Yousry El-Kassaby

Breeding conifer species for phenotypic improvement is challenging due to delayed expression of important phenotypes related to productivity and their late sexual maturity, causing long recurrent selection cycles. Genomic selection (GS) can address such shortcomings through early prediction of phenotypes based on large numbers of jointly considered genomic markers, typically, single nucleotide polymorphisms (SNPs). Additionally, current conifer breeding genetic evaluations are based on pedigree-based predictions. However, the maximization of genetic gain in breeding programs is contingent on the accuracy of the predicted breeding values and precision of the estimated genetic parameters, which can also be improved using GS.While GS has become a new paradigm in animal breeding, it is still in its infancy for tree improvement. Thus, GS requires validation before it can be operationally implemented. Collectively, this dissertation explores some of the challenges associated with the application of GS in forest tree improvement programs. Namely, the efficiency of GS compared to traditional phenotypic selection, methods to implement GS in a cost-efficient manner, and the prediction accuracy (PA) of phenotypes across generations, life-stages, and environments. To address thesechallenges I structured this dissertation into three analyses which use several GS methodologies,three genotyping platforms, and three conifer species.The first study explores the temporal decay and relative efficiency of GS PA for interiorspruce (Picea engelmannii × glauca). The second study investigates the use of single-step GS(ssGBLUP) to improve the precision and accuracy of genetic parameter estimates for white spruce(Picea glauca). The third study focuses on the combined use of ssGBLUP and climate data toimprove intra- and inter-generation PA in unobserved environments for Douglas-fir (Pseudotsugamenziesii). The results from these three studies demonstrated that: i) updating GS models requiresivphenotypic data at least mid-rotation age to accurately reflect mature growth traits, ii) the relativeefficiency of GS is greater than traditional selection assuming a 25% reduction in breeding cyclelength, iii) ssGBLUP is an effective tool for improvement in the genetic evaluation of openpollinated mating designs, and iv) inclusion of climate variables as environmental covariates in the GS models yields improvement in PA for unobserved environments.

View record

This study utilized both of leaf dimension measurements (leaf area, length, maximal width, and the ratio of length: width) and 12 microsatellite loci screened across 315 samples of the entomophilous mangrove species Avicennia marina (Forsk.) Vierh. The samples collected from 9 sites along the Saudi Arabian Red Sea coastline with an estimated sampling range of 1,345 kilometres. The study objectives were to examine the genetic diversity, population structure, and the field observed phenotypic leaf variation, and to inspect the influence of the distribution limit on the genetic compositions and the phenotypic leaf traits variation. The 12 loci detected a total of 89 alleles with an average allelic diversity of 7.42. Observed heterozygosity (Hₒ) was close to expected heterozygosities (Hₑ) for most sites, and the average (Hₒ) was 0.298. The levels of inbreeding ranged from negative 0.044 to positive 0.126, with an average inbreeding coefficient of 0.012. The component of variation among populations were (25%, 34%) (p
View record

Conventional tree breeding productivity (especially in conifers) is primarily constrained by late expression of commercially important traits, and late onset of sexual maturity. These characteristics conjointly correspond to lengthy testing phases prior to selection, which in turn restrains the accumulation of genetic gain. GS has the potential to increase genetic gain per unit time by allowing for the prediction and selection of traits at an earlier age.This dissertation investigates some facets of GS, specifically in relation to Douglas-fir (Pseudotsuga menziesii Mirb. (Franco)), and ‘real-world’ applications. Expressly: to compare pedigree based ABLUP and two GS methods; assess GS prediction accuracy over spatial and temporal deviations; validate the use of exome capture as a cost-effective genotyping platform for use in GS; use of GS to predict breeding values in the next generation; investigate the effect of relatedness on GS prediction accuracy; and to assess the number of makers required for GS in conifers. Chapter 2 utilizes exome capture as a genotyping platform to assess height and wood density GS prediction accuracies across space and time. A cross-generational GS analysis was performed in Chapter 3 using a progeny generation as an independent validation set. Chapter 4 investigates the effect of marker density on GS predictive accuracy in Douglas-fir and Interior spruce (Picea glauca (Moench) Voss x Picea engelmannii Parry ex Engelm.). The overriding conclusion, is that while some of the GS models’ prediction accuracies were high, the main driving force was the tracking of relatedness rather than LD. However with regard to the ability of the available markers to track pedigree, exome capture was found to be very competent. Knowing this, the following trends were observed: GS models performed similarly and were comparable to ABLUP; genotype x environment interactions are an important consideration for GS spatial analyses; height at 12 years was deemed an acceptable age at which accurate predictions can be made concerning future height and wood density; moderate to high cross-generational GS prediction accuracies were obtained, but were influenced by the relationship between training and validation sets; and increasing marker number increases GS prediction accuracy, for Douglas-fir and Interior spruce.

