Genetic basis of phenotypic correlations among growth traits in hybrid willow (Salix dasyclados×S. viminalis) grown under two water regimes


Author for correspondence: Martin Weih Tel: +46 18 672543 Fax: +46 18 673440 Email:


  • • Phenotypic correlations and quantitative trait loci (QTL) for important growth traits and a surrogate of intrinsic water-use efficiency (leaf δ13C) were analysed in a willow pedigree of 92 full-sibling clones grown under two water regimes. The major objective was to examine the genetic basis of the phenotypic correlations.
  • • Cuttings of Salix were glasshouse-grown during one growing season. The relative growth rate (RGR) and underlying traits were assessed. QTL analysis was conducted based on an available linkage map for Salix.
  • • Leaf area productivity and leaf nitrogen productivity were more important in determining RGR than leaf area ratio and specific leaf area. However, phenotypic correlations among growth traits partly varied between the two environments. QTL were detected for most growth traits, among them many common QTL for different traits. The QTL pattern reflected the phenotypic correlation pattern. None of the QTL for the complex traits was consistent across the different environments.
  • • The results demonstrate a genetic basis for phenotypic correlations among growth traits in Salix, and provide evidence for the existence of ‘master switches’ regulating some of the traits.


Growth and nitrogen-use efficiency (NUE) of plants reflect two major metabolic pathways, which are interrelated. The availability of water and water-use efficiency (WUE) are important determinants of growth, particularly in inherently fast-growing plants, and water economy greatly influences N economy (Wright et al., 2003). We therefore expect that growth, water and N economy, and the relationships among them vary with the availability of water to plants.

Classical growth analysis focuses on the assessment of various quantitative traits related to growth and N economy and the interrelationships among them. It seeks to analyse, at the phenotypic level, the mechanisms underlying variations in growth and biomass (or N) allocation of plants in response to different environmental conditions (Hunt, 1982). For example, the relative growth rate (RGR, g g−1 wk−1) is determined by the biomass allocation to leaves (leaf area ratio, LAR, m2 g−1) or within leaves (specific leaf area, SLA, m2 g−1) on the one hand; and leaf area productivity (LAP, g m−2 wk−1) on the other (Lambers et al., 1990). The relationships among quantitative traits have been investigated in numerous studies performed both among and within species (across genotypes). Many investigations indicate that variations in SLA often explain a large portion of the interspecific variation in RGR (Lambers et al., 1990; Cornelissen et al., 1996; Poorter et al., 2005), but the relationship is strongly influenced by environmental conditions (e.g. light and temperature: Poorter & Van der Werf, 1998; Loveys et al., 2002; Shipley, 2002).

In contrast to the phenotypic pattern, the genetic background of quantitative traits related to growth analysis, water economy and N economy has not been investigated extensively, and crossing studies focusing on both phenotype and genetics of these quantitative traits are few, especially regarding trees. Quantitative traits are under polygenic control and, during evolution, appropriate sets of genes responsible for quantitative traits could have been composed through the accumulation of numerous small adaptations. Alternatively, many quantitative traits could have been affected simultaneously by one or a few regulatory genes, which may suggest the existence of genetic links (trade-offs) between traits (Stearns, 1992; Chapin et al., 1993). Regulatory genes, also called ‘master switches’, have been suggested to control various metabolic processes in plants (Jaglo-Ottosen et al., 1998; Spoel et al., 2003) and also the suites of traits responsible for the fast or slow growth habit of plants adapted to different environments (Chapin et al., 1993).

Analysis of quantitative trait loci (QTL) enables the identification of specific regions on the genome that are responsible for variation in particular quantitative traits. The combination of growth and genetic (QTL) analysis performed on a large number of full-sibling relatives of plants is a promising tool to address the genetic basis of phenotypic correlations between growth traits and the existence of master switches. In a study performed on the herb Hordeum spontaneum exposed to a single type of environment, Poorter et al. (2005) found significant relationships between phenotypic traits such as SLA and RGR, but no evidence for the existence of master switches controlling a suite of quantitative traits. However, the fact that plants were exposed to only one kind of environment could possibly have caused the lack of evidence for master switches reported by Poorter et al. (2005). Our research focused on the genus Salix, of which some are fast-growing and used commercially for biomass production and phytoremediation (Verwijst, 2001). Breeding programmes have been established for Salix with the aim of increasing biomass production by phenotypic screening and recurrent selection (Gullberg, 1993), and linkage maps are available for Salix (Rönnberg-Wästljung, 2001; Tsarouhas et al., 2002; Rönnberg-Wästljung et al., 2003). Here we used the data from 92 full-sibling willow clones exposed to two contrasting environments, reported by Rönnberg-Wästljung et al. (2005), in which the aim was to identify (i) QTL for a number of growth traits; (ii) common QTL in the two environments differing in water availability; and (iii) clusters of genes involving various growth traits. Based partly on the results of Rönnberg-Wästljung et al. (2005) and partly on additional analyses, the objectives of this study were to examine phenotypic correlations among important (and mathematically related) growth traits; how water availability affects these correlations; and whether the pattern of phenotypic correlations between the traits reflects the pattern of overlapping or common QTL (clusters of genes).

