SEARCH

SEARCH BY CITATION

Keywords:

  • Pinus pinaster;
  • 13C;
  • growth;
  • heritability;
  • quantitative trait;
  • stable isotope;
  • tree rings;
  • water use efficiency

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Classical quantitative genetics and quantitative trait dissection analysis (QTL) approaches were used in order to investigate the genetic determinism of wood cellulose carbon isotope composition (δ13C, a time integrated estimate of water use efficiency) and of diameter growth and their relationship on adult trees (15 years) of a forest tree species (maritime pine). A half diallel experimental set-up was used to (1) estimate heritabilities for δ13C and ring width and (2) to decompose the phenotypic δ13C/growth correlation into its genetic and environmental components. Considerable variation was found for δ13C (range of over 3‰) and for ring width (range of over 5 mm) and significant heritabilities (narrow sense 0·17/0·19 for δ13C and ring width, respectively, 100% additivity). The significant phenotypic correlation between δ13C and ring width was not determined by the genetic component, but was attributable to environmental components. Using a genetic linkage map of a full-sib family, four significant and four suggestive QTLs were detected for δ13C, the first for δ13C in a forest tree species, as far as known to the authors. Two significant and four suggestive QTLs were found for ring width. No co-location of QTLs was found between δ13C and growth.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

In given environmental conditions, trees with high water-use efficiency at the leaf level (intrinsic WUE, defined as the ratio of net CO2 assimilation rate A to stomatal conductance for water vapour g) can maintain higher growth rates under water-limited conditions than trees with lower WUE (Sun et al. 1996; Nguyen-Queyrens et al. 1998). Measurements of plant carbon isotope composition (δ13C) provide time-integrated estimates of WUE (Farquhar, O’Leary & Berry 1982; Farquhar & Richards 1984) that can be applied to adult trees (Zhang & Marshall 1994; 1995; Guehl et al. 1995; Sun et al. 1996). Assessments of differences in WUE among- and within-tree species are facilitated by the crown- and time-integrative nature of tree ring δ13C.

Forest tree species are known to be among the most polymorphic species of the flora (Hamrick, Godt & Sherman-Broyles 1992). Genotypic differences in leaf δ13C of conifer species were found among provenances in common garden studies (Zhang, Marshall & Jaquish 1993; Zhang & Marshall 1994, 1995; Guehl et al. 1995; Nguyen-Queyrens et al. 1998). It has been suggested that provenance differences of δ13C might be determined by differences in stomatal sensitivity to changes in vapour pressure deficit (Zhang & Marshall 1995) and/or differences in plant hydraulic characteristics (Guehl et al. 1995). However, differences of δ13C among genetic families within provenances of Picea mariana (Mill.) were found to be mainly determined by differences in photosynthetic capacity (Johnsen & Major 1995; Major & Johnsen 1996). Similar indications were obtained for maritime pine (Pinus pinaster Ait.) by Guehl et al. (1995).

Tree growth is an important goal for forest tree breeding programmes. To avoid inadvertent negative selection for growth when selecting for high WUE, it is important to know if δ13C and growth are genetically linked. Positive but weak phenotypic relationships between δ13C and height or diameter growth (Flanagan & Johnsen 1995; Nguyen-Queyrens et al. 1998; Johnsen et al. 1999) have been found among trees within different forest tree species. Genetic parameters calculated for physiological or morphological traits can disentangle phenotypic relationships into genotypic and environmental components. Johnsen et al. (1999) found strong genetic correlations between δ13C and tree height or tree diameter. They concluded that A was determining δ13C and growth performance and thus constituted probably the link between the two traits. However, since δ13C as an indicator of WUE could be either controlled by A and/or by g, there is not necessarily a strong relationship between A and δ13C. This suggests that the existence of a genetic correlation between δ13C and growth is dependent on the factor by which WUE is controlled.

Adaptive traits, such as δ13C, are characterized by high phenotypic variation among and within populations of forest tree species (Meinzer et al. 1992; Zhang et al. 1993; Zhang & Marshall 1995; Flanagan & Johnsen 1995; Nguyen-Queyrens et al. 1998). Moreover, high heritabilities for δ13C have been found for non-woody (Matus, Slinkard & Van Kessel 1995; Asay, Johnson & Palazzo 1998) and woody species (Johnsen et al. 1999). The development of genetic mapping (Tanksley 1993) has made it possible to localize genetic factors controlling quantitative traits [quantitative trait loci (QTLs)]. High heritability of a trait is a favourable factor for quantitative trait dissection analysis.

In crop plant breeding, improvement of WUE has been an important aim and therefore the first QTLs for δ13C were detected in tomato (Martin et al. 1989). Mansur et al. (1993) found in a preliminary investigation of δ13C on soybean one large genomic region that could be responsible for as much as 53% of the observed variation. In a study of 3-week-old barley plants (Pakniyat et al. 1997), 12 AFLP markers were detected for δ13C, two of these markers alone accounted for 53·2% of the variation. QTLs for water use efficiency, as measured by the ratio of dry weight to water used, were found in soybean (Mian et al. 1996; Mian, Ashley & Boerma 1998).

