Introduction

Improving tree growth is an important breeding objective in many forest tree-breeding programmes. The heritability of tree growth is generally moderate to low (Hamilton and Potts 2008; Whiteman et al. 1992); however, growth QTLs have been identified in a range of tree species (reviewed in Butcher and Southerton (2007). Several studies have shown a significant genotype by environment interaction for growth traits (e.g. Hamilton and Potts (2008). In spite of these limitations, substantial progress has been made in improving tree growth through breeding programmes. This has been mainly achieved due to the large variation observed for these traits. However, molecular markers associated with growth traits may be useful for increasing selection efficiency for these low heritability traits.

Several studies have been conducted in eucalypts to identify QTLs associated with growth traits. The number of QTLs detected for growth traits is generally low compared with other traits such as wood quality. For example, in a recent study we detected 36 QTLs associated with several wood quality traits in Eucalyptus nitens (Thumma et al. 2010). In contrast, only three QTLs were observed for stem volume in Eucalyptus grandis (Grattapaglia et al. 1996). Similarly, three QTLs were identified for tree diameter in a hybrid cross between E. grandis and Eucalyptus urophylla (Verhaegen et al. 1997). In E. globulus, Bundock et al. (2008) identified one QTL for cumulative growth while Freeman et al. (2009) identified two QTLs for tree diameter. The low number of QTLs detected in these studies may reflect the low heritability generally associated with growth traits. QTL studies are useful approaches for exploring the genetic control of low heritability traits and may shed light on the molecular control of poorly understood phenotypes such as tree growth. Co-location of candidate genes to the QTL intervals may shed light on the potential genes controlling growth traits and further candidate genes underlying QTL positions can be identified using the eucalypt genome (http://eucalyptusdb.bi.up.ac.za/). However, to exploit the full potential of the genome sequence there is a need to conduct more QTL studies in different species using large populations.

Vegetative propagation through macro-propagation (stem cuttings) and/or micro-propagation (tissue culture) is an important feature of many tree-breeding programmes. Clonal propagation will help in the rapid deployment of selected or improved germplasm into commercial plantations. The ability to produce roots from stem cuttings has been shown to vary among different species. Conifer species such as Pinus (e.g. Pinus taeda, Pinus radiata) and Picea have been shown to be propagated easily through stem cuttings. In eucalypts, tropical species such as Eucalyptus camaldulensis, E. grandis, E. urophylla are easy to vegetatively propagate through stem cuttings. However, temperate species such as E. globulus and E. nitens are difficult to propagate clonally either through macropropagation by stem cuttings or micropropagation by tissue culture because of low root initiation (de Little et al. 1992; England and Borralho 1995; Hetherington and Orme 1989).

One approach to improve the vegetative propagation in E. nitens is to identify markers linked to propagation traits using QTL analysis. There are a few studies in eucalypts where progeny from eucalypt hybrids have been used for QTL analysis (Grattapaglia et al. 1995; Marques et al. 1999). Some of the large effect QTLs identified by Marques et al (1999) were validated in different progeny from the same family (Marques et al. 2005). However, as these studies were mainly based on dominant marker systems, it is not possible to compare the QTLs identified in these studies with other studies. While direct comparison of QTLs between different studies is difficult when dominant markers are used in linkage map construction, it should be possible to establish linkage group level homology of QTL effects if transferable markers such as microsatellites or single-nucleotide polymorphisms (SNPs) are mapped.

There are no studies in eucalypts where tissue culture data was used for QTL analysis, and there are no studies comparing QTLs detected for vegetative propagation traits from stem cuttings with tissue culture. In this study, we analysed the QTLs associated with growth and vegetative propagation traits in three full-sib families of E. nitens growing at two sites. The main objectives of this study are (a) to identify and compare the growth QTLs in different families growing at different sites, and (b) to detect QTLs associated with vegetative propagation traits based on stem cuttings and tissue culture methods.

Materials and methods

Full-sib family at Ridgley

The full-sib family at Ridgley was part of a three-generation out-crossed pedigree (Byrne et al. 1998). In 1993, seedlings of the third generation were planted in a field experiment in Northwest Tasmania, 10 km southwest of Ridgley. The design of the field experiment was a row/column layout consisting of 20 rows and 11 columns, with five-tree line plots from the same family in each row/column grid position. The growth traits, diameter at breast height (DBH-cm) and height (Ht-m) were measured at 2 years of age, (DBH-2 and Ht-2); at 3 years (DBH-3 and Ht-3); at 4 years (DBH-4 and Ht-4); and at 7 years (DBH-7).

