Quantitative genetic analysis of biomass and wood chemistry of Populus under different nitrogen levels


  • Evandro Novaes,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Luis Osorio,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Derek R. Drost,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    2. Plant Molecular and Cellular Biology Graduate Program, University of Florida, PO Box 110690, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Brianna L. Miles,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Carolina R. D. Boaventura-Novaes,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Catherine Benedict,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Christopher Dervinis,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Qibin Yu,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Robert Sykes,

    1. National Renewable Energy Laboratory, US Department of Energy, 1617 Cole Blvd., Golden, CO 80401, USA;
    Search for more papers by this author
  • Mark Davis,

    1. National Renewable Energy Laboratory, US Department of Energy, 1617 Cole Blvd., Golden, CO 80401, USA;
    Search for more papers by this author
  • Timothy A. Martin,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    Search for more papers by this author
  • Gary F. Peter,

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    2. Plant Molecular and Cellular Biology Graduate Program, University of Florida, PO Box 110690, Gainesville, FL 32611, USA;
    3. University of Florida Genetics Institute, University of Florida, PO Box 103610, Gainesville, FL 32611, USA
    Search for more papers by this author
  • Matias Kirst

    1. School of Forest Resources and Conservation, University of Florida, PO Box 110410, Gainesville, FL 32611, USA;
    2. Plant Molecular and Cellular Biology Graduate Program, University of Florida, PO Box 110690, Gainesville, FL 32611, USA;
    3. University of Florida Genetics Institute, University of Florida, PO Box 103610, Gainesville, FL 32611, USA
    Search for more papers by this author

Author for correspondence:
Matias Kirst
Tel:+1 352 846 0900
Email: mkirst@ufl.edu


  • • The genetic control of carbon allocation and partitioning in woody perennial plants is poorly understood despite its importance for carbon sequestration, biofuels and other wood-based industries. It is also unclear how environmental cues, such as nitrogen availability, impact the genes that regulate growth, biomass allocation and wood composition in trees.
  • • We phenotyped 396 clonally replicated genotypes of an interspecific pseudo-backcross pedigree of Populus for wood composition and biomass traits in above- and below-ground organs. The loci that regulate growth, carbon allocation and partitioning under two nitrogen conditions were identified, defining the contribution of environmental cues to their genetic control.
  • • Sixty-three quantitative trait loci were identified for the 20 traits analyzed. The majority of quantitative trait loci are specific to one of the two nitrogen treatments, demonstrating significant nitrogen-dependent genetic control. A highly significant genetic correlation was observed between plant growth and lignin/cellulose composition, and quantitative trait loci co-localization identified the genomic position of potential pleiotropic regulators.
  • • Pleiotropic loci linking higher growth rates to wood with less lignin are excellent targets to engineer tree germplasm improved for pulp, paper and cellulosic ethanol production. The causative genes are being identified with a genetical genomics approach.
Abbreviations: eQTL

expression quantitative trait locus


linkage group


logarithm of the odds ratio


pyrolyzer coupled to a molecular beam mass spectrometer


quantitative trait locus

S/G ratio

ratio of S- to G-lignin


simple sequence repeat


As a growing body of evidence supports the negative effects of accumulation of CO2 and other greenhouse gases in the atmosphere (Lenoir et al., 2008; Rosenzweig et al., 2008), society is increasingly turning to forests and forest management for the mitigation of atmospheric CO2 (Canadell & Raupach, 2008). Forests store c. 45% of terrestrial carbon (Bonan, 2008) and cellulosic ethanol production from wood has great potential to diminish the need for fossil fuels, limiting atmospheric CO2 accumulation (Sticklen, 2008). Therefore, increasing the productivity of plantation forests could have a significant impact on society by simultaneously enhancing carbon sequestration and meeting greater demands for renewable wood and bioenergy products.

The chemical composition of wood, cellulose (45–50%), hemicellulose (25%) and lignin (25–35%) (Plomion et al., 2001), is important for its conversion into products and for carbon sequestration. Because lignin is richest in carbon and the most recalcitrant component of wood, higher proportions can translate into more carbon stored for longer periods of time. However, higher lignin contents may be undesirable for the production of pulp/paper and cellulosic ethanol. For these applications, lignin needs to be extracted from wood with relatively harsh chemical treatments and high energy inputs (Li et al., 2003; Chen & Dixon, 2007; Sticklen, 2008). One strategy for increasing carbon sequestration and improving pulp, paper and cellulosic biofuel productivity is to raise the lignin content in nonharvested roots while reducing lignin in the woody stem. Therefore, the development of tree germplasm that is optimal for carbon sequestration and biomass conversion requires an understanding of the genetic regulation of growth, carbon allocation among plant organs and carbon partitioning into lignin, cellulose and hemicellulose within organs.

In woody plants, studies have identified a consistent, significant correlation between biomass growth and wood composition that is genetically regulated (Hu et al., 1999; Wu et al., 1999; Kirst et al., 2004; Yu et al., 2006). Pleiotropic genetic loci coordinating wood composition and stem growth are important targets for the enhancement of wood products. Populus is an excellent model species to identify these genetic elements, given the availability of several segregating pedigrees, easy clonal propagation, well-established transformation protocols and the genome sequence of a P. trichocarpa genotype (Tuskan et al., 2006). Previous studies have identified quantitative trait loci (QTLs) for biomass accumulation and allocation in segregating Populus families (Bradshaw & Stettler, 1995; Wu, 1998; Wu et al., 1998; Wullschleger et al., 2005; Rae et al., 2008). Above- and below-ground growth QTLs have also been mapped in poplar grown under ambient and elevated CO2 (Rae et al., 2007). However, few studies have attempted to map QTLs for wood chemical composition in poplar (Zhang et al., 2006) or, more importantly, have dissected the coordinate regulation of stem growth and wood chemistry.

