Poplar genes encoding fasciclin-like arabinogalactan proteins are highly expressed in tension wood

Authors

  • Florian Lafarguette,

    1. Equipe ‘Formation des Parois Lignifiées’, Unité Amélioration, Génétique et Physiologie Forestières, INRA Orléans, Avenue de la Pomme de Pin, BP 20 619 Ardon, F-45 166 Olivet Cedex, France;
    2. Present address: Forest Biology Research Center, Université Laval, Québec, QC, Canada, G1K 7P4;
    Search for more papers by this author
  • Jean-Charles Leplé,

    Corresponding author
    1. Equipe ‘Formation des Parois Lignifiées’, Unité Amélioration, Génétique et Physiologie Forestières, INRA Orléans, Avenue de la Pomme de Pin, BP 20 619 Ardon, F-45 166 Olivet Cedex, France;
    Search for more papers by this author
  • Annabelle Déjardin,

    1. Equipe ‘Formation des Parois Lignifiées’, Unité Amélioration, Génétique et Physiologie Forestières, INRA Orléans, Avenue de la Pomme de Pin, BP 20 619 Ardon, F-45 166 Olivet Cedex, France;
    Search for more papers by this author
  • Françoise Laurans,

    1. Equipe ‘Formation des Parois Lignifiées’, Unité Amélioration, Génétique et Physiologie Forestières, INRA Orléans, Avenue de la Pomme de Pin, BP 20 619 Ardon, F-45 166 Olivet Cedex, France;
    Search for more papers by this author
  • Guy Costa,

    1. Laboratoire de Chimie des Substances Naturelles, Groupe de Glycobiologie Forestière, 123, Avenue Albert Thomas, F-87 060 Limoges Cedex, France
    Search for more papers by this author
  • Marie-Claude Lesage-Descauses,

    1. Equipe ‘Formation des Parois Lignifiées’, Unité Amélioration, Génétique et Physiologie Forestières, INRA Orléans, Avenue de la Pomme de Pin, BP 20 619 Ardon, F-45 166 Olivet Cedex, France;
    Search for more papers by this author
  • Gilles Pilate

    1. Equipe ‘Formation des Parois Lignifiées’, Unité Amélioration, Génétique et Physiologie Forestières, INRA Orléans, Avenue de la Pomme de Pin, BP 20 619 Ardon, F-45 166 Olivet Cedex, France;
    Search for more papers by this author

Author for correspondence: Jean-Charles Leplé Tel: +33 2 38 41 78 37 Fax: +33 2 38 41 78 79 Email: Jean-Charles.Leple@orleans.inra.fr

Summary

  • • Fifteen poplar cDNA encoding fasciclin-like arabinogalactan proteins (PopFLAs) were finely characterized, whereas the presence of arabinogalactan proteins (AGPs) was globally assessed during wood formation.
  • • PopFLAs transcript accumulation was analysed through EST distribution in cDNA libraries, semi-quantitative RT-PCR, microarray experiment and Northern blot analysis. Similarly, AGPs contents were globally quantified by rocket electrophoresis. AGPs accumulation was further examined by Western blotting and immunocytolocalization.
  • • Ten PopFLAs were specifically expressed in tension wood (TW) and not expressed in the cambial zone. Rocket electrophoresis revealed important AGPs accumulation in TW xylem. An anti-AGPs specific antibody recognized two proteins preferentially present in the cell wall-bound fraction from TW. Immunocytochemistry revealed a strong labelling close to the inner part of the G-layer of TW fibres.
  • • PopFLAs are expressed in xylem and many are up-regulated in TW. It is suggested that some PopFLAs accumulating at the inner side of the G-layer may have a specific function in the building of this layer. PopFLAs expression may therefore be linked to the specific mechanical properties of TW.

Introduction

Trees are submitted to permanent environmental factors. Some of these stresses influence orientation and growth of trees by inducing deformations. In response to environment stimuli such as prevailing winds, snow, slope or asymmetric crown shape, leaning woody plants form an abnormal tissue, named reaction wood, which allows tree axis reorientation (Robards, 1969; Scurfield, 1973). In gymnosperm trees, the reaction wood is located on the lower side of branches and bent stem, is associated with increased compression strains and is therefore named compression wood. Compression wood is characterized by abnormal tracheids (Timell, 1986) with highly lignified cell walls enriched in p-hydroxyphenyl units and containing less cellulose than normal wood. In addition, the cellulose microfibril angle (with cell axis) is increased in the S2 sublayer of the secondary wall. By contrast, in angiosperm trees, reaction wood is named tension wood (TW), as it is associated with tensile strains and formed at the upper face of the inclined stems or branches (Timell, 1969). In poplar, as in numerous hardwoods, TW is mainly characterized by peculiar fibres, which are poorly lignified and have an additional thick layer in the secondary cell wall. This layer, called the gelatinous layer or G-layer, is mainly composed of crystalline cellulose microfibrils nearly parallel to the fiber axis (Norberg & Meier, 1966; Jourez, 1997; Jourez & Avella-Shaw, 2003). Tension wood is also characterized by a higher proportion of fibres, a lower proportion of vessels and a higher longitudinal shrinkage than normal wood (Jourez et al., 2001). Different models have been proposed in order to link these biochemical and anatomical characteristics to the mechanical properties of tension wood (Bamber, 2001). Okuyama et al. (1994) suggested a ‘unified hypothesis’ for emphasizing the importance of both the cellulose microfibril angle and lignification in tension strain generation. Supporting this model, several cellulose synthase genes have been shown to be up-regulated in aspen xylem during TW formation (Joshi, 2003). Likewise, some genes encoding enzymes from the lignin biosynthesis pathway appeared down-regulated during TW formation (G. Pilate et al., unpublished results).

Several phytohormones are involved in the regulation of tension wood formation. Recently, Moyle et al. (2002) reported that the expression of members of the Aux/IAA gene family was regulated during TW formation in poplar. However, alteration in the balance of endogenous auxin level is not sufficient to account for TW formation (Hellgren et al., 2004). Ethylene is another important developmental regulator as its production occurs asymmetrically within poplar stem upon bending. Moreover, one ACC oxidase gene is strongly regulated in TW (Andersson-Gunnerås et al., 2003).

Analyses of expressed sequence tags (ESTs) derived from multiple poplar xylem cDNA libraries have lead us to identify a set of genes that are highly regulated in TW including mainly members of the arabinogalactan proteins (AGPs) family (Déjardin et al., 2004). This striking finding was confirmed by using a cDNA-AFLP approach (F. Lafarguette et al. unpublished results). AGPs are wide-spread proteins throughout the plant kingdom. They are mostly found in the cell-wall, plasma membrane, and intercellular spaces (Fincher et al., 1983).