Materials and Methods

Plant material, growth conditions and harvest procedure

A detailed description of the plant material and growth conditions is given by Rönnberg-Wästljung et al. (2005). A tetraploid hybrid F2 population originating from a cross between a female diploid Salix viminalis (L.) clone ‘Jorunn’ and a male hexaploid Salix dasyclados (Wimm.) clone ‘SW901290’ was used. The clone Jorunn is bred for high productivity, while SW901290 is a natural and relatively frost-tolerant clone collected in Russia. Two randomly chosen individuals of the tetraploid F1 progeny were crossed in 1998 to produce the F2 pedigree used here. The grandparents, parents and F2 pedigree were propagated vegetatively and cuttings of all clones (96 in total) were used for the experiment.

Cuttings were planted in 1-l plastic pots filled with a mixture of 20% clay and 80% peat in mid-March 2001. All plants were grown in a ventilated glasshouse in Uppsala, central Sweden, at slightly (approx. 2–6 K) above-ambient temperature (mean temperature 20.3°C during the treatment period) and ambient light conditions (>12 h photoperiod). The mean daily maximum photosynthetically active radiation (400–700 nm) during the treatment period was 1176 µmol m−2 s−1. The pots were arranged in 25 blocks, and individual clones (96 in each block) were positioned randomly within the blocks (experimental unit). During the first 5 wk all plants were watered daily to field capacity of the substrate (‘well watered’). After 5 wk growth an initial biomass harvest (five blocks, 480 plants) was carried out, and two experimental treatments (drought and well watered) were started. Ten randomly chosen blocks were assigned to the drought treatment and 10 blocks were maintained as the well watered treatment. In the drought treatment, the plants were kept just above the wilting stage and received as little water as possible without allowing for leaf necrosis. The amount of water given to the plants in the two irrigation treatments was recorded regularly in 10 measurement pots placed randomly within each block. At the end of the treatment period, the drought-treated plants had been supplied with approx. 55% of the water supplied to the well watered plants. Individual plants were supplied with a complete fertilizer solution at a rate of 6 mg N wk−1 during the first 3 wk and 6 mg N per 0.5 wk during the remaining treatment period, to account for the increased nutrient demand of larger plants. The final biomass harvest began after 4 wk of treatment and pairs of two blocks, one from each treatment, were harvested successively during a 3-wk period.

Harvested plants were separated into leaves, stems, cuttings and roots. Leaf area was determined on fresh leaves with an area meter (LI-3100, LiCor Inc., Lincoln, NE, USA) and SLA was calculated as mean for all leaves of individual plants. All plant parts were dried at 70°C for 48 h and weighed. Nitrogen content of leaves was determined and a surrogate of intrinsic WUE was estimated based on the carbon-isotope ratio of leaves (Farquhar et al., 1989). For this purpose, three fully developed leaves from the top of individual plants, three blocks from the well watered treatment and three blocks from the drought treatment, were sampled immediately before the start of the final harvest. The foliar samples were dried, milled and analysed with an automated N- and C-analysis mass spectrometer (ANCA-MS, Europe Scientific Ltd, Crewe, UK) by Waikato Stable Isotope Unit, Hamilton, New Zealand. The C-isotope ratio of leaf samples (δ13Csample, ‰) was determined as:

δ13Csample = (Rsample/RPDB − 1) × 1000(Eqn 1)

where Rsample and RPDB are the 13C/12C molar abundance ratios of the leaf material and the PeeDee Belemnite standard (Craig, 1957). The leaf δ13C signature has been shown to correlate strongly with the intrinsic WUE by means of the CO2 assimilation per unit of water used (Farquhar et al., 1989), and more negative values indicate lower WUE.

Growth analysis and statistical analysis of phenological traits

Plant growth and N economy were analysed on a growing-season basis using the methods of classical growth analysis (Hunt, 1982) and based on biomass and leaf area changes between two consecutive harvests. Thus possible differences in relative growth rate (RGR, g g−1 wk−1) between clones were related to differences in leaf area ratio (LAR, m2 g−1); leaf area productivity (LAP, g m−2 wk−1); specific leaf area (SLA, m2 g−1); leaf biomass fraction (LW/W, %); leaf N content (LN/LA, mmol N m−2); and leaf N productivity (LNP, g [mol N]−1 wk−1).

Growth differences of individual clones between the two water regimes were computed as treatment indices (ratios of mean values in the drought and well watered treatments, d/w) separately for each growth trait. Correlations between treatment indices of different growth traits were assumed to reflect the pattern of genotype variation in drought response. The following relationships among traits of growth and N economy were utilized to analyse differences among the clones (Lambers et al., 1990):

RGR = LAR × LAP(Eqn 2)
LAR = SLA × LW/W(Eqn 3)
LAP = LN/LA × LNP(Eqn 4)

anova (Rönnberg-Wästljung et al., 2005), correlation and regression analyses were performed using the spss statistical software package (ver. 11.5, SPSS Inc., Chicago, IL, USA).

QTL analysis

The point of departure for QTL analysis is a linkage map based on segregation data generated from an amplified fragment-length polymorphism analysis according to Tsarouhas et al. (2002); the linkage map is based on 92 F2 progeny, reported by Rönnberg-Wästljung et al. (2003). QTL analysis was performed by composite interval mapping conducted with the program qtl cartographer ver. 1.15 (Basten et al., 2001), using mean values for each F2 clone in each water treatment and for the treatment index of all traits. To declare a putative QTL, the threshold was set to a genome-wide significance level of 0.05. The significances are presented as LOD values, where LOD is the logarithm of the odds ratio. The LOD threshold values corresponding to the above significance level were calculated for each trait using a series of 1000 permutations. The proportion of phenotypic variation explained by each significant marker was estimated as the coefficient of determination (R2) at the peak QTL position estimated by qtl cartographer. Further details of the QTL analysis are reported by Rönnberg-Wästljung et al. (2005).


Growth of F0 grandparents

The RGR of grandparents was similar, but the female grandparent (S. viminalis) was characterized by lower LAR and SLA, but higher LAP compared with the male grandparent (S. dasyclados) (equations 2,3; Table 1). In the well watered treatment, leaf C-isotope ratio was significantly higher (less negative) in the female grandparent (t-test, P = 0.024, N = 6). The grandparents varied in their response to drought treatment: the female grandparent showed a 12% increase in LN/LA along with a 24% decrease in LNP, while the male grandparent responded with a 48% higher LN/LA, but 45% lower LNP compared with the well watered treatment. The combined effect of increased LN/LA and decreased LNP was decreased LAP (equation 4) by 15% (female grandparent) and 20% (male). The LAR was decreased by the drought treatment in the female grandparent.

Table 1.  Abbreviations of the phenotypic traits measured and the mean values (± SD)* for the two F0 grandparents (female and male) exposed to two different water regimes (well watered and drought) in a glasshouse for 6 wk
AbbreviationTrait (units)Mean Well wateredMean Drought
  • *

    Calculations of complex growth traits are based on two consecutive biomass harvests and measures of variation therefore cannot be calculated simply.

  • Salix viminalis (Jorunn).

  • Salix dasyclados.

RGRRelative growth rate (g g−1 we−1)0.310.330.240.27
LARLeaf area ratio (m2 g−1)0.00580.00720.00530.0072
LAPLeaf area productivity (g m−2 wk−1)58.051.349.341.5
SLASpecific leaf area (m2 kg−1)15.7 ± 1.317.8 ± 3.114.4 ± 0.917.3 ± 0.3
LW/WLeaf biomass fraction (%)29.7 ± 0.131.0 ± 0.129.2 ± 0.130.6 ± 0.1
LN/LALeaf N content (mmol N m−2)113.4 ± 16.182.1 ± 13.8127.0 ± 14.7121.8 ± 18.7
LNPLeaf N productivity (g [mol N]−1 wk−1)511.6624.6388.5340.4
δ13CLeaf carbon-isotope ratio (‰)−28.2 ± 1.1−30.1 ± 0.4−26.4 ± 1.2−26.6 ± 0.6

Phenotypic pattern: relationships among growth traits

Across all clones (F2 offspring) and treatments (drought and well watered), RGR was significantly affected by both LAR and LAP (equation 2; Fig. 1). However, LAP explained a considerably greater part of the variation in RGR (29%) compared with LAR (10%) and SLA (6%) (equation 3; Fig. 1). The LAP was strongly correlated with LNP (Pearson r = 0.74, P < 0.001, N = 92), but uncorrelated with LN/LA (r = 0.10, P = 0.176) (equation 4). Thus LNP explained 60% of the variation in RGR (Fig. 1). The leaf C-isotope ratio was positively correlated with LN/LA (Pearson r = 0.55, P < 0.001, N = 92), but inversely correlated with RGR, LAR, SLA and LNP (−0.24 ≥ r ≥ 0.50, P ≤ 0.001).

