Quantitative trait loci for key defensive compounds affecting herbivory of eucalypts in Australia


Author for correspondence:
Jules Freeman
Tel:+61 3 6226 1828
Fax:+61 3 6226 2698
Email: jules.freeman@utas.edu.au


  • • Formylated phloroglucinols (FPCs) are key defensive compounds that influence herbivory by mammals and arthropods in eucalypts. However, the genetic architecture underlying variation in their levels remains poorly understood.
  • • Quantitative trait loci (QTL) analysis for the concentrations of two major FPCs, sideroxylonal A and macrocarpal G, was conducted using juvenile leaves from 112 clonally duplicated progenies from an outcross F2 of Eucalyptus globulus.
  • • Two unlinked QTL were located for macrocarpal, while another unlinked QTL was located for sideroxylonal. The sideroxylonal QTL collocated with one for total sideroxylonal previously reported using adult Eucalyptus nitens foliage, providing independent validation in a different evolutionary lineage and a different ontogenetic stage.
  • • Given the potential widespread occurrence of these QTL, their ontogenetic stability, and their impact on a range of dependent herbivores, it is possible that they have extended phenotypic effects in the Australian forest landscape.


Plant secondary metabolites play an important role in herbivore defence (Fraenkel, 1959; Harborne, 1991; Bennett & Wallsgrove, 1994; Foley & Moore, 2005; Donaldson & Lindroth, 2007). However, only recently has the full extent of the effects of heritable defensive compounds on dependent communities and ecosystem processes been demonstrated (Iason et al., 2005; Whitham et al., 2006). For example, genetic variation in condensed tannins in Populus have effects beyond the population level, influencing terrestrial (Bailey et al., 2004), endophytic (Bailey et al., 2005) and aquatic communities (Le Roy et al., 2006), as well as soil nitrogen mineralization (Schweitzer et al., 2004) and the aquatic decomposition of litter (Le Roy et al., 2006). Forest trees are important foundation species in many terrestrial ecosystems because of their longevity and dominance (Ellison et al., 2005). While studies of community and ecosystem genetics in forest trees have primarily focused on Populus species, the extended phenotypic effects of genetic variation on dependent communities has been demonstrated for other foundation species (e.g. Salix, Fritz & Price, 1988; Quercus, Stone & Cook, 1998; Pinus, Iason et al., 2005), including the Australian eucalypts (Whitham et al., 1999; Dungey et al., 2000).

Eucalypts dominate many of Australia's forests and woodlands and interact with a unique biota of co-adapted organisms (Williams & Woinarski, 1997). Members of the genus contain a wide variety of plant secondary metabolites including terpenes, phenolic-based compounds, cyanogenic glycosides and the formylated phloroglucinol compounds (FPCs) (Hillis, 1996; Li et al., 1996; Eschler et al., 2000; Goodger et al., 2006). Formylated phloroglucinol compounds are fully substituted, formylated acylphloroglucinols in their simplest form (Moore et al., 2004). Units of these make up dimeric forms of FPCs, such as the relatively well-studied sideroxylonals and form adducts with terpenes to form compounds such as macrocarpals (for more detail on the structure of FPCs see Lawler et al., 1999; Moore et al., 2004). The FPCs are the single most important compounds influencing herbivory by marsupial folivores in Eucalyptus (Moore et al., 2004). These compounds have proved in several eucalypt species to have deterrent effects to co-evolved marsupial herbivores, including koalas (Phascolarctos cinereus; Moore & Foley, 2005), ringtail possums (Pseudocheirus peregrinus; Lawler et al., 2000) and common brushtail possums (Trichosurus vulpecular; O’Reilly-Wapstra et al., 2004). There is also evidence to suggest that FPCs can influence consumption of Eucalyptus foliage by invertebrate herbivores (Floyd & Foley, 2001; Andrew et al., 2007).