In addition to genetics, environmental cues at the level of nutrient availability have a tremendous impact on tree growth, biomass allocation and wood composition. Nitrogen is generally the most limiting nutrient for tree growth and carbon sequestration (Oren et al., 2001; Finzi et al., 2007). Poplar trees exhibit extensive phenotypic plasticity in response to nitrogen. Nitrogen fertilization of young (< 1 yr old) poplar trees increases shoot biomass and wood cellulose content, while decreasing lignin content and the ratio of S- and G-lignin (S/G ratio) (Cooke et al., 2005; Pitre et al., 2007a,b). Nitrogen supply has also been observed to increase the photosynthetic rate in mature leaves (Cooke et al., 2005) and to alter xylem fiber anatomy by thickening and shortening (Pitre et al., 2007a). Changes in mRNA abundance have also been observed in poplar in response to nitrogen treatments (Cooke et al., 2003). Although these studies have contributed to a better understanding of the effects of nitrogen on tree anatomy and physiology, little is known about how genotypes interact with nitrogen and how these interactions are regulated in forest species.

The objective of this study was to define the genetic loci controlling the phenotypic variation in biomass accumulation, carbon allocation and partitioning in an interspecific Populus pedigree grown under two different nitrogen conditions. To identify these regions, a QTL mapping experiment with 396 clonally replicated genotypes was completed. Extensive phenotypic plasticity in response to nitrogen availability was observed. The experimental design allowed the analysis of genotype by nitrogen interaction for the first time in a tree QTL mapping population, resulting in the identification of 51 loci that control traits only under one nitrogen condition. Biomass accumulation, cellulose and lignin levels in wood were strongly correlated, genetically and phenotypically. More importantly, some QTLs for biomass growth and wood composition mapped to the same genomic region, delineating potential pleiotropic regulators that coordinate these traits.

Materials and Methods

Poplar pseudo-backcross pedigree phenotyping

An interspecific pseudo-backcross pedigree (family 52-124) composed of 396 genotypes was created by crossing the hybrid female clone 52-225 [P. trichocarpa (clone 93-968) × P. deltoides (clone ILL-101)] with the male P. deltoides clone D124. Six cuttings of each progeny genotype were planted in pots 41 cm deep (TPOT2, Stuewe & Sons, Inc., Corvallis, OR, USA) with Fafard 4MIX soil (Canadian Sphagnum Peat 40%, processed pine bark and vermiculite) in a glasshouse at the University of Florida. Plants were arranged in the glasshouse following a partially balanced incomplete block design with three biological replicates, six incomplete benches per replicate and two nitrogen treatments. Within each replicate, a coordinate system with 30 rows and 26 columns, where each plant was located in a row × column intersection, was utilized to account for possible systematic sources of variation across the glasshouse in our statistical model (see below). The total number of plants in the experiment was 2376 (i.e. 396 genotypes × 2 nitrogen treatments × 3 biological replicates). After potting, all plants were grown for 6 wk with 5 mM NH4NO3 supplied with Hocking's complete nutrient solution (Hocking, 1971) in a flood irrigation system of ebb-and-flow benches. Benches were flooded twice a day for approximately 30 min. During the 5th week of growth, initial diameters and heights were measured. At the start of week six, one half of the benches in each replicate was flooded with Hocking's solution supplemented with 25 mM NH4NO3 and the other half was flooded with the same solution, but without any nitrogen supplement.

After growing for 4 wk with and without nitrogen, during the late summer and early fall of 2006 (a total of 10 wk after potting), plants were harvested and phenotyped for total height, stem diameter at 5 cm above the bottom of the cutting, number of sylleptic (lateral) branches and number of internodes. After these measurements were taken, the plants were dissected into roots, main stem, leaves and sylleptic branches with leaves attached. By peeling off the bark, the main stem was further separated into two tissues: xylem and bark + phloem. Tissue samples were placed into barcode-labeled paper envelopes and dried in a 65°C drying room or a freeze-drier (Freezezone 18L Bulk Tray Dryer, Labconco, Kansas City, MO, USA), depending on the tissue. Dried tissues were conditioned in the laboratory for more than 1 wk before being weighed with high-precision balances. Weights of individual organs and tissues were summed to estimate above-ground (shoot) and total biomasses. Biomass allocation into root and shoot was calculated by the ratio of above-ground to root weights.

Wood chemical composition was measured in two biological replicates of the experiment from the 5–10 cm of bottom xylem of each plant's main stem. The xylem was ground in a Wiley mill to pass through a 20 mesh screen and re-dried. Two subsamples (approximately 4 mg each) of milled wood from each plant were scooped into 80 µl stainless steel sample cups that were subsequently covered with glass fiber paper (type a/d). The cups were automatically loaded into a pyrolyzer coupled to a molecular beam mass spectrometer (py-MBMS) at the National Renewable Energy Laboratory of the US Department of Energy (NREL – DOE), Golden, CO, USA. Samples were separated into two blocks and completely randomized. To evaluate the consistency and accuracy of the instrument, one loblolly pine and three Populus wood samples, previously characterized with wet chemistry (Browning, 1967), were analyzed in the py-MBMS systematically after every 44 runs with our samples. Each subsample was pyrolyzed for 30 s at 500°C. Vapors from pyrolysis were rapidly expanded under vacuum through a 0.012 in crystal orifice, creating a molecular beam that was directed to the Extrel™ Model TQMS C50 mass spectrometer, yielding a spectrum ranging from 30 to 450 mass-to-charge (m/z) ratio. Data were normalized against differences in total ions of each pyrolyzed sample. Peaks associated with each wood chemistry component, based on previous literature (Evans & Milne, 1987), were summed to produce a single estimate, as indicated in Table 2 and described elsewhere (Sykes et al., 2008).

Table 2. Estimates of clonal repeatability and average trait value for each of eight wood chemistry phenotypes estimated in two nitrogen treatments using a pyrolyzer coupled to a molecular beam mass spectrometer (py-MBMS)
TraitAcronym m/z peaks sumClonal repeatabilityNitrogen deficiency (0 mM)High nitrogen (25 mM)Prob > | t |
H 2 ± SEMean ± SEMean ± SE
  1. Standard error (SE) is depicted for each estimate. The last column contains P values of a t-test assessing the effect of nitrogen treatment in each phenotype.

Five-carbon hemicellulose sugarC557 + 73 + 85 + 96 + 1140.163 ± 0.03925.678 ± 0.03927.019 ± 0.0780
Six-carbon cellulose sugarC657 + 60 + 73 + 98 + 126 + 1440.174 ± 0.0432.888 ± 0.06535.540 ± 0.1370
Ratio C6/C5C6/C5 0.170 ± 0.0391.280 ± 0.0011.311 ± 0.0020
Syringyl lignin monomerS-lignin154 + 167 + 168 + 182 + 194 + 208 + 2100.338 ± 0.03813.03 ± 0.05410.153 ± 0.0483.81E-280
Guaiacyl lignin monomerG-lignin124 + 137 + 138 + 150 + 164 + 1780.150 ± 0.03913.156 ± 0.02911.160 ± 0.0399.44E-294
Ratio S-lignin/G-ligninS/G 0.378 ± 0.0310.993 ± 0.0040.917 ± 0.0048.24E-41
Total ligninLigninG-lignin + S-lignin + 120 + 152 + 180 + 1810.234 ± 0.03921.690 ± 0.0617.382 ± 0.0630
Ratio C6/ligninRatioCL 0.182 ± 0.041.543 ± 0.0072.119 ± 0.0150

Statistical analysis of phenotypic data

Before analyses of each phenotypic trait, the data were evaluated for the presence of outliers and removal or correction of measurements with recording errors. PROC INSIGHT (SAS Institute Inc. 9.1® 2004, Cary, NC, USA) was used to check the distribution of residuals. A square root transformation was applied to the number of sylleptic branches because of the nonnormal distribution of residuals.

Univariate analyses of each trait were performed using the SAS® System for Mixed Models (Littell et al., 1996) to separately account for the different sources of variation from our experiment. This analysis of variance allows an accurate estimation of the desired effects (i.e. clone and nitrogen treatment) by controlling the influence of undesired sources of variation (i.e. replicates, bench, row and column within replicate). The mixed model utilized was:

yijklmno = µ + ri + Tk + rtik + bj(i) + cl + rcil +tckl + pm(i) +qn(i)
       + eijklmno
(Eqn 1)

(yijklmno, response of the oth ramet of the lth clone in the kth treatment of the jth bench within the ith replication; µ, population mean; ri, random effect of replication, ∼ normally and independently distributed (NID) (inline image); Tk, fixed effect of the nitrogen treatment; rtik, effect of replication by treatment interaction, ∼ NID (inline image); bj(i), random effect of bench (incomplete block) within replication, ∼ NID (inline image); cl, random effect of clone, ∼ NID (inline image); rcil, random effect of replication by clone interaction, ∼ NID (inline image); tckl, random effect of treatment by clone interaction, ∼ NID (inline image); pm(i), random effect of row within replication, ∼ NID (inline image); qn(i), random effect of column within replication, ∼ NID (inline image); eijklmno, random error effect within the experiment, ∼ NID (0, R) and inline image.

The genetic analyses were modified to include different error variances for each replication. inline image and inline image are the variance components corresponding to replication, replication by treatment, bench within replication, clone, replication by clone, treatment by clone, row within replication, column within replication and residual effects of each replication, respectively.

Variance components and genetic parameters were estimated by restricted maximum likelihood, using ASReml (Gilmour et al., 2002). Least-square means used in the QTL analysis were calculated by including both clone effect and its interaction with treatment as fixed effects in the model.

Clonal repeatability for each trait in the univariate analysis was calculated by ASReml with the estimates of variance components as follows: inline image with inline image as previously defined.

To estimate phenotypic and genetic correlations, pair-wise traits were analyzed with the following bivariate model equation:

y i = Xibi + Z1ai + Z2di + Z3hi + Z4ki + Z5li + Z6mi + Z7ni + ei (Eqn 2)

where yi is the vector of observations for traits 1 (t1) and 2 (t2), bi is the vector of fixed effects (i.e. means and treatments) associated with the incidence matrix Xi, ai is the vector of random clonal effects associated with the incidence matrix Z1 and ai∼MVN (0, G), with

inline image

d i, hi, ki, li, mi, and ni are random effects corresponding to replication by treatment, bench within replication, replication by clone, treatment by clone, row within replication and column within replication associated with known incidence matrices Z2, Z3, Z4, Z5, Z6 and Z7, respectively. All random effects have zero mean and variance structure similar to G. ei is the vector of random error effects and ei∼MVN (0, R), with

inline image


inline image

where G and R are covariance–variance matrices corresponding to vectors ai and ei, respectively, and the R structure was associated with error effects of each replication. inline image are clonal variance components for trait 1, trait 2 and the covariance component between trait 1 and trait 2. inline image are error variance components and the covariance component, respectively, for trait 1 and trait 2 in replication 1. Replications 2 and 3 have similar descriptions of error variance and covariance components.

All traits were analyzed according to the full model (Eqn 2), but effects with no variation were dropped to fit the final model.

After estimating the variance and covariance components, genetic and phenotypic correlations were calculated for pairs of traits according to the following equation:

inline image

(inline image, genetic or phenotypic correlation between trait 1 and trait 2; inline image, clonal genetic covariance or phenotypic covariance between two traits; inline image, clonal genetic or phenotypic variances of trait 1 and trait 2.)

Genetic linkage map and QTL analysis

From a dense microsatellite and microarray-based genetic linkage map (Drost et al., 2009), we selected 181 evenly distributed markers that segregate in the hybrid female parent. The selection favored microsatellite (simple sequence repeat, SSR) markers (163 selected) because they were genotyped in all 396 individuals of the progeny, whereas microarray-based markers were characterized in only 154. Eighteen microarray markers were included to expand map coverage towards the flanks of some linkage groups (LGs) and to fill a gap in LG6 where no SSR was genotyped. MapMaker 3.0 (Lander et al., 1987) was utilized for the construction of the linkage map with a maximum recombination frequency of 45 cM using Kosambi's map function (Kosambi, 1944) and a minimum logarithm of the odds ratio (LOD) of three. Consistent with the haploid number of chromosomes of P. trichocarpa and P. deltoides, our map has 19 LGs, with a total length of 2889 cM and an average density of one marker every 16 cM. BLAST alignment of SSR primer and microarray probe sequences to the poplar genome sequence (JGI v.1.1) demonstrates that our map spans at least 85% of the assembled genome.

The linkage map was used for the identification of QTLs with composite interval mapping (Zeng, 1993) implemented in Windows QTL Cartographer v.2.5 (Wang et al., 2007). We utilized the standard model 6 with default settings for the selection of cofactor markers to account for background variance not associated with the locus being tested. Least-square means estimates for each individual in each nitrogen treatment were used in the analysis. For each phenotypic trait, QTL analyses for 1000 permutations were performed, establishing a null distribution of genome-wide maximum LOD scores, where the 95th percentile was defined as the significance threshold (Churchill & Doerge, 1994).