Figure 1.

Means of relative growth rate (RGR) for 92 full-sibling willow (Salix) hybrids plotted against means of (a) leaf area productivity (LAP); (b) leaf nitrogen productivity (LNP); (c) leaf area ratio (LAR); (d) specific leaf area (SLA). Plants were grown under two water regimes (drought, open symbols; well watered, closed symbols) in a glasshouse. Statistics (a,c,d, linear; b, quadratic regression): R2 = 0.29, P = 0.000 (a); R2 = 0.60, P = 0.000 (b); R2 = 0.10, P = 0.000 (c); R2 = 0.06, P = 0.001 (d). For abbreviations see Table 1.

The irrigation treatment and clone caused significant effects on all growth traits (anova, P < 0.001), but no significant treatment–clone interaction effects were found. The drought treatment resulted in decreased RGR (mean 14%), LNP (26%), LAR (12%) and SLA (11%), but increased LN/LA (29%) and leaf C-isotope ratio (7%) compared with the well watered treatment (t-tests, P < 0.001, N = 92 clones for all traits). The drought response in RGR by means of treatment index (drought/well watered) of individual clones varied between 0.6 and 1.1 across the clones and was explained by LAP (65%) and LNP (63%) to a greater extent compared with LAR (19%) and SLA (11%) (Fig. 2).

Figure 2.

Treatment indices (drought/well watered, d/w) for mean relative growth rate (RGR) of 92 full-sibling willow (Salix) hybrids plotted against treatment indices for (a) leaf area productivity (LAP); (b) leaf nitrogen productivity (LNP); (c) leaf area ratio (LAR); (d) specific leaf area (SLA). Statistics (a, quadratic; b–d, linear regressions): R2 = 0.65, P = 0.000 (a); R2 = 0.63, P = 0.000 (b); R2 = 0.19, P = 0.000 (c); R2 = 0.11, P = 0.001 (d). For abbreviations see Table 1.

Correlation analysis performed within the drought treatment revealed different results from the well watered treatment in 11 out of 28 cases (Table 2). The RGR was significantly correlated to LAR, SLA and LW/W only in the drought treatment. In the well watered treatment, clones with less negative C-isotope ratio (high intrinsic WUE) also had high NUE (sensu LNP), while there was no clear relationship in the drought treatment (Table 2; Fig. 3a). In contrast, the RGR was unrelated to the leaf C-isotope ratio in the well watered treatment (linear regression, R2 = 0.02, P = 0.229, N = 92), but negatively associated with leaf C-isotope ratio under drought conditions; clones with more negative leaf δ13C had higher RGR, and vice versa (linear regression, R2 = 0.11, P = 0.001, N = 92). Nevertheless, the overall pattern was a negative relationship between RGR and leaf δ13C, and a decrease in RGR, while leaf C-isotope ratio increased under drought compared with well watered conditions (Pearson r = −0.35, P < 0.001, N = 92, Fig. 3b). Also, clones with higher leaf δ13C (at WH) showed greater drought response (lower d/w) in RGR (Pearson r = −0.53, P < 0.001, N = 92).

Table 2.  Pearson correlation coefficients and significance levels (in parentheses, P ≤ 0.050 in italics) for phenological traits measured on 92 full-sibling willow (Salix) hybrids grown in a glasshouse under two water regimes (well watered and drought)
  1. Values in upper right corner refer to well watered plants; values in lower left corner to drought-treated plants (N = 92). Values in bold, different correlations in the two environments. For abbreviations see Table 1.

RGR −0.08 (0.422) 0.62 (0.000)−0.13 (0.210) 0.19 (0.063)−0.22 (0.038) 0.70 (0.000) 0.16 (0.130)
LAR 0.33 (0.001) −0.78 (0.000) 0.93 (0.000) 0.67 (0.000)−0.46 (0.000)−0.52 (0.000)−0.46 (0.000)
LAP 0.61 (0.000)−0.51 (0.000) −0.77 (0.000)−0.65 (0.000) 0.25 (0.015) 0.82 (0.000) 0.46 (0.000)
SLA 0.20 (0.048) 0.82 (0.000)−0.51 (0.000)  0.47 (0.000)−0.51 (0.000)−0.48 (0.000)−0.50 (0.000)
LW/W 0.33 (0.001) 0.69 (0.000)−0.33 (0.001) 0.34 (0.001) −0.17 (0.097)−0.54 (0.000)−0.35 (0.001)
LN/LA−0.26 (0.013)−0.46 (0.000) 0.18 (0.083)−0.50 (0.001)−0.31 (0.002) −0.33 (0.001) 0.04 (0.699)
LNP 0.73 (0.000)−0.21 (0.046) 0.81 (0.000)−0.18 (0.089)–0.11 (0.278)−0.42 (0.000)  0.43 (0.000)
δ13C−0.33 (0.001)−0.30 (0.003) 0.01 (0.920)−0.31 (0.002)−0.26 (0.011) 0.24 (0.019)−0.12 (0.244) 
Figure 3.

