Estimation of Genetic Parameters for 12 Fruit and Vegetative Traits in the University of Florida Strawberry Breeding Population

The University of Florida strawberry (Fragaria ·ananassa) breeding population has been continuously improved by recurrent selection since 1968. However, there is a lack of information on genetic parameters that may inform breeding decisions. Parameters were estimated in this population using 19 full-sib families from a 5 · 4 factorial mating design plus six additional biparental crosses and 14 control genotypes including some of the parents. During the 2010–11 season, clonal replicates of the seedling and parental genotypes were distributed within and among two field locations in west–central Florida. Twelve commercially important traits were measured including fruit chemical traits (soluble solids content and titratable acidity), other fruit and yield traits (early and total marketable yields, proportion of total cull fruit, proportion of misshapen fruit, proportion water-damaged fruit, and shape score), and vegetative traits (plant height and total runners). Heritabilities, genotype by environment interaction, and multiple correlations (phenotypic, genotypic, and genetic) were estimated using general mixed model analyses. Narrow-sense heritabilities varied from low to moderate (h = 0.13 ± 0.07 to 0.32 ± 0.09) except for shape score (h = 0.06 ± 0.04) and total average weight (h = 0.52 ± 0.07). Broad-sense heritabilities were larger (H = 0.18 ± 0.03 to 0.53 ± 0.04), and for more than half of the traits, over 50% of the total genetic variation was non-additive. Large genetic and genotypic correlations were found for some traits, most notably between soluble solids content and early marketable yield (–0.68 ± 0.22). Genetic gains for this pair of traits based on a Monte Carlo simulation illustrated the tradeoff between these two traits, showing that a 27% increase in early yield could be obtained through selection but at the expense of an 8% decrease in soluble solids. However, moderate gains can be made in both traits using the appropriate index coefficients. Florida is the primary source of strawberry fruit for the eastern United States and eastern Canada from December to late March. The state is second to California in total U.S. production with a harvested area of greater than 3600 ha during the 2010–11 season (U.S. Department of Agriculture, 2011). Production in Florida shares similarities with major production regions such as Australia, southern California, and Spain, where strawberries are grown using bare-root transplants in intensive, annualized systems for winter and spring markets. In 1968 the University of Florida (UF) started a strawberry breeding program (Whitaker et al., 2011), although some openpollinated seedling selection was performed before that time. Since that time, the breeding population has been continually improved for multiple plant and fruit traits through recurrent selection. Typically, 100 controlled crosses have been made each year among 30 or more parental genotypes in the main breeding population with additional crosses made for germplasm development efforts. Pedigree records have been maintained to monitor parentage and inbreeding, and full-sib crosses have almost always been avoided. To replicate commercial nursery conditions, each seedling genotype is asexually propagated through stolons (runners) in a temperate summer nursery to produce bare-root transplants for evaluation in the fruiting field. In this way multiple runner plants per seedling genotype may be evaluated; the original seedling plant is not evaluated. West–central Florida is characterized by periodic rainfall, high humidity, fluctuating temperatures, and occasional freezes, which inhibit pollination and fruit development, resulting in unmarketable fruit. Therefore, reducing the proportion of unmarketable fruit and thereby increasing marketable yield is an important breeding objective. The seasonality of fruit production for a strawberry cultivar is also of vital importance, in which an ideal pattern consists of large earlyseason yields from late November through January when the value of the crop is greatest and moderated late-season yields during February and March when overproduction can result in reduced market prices. Large average fruit size is also a breeding objective as well as favorable levels of traits that affect flavor perception such as soluble solids content (SSC) and titratable acidity (TA) (Joquand et al., 2008). In addition, the plant must be vigorous enough to establish well in the field and support high yields but no so large and dense as to restrict air movement and obscure the fruit from harvesters. A historical trial of cultivars and advanced selections from the UF strawberry breeding program revealed gains over time for fruit size and proportion of marketable fruit (Whitaker et al., 2011). Although SSC and TA varied widely among genotypes, clear trends over time could not be observed for these traits. Until recently, there have been no published reports of genetic parameters such as heritabilities and genetic correlations for the Received for publication 29 May 2012. Accepted for publication 2 July 2012. Corresponding author. E-mail: [email protected]. 316 J. AMER. SOC. HORT. SCI. 137(5):316–324. 