View record

Tree improvement programs are long-term and resource-demanding endeavors consisting of repeated cycles of breeding, testing, and selection and suffer from protracted testing phases. Phenotypic selection is commonly practiced and often requires trees reaching certain age and/or size resulting in slow accumulation of genetic gain. Open-pollinated (OP) family testing is the simplest and most economical means for screening, evaluating, and ranking large number of candidate parent trees but suffers from inflated additive genetic variance and heritability estimates. This dissertation investigates genomic selection (GS) and its applicability to forestry in selection and progeny testing evaluation.To address these two applications, I studied yield and wood traits from two white spruce populations, genotyped using Genotyping-by-Sequencing and SNPs array. I investigated the applicability of GS using the Ridge Regression Best Linear Unbiased Predictor (RR-BLUP) and the Generalized Ridge Regression (GRR)) algorithms and validated the derived predictive models in space across three progeny testing sites in interior British Columbia. Moreover, using principal component analysis (PCA), I fitted a multi-traits GS predictive model to address the inter-correlation among the studied attributes. Additionally, the Genomic Best Linear Unbiased Predictor (GBLUP) was used in genetic variance decomposition framework to unravel additive from non-additive genetic variances and I compared the results to that from the traditional pedigree-based (ABLUP) analysis. Differences between the RR-BLUP and GRR predictive models’ accuracies were observed indicating that the studied attributes’ genetic architecture is complex. Validating the GS’s predictive models in space clearly confirmed multi- to single-site superiority as they account for the genotype x environment interaction, commonly observed in forestry evaluation trials. When PCA scores used as multi-trait representatives, GS prediction models produced surprising results where the concurrent selection of negatively correlated traits such as wood density and growth is possible. The genetic variance decomposition indicated that the genomic-based approach outperformed that of the pedigree-based with the successfully separation of additive from non-additive genetic effects. This approach was demonstrated in a single- and extended to multi-site scenario, propelling OP testing to the forefront of forest trees genetic evaluation. In general, the effectiveness of GS was clearly demonstrated as an alternative selection and evaluation method.

View record

Life-history traits, known as fitness components, are related to the timing and success of development, reproduction, and senescence throughout the life cycle. Selection in variable environments may favor plants to defer germination until suitable conditions occur. Seed dormancy is an innate constraint on germination timing and prevents germination during periods that are ephemerally favorable. The timing of seed germination is the earliest life-history trait that is expressed and sets the context for the traits that follow. As such, seed dormancy may be construed as an adaptation for survival during bad seasons and can exert cascading selective pressures on subsequent life stages. Seed size is another important life-history trait linking the ecology of reproduction and seedling establishment with that of vegetative growth. As the two traits are, at modulations, regulated by hormone signaling cascades, evolve under correlated selective pressures, and exhibit co-varying phenotypes, this dissertation intended to elucidate their eco-evolutionary dynamics and possible genetic basis of adaptation. From an eco-evolutionary perspective, I demonstrated that dynamic climatic variables rather than constant geographic variables are the true environmental driving forces in seed dormancy and size variations in Pinus contorta Dougl. Evapotranspiration and precipitation in the plant-to-seed transition are the most critical climatic variables for seed dormancy and size variations, respectively. Unlike random temperature fluctuations between generations, wide temperature shifts considerably alter population structures and accelerate life-history evolution. Regarding the genetic basis of adaptation, environmental cues trigger different seed-set programming in Picea glauca and Arabidopsis by employing lineage-specific and deeply conserved microRNAs at different expression levels, respectively, to entrain phenotypical variations, such as dormancy intensities. Our findings additionally point to auxin as a key player that likely works in conjunction with the ABA and GA signal pathways previously investigated in mechanisms underpinning the seed-to-plant transition by chilling in Picea glauca seeds. This dissertation increases our understanding of plant evolution and persistence in the context of climate change and provides fundamental insight for understanding how microRNAs are at play in seed-set programs to regulate phenotypes, how winter chilling contributes to the timing of phenology, and how conifer life histories may develop under new climate scenarios.

View record

Seed orchards represent the link between breeding and silvicultural activities. They were expected to act as closed, panmictic populations in Hardy-Weinberg equilibrium, meaning that desired genes drafted during previous selection stage would be effectively transmitted from parental to offspring populations; however, extensive research has indicated that this expectation is not met. Scrutinizing seed orchards’ efficiency is of vital importance as it determines the genetic quality of future forest stands.Population genetics of four tree species’ seed orchards (western larch, Douglas-fir, lodgepole pine, and western redcedar) was studied using microsatellite DNA markers. Partial (family array) and full (bulk seed) pedigree reconstruction of offspring population (seed crops) were conducted using the likelihood-based parentage inference program CERVUS to estimate parental reproductive success, selfing rate, pollen contamination, effective number of parents (Ne), and seedlot genetic worth. Several simplified methods for predicting seed crops’ genetic quality and quantity were evaluated by comparing parental reproductive success with parental fecundities.In all species, the top 20% of males contributed approximately one half of successful within-orchard pollen, substantially reducing male Ne (45 to 62% of the orchards’ census numbers). Even larger distortion was observed among females (the top 20% of females produced 77% of seed crop in Douglas-fir), reducing female Ne to as little as 13% of the census. Selfing and pollen contamination rates were in the range of previously reported studies, with the exception of high (15.2%) and low (7.3%) selfing rates in Douglas-fir and western redcedar, respectively. Pollen bud production and seed-cone volume were found to be the most reliable proxies to parental reproductive success, genetic worth, and Ne estimates.An optimization protocol was developed for creating custom seedlots with maximized genetic gain at any Ne while collectively considering parental male and female fecundities, co-ancestry among parents, inbreeding, and variation in seed germination capacity. This protocol can be utilized in any generation’s seed orchard e.g. when seed supply exceeds demand, for mixing surpluses from multiple years, or if a given seed lot fails to meet minimum Ne requirements.

View record

The economic, social and biological value of our forests makes their sustainability essential to our well-being. To ensure their long-term health, it is critical to regularly and effectively monitor their inhabitants, as well as to detect non-indigenous species early and accurately. These programs rely on the precise diagnosis of species, which can be complicated for terrestrial arthropods by sizeable trap samples, damaged specimens, immature life stages and incomplete taxonomy. The recent advent of DNA barcoding, a technique that differentiates species using sequence variation in a standard gene region, shows tremendous promise for circumventing these obstacles. This dissertation evaluates the integration of barcoding into forest arthropod biomonitoring and biosurveillance programs with several investigations of nocturnal moths (Lepidoptera) in British Columbia, Canada. Barcode reference libraries are constructed for looper moths (Geometridae) and Lymantria (Erebidae) tussock moths, and are determined to successfully discriminate species in over 93% and 97% of cases, respectively. The libraries demonstrate how barcoding might enhance biosurveillance programs by flagging two new records for geometrid moths, and by successfully diagnosing 32 intercepted tussock moth specimens. These two libraries, and a multi-gene phylogeny constructed for Geometridae, are used to conduct faunal inventories in modified forest systems, and investigate the influence of disturbance on three levels of moth diversity—species, genetic, and phylogenetic. A first level inventory of Stanley Park, Vancouver, produces a preliminary list of 190 species, the detection of four new exotic species, and the discovery of two potentially cryptic species. Surveys conducted across several harvest treatments at two silvicultural research forests display no evidence of increased diversity at intermediate disturbance levels, but do reveal a correlation between species and genetic diversity. And lastly, three levels of moth diversity are estimated in ponderosa pine systems that differ widely in attack by Dendroctonus bark beetles, and demonstrate a negative association between species diversity and tree mortality. In combination, all projects suggest that DNA barcoding provides several advantages over traditional biosurveillance and biomonitoring, including the ability to rapidly sort specimens, a reduction in specialist time, the detection of species at low density, and the ability to appraise multiple levels of diversity.

View record

Genetic diversity is a key factor in species survival, evolution, and adaptation. It also reveals species genetic structure and provides insights into how different demographic forces shape species genetic variability. Although, black cottonwood (Populus trichocarpa Torr. & Gray) is the first tree to have its genome completely sequenced; however, information regarding its natural genetic diversity and population structure is lacking. I have investigated the extent of genetic diversity within and among 38 natural populations of P. trichocarpa sampled across British Columbia using 10 nuclear (nuSSR) and 12 chloroplast microsatellite (cpSSR) markers.CpSSR represents two haplotypes, clustering as northern and southern groups; however, aBayesian population structure analysis suggested the presence of three highly admixed groups supported by low population differentiation (low FST and RST). Monmonier’s spatial analysis suggested the presence of one genetic discontinuity dividing the studied area into northern and southern regions. These findings indicated that P. trichocarpa might have originated from two, northern and southern, glacial refugia that have experienced moderate contact through extensive gene flow. Nucleotide diversity for 10 candidate-gene loci involved in adaptive, defence, and housekeeping functions was abundant and varied across loci, with the majority showing neutral variations. Linkage disequilibrium (LD), decays rapidly to r² ≈ 0.18 within 700 base pairs (bp).Comparing the nucleotide diversity between P. trichocarpa and P. balsamifera L. to the EurasianP. tremula L. indicated that the two North American species had lower diversity (θw range 0.002 to 0.004) than the Eurasian poplar (θw = 0.005). The estimated time of divergence between the two North American and the Eurasian species indicated that the latter was five- to six-fold older compared to the two former species. The substitution rate was lower in North American species(0.4 x 10-⁸ per year) compared to the Eurasian poplar (2 x 10-⁸). Different association genetics models produced strikingly different results after the inclusion or exclusion of population structure, highlighting the importance of proper model construction.

View record