Detection of QTL on woody species, however, is still in development, due to the long generation time and therefore the lack of controlled crosses. Genetic maps have often to be constructed from F1 full-sib progenies. Carlson et al. (1991) were the first to show that randomly amplified polymorphic DNA (RAPD) primers could be screened for informative markers segregating in a 1 : 1 ratio in diploid tissue of full-sib progenies. Grattapaglia & Sederoff (1994) extended this idea in constructing parental maps of an interspecific eucalyptus hybrid family in a mapping strategy named ‘two-way pseudo-testcross’. It was further used in conifers (Kubisiak et al. 1996; Arcade et al. 2000) with RAPDs and amplified fragment length polymorphisms (AFLPs). For maritime pine the genome coverage required for linkage map construction and QTL analysis was achieved by using RAPD markers (Plomion et al. 1995a; Plomion, O’Malley & Durel 1995b) and AFLP analysis (Costa et al. 2000).

The objectives of the present study were: (1) to estimate the variability and heritability of δ13C and ring width in a forest tree species (Pinus pinaster Ait.) using a half-diallel experimental design; (2) to investigate the phenotypic correlation between δ13C and growth; (3) to separate the phenotypic correlation between δ13C and growth into a genetic and an environmental component; (4) to dissect δ13C and ring width into Mendelian inherited components (quantitative trait dissection analysis) using a P. pinaster Ait. full-sib family; and (5) to compare QTLs for δ13C with QTLs for growth.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Half diallel

A 12 by 12 half-diallel of maritime pine (Pinus pinaster Ait.) was used to estimate the variability, heritability and genetic correlations among the studied traits. Parental trees were crossed in 1980 and seeds from the controlled crosses sown in a nursery in spring 1982 and planted in autumn 1982. The 12 parents were trees phenotypically selected for stem growth and straightness in the local provenance of the Landes de Gascogne. The half-diallel was located in Cestas (Gironde, France, 44°44′ N, 0°44′ W) on a semi-humid podzolic soil. Spacing was 4 m between rows and 1·1 m between individual trees, i.e. 2272 trees/ha. No selfed crosses were analysed, therefore the half-diallel consisted of 66 families (12 female and 11 male parents) of 5–15 individuals each. Three families were not available, therefore only 63 families were analysed. A parentage test was performed using three microsatellites (Gerber et al. 2000) confirming the authenticity of the progenies used in the half diallel. The experimental design consisted of 74 incomplete randomized blocks (the large number of blocks is due to the fact that the presented half-diallel is part of a much larger complete diallel). For the present study, 564 trees were cut in March 1997 (trees were 15 years old). Discs were sampled, dried in a greenhouse and analysed for carbon isotopic composition (δ13C) and ring width as described below.

Full-sib family

A three-generation outbred pedigree comprising 202 15-year-old-trees was used to study the genetic architecture of the studied traits, i.e. the number, genome location and effect of Quantitative Trait Loci. The four grandparents were trees phenotypically selected for stem growth and straightness in the local provenance of the Landes de Gascogne and grafted in clonal archives. These grand parents were tested in a polycross progeny test and classified according to their breeding value as ‘Vigor +’ (for vigorous trees) and ‘Vigor –’ (for less vigorous trees). Each of the parental trees is the result of the cross of one ‘Vigor +’ and one ‘Vigor –’ grandparent. The two parental trees were crossed in 1980 and seeds from the controlled cross-sown in spring 1982. They produced progeny seedlings that were planted in autumn 1982. The family was located in Malente (Gironde, France, 44°30′ N, 0°47′ W) on a semi-humid podzolic soil. Spacing was 4 m between rows and 1·1 m between individual trees, i.e. 2272 trees/ha. The trees were felled in March 1997 and stem discs were cut, dried in a greenhouse and analysed for δ13C and ring width. From all the analysed families, 16 trees with no visible growth in the last 4 years before harvest were removed from the analysis.

Ring width measurements

Wood subsamples were taken from four positions on the circumference of the stem discs for the last 4 years of growth (1993–96). As the trees were cut in March 1997, this includes any growth during winter 1996/1997 utilizing reserve material from the summer 1996 growth period. Two different methods were used for ring width measurements. For the half diallel experiment, ring width was measured using the indirect X-ray-method first described by Polge (1966). For the full-sib family, the ring widths were measured at the four sampling points on the circumference using a semi-automatic system consisting of a digitizing tablet linked to a computer (precision 0·1 mm standard deviation). The width was averaged for each ring. To make the ring width data comparable with the δ13C measurements, for each tree the mean growth was calculated for the years 1993–1996 [mean ring width (MRW)], using an arithmetic mean.

Isotope measurements

The δ13C was measured of a bloc of four rings, which represents a ring-width weighted mean of the δ13C of each ring. The sampled blocks of wood were cut by hand into small pieces, pre-ground in a centrifugal mill (Tecator, Cyclotech 1093; Sample Mill, Höganäs, Sweden) and milled to a fine powder in a ball mill (Retsch, MM2000; Retsch, Haan, Germany). Cellulose was extracted after an acidic acid/nitric acid procedure described in Brendel, Iannetta & Stewart (2000). In brief, the method used a concentrated nitric acid/80% acetic acid 1 : 10 dilution (0·2 cm3 in 2 cm3) to digest lignin, proteins and hemicelluloses in 50 mg of powdered wood sample. The digested molecule fragments were then washed out using ethanol, any remaining acid was removed during a water wash. The samples were dried chemically with a pure ethanol/acetone progression and physically in a vacuum centrifugal evaporator (speed vac) at 100 hPa for 2 h. The original protocol (Brendel et al. 2000) was modified to include two extraction cycles, a 0·5 molar NaOH wash replacing the water wash to remove acids more thoroughly and prolonging the ethanol washes to 5 min at 60 °C. For δ13C analysis, 1 mg cellulose subsamples were combusted and analysed for 13C composition using a continuous flow isotope ratio mass spectrometer (Delta S; Finnigan MAT, Bremen, Germany). Carbon isotope composition was calculated relative to the Pee Dee Belemnite standard as (Craig 1957):

  • image(1)

where Rsa and Rsd are the 13C/12C ratios of the sample and the standard, respectively. The discrimination between the δ13C of atmospheric CO2 (δair≈− 8‰) and the δ13C of plant material (δplant) was calculated as (Farquhar & Richards 1984):

  • image(2)

Intrinsic WUE was estimated from discrimination using a modified equation from Farquhar et al. (1982):

  • image(3)

where ca is the atmospheric CO2 concentration (estimated as 360 × 10−6 mol mol−1), b is the net fractionation caused by carboxylation (27‰) and Δ is the discrimination between the δ13C of atmospheric CO2 and the δ13C of cellulose (Eqn 2).

Estimation of genetic parameters

The normality of the distribution of the traits for both experimental set-ups was tested using Smirnov–Kolmogorov test. Although δ13C was normally distributed, a small distortion from the normality was observed for ring width (P-value = 0·01). However, this distortion was considered to be too small to necessitate an adjustment.

Analyses of variance for block and family effects in the half diallel were carried out with the OPEP software (Baradat 1989; Baradat & Labbe 1995) according to the following model derived from the ‘Henderson III’ model (Searle 1971):

  • image(4)

where Yijk is the value of the trait for the individual k belonging to the family j, located in the block i, Bi is the fixed effect of the ith block, Fj is the random effect of the jth family and ɛijk is the random residual comprising: individual deviation from family mean and family–block interactions. When block and family effects were significant, data were adjusted to the block effect, prior to the decomposition of family effect. The half diallel analysis was carried out with OPEP using the model presented below, it is derived from the simplification of the random diallel model described by Garretsen & Keuls (1977) (Baradat & Desprez-Loustau 1997) which is adapted to non-orthogonal trials with reciprocal crosses:

  • image(5)

where Yijk is the value of the trait for the individual k corresponding to the cross between the male i and the female j, ai (aj) is the general combining ability (GCA) of the ith (jth) parent, sij is the specific combining ability (SCA) of the cross between the ith and the jth parent and ɛijk is the residual term. The additive and dominance variances are: σA2 = 4 σa2 and σD2 = 4 :σs2, whereas the phenotypic variance is: σP2 = σ 2(Yijk) = 2 σa2 + σs2 + σε2 . The narrow and broad sense heritabilities were calculated as hns2 = σA2P2 and hbs2 = (σA2 + σD2)/ σP2 , respectively. The percentage of additivity is calculated as the additive variance divided by the sum of additive plus dominance variances: σA2/( σA2 + σD2). Genetic and environmental correlations were computed with OPEP using a multitrait analysis of variance and covariance: (1) ‘estimated’ genetic and environmental correlations were calculated according to the additive and dominance effects assessed in the random model (Eqn 5; parental level), and (2) ‘predicted’ correlations were assessed from the individual breeding values of each tree (individual level; Fig. 2a–c). However, as the estimated and the predicted correlations gave similar results, only the correlation coefficients based on the prevalent ‘estimated’ results were used in the discussion. Standard errors of estimates of heritabilities were computed using the robust Jackknife method (Lebart, Morineau & Fénelon 1979).

image

Figure 2. Linear correlations between mean ring width (MRW) and δ13C for the half diallel experiment using data corrected with the individual tree breeding values (OPEP software): (a) phenotypic; (b) genetic (additive effects); and (c) environmental correlations; data for genetic and environmental correlations are centred and standardized by the mean.

Download figure to PowerPoint

QTL detection

Two genetic maps corresponding to the female and male parents of the full-sib family were established using AFLP markers, genotyped on a subset of 90 F1. The whole mapping population was further genotyped with evenly spaced markers to increase the statistical power of QTL detection (Chagnéet al. 2002). In order to reduce the intra-trial environmental background noise, the data were adjusted for the block effect. We used the two-way pseudo-test cross mapping strategy to construct the linkage map (Grattapaglia & Sederoff 1994). Twelve linkage groups were found for the female map, equalling the number of chromosomes for P. pinaster. For the male map 15 linkage groups were detected; however, the combination of the male and the female maps into a consensus map using three-quarter/one-quarter segregating markers yielded 12 linkage groups for each parent. For QTL analysis however, only the one-half/one-half segregating markers could be used.

For QTL analysis, MultiQTL software (A. Korol, http:www.multiqtl.com) was used. In a first step, QTLs were detected by interval mapping using a logarithmic odds ratio (LOD) threshold of 1·5 and a one-QTL-model (one QTL per linkage group). In a second step, these QTLs were taken as cofactors [composite interval mapping (CIM); introduced by Jansen & Stam 1994 and Zeng 1994], allowing individual QTL to be detected independently to the background noise. In a third step, a two QTL model (Korol et al. 1998) using CIM was applied, first testing if two QTLs were significant and then testing if two QTLs were more significant than one QTL. Standard deviations for the positions of the QTLs were calculated using a bootstrap method.

As there are difficulties involved when using asymptotic approximations of LOD statistics (fixed LOD level) for QTL detection (Doerge & Churchill 1996), a permutation approach was used to determine appropriate significance thresholds. Two theoretical critical thresholds were considered, the first corresponding to a per linkage group type I error of 5% allowing the detection of ‘suggestive’ QTL and the second corresponding to a genome wise type I error of 5% allowing the detection of ‘significant’ QTL. Theoretical critical threshold corresponding to a genome wise type I error of 5% were calculated for each chromosome taking into account the number of markers in each chromosome. If αm is the critical threshold at the marker level corresponding to a 5% genome wise type I error, the αc (critical threshold at the chromosome level) for a chromosome comprising n markers would be: αc = 1 − (1 − αm)n. These theoretical thresholds were compared to the thresholds associated with the LOD obtained by CIM at the chromosome level after 1000 permutations of the data. The proportion of phenotypic variance explained by each QTL was estimated using the coefficient of determination (R2, estimated by CIM, 1000 permutations), which is based on the partial correlation of a putative QTL with the trait adjusted for cofactors in the multilocus model.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Trait distributions

Means, ranges and variability for δ13C were very similar between the half diallel and the full-sib experimental designs (Table 1, Fig. 1). For MRW the growth was higher in the half diallel by 0·66 mm and also the range of observed values was larger (Table 1). However, the coefficient of variation was slightly higher for the full-sib experimental design than for the half diallel. This was also true when the coefficients of variation for MRW were calculated using the block effect adjusted data (data not shown).

Table 1.  Means, standard deviations (SD), coefficients of variation (SD/mean), ranges (maximum – minimum) and probability (P-value in percentage) of block and family effects (Eqn 4) of the two experimental designs for δ13C (‰) and mean ring width [MRW; (mm)]
  nMeanSDCoef. of var.RangeBlock effectFamily effect
  • *

    Significant at 5% level; NS, not significant.

δ13CHalf diallel564−26·210·610·023·642·8*2·3*
Full-sib186−26·480·630·023·230·0*
MRWHalf diallel564  2·881·100·386·77NS0·2*
Full-sib186  2·110·990·475·150·9*
image

Figure 1. Distribution of cellulose δ13C (raw data; not corrected for block effect) for (a) the half diallel experiment and (b) the full-sib experiment; parameters for the normal distributions as in Table 1.

Download figure to PowerPoint

Half diallel

Analysis of variance including family and block effect (Eqn 4) indicated variation among families for both δ13C and MRW, which justified the decomposition of the family effect according to Eqn 5. Taking into account the significant block effect for δ13C, the data were adjusted prior to the genetic decomposition. The narrow sense heritabilities were highly significant (P < 0·005) for δ13C and for MRW and close in their values (Table 2). No dominance effects were detected for these two traits, thus narrow sense and broad sense heritabilities are equal and additivities are 100%.

Table 2.  Half diallel: heritabilities for δ13C and mean ring width (MRW) with standard deviations and 95% confidence interval in parentheses. As there were no dominance effects detected and therefore additivity is 100%, the narrow sense (ns) and broad sense (bs) heritabilities are equal
 hns2 = hbs2% additivity
  • *

    Significant at the 5% level.

δ13C (‰)0·17 ± 0·06 (0·06–0·29)*100%
MRW (mm)0·19 ± 0·06 (0·06–0·31)*100%

The phenotypic correlation between mean ring width and δ13C associated faster growth with less negative δ13C values (higher WUE) and was significant with a coefficient of correlation of r = 0·45 (Table 3 and Fig. 2a). The correlation was not significant (Table 3) for the genetic component (additive effect), whereas the environmental component was highly significant with a strong correlation coefficient (r = 0·52).

Table 3.  Half diallel: correlations between mean ring width and δ13C; r is the correlation coefficient on the family level, estimated by OPEP software with standard deviation and 95% confidence interval in parentheses; r′ is the correlation coefficient estimated by linear regression analysis using the calculated individual tree breeding values which are shown in Fig. 2
 rr
  • *

    Significant at the 5% level; NS: not significant

Phenotypic correlation0·45 ± 0·057 (0·33–0·56)*0·46*
Genetic correlation0·27 ± 0·21 (0·07–0·15) NS0·02 NS
Environmental correlation0·52 ± 0·16 (0·20–0·83)*0·79*

Full-sib family

A significant positive phenotypic correlation between MRW and δ13C of the full-sib family (r = 0·39; P < 0·005; Fig. 3) was observed. For δ13C, eight QTLs were found on seven linkage groups (chromosomes) and for MRW six QTLs on four linkage groups (Table 4). Using the one-QTL model, six QTLs were detected for δ13C and two for MRW. With the two-QTL model, one pair of QTLs was found for δ13C and two pairs for MRW. For δ13C, two of the QTLs detected with the one-QTL model and the QTL-pair detected with the two-QTL model and for MRW one QTL-pair are ‘significant QTLs’ at a probability corresponding to a 5% genome wise type I error. All other detected QTLs are ‘suggestive QTLs’ at a probability corresponding to a 5% chromosome type I error. For δ13C, QTLs were detected on the male and the female maps; however, not on the same chromosomes. A multilocus model, including the male and female maps, explained 51·4% of the phenotypic variation of δ13C, the major QTL at chromosome 6 alone explaining 12·4%. For MRW, no QTLs were found on the female map and a multilocus model for the male map explained 42·9% of the observed phenotypic variation. No colocalization for a QTL of δ13C and of MRW was found.

image

Figure 3. Phenotypic linear correlation between mean ring width (MRW) and δ13C for the full-sib experiment (data adjusted for block effect).

Download figure to PowerPoint

Table 4.  Full sib: results of the composite interval mapping analysis for δ13C and mean ring width (MRW) using MultiQTL software. The P-value associated with the LODs were calculated using 1000 permutations of the data, standard deviation of position (SD) was calculated using the bootstrap method (1000 permutations); in case of a significant QTL-pair for a chromosome, LOD and P-value are given for the tests of (I) two QTLs versus no QTLs and (II) two QTLs versus one QTL (difference of LOD for two and for one QTL)
 MapChraNb Position ± SDc LOD P-valuedDir.eR2Rtotal2
  • a

    Chromosome ID.

  • b

    Number of full-sibs with available data for QTL detection.

  • c

    LOD score peak position (from the top) of the chromosome (cM).

  • d

    Probability for the null hypothesis of no QTL at the chromosome level.

  • ***

    Probability corresponding to a 5% genome wise type I error (significant QTL).

  • *

    Probability corresponding to a 5% chromosome type I error (suggestive QTL).

  • e

    Direction of influence of presence of the allele for each QTL.

δ13CMale 3a149   4·3 ± 13·2 1·78 0·021*+0·0470·268
Male 6164 102·7 ± 21·6 4·40 0·001***+0·124 
Male 8 85   0·0 ± 15·7 1·85 0·021*0·050 
Male 9183 104·6 ± 28·5 1·90 0·033*0·047 
Female 2 84 209·1 ± 56·4 2·30 0·019*+0·0650·246
Female 5164  99·9 ± 22·0 1·98 0·003***0·062 
Female121531:   0·0 ± 36·7I: 4·24 I: 0·002*** 0·119 
  2: 135·4 ± 27·1II:1·88II:0·036*+  
MRWMale 2b180 1:  47·2 ± 15·9 I: 2·49 I: 0·048* 0·1810·429
 1532: 51·8 ± 11·1II:1·57II:0·027*+  
Male 5145  56·8 ± 33·4 2·41 0·022*0·065 
Male 6174 1:   7·5 ± 31·0 I: 4·24 I: 0·001*** 0·124 
 1682:133·7 ± 22·0II:2·68II:0·002*+  
Male11165  11·6 ± 30·9 1·73 0·040*+0·059 

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Trait distributions

Standard deviations and ranges were similar for the half diallel and for the full-sib (Table 1). For δ13C the coefficients of variation were nearly the same, whereas the coefficient of variation for mean ring width was higher for the full-sib than for the half diallel (Table 1). Variability of traits might be expected to be higher in a half diallel with 12 parental trees, than in one full-sib family. However, it has to be taken into account that parental trees for both experiments were selected from the same provenance (Landes, Gascogne) and therefore might be genetically close. Further, with a polygenic complex trait such as δ13C, due to transgression, even parents with only a small difference in a measured trait can produce offspring with extreme values (Prioul et al. 1997). For δ13C, the average of family variation of the half diallel (0·55‰ standard deviation within a range of 0·0‰ to 1·0‰ standard deviations) was similar to the variation found for the 186 trees of the full-sib family (0·63‰ standard deviation). Similarly, for MRW, the average of standard deviations within half-diallel families was 0·94 mm, which is close to the 0·99 mm standard deviation found for the full-sib family. Using Eqns 2 and 3, the measured δ13C values transformed into a range of WUE of 67 µmol CO2 mol−1 H2O to 100 µmol mol−1 for the half-diallel and a range of 65 to 95 µmol mol−1 for the full-sib. This represents for the half diallel and the full-sib a variation from one to one-and-a half times the WUE.

Heritabilities and quantitative trait dissection analyses

The heritabilities for mean ring width and δ13C were found to be significant, similar between the two traits and of rather moderate value. Therefore selective crossings can improve growth and WUE. The heritabilities for ring width are comparable with values found in the literature for maritime pine or other conifers. Danjon (1994) found for maritime pine trees from the same provenance as used in the present study narrow-sense heritabilities for diameter growth ranging from 0 to 0·45 for different experimental set-ups (40–100 half or full-sib families). Blada (1999) found for a Pinus cembra L. 10 × 10 full-diallel narrow-sense heritabilities for diameter from 0·23 to 0·32 and broad-sense heritabilities from 0·50 to 0·59.

For δ13C there are no publications known to the authors that estimated heritability for maritime pine, and there are only a few publications of estimates of heritability for δ13C for other species. Narrow-sense heritability estimates by Johnsen et al. (1999) for Picea mariana are lower for diameter growth (0·14) than for δ13C (0·54). For non-woody species, heritabilities for δ13C can be high (broad-sense heritabilities for Lens culinaris Medikus 0·73, Matus et al. 1995 and for Agropyron desertorum (Fischer ex Link) Schultes 0·90, Asay et al. 1998), however, it has been shown that water stress could reduce the heritability of δ13C (Johnson et al. 1990; Ehdaie & Waines 1994). An explanation for the moderate heritabilities found in the present study could therefore be the integrative properties of δ13C measured on cellulose of several rings, together with the possibility of frequent water stress. The present study was located in the south-west of France, where summer drought is common (Nguyen-Queyrens et al. 1998). The half diallel was created from the descendants of trees selected for growth vigour. This might have restricted the genetic base compared with natural populations and hence lowered the detectable heritability of growth.