Full-sib families at Mt. Gambier

Two full-sib families with a common parent were selected from the Gunns Ltd. breeding programme. Family 1 (TO163 X SN382) had the largest number of progeny (327 individuals) and was used for mapping and discovering QTL for vegetative propagation and growth traits. Family 2 (SN382 X TO15) had a progeny size of 210 individuals and was used for validating QTLs discovered in family 1. The two families had a common parent SN382 which was used as a male parent in family 1 and as a female parent in family 2.

Seed from both families were germinated using standard medium. For each seedling, the stem was severed into two sections between the cotyledons and first pair of leaves. The apical tip was removed and used as a source for DNA extraction. The remainder of the upper portion comprising the leaf pairs was placed onto standard micropropagation medium (de Little et al. 1992). The lower segment including the root and cotyledons (ortet) was relocated to the glasshouse and new growth was used as the source of material for cutting assays.

Tissue culture—rooting assays

Cloned plants were transferred (i.e. subcultured) to fresh multiplication medium using standard techniques (de Little et al. 1992). The number of clones in family 1 was reduced from 327 to 300 due to ortet mortality. After about 12 weeks of multiplication, cultures were transferred from small containers to larger containers. The first rooting assay was conducted 24 weeks after germination when cultures had multiplied sufficiently. A maximum of ten shoots per clone per assay were tested for root induction. Shoots of good health and a minimum size of 1 cm were selected. Dominant shoots in a clump were selected where possible. Four rooting assays were conducted in family 1 and three rooting assays were conducted in family 2. Each shoot was scored for presence or absence of roots and the percentage of rooted shoots was obtained for each clone.

Stem cuttings—rooting assays

The ortets were transferred from root trainers into 125 mm pots and remained in the glasshouse over winter to promote shoot growth. Pruning was carried out periodically to remove apical dominance and to promote the maximum number of shoots. Shoots were harvested, dipped in auxin rooting hormone and planted in pots. Five rooting trials were conducted for family 1 and two trials were performed for family 2. Setting 2 in family 1 had some burning of cuttings due to lack of misting.

Field trial of ortets and stem cuttings from two families

The ortets and stem cuttings from the vegetative experiments were used to establish a field trial. The field trial was planted at Mt. Gambier, South Australia in August 1999. Two to five replications of rooted cuttings were established for each clone. The ortets were planted as a further replicate. Diameter at breast height was measured in 2002 (DBH-2) and 2008 (DBH-8), while a height measurement was taken in 2003 (Ht-3).

Molecular methods

In the Ridgley full-sib family, more than 290 markers were used for constructing a linkage map. Details of the different markers mapped on the linkage map were presented in Thumma et al. (2010).

In the Mt. Gambier full-sib families, the three parents were screened for polymorphic loci using RFLP and microsatellite markers. These markers were chosen to give full genome coverage based on the genetic linkage maps of E. nitens (Byrne et al. 1995) and E. globulus (Thamarus et al. 2002). In family 1, 106 markers comprising 96 RFLPs and 12 microsatellites were assayed in 327 seedlings. Thirty-eight markers were assayed in the 210 progeny of the validation family 2. As the number of markers genotyped in family 1 was low, a consensus linkage map was developed based on linkage information from previous maps of E. nitens (Byrne et al. 1995; Thumma et al. 2010) and E. globulus (Thamarus et al. 2002).

Statistical analyses of vegetative propagation and growth traits in Mt. Gambier full-sib families

Analyses were conducted in ASReml (Gilmour et al. 2006) using an individual-tree mixed effects model for vegetative propagation and growth traits in the Mt. Gambier families.

$$ {y_{ijklmn}} = \mu + {R_i} + {F_j} + R{F_{ij}} + {\hbox{ro}}{{\hbox{w}}_{k(i)}} + {\hbox{co}}{{\hbox{l}}_{l(i)}} + {\hbox{clon}}{{\hbox{e}}_{m(j)}} + {e_{ijklmn}}, $$
(1)

where y ijklmn is the individual tree measurement, μ is the overall mean, R i is the fixed effect of replicate (i = 1, 2, 3, 4, 5), F j is the fixed effect of family (j = 1, 2), rowk(i) is the random effect of row within replicate \( \sim N\left( {0,\widehat{\sigma }_{\rm{row}}^2} \right) \), coll(i) is the random effect of column within replicate \( \sim N\left( {0,\widehat{\sigma }_{\rm{col}}^2} \right) \), clonem(j) is the random effect of clone within family \( \sim N\left( {0,\widehat{\sigma }_{\rm{clone}}^2} \right) \), and e ijklmn is the random residual effect \( \sim N\left( {0,\widehat{\sigma }_{\rm{error}}^2} \right) \). Broad-sense within-family heritability for full-sib families (\( \widehat{H}_{WF}^2 \)) was estimated for each trait and theoretically contains one half the additive genetic variance, three fourths of the dominance genetic variance, and most of the epistatic genetic variance (Falconer and Mackay 1996).

$$ \widehat{H}_{WF}^2 = \frac{{\widehat{\sigma }_{\rm{clone}}^2}}{{\widehat{\sigma }_{\rm{clone}}^2 + \sigma_{\rm{error}}^2}}, $$
(2)

where \( \widehat{\sigma }_{\rm{clone}}^2 \) is the estimated variance among clones within full-sib family and the summation of \( \widehat{\sigma }_{\rm{clone}}^2 + \sigma_{\rm{error}}^2 \) is the phenotypic variance.

Additionally, repeatability of clonal means was estimated as

$$ \widehat{H}_{\overline C }^2 = \frac{{\widehat{\sigma }_{\rm{clone}}^2}}{{{\raise0.7ex\hbox{${\widehat{\sigma }_{\rm{clone}}^2 + \widehat{\sigma }_{\rm{error}}^2}$} \!\mathord{\left/{\vphantom {{\widehat{\sigma }_{\rm{clone}}^2 + \widehat{\sigma }_{\rm{error}}^2} t}}\right.}\!\lower0.7ex\hbox{$t$}}}}, $$
(3)

where t is the mean number of ramets per clone.

Approximate standard errors for heritability and clonal repeatability estimates were calculated using a Taylor series expansion method using ASReml (Gilmour et al. 2006). Clonal values were predicted in ASReml using BLUP (best linear unbiased predictions) procedures.

Additionally, the coefficient of genetic variation (CGV) was estimated in order to compare the genetic variances of the traits. The CGV expresses the genetic variance relative to the mean of the trait of interest and gives a standardised measure of the genetic variance relative to the mean of the trait. The higher the CGV for a trait, the higher is its relative variation.

$$ {\hbox{CGV}} = \frac{{{{\widehat{\sigma }}_{\rm{clone}}}}}{{\overline x }} \times 100\%, $$
(4)

where \( {\widehat{\sigma }_{\rm{clone}}} \) is the square root of the variance among clones and \( \overline x \) is the population mean for the trait.

QTL analysis

Both single marker analysis using analysis of variance and interval mapping were used to detect QTL in the Ridgley family and in family 1 of Mt. Gambier while only single-marker analysis was used in family 2 of Mt. Gambier. In the Ridgley family, interval mapping was done using QTL express (Seaton et al. 2002). The QTLs that were significantly associated with each trait were used as cofactors for searching for additional QTL with composite interval mapping (Zeng 1993). Significance thresholds were determined from permutation tests. Confidence intervals (95%) for QTL were estimated using the method of Darvasi and Soller (1997).

The clonal values representing the genetic effects adjusted for the fixed effects (replicate) were used for QTL analyses of Mt. Gambier families. As grand parental genotype information was not available, we used MapQTL 6.0 (Van Ooijen 2009) for interval mapping of the full-sib family 1 at Mt. Gambier. Genotype phase information required for QTL analysis was determined by using JoinMap (Van Ooijen 2006). Chromosome-wide and genome-wide significance levels of QTL positions were determined by permutation tests. The 95% confidence intervals were determined by the two-LOD support interval method. Multiple QTL model (MQM) mapping was used to detect additional QTLs by using the significant markers from interval mapping as co-factors. Regression approximation to the maximum likelihood algorithm implemented in MapQTL 6.0 was used in interval and MQM mapping. QTL positions were mapped using MapChart (Voorrips 2002).

Results

Trait variation in full-sib family at Ridgley

DBH on average ranged from 2.4 cm in 2-year-old trees to 16.5 cm in 7-year-old trees, while height ranged from 2.2 m in 2-year-old trees to 7.6 m in 4-year-old trees (Table 1). The correlation between early and later year growth traits was weaker compared with the correlation within early and later year growth traits (Table 2). Correlation between year 2 height and DBH was the highest, followed by the correlation between year 4 height and year 3 height and year 3 height and year 2 height. In general, correlations among height traits were higher compared with diameter traits (Table 2).

