Genomic Prediction of Complex Traits in Perennial Plants: A Case for Forest Trees.

This chapter provides an overview of the genomic selection progress in long-lived forest tree species. Factors affecting the prediction accuracy in genomic prediction are assessed with examples from empirical studies. Infrastructure and resources required for the implementation of genomic selection are evaluated. Some general guidelines are provided for the successful application of genomic selection in forest tree breeding programs.

A genome-wide SNP genotyping resource for tropical pine tree species.

We performed gene and genome targeted SNP discovery towards the development of a genome-wide, multispecies genotyping array for tropical pines. Pooled RNA-seq data from shoots of seedlings from five tropical pine species was used to identify transcript-based SNPs resulting in 1.3 million candidate Affymetrix SNP probe sets. In addition, we used a custom 40 K probe set to perform capture-seq in pooled DNA from 81 provenances representing the natural ranges of six tropical pine species in Mexico and Central America resulting in 563 K candidate SNP probe sets. Altogether, 300 K RNA-seq (72%) and 120 K capture-seq (28%) derived SNP probe sets were tiled on a 420 K screening array that was used to genotype 576 trees representing the 81 provenances and commercial breeding material. Based on the screening array results, 50 K SNPs were selected for commercial SNP array production including 20 K polymorphic SNPs for P. patula, P. tecunumanii, P. oocarpa and P. caribaea, 15 K for P. greggii and P. maximinoi, 13 K for P. elliottii and 8K for P. pseudostrobus. We included 9.7 K ancestry informative SNPs that will be valuable for species and hybrid discrimination. Of the 50 K SNP markers, 25% are polymorphic in only one species, while 75% are shared by two or more species. The Pitro50K SNP chip will be useful for population genomics and molecular breeding in this group of pine species that, together with their hybrids, represent the majority of fast-growing tropical and subtropical pine plantations globally.

Prediction ability of genome-wide markers in Pinus taeda L. within and between population is affected by relatedness to the training population and trait genetic architecture.

Genomic prediction has the potential to significantly increase the rate of genetic gain in tree breeding programs. In this study, a clonally replicated population (n = 2063) was used to train a genomic prediction model. The model was validated both within the training population and in a separate population (n = 451). The prediction abilities from random (20% vs 80%) cross validation within the training population were 0.56 and 0.78 for height and stem form, respectively. Removal of all full-sib relatives within the training population resulted in ∼50% reduction in their genomic prediction ability for both traits. The average prediction ability for all 451 individual trees was 0.29 for height and 0.57 for stem form. The degree of genetic linkage (full-sib family, half sib family, unrelated) between the training and validation sets had a strong impact on prediction ability for stem form but not for height. A dominant dwarfing allele, the first to be reported in a conifer species, was discovered via genome-wide association studies on linkage Group 5 that conferred a 0.33-m mean height reduction. However, the QTL was family specific. The rapid decay of linkage disequilibrium, large genome size, and inconsistencies in marker-QTL linkage phase suggest that large, diverse training populations are needed for genomic selection in Pinus taeda L.

Genomic prediction for fusiform rust disease incidence in a large cloned population of Pinus taeda.

In this study, 723 Pinus taeda L. (loblolly pine) clonal varieties genotyped with 16920 SNP markers were used to evaluate genomic selection for fusiform rust disease caused by the fungus Cronartium quercuum f. sp. fusiforme. The 723 clonal varieties were from five full-sib families. They were a subset of a larger population (1831 clonal varieties), field-tested across 26 locations in the southeast US. Ridge regression, Bayes B, and Bayes Cπ models were implemented to study marker-trait associations and estimate predictive ability for selection. A cross-validation scenario based on a random sampling of 80% of the clonal varieties for the model building had higher (0.71-0.76) prediction accuracies of genomic estimated breeding values compared with family and within-family cross-validation scenarios. Random sampling within families for model training to predict genomic estimated breeding values of the remaining progenies within each family produced accuracies between 0.38 and 0.66. Using four families out of five for model training was not successful. The results showed the importance of genetic relatedness between the training and validation sets. Bayesian whole-genome regression models detected three QTL with large effects on the disease outcome, explaining 54% of the genetic variation in the trait. The significance of QTL was validated with GWAS while accounting for the population structure and polygenic effect. The odds of disease incidence for heterozygous AB genotypes were 10.7 and 12.1 times greater than the homozygous AA genotypes for SNP11965 and SNP6347 loci, respectively. Genomic selection for fusiform rust disease incidence could be effective in P. taeda breeding. Markers with large effects could be fit as fixed covariates to increase the prediction accuracies, provided that their effects are validated further.

