Genomic selection in Picea abies

There are high expectations for tree breeding as a method for developing regeneration materials for forestry, which contributes to faster climatic adaptation, high biomass production and CO2 sequestration, high quality, and good health. New technology and knowledge about tree genomes provide Skogfrøverket, as a supplier of genetic materials for forestry, with new opportunities to develop competitive products in the form of seed lots with higher genetic quality. Genomic selection is a cornerstone in breeding livestock for food production, and the method is increasingly being used in plant breeding for crops. The method utilizes models where phenotypic (measured) traits are explained by variation in genetic markers to predict individuals' performance for traits important for quality, health and economic value. When the genomic models are well "trained" with phenotypic data, predictions can eventually be made without measuring the traits on the individuals themselves. In forest production, important traits like wood quality, disease resistance, and sustained growth and production, appear late in the rotation. These important traits are controlled by many genes spread across the genome, each showing relatively small effects. This means that genomic prediction should be able to give breeders good opportunities to increase the frequency of favorable alleles for production traits without compromising the preservation of genetic variation. The goal of the project was therefore to develop genomic prediction as a method for selection and control of genetic variation in the breeding of spruce, the most important tree species for commercial use in Norway. Tree breeding that in its operational daily work utilizes genomic data must be built on a well-developed workflow with protocols from sample collection, DNA extraction, genotyping, organization and quality assurance of data, to genomic analyses, and prediction of genomic breeding values. Central to the workflow is the breeding database where all data are organized and linked. This, that was established early in the in collaboration with BioBank AS in Hamar, is now fundamental infrastructure for all projects and operational breeding that involves genotyping at Skogfrøverket. Genomic prediction was conducted in two experimental series within the breeding zone for the lowlands in Eastern Norway. The results from the project are now being used to select individuals for the inclusion in the breeding population and as parents in the new seed orchards being established. The method for genomic prediction that was evaluated and implemented is based on the genomic relationship matrix (G matrix). The G matrix between all individuals in the relevant population is estimated based on the variation in approximately 39,000 single nucleotide polymorphisms (SNPs) from a SNP array developed at SLU in Sweden. Relationships and "genomic similarity" are then estimated accurately, so that even variation due to the "Mendelian sampling error term" of chromosomes during meiosis is captured. Previously, plant breeders had to rely on the so-called A matrix: the relationship between all individuals based on the crossing design, where 0 values refer to unrelated individuals, 0.25 to half-siblings, and 0.5 to full-siblings. In a G matrix, relationships vary around these mean values, but they contain more information about the genetic similarity of the genotypes. Prediction was carried out by explaining the observed traits, the phenotypes, using the G matrix and factors that explain environmental variation in the stand. The result was compared with models where the G matrix was replaced with the A matrix. The comparison of models with G and A matrix shows that the variation between trees for traits such as height growth and stem diameter, which usually show low heritability, is better explained by the G matrix, and that there is a clear gain in using this for calculating breeding values. For wood density, a trait with high heritability, the gain was smaller. The G matrix can therefore strengthen the selection from progeny tests that already exist in the breeding program if the individuals are both phenotyped and genotyped. To evaluate the strength of estimating the breeding values of non-phenotyped individuals from the same or a related population, cross-validation was carried out, where parts of the dataset were predicted based on the rest of the dataset. The predictive ability, the correlation between actual phenotypes and predicted phenotypes, was moderate for all traits (0.26 – 0.37). It remains to be seen whether this is sufficient to implement genomic selection without phenotyping in a breeding population. With the G matrix, the selection can be optimized in relation to breeding values and relationship before the final individuals are selected for the breeding population. This proves to be a very useful for making the best possible selection without unintentionally loosing genetic variation.

Prosjektet har brakt Skogfrøverket og skogbruket inn i den genomiske tidsalder. Planteforedlingen kan for første gang virkelig understøttes av målt variasjon i trærnes DNA. Dette har stor betydning for utviklingen av framtidig foryngelsesmateriale for gran og andre treslag der standardiserte genotypingsverktøy som SNP-arrays utvikles. SNP-array er utviklet for gran, furu og bjørk. I tillegg skal NIBIO utvikle et nytt array for eikeartene i et nystartet IPN-prosjekt. Foredling støttet av genomisk analyse bidrar til økt presisjon i utvalg i eksisterende foredlingspopulasjoner, mulighet for å predikere beslektede individers genetiske verdi uten at de selv har vært testet i feltforsøk, økt fleksibilitet i valg av strategi for å bringe foredlingen framover, og mulighet for å optimalisere og overvåke genetisk variasjon i forhold til genetisk gevinst. Skogfrøverket har vært rask med å implementere metodene både i gran- og furuforedlingen. Resultatene fra prosjektet blir nå brukt til å gjøre utvalg for to nye granfrøplantasjer. Foredlingsmålet er økt produksjon, stammer som er rettere med mindre skader og feil. For det ene utvalget ble også densitet vektlagt for å holdes uendret selv om der er en negativ korrelasjon med vekst. Variasjonen i avlsverdiene tilsier at utvalgene vil gi bedre produksjon tilsvarende en bonitetsøkning på 15 %. De nye frøplantasjene vil levere frø fra 2040. For skognæringen betyr dette raskere tilgang til frø som øker lønnsomheten i skogproduksjonen. Det betyr også gi mulighet for å endre foredlingsmål mer fleksibelt enn tidligere. Selv om vi ikke vet hvilke gener som styrer hvilke egenskaper, vil modellene kunne estimere avlsverdier for enhver fenotypet og genotypet populasjon med høy presisjon så lenge det eksisterer en viss slektskapsstruktur. Avlsverdier for deres genotypede avkom kan predikeres med moderat presisjon selv om disse ikke fenotypes. Dette er veldig nyttig når ny kunnskap om trærnes respons på klimaendringer, eller samfunnets behov for råstoff endres, og skal implementeres i foredlingsprogrammet. Foredlet foryngelsesmateriale kan anvendes over store områder til lav kostnad. For samfunnet betyr det at produksjonspotensialet i landskapet utnyttes bedre til produksjon av virke til industri og raskere gjenoppbygging av CO2-lagret i produksjonsskogen etter hogst. I Klimakur 2030 ble det beregnet at effekten av skogplanteforedlingen med moderat foredlingsgevinst kan bety 1 – 1.5 millioner tonn CO2 ekstra bundet i når vi nærmer oss år 2090. Metodene utviklet i dette prosjektet vil gi høyere foredlingsgevinst enn det som ble lagt til grunn da. Foryngelsesmaterialet i skogproduksjonen skal opprettholde genetisk diversitet treslaget. Genetisk diversitet henger sammen med frøplantasjenes effektive populasjonsstørrelse og innblanding av naturlig foryngelse i plantefeltene. Skogplanteforedlingen vil ha god kontroll på effektiv populasjonsstørrelse siden dette hele tiden kan måles med de genomiske dataene.

Breeding of conifer trees posses several challenges associated with their particular biology. These species posses long life cycles and thus investments are made several years before the eventual utilization of genetically improved material. Thus, this is susceptible to changes in the economic objectives of the products, market demands, management policies and ecological conditions. This is the particular case of breeding of Picea abies in Norway and other northern countries. Currently, there are emerging technologies to aid breeding programs of conifer trees with a variety of new genotyping tools and genomic information. Although genomic selection (GS) has been implemented on some domesticated animals and crops, this has not been operationally implemented on conifer trees. One of the reasons is that there is still a need to develop new knowledge on how these technologies could be implemented. In addition, the specific conditions of breeding of P. abies in Norway need further research. Thus, although there are nowadays available technology and analytical methods to implement GS and metrics of quality (genetic diversity) into the breeding programs of trees, specific R&D is required for their further operational implementation. This proposal is therefore focused on addressing the specific challenges for implementing GS and genetic diversity assessment into the breeding cycle of P.abies under the Norwegian conditions. Therefore, the overall goal of this proposal is to implement two new innovative steps into the breeding cycle of Picea abies with the purpose of reducing the cost of producing improved genetic material, as well as certifying the seedlots obtained. These two new implementations will increase the competitiveness of Skogfrøverket by providing improved at a lower cost and the necessary information to manage a sustainable genetic diversity on the breeding program of this species