Gain and Diversity in Multi-Generation Breeding Programs

Gain and Diversity in Multi-Generation Breeding Programs

Doctoral Thesis

Umeå 2019

ISSN 1401-6230

ISBN 91-576-5629-0

Abstract

Gain and diversity in multi-generation breeding programs. Doctors Dissertation.

ISSN 1401-6230, ISBN 91-576-5629-0

Progress in tree improvement comes from utilizing the genetic diversity found in unimproved forests. The balance between genetic gain and diversity is one of the most important considerations for all breeders. The sustainability in gain extraction over generations of the breeding program should be considered from its start. This thesis examines various strategies for selection in breeding. Using coancestry and its derivatives as a diversity measure, different methods are contrasted and compared for their efficiency in giving response to selection, considering the subsequent change in genetic diversity. It is concluded that restricted and unrestricted phenotypic selection and restricted (individual and family) index-selection, using data taking the performance of relatives into account, are fairly equal in terms of selection efficiency. However, a genuine and substantial improvement in selection response can be achieved by incorporating information on the population structure into the selection criterion. A possible way to enhance the efficiency in realized gain per unit decrease in diversity is to balance selection with relatedness. This can be seen as maximizing the allele containing capacity of the breeding population with regards to constraints on relatedness. Two ways to accomplish this, group merit selection and a linear programming method optimizing gain with a restriction on group coancestry are exemplified in this thesis. Benefits from coancestry-controlled selection are far from negligible, and can have a role to play in tree improvement. The breeding population should be seen as a dynamic entity regarding size and family contributions over time. A decision-model for infusion of fresh unrelated material is presented. The conclusion is that there often could be a place for refreshing the breeding population with new selections the first cycles of breeding. The diversity of regeneration material affects forests over the long term. It is concluded that diversity of species undergoing domestication must be monitored, with comparable measures throughout the whole breeding program, including seed producing stages.

Key words: breeding, coancestry, genetic diversity, selection, status number,

tree improvement.

Author’s address: Pranav Ahluwalia, Village Lauhat,Tehsil- GhumarwinDistrict- BilaspurHimachal Pradesh

To my family

Contents

Introduction

Goals of tree improvement

Problem

Objective of the thesis

Theoretical considerations and applications

The breeding population

Measures of diversity and effective population size

Coancestry-related diversity measures

Coancestry and its implications for selection

Effects of selection

Selection

Genetic improvement of desired traits

Genetic variance

Inbreeding

Drift occurs in small populations

Breeding efficiency

Stochastic simulation

Mating

Results and discussion

Summary of publications

Phenotypic versus Index selection

Coancestry-controlled selection

Infusion of unrelated material

Group coancestry as diversity measure

Accumulation of inbreeding

Time factor is not regarded

Breeding under tight constraints

Conclusions

Suggestions for future research

Literature cited

Acknowledgements

Introduction

A legitimate expectation of forest tree improvement is that erosion of genetic variability should be assayed and controlled. An obvious purpose (or constraint) of tree improvement is to prevent, or at least delay, the complications following management of small populations (e.g., Namkoong 1982, 1984; Ledig 1992; Wei 1995a; Yanchuck and Lester 1996). The goal over time can be set to get more progress in exchange for less diversity – sooner – without impairing the present value of the breeding program. For the manager of a tree improvement program, it would be the simplest thing to erode diversity for rapid genetic advancements, and the chief difficulty lies in optimizing breeding and selection activities so that genetic variation is maintained in the population (e.g., Kerr et al. 1998).

Tree breeding consists of several activities, of which selection for the next generation is the most apparent. Selection is a vital part of plant and livestock improvement, and has been the subject of many studies suggesting procedures to enhance genetic gain (e.g., Bos and Caligari 1995). Testing procedures for prediction of breeding values and estimating genetic variance are of great importance for decisions in a breeding program (e.g., Williams and Matheson 1994). A number of actions influence the rate of progress and the accuracy of decisions made. In tree breeding, accurate breeding values are time consuming to obtain because of the time required for trees to reach economic maturity and because of the high cost of establishing and maintaining progeny in precise field tests. Differences between testing environments and target environments for the improved regeneration material continues to challenge breeders.