Toward genomic selection in Pinus taeda: Integrating resources to support array design in a complex conifer genome.

PREMISE: An informatics approach was used for the construction of an Axiom genotyping array from heterogeneous, high-throughput sequence data to assess the complex genome of loblolly pine (Pinus taeda). METHODS: High-throughput sequence data, sourced from exome capture and whole genome reduced-representation approaches from 2698 trees across five sequence populations, were analyzed with the improved genome assembly and annotation for the loblolly pine. A variant detection, filtering, and probe design pipeline was developed to detect true variants across and within populations. From 8.27 million variants, a total of 642,275 were evaluated and 423,695 of those were screened across a range-wide population. RESULTS: The final informatics and screening approach delivered an Axiom array representing 46,439 high-confidence variants to the forest tree breeding and genetics community. Based on the annotated reference genome, 34% were located in or directly upstream or downstream of genic regions. DISCUSSION: The Pita50K array represents a genome-wide resource developed from sequence data for an economically important conifer, loblolly pine. It uniquely integrates independent projects that assessed trees sampled across the native range. The challenges associated with the large and repetitive genome are addressed in the development of this resource.

Major QTL confer race-nonspecific resistance in the co-evolved Cronartium quercuum f. sp. fusiforme-Pinus taeda pathosystem.

Fusiform rust disease, caused by the endemic fungus Cronartium quercuum f. sp. fusiforme, is the most damaging disease affecting economically important pine species in the southeast United States. Unlike the major epidemics of agricultural crops, the co-evolved pine-rust pathosystem is characterized by steady-state dynamics and high levels of genetic diversity within environments. This poses a unique challenge and opportunity for the deployment of large-effect resistance genes. We used trait dissection to study the genetic architecture of disease resistance in two P. taeda parents that showed high resistance across multiple environments. Two mapping populations (full-sib families), each with ~1000 progeny, were challenged with a complex inoculum consisting of 150 pathogen isolates. High-density linkage mapping revealed three major-effect QTL distributed on two linkage groups. All three QTL were validated using a population of 2057 cloned pine genotypes in a 6-year-old multi-environmental field trial. As a complement to the QTL mapping approach, bulked segregant RNAseq analysis revealed a small number of candidate nucleotide binding leucine-rich repeat genes harboring SNP associated with disease resistance. The results of this study show that in P. taeda, a small number of major QTL can provide effective resistance against genetically diverse mixtures of an endemic pathogen. These QTL vary in their impact on disease liability and exhibit additivity in combination.

Genomic Breeding for Diameter Growth and Tolerance to Leptocybe Gall Wasp and Botryosphaeria/Teratosphaeria Fungal Disease Complex in Eucalyptus grandis.

Eucalyptus grandis is one of the most important species for hardwood plantation forestry around the world. At present, its commercial deployment is in decline because of pests and pathogens such as Leptocybe invasa gall wasp (Lepto), and often co-occurring fungal stem diseases such as Botryosphaeria dothidea and Teratosphaeria zuluensis (BotryoTera). This study analyzed Lepto, BotryoTera, and stem diameter growth in an E. grandis multi-environmental, genetic trial. The study was established in three subtropical environments. Diameter growth and BotryoTera incidence scores were assessed on 3,334 trees, and Lepto incidence was assessed on 4,463 trees from 95 half-sib families. Using the Eucalyptus EUChip60K SNP chip, a subset of 964 trees from 93 half-sib families were genotyped with 14,347 informative SNP markers. We employed single-step genomic BLUP (ssGBLUP) to estimate genetic parameters in the genetic trial. Diameter and Lepto tolerance showed a positive genetic correlation (0.78), while BotryoTera tolerance had a negative genetic correlation with diameter growth (-0.38). The expected genetic gains for diameter growth and Lepto and BotryoTera tolerance were 12.4, 10, and -3.4%, respectively. We propose a genomic selection breeding strategy for E. grandis that addresses some of the present population structure problems.

