FLA genes all had their peak of expression at the late stage of secondary wall formation in upright trees, whereas the non-induced ones peaked at earlier stages of xylogenesis (Schrader et al., 2004) (Figure 3b).

Proteins containing fasciclin domains were shown to function as adhesion molecules in Drosophila melanogaster and Volvox carteri(Elkins et al., 1990; Huber and Sumper, 1994). In Arabidopsis, a mutation in the fasciclin domain of the SOS5 protein (AtFLA4) resulted in aberrant cell expansion (Shi et al., 2003). If AGPs act as molecular chaperones of different carbohydrates in the wall as speculated by Carpita and Gibeaut (1993), their abundance in different tissues and plant taxa would simply reflect differences in cell wall composition.

TW formation involves reprogramming of carbohydrate metabolism

The switch in cell wall biosynthesis from cellulose-, hemicellulose- and lignin-rich S-layers to cellulose-rich G-layers formed in the developing TW should involve a redirection of C metabolism. The transcriptome and metabolome data have been integrated into metabolic pathways to enable the key genes involved in this metabolic switch to be visualized. The data suggest increased activity for cellulose, lipid and glucosamine biosynthesis and pectin degradation, and decreased activity for C flux through guanosine 5′-diphosphate (GDP) sugars to mannans, the pentose phosphate pathway, lignin biosynthesis, and biosynthesis of cell wall matrix carbohydrates (Figure 4). These pathways are described in more detail below.

Figure 4.

Gene regulation of major carbon fluxes during tension wood (TW) formation. Pathways are from and Modification of transcript abundance (see Tables 1 and 3 and Supplementary Material for more information) and metabolites (Table 2) are indicated by color: red, increase; blue, decrease; gray, not present on the microarray; black, not affected. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. Genes significant for wood formation were identified according to signal strength and their relative abundance in wood-related EST libraries. The following genes (PU IDs) were included: α-amylase (PU07898); acetyl CoA C-acyltransferase (PU06448); β-fructosidase (PU06645); β-xylosidase, GH3_10 (PU02991) and GH3_24 (PU07336); cellulase, PttCel9A1-1 (PU02263), PttCel9A1-2 (PU06624), PttCel9B (PU00431); cellulose synthase, PttCesA1 (PU02217), PttCesA8-2 (PU05518), PttCesA3-2 (PU07525); UDP-D-Glc epimerase-reductase 1 (PU00967); fructokinase FRK1 (PU02482), FRK2 (PU06741) and FRK3 (PU06359); fructose-1,6-bisphosphate (PU09376); GDP-D-Man 3,5-epimerase (PU08282); GDP-Man 4,6-dehydratase MUR1 (PU12140); GDP-Man-pyro-phosphorylase 1 (PU06682), 2 (PU02562), and 3 (PU03052); glucose-6-P isomerase (PU12389); hexokinase (PU05835, PU09308); malate oxidoreductase (PU06751); malate dehydrogenase (PU00110); malate synthase (PU04017); pectate lyase, PL1_17 (PU01059), PL1_18 (PU003060), PL1_19 (PU00659), PL1_26 (PU00707), PL1_27 (PU01542); pectinesterase, CE8_66 (PU9266), CE8_87 (PU11859), PttPME1 (PU08008); PEP carboxykinase 2 (PU11304); phosphoglucomutase (PU04213); P-Man-mutase (PU08660); polygalacturonase, GH28_9 (PU09548), PttGH28A (PU01904); PttGT2A-1 (PU02290); PttGT8D (PU02889); UDP-D- glucose pyro-phosphorylase 2, PttUGP2 (PU07211); pyruvate kinase 1 (PU06747) and 2 (PU04889); pyruvatedehydrogenase (PU09583); SPP, sucrose phosphate phosphatase (PU04760); SPS, sucrose phosphate synthase (PU03689); sucrose synthase, PttSUS1 (PU07221), PttSUS2 (PU00401); transaldolase (PU05543); transketolase (PU02587); UDP-D-glucose 6-dehydrogenase (PU06807, PU08248, PU00484); UDP-D-glucuronic acid 4-epimerase 1-1 (PU11321), 1-4 (PU01508), and 6-2 (PU06427); UDP-D-glucuronic acid decarboxylase (PU02592); UDP-N-acetyl-glucosamine pyro-phosphorylase 1-1 (PU05407) and 1-2 (PU01077); UDP-N-acetylglucosamine:peptide N-acetylglucosaminyltransferase (PU00574); UDP-Xyl 4-epimerase 1-1 (PU06153) and 1-2 (PU07781). Fruc, fructose; Gal, galactose; Glc, glucose; GlcA, glucuronic acid; Man, Mannose; PEP, phosphoenolpyruvate; RGI, rhamnogalacturonan I; Rhm, rhamnose; Suc, sucrose.

Cellulose biosynthesis

The CesA gene family, which is involved in cellulose polymerization, has three specific members in Arabidopsis (AtCesA8, AtCesA7 and AtCesA4) which are responsible for secondary wall formation (Doblin et al., 2002). The corresponding genes in Populus spp., PttCesA3-1, PttCesA3-2 (PtrCesA1), PttCesA9 (PtrCesA2) and PttCesA1 (PtrCesA3), have been identified and demonstrated to be highly expressed during normal wood formation (Djerbi et al., 2004; Joshi et al., 2004). PttCesA3-2 and PttCesA8-2 [an additional CesA gene similar to Arabidopsis radial swelling 1 (rsw-1), also known as AtCesA1] transcripts were slightly increased in TW, whereas PttCesA1 transcripts were decreased (Figure 4, Table 3). The increase of PttCesA3-2 transcripts in TW was also indicated by real-time RT-PCR with gene-specific primers (Djerbi et al., 2004). The promoter of PtrCesA1 responded to mechanical stress in tobacco (Wu et al., 2000) and, although its induction by a gravitational stimulus was not studied by Wu et al., data from Djerbi et al. (2004) and from this report support the increase of PttCesA3-2/PtrCesA1 transcripts during TW formation. The increase of PttCesA8-2 and PttCesA3-2 transcripts and the decrease of PttCesA1 transcripts suggest that different CesA isoforms might be present in the biosynthetic rosettes of TW compared with normal wood. PttCesA1, PttCesA3-2 and PttCesA3-1 were among the most highly expressed CAZyme genes in both TW and normal wood, as determined by the intensity of the hybridization signal and EST frequency (Table 3), and therefore even a small modulation of their transcription may have meaningful consequences for cellular metabolism.