AGPs are implicated in diverse developmental roles such as differentiation, cell-cell recognition, somatic embryogenesis and programmed cell death (Showalter, 2001) but their exact functions remain unclear. They belong to the large family of hydroxyproline-rich glycoproteins (HRGPs), which also includes extensins, proline/hydroxyproline-rich glycoproteins and some lectins (Showalter, 1993; Gaspar et al., 2001). The carbohydrate moiety of AGPs, mainly large polysaccharide chains of type II arabinogalactan and galactan, accounts for approximately 90% of their composition (Nothnagel, 1997). These carbohydrates are able to react with a synthetic phenylazoglycoside dye called Yariv's reagent.

Numerous AGPs showing sequence similarities were identified in several plant species (Gaspar et al., 2001; Schultz et al., 2002). Deduced AGP protein backbones are designated as either ‘classical’ or ‘nonclassical’ AGP according to the organization into domains and motifs. Using this classification, Gaspar et al. (2001) described the fasciclin-like AGPs (FLAs), a subclass of classical AGPs containing domain(s) similar to Drosophila melanogaster fasciclins cell adhesion molecules. A number of genes encoding FLA protein backbones have been identified in Arabidopsis (Schultz et al., 2002) and in loblolly pine (Loopstra & Sederoff, 1995; Loopstra et al., 2000; Zhang et al., 2000). Some FLA proteins are associated with wood formation (No & Loopstra, 2000; Whetten et al., 2001; Lorenz & Dean, 2002). In loblolly pine, two AGP coding genes, PtX3H6 and PtX14A9, are up-regulated in the radial expanding hypocotyls, probably upon hormone regulation (No & Loopstra, 2000; No et al., 2000).

In this paper, we report the characterization of 15 highly homologous poplar genes encoding FLAs that are preferentially expressed in the differentiating xylem. Complete coding sequences were obtained from cDNA clones identified in the EST database (Déjardin et al., 2004). Deduced protein sequences are predicted to encode a class of fasciclin-like AGPs (named PopFLA1–15; AY607753 to AY607767). Further, we compared transcript accumulation for these structurally-related FLAs in the wood of poplar stems induced to form tension wood. Finally, using an immunocytochemical approach, we investigated the precise localization of AGPs in poplar wood.

Materials and Methods

Plant material and sampling

Poplar micropropagated shoots (Populus tremula × P. alba, section Populus, clone INRA 717–1-B4) were transferred to the glasshouse, potted in compost (3 l) and individually supplied with water and fertilizers by a drip system. At the onset of cambial growth early the following spring, shoots were artificially tilted by inclining the pots at 45° from the vertical using a rigid stick. Two months later, stems were harvested and samples from the cambial zone (CZ), differentiating (DX) and mature xylem (MX) were collected both on the upper and lower sides of the stem, which correspond to tension wood (TW) and opposite wood (OW). Tension wood is easily recognizable as a bright whitish crescent; this makes easy the split between TW and OW samples. Cambial zone samples were lightly scraped from the frozen peeled bark. While differentiating xylem samples were lightly scraped from the debarked stem, mature xylem samples correspond to wood chips collected after removal of the differentiating xylem. Wood tissues were immediately frozen in liquid nitrogen and stored at −80°C before use.

RNA extraction and Northern blot analysis

The frozen wood material was ground to a fine powder in liquid nitrogen. Total RNA was extracted from 250 mg (f. wt) of wood using QIAGEN RNEasy kit® and according to manufacturer's instructions. RNA concentration and quality were determined at 260 and 280 nm. Ten micrograms of total RNA was separated by electrophoresis in a denaturing 1.2% (w/v) agarose gel containing 0.4 m formaldehyde. Equal loading per lane was assessed by ethidium bromide staining. The RNA was transferred to a Hybond-N +® (Amersham Biosciences) nylon membrane by capillary blotting in 50 mm NaOH. Three distinct probes corresponding to PopFLA5, PopFLA10 and PopFLA14 were PCR-amplified from three cDNA clones, PtaXM0020G6G0614 (CF229065, 5′-GAAAAGCCAACAAAGGCAGT-3′ and 5′-TGGAATCTCAAAACTCAACTCAA-3′), PtaXM0018C8C0806 (CF228867, 5′-CCGGAAGTTCCCGCTTACTA-3′ and 5′-TTCGCAAAATCTCTCACCAA-3′) and PtaXM0006E1E0109 (CF227926, 5′-CCAGAAAAAGCCTAGGAACCA-3′ and 5′-AATTGCATTGGGGAGGCTAT-3′), respectively. Fifty nanograms for each probe was labelled with [32P]-dCTP (50 µCi) by random priming (Prime-a-gene® kit, Promega, Madison, WI, USA). Hybridization and subsequent washes were performed at 65°C following standard procedures (Sambrook et al., 1989) before autoradiography.

cDNA microarray

Insert fragments from 1465 cDNA clones were PCR-amplified using TriplexA (5′-CTCGGGAAGCGCGCCATTGTGTTGG-3′) and TriplexB1 (5′-ATACGACTCACTATAGGGCGAATTGGC-3′) primers. The PCR products were purified by using 96-well MultiScreen PCR plates (Millipore, Billerica, MA, USA). DNA concentration was determined by measuring absorbance at 260 nm and each sample was ajusted to 100 ng µl−1. Ten pl of each amplified cDNA was spotted in triplicate on polylysine-coated slides using a GMS 417 Arrayer. Labelling was performed using the CyScribe First-Strand cDNA Labelling Kit (Amersham Biosciences, Orsay, France) following the manufacturer's instructions. The differentiating xylem was sampled on three different bent trees either in the tension wood (DX-TW) or in the opposite wood (DX-OW) zones. RNA was extracted from each of these samples. Further, pooled DX-TW and DX-OW RNA samples were labelled using Cy3 and Cy5 dyes, respectively. Twenty µg of total RNA was used for each labelling reaction. RNA was removed from labelling reactions by alkaline hydrolysis and cDNA was subsequently purified on CyScribe GFX column (Amersham Biosciences). One hybridization was performed in a Corning hybridization chamber. The fluorescence intensity was monitored by using the array acquisition and analysis software for Genepix 4000B (Axon Instrument Inc., Molecular Devices Corporation, Union City, CA, USA). For each cDNA, the median intensity value for each triplicate was determined using Genespring 6.1 (SiliconGenetics, Redwood City, CA, USA). Spots corresponding to 3X SSC buffer were used as negative controls for background subtraction. The background median value was subtracted from the raw values for each gene. A Lowess normalization was applied taking 20.0% of the data to calculate the Lowess fit at each point. This curve was used to adjust the control value for each measurement.