2012. UF strawberry breeding population, which would be desirable for shaping breeding and selection strategies (Hasing et al., 2011). They provide an understanding of the effects of trait selection in the long term and the behavior of correlated traits that, if adverse, may hinder breeding progress if they are ignored during selection. Previous studies have reported genetic parameters for plant and fruit traits of strawberries in both annual and perennial production systems (reviewed by Galleta and Maas, 1990; Hancock et al., 2008). Information on genetic parameters for annual production systems have mainly been generated using the University of California–Davis breeding population. Narrowsense heritabilities for plant growth traits such as plant diameter have been low to moderate with little contribution of non-additive variance (Fort and Shaw, 2000; Shaw, 1993). Substantial amounts of additive variance for yield and fruit size have been demonstrated, although the relative proportions of additive and dominance variance have varied widely across testing environments and propagule types (Fort and Shaw, 2000; Pringle and Shaw, 1998; Shaw, 1989; Shaw et al., 1989; Shaw and Larson, 2005). Greater dominance variance was typically found for SSC and TA (Shaw et al., 1987). Gains from selection for SSC were predicted to be poor based on clonal trials of selected individuals, mainly as a result of large interactions with cultural environments and harvest dates (Shaw, 1988, 1990). Although these previous studies provide important benchmarks, they may not be reflective of the germplasm, population history, and testing environments of the UF strawberry breeding program. In this study, we explore the genetic basis of several important fruit and vegetative traits in the UF strawberry breeding population by conducting clonal tests of seedling, parental, and control genotypes across two environments and performing genetic analyses that incorporate pedigree records spanning 15 generations. Specifically we aim to: 1) obtain estimates of narrow-sense heritability, broad-sense heritability, and genotype by environment interactions; 2) estimate phenotypic, genotypic (additive plus non-additive genetic effects), and genetic (additive) correlations among the traits of interest; and 3) predict genetic gains from multivariate selection. Materials and Methods MATING AND FIELD DESIGNS. Twenty-five biparental crosses were generated for testing by controlled pollination among 17 parents. Nineteen biparental crosses were made among nine parents in a 5 · 4 factorial mating design (one cross missing). These parents were chosen to represent a broad range of phenotypic diversity present in the breeding program. Six additional biparental crosses were made among 10 different parents. These crosses were a random sample of the crosses already generated in the breeding program for evaluation during the 2010–11 season. Two parental genotypes were shared across the factorial crosses and the additional biparental crosses. All parents shared pedigree linkages and constituted a representative selection of named cultivars and advanced selections from the UF strawberry breeding program. Seventeen parents was considered a sufficient sample size to represent the main UF strawberry breeding population, which is maintained through controlled crosses among 25 to 30 parents each year and contains several connecting relatives from previous generations. Twenty seedlings were chosen at random from each cross for testing. In addition, 14 control genotypes were included, which arose from 23 different parents. These genotypes were either parents or other advanced selections. In 2010 all seedlings were germinated and transported to the breeding program’s summer nursery site in Monte Vista, CO (lat. 37 40#46.10$ N, long.106 8#10.83$ W) where they were clonally propagated by runners. Four bare-root runner plants were generated from each seedling and for each of the 13 additional parental and advanced selection genotypes. Before planting, transplants were individually weighed (grams) to determine initial runner plant weight. Two runner plants were established at the Gulf Coast Research and Education Center (GCREC) in Balm, FL (lat. 27 45#37.98$ N, long. 82 13#32.49$ W) on 11 Oct. 2010, and two runner plants were established at the test plots of the Florida Strawberry Growers Association in Dover, FL (lat. 28 0#55.55$ N, long. 82 14#5.24$ W) on 14 Oct. 2010. At each site the runner plants were arranged in a randomized block design (single-plant plots resulting in four total replications across sites) with two raised beds per replication in Balm and three beds per replication in Dover. Each site was prepared and maintained according to standard commercial practices, which are more fully described in Mackenzie et al. (2011). Briefly, beds were 91.5 m long, 71 cm wide, 15 cm high at the edges, and 18 cm high in the center and were fumigated with a mixture of telone and chloropicrin 1 month before transplanting. There were two offset rows of plants per bed. Plant spacing was 38 cm within rows and 28 cm between rows. After transplanting the runner plants were overhead-irrigated for 10 d during daylight hours to facilitate establishment. Once established, the plants were irrigated and fertilized exclusively through the drip tape. DATA COLLECTION. Data for all traits were gathered on an individual plant basis. Each runner plant was weighed to determine its initial weight in grams before establishment. Fruit harvests were made at weekly intervals beginning with the appearance of the first ripe fruit in late November and continuing until the end of January. As a result of restrictions in available labor for harvesting as yields increased, late-season fruit harvests were recorded every other week during February and March, similar to the partial records method