Genetic variation in the breeding population is the raw material for long-term breeding progress (e.g., Kang and Namkoong 1988), and, at the same time, it can provide superior genotypes to improve selection outcome (El-Kassaby 1992). The high variability in the breeding population is crucial for the formation of superior seed orchards with low relatedness among seed orchard trees (El-Kassaby 1995; Rosvall et al. 1999).

Since generation turnover is exceptionally long in conifers, a successful approach to effective tree improvement requires effective utilization of the genetic variability at hand (El-Kassaby 1992) and the creation of new variation by recombination. In the search for an adequate selection criterion, the realized genetic gain per unit diversity sacrificed is indeed an important target. This target deserves to be given more attention in the optimization of selection procedures (e.g., Toro and Pérez-Enciso 1990; Verrier etal. 1993; Wei and Lindgren 1994; Wray and Goddard 1994; Villanueva and Woolliams 1997; Lindgren and Mullin 1997). In order to accomplish this, all breeding activities must be considered, from selection and testing of plus trees, which basically is sampling genes for recombination (Cornelius 1994; III; IV), to the survival improved seedlings in the field (Ackzell and Lindgren 1994), as must time (Cotterill et al 1989) and cost (Lindgren et al. 1997a).

Most often it is assumed that there is no more refined information, such as information from genetic markers, on specific genotypic constitution available for breeding decisions apart from the breeding values or the phenotypic values and the relatedness between individuals subject to selection (Wei 1995a, and references therein). Other information affecting selection decisions is thinkable, but is generally incorporated as breeding population size, etc. For some time now, information derived from molecular data has had promising prospects (O’Malley and McKeand 1994; Williams and Hamrick 1995), but there are few examples of successful implementations in forest tree breeding, at least for quantitative traits, and there are reasons to believe that practical obstacles are not easily overcome (Strauss et al. 1992; Szmidt and Wang 1998).

Goals of tree improvement

We could categorize the purposes of forest tree improvement into four major goals:

1. Provide seed with a suitable physiology. Seed should be mature, have high germination percentage, and superior growth energy (El-Kassaby 1995). While seed orchard design and management are important factors in this respect, fertilization and horticultural practices can improve results. Selection criteria can include the ability to produce seed of acceptable quality.

1. Ensure adaptability of regeneration material. In many planting projects, field-survival is a key-trait of regeneration material (Fries and Lindgren 1986). The ability to survive critical abiotic factors is a variable character that shows a geographic pattern (e.g., Sorensen 1992; Xie and Ying 1993; Persson 1994). The rules of provenance transfer constructed for Sweden are of help when regional breeding populations are assembled.

1. Improve the genetics of commercial traits (e.g., White et al 1993, Wei 1995a). Progress in breeding is dependent on how well the material is known and how well this knowledge is incorporated into decisions made in the breeding program. Estimates of genetic variance, and thus the potential for progress, are improved by proper design of experiments (Williams and Matheson 1994).

1. Conserve genetic diversity in wild and improved forests (Namkoong 1984; Eriksson et al. 1993, 1995). The inevitable loss of genetic diversity following domestication must not be so severe as to reduce the adaptive potential of the improved material (White et al. 1993; Namkoong et al. 1988). Along with genetic improvement, there is a risk that genetic destitution will be the consequence following rather quickly from radical or shortsighted selection (Wei 1995a; Andersson et al. 1998a).

The two first objectives are well foreseen in sub-arctic and temperate areas by contemporary tree improvement. Tree improvement activities are directed by immediate needs, and today’s improved seeds posses the required ecological competence in areas where the climate is harsh and forest operations depend on a reliable regeneration. This is the case, at least regarding conifers in boreal and temperate areas (e.g., Savolainen 1996). Where not available, current methods appear to be reassuring to work out transfer rules.

Although large efforts have been invested in research, documentation and analysis, much remains to be done to fill the most immediate knowledge gaps in the aforementioned third and fourth breeding goals. These two objectives are the focus of much of today’s research and applied breeding. The third objective is complicated by the fact that desirable traits are often conflicting (i.e., wood density and growth rate) and by the fact that the biological mechanisms behind many traits are not readily understood. Objective number four has been brought into greater focus lately, promoted by development of new techniques and by a growing public concern (Ledig 1992, Szmidt and Wang 1998).

The impact on the potential adaptability of domesticated reforestation material, compared to unimproved material, has become the scope of investigation (Yang and Yeh 1992, Yanchuck and Lester 1996). Breeding programs are designed to assure inclusion of even rare alleles, and to maintain levels of heterozygosity (El-Kassaby and Ritland 1996). The results are not convincing that reforestation with improved material is accompanied by detrimental effects on forests and forest ecosystems (Stoehr and El-Kassaby 1997). On the contrary, improved survival rates, vigor and growth characterize many forests planted with improved material (e.g., Savolainen 1996) due to the increased diversity in early generation seed orchard seed (Szmidt and Muona 1985).

The diversity of many commercially interesting species have been described by a wealth of molecular techniques (e.g., Hamrick and Godt 1990; Williams and Hamrick 1995; Szmidt and Wang 1998, and references therein). Development of molecular markers and population genetics theory enables advances in this field, but much development of theory and tools remains to be done (Szmidt and Wang 1998). Other drawbacks are that molecular methods are too imprecise for application to most specific breeding management problems (Strauss et al 1992).

Problem

Work optimizing gain and diversity has made clear that the tradeoff between the two is the most important focus of breeding (e.g., Dempfle 1975; Wei 1995a,b; Brisbane and Gibson 1995; Zheng et al 1997; I-III). It is substantially easier to select for a single objective or an index criterion (e.g., Namkoong1970; Baker 1986; Kang and Namkoong 1988; Bos and Caligari 1995), than to optimize selection for contradictory goals (e.g., Quinton et al 1992; Namkoong 1982). To select simultaneously for both gain and diversity is a good example of such a dilemma. It becomes even more complex when a compromise between long-term goals and short-term goals is sought (Wei 1995a). Breeding programs, and consequently breeding populations, are dynamic entities where the qualities of the resulting regeneration material are the product of many decisions over a long period. A dynamic decision system in the breeding program, gradually incorporating new knowledge, will be even more important in advanced generations (Wray and Goddard 1994a; Kerr et al. 1998).

The main focus of this thesis is the complex issue of breeding efficiency in tree improvement and possible means towards its enhancement. The question arising is how an efficient breeding, accounting for all steps in a breeding program such as selection of plus trees, mating and selection should be defined. By efficiency is meant the level of realized gain compared to the amount of genetic variability lost in the process. Diversity Use Efficiency (DUE) used in II will be synonymous to breeding efficiency in most cases, but herein breeding efficiency is used to indicate that all steps in the breeding process should be included when evaluating strategy options in forest tree improvement.

Objective of the thesis

The objectives of this thesis were: (a) to construct a model describing the compromise between gain and diversity in multi-generation breeding, and (b) to develop methods of combining these goals in forest tree improvement.

Theoretical considerations and applications

Studies presented in this thesis share a common approach to some central issues in forest tree improvement. An attempt is made to discuss the meaning of some key concepts and their use throughout this work. The genetic status of a breeding program is influenced by many factors such as phenotypic and genotypic variability, allelic diversity and heterozygosity (e.g., Kitzmiller 1990; El-Kassaby 1992). These factors can be summarized as genetic diversity in a broad sense and the focus can vary depending on the viewpoint of the observer.

The genetic model frequently used in tree breeding is the infinitesimal genetic model.The basic idea is that the genotype is formed by an infinite number of individually inseparable genes, each with an infinitesimally small genotypic effect, and each contributing only a very small fraction of the genetic variance by the random, or random-like, transmission of alleles from generation to generation (Fisher 1918; Wright 1921). In breeding of commercial species, the infinitesimal genetic model has been, and still is, the prevailing concept dominating the theory of selection (Wricke and Weber 1986; Falconer and MacKay 1996; Lynch and Walsh 1998). Traits of commercial interest are generally considered to show a quantitative inheritance (e.g., Lynch and Walsh 1998). The use of allelic models is rare in practical breeding, and in this thesis, no direct consideration is given to allelic inheritance. Indirectly, group coancestry-derived diversity estimates presented herein are convertible and equally applicable as probabilities for transmission of alleles neutral to selection.

The model [1] used in publications I-IV is the quantitative genetic model of composite gene action with no regard to interactions among genes, where: Pis a population mean; xi is the deviation of a family from the mean of the population; and xi,j is the deviation of an individual from the mean of its family. The sum of these three terms represent the genetic value of an individual, Gi,j:

[1]

The phenotypic variance observed for the population is constituted by the within- and among-family variances

[2]

The genetic variance is simplified, consisting only of variation in additive gene effects; dominance effects and epistatic effects are considered to be absent. The values chosen for different variance components in I-IV reflect plausible heritabilities commonly reported from tree improvement programs.

In unrestricted phenotypic selection (PS, selection for phenotypic value), individuals are ranked on the basis of their phenotypic value, and the best are chosen irrespective of their relationship to other selected individuals. In restricted combined-index selection (CIS, selection for breeding value), estimates of individual breeding values are made more accurate by taking into account the performance of relatives, while the numbers of selected relatives is limited to some preset number.

In the genetic model used in publications I-IV, family means are given by the mean value of their progeny (Falconer and MacKay 1996). Mortality in the field, pests or similar catastrophic events are not specifically considered. Evaluations are generally conducted on predictions for a single quantitative trait, which can itself be a weighted index composed of several quantitative traits (e.g., a tree’s volume, straightness and health can be combined in a single breeding value).

In real-life breeding, individuals are not selected or rejected on the basis of their superiority or inferiority for height or growth. The basis for selection is sometimes modified by categorical assessments, so that individuals or families with inferior qualitative attributes, or prone to damage from pest or disease, are rejected. This negative selection is generally conducted without being specifically incorporated into selection indexes (Zobel and Talbert 1984).

The breeding population

By ”breeding population” is generally meant the individuals that contribute as parents to the next generation (e.g., Zobel and Talbert 1984), though it may be confused with the set of potential parents that still have chance to contribute with their genes. When breeders talk in general terms about the breeding population, they may also be referring to potential candidates for the next generation, prior to selection (e.g., Cotterill 1986).

A breeding population may or may not be structured. In some programs, individuals may form a single group of inter-mating parents (Zobel and Talbert 1984; Cotterill 1986). Alternatively, the population may be divided into sub-populations, or sublines (Burdon and Namkoong 1983; McKeand and Beineke 1980). Subdivision may be random, or on the basis of geographic origin or breeding value, for the purpose of managing genetic diversity (Williams et al 1995; Eriksson et al. 1993, 1995). The sub-lines may be bred to serve a specific purpose (e.g., Namkoong 1984; Namkoong et al. 1988; Burley 1994), or they may be randomly assigned replicates for the purpose of managing inbreeding in seed orchards (Lowe and van Buijtenen 1986; van Buijtenen and Lowe 1979). Other concepts for structured breeding populations include ”nucleus” breeding (Namkoong et al. 1988), and ”elite-line” breeding (Cotterill et al. 1989). The objective may be to retrieve genetically superior individuals from the elite (Namkoong 1982), while maintaining the necessary diversity in the main population.

The size of the breeding population is generally limited by budget constraints (Lindgren 1997a). The idea of a larger breeding population in the initial phase has been suggested in tree improvement by Lindgren et al (1997a) and for animal breeding by Verrier et al. (1993), among others. This concept is investigated further for a specific target level of group coancestry in publication III, where the concept of a variable breeding population size is discussed. The idea is to determine the level of group coancestry acceptable over the long term (Kerr et al 1998; Rosvall et al. 1999), and then to optimize the breeding population size accordingly.