Semi-quantitative reverse transcription-polymerase chain reaction (RT-PCR) analysis

Two µg of total RNA isolated as described above (see ‘RNA extraction and Northern blot analysis’) was reverse transcribed using 500 ng of oligo(dT)12–18 and 200 units of SuperScript™ II RT in a total volume of 19 µl following the manufacturer's instructions (Invitrogen™, Life Technologies, Cergy Pontoise, France). After first-strand cDNA synthesis, RNA complementary to the cDNA was removed by incubating the reaction with 2 units (1 µl) of E. coli RNAse H at 37°C for 20 min. PCR amplification was performed with 1/60 of the first-strand cDNA as template, 1 unit of Taq DNA polymerase (Invitrogen™, Life technologies) and 12 pmole of gene specific primers (Table 1) in a total volume of 25 µl. The cycling conditions were 94°C for 5 min for one initial step followed by 94°C for 45 s, 55°C for 1.5 min and 72°C for 1 min for 20 cycles. The PCR was terminated with one extra step at 72°C for 8 min and held at 4°C. The amplification was done using a GeneAmp® PCR System 9700 (Applied Biosystems, Courtaboeuf, France). As negative control to check for genomic DNA contamination, 33 ng total RNA not subjected to reverse transcription were PCR amplified. Ten µl of each PCR product was electrophoresed on a 1% agarose gel that was scanned under UV light after ethidium bromide staining.

Table 1.  Oligonucleotides sequences used in the semi-quantitative RT-PCR analysis
TemplatePredicted PCR product (bp)Name of the forward and reverse primersOligonucleotide sequences (5′–3′)
PopFLA1923cn2199AGTTTCTAGCATTATATCTTAACAAC
cn2199BCAGAATTATTCAACACACAATCGGC
PopFLA2881cn2200AGAAACACTTCTAAATAAGATC
cn2200BACACACCAAGCTTTCAATCCAAC
PopFLA3921cn2201ATTCAAGCTTTATATCAGGGACTG
cn2201BCGAAATTACTCAACAGATTCAATACAG
PopFLA4837cn2203ATGTACCACTACCTTTGCTCAGAC
cn2203BCATATTCAATATATACAGCTCAGAG
PopFLA5988cn2205AGAATCATATCAAATACTTCCATGCAC
cn2205BATTATTCAACACACAATCGGCCGGCG
PopFLA6967cn2208AATCAAACACTTTATGATTGAG
cn2208BTTTCAACAGGACTCCAAAGTCAAC
PopFLA7284cn2193AACCAGCAAAAGCACCTGCTGCAG
cn2193BTAAGAACGCCAGATTTGATAATGGC
PopFLA8828cn2139ACACTCCTAGCTACTTCAAGATGAG
cn2139BCATGAATATTTTATTGCCCCAAACGC
PopFLA9969cn2237AGTAGCAATCCAATTAAGATTTGGAAC
cn2237BAGATCTTTATCATCAGCACCATTG
PopFLA10753cn2148ATGTACAGTACAGGCCACACCTC
cn2148BATATATAGGCCAGCCCCGAAGAG
PopFLA11440cn1999AAGCACCTTCAGGTCCAACCAAC
cn1999BGATCCACCTGATACACAACTAACTG
PopFLA12534cn2011ACAATTTCTCTCTGAATTCTTGCTC
cn2011BGAATCAAGTTGCACGGTCCCAG
PopFLA13588cn2035AACGATGAGAAAGCAACTCCTCTC
cn2035BTGAAGGTGCAGGTGCTGGTGCTGCC
PopFLA14181cn2053ATCGCCATCAATTTAACCACA
cn2053BCCATTGTTTGTGTTGCCAAG
PopFLA15299cn2056ATATCCAAAAACCACTAAAATGGTGC
cn2056BAGGGTGCCTGATTTGAGACTTG
18SrRNA23218SA232CTTCGGGATCGGAGTAATGA
18SB232GCGGAGTCCTAGAAGCAACA

Protein extraction

Frozen tissues (200 mg) were ground to a fine powder in liquid nitrogen. Proteins were extracted 1 min at 4°C according to a modified method of Chabannes et al. (2001) in 1 ml of 0.1 m Tris-HCl pH 7.5, 2% PEG 6000 (w/v), 2% PVPP (w/v), 0.2 m sodium ascorbate. The crude extract was clarified by centrifugation (13 000 g/10 min) at 4°C and the supernatant collected to give fraction A, which contained mostly cytosolic and plasma membrane-bound proteins. Fraction B, which contained cell wall-bound proteins, was then extracted from the remaining pellet with 1 m NaCl during one night at 4°C under vigorous shaking. After centrifugation (13 000 g/10 min) at 4°C, the supernatant containing cell wall-bound proteins was recovered. Proteins were quantified according to Bradford's method (Bradford, 1976), aliquoted and stored at −20°C before use.

AGP quantitation on rocket (1D) Yariv agarose gel electrophoresis

The amounts of AGPs in the protein fractions A and B were quantified by rocket electrophoresis according to Schindler et al. (1995). Ten µg of protein extracts was analyzed in 1% (w/v) agarose gel (1.5 mm thick) prepared in Macllvaine buffer containing 30 µg ml−1 (β-D-Glc)3 Yariv reagent (Biosupplies Australia Pty LTD, Parkville, Australia). The electrophoresis was performed in the same buffer for 6 h at 4°C (5 V cm−1). After completion of electrophoresis, the gel was fixed overnight in a solution of 40% ethanol, 10% acetic acid, rinsed with water, impregnated for 10 min in 5% glycerol and air dried. Gum arabic (Sigma-Aldrich, Lyon, France) was used as standard.

Immunoblot analysis

The rat monoclonal antibody JIM14 was a kind gift of K. Roberts (Department of Cell Biology, John Innes Centre, Norwich, UK). The antiserum was supplied as a hybridoma culture supernatant supplemented with 0.02% sodium azide. The protein fractions were prepared from differentiating xylem and mature xylem samples. Equal amounts (15 µg) of fractions A and B were analyzed by SDS-polyacrylamide gel electrophoresis using an 8% acrylamide Tris-Glycine gel as described by Towbin & Gordon (1984). Proteins were visualized after Ponceau S red staining and transferred onto nitrocellulose membrane (0.45 µm, Amersham Biosciences). Western blot analysis was performed mainly according to the instructions provided with the BM chemiluminescence Western Blotting Kit (Roche Applied Science, Meylan, France). Briefly, the membrane was coated for 1 h at 20°C in Tris-buffered saline pH 7.5 containing 0.1% (v/v) Tween 20 (TBS-T) and 1% (v/v) blocking reagent. After three washes in TBS-T, the membrane was incubated for 1 h at 20°C with a rat JIM14 antibody at 1 : 1000 dilution in TBS-T supplemented with 0.5% blocking reagent. After two washes in TBS-T, the membrane was incubated for 30 min with mouse antirat IgG conjugated to horseradish peroxidase at a 1 : 1000 dilution in TBS-T supplemented with 0.5% blocking reagent. The membrane was then washed four times for 15 min each with TBS-T. Detection was performed following the manufacturer's instructions using Lumi Film for chemiluminescence detection (Roche Applied Science).

Sample preparation for microscopy

Sections of about 5 mm were cut from tension and opposite wood samples and were fixed for 4 h with 2.5% paraformaldehyde and 0.1% glutaraldehyde in 0.1 m McIlvaine citrate-phosphate buffer, pH 7.0. After dehydration in a graded series of ethanol, samples were embedded in medium grade LR White resin (London Resin Company Ltd, UK).