of Shaw (1989) where yield was recorded on alternate weeks. Two fruit chemical traits, SSC and TA, were assessed. The SSC trait was measured in the field four times between 11 Jan. and 30 Mar. 2011 and is expressed as the mean over time of all individual measurements. One or two ripe fruit were squeezed by hand until the expressed juice covered the prism of a handheld digital refractometer (PAL-1; Atago Co., Tokyo, Japan) that was calibrated with deionized (DI) water. On 8 to 9 Mar. 2011 juice samples were collected for measurement of TA. One or two fruit (depending on availability) per plant were squeezed by hand and the juice collected with a funnel into a 6-mL screw-capped plastic vial. The vials were placed on ice and transported to the GCREC laboratory where they were frozen at –20 C. At the conclusion of the season, the vials were thawed and 1 to 2 mL of juice was diluted with 50 mL DI water and titrated with 0.1 N NaOH to a pH 8.1 end point using a 719 S titraino and 738 stirrer (Metrohm USA, Westbury, NY). The ratio SSC/TA was calculated and included in the analysis as a separate trait because it is known that this ratio influences flavor perception in UF strawberry cultivars (Joquand et al., 2008). Seven additional yield and fruit traits were assessed. At each harvest, all ripe fruit were removed and counted. Marketable J. AMER. SOC. HORT. SCI. 137(5):316–324. 2012. 317 fruit were weighed (grams) to determine total marketable yield (TMY). Early marketable yield (EMY) was calculated as the marketable weight before the first harvest in February. The TMY was divided by the number of marketable fruit to estimate average fruit weight (AWT). The total number of unmarketable (cull) fruit was counted and expressed as a proportion of the total number of fruit (TC). Cull fruit were further rated into overlapping subcategories including total misshapen fruit (TM) and total water-damaged fruit (TWD), which were also expressed as a proportion of the total number of fruit. Marketable fruit were rated for shape during three different harvests between 11 Jan. and 23 Feb. Each fruit was subjectively categorized on a 1 to 3 pictorial scale, where a rating of ‘‘1’’ represented fruit with irregular shape and surface and a rating of ‘‘3’’ represented fruit with regular conical shape and minimal surface irregularities. A weighted mean shape score (SHP) was calculated by multiplying the number of fruit in each scale category by their scores and dividing by the total number of fruit in all categories. Two vegetative traits, total runners (TRs) and plant height (PHT), were also assessed. All runners produced between 30 Nov. 2010 and 4 Jan. 2011 were removed and counted. Plant height (centimeters) was measured on 25 to 27 Jan. 2011 using a straight ruler. STATISTICAL AND GENETIC ANALYSES. A covariance analysis across locations was carried out using ASReml software (Gilmour et al., 2009) to test the effect of including initial runner plant weight in grams as a fixed covariate in the model. Significance of the covariate was assessed by the incremental Wald statistic. Univariate and bivariate analyses for all traits were conducted using ASReml software (Gilmour et al., 2009). Univariate analyses were performed to generate variance components for the bivariate analysis and the type B genetic correlations between locations (Yamada, 1962). These correlations allow the estimation of genotype by environment interaction at both the additive and genotypic (additive plus non-additive) levels. Type B correlation values range from 0 to 1, and values close to 1 indicate a low genotype by environment interaction. The tested families had some relatedness among them causing some bias of the genetic estimates, but the bias was ameliorated by incorporating a pedigree structure to the model. The 14 control genotypes were included in the analyses to calculate variance components, and pedigree information was incorporated for each of them as well. The univariate analyses were undertaken following the general mixed model: y = Xb + Z1a + Z2c + Z3k + Z4l + e ð1Þ where y is the vector of observations; b is the vector including the population mean, sites, replications, and the covariate fixed effects; a is the vector of random additive effects; a ; NID 0, Asa c is the vector of non-additive genetic effects, including dominance plus epistasis and c ; NID 0, Csc ; k is the vector of random interactions between location and additive effects; and k ; NID 0, I ssas l is the vector of random interactions between location and non-additive effects, l ; NID 0, I sscs . Heterogeneity of the residual effects across locations was modeled as e ; NID 0, R ð Þ, wit IR = 4 Injsej where j is the location and 4 defines the direct sum operation. A is the matrix of additive relationships among genotypes, C is the matrix of non-additive effects, and Is is an identity matrix with s equal to the number of sites. sa, sc , sas, scs andsej are the additive genetic variance, non-additive genetic variance, additive by site interaction variance, non-additive by site interaction variance, and variances of random residual effects. X, Z1, Z2, Z3, and Z4 are known incidence matrices relating the observations in y to effects in b, a, c, k, and l. The bed effects were not included in the model because beds within replication effects had been proven not significant in a previous study at the same locations designed to account for spatial variability along the beds (unpublished data). Narrow sense heritability (I ) and broad sense heritability (H ) for each variable were estimated as follows: