Seasonal reorganization of the xylem transcriptome at different tree ages reveals novel insights into wood formation in Pinus radiata

Authors


Author for correspondence:
Xinguo Li
Tel: +61 2 6246 4848
Email: xinguo.li@csiro.au

Summary

  • Seasonal wood development produces earlywood (EW) and latewood (LW) with distinct properties. The molecular mechanisms controlling EW and LW formation at different tree ages are poorly understood.
  • Seasonal reorganization of the xylem transcriptome was investigated in Pinus radiata at four tree ages using cDNA microarrays. Transcriptome profiles were compared with seasonal wood variation measured by SilviScan (CSIRO, Clayton, Australia).
  • The xylem transcriptome was considerably reorganized during seasonal change, and this reorganization showed a maturation-related pattern. The greater reorganization occurred at the transition (30%) and juvenile (21%) stages, but it declined with tree maturity (11–13%). However, this pattern does not correlate well with maturation-related patterns of seasonal wood variation. In total, 319 seasonal-responsive xylem candidate genes were identified. Many transcripts involved in primary and secondary wall biosynthesis were preferentially accumulated in EW and LW, respectively. A large proportion (45–81%) of the candidate genes are preferentially regulated at a single age and their transcript abundance may influence maturation-related patterns of seasonal wood variation.
  • Seasonal reorganization of the xylem transcriptome is significantly affected by tree age. Physiological changes at the transition stage may contribute to its greater seasonal transcriptome reorganization. Identified stage-preferential xylem transcripts could influence seasonal wood variation at different tree ages.
Abbreviations
4CL

4-cinnamoyl CoA ligase

AGP

arabinogalactan protein

ARP

actin-related protein

C3H

p-coumarate 3-hydroxylase

C4H

cinnamate 4-hydroxylase

CAD

cinnamyl alcohol dehydrogenase

CCoAOMT

caffeoyl CoA O-methyltransferase

CCR

cinnamoyl CoA reductase

CDS

consistent direction selection

CesA

cellulose synthase

COMT

caffeic acid O-methyltransferase

CP

cell population

EST

expressed sequence tag

EM

early mature wood stage

EW

earlywood

FLA

fasciclin-like arabinogalactan protein

GO

gene ontology

JW

juvenile wood stage

LEA

late embryogenesis abundant protein

LM

late mature wood stage

LP

water deficit-inducible protein

LW

latewood

MIP

major intrinsic protein

PAR

photoassimilate-responsive protein

PCBER

phenylcoumaran benzylic ether reductase

PPBG

protective protein for beta-galactosidase

PRP

proline-rich protein

RTD

radical tracheid diameter

RT-MLPA

reverse transcriptase-multiplex ligation-dependent probe amplification

SAMS

S-adenosylmethionine synthetase

SNPs

single nucleotide polymorphisms

SS

specific surface

SuSy

sucrose synthase

TC

tracheid coarseness

TF

transcription factor

TTD

tangential tracheid diameter

TW

transition wood stage

TWT

tracheid wall thickness

UGT

UDP-glucose glucosyltransferase

ZF

zinc finger

Introduction

Wood is a major renewable resource for the timber, paper and emerging bioenergy industries. Wood formation derives from the cumulative annual activity of the vascular cambium. In temperate trees, dormant cambial initials are activated in early spring, and then undergo cell division and rapid expansion, during which primary cell wall synthesis dominates. Later in the growing season, a greater proportion of xylem cells undergo secondary wall synthesis and lignification, followed by programmed cell death. The seasonal dynamics of cambial activity gives rise to annual rings as seen in wood cross-sections. Annual rings are typically composed of two distinct parts, earlywood (EW) and latewood (LW). EW is formed in spring and early summer when temperature, rainfall and photoperiod are favourable for rapid growth, while LW is produced late in the growing season, mostly from midsummer to autumn (Paiva et al., 2008).

In conifers EW and LW are distinct in morphology (Uggla et al., 2001), chemical composition (Bertaud & Holmbom, 2004), wood properties (Ivkovich et al., 2002; Koubaa et al., 2002) and their roles in water transport (Domec & Gartner, 2002; Utsumi et al., 2003). EW tracheids are larger in diameter and have thinner walls than the tracheids of LW (Uggla et al., 2001). However, the mechanical strength of LW is much higher than that of EW (Ivkovich et al., 2002; Koubaa et al., 2002). Among different wood tissues within a tree, the largest variation in wood properties occurs during the transition from EW to LW (Provost et al., 2003). EW is particularly abundant in young trees, and is largely responsible for the poorer properties of juvenile wood (Ivkovich et al., 2002). Therefore, EW and LW formation during seasonal change significantly influences the overall wood properties of harvested logs.

Differential gene expression during EW and LW formation has previously been investigated in loblolly and maritime pine using cDNA microarrays, proteomics and other techniques (Provost et al., 2003; Egertsdotter et al., 2004; Gion et al., 2005; Yang & Loopstra, 2005; Paiva et al., 2008). Of 350–3512 xylem transcripts in pine microarrays, between 5 and 20% showed significant responses to seasonal change (Egertsdotter et al., 2004; Yang & Loopstra, 2005; Paiva et al., 2008). Genes preferentially expressed in EW formation were mostly involved in primary cell wall biosynthesis, while many genes overexpressed in LW played roles in secondary wall formation or had unknown functions (Egertsdotter et al., 2004; Yang & Loopstra, 2005). Compar-ison of two pine microarray studies (Yang & Loopstra, 2005; Paiva et al., 2008) suggested that the proportion and type of seasonal-responsive transcripts may change at different tree ages.