Light microscopy

Semi-thin sections (1 µm thick) were collected with a diamond knife (Diatome, Biel, Switzerland) installed on an ultracut R (Leica, Rueil-Malmaison, France), placed on silanized slides (Dako Cytomation, Trappes, France) and fixed by heating at 55°C for 3 min. To prevent liquid evaporation, all subsequent incubations were conducted in a moist chamber. Sections were successively incubated for 1 h with a droplet of blocking solution containing 3% bovine serum albumin (BSA) and 0.01% Tween-20 in 10 mm Tris Buffered Saline (TBS) buffer, pH 7.5, and then twice for 5 min with droplets of dilution buffer comprising 0.3% BSA and 0.01% Tween-20 in 10 mm TBS, pH 7.5. The primary antibody reaction for AGP detection was performed with the rat JIM14 monoclonal antibody (Knox et al., 1991) in dilution buffer (1 : 200). The sections were then washed with TBS buffer before incubation with 15 nm gold-conjugated goat antirat serum (British Biocell International, Cardiff, UK) as secondary antibody, in dilution buffer (1 : 50). Control sections were obtained by omitting the primary antibody. Sections were subsequently poststained with a mix of methylene blue/Azur II (Richardson et al., 1960) and examined under Leica DMR light microscopy. Immunogold labelling was enhanced by the Silver Enhancing kit (British Biocell International, Cardiff, UK) according to the manufacturer's instructions.

Transmission electron microscopy

Electron microscopic observations were made at 80 KV using a Philips CM110 electron microscope (Philips, Eindhoven, the Netherlands). Eighty nanometre-thick sections were cut from the resin blocks and harvested on 300-mesh formvar-coated nickel grids. The grids were successively floated for 1 h in a blocking solution containing 3% bovine serum albumin (BSA) and 0.01% Tween-20 in 50 mm Tris Buffered Saline (TBS) buffer, pH 7.5, and then twice for 5 min in a dilution buffer containing 0.3% BSA and 0.01% Tween-20 in 50 mm TBS, pH 7.5. The primary antibody reaction for AGP detection was performed by floating the grids on drops of dilution buffer containing rat JIM14 monoclonal antibody (1/10) (Knox et al., 1991). The grids were then washed in TBS buffer and incubated with 15-nm gold-conjugated goat antirat serum in dilution buffer (1/50) (British Biocell International, Cardiff, UK) as secondary antibody. Control sections were obtained by omitting the primary antibody. Grids were subsequently floated for 25 min in 3% aqueous uranyl acetate solution for additional contrast. Immunogold labelling was enhanced by using the Silver Enhancing kit (British Biocell International, Cardiff, UK) according to the manufacturer's instructions.

Results

Fifteen fasciclin-like AGPs are differentially represented in poplar wood cDNA libraries

In a recent study, we analysed the distribution of 10 062 poplar EST derived from four nonnormalized cDNA libraries. As cDNA clones were randomly picked, this distribution is assumed to reflect the level of gene expression in the different wood zones. From analyses of sequence homologies, all the EST have been organized into clusters divided into contigs, themselves subdivided into consensus group. EST assembled in consensus group theoretically correspond to a gene (for details, see Déjardin et al., 2004). Interestingly, some of the largest EST clusters were similar to AGP genes. Moreover, five of these AGP clusters were over-represented in the differentiating and mature xylem (DX and MX, respectively) cDNA libraries when compared with the cambial zone (CZ) one. In addition, three of them were only represented in TW (Déjardin et al., 2004). Further computational analysis allowed identification of five additional smaller AGP clusters. For each of these 10 AGP clusters, one consensus group was selected for in depth sequence analyses, whereas for one cluster (cl1592) analyses were performed on six consensus groups. Altogether, the complete sequence of 15 cDNAs (referred to as PopFLA) was determined using the longest cDNA clone available for each consensus. The origin of each PopFLA is summarized in Table 2.

Table 2.  Fasciclin-like AGP identified from the assembly of poplar EST
PopFLA (accession #)Longest EST clone (accession #)Cluster IDConsensus ID selectedEST in DX-TWEST in DX-OWbDX-TW/DX-OWEST in MXbMX/DX
  • a

    EST clusters presented in Déjardin et al. (2004). No EST originating from the cambiale zone (CZ) cDNA libmrary were found in any of the consensus groups presented here.

  • Significance level of differences in ESTs distributions were calculated according to the Audic and Claverie's test (Audic & Claverie, 1997, http://igs-server.cnrs-mrs.fr/~audic/significance.html). bindicates a differential expression between the two conditions with a probability P, ***indicates significance at 5% level. Number of ESTs from DX-TW = 2810, DX-OW = 2562 and MX = 2359.

PopFLA1
(AY607753)
PtaXM0028A2A0202
(CF229653)
a15922199 5 00.96 > P > 0.95*** 50.9 > P > 0.8
PopFLA2
(AY607754)
PtaxmjxtD2D0208
(CF229834)
a15922200 4 00.96 > P > 0.95*** 60.96 > P > 0.95***
PopFLA3
(AY607755)
Ptajxtjxt2F4F0412
(CF237148)
a15922201 7 00.99 > P > 0.98*** 60.8 > P > 0.7
PopFLA4
(AY607756)
PtaJXT4C1C0105
(CF236556)
a1592220310 00.999 > P > 0.998*** 40.1 > P
PopFLA5
(AY607757)
PtaXM0027E2E0210
(CF229610)
a15922205 6 00.98 > P > 0.97***100.993 > P > 0.992***
PopFLA6
(AY607758)
PtaJXT0020C11C1105
(CF235220)
a1592220813 0P > 0.999*** 90.8 > P > 0.7
PopFLA7
(AY607759)
PtaJXT3F12F1212
(CF236504)
a1588219323 0P > 0.999***26P > 0.999***
PopFLA8
(AY607760)
PtaXM0020E10E1010
(CF229042)
a1549213923 1P > 0.999***220.99 > P > 0.98***
PopFLA9
(AY607761)
PtaXM0012H5H0515
(CF228432)
16202237 6 00.98 > P > 0.97*** 30.3 > P > 0.2
PopFLA10
(AY607762)
PtaJXT0016A8A0802
(CF234864)
a15552148 3 00.9 > P > 0.8 30.8 > P > 0.7
PopFLA11
(AY607763)
PtaJXT0015B7B0703
(CF234790)
14371999 6 60.2 > P > 0.1 00.99 > P > 0.98***
PopFLA12
(AY607764)
PtaJXT0026H11H1115
(CF235776)
14462011 1 10.1 > P 00.4 > P > 0.3
PopFLA13
(AY607765)
PtaJXO0022C7C0705
(CF233355)
a14652035 7100.7 > P > 0.6 00.998 > P > 0.997***
PopFLA14
(AY607766)
PtaJXO0024D6D0608
(CF233507)
14792053 4 40.2 > P > 0.1 10.8 > P > 0.7
PopFLA15
(AY607767)
PtaXM0001H4H0416
CF227583
14822056 6 70.4 > P > 0.3 10.94 > P > 0.93

For all the analyzed PopFLA, the deduced protein sequence possesses the features of fasciclin-like AGPs with a N-terminal secretion signal and a fasciclin-like domain flanked by two AGP-like regions rich in Ala, Pro, Ser and Thr (Fig. 1). PSI-Blast searches (Altschul et al., 1997) of the 15 PopFLAs sequences against nonredundant GenBank database revealed the existence of similar genes in poplar (Pop14A9; AF183809), Arabidopsis (FLA12; At5g60490), cotton (GhAGP1; AY218846) and loblolly pine (p14A9; U09556). The alignment of all of these FLA sequences revealed a conserved structural organisation with only some discrepancies in the length and composition of the AGP-like domains (Fig. 1). Indeed, PopFLA1–10, PopFLA12 and PtX14A9 display a longer proximal domain with predominantly amino acid residues like Ala, Val, Ser, Pro and Thr. Proline residues are most often arranged in small noncontiguous clusters of Pro-Ser, Pro-Ala, Pro-Thr. Such motifs were found as a sequence-based glycosylation code by Kieliszewski and coworkers (Shpak et al., 1999; Kieliszewski & Shpak, 2001; Kieliszewski, 2001; Tan et al., 2003). Noncontiguous Pro with flanking residues Ala, Val, Thr or Ser are good candidates for hydroxylation by proline hydroxylase and may serve as arabinogalactosylation sites. For all the analysed sequences, a similar ‘AGP-like’ domain is also present in the C terminal region. Interestingly, PopFLA2 differs from the other sequences due to its shorter distal AGP-like domain. Interpro database searches revealed the presence of a fascilin-like domain (IPR000782 interpro domain) between the two ‘AGP-like’ regions in all sequences, with the characteristic H1 and H2 conserved motifs, as determined by Kawamoto et al. (1998) (Fig. 1). Likewise, an N-terminal signal peptide for secretion is predicted in all the sequences. In addition, nine PopFLAs are predicted to have a complete C-terminal signal for the addition of the glycosylphosphatidylinositol (GPI) anchor (BigPI Plant prediction program; Eisenhaber et al., 2003). These sequences share a protein domain organization similar to the Arabidopsis FLAs group A composed of FLA 6, 7, 9, 11, 12 and 13 (Johnson et al., 2003). PopFLA 3, 4, 6, 8, 9 and 10 are not predicted to be GPI-anchored and may constitute a new subclass.

Figure 1.

Sequence alignment of 15 PopFLAs and four related sequences from poplar (Pop14A9; AF183809), Arabidopsis (FLA12; At5g60490), cotton (GhAGP1; AY218846) and loblolly pine (Ptx14A9; U09556). Pop14A9 and PopFLA15 are 93% identical at the nucleic level and correspond most likely to the same gene (not shown). The alignment was generated by ClustalW (http://www.ebi.ac.uk/clustalw/) and manually edited. Dashes represent gaps introduced to maximize the alignment. Signal peptide for excretion was predicted with TargetP and is boxed (http://www.cbs.dtu.dk/services/TargetP/). Noncontiguous Pro, potentially arabinogalactosylated (Kieliszewski, 2001) within the AGP-like domains (boxed) are shaded in light grey. A fasciclin-like domain was predicted using interproscan (http://www.ebi.ac.uk/interpro/scan.html; IPR00782 domain). The HI and H2 conserved regions characteristic of this domain (Kawamoto et al., 1998) are indicated above the alignment. Highly conserved amino acid residue within H1 and H2 are in bold. Amino acids thought to be involved in adhesion (Tyr-His or Phe-His, Ile, Leu and Val; Kim et al., 2000; Kim et al., 2002) are underlined. The position of GPI anchor signal sequence was predicted using Big-PI plant predictor (http://mendel.imp.univie.ac.at/gpi/plant_server.html; Eisenhaber et al., 2003). Predicted GPI-modification site is shaded in dark grey. N-glycosylation sites (Asp-X-Ser/Thr) are predicted using the program NetNGlyc taking only (++) predictions (and better) as positive result (http://www.cbs.dtu.dk/services/NetNGlyc/).

The PopFLA genes are differentially expressed in wood tissues

The representation of each PopFLA gene in DX-TW, DX-OW and MX cDNA libraries is reported in Table 2. The distribution of ESTs corresponding to PopFLA1–9 is significantly different between DX-TW and DX-OW (at the 95% confidence level). Moreover, PopFLA2, 5, 7 and 8 are more represented in MX than in DX. By contrast, PopFLA11 and PopFLA 13 are more represented in DX. No EST corresponding to PopFLA genes was detected in the CZ library. We complemented these data by further investigating the expression of these genes in the different wood tissues, using three experimental approaches: semi-quantitative RT-PCR, microarray and Northern blot analyses.

Semi-quantitative RT-PCR study was performed by using primers designed to enable gene specific amplification of each of the 15 PopFLA sequences (Table 1). Expression levels of those transcripts were assessed in CZ, DX and MX from TW and OW (Fig. 2). Using standard conditions, no PopFLA transcript amplification could be detected in the CZ. However, raising the number of PCR cycles to 30, faint signals were detected for PopFLA11 to PopFLA15 (data not shown). The expression of PopFLA1–10 was specific of tension wood xylem tissues, while PopFLA11–15 was present both in opposite and tension wood zones. Furthermore, the expression level of PopFLA11, 13, 14 and 15 appeared higher in differentiating xylem than in mature xylem. All these results were consistent with the ESTs’ distribution in the studied wood cDNA libraries (Table 2).

Figure 2.

Semi-quantitative polymerase chain reaction (PCR) on reverse transcribed cDNAs from different wood tissues. PopFLAs expression was assessed on the cambial (CZ), differentiating xylem (DX) and mature xylem (MX) tissues sampled in tension (TW) and opposite (OW) wood areas. The size of each expected PCR product is indicated on the right. Control amplifications have been done using a primer set amplifying a poplar 18S rRNA housekeeping gene.

Based on a microarray analysis, we compared the expression patterns in DX-TW and DX-OW of a set of 1465 genes, including the 15 PopFLAs, except PopFLA11 (Table 3). In agreement with RT-PCR results, PopFLA1–10 were significantly over-expressed in tension wood, whereas the others were not.

Table 3.  Intensity values of several PopFLAs obtained from microarray experiment
aEST spotted, Accession numberAnnotationbRatio DX-TW/DX-OWDX-OWDX-TWt-test P-value
  • a

    when available we selected an EST clone localized at the 3 prime end of the PopFLA cDNAs.

  • b

    ratio of median intensity signal from tension wood differentiating xylem (DX-TW) vs. opposite wood differentiating xylem (DX-OW).

  • ***

    indicates that difference in intensity between DX-TW and DX-OW is significant at 1% level.

PTAJXT0023G11G1113, CF235512PopFLA119.54*** 182.23560.70.00058
PTAJXT8A12A1202, CF236851PopFLA213.62*** 171.32332.74.15E-06
PTAXM0021A9A0901, CF229086PopFLA3 4.492*** 166.3 7470.00042
PTAXM0017B8B0804, CF228769PopFLA4 5.604*** 166.8 9352.00E-05
PTAXM0020G6G0614, CF229065PopFLA516.24*** 157.725615.22E-06
PTAXM0005D6D0608, CF227846PopFLA623.56*** 2285372.31.95E-05
PTAJXT0017B9B0903, CF234959PopFLA717.11*** 360.96174.75.95E-06
PTAJXT0013C11C1105, CF234631PopFLA819.61*** 307.26024.78.70E-07
PTAJXT2C9C0905, CF236399PopFLA9 3.388*** 264.7 896.70.00552
PTAXM0018C8C0806, CF228867PopFLA1011.44*** 321.436780.00396
PTAJXT0026H11H1115, CF235776PopFLA12 0.7341384.351016.30.63
PTAJXO0017H11H1115, CF232967PopFLA13 0.831594.8161323.70.718
PTAXM0006E1E0109, CF227926PopFLA14 0.8993494.893140.3330.666
PTAJXO1B6B0604, CF233995PopFLA15 0.8771797.0315760.823

In addition, the expression of three PopFLA genes was verified by Northern blot hybridisation. Thus, we confirmed the preferential accumulation of PopFLA5 and 10 transcripts in the tension wood differentiating xylem. PopFLA14 transcript levels did not seem to change upon TW formation (Fig. 3). It should be pointed out that two bands were detected with the PopFLA14 probe, probably resulting from the occurrence of a transcript very similar to PopFLA14.

Figure 3.

Relative amounts of mRNA corresponding to three classes of AGP in differentiating xylem from opposite wood (DX-OW) and tension wood (DX-TW).Ten micrograms of total RNA from DX-OW and DX-TW sampled on three individual bent poplar stems (1, 2, 3) were separated by electrophoresis, stained with ethidium bromide (c, e) and blotted onto a nylon membrane. One membrane was hybridized sequentially with probes specific to PopFLA10 (a) and PopFLA14 (b). The previous probe was stripped off the membrane before any new hybridization. The second membrane was hybridized with a probe specific for PopFLA5 (d). Results were revealed by autoradiography after 12–24 h.

Differentiating and mature xylem contain higher amounts of AGPs in tension wood zone

We further investigated whether an abundance of PopFLAs transcripts was related to an increase in AGPs accumulation. The presence of AGPs at the protein level was assessed using two techniques, based on first the complexation reaction of (β-D-Glc)3 Yariv's reagent with AGPs, and second the ability of the monoclonal antibody JIM14 to specifically recognize an AGP epitope (Knox et al., 1991). Proteins were extracted from partly and completely differentiated xylem sampled in tension and opposite wood areas, which led to the purification of protein fractions either originating from the plasma membrane and cytosol (fraction A) or bound to the cell wall (fraction B). The relative quantities of AGPs in these samples were estimated by rocket electrophoresis. AGPs were highly abundant in the tension wood xylem area (Fig. 4). Ten µg of fraction A from TW contained more than 1.25 µg Arabic gum equivalent, whereas fraction B contained less than 0.8 µg gum arabic equivalent. Only trace levels were detected in OW.

Figure 4.

Higher amounts of AGPs in tension wood compared with opposite wood as revealed by rocket electrophoresis. Plasma membrane-bound and cytosolic proteins (Fraction A) and cell wall-bound proteins (Fraction B) extracted from differentiating (DX) and mature xylem (MX) were subjected to rocket electrophoresis in agarose gels containing 30 µg ml−1 β-glycosyl Yariv reagent [1,3,5-tri-(p-β-D-glucosylsyloxyphenyl-azo)-2-4-6-trihydroxybenzene], a deep red colored dye that binds specifically to AGPs (Schindler et al., 1995). There is a linear relationship between the area of precipitation peak and concentration of AGP in the samples. Various amounts of gum arabic were used for calibration and results were expressed as gum arabic equivalent. AGP-containing fractions were extracted from differentiating xylem (a) and mature xylem (b).

JIM14 AGP epitope accumulates preferentially in tension vs opposite wood

JIM14, a monoclonal antibody that recognizes an epitope specific to AGPs (Knox et al., 1991), was used to examine AGPs expression by Western blot during tension wood formation. Two polypeptides with an apparent molecular mass of 100 and 200 kDa were detected by JIM14. The observed broad bands typical of glycosylated proteins were present in all samples but appeared a lot stronger in TW fraction B from DX or MX (Fig. 5).

Figure 5.

JIM14 antibody labels two polypeptides in Western blot analysis. Plasma membrane-bound and cytosolic proteins (Fraction A) and cell wall-bound proteins (Fraction B) were extracted from differentiating (DX) and mature xylem (MX). Protein fractions (15 µg) were separated by SDS-PAGE. Arrows indicate the polypeptides recognized by the antibody. Molecular weights in kDa are indicated in the left margin.

Immunocytolocalization of the JIM14 AGP epitope in poplar stem tissues

JIM14 was further utilized to localize AGPs in the tension wood zone of stem sections collected from artificially tilted poplar plants. Strong labelling of the G-fibres was observed in JIM14 treated sections (light microscopy analyses, Fig. 6). No significant labelling was observed in control sections treated only with the secondary antibody confirming the specificity of the immunogold labelling (data not shown). In addition, a uniform but moderate labelling was clearly visible on the middle lamella and primary cell wall of fibres, ray-cells and vessels from both opposite (Fig. 6a,b) and tension wood zones (Fig. 6c,d). Furthermore, JIM14 significantly labelled the cytoplasm and the inner part of ray-cell secondary wall (Fig. 6a,c).

Figure 6.

Immunocytolocalization of JIM14 epitopes in the crosssections of bent poplar stems. Transverse semi-thin sections of 2-yr-old poplar stems tilted for 2 months were analysed by light microscopy. A moderate labelling is visible in the middle lamella and primary cell wall of fibres (F), ray parenchyma cells (R) and vessels (V) within opposite wood (a,b) and tension wood (c,d). JIM14 antibodies line the edge of the G-layer (G) in tension wood (c,d). Secondary cell walls are significantly thinner in early wood (a) than in late wood (b). In picture (d), a clear transition is observed between the late wood (LW) formed before tilting of the tree and the early wood (EW) formed during inclination and corresponding to tension wood (TW).

In the G-fibres, JIM14 epitope was specifically detected close to the inner surface of the G-layer (Fig. 6c,d). We did not observe any labelling in the secondary wall of opposite wood fibres, whether it was in early wood (Fig. 6a) or in late wood (Fig. 6b), even though the latter exhibited a thicker secondary cell wall. A clear transition is visible between late wood tissues formed before tilting the tree and the early wood displaying G-fibres formed in the tilted stem (Fig. 6d). Again, JIM14 epitope decorates the primary cell wall and middle lamella in both types of wood and markedly labels the inner surface of the G-layer in the tension wood zone.

Transmission electron micrograph observations confirmed that the labelling occurred preferentially near the inner surface of the G-layer of tension wood fibres, whereas it was not or was only very weakly detected within the G-layer (Fig. 7). JIM14 epitope was observed similarly in the middle lamella and primary cell wall of both opposite (Fig. 7a,b) and tension wood (Figs 7c.d) tissue sections. Both in tension and opposite wood, a significant labelling was detected in the cytoplasm of ray cells near the inner side of the plasma membrane (Fig. 7e,f). Interestingly, the poorly contrasted structure of the inner side of the G-layer that may be involved in its edification appeared also decorated with JIM14 epitope (Fig. 7g).

Figure 7.

Cellular localization of JIM14 epitope in bent poplar stem. Transverse ultra thin sections of 2-yr-old poplar stem tilted for 2 months were analysed under a transmission electron microscope. Opposite wood (a, b) and tension wood sections (c, d), at two different magnifications. Magnification of (e) a ray cell (f) the cell wall between a fibre (F) and a ray cell (R) and (g) a G-layer. (h) A section presenting the ultrastructural organisation of tension wood fibre cell wall. We observed a uniform and moderate labelling in the middle lamella and primary wall (PW) from both opposite and tension wood sections (a, b, c, d). A significant labelling is visible on the edge of the G-layer (G) but no labelling is detectable within the G-layer (c, d, f, g). An important labelling is visible in a poorly contrasted structure at the inner side of the G-layer (g, see arrow). OW, opposite wood; TW, tension wood.

Discussion

Several PopFLAs are highly and specifically represented in tension wood

Recently, Schultz et al. (2002) classified 47 Arabidopsis glycosylphosphatidylinositol (GPI)-anchored AGPs into four groups: classical AGPs, AG-peptides, basic AGPs and fasciclin-like AGPs. Compared with the three other groups, the FLAs group is quite heterogeneous as it includes AGPs with one or two AGP-like domains and one or two fasciclin-like domains. Johnson et al. (2003) proposed to subdivide the Arabidopsis FLAs into groups named A to D. In this study, we reported 15 poplar full-length cDNA encoding fasciclin-like AGPs, named PopFLA1 to PopFLA15. Nine of these PopFLAs were similar to the group A of FLAs from Arabidopsis. These protein sequences were predicted to have a N-terminal secretion signal, a GPI anchor and one fasciclin-like domain flanked by two AGP-like regions.

The 15 PopFLAs were identified among a set of 10 062 ESTs sequenced from four different cDNA libraries prepared from different wood areas (CZ, DX-TW, DX-OW and MX) (Déjardin et al., 2004). Among the 15 PopFLAs, six were closely related sequences, classified into six different contigs within the cluster 1592, whereas the others originated from nine different clusters. The examined PopFLAs genes were highly expressed during wood formation. Interestingly, none of these PopFLAs transcripts were detected in CZ samples. CZ samples have been harvested by scraping the bark surface with a scalpel and they contain not only cambium cells but also very young expanding xylem and phloem cells. This suggests that PopFLAs are most likely expressed during xylem differentiation once the xylem cells have started to build their secondary cell wall. Several AGP genes were also shown to be highly expressed in differentiating xylem from loblolly pine (Loopstra & Sederoff, 1995; Loopstra et al., 2000; Zhang et al., 2003). As with pine AGPs, the observed specific expression of the PopFLAs in differentiating and mature xylem suggests an important but still undetermined function for these proteins during xylogenesis and secondary cell wall thickening.

Moreover, expression analyses revealed that 10 of the PopFLAs genes, PopFLA1 to PopFLA10, were very strongly expressed in DX-TW and MX-TW, while no or very low expression could be detected in the opposite wood (Fig. 2). This expression profile was consistently accessed by using independant approaches such as ESTs distribution in cDNA libraries, Northern blot hybridisations, cDNA microarray or PCR. In addition, RNA accumulation for these 10 PopFLAs except PopFLA9, appears slightly higher in mature xylem cells than in differentiating xylem (Fig. 2). In loblolly pine, several AGP genes also appeared to be regulated in compression wood (Loopstra & Sederoff, 1995; Zhang et al., 2000). Indeed, based on the analysis of ESTs distribution in pine cDNA libraries, three of the six available pine AGP genes were over-represented in compression wood (Whetten et al., 2001). Compression wood has a mechanical function similar to tension wood, however, it is very different chemically and anatomically. It remains to be elucidated how AGPs act during secondary cell wall differentiation.

Could the specific expression of these 10 PopFLAs be related to features specific to tension wood? TW differs mainly from normal wood by an increased rate in cambial cell division and the differentiation of G-fibres. As PopFLAs are almost not expressed in cambial cells, it is likely that they play a role in the differentiation of G-fibres.

Although PopFLA11–15 are closely related to the other 10 PopFLAs sequences, they exhibit different expression patterns. These five PopFLAs are also specifically expressed in xylem tissues, as are the others; however, their expression wasn't regulated in tension wood. In addition, the expression of four of them, PopFLA11, 13, 14 and 15 (PopFLA15 has been previously published as Pop14A9; AF183809), appears higher in differentiating xylem than in mature xylem. It remains to be determined whether this specific expression pattern accounts for a specific function during cell wall differentiation (Fig. 2). From sequence alignment, we observed that PopFLA11 and 13 were very similar and appeared to merge to the Arabidopsis FLA12 and to the cotton GhAGP1 (Fig. 1). Furthermore, PopFLA11–15, FLA12 and GhAGP1 have a very short proximal AGP-like domain. Pop14A9 was isolated from differentiating poplar xylem and the accumulation of GhAGP1 was reported to increase as soon as 5 d post anthesis during cotton fibre elongation and differentiation (Ji et al., 2003).

Rocket electrophoresis indicated that AGPs accumulated specifically in TW compared with OW either from DX or MX samples (Fig. 4). These results correlate well with the expression pattern of PopFLA1–10. Moreover AGPs appear more abundant in fraction A (Fig. 4), which contains most likely the cytoplasmic cell compartment together with the plasma membrane-bound proteins. AGPs may also be imbedded within the cell wall, or weakly attached to the cell wall.

We investigate the cellular localization of those AGPs with the JIM14 antibody known to recognize epitope from AGPs associated to the cell wall (Knox et al., 1991). This antibody reacts with a carbohydrate epitope that is potentially present on various protein backbones of AGPs. Therefore, it is not specific of the fasciclin-like AGPs, and in poplar wood samples, it mainly binds two polypeptides of high molecular weight (Fig. 5). As the two signals increase markedly in DX-TW and MX-TW compared with OW samples and because the PopFLAs are among the most accumulated transcript in TW, we can reasonably assume that some TW-regulated PopFLAs are recognized by JIM14. It should be pointed out that with JIM14 antibody, the stronger signal obtained in the Western blot analysis occurs in the TW fractions composed of cell wall-bound proteins (Fig. 5 Fraction B). In agreement with this, immunocytolocalization study indicates that JIM14 decorates the inner surface of the G-layer. Nevertheless, recognition of TW-regulated PopFLAs by JIM14 remains to be unequivocally demonstrated. The labelling observed in the middle lamella, the primary cell wall and the ray parenchyma cells may correspond to other AGPs like PopFLA11–15 or even other uncharacterized AGPs. RNA expression analysis, biochemical investigations and immunolocalization support the hypothesis of the high expression of some PopFLAs in tension wood and raise the question of their function in the formation of this tissue.