Table 1 Variation in growth traits in Eucalyptus nitens full-sib family at Ridgley
Full size table
Table 2 Phenotypic trait correlations among growth traits in Eucalyptus nitens full-sib family at Ridgley
Full size table

Trait variation in full-sib families at Mt. Gambier

Variation and heritability of growth traits

Significant clonal variation was detected (p < 0.05) for all of the growth traits measured in the two full-sib families at Mt. Gambier with greatest genetic variation detected for DBH-8 compared with growth measurements at younger ages. Additionally, highly significant differences (p < 0.001) were detected between the two-full sib families for DBH-2 and Ht-3, but this source of variation disappeared for DBH-8 (p = 0.523). The coefficient of genetic variation ranged from 5.1% for Ht-2 to greater than 10% for the two diameter traits (Table 3). Broad-sense heritability estimates for growth traits were small and ranged from 0.06 for DBH-2 to 0.16 for DBH-8 (Table 3). Similarly, clonal repeatability estimates ranged from 0.17 for DBH-2 to 0.36 for DBH-8. Within different age groups, heritability and clonal repeatability estimates for early ages, i.e. DBH-2 and Ht-3 were lower than that of later years, i.e. DBH-8.

Table 3 Mean, percentage coefficient of genetic variation (%CGV) within family broad-sense heritability (\( \widehat{H}_{WF}^2 \)) and clonal repeatability (\( \widehat{H}_{\bar{C}}^2 \)) for traits in Eucalyptus nitens full-sib families at Mt. Gambier
Full size table

Variation and heritability of vegetative propagation traits

In both families, the percentage rooting was consistently higher in stem cuttings (40.8%) compared with rooting in tissue culture (25.5%) (Table 4). The mean percentage of rooting in family 1 for tissue culture was 24.6, and for stem cuttings it was 34.5. In family 2, the mean percentage of rooting in tissue culture was 22.5 and 47.2 for rooting from stem cuttings. There was a decline in rooting percentage over time particularly in tissue culture. The second setting of stem cuttings for the family 1 pedigree was burned from water stress, which probably explains the lower percentage of rooting. The increase in rooting percentage in setting 4 was unexpected. It appears that generally the decline in rooting over time is slower in stem cuttings compared with tissue culture with the percentage of rooting dropping by 50% after three rounds of setting in tissue culture in family 2.

Table 4 Percentage of rooting in tissue culture and in stem cuttings at each setting for both Eucalyptus nitens families at Mt. Gambier
Full size table

Clonal variation for vegetative propagation (% rooting) for both tissue culture and stem cuttings was highly significant (p < 0.0001). Significant family differences were detected for % rooting in stem cuttings (p < 0.001) but not in tissue culture (p = 0.27). Broad-sense heritability for percentage of roots produced with tissue culture was higher (0.43) compared with stem cuttings (0.33) (Table 3). In general, heritability was higher for vegetative propagation traits compared with growth traits reflecting the higher percentage of genetic variation associated with vegetative propagation traits compared to growth traits (Table 3).

QTL analysis of growth traits in Ridgley full-sib family

QTL analysis of growth at different ages was performed using interval mapping. The number of QTLs detected for growth traits was lower (Table 5; Fig. 1) than those detected for wood quality traits in an earlier study using the same family (Thumma et al. 2010). The number of QTLs detected ranged from one for DBH-2 to three for DBH-7 and Ht-3. No significant QTLs were detected for Ht-2. Composite interval mapping using the significant markers as co-factors did not reveal any further QTLs. The number of QTLs detected for later year traits was higher than early year traits. This may reflect the higher heritability of later year traits compared with earlier years as seen in the Mt. Gambier families. A cluster of QTLs for year 3 and year 4 height and DBH was observed on LG5, possibly reflecting a high correlation among these traits. Similarly, QTLs for DBH-3 and DBH-4 and QTLs for Ht-3 and DBH-7 were clustered on LG6 and LG12, respectively. The percentage of variance explained by each QTL was small and varied between 2.7% and 6.4%. Most of the QTL effects were paternal in origin while the QTL effects for DBH-7 on LG1 and Ht-3 on LG12 were maternal in origin. The QTL effects for Ht-3 on LG4, DBH-2 on LG8 and Ht-4 on LG11 were due to interaction between maternal and paternal alleles. At the two common QTLs, the QTL effects for height and DBH traits on LG5 and the QTL effects for DBH-3 and DBH-4 on LG6 were paternal in origin. However, for the common QTL on LG12, the QTL effect for Ht-3 was maternal while the QTL effect for DBH-7 was paternal in origin. Three candidate genes co-located to QTL intervals for different growth traits. The flowering gene ELF1 co-located to a QTL for Ht-4 on LG11, a fasciclin-like AGP gene (FLA3) co-located to a QTL for DBH-7 on LG1 and an EST (eUPSE10) co-located to a common height QTL on LG5 (Fig. 1).