Genetic control of phenotypic traits and the effect of nitrogen treatments

For brevity, all growth and allocation traits depicted in Table 1 are referred to as phenotypic traits, whereas wood composition and partitioning traits depicted in Table 2 are called wood chemistry traits. Clonal repeatability (or within-family broad-sense heritability) was estimated for each trait across the entire dataset (i.e. all replicates and both nitrogen treatments). Estimates of clonal repeatability were moderate for all 12 phenotypic traits, ranging from 0.25 for sylleptic branch biomass to 0.38 for sylleptic branch count (Table 1). The effect of nitrogen fertilization was highly significant for all traits, increasing diameter, height, number of internodes, number of sylleptic branches and above-ground biomass, whereas root biomass was significantly decreased when compared with the nitrogen deficiency treatment (Table 1, Fig. 1). Carbon allocation favored shoot over root biomass, when nitrogen fertilization was applied. The average above- to below-ground biomass ratio increased significantly from 5.07 to 11.90 under 0 to 25 mM NH4NO3 treatments.

Table 1. Estimates of clonal repeatability and average trait value for each of 12 phenotypes measured in two nitrogen treatments
TraitAcronymClonal repeatabilityNitrogen deficiency (0 mM)High nitrogen (25 mM)Prob > |t|
H 2 ± SEMean ± SEMean ± SE
  1. Standard error (SE) is depicted for each estimate. The last column contains P values of a t-test assessing the effect of nitrogen treatment in each phenotype.

Diameter (cm)diam0.328 ± 0.0254.41 ± 0.0315.087 ± 0.0430
Height (cm)ht0.325 ± 0.02767.751 ± 0.71278.9 ± 0.7390
Height live crown (cm)htlc0.338 ± 0.02661.517 ± 0.73574.958 ± 0.7490
Internodes countint0.344 ± 0.03329.163 ± 0.219 34.21 ± 0.241 0
Leaf biomass (g)tleaf0.305 ± 0.0274.86 ± 0.0926.734 ± 0.120
Stem biomass (g)tstem0.298 ± 0.0263.814 ± 0.084.469 ± 0.0961.69E-07
Sylleptic branch countsylf0.382 ± 0.0281.819 ± 0.084.097 ± 0.171 0
Sylleptic branch biomass (g)sylwt0.249 ± 0.040.8 ± 0.0572.053 ± 0.0970
Above-ground biomass (g)abbio0.314 ± 0.0259.064 ± 0.18612.462 ± 0.2440
Root biomass (g)rootwt0.303 ± 0.0252.204 ± 0.0571.265 ± 0.0334.14E-44
Ratio abbio/rootwtrabbe0.312 ± 0.0295.073 ± 0.076 11.896 ± 0.137 0
Total biomass (g)tbio0.31 ± 0.02611.267 ± 0.24113.742 ± 0.2782.26E-11
Figure 1.

Effect of nitrogen fertilization on shoot biomass accumulation on one genotype of the Populus pseudo-backcross population. After establishment, the plant on the left was grown under nitrogen deficiency (0 mM of NH4NO3), whereas the plant on the right was treated with 25 mM NH4NO3. A height scale (in cm) is depicted in each photograph.

Estimates of clonal repeatability for wood chemistry traits ranged from 0.15 for G-lignin to 0.38 for partitioning between S- and G-lignin (S/G ratio). Most of the clonal repeatability estimates for wood chemistry traits were lower than those calculated for phenotypic traits, except for S-lignin (0.34) and the S/G ratio (0.38). The effect of nitrogen fertilization was highly significant for wood chemistry traits, increasing cellulose and hemicellulose and decreasing lignin when compared with plants grown under limiting nitrogen. In response to nitrogen, the increase in carbohydrates was greater for cellulose compared with hemicellulose, as observed in the C6/C5 ratio. The decrease in mean lignin content in response to nitrogen was 7% greater for S-lignin than G-lignin, as observed in the S/G ratio (Table 2).

Analysis of genetic and phenotypic correlations among traits

Pair-wise phenotypic and genotypic correlations were estimated for all traits across the entire experiment, that is combining data from the two nitrogen treatments (Table 3; Table S1, see Supporting Information). Only genetic correlations are depicted in parentheses when a specific pair-wise relationship is described in the text. Most of the morphological and biomass traits were positively correlated phenotypically and genetically with each other – for example, height was strongly and positively correlated with diameter (0.78), stem biomass (0.91), leaf biomass (0.69), above-ground biomass (0.79) and total biomass (0.78). All biomass traits were also positively correlated with each other. For example, root biomass was strongly genetically correlated with leaf (0.86), stem (0.81), above-ground (0.86) and total (0.90) biomass. As expected, plants growing more rapidly tended to accumulate biomass at higher rates in all vegetative tissues. The ratio between above- and below-ground biomass was the only trait that was not strongly correlated with any other phenotypic trait, except for its autocorrelation with root biomass (−0.63).

Table 3. Pair-wise estimates of phenotypic (below diagonal) and genotypic (above diagonal) correlations between traits (see Tables 1, 2 for acronym description)
  1. Bold type shows pair-wise traits with correlation > | 0.60 |.

htX 0.78 0.22 0.70 0.96 0.42 0.69 0.91 0.78 0.80 −0.13 0.67 0.260.360.00−0.130.06−0.500.270.20
diam 0.75 X0.12 0.76 0.73 0.35 0.74 0.85 0.80 0.79 −0.260.500.110.18−0.07−0.030.10−0.340.220.13
sylwt0.150.19X0.430.400.800.350.260.460.45−0.030.380.530.46 0.63 −0.60−0.50−0.45−0.250.56
rootwt 0.70 0.74 0.49X 0.74 0.43 0.86 0.81 0.90 0.86 −0.63 0.590.550.600.36−0.37−0.17−0.610.070.53
htlc 0.94 0.70 0.25 0.72 X0.56 0.73 0.89 0.83 0.84 −0.14 0.73 0.350.430.13−0.23−0.04−0.510.180.33
sylf0.260.31 0.70 0.420.39X0.340.400.500.500.040.580.460.420.46−0.50−0.41−0.38−0.270.48
tleaf 0.74 0.75 0.28 0.80 0.75 0.35X 0.90 0.96 0.96 −0.220.570.500.530.33−0.45−0.29−0.56−0.030.52
tstem 0.86 0.81 0.23 0.76 0.83 0.32 0.90 X 0.93 0.94 −0.19 0.67 0.460.520.25−0.27−0.08−0.530.140.45
tbio 0.76 0.77 0.44 0.84 0.78 0.46 0.95 0.92 X 1.00 −0.24 0.67 0.58 0.62 0.38−0.48−0.29−0.61−0.610.58
abbio 0.77 0.77 0.43 0.81 0.79 0.46 0.96 0.93 0.99 X−0.15 0.67 0.530.570.36−0.47−0.29−0.59−0.590.54
int 0.70 0.540.32 0.61 0.71 0.44 0.62 0.64 0.65 0.66 −0.07X0.440.480.30−0.43−0.30−0.46−0.110.45
C60.230.160.310.430.310.420.420.310.430.43−0.200.37X 0.95 0.89 −1.00 −0.85 −0.72 −0.53 0.99
C50.300.220.320.470.350.390.440.360.450.45−0.240.39 0.97 X 0.75 −0.87 −0.68 −0.77 −0.35 0.92
C6/C50.−0.110.24 0.81 0.59X −0.95 −0.93 −0.48 −0.81 0.97
Lignin−0.23−0.27−0.45−0.49−0.33−0.46−0.47−0.37−0.48−0.480.19−0.42 −0.78 −0.71 −0.73 X 0.93 0.57 0.69 −0.98
S-lignin−0.02−0.07−0.38−0.27−0.13−0.41−0.30−0.16−0.29−0.300.07−0.27 −0.63 −0.51 −0.72 0.90 X0.23 0.91 −0.92
G-lignin−0.46−0.45−0.29 −0.61 −0.49−0.33−0.55−0.51−0.55−0.550.27−0.46 −0.64 −0.68 −0.40 0.72 0.34X−0.27 −0.70
S/G0.280.21−−−0.55−0.55−0.080.01−0.27−0.11−0.510.470.80−0.19X −0.63
RatioCL0.250.240.350.520.370.480.470.410.480.49−0.220.41 0.96 0.89 0.83 −0.89 −0.75 −0.68 −0.35X

Wood chemistry traits were genetically and phenotypically correlated among themselves (Table 3). Cellulose (C6) levels were strongly positively correlated with the amount of hemicelluloses (C5) (0.95). By contrast, both C5 and C6 amounts were strongly negatively correlated with total lignin (−0.87 and −0.99), G-lignin (−0.77 and −0.71) and S-lignin (−0.68 and −0.85) contents. The almost perfect negative genetic correlation between cellulose and lignin content (−0.99) decreased to −0.78 at the phenotypic level. The ratio between wood carbohydrates (C6/C5) was negatively correlated with total lignin (−0.95) and S/G ratio (−0.81). This suggests that, when more carbon is partitioned into cellulose relative to hemicellulose, the total amount of lignin decreases and a disproportionate reduction in syringyl relative to guaiacyl monomers is observed. Lignin with a higher proportion of syringyl monomers contains less carbon–carbon cross-links and therefore is more easily extracted than guaiacyl-rich lignin (Chang & Sarkanen, 1973). Thus, in this population, carbon partitioning into carbohydrates (C6 > C5) is associated with higher proportions of a less extractable type of lignin monomer. Variation in these traits was very high in both nitrogen treatments. The proportions of C5 and C6 carbohydrates were in the ranges 21–34% and 26–48%, respectively. For lignin syringyl monomers, the proportion ranged from 7 to 19%.

Biomass and wood chemistry were also strongly correlated (Table 3). In general, wood carbohydrate contents (C5 and C6) were positively correlated with the biomass of all vegetative tissues, whereas higher lignin was associated with lower levels of biomass accumulation. For example, total biomass was genetically correlated with C6 (0.57), C5 (0.62) and lignin (−0.48). The ratio between wood carbohydrates (C6/C5) was positively correlated with biomass traits, especially with sylleptic branches (0.63), but not correlated with height and diameter. This result indicates that, in the existing experimental conditions, carbon partitioned to cellulose may lead to higher biomass accumulation than carbon partitioned to hemicelluloses. However, it is difficult to establish any relationships of cause and effect, and it is also possible that the higher cellulose accumulation is the result of superior growth ability. The ratio between lignin monomers (S/G) was negatively correlated with above-ground biomass (−0.59), but not genetically correlated with below-ground (root) biomass (0.06). Therefore, the higher the above-ground biomass, the lower the extractability of lignin in woody tissues.

QTL mapping in each nitrogen treatment

A total of 63 QTLs was identified in the two nitrogen treatments (0 and 25 mM) using α = 0.05 as threshold defined with 1000 permutations (Table 4, Fig. 2). S-Lignin was the only trait for which no QTL was mapped in either nitrogen treatment. Because of the trait's heavy skewed distribution, QTL for sylleptic branch count (sylf) reported here was analyzed elsewhere with a Poisson statistical model (Ma et al., 2008). Under high nitrogen, 31 QTLs were identified, 21 for 11 phenotypic traits and 10 for seven wood chemistry traits. Under the nitrogen deficiency treatment, 32 QTLs were detected, 19 for 10 phenotypic and 13 for five wood chemistry traits. The origin of the QTL alleles positively affecting the traits was balanced, with 34 coming from P. trichocarpa and 29 from P. deltoides grand-parents of the pedigree. Under high nitrogen, the most significant QTL (LOD = 6.93) was detected on LG14 for biomass allocation between above and below ground. In the nitrogen-limiting treatment, the most significant QTL (LOD = 6.33) was observed on LG1 for the number of internodes (Table 4). The percentage of the phenotypic variance explained by the QTLs ranged from 3.58% for the C6/C5 ratio to 11.36% for the same trait under high nitrogen treatment. Under high nitrogen treatment, LG13 had the most QTLs with 12, whereas, under nitrogen deficiency, LG1 had the most QTLs mapped, with loci controlling 10 different traits. On these two LGs, large fractions of the QTLs were detected in the same regions (Fig. 2). Five of the 19 LGs had no QTLs mapped in our study.

Table 4. Linkage group (LG) and flanking marker localization for the 63 quantitative trait loci (QTLs) identified under the two nitrogen treatments (H, high; D, deficiency)
QTLNitrogen treatmentTrait acronymLGFlanking markersOrigin of positive alleleLOD peakPhenotypic variance explained
Marker 1Marker 2
1Habbio12S170P2737 P. deltoides 3.314.36%
2Habbio13P2847G2218 P. trichocarpa 3.057.04%
3Hdiam 3G3465P2481 P. deltoides 3.835.32%
4Hdiam12S170P2737 P. deltoides 4.205.25%
5Hdiam13P2847G2218 P. trichocarpa 3.187.03%
6Hhtlc 2O461P422 P. deltoides 3.744.85%
7Hint 1G3205G834 P. trichocarpa 4.727.26%
8Hint 6G3600O50 P. trichocarpa 5.666.69%
9Hint13P2847G2218 P. trichocarpa 4.719.20%
10*Hsylf10G2431P2855 P. trichocarpa 3.017.20%
11Hsylwt 1G124G1568 P. trichocarpa 4.985.70%
12Hsylwt 1P2731G1782 P. trichocarpa 4.304.90%
13Hsylwt17G641G3580 P. trichocarpa 4.986.18%
14Hrabbe 3G1470G3365 P. deltoides 4.975.55%
15Hrabbe14G1866G1177 P. deltoides 6.939.75%
16Hrootwt13P2847G2218 P. trichocarpa 3.117.35%
17Htbio12S170P2737 P. deltoides 3.424.51%
18Htbio13P2847G2218 P. trichocarpa 3.187.52%
19Htleaf12S170P2737 P. deltoides 3.895.11%
20Htleaf13P2847G2218 P. trichocarpa 3.869.52%
21Htstem12S170G2682 P. deltoides 3.474.60%
22HC513P14P2658 P. trichocarpa 3.175.03%
23HC613G2577G2218 P. trichocarpa 4.7811.24%
24HC6/C5 1P575G1782 P. trichocarpa 2.973.58%
25HC6/C513G2577G2218 P. trichocarpa 5.2311.36%
26HG-lignin13P2847G2218 P. deltoides 3.246.78%
27HLignin13G2577G2218 P. deltoides 4.519.61%
28HRatioCL13G2577G2218 P. trichocarpa 4.6910.17%
29HS/G 1P2731G1782 P. deltoides 3.113.88%
30HS/G10G1946G938 P. deltoides 3.634.27%
31HS/G15G4047P2585 P. trichocarpa 3.395.03%
32Dabbio 1P575P2786b P. trichocarpa 3.526.10%
33Ddiam 3G3465P2481 P. deltoides 3.364.61%
34Ddiam 8P2598G3578 P. trichocarpa 3.294.42%
35Dht 2G876P422 P. deltoides 3.214.50%
36Dhtlc 1G3205G834 P. trichocarpa 4.706.49%
37Dhtlc 2G876P422 P. deltoides 3.023.94%
38Dint 1G3205G834 P. trichocarpa 6.339.91%
39Dint 6G2281O50 P. trichocarpa 3.443.79%
40Dsylwt 5G1063G3627 P. trichocarpa 3.585.10%
41Dsylwt15G4047P2585 P. deltoides 3.304.62%
42Drabbe 1G825P575 P. trichocarpa 3.304.24%
43Drabbe 3G1470G3365 P. deltoides 3.934.38%
44Drabbe 9G3457G588 P. trichocarpa 3.013.64%
45Drabbe14G1866G1177 P. deltoides 2.953.95%
46Drootwt 9G3457G588 P. deltoides 3.334.34%
47Dtleaf 1P575P2786b P. trichocarpa 3.376.98%
48Dtleaf 3G3465G1688 P. deltoides 3.114.90%
49Dtleaf13P2658G2577 P. trichocarpa 3.074.42%
50Dtstem 1P575P2786b P. trichocarpa 3.103.97%
51DC5 1G3205G834 P. trichocarpa 5.848.06%
52DC5 6S18P2221 P. deltoides 4.098.79%
53DC518G1089G1244 P. trichocarpa 3.636.85%
54DC6 1G3205G834 P. trichocarpa 5.807.96%
55DC6 6S613G2126 P. deltoides 2.943.85%
56DC6 6S18E538 P. deltoides 3.287.38%
57DC6/C5 6S613G2126 P. deltoides 3.584.73%
58DG-lignin 1G3205G834 P. deltoides 3.904.95%
59DG-lignin 6E918G139 P. trichocarpa 3.586.97%
60DG-lignin13G2577G2218 P. deltoides 4.017.51%
61DG-lignin17G641P648 P. deltoides 3.626.24%
62DRatioCL 1G3205G834 P. trichocarpa 4.115.30%
63DRatioCL 6S613G2126 P. deltoides 3.354.42%
*QTL for sylf was mapped elsewhere (Ma et al., 2008).
Also depicted for each QTL are the grand-parent origin of the allele with positive effect on the trait, logarithm of the odds ratio (LOD) score and percentage of the phenotypic variance explained.
Figure 2.

Map location of all 63 quantitative trait loci (QTLs) identified in our experiment. QTLs mapped under nitrogen deficiency treatment are represented on the left of each linkage group, whereas QTLs mapped under high nitrogen are on the right. Colored bars indicate map regions in which the logarithm of the odds ratio (LOD) profile is above statistical threshold, and black arrows mark the position of the LOD peak for each QTL. QTLs spanning < 10 cM above threshold are represented by 10 cM bars. Microsatellite markers are identified in black and array-based markers in red. Linkage groups in which no QTLs were mapped are not shown. *QTL for sylf was mapped elsewhere (Ma et al., 2008). (See Tables 1, 2 for trait acronyms.)

Phenotypic plasticity in response to nitrogen treatments

Only six QTLs for four phenotypic traits were mapped in the same genomic region under both nitrogen treatments (compare QTLs on both sides of each LG in Fig. 2). For the number of internodes, two QTLs (LG1 and LG6) were co-localized in both treatments. Similarly, for the ratio between above- and below-ground biomass, QTLs were detected on LG3 and LG14. Also mapped consistently under both nitrogen conditions were QTLs for the height of live crown (LG2) and diameter (LG3). The grand-parent origin of the positive allele was the same under both nitrogen treatments for all QTLs, indicating that the same genetic elements are likely to control the traits independent of nitrogen availability. All 51 other QTLs identified in our pedigree were treatment specific, suggesting that, in Populus, a large fraction of the interspecific variance in growth, carbon allocation and carbon partitioning traits is highly responsive to the level of nitrogen available in the environment. For wood chemistry traits, none of the QTLs was mapped under both nitrogen treatments.

Co-localization of QTLs for different traits

In this interspecific Populus pedigree, we found eight genomic regions that controlled part of the variation for at least two different traits under the same nitrogen treatment. For example, under high nitrogen, there were two regions with QTLs co-localized for several traits – one on LG12 between markers S170 and G2682 and another on LG13 between P2847 and G2218 (Fig. 2). The traits with QTLs co-localized on LG12 were leaf, stem, above-ground and total biomass, as well as diameter. These traits were all positively, highly genetically correlated with each other and, as such, the QTL alleles had the same direction of effect in all traits – that is the P. deltoides QTL allele positively affected all traits. On LG13, there was a hot spot of co-localized QTLs for both phenotypic and wood chemistry traits. The traits with co-localized QTLs on LG13 were above-ground, below-ground, leaf and total biomass, diameter, number of internodes, cellulose, ratio between cellulose and hemicelluloses, lignin and ratio between cellulose and lignin. Consistent with the direction of correlations, the P. trichocarpa allele positively affected the total biomass (above- and below-ground), wood cellulose and hemicellulose, but negatively affected wood lignin content. Another co-localization of QTLs under high nitrogen occurred on LG1 for sylleptic branch biomass and S/G ratio.

Under nitrogen deficiency, five regions had co-localized QTLs for different traits (Fig. 2). Two of the regions occurred on LG1. The region between markers G3205 and G834 contained QTLs for height of the live crown, number of internodes, wood cellulose, hemicelluloses, guaiacyl lignin and ratio between cellulose and lignin, and the region between P575 and P2786b contained QTLs for above-ground, leaf and stem biomass. As in the hot spot on LG13, the QTL alleles had opposite effects for wood carbohydrates and lignin, but the same direction of effect for height, number of internodes, biomass traits and wood carbohydrates. Between markers S613 and G2126 on LG6, QTLs were identified for the following wood chemistry traits: cellulose, guaiacyl lignin, ratio between cellulose and hemicellulose, and ratio between cellulose and lignin. Also on LG6, between markers S18 and E538, there were co-localized QTLs for wood cellulose and hemicellulose content. Co-localized QTLs for root weight and ratio between above- and below-ground biomass were mapped between markers G2541 and G588 of LG9.


In this article, we report the widespread effects of nitrogen fertilization on the genetic regulation of growth and wood chemistry traits in the progeny of an interspecific pseudo-backcross of P. deltoides and P. trichocarpa. Nitrogen supply affected all growth and wood chemistry traits, confirming the extensive phenotypic plasticity of poplar in response to this essential nutrient. Studies have shown that nitrogen positively influences poplar tree height, diameter, number of leaves, number of sylleptic branches and shoot biomass (Pregitzer et al., 1990; Ibrahim et al., 1997; Kleiner et al., 1998; Cooke et al., 2005). However, contrary to our results, the same studies reported an increase in root biomass in response to nitrogen supply. Although statistical significance was either not observed (Cooke et al., 2005) or not tested (Pregitzer et al., 1990; Ibrahim et al., 1997; Kleiner et al., 1998), this disparity might reflect differences in experimental conditions. For example, in our trial, nutrients were supplied with a flood irrigation system in ebb-and-flow benches, whereas, in previous work, plants were fertilized with irrigation on the top of the pots. Despite differences in total root biomass, inarguably nitrogen fertilization caused significant changes in biomass allocation, decreasing root/shoot biomass ratio in poplar. Wood chemistry also responded to different nitrogen availability. Confirming previous studies with poplar trees (Pitre et al., 2007a,b), we observed that nitrogen fertilization significantly increased wood cellulose content and decreased lignin content and its extractability, measured as the S/G ratio. We also found that nitrogen supply increased hemicelluloses at a significantly lower rate (5%) when compared with the increase in cellulose sugars (8%).

The extensive phenotypic plasticity in response to nitrogen availability observed in Populus may be triggered by changes in gene regulation (Sultan, 2000). Supporting this hypothesis in poplar, nitrogen significantly changed the transcript abundance of 52 cDNA clones identified by differential display (Cooke et al., 2003). Consistent with a nitrogen-mediated decrease in wood lignin, two of the differentially expressed cDNA clones were from putative lignin biosynthesis genes, and were down-regulated in response to nitrogen. Similarly, studies that assessed the effects of nitrogen fertilization on the whole transcriptome of Arabidopsis also demonstrated that nitrogen represses most of the genes from the phenylpropanoid pathway (Scheible et al., 2004; Gutierrez et al., 2008). Some genes involved in cell wall growth and modification, including expansins and xyloglucan:xyloglucosyl transferase, were induced in response to nitrogen. Nitrogen induction was also detected in most genes involved with photosynthesis, the Calvin cycle and photorespiration (Scheible et al., 2004). Changes in gene expression in response to nitrogen fertilization may be the result of regulatory signals that balance carbon and nitrogen metabolism (Koch, 1997; Palenchar et al., 2004; Weigelt et al., 2008). The assimilation of nitrogen into amino acids depends on the availability of carbon skeletons, and carbon is stored in starch if nitrogen is deprived (Calenge et al., 2006). When nitrogen is luxuriant (relatively low carbon to nitrogen ratio), plant metabolism shifts resources towards the absorption of carbon through photosynthesis and develops more shoot than root. Conversely, when nitrogen is limited (high carbon to nitrogen ratio), metabolism favors the allocation of biomass to root over shoot (Koch, 1997).