Evaluation of the efficiency of genomic versus pedigree predictions for growth and wood quality traits in Scots pine.

BACKGROUND: Genomic selection (GS) or genomic prediction is a promising approach for tree breeding to obtain higher genetic gains by shortening time of progeny testing in breeding programs. As proof-of-concept for Scots pine (Pinus sylvestris L.), a genomic prediction study was conducted with 694 individuals representing 183 full-sib families that were genotyped with genotyping-by-sequencing (GBS) and phenotyped for growth and wood quality traits. 8719 SNPs were used to compare different genomic with pedigree prediction models. Additionally, four prediction efficiency methods were used to evaluate the impact of genomic breeding value estimations by assigning diverse ratios of training and validation sets, as well as several subsets of SNP markers. RESULTS: Genomic Best Linear Unbiased Prediction (GBLUP) and Bayesian Ridge Regression (BRR) combined with expectation maximization (EM) imputation algorithm showed slightly higher prediction efficiencies than Pedigree Best Linear Unbiased Prediction (PBLUP) and Bayesian LASSO, with some exceptions. A subset of approximately 6000 SNP markers, was enough to provide similar prediction efficiencies as the full set of 8719 markers. Additionally, prediction efficiencies of genomic models were enough to achieve a higher selection response, that varied between 50-143% higher than the traditional pedigree-based selection. CONCLUSIONS: Although prediction efficiencies were similar for genomic and pedigree models, the relative selection response was doubled for genomic models by assuming that earlier selections can be done at the seedling stage, reducing the progeny testing time, thus shortening the breeding cycle length roughly by 50%.

Quantitative Genetics and Genomics Converge to Accelerate Forest Tree Breeding.

Forest tree breeding has been successful at delivering genetically improved material for multiple traits based on recurrent cycles of selection, mating, and testing. However, long breeding cycles, late flowering, variable juvenile-mature correlations, emerging pests and diseases, climate, and market changes, all pose formidable challenges. Genetic dissection approaches such as quantitative trait mapping and association genetics have been fruitless to effectively drive operational marker-assisted selection (MAS) in forest trees, largely because of the complex multifactorial inheritance of most, if not all traits of interest. The convergence of high-throughput genomics and quantitative genetics has established two new paradigms that are changing contemporary tree breeding dogmas. Genomic selection (GS) uses large number of genome-wide markers to predict complex phenotypes. It has the potential to accelerate breeding cycles, increase selection intensity and improve the accuracy of breeding values. Realized genomic relationships matrices, on the other hand, provide innovations in genetic parameters' estimation and breeding approaches by tracking the variation arising from random Mendelian segregation in pedigrees. In light of a recent flow of promising experimental results, here we briefly review the main concepts, analytical tools and remaining challenges that currently underlie the application of genomics data to tree breeding. With easy and cost-effective genotyping, we are now at the brink of extensive adoption of GS in tree breeding. Areas for future GS research include optimizing strategies for updating prediction models, adding validated functional genomics data to improve prediction accuracy, and integrating genomic and multi-environment data for forecasting the performance of genetic material in untested sites or under changing climate scenarios. The buildup of phenotypic and genome-wide data across large-scale breeding populations and advances in computational prediction of discrete genomic features should also provide opportunities to enhance the application of genomics to tree breeding.

Improving wood properties for wood utilization through multi-omics integration in lignin biosynthesis.

A multi-omics quantitative integrative analysis of lignin biosynthesis can advance the strategic engineering of wood for timber, pulp, and biofuels. Lignin is polymerized from three monomers (monolignols) produced by a grid-like pathway. The pathway in wood formation of Populus trichocarpa has at least 21 genes, encoding enzymes that mediate 37 reactions on 24 metabolites, leading to lignin and affecting wood properties. We perturb these 21 pathway genes and integrate transcriptomic, proteomic, fluxomic and phenomic data from 221 lines selected from ~2000 transgenics (6-month-old). The integrative analysis estimates how changing expression of pathway gene or gene combination affects protein abundance, metabolic-flux, metabolite concentrations, and 25 wood traits, including lignin, tree-growth, density, strength, and saccharification. The analysis then predicts improvements in any of these 25 traits individually or in combinations, through engineering expression of specific monolignol genes. The analysis may lead to greater understanding of other pathways for improved growth and adaptation.

Correction: Linkage and Association Mapping for Two Major Traits Used in the Maritime Pine Breeding Program: Height Growth and Stem Straightness.

[This corrects the article DOI: 10.1371/journal.pone.0165323.].

Linkage and Association Mapping for Two Major Traits Used in the Maritime Pine Breeding Program: Height Growth and Stem Straightness.

BACKGROUND: Increasing our understanding of the genetic architecture of complex traits, through analyses of genotype-phenotype associations and of the genes/polymorphisms accounting for trait variation, is crucial, to improve the integration of molecular markers into forest tree breeding. In this study, two full-sib families and one breeding population of maritime pine were used to identify quantitative trait loci (QTLs) for height growth and stem straightness, through linkage analysis (LA) and linkage disequilibrium (LD) mapping approaches. RESULTS: The populations used for LA consisted of two unrelated three-generation full-sib families (n = 197 and n = 477). These populations were assessed for height growth or stem straightness and genotyped for 248 and 217 markers, respectively. The population used for LD mapping consisted of 661 founders of the first and second generations of the breeding program. This population was phenotyped for the same traits and genotyped for 2,498 single-nucleotide polymorphism (SNP) markers corresponding to 1,652 gene loci. The gene-based reference genetic map of maritime pine was used to localize and compare the QTLs detected by the two approaches, for both traits. LA identified three QTLs for stem straightness and two QTLs for height growth. The LD study yielded seven significant associations (P ≤ 0.001): four for stem straightness and three for height growth. No colocalisation was found between QTLs identified by LA and SNPs detected by LD mapping for the same trait. CONCLUSIONS: This study provides the first comparison of LA and LD mapping approaches in maritime pine, highlighting the complementary nature of these two approaches for deciphering the genetic architecture of two mandatory traits of the breeding program.

Quantitative trait loci influencing forking defects in an outbred pedigree of loblolly pine.

BACKGROUND: The use of wood as an industrial raw material has led to development of plantation forestry, in which trees are planted, managed, and harvested as crops. The productivity of such plantations often exceeds that of less-intensively-managed forests, and land managers have the option of choosing specific planting stock to produce specific types of wood for industrial use. Stem forking, or division of the stem into two or more stems of roughly equal size, is a character trait important in determining the quality of the stem for production of solid wood products. This trait typically has very low individual-tree heritability, but can be more accurately assessed in clonally-replicated plantings where each genotype is represented by several individual trees. We report results from a quantitative trait mapping experiment in a clonally-replicated full-sibling family of loblolly pine (Pinus taeda L.). RESULTS: Quantitative trait loci influencing forking defects were identified in an outbred full-sibling family of loblolly pine, using single-nucleotide polymorphism markers. Genetic markers in this family segregated either in 1:2:1 (F2 intercross-like segregation) or 1:1 ratio (backcross-like segregation). An integrated linkage map combining markers with different segregation ratios was assembled for this full-sib family, and a total of 409 SNP markers were mapped on 12 linkage groups, covering 1622 cM. Two and three trait loci were identified for forking and ramicorn branch traits, respectively, using the interval mapping method. Three trait loci were detected for both traits using multiple-trait analysis. CONCLUSIONS: The detection of three loci for forking and ramicorn branching in a multiple-trait analysis could mean that there are genes with pleiotropic effects on both traits, or that separate genes affecting different traits are clustered together. The detection of genetic loci associated with variation in stem quality traits in this study supports the hypothesis that marker-assisted selection can be used to decrease the rate of stem defects in breeding populations of loblolly pine.

Performance of genomic prediction within and across generations in maritime pine.

BACKGROUND: Genomic selection (GS) is a promising approach for decreasing breeding cycle length in forest trees. Assessment of progeny performance and of the prediction accuracy of GS models over generations is therefore a key issue. RESULTS: A reference population of maritime pine (Pinus pinaster) with an estimated effective inbreeding population size (status number) of 25 was first selected with simulated data. This reference population (n = 818) covered three generations (G0, G1 and G2) and was genotyped with 4436 single-nucleotide polymorphism (SNP) markers. We evaluated the effects on prediction accuracy of both the relatedness between the calibration and validation sets and validation on the basis of progeny performance. Pedigree-based (best linear unbiased prediction, ABLUP) and marker-based (genomic BLUP and Bayesian LASSO) models were used to predict breeding values for three different traits: circumference, height and stem straightness. On average, the ABLUP model outperformed genomic prediction models, with a maximum difference in prediction accuracies of 0.12, depending on the trait and the validation method. A mean difference in prediction accuracy of 0.17 was found between validation methods differing in terms of relatedness. Including the progenitors in the calibration set reduced this difference in prediction accuracy to 0.03. When only genotypes from the G0 and G1 generations were used in the calibration set and genotypes from G2 were used in the validation set (progeny validation), prediction accuracies ranged from 0.70 to 0.85. CONCLUSIONS: This study suggests that the training of prediction models on parental populations can predict the genetic merit of the progeny with high accuracy: an encouraging result for the implementation of GS in the maritime pine breeding program.

Genomic selection in maritime pine.

A two-generation maritime pine (Pinus pinaster Ait.) breeding population (n=661) was genotyped using 2500 SNP markers. The extent of linkage disequilibrium and utility of genomic selection for growth and stem straightness improvement were investigated. The overall intra-chromosomal linkage disequilibrium was r(2)=0.01. Linkage disequilibrium corrected for genomic relationships derived from markers was smaller (rV(2)=0.006). Genomic BLUP, Bayesian ridge regression and Bayesian LASSO regression statistical models were used to obtain genomic estimated breeding values. Two validation methods (random sampling 50% of the population and 10% of the progeny generation as validation sets) were used with 100 replications. The average predictive ability across statistical models and validation methods was about 0.49 for stem sweep, and 0.47 and 0.43 for total height and tree diameter, respectively. The sensitivity analysis suggested that prior densities (variance explained by markers) had little or no discernible effect on posterior means (residual variance) in Bayesian prediction models. Sampling from the progeny generation for model validation increased the predictive ability of markers for tree diameter and stem sweep but not for total height. The results are promising despite low linkage disequilibrium and low marker coverage of the genome (∼1.39 markers/cM).

Genome-wide distribution of genetic diversity and linkage disequilibrium in a mass-selected population of maritime pine.

BACKGROUND: The accessibility of high-throughput genotyping technologies has contributed greatly to the development of genomic resources in non-model organisms. High-density genotyping arrays have only recently been developed for some economically important species such as conifers. The potential for using genomic technologies in association mapping and breeding depends largely on the genome wide patterns of diversity and linkage disequilibrium in current breeding populations. This study aims to deepen our knowledge regarding these issues in maritime pine, the first species used for reforestation in south western Europe. RESULTS: Using a new map merging algorithm, we first established a 1,712 cM composite linkage map (comprising 1,838 SNP markers in 12 linkage groups) by bringing together three already available genetic maps. Using rigorous statistical testing based on kernel density estimation and resampling we identified cold and hot spots of recombination. In parallel, 186 unrelated trees of a mass-selected population were genotyped using a 12k-SNP array. A total of 2,600 informative SNPs allowed to describe historical recombination, genetic diversity and genetic structure of this recently domesticated breeding pool that forms the basis of much of the current and future breeding of this species. We observe very low levels of population genetic structure and find no evidence that artificial selection has caused a reduction in genetic diversity. By combining these two pieces of information, we provided the map position of 1,671 SNPs corresponding to 1,192 different loci. This made it possible to analyze the spatial pattern of genetic diversity (He) and long distance linkage disequilibrium (LD) along the chromosomes. We found no particular pattern in the empirical variogram of He across the 12 linkage groups and, as expected for an outcrossing species with large effective population size, we observed an almost complete lack of long distance LD. CONCLUSIONS: These results are a stepping stone for the development of strategies for studies in population genomics, association mapping and genomic prediction in this economical and ecologically important forest tree species.