The strong differences in the expression pattern of the different PopFLAs probably reflect different cellular functions that remain to be elucidated. Based upon their predicted structure, potential role of these proteins during tension wood formation may be related to adhesion properties and signalling.

From deduced structure to predicted function for PopFLAs in tension wood formation

Nine PopFLAs proteins belong to the subclass of FLAs with a single fasciclin domain bordered by two domains predicted to be highly glycosylated usually found in arabinogalactan proteins (AGP-like domains) and a potential anchoring to the glycosylphosphatidylinositol (GPI). Presently, direct evidence for GPI anchoring is still lacking for a number of predicted GPI-anchored proteins and most of them are inferred from sequence analyses. GPI anchoring is believed to enable a great mobility of the protein at the surface of plasma membrane and permits localized delivery of the protein after GPI cleavage by C phospholipases. In silico analysis indicated that all but six PopFLAs are suitable candidates for plasma membrane-bound GPI-anchored proteins. The six FLAs without GPI anchorage (PopFLA3, 4, 6, 8, 9 and 10) are up-regulated in tension wood. These PopFLAs are likely secreted or cell wall-bound proteins and are potential candidates for the AGPs detected by JIM14 in the border of the G-layer. Although many plant GPI-anchored proteins have been documented (Borner et al., 2002; Borner et al., 2003), very few have been functionally studied. However, functional data have been recently reported for three Arabidopsis GPI-anchored proteins encoded by cobra, sku5 and sos5 (Schindelman et al., 2001; Sedbrook et al., 2002; Shi et al., 2003) and for a rice COBRA-like protein (Li et al., 2003). All these GPI-anchored proteins seem to be involved in the orientation of cell elongation and/or in cell-to-cell adhesion. In our study, JIM14 decorates the plasma membrane region and the inner side of the G-layer. This suggests that some GPI-anchored PopFLAs, namely PopFLA1, 2, 5 and 7, are mobilized even transiently to the G-fibre plasma membrane. Their possible function may be related to the establishment of connections between cell wall and cytoskeleton during fibre expansion and elongation.

All the deduced PopFLAs proteins share a central β-immunoglobulin (Ig)-H3/fasciclin domain known to occur as multiple repeats in a limited number of proteins including Drosophila fasciclin I and TGF-β induced protein Ig-H3 domain. β-Ig-H3/fasciclin domain is present in Algal-CAM, a Volvox glycoprotein, and seems to play an important role in cell adhesion (Huber & Sumper, 1994). All 15 PopFLAs exhibit two highly conserved motifs, namely H1 and H2, on the edges of the fasciclin domain (Kawamoto et al., 1998, Fig. 1). Flanking these two motifs, several residues have been shown to be important for mediating adhesion properties (Kim et al., 2000; Kim et al., 2002). In human cells, DI and YH motifs are thought to interact with several components of the extracellular matrix such as integrins. PopFLAs do not contain a DI motif, which is not surprising as this motif is only present in Arabidopsis FLAs from group B (Johnson et al., 2003). However, as in group A of Arabidopsis FLAs, FH (Phe His) residues are often found in PopFLAs in place of the YH (Tyr His) conserved motif. There is still lack of experimental evidence in plants for FLAs adhesion properties. Recently, Shi et al. (2003) published the first report on the functional analysis of a fasciclin-like protein in an Arabidopsis mutant. The mutated gene, FLA4 (also named sos5), encodes a putative cell surface adhesion protein containing two fasciclin-like domains. The sos5 mutant is characterized by abnormal root cell expansion and defective cell wall structure. From this phenotype it has been suggested that FLA4 functions through noncovalent ionic interactions between its fasciclin domains and other FLAs (Johnson et al., 2003). In order to regulate cell expansion, such interactions may occur at the plasma membrane and/or in the cytoplasm.

Alternatively, FLAs like AGPs containing N-acetylglucosamine can be a substrate for chitinase (Rojas-Herrera & Loyola-Vargas, 2002) leading to the release of oligosaccharide signal molecules. Endochitinase activity may either operate through the inhibition of specific AGP signalling molecules, and/or upon the liberation of oligosaccharides acting as Nod factor (Domon et al., 2000). In such a system, already described on a plant embryogenesis model (McCabe et al., 1997), oligosaccharin transmits a signal necessary to induce somatic embryo formation. A similar process may occur during tension wood formation as suggested by the high expression of an endochitinase gene in the DX-TW library (Déjardin et al., 2004). Such a signal could potentially be transmitted through the ray parenchyma cells toward the cambium.

This study demonstrates that many FLAs are expressed in woody tissues, whereas some of them are specifically regulated during tension wood formation and are among the most highly expressed genes in this reaction wood. Likewise, preliminary studies at the protein level revealed an accumulation of AGPs in the tension wood zone. Moreover, high-resolution localization studies with an antibody specifically directed against AGPs indicated that some AGP accumulation occurs close to the plasma membrane on the internal side of the G-layer, a characteristic feature of tension wood fibres. Immunocytological studies with FLAs specific antibodies help delineate the specific localisation of the different PopFLAs. Toward this end, antibodies against specific domains of the protein backbone could be raised, although the extensive glycosylation of these proteins may interfere with recognition between such antibodies and native FLAs in poplar wood. Alternatively, antibodies directed toward wood purified FLAs fractions could circumscribe this problem. Finally, strategies are currently being explored to investigate the function of PopFLAs during wood formation, through the production of transgenic poplars with altered PopFLAs expression.

Acknowledgements

We are grateful to Pr. K. Roberts for kindly providing the JIM14 polyclonal antibodies. We also thank B. Delaleux (Laboratory of Electron Microscopy P.R.C., INRA, Tours, France) for her help in transmission electron microscopy, Dr Michel Duclos (Station de Recherche Avicole, INRA, Tours, France) for kindly giving us access to the array of acquisition and analysis software, V. Guerin for technical assistance in molecular work, N. Millet, R. Blanluet and H. Miteul for plant care. This work was supported by the INRA AIP Transversalité (‘Génomique fonctionnelle des ligneux: Lignome’), by the AGRICE-ADEME (‘Plantes lignocellulosiques et herbacées: modulation de leur lignification et propriétés d’usage par génie génétique’) and by the fund provided by the DERF (grant n°01.40.40/99). F. Lafarguette was supported by the Conseil Régional de la Région Centre. The authors are indebted to Dr John MacKay for correcting the English.

Ancillary