Consistent with the widespread phenotypic and gene expression plasticity of plants in response to nitrogen, our study detected extensive changes in the genetic control of carbon allocation and partitioning with nitrogen supply. Most of the QTLs identified in our study (51 of 63) are nitrogen treatment specific. In other plant species, significant QTL by nitrogen interactions have been reported for disease resistance and root traits in rice (Talukder et al., 2005; MacMillan et al., 2006), carbohydrate content and plant biomass in Arabidopsis (Rauh et al., 2002; Calenge et al., 2006), and grain yield and quality in wheat (Laperche et al., 2007). Significant QTL by nitrogen treatment interaction indicates that fast- and slow-growing genotypes are not the same under limiting and luxuriant nitrogen conditions. Spearman correlation of phenotypic values under the two nitrogen treatments indicates extensive changes in rank among genotypes (Table S2, see Supporting Information). The even lower Spearman correlations between nitrogen treatments for wood chemistry traits indicate that most of these traits are more responsive to nitrogen supply than are biomass and growth traits. This agrees with the lower clonal repeatability observed for most wood components when compared with phenotypic traits (Tables 1, 2), which was not observed when clonal repeatability was estimated for each nitrogen treatment separately (Table S3, see Supporting Information). Supporting the idea that wood chemistry traits are more reactive to nitrogen availability, no QTLs for wood traits were consistently mapped in both nitrogen treatments, whereas six QTLs for phenotypic traits were mapped independently of nitrogen level.

Our results support the hypothesis that growth co-varies with wood chemistry traits in trees. As demonstrated previously with mutants of Populus (Hu et al., 1999) and Pinus (Wu et al., 1999; Yu et al., 2006), and in a segregating pedigree of Eucalyptus (Kirst et al., 2004), we observed that fast-growing poplar trees have higher cellulose and lower lignin content when compared with slow-growing trees. To genetically dissect the association between growth and wood quality traits, we searched for common pleiotropic loci that regulate variation in both trait groups. We identified two common QTLs for growth and wood chemistry. One QTL cluster detected in our study occurs on LG13, between markers G2577 and G2218, and regulates growth, biomass and wood composition traits. Specifically, trees that inherited the P. trichocarpa allele at this locus grew faster, contained more cellulose and less lignin when compared with those that inherited the P. deltoides allele. Genomic regions in the vicinity of this LG13 QTL cluster control the abundance of metabolites derived from the phenylpropanoid pathway in an independent pedigree of Populus (Morreel et al., 2006). The other QTL cluster coordinately controls the height of the live crown and wood composition under nitrogen deficiency and is located on LG1, between markers G3205 and G834. In this same region of LG1, QTLs have been mapped previously for height, stem circumference, stem volume, root density and root growth in family 331, which shares a common grand-parent (clone 93-968) with family 52–124 analyzed here (Rae et al., 2007, 2008).

Other QTLs detected in our study were also detected in family 331. For example, both QTLs for the height of the live crown (LG1 and LG2), mapped under limiting nitrogen treatment, co-localized with the QTL for stem height identified by Rae et al. (2008). The QTL identified for sylleptic branch count under high nitrogen (LG10) from our experiment (Ma et al., 2008) was also mapped in family 331 (Rae et al., 2008). Two QTLs for above-ground biomass from our experiment (LG1 under limiting nitrogen and LG12 under high nitrogen) co-localized with QTLs for stem height and diameter in family 331 growing under elevated (LG1) or ambient (LG12) CO2 (Rae et al., 2007). Two of our QTLs for total leaf biomass (LG3 under limiting nitrogen and LG12 under luxuriant nitrogen) mapped in the same region of previously identified QTLs controlling leaf number (Rae et al., 2006).

A number of interesting a priori candidate genes are physically positioned within the marker intervals flanking potential pleiotropic quantitative loci on LG1 and LG13. An immediately apparent candidate gene underlying the QTL cluster on LG13 is cinnamate-4-hydroxylase, encoding an enzyme that regulates the conversion of cinnamate to 4-coumarate in the phenylpropanoid biosynthesis pathway, and a transcript for which we previously identified a strong correlation with growth in the Eucalyptus segregating population (Kirst et al., 2004). Similarly, the QTL cluster for growth and wood composition on LG1 harbors a cluster of three tandemly duplicated genes encoding putative cinnamoyl-CoA reductases, which regulate the conversion of feruloyl-CoA to coniferaldeyhyde in the phenylpropanoid pathway. Additional components (chalcone synthase, 4-coumarate:CoA ligase) and potential regulators [AtMYB12 homolog (Mehrtens et al., 2005) and other uncharacterized R2R3-MYB-type transcription factors] of phenylpropanoid metabolism are also conceivable candidate genes within these intervals. Furthermore, several C5/C6 metabolic enzymes, including xyloglucan fucosyltransferase, glucose-6-phosphate-1-dehydrogenase, xyloglucan endotransglycosylase and cellulose synthase, are encoded by genes within these intervals. Finally, numerous genes of unknown function, or with no homology to known Arabidopsis or Oryza coding sequences, are prevalent among genes within the intervals, leaving open the possibility that some of these uncharacterized elements might coordinately regulate growth and biomass composition in woody plants. The gene models within marker intervals that flank most of our QTLs, including those clusters on LG1 and LG13, are available as Supporting Information (Table S4). Gene models within some QTL intervals (eight of 29) could not be found, because at least one of the flanking markers either did not have a BLAST hit in the genome or hit an unmapped scaffold.

The number of candidate genes within these QTLs is certainly too large to be tested with functional analysis – our average QTL interval is 31 cM wide and contains approximately 426 genes, 45% of which have unknown function (Table S4). Although a reduction in the size of the candidate gene pool is necessary, the traditional positional cloning-based isolation of the regulatory locus would be challenging given the breadth of the QTL intervals, low heritability of quantitative traits and the typical low resolution of maps in forest species. Nevertheless, the availability of the P. trichocarpa genome sequence (Tuskan et al., 2006), allied to gene expression data that are being assayed with microarrays from three tissues (root, xylem and leaf) of our experiment, will form an excellent foundation to narrow down the candidate genes within QTLs. Integration of gene expression with genotypic and phenotypic data in a genetical genomics setting (Jansen & Nap, 2001) will uncover regions containing regulators of gene expression (eQTL) and of phenotypes. The identification of candidate genes for an eQTL/QTL hot spot can be achieved by elucidating genetic networks controlled by the region, as has been demonstrated in mammals (Mehrabian et al., 2005; Meng et al., 2007). Pinpointing genetic regulators linking tree growth to wood quality will ultimately enhance our ability to breed for and engineer genotypes improved for pulp, paper, below-ground carbon sequestration and cellulosic ethanol production.


We thank Dr Dudley A. Huber for advice on the statistical analysis of the data. This work was supported by the Department of Energy, Office of Science, Office of Biological and Environmental Research, Grant Award No. DE-FG02-05ER64114 (to M. K.).