Incorporating non-key traits in selecting the Pinus radiata production population
© Kennedy et al.; Licensee Springer 2014
Received: 7 February 2013
Accepted: 26 March 2014
Published: 2 July 2014
Three of the traits considered of most economic importance in the genetic improvement of Pinus radiata D. Don, termed as ‘key’ traits, are tree diameter (a proxy for stem volume), wood density and wood stiffness. A number of other traits (non-key traits) may be of equal importance to growers depending on where and for what purpose the trees are being grown. Two such non-key traits are: resistance to the needle blight caused by Dothistroma septosporum (Dorog.) M. Morelet, and reduced heartwood content.
Data from two trial series (each planted at two sites, from a total of some 330 families (189 were half-sib and 142 were full-sib) with 10 to 30 individuals assessed per family) were analysed to determine the effect forwards selection of key traits had upon genetic gains of these two non-key traits.
Multivariate analysis for each trial provided estimates of trait narrow-sense heritabilities (h2) and genetic correlations between traits. Density was the most heritable trait assessed (ĥ2 0.50 to 0.72) with heartwood ring number (0.21 to 0.35), acoustic velocity (0.40), resistance to Dothistroma septosporum (0.20 to 0.34) and stem diameter (0.11 to 0.25) being less heritable. Wood properties were adversely correlated with growth rate to varying degrees.
To estimate the impact of differing technical weights on multiple traits a selection index model was used. Strong positive genetic correlations (r g 0.69 to 0.87) between resistance to Dothistroma septosporum and stem diameter means that strong selection for stem diameter after severe Dothistroma attack assures genetic gain in resistance to Dothistroma septosporum. Strong selection for stem diameter compromised wood properties due to adverse correlations between the two. Heartwood ring number was almost uncorrelated with the other key traits, density and stiffness, meaning that zero or slight negative gain would be expected via indirect selection.
In such instances, it is advocated that key traits be selected for in the breeding population using a selection index and that non-key traits such as heartwood be selected using independent culling in the production population.
KeywordsDothistroma septosporum Heartwood ring number Independent culling Pinus radiata Index selection
Tree breeding primarily involves selection of superior genotypes from breeding populations that are expected to improve a number of economic traits in the production populations. Selection indices are commonly used in tree breeding to improve multiple traits simultaneously (Cotterill and Dean ). To understand the potential trade-offs made during selection for a suite of traits, it is important to understand how such traits are correlated. Selection of measurement traits are constrained by both their cost of measurement, and by their ability to affect the profitability of breeding objectives that are based on defined products and/or industrial processes (e.g., for Pinus radiata D. Don; see Apiolaza and Garrick , Ivković et al. ). The three key traits (KTs) for P. radiata now considered to be of most economic importance are diameter (DBH, a proxy for stem volume), density (DEN) and stiffness (PME) (Pers. Com. John Butcher). However, depending on the location of the stand or the intended end-product, one or several other traits referred to as non-key traits (NKTs) may also be of importance to the grower. The expected impacts of forwards selection for these three KTs on two NKTs, resistance to Dothistroma septosporum (Dorog.) M. Morelet (DTR) and heartwood ring number (HWRN), are examined in this paper.
Dothistroma needle blight causes foliage loss and, in moist warm climates, results in growth loss of up to 70 percent (Shaw and Toes ) making it an important selection trait for plantings in regions at high risk of infection. Heritability estimates for DTR have been extensively reported. For example, Ivcović et al. () reported narrow-sense heritability estimates ranging from 0.05 to 0.69, with a median of 0.36 in a series of 16 P. radiata trials assessed in Australia for DTR. A positive genetic correlation between DTR and growth at a later age was also reported, with a median value of 0.39, however no correlation was found between density and DTR (Ivcović et al. ).
Heartwood in P. radiata is undesirable due to its darker appearance, unreliable durability and difficulty to treat chemically. These properties combine to devalue wood containing heartwood, making increased sapwood content of interest to wood processors. A prior paper, involving the same genetic material, investigated the genetic relationship of heartwood/sapwood with diameter growth (Kennedy et al. ) finding sapwood traits to be moderately heritableand sapwood cross-sectional area to be strongly correlated (r g = 0.9) with diameter growth. Heartwood content is strongly correlated phenotypically with diameter growth in conifers (Climent et al. ), but evidence of some independent genetic control has been found in P. radiata (Nyakuengama et al. ) particularly if assessing HWRN as used in this paper (Kennedy et al. ).