Table 3.  Differentially regulated and most highly expressed genes in developing TW coding for carbohydrate-active enzymes
CAZY famIDDescriptionPU IDMean ratio TW/CSign levelMean signal TWMean signal CTension woodCambial regionWood cell deathother librariesAGI IDScore
  1. Genes significant for wood formation were selected according to signal strength and their relative abundance in wood-related EST libraries. Stars indicate sgnlficantly differentially expressed genes (ANOVA) at P ≤ 0.05 - *, P ≤ 0.01 - **, or P ≤ 0.001. - ***, Number of EST clones is given for wood related and other libraries present in Populus DB.

CE879Pectinesterase, PttPME1PU080081.20*176715981102At3g 143101357
CE8 87 -PectinesterasePU118590.80*291237512109At3g492202154
EXPN11Expansin, putative (AtEXPA4), PttExpa5PU077431.18*700860495002At2g397001201
GH133Glycosyl hydrolasePU119480.62***341854760108At5g368901895
GH912Endo-1,4-beta-glucanase(KOR), PttCel9A1-2PU066240.82 9913125763002At5g497202807
GH913Endo-1,4-beta-glucanase(KOR), PttCel9A1-1PU022630.91 2510228365203810At5g497202835
GH917Endo-1,4-beta-glucanase, PttCel9BPU004311.31*10947895105At1g713801985
GH1316Alpha-amylase, PttAAMY2PU078983.53***220062010011At1g698303096
GH1634Xyloglucan endotransglycosylase, PttXTH16APU066810.88*95411011107At5g138701316
GH1635Xyloglucan endotransglycosylasee, PttXTH16CPU074481.25*4244336163039At4g032101174
GH1638Xyloglucan endotransglycosylase, PttXTH16TPU068730.71***90813043000At1g105501037
GH1915Chitinase-like protein 1 (AtCTL1)PU015071.11 5361478325012At1g058501257
GH1917Chitinase-like, PttGH19APU015170.93 3818641679181120At3g169201393
GH2886Polygalacturonase, PttGH28APU01 19041.38***706949832102At1g191702061
GH3522Beta-galactosidasePU120901.01 11471119390001At5g638102919
GH3510Beta-galactosidase, PttGal35BPU067830.77**517567012024At3g137503432
GH513Alpha-L-arabinofuranosidase, PttGH51A-2PU 100640.72*124619270011At3g107402558
GH512Alpha-L-arabinofuranosidase, PttGH51A-1PU122630.68*112918480002At3g107402434
GT1244UDP-glucoronosyl/ UDP-glucosyl transferasePU045590.84**91211060002At4g010701446
GT1250UDP-glucoronosyl/ UDP-glucosyl transferasePU041300.84*101412530001At3g 165201032
GT1258UDP-glucoronosyl/ UDP-glucosyl transferase, PttGT1BPU105910.51**4912962000213At4g010701625
GT249Cellulose synthase, PttCesA3-1PU012521.32 223711782712190At4g187804426
GT250Cellulose synthase, PtCesA3-2, PtrCesA1PU075251.30*380843046912190At4g187804193
GT257Cellulose synthase (AtCesA1), PtCesA8-2; PtrCesA4PU055181.19*916578852001At4g324105095
GT251Cellulose synthase (AtCesA7), PtCesA9-1PU067571.08 671464622020At5g174204837
GT268Cellulose synthase (AtCesA2),PtCesA4-1; PttCesA4; PtrCesA7PU062491.07 578253B72111At4g393505017
GT260Cellulose synthase(AtCesA3), PtCesA5-1; PtrCesA5PU091261.04 969794460002At5g051705019
GT253Cellulose synthase (AtCesA4), PtCesA1 ; PatCel1, PttCesA1; PtrCesA3PU022170.86*20390237617861At5g440304491
GT229Mannan synthase (AtCSLA9), PtGT2A-1PU029900.23***2593118482100At5g037602285
GT426Sucrose synthase, PttSUS1PU072211.57***26166172244813917At3g431903643
GT427Sucrose synthase, PttSUS2PU004011.39***446293340237712At3g43l903678
GT829Alpha-glycosyltransferase, PttGT8F-1PU104850.67**6340101501000At1g193001466
GT826Alpha-glycosyltransferase, PttGT8EPU044580.67***176627210001At1g193001202
GT84Alpha-glycosyltransrerase, PttGT8F-2PU062020.65***8266128901000At1g193001319
GT818Alpha-glycosyltransferase, PttGT8B-2PU063940.62***7879127962200At3g186602243
GT852Alpha-glycosyltransferase, PttGT8DPU028890.62***6691108173222At5g546902363
GT861Alpha-glycosyltransferase, PttGT8CPU020000.60***416071920100At3g186602391
GT1411Beta-glycosyltransferasePU064071.40*344724741000At5g 150501773
GT143Beta-glycosyltransferase, PttGT14APU000711.34*178113190101At3g036901363
GT146Beta-glycosyltransferase, PttGTi14BPU029441.20 675558120101At1g71070858
GT3132Galactosyltransferase, PttGT31BPU064191.57***10156331103Atlg329301645
QT3129Galactosyltransrerase, PttGT31BPU006051.39***12158761103At1g329301644
GT414UDP-N-acetylglucosamine:peptide N- acetylglucosaminyltransferasePU005741.12*367933380100At3g042404302
GT431Beta-glucuronytfransferase, PttGT43A-2,PU027401.00 11867116451601At2g37090681
GT4760Bela-glycosyltransferase, PttGT47BPU002252.34***262711170111At3g454001545
GT4773Beta-glucuronyltransferase, PttGT47A-1PU006431.25*12101990914444At1g274402036
GT472Beta-glucosyrtransferasa, PttGT47D-3PU080131.15*500743201003At5g61840888
GT4752Beta-glucosyltransferase, PttGT47CPUQ15290.75**575578762312At2g281101595
PL119Pectin/pectate lyasePU006591.65***13798421400At4g247801776
PL117Pectin/pectate lyasePU010591.61***173210802602At4g247801850
PL118Pectitn/pectate lyasePU030601.42***1999140533011At4g247801863
PL126Pectin/pectate lyasePU007071.34**392329593106At4g137101854
PL127Pectin/pectate lyasePU015421.23*450236035500At4g137101825
PL44Rhamnogalacturonan lyase (AtMYST6)PU081056.95***34585074110At2g226202103
X6937Pectin methylesterase inhibitorPU071191.23***251020751012At1g14890555