Existing QTLs for maritime pine were localized for traits related to growth (Plomion, Durel & O’Malley 1996; Gerber, Lascoux & Kremer 1997). We were able to provide here the first example of QTL observations for δ13C in a forest tree species. The four significant QTLs found for δ13C explained nearly one-third of the phenotypic variation observed for this trait. Several experiments (Prioul et al. 1997) have shown that even for complex traits, such as growth or carbon isotope discrimination, the expected number of major loci is quite small, a small number of genetic factors predominantly determine a quantitative trait. No co-localizations of QTLs for δ13C and QTLs for MRW were detected, suggesting no common genetic control for these two traits. However, underestimation of the number of QTL is inherent to the methodology of QTL detection.

Relationships between δ13C and MRW

The phenotypic correlations between δ13C and growth (ring width) found for the half diallel and full-sib experimental designs are significant with moderate coefficients of correlation (Figs 2a & 3, Table 3) and the estimated regressions are similar in slope and intercept. Among trees in the same environmental conditions, this suggests that an increased growth relates to a higher WUE. Depending on the physiology of a plants, a difference in WUE could be predominantly determined by stomatal conductance and/or by assimilation rate. The Farquhar model of carbon isotope discrimination (Farquhar, Ehleringer & Hubick 1989) predicts that an increasing photosynthetic capacity will decrease Δ. Positive as well as negative correlations have been found between photosynthesis and growth (Johnsen & Major 1995); however, when assuming a positive correlation between photosynthetic capacity and growth, a positive correlation between δ13C and growth could suggest a predominantly assimilation rate-based control of δ13C. This is in agreement with results for black spruce: differences among families were found to be mainly determined by differences in photosynthesis (Johnsen & Major 1995), whereby differences in photosynthesis were rather the result of non-stomatal limitations than of stomatal limitation (Major & Johnsen 1996). If δ13C were to be controlled by stomatal conductance, the Farquhar model predicts a negative correlation between δ13C and growth. Therefore, the positive correlation between δ13C and growth suggests that the variation of WUE among the measured trees is rather controlled by assimilation than by stomatal conductance. This was the case for the half diallel as well as for the full-sib family.

The calculated genetic and environmental correlations between δ13C and MRW indicate that the phenotypic correlation is mainly based on environmental influence. The block effect was included in the estimation of the genetic parameters, hence the environmental correlation is probably due to microenvironmental influences on each individual tree. As the model also accounts for any type of genetic effect, including intra-family genetic variation, the observed large environmental variation therefore suggests for the two measured traits a high sensitivity to microenvironmental conditions. This also suggests for growth and water use efficiency a high non-genetic plasticity to adjust to environmental conditions. The strong environmental correlation that was found for the two traits is therefore probably due to a substantial environmental influence of less negative δ13C with increased growth and vice versa.

The non-significant genetic correlation in the present study is in opposition to the strong genetic correlation between δ13C and tree growth (height and diameter) found for black spruce (Johnsen et al. 1999). Several factors might have contributed to this discrepancy. First, Johnsen et al. (1999) suggested assimilation rate as common control for δ13C and growth. However, correlations between assimilation rate and growth found in the literature range from negative over non-significant to positive relationships (Johnsen & Major 1995). Therefore, even if δ13C is determined by assimilation rate, if growth is not determined by assimilation rate, then there might be no correlation between δ13C and growth. Second, genetic control was rather moderate for both traits, which might have lowered the significance of a genetic correlation. Third, it has also to be taken into account that in the present study carbon isotope discrimination measured on cellulose of main stem wood was compared to the average diameter growth of 4 years, whereas Johnsen et al. (1999) compared δ13C of needle material with height or trunk diameter. These complex traits might include the action of a number of different genes and a common genetic control might exist for the combination of needle δ13C to height or diameter and not for the combination of trunk cellulose to mean annual growth of the same growth period.

The lack of a genetic correlation between δ13C and MRW found in the half-diallel is in agreement with the lack of co-localizations between QTLs for δ13C and for MRW found in the full-sib experimental design. Albeit the lack of a genetic correlation, both traits were found to be heritable and significant QTLs were detected. This opens new perspectives for the investigation of the genetic determinism of water use efficiency and the identification of groups of genes involved in drought responses.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Financial support for this work was provided by a post-doctoral grant INRA/Région Lorraine, a contract of Région Aquitaine and funding by Action structurante INRA Ecogene. We thank Claude Bréchet for the isotopic measurements, Elvire Hatch, Régis Burlett and Michel Leprêtre for help with sample preparation.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  • Arcade A., Anselin F., Faivre Rampant P., Lesage M.C., Laurans F., Paques L.E. & Prat D. (2000) Application of AFLP, RAPD and ISSR markers to genetic mapping of European larch and Japanese larch. Theoretical and Applied Genetics 100, 299307.
  • Asay K.H., Johnson D.A. & Palazzo A.J. (1998) Parent-progeny relationships for carbon isotope discrimination and related characters in crested wheatgrass. International Journal of Plant Sciences 159, 821825.
  • Baradat P. (1989) Amélioration génétique des arbres forestiers, éléments méthodologiques. pp. 1204. INRA, Bordeaux, France.
  • Baradat P. & Desprez-Loustau M.L. (1997) Analyse diallèle et intégration de la sensibilitéà la rouille courbeuse dans le programme d’amélioration du pin maritime. Annales des Sciences Forestieres 54, 83106.
  • Baradat P. & Labbe T. (1995) Traitement statistique des essais de sélection. In: Opep. Un logiciel intégré pour l’amélioration des plantes pérennes, pp. 303330. Cirad, Montpellier, France.
  • Blada I. (1999) Diallel crossing in Pinus cembra III. Analysis of genetic variation at the nursery stage. Silvae Genetica 48, 179187.
  • Brendel O., Iannetta P.P.M. & Stewart D. (2000) A rapid and simple method to isolate pure alpha-cellulose. Phytochemical Analysis 11, 710.DOI: 10.1002/(sici)1099-1565(200001/02)11:1<7::aid-pca488>3.0.co;2-u
  • Carlson J.E., Tulsieram L.K., Glaubitz J.C., Luk V.W.K., Kauffeld C. & Rutledge R. (1991) Segregation of random amplified DNA markers in F1 progeny of conifers. Theoretical and Applied Genetics 83, 194200.
  • Chagné D., Lalanne C., Madur D., et al. (2002) A high density genetic map of Maritime pine based on AFLPs. Annals of Forest Science in press.
  • Costa P., Pot D., Dubos C., Frigerio J.-M., Pionneau C., Bodénès C., Bertocchi E., Cervera M., Remington D.L. & Plomion C. (2000) A genetic map of Maritime pine based on AFLP, RAPD and protein markers. Theoretical and Applied Genetics 100, 3948.
  • Craig H. (1957) Isotopic standards for carbon and oxygen and correction factors for massspectrometric analysis of carbon dioxide. Geochimica et Cosmochimica Acta 12, 133149.
  • Danjon F. (1994) Heritabilities and genetic correlations for estimated growth curve parameters in maritime pine. Theoretical and Applied Genetics 89, 911921.
  • Doerge R.W. & Churchill G.A. (1996) Permutation tests for multiple loci affecting a quantitative character. Genetics 142, 285294.
  • Ehdaie B. & Waines J.G. (1994) Genetic analysis of carbon isotope discrimination and agronomic characters in a bread wheat cross. Theoretical and Applied Genetics 88, 10231028.
  • Farquhar G.D. & Richards P.A. (1984) Isotopic composition of plant carbon correlates with water-use efficiency of wheat genotypes. Australian Journal of Plant Physiology 11, 539552.
  • Farquhar G.D., Ehleringer J.R. & Hubick K.T. (1989) Carbon isotope discrimination and photosynthesis. Annual Review of Plant Physiology and Molecular Biology 40, 503537.
  • Farquhar G.D., O'Leary M.H. & Berry J.A. (1982) On the relationship between carbon isotope discrimination and the intercellular CO2-concentration in leaves. Australian Journal of Plant Physiology 9, 121137.
  • Flanagan L.B. & Johnsen K.H. (1995) Genetic variation in carbon isotope discrimination and its relationship to growth under field conditions in full-sib families of Picea mariana. Canadian Journal of Forest Research 25, 3947.
  • Garretsen F. & Keuls M. (1977) A general method for the analysis of genetic variation in complete and incomplete dialleles and North Caroline 2 designs. Part I. Procedures and general formulas for the random model. Euphytica 26, 537551.
  • Gerber S., Lascoux M. & Kremer A. (1997) Relation between protein markers and quantitative traits in maritime pine (Pinus pinaster Ait.). Silvae Genetica 46, 286291.
  • Gerber S., Mariette S., Streiff R., Bodénès C. & Kremer A. (2000) Comparison of microsatellites and amplified fragment length polymorphism markers for parentage analysis. Molecular Ecology 9, 10371048.DOI: 10.1046/j.1365-294x.2000.00961.x
  • Grattapaglia D. & Sederoff R. (1994) Genetic linkage maps of Eucalyptus grandis and Eucalyptus europhylla using a pseudo testcross: mapping strategy and RAPD markers. Genetics 137, 11211137.
  • Guehl J.-M., Nguyen-Queyrens A., Loustau D. & Ferhi A. (1995) Genetic and environmental determinants of water use efficiency and carbon isotope discrimination in forest trees. In EUROSILVA Contribution to Forest Tree Physiology (eds H. Sandermann & M. Bonnet-Masimbert), pp. 297321. INRA, Paris, France.