Systems biology of lignin biosynthesis in Populus trichocarpa: heteromeric 4-coumaric acid:coenzyme A ligase protein complex formation, regulation, and numerical modeling.

As a step toward predictive modeling of flux through the pathway of monolignol biosynthesis in stem differentiating xylem of Populus trichocarpa, we discovered that the two 4-coumaric acid:CoA ligase (4CL) isoforms, 4CL3 and 4CL5, interact in vivo and in vitro to form a heterotetrameric protein complex. This conclusion is based on laser microdissection, coimmunoprecipitation, chemical cross-linking, bimolecular fluorescence complementation, and mass spectrometry. The tetramer is composed of three subunits of 4CL3 and one of 4CL5. 4CL5 appears to have a regulatory role. This protein-protein interaction affects the direction and rate of metabolic flux for monolignol biosynthesis in P. trichocarpa. A mathematical model was developed for the behavior of 4CL3 and 4CL5 individually and in mixtures that form the enzyme complex. The model incorporates effects of mixtures of multiple hydroxycinnamic acid substrates, competitive inhibition, uncompetitive inhibition, and self-inhibition, along with characteristic of the substrates, the enzyme isoforms, and the tetrameric complex. Kinetic analysis of different ratios of the enzyme isoforms shows both inhibition and activation components, which are explained by the mathematical model and provide insight into the regulation of metabolic flux for monolignol biosynthesis by protein complex formation.

Genomic estimated breeding values using genomic relationship matrices in a cloned population of loblolly pine.

Replacement of the average numerator relationship matrix derived from the pedigree with the realized genomic relationship matrix based on DNA markers might be an attractive strategy in forest tree breeding for predictions of genetic merit. We used genotypes from 3461 single-nucleotide polymorphism loci to estimate genomic relationships for a population of 165 loblolly pine (Pinus taeda L.) individuals. Phenotypes of the 165 individuals were obtained from clonally replicated field trials and were used to estimate breeding values for growth (stem volume). Two alternative methods, based on allele frequencies or regression, were used to generate the genomic relationship matrices. The accuracies of genomic estimated breeding values based on the genomic relationship matrices and breeding values estimated based on the average numerator relationship matrix were compared. On average, the accuracy of predictions based on genomic relationships ranged between 0.37 and 0.74 depending on the validation method. We did not detect differences in the accuracy of predictions based on genomic relationship matrices estimated by two different methods. Using genomic relationship matrices allowed modeling of Mendelian segregation within full-sib families, an important advantage over a traditional genetic evaluation system based on pedigree. We conclude that estimation of genomic relationships could be a powerful tool in forest tree breeding because it accurately accounts both for genetic relationships among individuals and for nuisance effects such as location and replicate effects, and makes more accurate selection possible within full-sib crosses.

The distribution of genetic parameter estimates and confidence intervals from small disconnected diallels.

The distributions of genetic variance components and their ratios (heritability and type-B genetic correlation) from 105 pairs of six-parent disconnected half-diallels of a breeding population of loblolly pine (Pinus taeda L.) were examined. A series of simulations based on these estimates were carried out to study the coverage accuracy of confidence intervals based on the usual t-method and several other alternative methods. Genetic variance estimates fluctuated greatly from one experiment to another. Both general combining ability variance (sigma(2) (g)) and specific combining ability variance (sigma(2) (s)) had a large positive skewness. For sigma(2) (g) and sigma(2) (s), a skewness-adjusted t-method proposed by Boos and Hughes-Oliver (Am Stat 54:121-128, 2000) provided better upper endpoint confidence intervals than t-intervals, whereas they were similar for the lower endpoint. Bootstrap BCa-intervals (Efron and Tibshirani, An introduction to the bootstrap. Chapman & Hall, London 436 p, 1993) and Hall's transformation methods (Zhou and Gao, Am Stat 54:100-104, 2000) had poor coverages. Coverage accuracy of Fieller's interval endpoint(J R Stat Soc Ser B 16:175-185, 1954) and t-interval endpoint were similar for both h(2) and r(B) for sample sizes n