Eucalyptus globulus Labill. is a dominant tree species of ecological importance in native forests of south-eastern Australia and is also one of the most important hardwood plantation species in temperate regions of the world. As a result, it has been the subject of extensive genetic research (reviewed by Potts et al., 2004), making it a model for tree genetic research. Sideroxylonals and macrocarpals are quantitatively prominent FPC compounds in E. globulus (Moore et al., 2004). Quantitative genetic studies have demonstrated significant genetic based intra-specific variation in resistance of E. globulus to marsupial herbivores and correlated variation in the levels of sideroxylonals and macrocarpals (O’Reilly-Wapstra et al., 2002, 2004, 2005a,b).

Quantitative trait loci (QTL) analysis has the potential to complement traditional quantitative genetic studies, by providing information regarding the number, location and magnitude of effects of loci influencing quantitative traits. Several QTL influencing the concentration of FPCs have been identified for the first time recently in an Eucalyptus nitens cross (Henerey et al., 2007). However, putative QTL in outcrossed forest trees are often restricted to the pedigree in which they are located (Sewell & Neale, 2000) and require validation in other pedigrees to demonstrate their broad applicability. In this study we report on QTL for two FPCs in E. globulus– sideroxylonal A and macrocarpal G – and demonstrate that a QTL for sideroxylonal discovered in E. globulus is also associated with variation in levels of this compound in E. nitens.

Materials and Methods

Genetic material and trial design

Linkage mapping and QTL analysis were performed in a large outcross F2 family (T7/KI157//KI5/T144), with grandparents originating from King Island (KI; situated in Bass Strait, to the northwest of the island of Tasmania) and Taranna (T; in the southeast of Tasmania). A total of 112 genotypes were used for map construction (see Freeman et al., 2006). Each genotype was clonally replicated (two trees per genotype) and genotype means were used for QTL analysis. The use of genotype means removed some of the environmental effects and added power for QTL detection. The field trial was planted at Woolnorth in northwest Tasmania in May 1998 (see Milgate et al., 2005 for full description of the trial design). The two clones for each genotype were placed in separate replicates at random.

Analysis of formylated phloroglucinol compounds and heritability

Juvenile leaves were harvested from a single branch, at breast height on the northeast side of each tree, in July 2000. Subsequently, the leaves were freeze-dried and stored in plastic bags at 4°C. From each sample, 25–50 mg (fresh weight) of leaf tissue was ground and analysed chemically using high-performance liquid chromatography (HPLC), following Wallis & Foley (2005). Using a pure standard of sideroxylonal A (kindly provided by Prof. William Foley, Australian National University), results for this compound are expressed as mg g−1 dry matter (DM) and results for macrocarpal G are expressed as mg g−1 DM equivalents of macrocarpal G (using a macrocarpal G standard provided by Prof. William Foley). To verify the genetic basis to variation in each compound, the broad-sense heritability (H2; i.e. clonal repeatability in Falconer, 1989) and its standard error were calculated using a one-factor random model in asreml (Gilmour et al., 2006).

QTL analysis

The distribution of genotype means for sideroxylonal A departed significantly (Shapiro–Wilk, P < 0.05) from normality, and were therefore log10-transformed to normalize their distribution for QTL analysis. The distribution of genotype means for macrocarpal G did not depart significantly from normality. The QTL analysis was conducted with mapqtl 4.0 (van Ooijen et al., 2002), using the consensus linkage map constructed by Freeman et al. (2006). Putative QTL were declared at two different levels, significant (genome-wide type I error < 0.05) and suggestive (chromosome-wide type I error rate < 0.05). The LOD threshold for genome-wide significance was determined by permutation testing (1000 replications; Churchill & Doerge, 1994). An average chromosome-wide LOD threshold (LOD 3) for suggestive QTL was determined by empirical simulations (van Oijen, 1999). Interval mapping was performed initially using the default parameters of mapqtl. Map regions exceeding the suggestive threshold in interval mapping were selected as cofactors for analysis using the multiple QTL model (MQM) technique. Subsequently MQM mapping was performed, using forward selection of cofactors and an iterative approach until a stable set of cofactors was found (van Ooijen et al., 2002). Where more than one unlinked putative QTL were discovered for a trait, the closest marker to each QTL was tested for epistasis (i.e. a significant interaction effect) using PROC MIXED of SAS.