  • Hamrick J.L., Godt M.J.W. & Sherman-Broyles S.L. (1992) Factors influencing levels of genetic diversity in woody plant species.New Forests 6, 95124.
  • Jansen R.C. & Stam P. (1994) High resolution of quantitative trait into multiple loci via interval mapping. Genetics 136, 14471445.
  • Johnsen K.H., Flanagan L.B., Huber D.A. & Major J.E. (1999) Genetic variation in growth, carbon isotope discrimination, and foliar N concentration in Picea mariana: analyses from half-diallel mating design using filed-grown trees. Canadian Journal of Forest Research 29, 17271735.DOI: 10.1139/cjfr-29-11-1727
  • Johnsen K.H. & Major J.E. (1995) Gas exchange of 20-year-old black spruce families displaying a genotype–environment interaction in growth rate. Canadian Journal of Forest Research 25, 430439.
  • Johnson D.A., Asay K.H., Tieszen L.L., Ehleringer J.R. & Jefferson P.G. (1990) Carbon isotope discrimination: potential in screening cool-season grasses for water-limited environments. Crop Science 30, 338343.
  • Korol A., Ronin Y., Hayes P. & Nevo E. (1998) Multi-interval mapping of correlated trait complexes: simulation analysis and evidence from barley. Heredity 80, 273284.
  • Kubisiak T.L., Nelson C.D., Nance W.L. & Stine M. (1996) RAPD linkage mapping in a longleaf pine × slash pine F1 family. Theoretical and Applied Genetics 90, 11101127.
  • Lebart L., Morineau A. & Fénelon J.P. (1979) Traitements des Données Statistiques. Dunod, Paris, France.
  • Major J.E. & Johnsen K.H. (1996) Family variation in photosynthesis of 22-year-old black spruce: a test of two models of physiological response to water stress. Canadian Journal of Forest Research 26, 19221933.
  • Mansur L.M., Lark K.G., Kross H. & Oliveira A. (1993) Interval mapping of quantitative trait loci for reproductive, morphological, and seed traits of soybean (Glycine max L.). Theoretical and Applied Genetics 86, 907913.
  • Martin B., Nienhuis J., King G. & Schaefer A. (1989) Restriction fragment length polymorphisms associated with water use efficiency in tomato. Science 243, 17251728.
  • Matus A., Slinkard A.E. & Van Kessel C. (1995) Carbon isotope discrimination and indirect selection for seed yield in lentil. Crop Science 35, 679684.
  • Meinzer F.C., Rundel P.W., Goldstein G. & Sharifi M.R. (1992) Carbon isotope composition in relation to leaf gas exchange and environmental conditions in Hawaiian Metrosideros polymorpha populations. Oecologia 91, 305311.
  • Mian M.A.R., Ashley D.A. & Boerma H.R. (1998) An additional QTL for water use efficiency in soybean. Crop Science 38, 390393.
  • Mian M.A.R., Bailey M.A., Ashley D.A., Wells R., Carter T.E., Parrot J.W.A. & Boerma H.R. (1996) Molecular markers associated with water use efficiency and leaf ash in soybean. Crop Science 36, 12521257.
  • Nguyen-Queyrens A., Ferhi A., Loustau D. & Guehl J.-M. (1998) Within-ring δ 13C spatial variability and interannual variation in wood cellulose of two contrasting provenances of Pinus pinaster. Canadian Journal of Forest Research 28, 766773.DOI: 10.1139/cjfr-28-5-766
  • Pakniyat H., Powell W., Baird E., Handley L.L., Robinson D., Scrimgeour C.M., Nevo E., Hackett C.A., Caligari P.D.S. & Forster B.P. (1997) AFLP variation in wild barley (Hordeum spontaneum C. Koch) with reference to salt tolerance and associated ecogeography. Genome 40, 332341.
  • Plomion C., Bahrman N., Durel C.-E. & O'Malley D.M. (1995a) Genomic mapping in Pinus pinaster (maritime pine) using RAPD and protein markers. Heredity 74, 661668.
  • Plomion C., O'Malley D.M. & Durel C.-E. (1995b) Genomic analysis in maritime pine (Pinus pinaster). Comparison of two RAPD maps using selfed and open-pollinated seeds of the same individual. Theoretical and Applied Genetics 90, 10281034.
  • Plomion C., Durel C.-E. & O'Malley D.M. (1996) Genetic dissection of height in maritime pine seedlings raised under accelerated growth conditions. Theoretical and Applied Genetics 93, 849858.
  • Polge H. (1966) Etablissement des courbes de variation de la densité du bois par exploration densitométriques de radiographies d’échantillons prélevés à la tarière sur des arbres vivants. Annales Des Sciences Forestieres 23, 1206.
  • Prioul J.-L., Quarrie S., Causse M. & Vienne D.D. (1997) Dissecting complex physiological functions through the use of molecular quantitative genetics. Journal of Experimental Botany 48, 11511163.
  • Searle S.D. (1971) Linear Models, p. 532. John Wiley, New York, USA.
  • Sun Z.J., Livingston N.J., Guy R.D. & Ethier G.J. (1996) Stable carbon isotopes as indicators of increased water use efficiency and productivity in white spruce (Picea glauca (Moench) Voss) seedlings. Plant, Cell and Environment 19, 887894.
  • Tanksley S.D. (1993) Mapping polygenes. Annual Review of Genetics. 27, 205233.DOI: 10.1146/annurev.genet.27.1.205
  • Zeng Z.-B. (1994) Precision mapping of quantitative trait loci. Genetics 136, 14571468.
  • Zhang J. & Marshall J.D. (1994) Population differences in water-use efficiency of well-watered and water-stressed western larch seedlings. Canadian Journal of Forest Research 24, 9299.
  • Zhang J. & Marshall J.D. (1995) Variation in carbon isotope discrimination and photosynthetic gas exchange among populations of Pseudotsuga menziesii and Pinus ponderosa in different environments. Functional Ecology 9, 402412.
  • Zhang J., Marshall J.D. & Jaquish B.C. (1993) Genetic differentiation in carbon isotope discrimination and gas exchange in Pseudotsuga menziesii. Oecologia 93, 8087.