The practicality of simultaneously selecting for DTR and HWRN, and the three KTs, is dependent largely on the heritability of each trait, and the genetic correlations between them. It may eventuate that some NKTs are favourably correlated with some of the KTs, and in this situation the correlated response of the NKT to selection on the KTs means that the NKT may also benefit indirectly. Conversely, if adverse correlations between the KTs and NKTs exist, simultaneous selection for both categories will prove more difficult.
A multi-trait selection index model was produced to illustrate the generalised benefits and constraints of particular selection strategies, to identify situations where alternative selection approaches might need to be considered and to test a number of selection scenarios among KTs and NKTs. The implications of selecting for KTs and the NKTs of HWRN and DTR on the gain that might be achievable in the production population was examined using individual-tree breeding values generated from mixed-model analysis.
The field trials
Data from two Pinus radiata field-trial series were analysed in this study, the ‘Female Tester’ series and the ‘Dothistroma’ series.
The Female Tester series was planted at two sites in New Zealand, one at Esk Forest, Hawkes Bay (39° 15’ S, 176° 42’ E) and the other at Woodhill Forest, Northland, (36° 46’ S, 174° 21’ E). The trials were of single-tree-plot, sets-within-reps design consisting of 30 replicates. Each trial was made up of five sets, containing 36 seedlots, of which five in each set were controls with Growth and Form (GF, see Vincent and Buck ) ratings varying from 6 to 27. The other seedlots were Female Tester crosses; such crosses were composed of a number of pollen parents that had been crossed factorially with five female parents. This was adopted to better compare the top-ranking parents.
Two Dothistroma needle blight screening trials were planted at Kinleith Forest, Central North Island, New Zealand. The trials were planted in 1985 at Phoenix Road (38° 23’ S, 176° 13’ E) and Bobcat Road (38° 24’ S, 175° 99’ E). The design of these trials was a single-tree-plot, sets-within-reps design consisting of 30 replicates. The seedlots were divided into seven sets, each consisting of around 25 seedlots, with all seedlots in set A (not included in the analysis) being self-pollinated. The remainder of the sets were made up of open-pollinated families. Between three and eight controls were included in each set.
Sampling and assessment
Trait abbreviations and units of measurement
Diameter at breast height
Basic density at breast height
Resistance to Dothistroma septosporum
Number of heartwood rings
Predicted modulus of elasticity
Summary of the ages from planting and numbers of individuals and families measured for each trait at the four trial sites
The green density of a standing tree was assumed to be a constant, and so a direct relationship between V2 and PME was assumed for the analyses conducted here.
Estimated trait individual narrow-sense heritabilities and means for the two Dothistroma series trials and estimated Type B genetic correlations ( r B ) between the two sites for each trait; approximate standard errors are in parenthesis
Multi-trait selection model
An index model was constructed in Microsoft Excel software based on the traditional ‘Smith-Hazel’ index model (Cotterill and Dean ) to explore the dynamics of selecting for the KTs and NKTs when different technical weights (level of importance) were imposed. An outline of the model, using notation adapted from Burdon (), is presented in the equations below:
b is the column vector of index weights 1, 2…n;
P is the phenotypic variance-covariance matrix;
G is the genotypic variance-covariance matrix;
b is the column vector of technical weights.
Where i is the selection intensity (e.g. 2.42 = a selection rate of 2%).
The individual-tree breeding values estimated from the linear mixed-effects multi-trait analysis of the Dothistroma series trials were used to determine the effects of independent culling upon the NKTs after index selection of the KTs. A selection index was constructed for the two KTs assessed (DBH and DEN). The index took the form of Eq. 6, with appropriate weights within the multi-trait selection model applied to DBH and DEN to provide gain estimates in line with breeding objectives1 for this hypothetical round of the breeding cycle (10% increase in DBH and a 10 kg m−3 increase in DEN; note: this approximately reflected a weight of 2 on DBH and 1 on DEN). Non-key traits were omitted from the selection model, as they would not be considered at this stage of selection. Technical weights were multiplied by the individual-tree breeding values, as calculated by Eq. 1. The top 2% of individuals (252 from 12600) for the index were selected. The 10 best individuals for HWRN were then selected from this top 2%.