Pinus radiata (radiata pine) is an important commercial forest tree species in Australia, New Zealand, Chile and several other countries. Over the last five decades two generations of radiata pine breeding in Australia have increased growth rates by 43%; however, about one-third of the wood in harvested logs is juvenile wood of poor quality (Wu et al., 2007). In order to understand the molecular mechanisms of wood formation in radiata pine, we have focused our attention on seasonal wood development at different stages of tree maturation. We constructed radiata pine cDNA microarrays containing 18 432 clones from six developing xylem libraries. Among these clones, 6194 were randomly sequenced and then assembled into 3320 xylem unigenes (Li et al., 2009). These microarrays were used to study seasonal reorganization of the xylem transcriptome at four tree ages (5, 9, 13 and 30 yr) spanning a typical radiata pine rotation. These tree ages represent four distinct phases of maturation, namely juvenile (JW), transition (TW), early mature (EM) and late mature (LM) wood stages (Gapare et al., 2006; Wu et al., 2007). Wood property variation during seasonal change was also measured at the four stages. We observed that seasonal reorganization of the xylem transcriptome varies at different tree ages, and changes of transcript abundance may correlate with maturation-related seasonal wood variation.

Materials and Methods

Plant material and sampling

Three Pinus radiata D.Don trees at each of JW (5 yr), TW (9 yr) and EM (13 yr) ages (based on wood ring number at breast height) were sampled in commercial plantations at Bondo, NSW (35°16′44.04″S, 148°26′54.66″E). The sampled trees were growing relatively close to each other in a similar field environment. Two radiata pine trees aged 30 yr were sampled at Yarralumla, ACT (35°18′27″S, 149°7′27.9″E). Each sampled tree was used to collect EW and LW tissues to enable comparisons in the same genetic background. All EW and LW samples were collected on the same (or similar) dates in spring (October) and autumn (March), respectively. The radiata pine EW tissues were primarily undergoing cell division and radial expansion (primary wall synthesis), and LW tissues were predominantly laying down secondary walls according to a previous study (Skene, 1969).

Developing xylem tissues were scraped with a sharp chisel from the exposed xylem surface at breast height (1.3 m) after removing the bark of the sampling area. To avoid the presence of compression wood, we collected EW and LW samples from the same tree on opposite sides of the trunk perpendicular to the prevailing wind direction. Samples were collected in the morning and immediately placed into liquid nitrogen in the field until stored at −80°C.

Microarray construction and characterization

About 3000 cDNA clones from each of the six developing xylem libraries in radiata pine (Li et al., 2009) were randomly isolated using toothpicks or a Versarray colony picker (Bio-Rad). The clones and an additional 35 reference genes were PCR-amplified using the universal M13 primers, followed by purification with ethanol. The purified PCR products were transferred into 384-well plates using a TECAN GENESIS workstation 200 (Tecan Group Ltd, Mannedorf, Switzerland). A total of 18 432 purified PCR products were spotted on GAP II coated slides (Corning Incorporated, New York, USA) using an SDDC2 arrayer (ESI, Toronto, Canada) and Chipmaker 3 pins (Telechem International, Sunnyvale, USA) at 20°C and 60% humidity. Configuration of the microarrays consisted of 48 blocks arranged in 12 rows and four columns. Microarray slides were stored in a slide holder at room temperature and treated with UV cross-linking at 300 mJ cm−3 before use for hybridization.

Of 18 432 spots in the radiata pine cDNA microarrays, 5952 were previously sequenced (Li et al., 2009). Expressed sequence tags (ESTs) of an additional 217 clones were recently registered in GenBank (accession numbers: GO269326GO269542). Thus, a total of 6194 spots (including 35 reference genes) had EST sequences. Assembly of these ESTs using the BIO301 system with default settings (BIO301) resulted in 3320 xylem unigenes, including 986 contigs and 2334 singletons.

Microarray experiments and data normalization

Transcript abundance in EW and LW tissues was directly compared for each tree sampled. Three biological replicates were used for comparisons at JW, TW and EM stages, and two biological replicates used for the LM stage. Each biological replicate included dye swaps, resulting in a total of six or four replicates in each experiment. Total RNA was extracted using a CTAB method (Chang et al., 1993) with slight modifications. Equal amount of total RNA (20 μg) from the two samples being compared was used as the starting template in microarray experiments. Syntheses of cDNAs and probe labelling were performed using the SuperScript™ Plus Direct cDNA Labelling System (Invitrogen). Purified probes from the two samples being compared were combined and dried, and then resuspended in 60 μl hybridization buffer (formamide 500 μl, 20 × SSC 250 μl, 20% SDS 5 μl, sonicated salmon sperm DNA 0.1 mg, dissolved in ddH2O to 1 ml). Microarray hybridizations were incubated at 42°C overnight, followed by washing procedures as described in the manufacturer instructions (Corning Incorporated).

Hybridized microarrays were scanned using a GenePix Personal 4100A microarray scanner (Axon Instruments, Sunnyvale, USA). Wavelengths were adjusted to account for dye bias and acquire optimal spot recognition at minimum and similar saturation for both channels. Images were preprocessed using GenePix® Pro 4.0 and Acuity software (Axon Instruments). Median values of fluorescence intensity of the red and green colours were used to generate a ratio representing differential transcript accumulation of genes in the two samples being compared. The log-2 ratio values of all 22 hybridized microarrays were normalized at both print-tip and slide scale levels using GEPAS v3.1 (Montaner et al., 2006; Tarraga et al., 2008). The initial dataset of the 22 microarrays was registered in the NCBI database with GEO accession number GSE15230.

Analysis of radiata pine cDNA microarrays

A three-step approach was used for the analysis of the radiata pine cDNA microarrays (see ‘the Results section’ for details). Differentially accumulated clones were identified using P- or Q-values (≤ 0.05) calculated with Cyber-T (Baldi & Long, 2001) or SAM (Tusher et al., 2001; Larsson et al., 2005), respectively. Normalized log-2 ratio values of the differentially accumulated transcripts (unigenes) identified at each maturation stage were combined as a single data file for a cluster analysis using the ‘CLICK’ algorithm of the Expander software with an expected mean homogeneity > 0.9 (Sharan et al., 2003; Shamir et al., 2005). Functional categories of each cluster were further analysed by DAVID Bioinformatics Resources 2008 (Huang et al., 2007a,b). This single data file was also used for constructing a hierarchical clustering dendrogram using complete linkage and correlation-based distance of the EPCLUST software (Kapushesky et al., 2004).

Wood property measurements

Wood properties were measured using SilviScan 2 (Evans, 1994; Evans et al., 2000). Six trees aged 13 were sampled from the same plantation as the trees sampled for the EM stage in microarray experiments, while 10 trees aged 25 yr were sampled from an adjacent plantation in the same locality (Bondo, NSW). One 12-mm-diameter wood core was drilled through each tree at breast height. Wood cores were trimmed from the pith to the bark to produce a strip 2 mm thick (tangential) and 7 mm wide (longitudinal). This wood strip was measured at every 25 μm by SilviScan 2, including ring width, radial tracheid diameter, tangential tracheid diameter, cell population, specific surface, wall thickness, tracheid coarseness and wood density. SilviScan profiling of each property trait was partitioned into EW and LW at the four tree maturation stages. Based on the wood density profiles transition phases of the six (13 yr) and 10 (25 yr) trees were 8–9 and 7–9 yr, respectively. Therefore, each wood property was grouped into EW and LW tissues at JW (1–7 or 1–6 yr), TW (8–9 or 7–9 yr), EM (10–13 or 10–15 yr) and LM (16–25 yr) stages.

Multiplex ligation-dependent probe amplification (MLPA)

Microarray transcript abundance for selected candidate genes was validated using the reverse transcriptase-multiplex ligation-dependent probe amplification (RT-MLPA) method (Schouten et al., 2002). Transcripts of 13 candidate genes selected from microarray experiments with trees sampled at the JW stage (5 yr) were validated, including cellulose synthase 3 (CesA3), CesA11, caffeoyl CoA O-methyltransferase (CCoAOMT), chitinase-like, cinnamate 4-hydroxylase (C4H), protective protein for beta-galactosidase (PPBG), basic blue protein, arabinogalactan protein 4 (AGP4), aquaporin, expansin, cellulase, peroxidase and an unknown (PGI.072705#TC66804). Another 11 candidate genes identified from microarray experiments with trees sampled at the TW stage (9 yr) were also validated, including annexin 1c, PPBG, dehydrin, C4H, basic blue protein, aquaporin, AGP4, photoassimilate-responsive protein (PAR), proline-rich protein (PRP), UDP-glucose glucosyltransferase (UGT) and peroxidase.

The transcript abundance of these 24 candidate genes was measured using RT-MLPA with EW and LW tissues collected from eight trees at JW (5 yr) and four trees at TW (9 yr) stages, respectively. Four technical replicates were used for each biological replicate. Approximately 400 ng of DNase-treated total RNA was reverse transcribed into cDNA using the ImProm-II Reverse Transcription System (Promega) and oligo (dT)15 primer. Synthesized cDNA was hybridized with mixed left probe oligo (LPO) and right probe oligo (RPO) probes (Supporting Information, Table S1) designed for the selected candidate genes. Hybridization was performed at 60°C overnight, follo-wed by ligation and PCR amplification with SALSA D4 primer. PCR products from multiple genes were separated using the CEQ™ 8000 Genetic Analysis System and analysed using its built-in software (Beckman Coulter, Brea, USA).

Results

Statistical analysis of the radiata pine microarrays

We took a three-step approach to analyse our redundant microarrays. We first identified differentially accumulated cDNA clones based on their P- or Q-values (≤ 0.05). However, transcript accumulation of a given gene involved in controlling a quantitative trait is expected to be variable among biological replicates. Hence selection based on stability of transcript abundance (P-value) or the false-positive rate (Q-value) alone may eliminate some potential candidate genes. On the other hand, statistics of P- or Q-values require an independent random model for t- or F-tests and a sufficient number of replicates (Pawitan et al., 2005; Jorstad et al., 2008). These requirements are not suitable in redundant microarrays, particularly with limited replicates. To overcome this challenge, an alternative method, which we term ‘consistent direction selection’ (CDS) was used for the selection of differentially accumulated clones. When a clone is consistently up- or down-regulated in a particular wood sample among all replicates with at least a 20% change of the mean abundance value, it will be selected as a differentially accumulated clone regardless of the stability of its abundance. To minimize the risk of eliminating false-negatives, differentially accumulated clones selected using P-values, Q-values and the CDS methods were pooled for further analyses except for those without sequence information.

In the second step, we shortlisted differentially accumulated transcripts from contigs containing one or more differentially accumulated clones using two thresholds (Table 1), whereas differentially accumulated singletons were directly included in the shortlist. The first threshold (A) was used to select a contig if the majority of its redundant clones showed a similar accumulation pattern in at least half the total replicates. This threshold identified redundant clones in a contig with nonrandom accumulation. From the contigs passing threshold A, the second threshold (B) selected contigs with sufficient redundant clones having consistent up- or down-regulation in at least all but one replicate. All differentially accumulated transcripts identified from contigs and singletons were used for the analyses of global transcript accumulation patterns. In the final step, candidate genes were selected from differentially accumulated contigs to ensure a higher degree of confidence. Selected genes had at least a 25% change in average accumulation ratio (or log-2 ratios at ± 0.322) in consistently up- or down-regulated clones of a differentially accumulated contig. This threshold was arbitrarily chosen in an attempt to balance inclusion of biologically significant genes of small differential accumulation (such as transcription factors) with rejection of genes showing no significant differential accumulation. A summary of selected transcripts (at the levels of clones, unigenes and candidate genes) differentially accumulated in EW and LW are listed in Table 2.

Table 1.   Thresholds for differentially accumulated transcripts (unigenes)
Redundant clones in unigeneA%aB%b
  1. aA% represents the proportion of redundant clones in a contig with relatively similar accumulation pattern in at least half of the total replicates.

  2. bB% represents the proportion of redundant clones in a contig with consistent up- or down-regulation in at least all but one replicate.

1100100
2100100
310066.7
4100≥ 70
5–9≥ 80≥ 60
10–19≥ 80≥ 50
20–29≥ 80≥ 40
30–39≥ 80≥ 35
40–81≥ 75≥ 30
Table 2.   Number of selected transcripts preferentially accumulated in earlywood (EW) and latewood (LW)a
TranscriptsMethodbEWJLWJEWTLWTEWMLWMEWRLWRTotalf
  1. aEWJ, LWJ, EWT, LWT, EWM, LWM, EWR and LWR represent earlywood (EW) and latewood (LW) tissues collected at juvenile (5 yr), transition (9 yr), early mature (13 yr) and late mature (or rotation) (30 yr) ages, respectively.

  2. bSee relevant portions of the Materials and Methods as well as the Results sections for a description of the selection approach.

  3. cThe total number of clones on the radiata pine microarrays is 18 432.

  4. dThe total number of unigenes from the 6194 sequenced clones on the radiata pine micorarrays is 3320.

  5. eCandidate genes were only selected from contigs. The total number of contigs from assembly of the 6194 sequenced clones is 986.

  6. fTotal numbers from EWJ, LWJ, EWT, LWT, EWM, LWM, EWR and LWR after the redundancies were removed.

ClonescP-values12381646231018776728181972567044
Q-values11100825087041152003936
CDS5071271184011112145354696145335
Pooled12921773271219467128805076507947
Unigenesd 3972905194762301941602211710
Candidate genese 66681067319264568336

Seasonal reorganization of the xylem transcriptome at different tree ages

We investigated xylem transcriptome reorganization during seasonal change at four distinct stages of wood maturation. Among transcripts of 3320 xylem unigenes analysed in the radiata pine microarrays, between 11.5 and 29.9% are preferentially accumulated in either spring (EW) or autumn (LW) at different maturation stages (Fig. 1a). This suggests that seasonal wood development involves considerable xylem transcriptome reorganization and this reorganization is significantly affected by tree age. The greatest seasonal reorganization of the xylem transcriptome occurs in trees at the TW stage (9 yr; 29.9%), followed by the JW stage (5 yr; 20.7%), whereas only 11.5 and 12.7% of the xylem transcriptome are responsive to seasonal change at EM (13 yr) and LM (30 yr) stages, respectively (Fig. 1a).

Figure 1.

 Xylem transcriptome reorganization during seasonal change in radiata pine. (a) Percentage of transcripts (unigenes) preferentially accumulated in earlywood (EW, grey bars with white stripes) and latewood (LW, black bars with white dots) tissues at four tree ages: 5 yr, juvenile age; 9 yr, transition age; 13 yr, early mature age; 30 yr, late mature age. (b) Hierarchical clustering dendrogram of transcripts (unigenes) differentially accumulated in EW and LW tissues at the four tree ages.

We used hierarchical clustering dendrograms to compare the seasonal-responsive xylem transcripts (unigenes) and their abundance at the four stages (Fig. 1b). Greater similarity of seasonal reorganization of the xylem transcriptome was observed between JW (5 yr) and TW (9 yr) trees, and between EM (13 yr) and LM (30 yr) trees. This hierarchical pattern was also observed in analyses using 7947 clones or 626 contigs differentially accumulated in EW and LW (data not shown), respectively. Interestingly, while the first three tree ages are relatively close (5, 9 and 13 yr), transcriptome reorganization in the EM trees (13 yr) most closely resembled the reorganization taking place in LM trees (30 yr). We conclude that seasonal transcriptome reorganization in a typical radiata pine rotation has two distinct phases: an early phase before the EM stage (including JW and TW) and a late phase after wood has assumed mature properties (including EM and LM). The largest seasonal transcriptome reorganization takes place during the transition from juvenile to mature wood.

Co-regulation patterns of seasonal-responsive xylem unigenes

We identified a total of 2487 unigenes differentially regulated in either EW or LW at the four maturation stages. These unigenes represented 1710 distinct unigenes after removing the redundancies (Table 2). Abundance of these distinct unigenes was clustered into 12 groups, which revealed apparent co-regulation patterns across the wood maturation process (Fig. 2). The two largest groups included transcripts preferentially accumulated in EW (cluster 1, 461 genes) and LW (cluster 2, 294 genes) at the TW stage (9 yr). This further suggested that the xylem transcriptome at the TW stage is particularly responsive to seasonal change. Transcripts preferentially accumulated in EW (cluster 3, 249 genes) and LW (cluster 4, 191 genes) at the JW stage (5 yr) are also highly represented. However, the later stages of wood maturation (13 and 30 yr) have relatively few transcripts responsive to seasonal change. We also observed some transcripts (cluster 8, 65 genes) preferentially accumulated in LW formation at the two earlier stages (5 and 9 yr). Interestingly, cluster 11 (21 genes) included transcripts preferentially accumulated in EW at younger ages (5 and 9 yr) but then preferentially accumulated in LW of mature trees (13 and 30 yr).

Figure 2.

 Co-regulation patterns of differentially accumulated transcripts in earlywood (EW) and latewood (LW). A total of 1710 transcripts (unigenes) differentially accumulated in EW and LW at the four tree ages were clustered into 12 groups using the ‘CLICK’ function of the Expander software. Average homogeneity of the 12 groups is 0.966 and the average separation score is −0.172. Only one unigene was not clustered in any group. Mean log-2 ratio (LW/EW) of transcripts in a cluster is presented on the y-axis at the tree ages shown on the x-axis. Error bars represent the standard deviation of the mean log-2 ratio.

A large proportion (1641, 96%) of the 1710 distinct unigenes had matches in the UniProt (65.9%) and TIGR (additional 30.1%) databases. These distinct unigenes included 818 and 892 transcripts preferentially accumulated in EW and LW, respectively. From their matches showing unique accession numbers, the 818 and 892 transcripts may represent at least 526 and 479 distinct genes preferentially regulated in EW and LW, respectively. A DAVID functional annotation chart (Huang et al., 2007a,b) of the distinct genes revealed some unique gene ontology (GO) terms (Table S2), which are preferential to EW or LW formation. For example, transport, proteolysis, catabolic process, cellular component organization and membrane are preferential to EW formation. By contrast, catalytic activity, purine nucleotide binding and cytoplasmic parts are preferential to LW formation.

Identification of seasonal-responsive xylem candidate genes

From 986 xylem contigs present in the radiata pine microarrays, we identified a total of 443 transcripts preferentially accumulated in either spring (EW) or autumn (LW) at one or more of the four tree ages. After removing the redundancies at different tree ages, 336 transcripts remained as distinct candidate genes. All but one of these genes had matches in the UniProt and TIGR databases, which yielded 319 unique accession numbers, representing seasonal-responsive xylem candidate genes (Table S3). From these candidate genes, 13 and 11 genes identified at juvenile and transition ages, respectively, were selected for validation by RT-MLPA. Our results showed that RT-MLPA data were relatively consistent with microarray transcript abundance (Fig. 3a,b). We concluded that the microarray experiments in this study were sufficiently reliable for the identification of candidate genes responsive to seasonal change.

Figure 3.

 Validation of selected candidate genes using reverse transcriptase-multiplex ligation-dependent probe amplification (RT-MLPA). (a) Thirteen genes selected from microarray experiments at juvenile age: 1, cellulose synthase 3 (CesA3); 2, CesA11; 3, caffeoyl CoA O-methyltransferase (CCoAOMT); 4, chitinase-like; 5, cinnamate 4-hydroxylase (C4H); 6, protective protein for beta-galactosidase (PPBG); 7, basic blue protein, 8, arabinogalactan protein 4 (AGP4); 9, aquaporin; 10, expansin; 11, cellulase; 12, peroxidase; 13, unknown (PGI.072705#TC66804). (b) Eleven genes selected from microarray experiments at transition age: 1, annexin 1c; 2, PPBG; 3, dehydrin; 4, C4H; 5, basic blue protein; 6, aquaporin; 7, AGP4; 8, photoassimilate-responsive protein (PAR); 9, proline-rich protein (PRP); 10, UDP-glucose glucosyltransferase (UGT); 11, peroxidase. Log-2 ratio (LW/EW) > 0 and < 0 represent transcripts preferentially accumulated in latewood (LW) and earlywood (EW), respectively. Error bars represent the standard deviation of the mean log-2 ratio of transcript abundance. Grey bars, microarrays; black bars with white dots, RT-MLPA.

A large proportion of the 319 candidate genes are involved in cell wall biosynthesis (85 genes, 26.6%), suggesting cell wall-related genes are particularly responsive to seasonal change. Cell wall-related transcripts preferentially accumulated in EW (Table S3) fall into several functional groups, including cell division (cyclin-like F-box and profiling-1), differentiation (clavata-like receptor), expansion (expansins), actin skeleton (actins and ARP6), cell wall components (AGP4, FLA16, AGP-like and PRPs) and pectin pathway (pectate lyases, pectinesterase and UDP-apiose/xylose synthase). By contrast, cell wall-related transcripts preferentially accumulated in LW are mostly involved in cellulose synthesis (CesA3, CesA7, SuSy, callose synthase-like and cellulase), lignification (4CL, C3H, CAD, CCoAOMT, COMT, laccase, PCBER, endochitinase and chitinase-like), microtubules (tubulins), cell wall components (AGP5), cell wall degradation (beta-1,3-glucanase) and cell death (metacaspase type II).

Some transcripts responsive to auxin and ABA signals (auxin-regulated protein, cullin 1A, cullin-like, 14-3-3 and 14-3-3-like) are preferentially accumulated in EW, whereas ethylene-responsive transcripts (ethylene responsive element binding factor and ethylene-forming enzyme) are more abundant in LW. Transcripts involved in water transport and drought stress have diverse regulation patterns. For example, aquaporins, major intrinsic protein 2 (MIP-2) and water deficit-inducible protein (LP3) transcripts are more present in EW, while dehydrins, aquaporin-like, late embryogenesis abundant protein (LEA), desiccation-related protein and LP6 are preferentially accumulated in LW. Transcripts of some transcription factors are preferentially accumulated in EW, including genes encoding homeodomain-leucine zipper (HB-3), class III HD-Zip (HDZ31), pollen-specific LIM domain (LIM) and zinc finger-like (ZF-like) proteins, while transcripts of GATA13, tubby-like and ZF genes are preferentially accumulated in LW. Interestingly, a number of transcripts are consistently up-regulated in EW (AGP4, peroxidase, LP3 and serine/threonine protein kinase BRI1-like 2) or LW (dehydrin) in at least three of the four tree ages. In addition, a significant proportion of seasonal-responsive candidate genes (71, 22.3%) are functional unknowns, highlighting our poor understanding of seasonal wood development at the molecular level.

Differential regulation of seasonal-responsive xylem candidate genes at different tree ages

Among 319 seasonal-responsive xylem candidate genes, 137 and 131 are preferentially responsive to spring (EW) and autumn (LW), respectively, at different tree ages. More seasonal-responsive candidate genes were identified at the JW (101) and TW (125) stages than at the EM (22) and LM (80) stages. Between 45.5 and 81.2% (61.0% on average) of the seasonal-responsive xylem candidate genes are differentially regulated at a single age (Fig. 4). In total, 217 seasonal-responsive xylem candidate genes were identified as stage-preferential transcripts from the four tree ages (Table S3).

Figure 4.

 Differential regulation of seasonal-responsive xylem candidate genes at different tree ages. The number of transcripts (candidate genes) preferentially accumulated in earlywood (EW) (a) and latewood (LW) (b) at different tree ages. Seasonal-responsive candidate genes common to two or more tree ages are indicated. The percentage (%) of transcripts preferential to a particular tree age is shown. One common transcript from candidate genes selected at 9 and 13 yr in LW was not indicated in panel (b) to keep it simple.

A number of transcripts involved in cell division, expansion, pectin pathway, hormone signalling (auxin and ABA) and transcription are preferentially accumulated in EW formation at the JW and TW stages, although the responding genes in each category differ between the two stages. Transcripts involved in actin skeleton development, water transport and cell wall structural proteins are also preferentially accumulated in EW at the TW stage. By contrast, some transcripts involved in tubulin skeleton development, disease resistance or other stress responses are preferential to EW formation at the EM or LM stages. During LW formation, several transcripts involved in drought stress and the lignin pathway are preferential to the JW and TW stages, but cellulose synthesis transcripts are only preferential to the JW stage. By contrast, transcripts related to both cellulose synthesis and lignin pathway are not preferential to LW formation at the EM and LM stages. In addition, transcripts involved in ethylene, light, calcium and ubiquitin signals were observed to be preferential to LW formation at the LM stage. Interestingly, transcripts of the two ethylene signalling genes preferential to LW at the LM stage contrast with the transcripts of auxin and ABA signalling genes, which were preferential to EW at the JW and TW stages.

Seasonal wood variation during wood maturation process

SilviScan profiling of two sets of radiata pine trees aged 13 and 25 yr (based on ring number at breast height) revealed marked differences in wood properties between EW and LW tissues at the four maturation stages (Figs S1, S2). A large proportion of wood tissues within each ring is EW, particularly in trees at the JW stage. EW tracheids are wider in both radial and tangential diameters (RTD and TTD) and larger in specific surface (SS) than LW tracheids. The walls of tracheids in LW are much thicker than EW tracheids and LW has more tracheids per unit area (cell population, CP) than EW. These morphological differences directly correspond with seasonal wood development, given that tracheid cell growth and elongation are active in spring, and secondary cell walls are primarily formed later in the growing season. The tracheids of EW are less coarse (mass per unit tracheid length), a positive characteristic for paper-making. In terms of mechanical strength, LW has greater wood density than EW, which is consistent with the smaller size and thicker walls of LW tracheids.

Most wood properties of either EW or LW gradually change with tree age (Figs S1, S2), except for a few properties which are similar between EM and LM stages (Fig. S2). Comparisons of wood property profiles from the two data sets revealed relatively consistent patterns of seasonal wood variation with tree age (Fig. 5). Seasonal variation of tangential tracheid diameter (TTD), SS, wall thickness (TWT), coarseness (TC) and wood density showed a similar pattern, in which their seasonal variations are smaller at the JW and LM stages, but greater at the TW and EM stages, whereas seasonal variation of EW and LW width, RTD and CP gradually changed with tree age.

Figure 5.

 Seasonal wood variation at different stages of wood maturation. Two sets of radiata pine wood samples from trees at age 13 (6 trees) and 25 yr (10 trees) were measured by SilviScan. Eight wood property traits were analysed: width of earlywood (EW) and latewood (LW) within a ring, radical tracheid diameter, tangential tracheid diameter, specific surface, cell population, tracheid wall thickness, tracheid coarseness and density. Seasonal variation for a wood property trait was calculated as a percentage value using (LW − EW)/EW×100. SilviScan profiling data was grouped into three or four stages of wood maturation based on the transition age (8–9 or 7–9 yr for trees at 13 or 25 yr, respectively), including juvenile (JW), transition (TW), early mature (EM) and late mature (LM) wood stages. Black bars with grey strips, trees aged 25 yr; grey bars with white dots, trees aged 13 yr.

Discussion

Redundant cDNA microarrays

Anonymous (Hayward et al., 2000; Hegarty et al., 2005; Cannon et al., 2006; Wintz et al., 2006; Park et al., 2008; Qiu et al., 2008) and partially anonymous microarrays (Lehnert et al., 2004; Donaldson et al., 2005; Wu et al., 2005) have been widely used for transcript profiling because of their lower cost in construction (Hegarty et al., 2005; Kammenga et al., 2007) and internal technical replicates from redundant clones (Lehnert et al., 2004; Hegarty et al., 2005; Wintz et al., 2006). Transcript abundance of redundant clones with lower redundancy is usually consistent (Lehnert et al., 2004; Aspeborg et al., 2005) and highly reproducible (Hegarty et al., 2005); however, conflicting accumulation of redundant clones was observed in Eucalyptus (Qiu et al., 2008) and maritime pine (Gion et al., 2005). In this study, between 5.2 and 16.8% (12.3% on average) of transcripts (contigs) (containing up to 81 redundant clones) showed opposite accumulation patterns at different maturation stages (data not shown). Diverse accumulation patterns of redundant clones may be caused by differences in DNA concentration, flanking gene regions, fragment length and position on microarrays; uneven hybridization and washing within and/or among microarray slides; multiple PCR products in a microarray spot; cross-hybridization among gene family members and/or gene duplications; indels and single nucleotide polymorphisms (SNPs) in redundant clones; and random errors in data normalization, statistical analysis, EST sequencing and assembly.

Diverse accumulation of redundant clones and its impact on the selection of differentially expressed genes have been recognized previously (Anderssen et al., 2004; Lehnert et al., 2004; Hegarty et al., 2005). However, analyses of redundant microarrays have largely relied on statistical methods developed for nonredundant microarrays (Hayward et al., 2000; Hegarty et al., 2005; Cannon et al., 2006; Wintz et al., 2006; Park et al., 2008; Qiu et al., 2008). The CDS approach we used to analyse radiata pine redundant cDNA microarrays identified differentially accumulated clones that largely overlapped with those identified using P- and Q-values (Fig. S3). We therefore pooled all clones identified by the three methods for the identification of differentially accumulated transcripts (unigenes), aiming to minimize the risk of eliminating genuinely differentially accumulated clones. Our results suggested that the three-step approach is useful in the statistical analysis of redundant microarrays.

Seasonal-responsive xylem genes

Seasonal wood development is induced and controlled by hormone signals (Sundberg et al., 1991; Uggla et al., 1998, 2001; Israelsson et al., 2005; Mwange et al., 2005; Hou et al., 2006; Nilsson et al., 2008). Several hormone-responsive transcripts were differentially accumulated in response to seasonal change in radiata pine, including transcripts involved in auxin, ABA, GA, cytokinin and ethylene signalling. For example, some auxin-related transcripts were preferentially accumulated in EW of radiata pine, whereas ethylene-responsive transcripts are more accumulated in LW formation. Differential transcript accumulation of hormone signalling-related genes has also been observed in aspen (Moyle et al., 2002) and loblolly pine (Egertsdotter et al., 2004).

Seasonal-responsive xylem candidate genes of radiata pine include a large proportion (26%) of cell wall-related genes. Many transcripts involved in primary and secondary cell wall biosyntheses are preferentially accumulated in EW and LW, respectively, which is broadly consistent with earlier findings in various tree species (Egertsdotter et al., 2004; Gion et al., 2005; Yang & Loopstra, 2005; Paiva et al., 2008). Transcripts involved in cell wall structural protein have particularly diverse accumulation patterns in EW and LW formation. An arabinogalactan protein transcript, AGP4, is consistently overaccumulated in EW at all four age classes of radiata pine, as was observed in maritime pine (Gion et al., 2005; Paiva et al., 2008) and loblolly pine (Yang et al., 2005). Two other AGPs (AGP-like and FLA16) are also preferentially accumulated in EW of radiata pine. Preferential transcript accumulation of two PRP genes in EW of radiata pine is consistent with findings in loblolly pine (Egertsdotter et al., 2004). By contrast, AGP5 transcript is preferentially accumulated in LW of radiata (this study) and loblolly pine (Yang & Loopstra, 2005; Yang et al., 2005). Therefore, different AGPs appear to play crucial roles in either primary or secondary cell wall formation. The recent observation that fasciclin-like AGPs play a key role in the mechanical properties of Arabidopsis and eucalypt stems suggests a functionally conserved role in conifers (MacMillan et al., 2010).

Cell wall-related transcripts differentially accumulated in EW and LW may play key roles in conferring distinct wood properties during seasonal change. Preferential transcript accumulation of genes involved in cell division, differentiation, expansion and the pectin pathway during EW formation coincides with the greater width of EW within an annual ring, the wider diameter of EW tracheids and their larger specific surface (Figs S1 and S2). By contrast, transcripts preferentially accumulated in LW are mostly involved in cellulose synthesis, lignification, tubulin skeleton development, cell wall degradation and cell death. These genes would probably be involved in the deposition of S2 layers of secondary cell walls in LW tracheids and thus confer stronger mechanical strength to LW tissues (Figs S1, S2).

Maturation-related seasonal transcriptome reorganization

Our current study revealed a maturation-related pattern of seasonal transcriptome reorganization, with the greatest seasonal reorganization occurring at the TW stage (Fig. 1). There are several factors which may account for this increased reorganization at the transition phase. Radiata pine trees at transition ages (8–11 yr based on LW density) (Gapare et al., 2006) are usually in the rapidly growing (exponential) phase of their lifetime. The maturation-related seasonal transcriptome reorganization correlates well with the growth curve of radiata pine. Transcript accumulation of genes involved in cell division, differentiation, expansion, energy metabolism, cellular transport and transcription at the transition phase, particularly in spring, coincides with its greater growth rate. Maintenance of a rapid growth rate may contribute largely to the increased seasonal reorganization at the transition phase.

The wood density-based transition ages typically coincides with a shift from free growth to increased competition from neighbouring trees. Consequently, transition-aged trees are likely to be influenced by tree spacing, growth rate, rainfall, wind, soil and other environmental factors. Physiological responses to these changes may contribute significantly to the greater seasonal transcriptome reorganization observed in trees at the transition phase. For example, several genes related to light, heat and water status are responsive to seasonal change at the TW stage (Table S3). Transition ages also coincide with the appearance of reproductive tissues. During this phase trees initiate flowering and eventually achieve maximal reproductive capacity. Hormone signals involved in reproductive growth are transported through xylem (Robert & Friml, 2009). Thus, flowering in spring and fruiting in autumn in trees in the transition phase may contribute to the seasonal transcriptome reorganization compared with the JW stage dominated by vegetative growth. For example, transcripts responding to auxin which are preferential to the TW stage may have roles in flowering. The increased seasonal reorganization may also be influenced by variations in the timing of the transition phase among the sampled trees at 9 yr.

On the other hand, the lower seasonal reorganization of the xylem transcriptome in trees at the LM stage may be related to the lower variation of EW and LW properties, as shown in Fig. 5. The lower seasonal reorganization at the mature wood stages (EM and LM) correlates with the more stable surrounding environment that mature aged trees experience after canopy closure. In addition, it is also related to the decreased growth rate in mature aged trees of radiata pine.

Correlation of transcript abundance and wood properties

The xylem transcriptome is particularly responsive to seasonal change at earlier stages of tree maturation (TW and JW) (Fig. 1a,b). By contrast, seasonal wood variation tended either to show a gradual directional change with age for most property traits, or was very similar between TW and EM stages for other properties (Fig. 4). This suggested a relatively poor correlation between global changes in the xylem transcriptome and wood properties. Thus, the magnitude of seasonal transcriptome reorganization alone does not explain seasonal wood variation at the four tree ages. This is probably because only 30% of the seasonal-responsive xylem unigenes identified in this study are involved in cell wall formation. The other 70% are either functional unknowns (31%) or have no clear roles in cell wall biosynthesis. These poorly understood unigenes may mask the link between the xylem transcriptome and wood properties in response to seasonal change at different maturation stages.

We found that about 61% of the seasonal-responsive xylem candidate genes are preferential to a single tree age (Fig. 4). Among these stage-preferential genes, more cell wall-related genes are present at earlier stages (33.3% at 5 yr and 27.8% at 9 yr) than at the mature stage (10.6% at 30 yr) (Table S3). These stage-preferential cell wall genes may provide molecular clues for the seasonal wood variations at a particular tree age. Frequently, genes belonging to the same gene family were differentially regulated at all or most tree ages; however, more often, different family members were differentially regulated at particular tree ages. This suggests that some radiation within gene families may be driven by maturation-related processes. Further functional analysis of the stage-preferential genes, particularly genes involved in cell wall formation, could shed more light on age-related changes of EW and LW variation. The new knowledge gained will be invaluable for the development of molecular tools aimed at improving the wood properties of radiata pine.

Acknowledgements

The authors would like to thank Bala Thumma and Charlie Bell for their technical advice with RT-MLPA, Shannon Dillon for sharing wood property data, Iain Wilson for help with the microarray robot and scanner, and Chris Boland for assistance with the colony picking robot. We also thank Dr Sally-Ann Walford and two anonymous reviewers for valuable criticism of the manuscript. This work was supported by funds from Forest and Wood Products Australia (FWPA), ArborGen LLC, the Southern Tree Breeding Association (STBA), Queensland Department of Primary Industry (QDPI) and the Commonwealth Scientific and Industrial Research Organization (CSIRO).

X.L. carried out cDNA library construction, microarray construction, microarray experiments, RT-MLPA validation, statistical analysis and manuscript preparation. S.S. and H.W. proposed the research project and guided the research process. All the authors have read and approved the final manuscript.

Ancillary

Advertisement