Table 5 Interval mapping of growth QTL in Eucalyptus nitens full-sib family at Ridgley
Full size table
Fig. 1
figure 1

QTL positions for growth traits in Eucalyptus nitens full-sib family at Ridgley (RD). Bars indicate 95% confidence intervals. Scale on left indicates distance in centiMorgans (cM). Markers mapping nearest to the centre of the QTL interval are shown in bold. Trait abbreviations as in Table 1. Numbering of linkage groups is based on the Eucalyptus consensus map (Brondani et al. 2006)

Full size image

QTL analysis of growth traits in Mt. Gambier full-sib families

Interval mapping of growth QTLs in family 1

A consensus map developed by comparison with the linkage map from the full-sib family at Ridgley was used for QTL analysis in family 1 at Mt. Gambier. Interval mapping in family 1 using the predicted genetic values revealed two QTLs each for DBH-8 on LG3 and LG11, and for Ht-3 on LG8 and LG9 (Table 6; Fig. 2). However, no QTLs were identified for DBH-2. Further analysis with MQM mapping did not reveal any additional QTL. The percentage of genotypic variance explained by the QTL is only marginally higher than the percentage of phenotypic variance explained by the QTL in the full-sib family at Ridgley (8.5% vs 6.4%). The percentage of genotypic variance explained varied between 6 and 8.5% in family 1 at Mt. Gambier.

Table 6 QTL analysis of growth traits in Eucalyptus nitens full-sib families at Mt. Gambier
Full size table
Fig. 2
figure 2

QTL positions for growth and vegetative propagation traits in Eucalyptus nitens full-sib family 1 at Mt. Gambier (MG). Bars indicate 95% confidence intervals. Scale on left indicates distance in cM. Markers mapping nearest to the centre of the QTL interval are shown in bold. a QTL positions of growth traits; b QTL positions of vegetative propagation traits. Trait abbreviations are as in Table 3

Full size image

Single-marker analysis of growth QTLs in family 2

As the number of markers genotyped in family 2 was small, QTL analysis was performed using single marker analysis only. Family 2 was primarily used to validate the results from family 1. Two markers on LG11 (g482 and g417) which map to the 95% confidence interval of QTL for DBH-8 observed in family 1 (Fig. 2) were significantly associated with DBH-8 (Table 7) in family 2. Similarly, g243 at 225.4 cM on LG3 was significantly associated with DBH-8. However, this marker is 182 cM away from the QTL for DBH-8 observed on LG3 (Fig. 2). One marker on LG8 at 118 cM (c104) was significantly associated with Ht-3 which was close to the QTL for Ht-3 observed in family 1. Similarly, two markers (c378 and g010) on LG9 near the QTL for Ht-3 in family 1 were significantly associated with Ht-3 in family 2. Analysis of the parental contribution to the QTL effect at marker g482 on LG11, which showed a significant effect on DBH-8 by single marker analysis in both families, revealed that the QTL effect was derived from the common parent (SN382) in both families.

Table 7 Single-marker analysis of growth traits in family 2
Full size table

Comparison of QTLs of growth traits between Ridgley and Mt. Gambier families

The QTL for DBH-7 on LG8 in the Ridgley family overlaps with the QTL for Ht-3 in the Mt. Gambier family. Similarly, the QTL for Ht-4 on LG11in the Ridgley family overlaps with the QTL for DBH-8 in Mt. Gambier family (Fig. 3). The QTL for DBH-8 on LG11 observed in family 1 of Mt. Gambier was validated in family 2 using single marker analysis (Table 7). The flowering gene, ELF1 mapped to the QTL on LG11.