Means of (a) leaf nitrogen productivity (LNP); (b) relative growth rate (RGR) for 92 full-sibling willow (Salix) hybrids plotted against means of leaf carbon-isotope ratio. Plants were grown under two water regimes (○, WL, drought; •, WH, well watered) in a glasshouse. Statistics (quadratic regressions): R2 = 0.23, P = 0.000 (a, WH); R2 = 0.08, P = 0.027 (a, WL); R2 = 0.20, P = 0.000 (b).

Genotypic pattern: overlapping QTL for different growth traits, clusters

Among the eight growth traits studied, we detected most QTL for RGR (two for well watered, six for drought, eight for d/w index) and LNP (0, 3, 7) (Table 3). Fewer than 10 QTL in total were identified for LN/LA (2, 4, 1); leaf δ13C (4, 2, 3); LAP (2, 0, 5); LAR (0, 3, 3); and SLA (1, 0, 4); no QTL were found for LW/W. A higher number of QTL was found for the treatment index (mean 3.9) compared with drought (mean 2.3) and well watered treatments (mean 1.4).

Table 3.  Quantitative trait loci (QTL) for various growth traits and treatment indices measured on 92 full-sibling willow (Salix) hybrids grown in a glasshouse under two different water regimes (well watered and drought)
Trait*Chrom.MarkerWell watered treatmentDrought treatmentTreatment index (d/w)
Position (cM)Add. effectLODR2Position (cM)Add. effectLODR2Position (cM)Add. effectLODR2
  • *

    QTL for traits RGR and δ13C from Rönnberg-Wästljung et al. (2005).

  • LOD values in parentheses indicate threshold values for P = 0.050.

  • Numbers in parentheses indicate chromosome number.

  • §

    QTL analysis was performed on 1/x transformed data because of non-normal frequency distributions of untransformed data. For trait abbreviations see Table 1.

LARIIIM1417    28.5 −0.0014.22 (3.24)0.13    
CM652    21.7 −0.0013.72 (3.24)0.08    
Doublet (40)M1210        12−0.104.60 (3.18)0.17
eF155    14.9 −0.0025.31 (3.02)0.09    
jF485        14.11−0.083.78 (3.17)0.13
sF562        34.2 0.084.06 (3.17)0.16
LAPIIIM141724.5 10.403.85 (3.09)0.13        
Triplet (30)M203 6 −8.893.41 (3.09)0.15        
NM566        45.8 0.276.00 (3.04)0.15
YM292        16−0.224.61 (3.04)0.11
Triplet (44)M367         2−0.244.78 (3.04)0.19
IVF275         7.4−0.193.20 (2.90)0.10
VIF1420        26.4 0.182.94 (2.90)0.12
SLA§NM258        31.3 0.146.50 (3.28)0.11
Triplet (33)M693         0 0.103.74 (3.28)0.10
Doublet (40)M1210        12 0.135.46 (3.28)0.19
tF251038.9 −9.104.70 (2.92)0.20        
tF248        41.2 0.114.76 (3.14)0.08
LW/WNo significant QTL was detected for this trait
LN/LAOM43516.2−10.564.57 (3.27)0.15        
Triplet (45)M61111  8.903.36 (3.27)0.11        
IIIM1417    18.5 12.13.33 (3.20)0.11    
KM493    22.6−14.144.04 (3.20)0.09    
OM432        19.5 0.174.62 (3.23)0.12
eF155    12.9 12.535.05 (3.06)0.29    
Triplet (29)F3913    26 19.815.11 (3.06)0.18    
LNPTriplet (30)M557     9−67.23.35 (3.03)0.14    
IIIM1417        32.5−0.183.92 (2.92)0.20
NM566        45.8 0.246.84 (2.92)0.08
TM564        19.8−0.194.31 (2.92)0.14
YM292        16−0.236.30 (2.92)0.14
Doublet (31)M6312         0−0.194.26 (2.92)0.14
Triplet (44)M367         0−0.225.68 (2.92)0.08
mF1811    6 61.463.78 (3.30)0.29    
qF213    0−58.553.35 (3.30)0.12    
nF281         0−0.192.64 (2.48)0.07

Three common QTL for different environments were detected: QTL for leaf δ13C (linkage group III) were common for the well watered and drought treatments; QTL for SLA (t) and LN/LA (O) were found for both the well watered treatment and the treatment index. However, none of the QTL for the complex traits (RGR, LAP, LAR, LNP) was consistent across the two environments. The LOD values for QTL ranged from 2.6 to 7.3 and the QTL explained between 7 and 29% of the total phenotypic variation (Table 3).