The broad-sense heritability was moderate to high for both macrocarpal G (H2 = 0.51 ± 0.07) and sideroxylonal A (H2 = 0.79 ± 0.04) concentration, indicating that the phenotypic variation in these traits has a genetic basis. Two putative QTL explained a large proportion (67.1%) of the phenotypic variance (i.e. the variance in genotype means) for macrocarpal G under MQM analysis. One highly significant QTL was located on linkage group 6 (QmacG1, LOD 7.35; Table 1, Fig. 1), and the other was located on linkage group 11 (QmacG2, LOD 3.98; Table 1, Fig. 1). The interaction between these QTL was nonsignificant (F9,85 = 1.61; P = 0.13), suggesting that there was no epistasis between them. QmacG1 segregated solely from the female parent, while QmacG2 segregated from both parents. The genotype means for QmacG2 suggested that there may be a single dominant gene affecting macrocarpal G in this region, with high concentrations of the compound recessive (Table 2). Individuals with both the most favourable genotype at QmacG2 (bd) and the favourable allele at QmacG1 (b) had a mean level of macrocarpal G that was 2.64 mg g−1 DM greater than individuals with the least favourable allele at QmacG1 (a) in combination with genotypes ac, ad, or bc, at QmacG2 (data not shown).

Table 1.  Putative quantitative trait loci (QTL) for formylated phloroglucinols in the foliage of Eucalyptus globulus
QTLLinkage groupAdjacent microsatellite markerMap position (cM)MQM mapping LODa % Expb
  • a

    Peak LOD score for each QTL, (

  • ***

    , significant, genome-wide type I error P < 0.001;

  • s

    , suggestive, chromosome-wide type I error P < 0.05).

  • b

    Percentage of phenotypic variation explained by each QTL peak.

QmacG1 6Emb17375.97.35***53.8
QsidA1 1Emb124.23.08s12.6
Figure 1.

The location of putative quantitative trait loci (QTL) for sideroxylonal A and macrocarpal G in the foliage of Eucalyptus globulus. Distance between markers in Kosambi centimorgans is shown to the left of each linkage group, while marker names are to the right. Linkage groups are as described by Freeman et al. (2006), with microsatellite markers named starting with Emb or CRC. The nearest marker to each QTL peak (shown in bold) was a microsatellite. Solid bars and lines represent one- and two-LOD support interval, respectively. The two-LOD support interval corresponds approximately to a 95% confidence interval (Lander & Botstein, 1989).

Table 2.  Genotype means at the closest marker to each quantitative trait loci (QTL) for formylated phloroglucinols in Eucalyptus globulus
TraitQTLMicrosatellite markerParental genotypesaGenotype means (mg g−1 DM)Family mean ±  SE (mg g−1 DM)
  1. a Significance of segregation from each parent: s , suggestive, P < 0.1; **, P < 0.01; ***, P < 0.001.

Macrocarpal GQmacG1Emb173ab***cdacadbcbd6.07 ± 0.141
Sideroxylonal AQsidA1Emb12ab**cd**acadbcbd2.50 ± 0.085

A single putative QTL for sideroxylonal A was located on linkage group 1 (QsidA1, LOD 3.08; Table 1, Fig. 1), explaining an estimated 12.6% of the phenotypic variance. QsidA1 segregated from both parents, the differences between genotype means for individuals with the most favourable genotype (ad), versus the least favourable genotype (bc) was 0.96 mg g−1 DM (Table 2). The genotype means suggested that the effects of different alleles at this QTL were additive (Table 2).


Two putative QTL were identified for macrocarpal G, while a single QTL was identified for sideroxylonal A. The moderate to high broad-sense heritability for both macrocarpal G and sideroxylonal A in this study, in combination with substantial intraspecific genetic variation in the levels of both of these FPCs in E. globulus (O’Reilly-Wapstra et al., 2002, 2004, 2005b), aided our ability to detect QTL influencing these traits. Our findings, together with those of Henerey et al. (2007), suggest that there are loci in eucalypt genomes that have a large effect on FPC concentration and thus potential effects on marsupial herbivores. For example, the variability evident between different genotypes at the QTL reported in this study (2.64 mg g−1 DM for macrocarpal G and 0.96 mg g−1 DM for sideroxylonal A) is likely to affect consumption by marsupial herbivores, such as the common brushtail possum (Trichosurus vulpecula) and the red-bellied pademelon (Thylogale billardierii; O’Reilly-Wapstra et al., 2002, 2004). The phenotypic range in sideroxylonal and macrocarpal concentrations in our study were 54% and 56%, respectively, of the phenotypic range observed in a diverse collection of E. globulus germplasm grown in a common environment (O’Reilly-Wapstra et al., 2004). Over these ranges of FPC concentrations the feeding response is essentially linear, with an increase in FPC concentration leading to a decrease in foliage consumption by marsupial herbivores (Lawler et al., 2000; O’Reilly-Wapstra et al., 2004).

An important finding from the present study was the collocation of a QTL for sideroxylonal A (QsidA1) in juvenile leaves of E. globulus, with a QTL for total sideroxylonal in adult leaves of E. nitens (Henerey et al., 2007), both of which were associated with microsatellite marker Emb12. Macrocarpals have not as yet been reported in E. nitens. Assuming that the collocated sideroxylonal QTL represent the effects of the same gene(s) in each species, two key inferences can be made. First, the collocation suggests that this QTL affects sideroxylonal concentration across ontogenetic stages, despite the marked changes in leaf morphology that occurs in both of these species (Jordan et al., 2000; Lawrence et al., 2003). This finding is consistent with quantitative genetic studies reporting significant population-level correlation across these ontogenetic stages for sideroxylonal content in E. globulus (O’Reilly-Wapstra et al., 2007). Second, the validation of this QTL across two different eucalypt species implies that it may be important in many eucalypt species. The FPCs appear to be constitutive defences (Rapley et al., 2007) and if there are costs and benefits associated with their production, as with many constitutive chemical defences (Siemens et al., 2003), it is possible that balancing selection (Andrew et al., 2007) is maintaining ancestral polymorphisms across eucalypt species. The available evidence, however, does not provide the resolution to determine whether this QTL results from the segregation of the same gene(s) in both species. Nevertheless, the collocation of this QTL across species suggests that it should be a primary target for identifying the underlying gene(s). Very little is known about the biosynthesis of FPCs (Moore et al., 2004; Henerey et al., 2007), thus identification of QTL for these compounds provides a step towards the identification of candidate genes underlying their variation.

Most community genetic studies have focused on gross, probably multigenic, differences between genotypes (Dungey et al., 2000; Bangert et al., 2006; Whitham et al., 2006) and specific genes, or QTL, with significant community or ecosystem effects have yet to be identified in nature. Given the potential widespread occurrence of our QTLs, their ontogenetic stability and impacts on mammalian (Lawler et al., 2000; Moore & Foley, 2005) and arthropod (Floyd & Foley, 2001; Andrew et al., 2007) herbivores, it is possible that they have extended phenotypic effects in the Australian forest landscape and are ideal targets for future community genetic studies.


We thank Paul Tilyard, Hugh Fitzgerald and Luke Rapley for their assistance with this study. We also thank Robert Barbour and Dorothy Steane for comments on the manuscript. This work was funded by the Australian Research Council, Grant DP0773686 and contributes to collaborative research supported by NSF FIBR Grant DEB-0425908.