Results and discussion
Estimated multivariate genetic correlations-heritability matrices (upper right and diagonals in bold, respectively), with approximate standard errors in parenthesis, for KTs and NKTs in: (1) the two Dothistroma series trials, (2) the Esk Forest trial and (3) the Woodhill Forest trial
1. Phoenix & Bobcat
(Dothistroma series trials)
367.1 (Kg m−3)
(Female tester series)
344.3 (Kg m−3)
(Female tester series)
For a given set of weights, the predicted gain for each trait calculated using Eq. 8 was plotted, to help build up the picture of the trade-offs when employing multi-trait selection. The graphs in Figures 1, 2 and 3 illustrate the dynamic process that gain is allocated to each trait under multi-trait selection. Across all sites, it is apparent that HWRN cannot be reduced (i.e. a negative gain achieved) without a large reduction in optimal DBH gain. Heartwood ring number is almost uncorrelated to any of the KTs, making it difficult to achieve large gains in both.
The pattern of relative gain estimated for the two trial series were similar, but gains were lower for DBH in the Female Tester series trials largely due to the relatively low heritabilities at both the Esk and Woodhill sites (i.e. <0.16). The parameter DTR03 is favourably correlated with DBH06; the bigger the tree the lower the incidence of DTR infection. Owing to this favourable genetic correlation, selection for growth alone will result in an increase in DTR resistance. Strong selection for DBH will, however, compromise wood properties such as DEN and HWRN (Figure 3). As HWRN is not strongly correlated to the two other KTs, DEN and PME, selection for the three KTs is unlikely to decrease HWRN.
Selection indices and independent culling
The basic approach for selection of NKTs used here was to restrict consideration of these traits to the production population using independent culling to provide targeted seedlots that meet individual growers’ needs. An additional advantage of this approach would be that only trees among the prospective production population need be assessed for NKTs of interest, making considerable cost savings. An example of how this approach might work follows.
This approach of using independent culling to improve NKTs not strongly correlated with any of the KTs is only likely to lead to gains in the current production population, as gene frequencies that improve the NKTs are not increased in the breeding population. The trade-offs in gain selecting for just a few uncorrelated traits, demonstrated earlier, illustrate the importance of concentrating improvement upon only a few traits of most economic importance. As P. radiata is grown on a large number of site types and has such a wide variety of uses, it has been a challenge to keep the breeding objectives down to only a few traits (Shelbourne ). Calculation of economic weights for NKTs is likely to prove difficult, as a particular trait may be of importance for some growers and producers and for others not, for example, with widely varying incidence of Dothistroma infection across different regions. Earlier generations of the New Zealand breeding programme partitioned the main breeding population into several breeds that concentrated on improving specific traits (Jayawickrama and Carson ). These breeds were later combined to form the current breeding population which had the benefit of reducing costs and increasing selection intensity. Identification of other NKTs uncorrelated to the KTs used for selection may see the return of specific breeds if enough demand for their improvement exists.
If reliable estimates of genetic correlations between the three identified KTs and a NKT of particular interest can be obtained, then it would be possible to increase the weight given to the NKT in question, elevating potential gain from selection. The main drawback to such an approach would lie in the situation whereby the NKT was unfavourably correlated with a number of the KTs. Considerable penalty would be incurred in the unfavourably correlated KTs if any gain was to be achieved through selection for the NKT(s) in question. Under normal circumstances, if such a situation was to arise, one would select for this NKT based on independent culling, but absence of measurement data means this cannot generally be considered at present. More likely, the trees selected for the three KTs would be assessed for the NKTs of interest, reducing operational measurement costs. Non-key traits are best selected from within the production population rather than the breeding population.
Sensitivity analyses using the Smith-Hazel selection index model enabled a general understanding of the levels of gain and trade-offs among key and non-key traits, and the effects that choosing different populations can have on gain. While the desired technical weights, derived from the traditional mass-selection index approach, cannot be directly applied to multivariate BLUP breeding values (and yield the same levels of genetic gains) they do provide an important ‘starting point’ for further sensitivity evaluations. Both approaches can provide insights and good quantification of what may be coming out of production populations when independent culling is applied to NKT’s. As the results presented here indicate, positive gains can still be made in a trait that is not correlated with the key traits under selection.
The authors thank the Radiata Pine Breeding Company for providing the data and funding for this work and the anonymous referees for their helpful suggestions.
- Apiolaza LA, Garrick DJ: Breeding objectives for three silvicultural regiemes of radiata pine. Canadian Journal of Forest Research 2001,31(4):654–662. 10.1139/x00-200View ArticleGoogle Scholar
- Burdon RD: Genetic correlation as a concept for studying genotype-environment interaction in forest tree breeding. Silvae Genetica 1977, 26: 168–175.Google Scholar
- Burdon RD: Early selection in tree breeding: principles for applying index selection and inferring input parameters. Canadian Journal of Forest Research 1989, 19: 499–504. 10.1139/x89-076View ArticleGoogle Scholar
- Climent J, Chambel MR, Pérez E, Gil L, Pardos J: Relationship between heartwood radius and early radial growth, tree age, and climate in Pinus canariensis . Canadian Journal of Forest Research 2002, 32: 103–111. 10.1139/x01-178View ArticleGoogle Scholar
- Cotterill P, Dean CA: Successful tree breeding with index selection. CSIRO Division of Forestry and Forest Products, Melbourne Australia; 1990.Google Scholar
- Gilmour AR, Gogel BJ, Cullis BR, Thompson R: ASReml User Guide Release 3.0. VSN International Ltd., Hemel Hempstead, HP1 1ES, UK; 2009.Google Scholar
- Ivković M, Wu HX, McRae TA, Mike BP: Developing breeding objectives for radiata pine structural wood production. 1. Bioeconomic model for economic weights. Canadian Journal of Forest Research 2006, 36: 2920–2931. 10.1139/x06-161View ArticleGoogle Scholar
- Ivcović M, Baltunis B, Gapare WJ, Elms S, Sasse J, Dutkowski G, Elms S, Wu H: Breeding against Dothistroma needle blight of radiata pine in Australia. Canadian Journal of Forest Research 2010, 40: 1653–1660. 10.1139/X10-097View ArticleGoogle Scholar
- Jayawickrama KJS, Carson MJ: A breeding strategy for the New Zealand radiata pine breeding cooperative. Silvae Genetica 2000,49(2):82–90.Google Scholar
- Kennedy SG, Yanchuk AD, Jefferson PA: Relationship of heartwood traits with diameter growth, implications for genetic selection in Pinus radiata. Tree Genetics and Genomes 2013, 9: 1313–1319. 10.1007/s11295-013-0639-3View ArticleGoogle Scholar
- Lynch M: Genetics and analysis of quantitative traits. Sinauer Associates, Sunderland; 1998.Google Scholar
- Nyakuengama JG, Matheson AC, Spencer DJ, Evans R, Vinden P: Genetics of heartwood formation and moisture status in Pinus radiata . Appita Journal 2000,53(1):30–35.Google Scholar
- Schneeberger M, Barwick SA, Crow GH, Hammond K: Economic indices using breeding values predicted by BLUP. Journal of Animal Breeding Genetics 1992, 109: 180–187. 10.1111/j.1439-0388.1992.tb00395.xView ArticleGoogle Scholar
- Shaw CG III, Toes EHA: Impact of Dothistroma needle blight and Armillaria root rot on diameter growth of Pinus radiata . Phytopathology 1977, 67: 1319–1323. 10.1094/Phyto-67-1319View ArticleGoogle Scholar
- Shelbourne CJA: Genetics of adding value to the end-products of radiata pine. In IUFRO ‘97 Genetics of Radiata Pine. Proceedings of NZFRI-IUFRO conference, Rotorua, New Zealand 1–4 December and Workshop 5 December 1997. Edited by: Burdon RD, Moore JM. New Zealand Forest Research Institute, Rotorua; 1997:129–141. For. Res. Inst. Bull. 203Google Scholar
- Smith DM: Maximum moisture content method for determining the specific gravity of small wood samples. United States Forest Service, Forest Products laboratory, Madison, Wisconsin; 1954.Google Scholar
- Vincent G, Buck K: Radiata pine seedlots – GF Plus. New Zealand Tree Grower 1998,19(2):33–35.Google Scholar
- Wang X, Ross RJ, McClellan M, Barbour RJ, Erickson JR, Forsman JW, McGinnis GD: Nondestructive evaluation of standing trees with a stress wave method. Wood Fibre Science 2001,33(4):522–533.Google Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.