Common or largely overlapping QTL for different traits were frequently detected, especially within chromosomes III and Y (male map) and t (female map) (Table 4). Several common QTL for RGR vs LAP (five) and RGR vs LNP (five) were identified, involving groups 30, 44, Y (male map) and IV, VI, m, q (female map). No common QTL were found for RGR vs LAR and RGR vs SLA, but one QTL was in common for LAR vs SLA regarding the treatment index. Common QTL were also detected for leaf δ13C vs RGR and LNP, here involving groups Y (male map) and IV, i, n, t (female map). The QTL analysis suggests the existence of several clusters: QTL located on chromosome III (male map) appear to control a whole suite of traits including LAP, LAR, LNP and LN/LA (Fig. 4). Groups 30, 44 and Y (male map) seem important for the control of RGR, LNP and LAP. Group t (female map) might have great importance for the expression of LNP, SLA and leaf δ13C.

Table 4.  Common quantitative trait loci (QTL) for growth traits and treatment indices measured on 92 full-sibling willow (Salix) hybrids grown in a glasshouse under two water regimes (w, well watered; d, drought)
  1. As only one QTL per chromosome was detected, chromosome symbols (chromosome numbers for doublets and triplets) represent the corresponding QTL. For trait abbreviations see Table 1; for markers associated with the respective chromosomes see Table 3.

wRGR30 (M)   30 (M)       
wLAP  III (M)III (M)30 (M)     III (M) 
wSLA t      t   
wLN/LA         O  
wδ13C     III (F)i, YYt Y 
dRGR    m, q  VI (F)    
dLAR   III (M), e      III (M) 
dLN/LA          III (M) 
dδ13C      IV (F)IV (F)    
d/wRGR       IV (F), Y44 (M)  Y, 44 (M) 
d/wLAR        40 (M)   
d/wLAP        N N, Y44 (M) 
d/wSLA          N 
d/wLNP           n
Figure 4.

Important quantitative trait loci for growth analysis traits in Salix found in the male (a) and female (b) maps, based on maps published by Rönnberg-Wästljung et al. (2005). Markers are indicated to the left of the linkage groups; see Rönnberg-Wästljung et al. (2003) for further details. (DW, above-ground dry weight; LNC, mass-based leaf N concentration; rDW, root dry weight; shd, shoot basal diameter; shl, main shoot length; for other abbreviations see Table 1).


The grandparents of the pedigree studied had similar RGR, but varied considerably in the growth traits determining RGR and in the response of these to contrasting water regimes. A high heterozygosity in the undomesticated S. dasyclados grandparent apparently caused sufficient variation in RGR and its determinants across the 92 clones of the F2 population to detect QTL (Rönnberg-Wästljung et al., 2005). In general, the precision of the phenotypic screening was rather high because of vegetative propagation of individual genotypes and many replications. We were not able to control physiological drought intensity accurately in the 1920 individuals (of partly different size) used for the final harvest. The levels of soil-water depletion that plants could sustain without wilting probably varied among the 96 genotypes (Wikberg & Ögren, 2004), and this genotypic variation in physiological drought resistance might have generated a great portion of unexplained variation in the statistical models. In addition, vegetative propagation infers nonrandom variation caused by differences in cutting quality and other sources of variation (Schwaegerle, 2005), which undoubtedly affected the precision of our estimates especially for the complex growth analysis traits that are based on two independent individuals harvested at different points in time and, even more, treatment indices (e.g. d/w treatment indices >1 for some clones, Fig. 2). Thus the absence of any significant clone–treatment interaction effect for growth analysis traits might have been caused partly by reduced precision and associated low statistical power, particularly for the interaction term. We therefore focused not on treatment responses of individual clones, but on general patterns arising from the relationships between the traits across the 92 genotypes and two irrigation treatments. To our knowledge, this is the first study comparing growth analysis traits across a relatively large number of closely related tree genotypes grown in contrasting environments.