Fig. 3
figure 3

Comparison of growth QTL between Eucalyptus nitens full-sib families at Ridgley (RD) and Mt. Gambier (MG)

Full size image

QTL analysis of vegetative propagation traits

Interval mapping of vegetative propagation traits in family 1

QTL analysis was performed using the predicted genetic values of mean percentage of roots produced from all settings. Four QTLs were observed for percentage of roots produced by both stem cutting and tissue culture methods (Table 8; Fig. 2). The percentage of genotypic variation explained by the QTL for percentage of roots produced ranged from 4.9% to 15.4%. On LG1 there was a common QTL for percentage of roots produced by both methods. The QTL on LG3 explained the highest percentage of genetic variation (15.4%) in roots produced by tissue culture while the QTL on LG7 explained the highest percentage of genetic variation (11.1%) in roots produced by the stem cutting method. The percentage of genotypic variation explained by vegetative propagation traits was generally higher than growth traits. Similarly, the LOD scores associated with vegetative propagation QTLs were higher than those of the growth traits. Analysis of the QTL effect at the common QTL on LG1 using the flanking markers revealed that the effect at this position was mainly derived from the paternal parent (common parent, SN382).

Table 8 QTL analysis of vegetative propagation traits in Eucalyptus nitens full-sib families at Mt. Gambier
Full size table

Single-marker analysis of vegetative propagation traits in family 2

Single-marker analysis of vegetative propagation traits in family 2 revealed that two markers (g095 on LG7, and c077B on LG11) were significantly associated with percentage of roots produced by tissue culture (Table 9). Similarly, two markers (g417 and c115) on LG11, one marker (Es140) on LG9 and one marker (c288) on LG5 were significantly associated with percentage of roots produced by stem cuttings. None of the QTLs for percentage of roots produced by tissue culture in family 1 could be validated in family 2. However, two markers (g095 on LG7 and c077B on LG11) which were significantly associated with percentage of roots produced by tissue culture, co-located to QTLs for percentage of roots produced by stem cuttings on LG7 and LG11, respectively in family 1. Of the four QTLs for percentage of roots produced by stem cuttings in family 1, at least one QTL was validated in family 2. Two markers (g417 and c115) on LG11, which were significantly associated with roots produced by stem cuttings, co-located with QTL for the same trait on LG11 in family 1. The QTL effect for percentage of roots produced by stem cuttings on LG11 in family 1 was derived mainly from the common parent (SN382), while in family 2, the QTL effect was from the paternal parent (TO15).

Table 9 Single-marker analysis of vegetative propagation traits in family 2
Full size table

Discussion

Heritability and the number of growth QTLs

The number of QTLs detected for growth traits in the Ridgley full-sib family was lower than the number of QTLs detected for wood quality traits using the same family in an earlier study (Thumma et al. 2010). However, the number of QTLs detected for growth traits in this study is higher than in comparable studies (Bundock et al. 2008; Freeman et al. 2009). The number of QTLs detected for later year traits was higher than early year growth traits (Table 5). For some early year growth traits (Ht-2) no QTLs were detected. The number of QTLs detected, therefore, appears to be directly related to the heritability of the traits. Heritability of the early year growth traits was lower than later year growth traits in both Mt. Gambier families (Table 3). Similarly, the higher number of wood quality QTLs detected in several studies may reflect the higher heritability of these traits compared with growth traits. However, heritability of a trait and the percentage of phenotypic variation explained by the QTL appear to be poorly correlated. The estimated percentage of phenotypic variation explained for growth traits reported in a number of studies is moderate (<15%; (Grattapaglia et al. 2009). Some of these values may have been overestimated as the size of the populations used in these studies was small. In a recent study Buckler et al. (2009) identified more than 30 QTLs for flowering time in maize using several large families. The percentage of phenotypic variation explained by each of these QTLs is generally less than 3% even though flowering time is highly heritable. This is in contrast to inbred species such as Arabidopsis and rice where a few QTLs explained a large percentage of variation in flowering time (El-Lithy et al. 2006; Izawa et al. 2003). Similarly, several SNPs each explaining less than 0.3% of phenotypic variation were associated with human height, although the heritability of height is very high (Weedon et al. 2007). Even though the individual QTL effect was small, flowering time in the maize parental lines was accurately predicted (R 2 > 0.87) by considering all the significant QTLs together. Buckler et al. (2009) suggested that mating system and demographics influence the genetic architecture of adaptive traits. Similar small effect QTL may underlie many economically important traits in forest trees irrespective of their heritability. This is supported by results from recent association studies which show that the percentage of variation explained by SNP-trait associations is less than 5% for wood quality and adaptive traits such as cold tolerance (Eckert et al. 2009; Gonzalez-Martinez et al. 2007; Thumma et al. 2009; Thumma et al. 2005). However, together the significant SNPs may explain a substantial proportion of trait variation.

Comparison of growth QTLs

The QTL for DBH-7 on LG8 and for Ht-4 on LG11 found in Ridgley overlap with the QTL intervals for Ht-3 on LG8 and DBH-8 on LG11 in the Mt. Gambier family. Single marker analysis results from family 2 of Mt. Gambier also suggest a QTL for DBH-8 on LG11. These QTLs, therefore, may represent the effects of common genes underlying the growth traits.

There are few studies analysing QTLs associated with growth traits in eucalypts and even fewer studies with common markers mapped between different studies making it difficult to compare the results from this study with others. Grattapaglia et al. (1996) observed several QTLs for growth traits including a QTL for DBH on LG5 in E. grandis. Freeman et al. (2009) also identified a QTL for DBH on LG5 in E. globulus. Similarly, in the present study a number of QTLs was observed on LG5. Even though all these studies suggest a growth QTL on LG5, within the linkage group the position of the QTL may not be the same. Fine mapping with several markers may determine the position of the QTL and candidate genes underlying the QTL positions.

The homolog of the Arabidopsis LFY flowering gene ELF1 maps to the common growth QTL found on LG11 (Fig. 1). The co-location of the ELF1 gene with a growth QTL may reflect the correlation between flowering time and growth observed in several Eucalyptus species (Mora et al. 2009; Wiltshire et al. 1998). Southerton et al. (1998) found that the ELF gene was expressed in floral primordia and leaf primordia in vegetative meristems of E. globulus. Similarly, Populus flowering locus (FT2) initiates the onset of flowering thus changing from the vegetative phase to the reproductive phase, and regulates seasonal growth and flowering (Hsu et al. 2006). The co-location of ELF gene with growth QTL observed in this study and the correlation between flowering time and growth seen in several Eucalyptus species may be clues to the potential molecular mechanism underlying the QTL. However, identification of candidate genes based on co-location to the QTL regions needs to be treated cautiously. Within a QTL position, there could be several linked genes. Further studies are needed to confirm the role of a gene controlling the trait.

Establishing the identity of ELF genes

We mapped ELF gene on E. nitens linkage map using an RFLP probe. This probe detected two loci segregating in the Ridgley family. These two loci mapped to LG11: one at 82 cM, and one at 116 cM, close to the QTL interval for Ht-4 (Fig. 1). Southerton et al. (1998) identified two copies of ELF in E. globulus. One of the ELF genes, ELF2, was found to be a pseudo gene. In order to find out if the ELF gene mapping to the QTL interval is the functional gene, we sequenced the nearby genomic marker (g482). We then BLAST searched the E. grandis genome sequence (http://eucalyptusdb.bi.up.ac.za/) with the two ELF sequences of E. globulus and the g482 sequence. This revealed a scaffold which contained the homologs for both ELF and the g482 sequences. We were able to determine that the ELF locus near g482 is the functional copy of the gene based on a 37 bp insertion in the 3′ region of the gene which is absent in the pseudo gene. The physical distance between the ELF gene and g482 on the scaffold is 633 kb. This translates into a genetic distance of about 1.60 cM between these two markers by considering 1 cM as equivalent to a mean physical distance of 395 kb (Grattapaglia and Bradshaw 1994). The genetic distance observed between these two markers on the E. nitens linkage map is 5.7 cM (Fig. 1). Increasing the marker density should improve the discrepancy observed between these two estimates. This analysis represents one of the first studies to utilise the public Eucalyptus genome resource (http://eucalyptusdb.bi.up.ac.za/) to establish the identity of a gene underlying a QTL interval.

Comparison of growth and wood quality QTLs in Ridgley family

Comparison of the growth QTLs detected in the Ridgley family with wood quality QTLs detected in an earlier study (Thumma et al. 2010) revealed several common QTLs between the two traits (Electronic supplementary material, Figure 1). For some shared QTL positions, the QTL effects are in a favourable direction. For example a QTL for cellulose content in the earlier study was observed near the QTL for DBH-7 on LG1. At this common interval, the QTL effect for both traits was derived from the maternal parent and the QTL effect was positive for both traits. Similarly near the common QTL for height and DBH traits on LG5, a QTL for microfibril angle (MFA) was observed. The QTL effect at this position was negative between growth traits and MFA indicating higher growth is associated with lower MFA, a favourable condition from a breeding perspective. However, there appears to be a negative or unfavourable correlation between growth and density QTL. On LG6, 11 and 12 a QTL for density was observed near the QTL for growth traits. At all these positions the QTL effects for growth and density were negative indicating higher growth was associated with lower density. All these correlations between QTL effects may reflect the phenotypic correlations observed between the growth and wood quality traits in the Ridgley family. For example, a significant negative relationship (r = −0.22) was observed between DBH-7 and wood density in the Ridgley family. The negative QTL effects between growth and density traits observed in this study reinforce similar findings from earlier studies in eucalypts (Bundock et al. 2008; Grattapaglia et al. 1996; Verhaegen et al. 1997). These results suggest that selection for growth may have some adverse impact on wood density. However, in a recent study Stackpole et al. (2009) observed that the adverse correlation found between density and growth traits at selection age (4–5 years) in E. globulus breaks down at harvest age (16–17 years). QTL analysis of density and growth traits collected at harvest age may therefore reveal QTLs independent of each other.

Vegetative propagation QTLs in Mt. Gambier families

The higher heritability of vegetative propagation traits compared with growth traits resulted in the detection of more QTLs for vegetative propagation traits. The percentages of variation explained by the vegetative propagation QTLs are also higher than for growth traits. Co-location of QTLs for percentage of roots produced by both tissue culture and stem cutting methods on LG1 suggests common genes underlying these traits. The QTL with the strong effect on LG7 was validated by single marker analysis in family 2. However, in family 1 this QTL was associated with roots produced by the stem cutting method, while in family 2 it was associated with roots produced by tissue culture. Similarly, the QTL on LG11 was validated in family 2 for percentage of roots produced by both methods. Previous studies have reported a number of putative QTL for the ability to produce roots from stump sprout cuttings (Grattapaglia et al. 1995; Marques et al. 1999). Marques et al. (2002) by comparing different studies identified a homologous linkage group (LG1) containing a putative QTL for vegetative propagation. The QTL for roots produced by stem cutting and tissue culture observed on LG1 in the present may be the same as the common QTL observed in other studies.

Grattapaglia et al. (1995) suggested that a small number of QTLs of large effect may underlie vegetative propagation traits in eucalypts (Grattapaglia et al. 1995). This is in contrast to the results from the present study. It seems likely that the large effects are in part due to the small sample size employed in the study (Beavis 1998). In an Eucalyptus tereticornis x E. globulus pedigree the QTL effects for the same traits tended to be smaller and the sample size used was much larger (Marques et al. 1999). Comparison of QTLs between different studies is difficult as QTL detection is influenced by a number of factors such as population size, experimental design and the analytical techniques employed.

Conclusions

We have identified several QTLs for growth and vegetative propagation traits using three families growing at two sites. QTL analysis in Mt. Gambier families had more power as genetic data predicted from clonal replications was used. At two positions, QTLs for growth are conserved in Ridgley and Mt. Gambier families. Two vegetative propagation QTLs on LG7 and LG11 were validated in a separate family. However, these results need to be treated cautiously. While clonal replicates were used to obtain the phenotype data, fewer markers were genotyped in the Mt. Gambier families compared with the Ridgley family. In Ridgley family 1, the other hand the trait data was obtained from single trees.

Currently, there are few candidate genes for complex traits such as growth in eucalypts. The flowering gene ELF2 on LG11 could be considered a candidate gene controlling growth. It is difficult to compare QTL positions from this study with other Eucalyptus studies. There are few QTL reports examining growth traits and even fewer on vegetative propagation traits in eucalypts and the number of common markers between the different studies is very low. There is a clear need for the use of common gene-based markers to better understand the genetic control of complex traits such as growth and to compare results from different studies. At present, the candidate gene markers on the E. nitens linkage map are primarily related to cell wall biosynthesis. Currently, we are mapping several candidate genes for adaptive traits on the E. nitens linkage map. The public Eucalyptus genome resource (http://eucalyptusdb.bi.up.ac.za/) will be very useful for revealing candidate genes underlying QTL intervals identified in this study. QTL mapping represents the first step in identifying candidate genes controlling complex but important traits such as growth and vegetative propagation traits. Marker-assisted selection is expected to be very useful for growth traits as phenotypic correlations between early and later year growth traits is low and markers linked to growth traits may be used for increasing selection efficiency at an early age.