Transcripts of two genes putatively involved in cellulose biosynthesis, PttCel9A1-1 and PttGH19A, were not increased in TW. Both were strongly expressed in wood-forming tissues according to signal intensity on the microarray and EST frequency (Figure 4a, Table 3), and specifically in the secondary wall-forming tissues (Aspeborg et al., 2005). PttCel9A1-1 is homologous to the endoglucanase KORRIGAN1 (KOR1) gene of Arabidopsis thought to be involved in cellulose biosynthesis, although its exact function and interaction with CesA proteins is a matter of debate (Doblin et al., 2002). The PttCel9A1-1 expression product obtained in Pichia pastoris was shown to digest amorphous cellulose but not xyloglucan (Master et al., 2004), consistent with a role in cellulose biosynthesis. PttGH19A has sequence similarity to chitinases from the GH19 family. However, the encoded protein lacks the chitinase catalytic machinery and is likely to have acquired another, but as yet unknown function (Aspeborg et al., 2005). It is similar to Arabidopsis chitinase-like gene AtCTL1 whose mutant phenotype shows a severe cellulose deficiency (Mouille et al., 2003; Zhong et al., 2002). It is also similar to GhCTL1/GhCTL2, which has been demonstrated to be a catalytically unique chitinase-like protein that is preferentially expressed during cotton secondary wall biosynthesis (with a content of 95% cellulose) (Zhang et al., 2004). However, its precise involvement in the cellulose biosynthetic process is still unknown.

CesA proteins in the cellulose biosynthetic rosettes use cytosolic uridine diphosphate (UDP)-glucose as substrate, which is provided directly by particulate sucrose synthase (P-SuSy) (Haigler et al., 2001). This enzyme produces UDP-glucose and fructose from sucrose and UDP. Transcripts of several SuSy genes were increased in TW. Of these, PttSUS1 and PttSUS2 seem to be the most expressed, based on the EST frequencies and signal intensity levels (Figure 4a, Table 3). The increase of SuSy transcripts supports the idea of increased C flux to cellulose in TW. Consistent with the increased sucrose utilization, the metabolome analysis detected a decreased amount of sucrose in TW compared with control trees (Table 2). In addition to sucrose, starch might constitute an alternative temporary C source in TW, as indicated by an increase in α-amylase transcripts (Figure 4a, Table 3). The resulting glucose would be phosphorylated in the plastids by a hexokinase (Neuhaus and Emes, 2000) and transported out to the cytosol by a glucose-6-phosphate translocator, the transcript of which was also increased in TW (Table 1b). The cytosolic pool of glucose-6-phosphate can feed subsequent pathways to glycolysis, the pentose phosphate pathway, and sucrose and starch metabolism, as discussed below (Figure 4a). Taken together, our data suggest that more C is allocated to cellulose during TW formation compared with normal wood formation, but transcripts coding for proteins involved in the cellulose biosynthesis machinery were not generally increased. This implies that the abundance of these transcripts is not limiting for cellulose biosynthesis, or that the activity of cellulose biosynthesis is regulated at another level.

Alternative fluxes of C between cellulose, lignin and AGPs

In addition to providing UDP-glucose for cellulose biosynthesis, another consequence of high SuSy activity is the accumulation of fructose in the cytoplasm, which requires activation of fructose metabolism. In TW, this fructose appears to be phosphorylated by fructokinase as the corresponding genes were among the most highly expressed, and their transcript abundance was significantly increased, in TW (Figure 4a, Table 1a). This is consistent with the model of C partitioning during cellulose biosynthesis in cotton proposed by Delmer and Haigler (2002). The fructokinase product fructose-6-phosphate could be used for several purposes: (i) energy production in glycolysis, (ii) metabolism through the pentose phosphate pathway, (iii) GDP-mannose biosynthesis, (iv) N-acetyl-glucosamine biosynthesis or (v) conversion back to sucrose by sucrose phosphate synthase (SPS) and glucose-6-phosphate isomerase. Our transcriptome analysis suggested that, during TW formation, excess fructose-6-phosphate is diverted to N-acetyl glucosamine biosynthesis and sucrose regeneration, while during normal wood formation more fructose-6-phosphate is used for lignin biosynthesis, via the pentose phosphate pathway and the shikimic biosynthetic pathway, and for hemicellulose biosynthesis. Indeed, transcripts encoding plastidic pentose phosphate pathway enzymes such as transaldolase and transketolase, and shikimate pathway enzymes leading to lignin biosynthesis were coordinately decreased in TW, while those encoding UDP-N-acetylglucosamine pyrophosphorylase, a major enzyme in the N-acetyl glucosamine pathway, were increased (Figure 4b). N-acetyl-glucosamine could be used for glycosylation of AGPs (van Hengel et al., 2001) and transcripts encoding one UDP-N-acetylglucosamine:peptide N-acetylglucosaminyltransferase, GT41_4, were slightly increased in TW (Figure 4b, Table 3). Overall, the changes described above reflect decreased lignin biosynthesis and increased cellulose and AGP biosynthesis in TW. We propose that fructose-6-phosphate represents a connection between cellulose and lignin biosynthetic pathways, and changes in C flux through fructose-6-phosphate could explain how reduction of lignin formation could promote cellulose biosynthesis, as observed in transgenic poplar (Hu et al., 1999).

Sucrose salvage pathway might compromise pectin and hemicellulose biosynthesis

The sucrose salvage pathway, which uses fructose-6-phosphate through the action of SPS, requires UDP-glucose formed by glucose-6-phosphate isomerase, phosphoglucomutase and UDP-glucose pyrophosphorylase (UGPase) to make sucrose phosphate, and then sucrose is made by sucrose phosphate phosphatase (SPP). In TW, only UGPase transcripts were significantly increased while transcripts encoding the remaining enzymes were not affected (Figure 4a, Table S1). As the coordinate action of these enzymes is necessary for the formation of sucrose, it is possible that their regulation is mostly exerted at the activity level and not at the transcript level (Winter and Huber, 2000). In this scheme, SPS would compete for UDP-glucose (formed by UGPase) with enzymes involved in nucleotide sugar interconversion for matrix biosynthesis. If UDP-glucose formed from sucrose via P-SuSy was the sole source of C for cellulose biosynthesis, as proposed by Haigler et al. (2001), then SPS would be a key enzyme regulating the flux of C between cellulose and matrix components.

The nucleotide–sugar interconversion gene family has been recently characterized in Arabidopsis (Seifert, 2004), and we identified poplar gene models with homologous sequences. The initial C flux into various interconverting reactions occurs through UDP-glucose. In TW, transcripts encoding several enzymes responsible for these reactions were decreased, indicating decreased C flux to arabinan, xylan, pectin and xyloglucan (Figure 4c, Table 1a, Table S1). Similarly, the flux through GDP-mannose is likely to be reduced in TW by a marked decrease in the transcripts encoding three GDP-mannose-pyrophosphorylase proteins (Figure 4d, Table 1a). This in turn would have a further impact on the formation of carbohydrate polymers containing mannose and fucose, such as mannans, xyloglucan and rhamnogalacturonan II.

Not only were the transcripts encoding nucleotide–sugar conversion enzymes decreased, but so were several of those encoding Golgi glycosyl transferases (GTs), which are probably involved in the biosynthesis of hemicelluloses, such as xylans and mannans (Table 3). The precise identity of these GTs is in most cases still unknown, but the decrease in their transcript abundance during TW formation further supports their function in the biosynthesis of hemicelluloses, which are decreased in TW (Timell, 1969). GTs from the CAZy family GT8 make alpha glycosidic bonds. Several GT8-encoding transcripts, including PttGT8B-2 (reported as PttGT8B by Aspeborg et al., 2005), PttGT8C, PttGT8D, PttGT8E and PttGT9F, were decreased in TW (Table 3). PttGT8B-2 and PttGT8C display distant similarity to animal glycogenins and might act as priming enzymes in polysaccharide biosynthesis (Aspeborg et al., 2005). PttGT8D is a member in the same subgroup as Arabidopsis QUASIMODO1 (QUA1), whose mutant phenotype shows homogalacturonan deficiency (Bouton et al., 2002), indicating a likely involvement in pectin biosynthesis. GTs from the CAZy family GT2 are mostly enzymes involved in the biosynthesis of the backbone of cell wall polymers. Transcripts encoding PttGT2A-1 (known as PttGT2A), which is possibly involved in glucomannan backbone formation (Aspeborg et al., 2005), and another enzyme of unknown function, GT2_38, were decreased in TW. Family GT47 GTs include ‘inverting’ enzymes, i.e. transferases that make glycosidic bonds of stereochemistry opposite to that of the sugar nucleotide. PttGT47A, PttGT47B and PttGT47D-3 (formerly known as PttGT47D; Aspeborg et al., 2005) transcripts were increased in TW. These enzymes might be involved in the biosynthesis of arabinogalactan antennae of AGPs (Showalter, 2001). PttGT47A could be specifically involved in glucuronic acid transfer to the antennae as it resembles a β-glucuronyl transferase NpGUT1 (Iwai et al., 2003). In contrast, PttGT47C transcripts were decreased in TW, suggesting a role in hemicellulose biosynthesis, but the substrate specificity of this GT47 subfamily is yet to be determined. Finally, the GT31 family includes animal β(1,3)-N-acetylglucosaminyltransferases, β(1,3)-galactosyltransferases and β(1,3)-N-acetylgalactosaminyltransferases, but no plant member has yet been characterized, making the specificity of the latter hard to predict. Transcripts encoding three enzymes of this family were increased in TW (GT31_4, GT31_29, and GT31_32) and these could be involved in AGP biosynthesis.

An unexpected result from the microarray analysis was the increase of transcripts encoding cell wall degrading and modifying enzymes in TW (Figure 4c, Table 3). Some of these genes were also among the most abundantly represented in the EST library of TW (Figure 1). Pectin degradation was indicated by an increase of transcripts encoding pectin methyl esterase (family CE8), polygalacturonase (family GH28), and several pectate lyases of the PL1 family. In addition, transcripts encoding a member of polysaccharide lyase family 4, which includes fungal rhamnogalacturonan I lyases, similar to the Arabidopsis AtMYST6 gene, were highly increased in TW. Finally, the decrease of two family GH3 β-xylosidase transcripts indicates down-regulation of xylan metabolism.

Transcripts of genes involved in lignin biosynthesis are coordinately decreased in TW formation

In TW of Populus spp., the lignin content is reduced by about 20% of cell wall dry weight compared with normal wood (Timell, 1969). Carbohydrates are shuttled into monolignol biosynthesis via the shikimate pathway, both as shikimate and as phenylalanine (Boerjan et al., 2003). Indeed, consistent down-regulation of the shikimate pathway was observed in TW-forming tissues (Figure 5). The phenylpropanoid and monolignol biosynthesis pathways have been the subject of intensive research, and several of the poplar genes in this pathway have been cloned and demonstrated to occur in different isoforms, some being wood specific (Boerjan et al., 2003; Hu et al., 1998; Kao et al., 2002). To add to the picture of the major isoforms of early phenylpropanoid and monolignol biosynthesis genes expressed during wood formation, we identified the genes with a high signal on the microarray combined with the occurrence of EST sequences in the wood-forming libraries. Furthermore, a decrease of transcript abundance during TW formation would suggest a function for the corresponding gene in lignification (Figure 5, Table S3). The emerging picture is that transcripts of only some of the early phenylpropanoid pathway genes were decreased, consistent with the multifunctional purpose of this pathway, whereas transcripts of genes in the monolignol-specific branch downstream of caffeoyl-CoA were consistently decreased. Interestingly, two wood-expressed genes encoding phenylalanine-lyase (PAL) were identified, but only PttPAL1 transcripts were decreased during TW formation. Similarly, three caffeoyl-CoA O-methyltransferase (CCoAOMT) genes were found. PttCCoAOMT1 and 2 were homologs of two genes previously described in poplar (Chen et al., 2000) and their transcripts were decreased in TW, whereas PttCCoAOMT3 was not affected. Possibly genes not affected are only expressed during early S2 formation which also takes place in TW. At least two PAL genes have different expression profiles across the wood-forming tissues (Schrader et al., 2004), with PttPAL2 peaking in expression at an earlier stage than PttPAL1. Promoter-GUS studies of the PttCCoAOMT1 and 2 homologs in poplar showed expression in vessels and neighboring parenchyma cells (Chen et al., 2000), which would explain their down-regulation in TW as vessel number is decreased. However, it is surprising that these major CCoAOMT genes are not expressed in fibers were most lignin is deposited. However, bending the stem (90°) induced the genes in fibers at the point of bending (Chen et al., 2000), but this is most likely related to wound-induced lignin and not to TW induction. For the other wood-specific genes in the early phenylpropanoid pathway, three representatives of cinnamate 4-hydroxylase (C4H) genes were present but not affected in TW, whereas one representative of 4-coumarate:coenzyme A ligase (4CL) genes, two of hydroxycinnamoyl transferase (HCT) genes and one of p-coumarate 3-hydroxylase (C3H) genes showed decreased transcript abundance. Although HCT and C3H have not been cloned in Populus, these genes have high similarity (over 80% identical) to genes described in Nicotiana tabacum (L.) (Hoffmann et al., 2003) and Arabidopsis (Schoch et al., 2001) that were shown to function in lignification.

Figure 5.

Regulation of genes leading to lignin biosynthesis during tension wood (TW) formation. The figure illustrates the shikimate, phenylpropanoid and monlignol biosynthesis pathways according to Herrmann and Weaver (1999) and Boerjan et al. (2003). A decrease in transcript abundance is accompanied by asterisks indicating the significance level (analysis of variance): *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. For the shikimate pathway the significance level for the most decreased representative of each enzymatic function is given. For the phenylpropanoid and monolignol pathways genes significant for wood formation were identified according to signal strength and their relative abundance in wood-related ESTlibraries (Table S3). The following genes were included: DAPH synthase (PU02218); DHQ dehydratase (PU11655); shikimate synthase (PU01648); EPSP synthase (PU08322); chorismate synthase (PU00151); prephenate dehydratase (PU07029); phenylalanine lyase, PAL1 (PU02719) and PAL2 (PU07826); cinnamate 4-hydroxylase, C4H1 (PU01570) and C4H2 (PU06440); 4-coumarate:coenzyme A ligase, 4CL (PU01475); hydroxycinnamoyl transferase, HCT1 (PU07343) and HCT2 (PU02647); p-coumarate 3-hydroxylase, C3H (PU05862); caffeoyl-CoA O-methyltransferase, CCoAOMT1 (PU01811), CCoAOMT2 (PU02461) and CCoAOMT3 (PU02140); cinnamoyl-CoA reductase, CCR1 (PU00591); ferulate 5-hydroxylase, F5H (PU01148); caffeic acid O-methyltransferase, COMT1 (PU02638) and COMT2 (PU01944); cinnamyl alcohol dehydrogenase, CAD (PU0003). Shikimate was significantly decreased in TW (Table 2). E4P, erythrose 4-phosphate; PEP, phosphoenolpyruvate; DAHP, 3-deoxy-D-arabino-heptulosonate 7-phosphate; DHQ, 3-dehydroquinate; DHS, 3-dehydroskikimate; S3P, shikimate 3-phosphate; EPSP, enolpyruvylshikimate 3-phosphate; SAD, sinapyl alcohol. DHQ synthase and SAD were not spotted on the microarray.

Transcripts in the monolignol-specific pathway encoding cinnamoyl-CoA reductase (CCR), ferulate 5-hydroxylase (F5H) and caffeic acid O-methyltransferase (COMT) (one, one and two genes, respectively) were all decreased. Several cinnamyl alcohol dehydrogenase (CAD) genes were identified, but only one was highly expressed in wood-forming tissues and its transcript was weakly decreased in TW. The sinapyl alcohol dehydrogenase (SAD) suggested to be responsible for the final step in sinapyl alcohol biosynthesis (Li et al., 2001) was not represented on the microarray. This means that no SAD homolog was sequenced in the wood-forming libraries, suggesting low expression during wood formation.

After their biosynthesis, monolignols are transported from the cytoplasm to the cell wall, and polymerized to a lignin matrix. The molecular mechanisms and the proteins responsible for transport and polymerization are not known. However, consistent with the idea of coordinated transcriptional regulation of lignin biosynthesis (Rogers and Campbell, 2004), it seems likely that these are down-regulated in TW. It has been hypothesized that the monolignols are glycosylated by GTs to facilitate transport across the membrane, and then released by glycoside hydrolases (GHs) before polymerization in the cell wall (Boerjan et al., 2003). Indeed, GTs with activity against monolignols, and their corresponding aldehydes, have been identified in Arabidopsis (Lim et al., 2001, 2005). We found several genes within the GT1 and GH1 families active on secondary metabolites, with decreased transcript abundance in TW (Table 3). However, none of the GT genes was highly similar to the identified Arabidopsis genes.

In the cell wall, the monolignols are oxidized to their radicals and polymerized. Laccases, peroxidases and other phenol oxidases have long been thought to be involved in this polymerization (Baucher et al., 2003), but conclusive evidence for their role is still lacking. In Populus spp., several laccases (Ranocha et al., 1999) and peroxidases (Christensen et al., 2001) have been cloned and characterized. We found at least eight different laccases with a high signal on the microarray, and among these were the lac1, lac3, lac90 and lac110 previously identified in P. trichocarpa (Table S3) (Ranocha et al., 1999). Interestingly, the only laccase with decreased transcript abundance in TW was lac 3, which also was the only laccase resulting in a xylem phenotype among antisense lac3, lac 90 and lac110 trees (Ranocha et al., 2002). Although the lignin content was not affected in antisense lac3 trees, the amount of soluble phenolics and the structure of the secondary wall were altered. The finding that lac3 was co-regulated with lignin biosynthesis genes in TW supports its role in lignin polymerization. Among the peroxidases, some genes had increased transcript levels, whereas none was significantly decreased in TW (Table 1). Christensen et al. (2001) cloned the major anionic peroxidase genes responsible for syringylaldazine activity in developing xylem as key candidates for involvement in lignification. However, no ESTs for these genes were sequenced from the wood-related libraries, confirming their low expression in developing wood, a finding also obtained by Christensen et al. (2001).

A hypothesis for stereospecific coupling of monolignols by the dirigent protein was suggested after its discovery in Forsythia×intermedia Zabel and the demonstration that it mediated coupling of lignans (Burlat et al., 2001). The closest Populus homolog to the Forsythia protein was represented by several ESTs in wood-forming tissues (Table S3), but its transcript level was not affected during TW formation. However, transcripts of phenylcoumaran benzylic ether reductase, another enzyme involved in lignan biosynthesis which is highly abundant in maturing xylem, were decreased in TW (Table S3) (Mijnsbrugge et al., 2000). This raises the question of whether lignan biosynthesis is coordinately down-regulated with lignin biosynthesis during TW formation.

Signals and transcription factors

Perception of the gravitational stimulus and induction of the G-layer in the developing TW fibers must be transmitted by signals, such as plant hormones, which are likely to trigger transcriptional regulators redirecting the C flux from lignin and hemicellulose to cellulose. Developing fibers will respond to the gravitational signal up to a rather late stage of secondary wall formation (Timell, 1986). Therefore, factors required for G-layer formation should be present in the tissues investigated in this study. Indeed, a rather large set of transcription factors were differentially regulated in developing TW (Table 1a). Both MYB and LIM transcription factors have been demonstrated to coordinately regulate the biosynthesis of monolignols by binding to AC elements in the promoter (Rogers and Campbell, 2004). A MYB transcription factor (PttMYB21a) (Table 1a) previously demonstrated to be up-regulated in TW and to repress lignin (Karpinska et al., 2004) also showed increased transcript levels in this study. We also found three LIM transcription factors similar to NtLIM1 of N. tabacum, which positively stimulated monolignol biosynthesis in transgenic plants (Kawaoka et al., 2000). However, the transcript abundance of all three LIM genes was increased in TW, and their products are therefore not likely to have a similar function to NtLIM1 (Table 1a, Table 3).

Among the genes related to hormonal signaling that were affected in TW, auxin and ethylene signaling pathways clearly dominated (Table 4). Auxin has long been implicated in the TW response, and application of exogenous auxin or auxin transport inhibitors can in fact induce G-fibers (Mellerowicz et al., 2001). However, measurements of endogenous auxin did not reveal any obvious changes of the auxin balance in the cambial region tissues when TW was formed (Hellgren et al., 2004). Nevertheless, a rather large set of auxin-related genes were differentially expressed in developing TW. Two genes previously demonstrated to be strongly up-regulated during secondary wall formation (Schrader et al., 2003) showed a significant decrease in transcript abundance during TW formation; the putative auxin influx protein PttLAX1 and the auxin-responsive protein PttIAA5. In addition, transcript levels for several auxin response factors were also decreased, whereas the transcript level for one auxin-responsive (aux/IAA) gene was increased. Several aux/IAA genes earlier suggested to respond to gravitational stimuli (Moyle et al., 2002) were weakly expressed during secondary wall formation and could therefore not be confirmed in our study. Possibly PttLAX1 plays a role in modifying auxin levels in wood fibers at a late stage of development. The levels of auxin in wood fibers forming secondary walls are low, but may still be important, and small modifications in these tissues may have escaped detection in the study by Hellgren et al. (2004). Alternatively, the auxin response pathway may be triggered by other signals. Ethylene is well known to interact with auxin signaling, and is indeed induced during TW formation. We have previously demonstrated the strong induction of an ACC oxidase (ACO1) in developing TW (Andersson Gunnerås et al., 2003), and the microarray data confirm this observation and also show an increase in transcripts encoding ethylene receptors and signaling proteins (Table 4). However, a role for ethylene in G-fiber formation was not supported in work in our laboratory using ethylene-insensitive birch and aspen trees (J. Love, S. Björklund, H. Tuominen and B. Sundberg, unpublished results).

Table 4.  Differentially regulated genes related to auxin and ethyiene
PU IDDescriptionMean ratio TW/CSign levelMean signal level TWMean signal CTension woodCambial regionWood cell deathOther libraryAGI IDScore
  1. Genes significant for wood formation were selected according to signal strength and their relative abundance in wood related EST libraries. Stars indicate significantly differentially expressed genes (ANOVA) at P ≤ 0.05-*, P ≤ 0.01 -**, or P ≤ 0.001. -***, Number of EST dories is given (or wood related and other libraries present in Populus DB.

Ethylene-related genes
PU01604ACC oxidase, PttACO117.06***137697808300At1g773301173
PU06991EIN3-binding F-box protein, EBF1-like2.50***132454220927At2g254901638
PU08374Ethylene-insensitive 3, EIN31.82***277515350004At3g207701633
PU07354Ethylene response sensor, ERS1.75***9375411110At2g409402359
PU09099Ethylene receptor, ETR21.58***14339510102At3g231502366
PU01574Serine/threonine protein kinase, CTR11.19**6044990101At5g037301895
Auxin-related genes
PU03374Auxin-responsive protein, PttlAA111.44*507837660002At2g22670749
PU07504Auxin-responsive protein-related1.16*833272111000At3g617501086
PU02667Auxin-responsive protein-related0.90*4174740100At3g62100413
PU06171Auxin-responsive factor AUX/1AA-related0.89**8029012001At1g192201965
PU01813Auxin-responsive protein, PttIAA80.87**158418371201At5g43700559
PU02111Auxin-responsive protein, IAA30.81*98712530202At1g04240600
PU12989Auxin-responsive factor, ARF60.76*110715030002At1g303303020
PU02764Auxin-responsive protein-related0.75*7189960100At1g19840520
PU07451Auxin-responsive protein, PttlAA50.74*670390771013At1g04100441
PU01758Auxin-responsive factor, ARF60.72***71810061100At1g303302215
PU10683Auxin-responsive factor, ARF10.64***108318800009At5g620002946
PU07219Auxin-responsive protein-related0.63***337156012001At3g25290734
PU11293Auxin-responsive protein, IAA160.63***89814930004At3g04730372
PU04390AUX1-like protein, PttLAX10.54***339261920325At2g381202216
PU06643Similar to auxin downregulated protein ARG100.37***740203710115At4g27450 .1101

Final comments

Global transcriptome and metabolome analysis obviously does not reveal the regulation of gene transcription, protein abundance or protein activity, and will therefore only tell us part of the story behind G-layer biosynthesis in TW. However, with the integration of transciptome and metabolite data into biochemical pathways, and considering the large modification in the biosynthesis of cell wall components that takes place during TW formation, our analysis should serve as a roadmap and a stimulus for further functional analysis of genes involved in making matrix polysaccharides, cellulose and lignin in secondary walls.

Experimental procedures

Plant material

For the EST library, hybrid aspen (Populus tremula L. × tremuloides Michx.) were raised in a glasshouse to a height of 2 m. TW was induced by leaning and supporting the tree at an angle of about 45° and tissues were sampled after 3 weeks of induction.

For the microarray, Northern blot analysis and metabolic profiling, field-grown aspen trees (P. tremula L.) 4.5 to 5.5 m in height and 3 cm in diameter at breast height were selected from a natural stand near Umeå, Sweden (63°50′N, 20°20′E). TW was induced during the most active period of growth by bending and fixing the trees with string, so that the midpoint of the stem was at an angle of about 45°. Upright trees were used as the control. Stem segments were collected at the midpoint of the stem after 11 days of bending, immediately frozen in liquid nitrogen and transported to the laboratory on dry ice to be stored at −70°C. Anatomical investigations confirmed that TW was formed in all bent trees, and that normal wood devoid of G-fibers was formed in the upright control trees. Developing TW was obtained from the upper (TW side) quarter of the circumference of the bent trees, and all around the circumference of upright trees. Tissues were collected by peeling the bark and scraping the exposed xylem side with a scalpel. The obtained tissues were thus enriched in developing xylem at the stage of secondary wall formation, while the cambial zone and the majority of expanding xylem cells were excluded (see Gray-Mitsumune et al., 2004). The experiment was performed twice, in two different years, to obtain independent biological replication.

Tension wood cDNA library and the POP1 microarray

Construction of the TW cDNA library and the POP1 microarray has been described previously (Andersson et al., 2004; Sterky et al., 2004). The POP1 microarray is based on a unigene set of 12 375 clones mainly from wood-related libraries. Annotation of the unigene is based on the best Arabidopsis hit from the best gene model. The identity of genes discussed in any detail was verified by blastsearches in the Arabidopsis database ( Functional classification was based on MIPS (Munich Institute for Protein Sequences) classification with modifications.

Experimental design, target preparation and hybridization for microarray analysis

We used a direct comparison design, where hybridizations were performed with RNA from TW and control trees from pooled samples of three trees. Four replicate hybridizations, including two dye-swaps, were performed. Taking into account the duplicates on the slides, eight data points for each clone were obtained. The experiment was performed twice, in two different years, giving 16 data points for each clone.

The tissue was homogenized in liquid nitrogen using a mortar and pestle, and total RNA was prepared using the method described by Chang et al. (1993). After precipitation with LiCl overnight, the RNeasy Plant Mini Kit (Qiagene, Hilden, Germany) was used for the wash and elution steps. RNA from each sample was reverse-transcribed into aminoallyl-labeled cDNA, using 30 μg of total RNA primed with 10 μg of oligo-dT primer (CyberGene, Huddinge, Sweden), and 300 U of Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA). The cDNA was purified using MiniElute spin-columns (Qiagen). The cDNA was eluted with 15 μl of 0.1 m NaHCO3 (pH 9.0) directly into dried pre-aliquoted Cy3- or Cy5-esters (Amersham Biosciences AB, Uppsala, Sweden). Excess dye was removed using the Qiaquick PCR purification kit (Qiagen). Then 26 μg of oligo(dA)80 (CyberGene), 10 μg of tRNA (Invitrogen) and hybridization mix [4 × saline sodium citrate (SSC), 0.3% sodium dodecyl sulfate (SDS) and 25% formamide] were added to the purified Cy3- and Cy5-labeled targets, to a final volume of 30 μl. The mixture was denatured for 2 min at 95°C, chilled on ice for 30 sec, and then put on the slide that had been pre-hybridized for 30 min at 42°C in a buffer containing 5 × SSC, 5 × Denhardt's solution, 100 μg ml−1 CT-DNA and 50% formamide. The slide was placed in a hybridization chamber and incubated in a water bath at 42°C for 18–20 h. After hybridization, the slide was washed for 5 min in 1 × SSC and 0.03% SDS, then in 0.2 × SSC and finally in 0.05 × SSC. All washing steps were carried out at room temperature. The slides were dried with N2 gas, prior to scanning.

Populus genome data sources

Gene models have been predicted by several prediction programs in parallel with the Populus Genome Project ( We have worked with the filtered model set released in September 2004 (denoted ‘version 1’) and the filtered set prepared for the genome annotation jamboree in December 2004 (denoted ‘version 2’). Genes of importance for the discussion were checked manually and in few cases a more appropriate gene model was selected.

Data treatment and statistical analysis

Slides were scanned using a ScanArray 4000 scanner (Perkin Elmer Life Sciences, Boston, MA, USA) at 5 μm. Laser settings and the photomultiplier tube were adjusted to obtain overall similar signal strengths in the two channels and to obtain 1–2% of saturated spots. Spot intensities were quantified using Genepix Pro 4.1 (CBS; Axon Instruments, Union City, CA, USA). Spots with high background or dust speckles were manually flagged as bad. The median background pixel intensity was subtracted from the mean spot pixel intensity. For the calculation of the background, the local method was chosen. All further data manipulations and computations were carried out with the SAS system for Windows V8 (SAS Inc., Cary, NC, USA). First, spots that were 10% or more saturated in both channels were deleted, as the ratio would be unreliable (Wang et al., 2001). ‘Treatment’ significance for each gene was estimated using the mixed model of analysis of variance (anova) developed by Wolfinger et al. (2001). Briefly, lowess transformed log2 data (yijk) were subjected to a normalization according to the model: yijk = μ + Ai + Pij + Pok + (A × Pi)ij + (Pi × Poi)jk + eijk, where μ is the sample mean, Ai is the effect of the microarray, Pij is the effect of the pin and Pok is the effect of the position. (A × Pi)ij is the interaction of the microarray and pin effects, (Pi × Po)jk is the interaction of the pin and position effects, and eijk is the stochastic error. The residual values from this model were then fitted into a gene-specific model: rijk = μ + Ai + Dl + Tj + Ym + eijk, where Dl is the effect of dye, Tj is the treatment effect (control and TW), and Ym is the effect of the year. Both models were implemented using proc mixed in SAS (SAS/STAT Software version 8; SAS Institute). For the gene model, type 3 F-values and P-values were calculated for the terms dye, treatment and year. A gene was considered to be differentially expressed when the treatment effect was significant at P ≤ 0.001. For selected groups of genes the significance level was lowered to P ≤ 0.01 or to P ≤ 0.05. Data were deposited at UPSC-BASE ( experiment number 005.

Northern blot analysis

Total RNA was prepared from the same tissues using the same method as above. A sample of 30 μg of total RNA was separated on a formaldehyde agarose gel according to Sambrook et al. (1989) and blotted onto a Hybond-N nylon filter (Amersham, Little Chalfont, UK). Specific probes for each cDNA clone were digested from pBluescript vectors using restriction enzymes Pst1 and Xho1. Radiolabeling was performed with [α-32P] dATP using a Strip-EZTM DNA (Ambion, Austin, TX, USA) probe synthesis kit, following the supplier's instructions. Unincorporated nucleotides were removed by Nick columns (Pharmacia Biotech, Uppsala, Sweden). Hybridization was performed overnight at 65°C in Church buffer (Church and Gilbert, 1984) and the stringency of final washes was 0.1 × SSC and 0.1% SDS, at 65°C. The radioactivity on the membrane was detected using a GS-525 Molecular Imager (BioRad, Solna, Sweden).

CAZyme annotation and functional classification

The CAZy (Carbohydrate Active enZYmes) database ( describes the families of structurally related catalytic and carbohydrate-binding modules (or functional domains) of enzymes that degrade, modify or create glycosidic bonds. Poplar gene models ( were searched with the tools used to update the CAZy database, resulting in the identification of about 1600 CAZyme genes (BH and PMC, unpublished results). All genes identified were subsequently manually verified and aligned with closely related CAZyme-encoding genes to refine their functional annotation.

The remaining genes were classified according to their putative function using a modified MIPS classification scheme. The following exclusive functional classes were considered: (i) cell wall, (ii) hormones, (iii) communication (MIPS class 10), (iv) transcription (MIPS class 04), (v) defense (MIPS class 11), (vi) cell cycle (MIPS class 03), (vii) transport (MIPS classes 08 and 67), (viii) Protein synthesis (MIPS class 05), (ix) protein fate (MIPS class 14), (x) carbohydrate metabolism other than CAZYmes (MIPS class 01.05), (xi) other metabolism (MIPS class 01), (xii) energy (MIPS class 02), and (xiii) no class (MIPS classes 98, 99 and no hit).

Metabolite analysis, extraction, derivatization and GC/MS analysis

Ten mg of sample and 1 ml of extraction medium (chloroform:MeOH:H2O, 2 : 6 : 2) were extracted, derivatized and analyzed by GC/TOF-MS according to Gullberg et al. (2004).

Analysis of GC/MS data

For comparison of GC/MS data, non-processed mass spectrometry files from GC/TOFMS analysis were exported in CSV format to matlab software 6.5 (Mathworks, Natick, MA, USA), where all data pretreatment procedures, such as baseline correction and chromatogram alignment, were performed using custom scripts allowing rapid comparison of metabolite profiles (Jonsson et al., 2004, 2005). Statistical analysis of differences between samples was carried out by partial least squares to latent structures discriminate analysis (PLS-DA). Determination of significant metabolites was performed by interpretation of the first weight vector (w1) from the PLS-DA model, as described by Trygg and Wold (2002), together with the 99% confidence intervals calculated using jack-knifing (Martens and Martens, 2000). The analysis was performed with simca-p + software (Umetrics AB, Umeå, Sweden).

Mass spectra deconvolution and quantification were performed by processing the GC/TOFMS analysis using chromatof (2.12) software (Leco Corp., St Joseph, MI, USA). Automatic peak detection and mass spectrum deconvolution were performed using a peak width set to 2.0 sec. To obtain accurate peak areas for use in quantification, unique quantification masses for each metabolite were specified and the samples were reprocessed.


This research was supported by Formas, the Swedish Research Council, the European project Eden QLK5-CT-2001-00443 and the Wood Ultrastructure Research Centre.