Phenotypic pattern

Our data demonstrate that phenotypic relationships among growth traits may vary depending on irrigation conditions, not only light (Poorter & Van der Werf, 1998; Shipley, 2002) and temperature conditions (Loveys et al., 2002). As drought stress was imposed, an increase in the importance of LAR and SLA for explaining RGR was observed compared with the well watered treatment. In parallel, most genotypes decreased LAR, but about half the genotypes increased LAP (also called net assimilation rate) in response to drought (treatment indices >1, Fig. 2a). Drought-induced increase in LAP was reported for fast-growing perennials in a study comparing various species belonging to different growth forms (Galmés et al., 2005), and can be seen in the light of increased area-based leaf N and greater photosynthetic capacity at given stomatal conductance in plants grown in drought compared with well watered environments (Wright et al., 2003).

In general, when comparing herbaceous and woody species originating from a wide range of natural environments, many investigators report SLA (and LAR) as more important in determining RGR compared with LAP: LAR and SLA are usually tightly linked and positively correlated with RGR (Lambers et al., 1990; Cornelissen et al., 1996; Poorter et al., 2005). Across the whole data set, we found LAP and LNP to be more important in determining RGR, in terms of both variations among genotypes (Fig. 1) and phenotypic response to drought (Fig. 2). This pattern is in agreement with other intraspecific comparisons among Salix genotypes, at both laboratory and field scale, where a clear relationship between SLA and growth rate was lacking (Weih, 2001; Robinson et al., 2004; Tharakan et al., 2005; Weih & Nordh, 2005).

We also found that relationships between growth, water- and N-economy traits varied depending on the irrigation regime. In general, NUE and WUE may trade off, because plants tend to reduce stomatal conductance under water stress so that WUE is maximized at the expense of NUE, and because increased NUE is frequently achieved by greater stomatal conductance, which inevitably increases water losses (Field et al., 1983; Wright et al., 2003). Our data provide evidence for a phenotypic link between NUE (sensu LNP; Hirose, 1984) and intrinsic WUE (sensu leaf δ13C), but also for shifts in the NUE–WUE relationship when plants are grown under well watered compared with water-stressed conditions. A positive correlation between NUE and intrinsic WUE was found in the well watered environment; under drought conditions, clones characterized by high intrinsic WUE (less negative leaf δ13C) compromised NUE, suggesting a trade-off between NUE and WUE (Fig. 3).

Intrinsic WUE (sensu leaf δ13C) may be related to RGR in a similar way compared with LNP, because the latter (or plant N productivity) is a major determinant of RGR in this and other studies comparing Salix genotypes (Weih, 2001; Weih & Nordh, 2005). However, a combination of the nonlinear relationship between RGR and LNP; decreased LN/LA with increased RGR and LNP; and compromised LNP in favour of WUE under drought conditions caused RGR and WUE to trade off under drought conditions (Fig. 3b). This means that clones characterized by high intrinsic WUE are drought-avoiders, which is reflected by the high drought response of RGR in this material and the negative correlation between leaf δ13C and d/w RGR. Our data provide additional evidence for a phenotypic link between intrinsic WUE and RGR suggested previously by Weih (2001): the link is more apparent under drought compared with well watered conditions. In an experiment evaluating biomass production and intrinsic WUE across 29 poplar clones grown under optimal irrigation, Monclus et al. (2005) found no correlation between the two traits, and hence, suggested that it should be possible to select clones combining high productivity and high WUE. Our results indicate that this conclusion may be valid only for plants grown under optimal water supply, because none of the clones in our study combined high RGR with high intrinsic WUE under water-stress conditions. Such a pattern is not surprising as advantages of increased WUE are typically seen in drought environments (Condon et al., 2004).

QTL analysis

More than one QTL were detected for most growth traits, which demonstrates that these traits are under polygenetic control (Tanksley, 1993). Most of the QTL found here for the eight growth traits each explained between 10 and 20% of the phenotypic variation, which is within the range of other traits found for Salix (Tsarouhas et al., 2002; Rönnberg-Wästljung et al., 2005). In contrast to many simpler traits in this study (leaf δ13C, SLA, LN/LA) and Rönnberg-Wästljung et al. (2005), none of the QTL for the complex traits was consistent across different environments, and a majority of QTL detected here (95%) were specific for one of the treatments or the treatment index. Thus for all traits except LN/LA, between 38% (leaf δ13C) and 75% (SLA) of the QTL were specific to the treatment index, which suggests a complex genetic background for drought responses and drought resistance in Salix (Rönnberg-Wästljung et al., 2005), and in particular for the complex growth analysis traits. A minority of QTL and few clusters of QTL for different traits (two out of 33 in total) were detected within the well watered treatment of this study. The pattern is interesting with respect to the results from a growth and QTL analysis on Hordeum genotypes performed in a single and well watered environment (Poorter et al., 2005), where the authors found poor evidence for clusters. Our results suggest that experiments under contrasting environments might provide a better basis for the detection of QTL for complex traits and, particularly, clusters of overlapping QTL for these traits compared with investigations performed in single environments.

Relationships between phenotype and QTL pattern

In general, QTL analysis reflected the phenotypic pattern observed: more QTL were detected for LAP and the underlying traits (LNP and LN/LA) compared with LAR and its determinants (SLA and LW/W; equations 2–4); several QTL for RGR were overlapping with QTL for LAP or LNP, while no overlapping QTL were found with QTL for either LAR or SLA. In addition, the divergence of the phenotypic correlation pattern between the two irrigation treatments matches the absence of consistent QTL for the complex traits across the different environments. The accordance of the phenotypic and QTL pattern demonstrates a genetic basis for the phenotypic correlations and also provides evidence for the existence of ‘master genes’ that simultaneously control several traits. This result is contrasts with the findings of Poorter et al. (2005), which showed weak coincidence between QTL and growth analysis pattern and no indications for ‘master switches’ in Hordeum. Overlapping QTL for several traits, including RGR and total plant biomass at final harvest, were previously detected in Salix using the same F2 population (Rönnberg-Wästljung et al., 2005). This study corroborates our previous results and further links important phenotypic growth traits and their mathematical relationships (equations 2–4) to a genetic basis. These growth analysis traits are partly located in the same clusters as previously investigated simple traits such as shoot length and final plant biomass (e.g. groups Y and IV, Fig. 4). The groups N and III (male map) suggest common genetic bases for LAP vs SLA and LAP vs LAR, respectively, which reflect negative phenotypic correlations between these traits observed in our data (Table 2) and also other studies comparing different species and genotypes (Poorter et al., 2005 and references therein). In particular, the QTL at group III (male map) appears an interesting candidate for a ‘master switch’ controlling drought acclimation of LAP and its determinants (equation 4): it regulates LAP in the well watered treatment, LN/LA under drought conditions, and the drought response of the ratio between the two (LNP for the treatment index).

In contrast to forest trees, crop species and the model plant Arabidopsis thaliana have frequently been used in QTL studies on the drought response of plants related to WUE and NUE (Specht et al., 2001; Price et al., 2002; Hausmann et al., 2005). For example, Hausmann et al. (2005) found genetic correlation and QTL colocalization between intrinsic WUE (leaf δ13C) vs shoot biomass and WUE vs leaf N concentration in Arabidopsis, which they interpret in terms of genetic trade-offs between the efficiencies of water and N use. Our data indicate a genetic basis for a link between water and N economy in Salix, because phenotypic correlations coincided with two overlapping QTL for leaf δ13C vs LNP (groups Y, male map; n, female map) and three colocalizations for leaf δ13C vs RGR (groups Y, male map; i, IV, female map) mainly regarding the treatment indices. Within the same groups we also found evidence for colocalization of QTL for leaf δ13C vs LAP in two cases, which reflects positive phenotypic correlation between the traits in the drought treatment. Our data suggest that many phenotypic links (or trade-offs) between intrinsic WUE (sensu leaf δ13C) and growth traits have a genetic basis in Salix. From a breeders’ perspective, the elimination of these partly undesirable trade-offs by biotechnology tools might therefore be difficult (Weih, 2003).

Based on a relatively large number of genotypes and QTL analysis, the study elucidated the genetic basis for phenotypic correlations between important growth analysis traits in Salix. Although we identified a number of QTL for the different growth traits, we may have missed out some important QTL because the genome coverage of the linkage maps for the tetraploid F1 parents is still incomplete (Rönnberg-Wästljung et al., 2003). In addition, we detected a number of QTL candidates with LOD values just below the critical significance level (e.g. LAR in group III; LNP in group t; well watered treatment, data not shown), which might have become significant with a higher number of F2 lines and/or greater precision. On the other hand, the tetraploid nature of the F1 parents includes the possibility that some of the QTL identified here could be alleles of the same QTL, but on homologous linkage groups. We were not able to identify homologous linkage groups in this study because we used a dominant marker system with markers segregating 1 : 1. Future QTL studies will apply fully informative markers such as microsatellites, which will improve our capability to investigate the inheritance of growth traits.


We thank two anonymous referees for constructive comments on a previous version of this paper. M.W. and A.C.R.-W. would like to acknowledge funding from the Swedish Energy Agency; C.G. was funded by the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning.