Abstract
Theoretical studies suggest that marker-assisted selection (MAS) has case-specific advantages over phenotypic selection (PHE) for selection of quantitative traits. However, few studies have been conducted that empirically compare these selection methods in the context of a plant breeding program. For direct comparison of the effectiveness of MAS and PHE, four cucumber (Cucumis sativus L.; 2n = 2x = 14) inbred lines were intermated and then maternal bulks were used to create four base populations for recurrent mass selection. Each of these populations then underwent three cycles of PHE (open-field evaluations), MAS (genotyping at 18 marker loci), and random mating without selection. Both MAS and PHE were practiced for yield indirectly by selecting for four yield-component traits that are quantitatively inherited with 2–6 quantitative trait loci per trait. These traits were multiple lateral branching, gynoecious sex expression (gynoecy), earliness, and fruit length to diameter ratio. Both MAS and PHE were useful for multi-trait improvement, but their effectiveness depended upon the traits and populations under selection. Both MAS and PHE provided improvements in all traits under selection in at least one population, except for earliness, which did not respond to MAS. The populations with maternal parents that were inferior for a trait responded favorably to both MAS and PHE, while those with maternal parents of superior trait values either did not change or decreased during selection. Generally, PHE was most effective for gynoecy, earliness, and fruit length to diameter ratio, while MAS was most effective for multiple lateral branching and provided the only increase in yield (fruit per plant).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
Theoretical-based simulation studies suggest that the effectiveness of marker-assisted selection (MAS) for polygenic traits can be greater than traditional trait-based selection (Lande and Thompson 1990; Zhang and Smith 1992; Edwards and Page 1994; Gimelfarb and Lande 1994a; Gimelfarb and Lande 1994b). In general, these studies agree that MAS efficiency is enhanced when markers are tightly linked (<5.0 cM) to quantitative trait loci (QTL), selection is performed in early generations prior to recombination between markers and QTL, large populations are used, and selection is practiced on traits with low heritability. However, the underlying assumptions of simulation studies may not be upheld in practice which may reduce their applicability to empirical studies (van Berloo and Stam 2001). In practice, MAS has been effective for the introgression of simple traits or a small number of genes in several crop species including disease resistance in common bean (Phaseolus vulgaris L.; de Oliveira et al. 2005), disease resistance and grain protein in wheat (Triticum aestivum L.; Kuchel et al. 2007), and root traits in rice (Oryza sativa L.; Steele et al. 2006). However, MAS appears less effective for complex traits such as yield (Francia et al. 2005; Collard and Mackill 2008; Xu and Crouch 2008). Although MAS has been successfully reported in commercial breeding programs, details of these successes are limited and implementation of MAS in public breeding programs has been slow (as reviewed by Xu and Crouch 2008). Empirical comparisons of MAS to phenotypic selection (PHE) are scarce and often conflicting (Zhang et al. 2006; Moreau et al. 2004; Davies et al. 2006) suggesting the need for a direct comparison of the effectiveness of MAS and PHE for complex traits.
Yield is a complex trait that has been a focus of cucumber (Cucumis sativus L.; 2n = 2x = 14) breeding for over 50 years (Lower and Edwards 1986; Wehner 1989; Wehner et al. 1989). Although the yield of US processing cucumber increased steadily from 1950 to 1980, it has reached a plateau since the early 1980s (Shetty and Wehner 2002). Selecting directly for yield is difficult which is partially due to its low narrow-sense heritability (0.07–0.25) and the dramatic influence of the environment on trait expression (as reviewed by Wehner 1989). The most effective breeding approach for yield improvement in cucumber may be selection for traits directly related to yield (Wehner 1989; Cramer and Wehner 1998; Cramer and Wehner 2000b).
Four important yield components in cucumber are earliness, gynoecious sex expression (gynoecy), fruit length to diameter ratio, and multiple lateral branching (Cramer and Wehner 2000a; Fazio et al. 2003a). Earliness, gynoecy, and multiple lateral branching have been shown to be positively correlated with the number of fruit per plant (Cramer and Wehner 2000a; Cramer and Wehner 2000b; Fazio 2001), and length to diameter ratio is an important determinant of marketable fruit yield (Serquen et al. 1997a). Each of these traits is under the control of two to six major genes with relatively large effects. The narrow-sense heritabilities (h 2) of each trait range from 0.14 to 0.48 depending on trait and environment (Serquen et al. 1997b; Fazio et al. 2003b). The negative correlations that exist between these yield components (e.g., gynoecy with multiple lateral branching and earliness with length to diameter ratio) make the simultaneous improvement of these traits a challenge.
The use of MAS in cucumber breeding has potential for increasing the efficiency and effectiveness of selection for yield components through line and population improvement. Moderately saturated linkage maps have been developed for cucumber and genomic regions have been identified that have proven useful for selection of yield components by MAS during backcross breeding (Fazio et al. 2003a; Fan et al. 2006). These studies utilized yield-associated QTL identified initially by Serquen et al. (1997b) and then by Fazio et al. (2003b) in separate mapping populations derived from a cross between lines Gy-7 (synom. G421) and H-19. Fazio et al. (2003b) confirmed the marker-QTL linkages of a single trait, multiple lateral branching, and demonstrated that selection with these markers increased the number of branches equal to phenotypic selection during two cycles of backcrossing. Fan et al. (2006) evaluated the effectiveness of MAS for multiple yield components by backcrossing after three cycles of PHE by recurrent selection. Markers utilized for MAS were linked to QTL for earliness (LOD ≥ 4.1), gynoecy (LOD ≥ 3.0), length to diameter ratio (LOD ≥ 4.2), and multiple lateral branching (LOD ≥ 3.0). Selection by PHE improved multiple lateral branching and length to diameter ratio and MAS continued improvement of these traits as well as gynoecy. These two studies indicate that MAS is effective for selecting yield components in cucumber by backcross breeding typical of breeding line development. However, the efficacy of MAS for yield components in cucumber population improvement typical of breeding programs has not yet been established.
Given the potential utility of MAS, a study was designed to increase cucumber yield by simultaneous selection of multiple yield components employing MAS and PHE, and to directly compare these methods for response to recurrent selection. In order to test their efficacy, both methods used the same selection scheme, which was designed to overcome previously documented negative correlations between yield components. Four populations were created by intermating four inbred lines, and then each population underwent three cycles of recurrent selection by PHE and MAS, as well as random mating without selection (RAN). This study will allow for the development of appropriate breeding strategies for the use of PHE and MAS in cucumber.
Materials and methods
Germplasm and population development
Four inbred lines were chosen as parents from the US Department of Agriculture (USDA) cucumber breeding program, Madison, WI, to develop four separate populations (Table 1; Fig. 1). Lines 6996A and 6995C were drawn from a recombinant inbred line (RIL) population (Gy-7 × H-19, F9; Staub et al. 2002). Line 6823B originated from a cross between the RIL parent H-19 and a USDA elite processing line whose progeny were then selected for H-19 attributes. Line 6632E is morphologically similar to the RIL parent Gy-7, but does not have either parent in its pedigree (Staub and Crubaugh 2001). These lines were specifically chosen because their complementary phenotypes (Table 1) provided the basis for selection of earliness, gynoecy, multiple lateral branching, and length to diameter ratio.
The four parental inbreds were intermated to create four distinct populations that subsequently underwent selection (Fig. 1). Crosses were made in a greenhouse in Madison, WI, in 2000 by pollinating female flowers of each inbred with bulked pollen from the other three lines. The resulting seeds were bulked by maternal parent to create four populations (i.e., Pop. 1–4; Table 1; Fig. 1) which had not undergone selection and were designated as cycle 0 (e.g., Pop. 1 C0). Each of these populations subsequently underwent PHE, MAS, and RAN for three cycles (C1–C3). All selection and mating was performed within each of the four populations, independent of the other three populations (i.e., intrapopulation improvement only). PHE was performed based on phenotype alone (i.e., without marker information), and MAS was applied without regards to phenotypic information (i.e., marker information only).
Selection scheme
Phenotypic selection for earliness, gynoecy, multiple lateral branching, length to diameter ratio, and standard-leaf type (leaf area > 40 cm2; Staub et al. 1992) was practiced under open-field conditions at the University of Wisconsin Experiment Station, Hancock, WI (UWESH) in soil classified as Plainfield loamy sand (Typic Udipsamment; sandy, mixed, mesic). Data were taken on individual plants, where leaf type was classified as standard (LL) or little leaf (ll = 30–40 cm2; Staub et al. 1992). Earliness was assessed as the number of days from planting to anthesis of the first female flower. Sex expression was measured as the percentage of the first ten flowering nodes bearing female flowers (nodes with both male and female flowers were classified as male) where 100% was designated gynoecious, 50–90% was considered predominantly female (PF), and less than 50% was classified as monoecious. Fruit length to diameter ratio was estimated by visual inspection of at least four immature fruit (USDA grade size 3A–3B; 3.0–5.0 cm in diameter). Multiple lateral branching was recorded at or after anthesis as the number of lateral branches (at least three internodes in length) in the first ten nodes of the mainstem.
Phenotypic selection was accomplished in two stages within each cycle of selection (Fig. 1) using minimum trait thresholds for the first stage, and index selection for the second stage. For Stage 1, a total of 400 C0, 600 C1, or 600 C2 plants from each population were evaluated in 2001, 2002, and 2003, respectively. Individual plants were first evaluated for leaf type, earliness, gynoecy, and multiple lateral branching, since these are the first traits to be expressed developmentally in cucumber. Only individuals that met pre-established thresholds of standard-sized leaves, earliness as <48 days to the first female flower, gynoecy as >50% female flowers, and multiple lateral branching as >3 branches were evaluated for fruit length to diameter ratio. Those plants with a length to diameter ratio above the threshold (2.8) were designated selections at Stage 1. A subjective index was employed at Stage 2 of PHE where multiple lateral branching and earliness were weighted approximately 2 (multiple lateral branching) and 1.5 (earliness) times that of gynoecy and length to diameter ratio, which were weighted equally. These weights were based on their relative importance to early, uniform yield. The relative weights among the traits are illustrated by the selection differentials (difference in trait means of the selections from Stage 2 and the selections from Stage 1) for each trait in each population (Table 2). Individual plants were ranked by their values of multiple lateral branching, then earliness, and the values of gynoecy and length to diameter ratio were used to make Stage 2 selections among the highest ranked individuals. Twenty plants were selected from Stage 2 in each cycle (C1–C3) of PHE within each population, representing a standardized selection intensity (i) of 2.063, 2.219, and 2.219 for C1, C2, and C3, respectively.
The efficient recombination of Stage 2 selections required intermating in the greenhouse. The short growing season of Wisconsin did not allow enough time in the field for chemical induction of male flowers and intermating to produce mature seed from Stage 2 selections with mostly female flowers. Thus, at least two meristems of each Stage 2 selection were taken to a greenhouse in Madison, WI. Once these cuttings were rooted, they were transplanted and were allowed to establish for at least 1 week. The apical meristems and surrounding leaves were then treated with two applications (7 days apart) of 3 mM silver thiosulfate [Ag(S2O3)2]3− as a foliar spray to induce male flower production (Nijs and Visser 1980). Selections were then randomly mated by pollination of each female flower with five random male flowers.
To perform MAS, a total of 18 markers linked to F (femaleness), de (determinate), ll, and previously identified QTL (Serquen et al. 1997b; Fazio et al. 2003b) for earliness, gynoecy, multiple lateral branching, and length to diameter ratio were selected (Table 3; Fig. 2). All markers employed were drawn from Fazio et al. (2003b), except AJ6SCAR, and M8SCAR which were SCARs converted from previously mapped RAPDs (Nam et al. 2005; Robbins 2006). The strategy used to select markers for use in MAS is outlined in Robbins et al. (2002) and Robbins and Staub (2004). Briefly, many factors were taken into consideration when selecting markers such as marker type, marker inheritance (i.e., dominant or codominant), genetic distance from QTL, and number of QTL in proximity to the marker. The QTL identified for selection in this study had a relatively large effect (cumulative R 2 > 37–85%), high LOD scores for marker linkages (>3.0; Table 3; Fig. 2), and were consistent over several environments (Serquen et al. 1997b; Fazio et al. 2003b). The majority of marker-QTL associations in this study were <5.0 cM (Fazio et al. 2003b). In most cases, markers flanking the QTL of interest (Edwards and Page 1994) were employed, especially in regions where marker-QTL associations were >5 cM (e.g., AK5SCAR and M8SCAR for multiple lateral branching; Table 3; Fig. 2). Codominant markers tightly linked to QTL were given preference. Where available, SCAR, SNP, and SSR markers were chosen over RAPD and AFLP markers because of their inherent robustness, ease of use, and ability to be multiplexed (Polashock and Vorsa 2002; Tang et al. 2003; Mohring et al. 2004; Staub et al. 2004). Once the markers were chosen, the desired allele at each marker locus was identified. The desired allele was the Gy-7 allele, the H-19 allele or both Gy-7 and H-19 alleles (heterozygous) since all four parental lines carried only Gy-7 or H-19 alleles at each marker locus. The combination of desired alleles over all marker loci was identified as the ideal genotype, or ideotype (i.e., allele selected column of Table 3).
The selected markers were used to genotype individuals to make selections for each cycle of MAS. Marker genotyping, including DNA extraction, polymerase chain reaction (PCR) amplification, and agarose gel electrophoresis, was conducted according to Fazio et al. (2003b). To increase marker efficiency, the markers were multiplexed in empirically determined groups (Table 3) according to Staub et al. (2004) and Robbins (2006). All individuals within a population were genotyped at each marker locus. Those plants whose genotype matched the ideotype at the greatest number of marker loci were selected and intermated to produce the next generation. For each cycle of MAS within each of the four populations, the number of individuals tested, the selection intensity, and crossing scheme were identical to that of PHE.
Random mating was accomplished by first sowing 20 random seeds from each of the four C0 populations. Then gynoecious plants were chemically induced to produce male flowers and all plants were intermated using the same scheme as that for MAS and PHE to create C1. The resulting seeds were equally bulked, and 20 random C1 plants were intermated in the same manner to create C2 seed and the same procedure was used to create the C3 population.
Open-field evaluation of selection
Response to selection was evaluated in an open-field trial at UWESH in the summer of 2004 with all entries replicated in two planting dates. Seeds were sown in a greenhouse in Madison, WI on June 4, 2004 and June 16, 2004, then transplanted on June 23, 2004 and July 7, 2004, respectively. Each planting date was arranged in a split-plot design with eight replications of each population (whole plot factor) in randomized complete blocks, with a combination of cycle (i.e., C0–C3) and method of selection (i.e., MAS, PHE, and RAN) completely randomized as subplots with ten plants per subplot. Plots were arranged in single rows with 18 cm between plants and 1.5 m between rows (~37,000 plants/ha). This plant density was chosen because it optimized potential yield in highly branched genotypes in multiple harvest operations in Wisconsin (Fredrick and Staub 1989; Staub et al. 1992). The four inbred lines that served as parents, as well as Gy-7, H-19, and the commercial cultivar ‘Vlasset’ (Seminis Vegetable Seeds, Inc, Oxnard, Calif.) were included as controls for comparison.
All traits under selection were evaluated as well as yield, which was measured as the number of fruit per plant. Yield was recorded for each of four harvests at 59, 66, 76, and 96 (first planting date) and 54, 64, 75, and 91 (second planting date) days after planting to calculate four-harvest means. Each of the four harvests occurred when two to three oversized fruit (>51 mm in diameter) were observed within a plot (Wehner et al. 1989). All immature fruits >20 mm in diameter and >10 cm in length were included in total fruit number. Both multiple lateral branching and gynoecy were evaluated on each plant exactly as during PHE. Mean fruit length to diameter ratio was obtained per plot by measuring the length and diameter of 5–10 randomly selected fruits (USDA 2B–3A grade; 2.5–3.0 cm in diameter), and then averaging over three harvests. Earliness was defined as the average number of fruits per plant in the first harvest.
Statistical analysis
All response variables were initially analyzed by analysis of variance (ANOVA) using PROC GLM of SAS (2003) to determine treatment effects. Treatments of planting date, populations, cycles, and methods were considered fixed effects, while blocks were considered random. Specific single-degree of freedom contrasts within analyses of variance were employed to determine general response to selection for biologically important comparisons (e.g., PHE and MAS). Selection responses (linear and quadratic effects) were computed by regression of trait means on selection cycles within each population for each selection method by employing single-degree of freedom contrasts within ANOVA (Steele et al. 1996). To determine the relationship between the traits under selection, phenotypic correlations among traits were calculated by Pearson correlation using PROC CORR of SAS (2003).
Results
We conducted a replicated trial to determine the effect of three selection methods (MAS, PHE, and RAN) for four quantitative traits (earliness, gynoecy, fruit length to diameter ratio, and multiple lateral branching) over three selection cycles in four cucumber populations. The ANOVA of data obtained from the trial indicated that all main effects (planting date, populations, and combinations of cycles and selection methods) were highly significant (P < 0.001) for all traits. In general, planting date affected the magnitude of the mean value of a trait and not the entry ranking in response to selection over cycles. The means of all traits were higher for all populations in the first planting than the second, except for multiple lateral branching, which was lower. Although the planting date by population interaction was significant for length to diameter ratio (P = 0.01) and earliness (P = 0.001), general trends over cycles were the same for both plantings for all traits. Selection was performed in each of the four populations independent of each other, and response to selection varied by population. Therefore, results are presented by population with both plantings combined (Fig. 3; Supplementary Table). Results of selection for yield components, indirect selection for yield, correlated response to selection, and temporal efficiency of selection are presented with regards to responses in each of the four populations and to evaluate the three different types of selection using population performances for comparative analyses.
Selection of yield components
Population response varied with selection method and the trait being evaluated (Fig. 3). The effectiveness of MAS was determined by comparing the regression slope of MAS to that of RAN (Fig. 3). The effectiveness of PHE was determined by a similar comparison to RAN. Comparisons were made separately within each Population for each trait (Fig. 3). In Population 1, PHE was more effective than MAS. The values of two traits increased after MAS (multiple lateral branching and length to diameter ratio) while PHE increased the values of three traits (multiple lateral branching, length to diameter ratio, and earliness). The most effective selection method for Population 2 was PHE. Two trait values increased after PHE (earliness and gynoecy), while none were effectively increased by MAS. The increase in length to diameter ratio and multiple lateral branching after MAS cannot be attributed to selection since a similar increase was observed after RAN. Selection from PHE was least effective in Population 3. Only one trait value effectively increased after PHE (multiple lateral branching) compared to RAN, but trait values of earliness and gynoecy decreased. After MAS, multiple lateral branching values increased and gynoecy decreased similar to PHE, but earliness did not change. These results indicate that MAS was slightly more effective than PHE in Population 3. In Population 4, PHE increased values for earliness while MAS and PHE provided similar results for all other yield component traits. Thus, PHE was more effective than MAS in this population.
In some cases, population response to selection was dependent on the phenotypic difference between parental lines. The four inbred lines (Table 1; Fig. 1) used as parents in this study were specifically chosen because high values for some of the traits under selection complimented low values found in other lines (e.g., 6632E is high for gynoecy and earliness, but low for multiple lateral branching and length to diameter ratio; Table 1; Fig. 3). This disparity among trait values was predictably minimized in the C0 populations (Fig. 3; Supplementary Table). In general, however, populations responded favorably to selection for traits that were inferior in maternal parents, while traits with superior values in the maternal parents either did not change or decreased after selection. For example, in Population 4 where the inbred parent (6995C) was inferior for earliness and gynoecy but superior for length to diameter ratio and multiple lateral branching, both earliness and gynoecy increased after PHE while length to diameter ratio and multiple lateral branching decreased (Fig. 3; Supplementary Table). Populations responded better overall to PHE than MAS at increasing inferior trait values. In contrast, trait values generally decreased or remained unchanged after RAN, regardless of maternal parent values. The three exceptions were an increase in trait values after RAN for length to diameter ratio in Population 2, multiple lateral branching in Population 2, and length to diameter ratio in Population 3.
Indirect selection for yield
Selection for yield components by either MAS or PHE did not increase yield (number of fruit per plant) in the majority of the populations (Fig. 3; Supplementary Table). Indirect selection by PHE was most effective at maintaining yield (Populations 1 and 2), while the only increase in yield came from MAS (Population 3).
Correlated response to selection
Phenotypic trait correlations are important since they can have a dramatic effect on cucumber fruit yield (i.e., source/sink relationships) depending on plant architecture, and the type and intensity of selection. Strong positive and negative phenotypic correlations (r) between yield components were identified after both MAS and PHE and are presented by population in Table 4. Consistent, positive correlations were detected for earliness with gynoecy and yield (r = 0.25–0.70), but earliness was always negatively correlated with multiple lateral branching (r = −0.26 to −0.54). Earliness and length to diameter ratio were usually not correlated. Negative correlations were generally detected for gynoecy with length to diameter ratio and multiple lateral branching (r = −0.07 to −0.64), but gynoecy was typically not correlated with yield. Generally, length to diameter ratio was positively correlated with multiple lateral branching (r = 0.06–0.38). Consistent, positive correlations were identified between length to diameter ratio and yield only in Populations 2 and 4 (r = 0.28–0.34). Yield and multiple lateral branching were normally not correlated.
Temporal efficiency of selection
There can be dramatic differences in the cost of selection (i.e., labor and time) given the life cycle time of cucumber. For instance, selection by MAS required less time to complete than PHE (Fig. 1). Evaluation by PHE under Wisconsin conditions required open-field evaluations of mature plants during the growing season. All four populations were evaluated by PHE simultaneously for each cycle and required 3 months (June–August) from seeding until all data were collected. Recombination required 1 month to establish roots and transplant cuttings and 3 months to induce male flowers and intermate selections to obtain mature seed. Since a field season was necessary for each cycle of PHE, 31 months (June 2001–December 2003) were required to complete three selection cycles. Evaluation by MAS generally required 1 month from seeding to collect all genotypic information. Populations were too large to genotype simultaneously, so they were offset such that genotyping usually occurred in one population while other populations were intermated. Selections from MAS required 3 months for transplanting, induction of male flowers, and intermating to obtain mature seed. In contrast to PHE, all three cycles of MAS were performed for all four populations in a total of 19 months (September 2002–March 2003 and June 2003–May 2004).
Discussion
Empirical studies comparing MAS and PHE for increasing gain from selection in various plant species have provided mixed results. In some cases, MAS was more effective and/or efficient than PHE (e.g., Yousef and Juvik 2001; Yu et al. 2000; Fazio et al. 2003a; Zhang et al. 2006). In other studies, the two methods were considered equal (e.g., Stromberg et al. 1994; Romagosa et al. 1999; Van Berloo and Stam 1999; Willcox et al. 2002; Moreau et al. 2004). In additional studies, MAS was not as effective and/or efficient as PHE (e.g., Hoeck et al. 2003; Davies et al. 2006). In other comparisons, the effectiveness of MAS and PHE varied within the same study (e.g., Eathington et al. 1997; Schneider et al. 1997; Flint-Garcia et al. 2003). Most of these studies, however, did not evaluate selection methods for their efficacy in the improvement of multiple, quantitatively inherited traits over multiple cycles of recurrent selection. We present data herein that provide a comprehensive, comparative evaluation of MAS and PHE for quantitative traits in a vegetable crop species using a selection scheme that is representative of a breeding program.
Considerations for MAS
When selecting for multiple, quantitative traits, the determination of which marker-QTL associations to use in selection may affect the outcome of MAS. In several instances, QTL were so tightly clustered that multiple QTL for different traits were located between adjacent marker loci (e.g., QTL for all traits were linked to CSWCT28 and L18-SNP-H19 as well as OP-AD12-1; Table 3; Fig. 2). As the desired QTL allele came from different parental lines for separate traits (e.g., earliness and gynoecy from Gy-7; multiple lateral branching and length to diameter ratio from H-19 at CSWCT28), strategic decisions were made based on QTL effects and neighboring genes to determine the most appropriate parental type for each marker locus. For CSWCT28, a codominant marker, heterozygotes were selected in an effort to carry QTL alleles from both parents in this region. In another example, the Gy-7 allele was selected at OP-AD12-1, the marker linked to the little leaf gene (ll) from H-19, in order to avoid the deleterious effects of the little leaf type on gynoecy and earliness (Fazio et al. 2003b). Little leaf types, however, typically have more branches than standard leaf types, and the QTL (from H-19) with the greatest effect on multiple lateral branching (LOD = 32.9, R 2 = 32%) is tightly linked (0.7 cM) to ll (Fazio et al. 2003b). Selection of the H-19 allele at OP-AD12-1, therefore, may have resulted in greater gains in multiple lateral branching from MAS, but may, in turn, have negatively affected earliness, gynoecy, and length to diameter ratio, which are associated with the Gy-7 allele.
Another consideration for MAS is marker type. The majority of the RAPD markers used in this study were repeated to provide certainty during genotyping. In contrast, all but one (M8SCAR) of the SNP and SCAR markers could be multiplexed (Table 3), allowing for increased genotyping efficiency. The low repeatability of RAPDs and the advantage of multiplexing for high-throughput genotyping demonstrate the need for SNP, SCAR, and SSR markers for more efficient MAS in cucumber.
Effectiveness of selection for yield components
Each of the four base populations underwent RAN to provide four estimates of genetic drift. Since crosses from RAN followed the same mating scheme as MAS and PHE, RAN serves as a reference to determine the effectiveness of selection by MAS or PHE. When considering all five traits in each of the four populations, 15 of the 20 slopes were significant after RAN, but in only 3 instances did trait values increase (Fig. 3; Supplementary Table). These changes in trait values are most likely due to genetic drift or physiological factors such as source–sink relationships. In the three instances where trait values increased, the similar increase from MAS or PHE cannot be attributed to selection. Although trait values were generally not static in the absence of selection, their general reduction indicates that increases after MAS or PHE can be attributed to a response from selection.
Both MAS and PHE provided improvements in all traits under selection in at least one population, except earliness by MAS (Fig. 3; Supplementary Table). Generally, PHE was most effective for gynoecy, earliness, and length to diameter ratio, while MAS was slightly more effective for multiple lateral branching. Both PHE and MAS were generally effective at improving populations with inferior traits, but not as effective at maintaining traits with high values. Based on trait value changes in response to selection, PHE was more effective than MAS in Populations 1, 2, and 4, but MAS was slightly more effective than PHE in Population 3. Thus, the choice of selection methods for cucumber improvement through plant architectural manipulation (i.e., yield components) will depend upon the populations and traits under selection.
Our results are complementary to those of Fazio et al. (2003b) and Fan et al. (2006), which explore different aspects of incorporating MAS into cucumber breeding programs. Fazio et al. (2003a) compared MAS and PHE for a single trait during backcrossing typically utilized in developing superior cucumber breeding lines. Both selection methods equally improved multiple lateral branching, but MAS was more efficient. We observed a similar response in multiple lateral branching from MAS and PHE in four populations even though we were selecting for multiple traits in a recurrent selection scheme focused on population development. Fan et al. (2006) tested whether MAS could improve multiple yield component traits during backcrossing after two cycles of PHE. The base population and the three cycles of PHE by recurrent selection reported by Fan et al. (2006) are the same as Population 1 described herein. Specific selections from PHE cycle 2 of Population 1 were utilized by Fan et al. (2006) to test the effectiveness of MAS for two backcrossing cycles after PHE. The structure and focus of the Fan et al. (2006) study (tandem selection of PHE in one population by recurrent selection then MAS during backcrossing for breeding line development) is distinct from that presented herein (MAS and PHE in parallel for direct comparison in four populations to mitigate trait correlations using recurrent selection for population improvement). Both studies, nevertheless, indicate the potential of MAS. After two cycles of PHE improved multiple lateral branching and length to diameter ratio in Population 1, Fan et al. (2006) demonstrated that subsequent use of MAS continued to improve these two traits during backcrossing. Although gynoecy was not improved by PHE in Population 1, Fan et al. (2006) showed that MAS can increase femaleness during backcrossing. In this study, MAS increased gynoecy, length to diameter ratio, and multiple lateral branching in at least one population. These combined results confirm the potential value of the marker-QTL associations for selection of these three traits using several breeding strategies. However, the effectiveness of MAS is population-dependent, especially during recurrent selection.
Effectiveness of indirect selection for yield
Yield was not under direct selection in this study, but was evaluated to test the efficacy of indirect improvement by selection for yield components. Indirect selection by MAS or PHE was generally not effective at increasing yield. Nevertheless, the hypothesis that yield increases with the improvement of all four yield components cannot be rejected, since in no instance did improvement of all four traits occur. The challenge to improve yield in cucumber will likely be the simultaneous improvement of yield components using both MAS and PHE.
The simultaneous increase in all four traits under selection in this study is predictably difficult given the negative correlations among some yield component traits. The strength and direction of these correlations have been documented in a wide range of genetic backgrounds (Kupper and Staub 1988; Serquen et al. 1997a; Cramer and Wehner 1998; Cramer and Wehner 1999; Cramer and Wehner 2000b; Fazio et al. 2003b). To mitigate negative correlations among yield components, we intermated four parental inbred lines and employed recurrent selection in four different populations. This strategy was generally ineffective, however, because gynoecy and earliness were positively correlated as were multiple lateral branching and length to diameter ratio in all four populations (Table 4). The correlations among these yield components are most likely due to a combination of pleiotropy with the F, de, and ll genes (Fazio et al. 2003b), and linkage among individual QTL (Robbins and Staub 2004). Fine mapping in regions with clustered QTL would assist in determining the extent of linkage between QTL and identifying molecular markers that could be useful for selecting recombinants between tightly linked QTL (Nam et al. 2005).
We observed a heterotic yield effect in the cucumber populations examined. Yield was higher in every C0 population than the maternal parent that produced it, except in Population 2, which was derived from the highest yielding parent, 6823B (Table 1; Fig. 3). Cucumber is considered a cross-pollinated crop, and although it exhibits little inbreeding depression, heterosis for yield has been observed in a number of cases (Wehner 1989). Using the mean of the four parents (1.81) as the mid-parent value, the mid-parent percent heterosis for yield is 22, 12, 2.6, and 27% for Populations 1–4, respectively. These values are similar to those reported for fruit number in previous studies (Wehner 1989). Given this heterotic yield effect, and the difficulty of simultaneously increasing several yield components, inbred lines with high values for specific yield component combinations could be developed in parallel, and then crossed to create high yielding hybrids. Our results indicate that, while it is difficult to improve all four yield component traits simultaneously, both MAS and PHE can be utilized to improve specific trait combinations such as length to diameter ratio with multiple lateral branching or gynoecy with earliness. This approach would involve extensive combining ability or test cross evaluation of inbred lines in multiple environments, and would likely be population specific.
Efficiency of selection methods in breeding programs
For MAS to be employed in plant improvement programs, it must provide resource (cost/benefit), technical (improved effectiveness), or temporal (efficiency) advantages over PHE. In this study, the cumulative time required to complete three cycles of MAS in all four populations was 19 months as compared to 31 months for PHE (Fig. 1). The increased efficiency of MAS may, in some cases, be an advantage over PHE. For example, the improvement of gynoecy per year in Population 4 was similar between MAS (4.9% per cycle × 3 cycles per year = 14.7% per year) and PHE (12.6% per cycle × 1 cycle per year = 12.6% per year). The efficiency of MAS could be further improved by the use of codominant, single-copy markers that can be multiplexed, such as SCARs, SNPs, and SSRs in combination with high-throughput technologies such as robotics, gel-less assays, microarrays, and pyrosequencing (Gupta et al. 2001; Collard et al. 2005). The resources and methods available at the inception of this project have changed dramatically. A large amount of genomic resources will soon be available for cucumber (Huang et al. 2008) that could greatly increase the efficiency of MAS in marker-assisted recurrent selection (MARS) or genome-wide selection approaches (Bernardo and Charcosset 2006; Bernardo and Yu 2007). However, substantial investments required for high-throughput technologies are currently cost limiting for minor crops such as cucumber. The efficiency of MAS will most likely increase as these genomic tools become more available and affordable.
Recurrent selection is the method of choice for traits with low heritability and has been used extensively for yield improvement in cucumber (Lower and Edwards 1986; Wehner 1989; Cramer and Wehner 1998). Two important considerations for recurrent selection are selection intensity and genetic drift. Selection intensity must be stringent enough to increase desired allele frequencies (make gain from selection), but modest enough to allow diversity to continue improvement in subsequent cycles of selection (Casler 1999; Bernardo 2002). Our results from RAN indicated that selecting 20 out of 600 individuals to obtain high selection intensities resulted in genetic drift for some traits (Fig. 3; Supplementary Table). The evaluation of 600 individuals in each population was the maximum allowable for each method with the resources available in this study. However, evaluating 600 individuals by MAS and 600 by PHE in the same cycle and intermating 40 selections is possible. Using this approach, high selection intensities are maintained while intermating more individuals may overcome genetic drift. In addition, evaluating a greater number of individuals may allow for recombination among tightly linked QTL to overcome negative correlations among traits due to linkage. Thus, selection for improved yield in cucumber may be most effective by combining both MAS and PHE, a conclusion that is supported by previous studies comparing MAS and PHE (Eathington et al. 1997; Bohn et al. 2001; Davies et al. 2006).
References
Bernardo R (2002) Breeding for quantitative traits in plants, 1st edn. Stemma Press, Woodbury
Bernardo R, Charcosset A (2006) Usefulness of gene information in marker-assisted recurrent selection: a simulation appraisal. Crop Sci 46:614–621
Bernardo R, Yu J (2007) Prospects for genomewide selection for quantitative traits in maize. Crop Sci 47:1082–1090
Bohn M, Groh S, Khairallah MM, Hoisington DA, Utz HF, Melchinger AE (2001) Re-evaluation of the prospects of marker-assisted selection for improving insect resistance against Diatraea spp in tropical maize by cross validation and independent validation. Theor Appl Genet 103:1059–1067
Casler MD (1999) Phenotypic recurrent selection methodology for reducing fiber concentration in smooth bromegrass. Crop Sci 39:381–390
Collard BCY, Mackill DJ (2008) Marker-assisted selection: an approach for precision plant breeding in the twenty-first century. Phil Trans R Soc B 363:557–572
Collard BCY, Jahufer MZZ, Brouwer JB, Pang ECK (2005) An introduction to markers, quantitative trait loci (QTL) mapping and marker-assisted selection for crop improvement: The basic concepts. Euphytica 142:169–196
Cramer CS, Wehner TC (1998) Fruit yield and yield component means and correlations of four slicing cucumber populations improved through six to ten cycles of recurrent selection. J Am Soc Hort Sci 123:388–395
Cramer CS, Wehner TC (1999) Little heterosis for yield and yield components in hybrids of six cucumber inbreds. Euphytica 110:99–108
Cramer CS, Wehner TC (2000a) Path analysis of the correlation between fruit number and plant traits of cucumber populations. HortScience 35:708–711
Cramer CS, Wehner TC (2000b) Fruit yield and yield component correlations of four pickling cucumber populations. Cucurbit Genet Coop Rpt 23:12–15
Davies J, Berzonsky WA, Leach GD (2006) A comparison of marker-assisted and phenotypic selection for high grain protein content in spring wheat. Euphytica 152:117–134
de Oliveira EJ, Alzate-Marin AL, Borem A, Fagundes SD, de Barros EG, Moreira MA (2005) Molecular marker-assisted selection for development of common bean lines resistant to angular leaf spot. Plant Breed 124:572–575
Eathington SR, Dudley JW, Rufener GK (1997) Usefulness of marker-QTL associations in early generation selection. Crop Sci 37:1686–1693
Edwards MD, Page NJ (1994) Evaluation of marker-assisted selection through computer simulation. Theor Appl Genet 88:376–382
Fan Z, Robbins MD, Staub JE (2006) Population development by phenotypic selection with subsequent marker-assisted selection for line extraction in cucumber (Cucumis sativus L.). Theor Appl Genet 112:843–855
Fazio G (2001) Comparative study of marker-assisted and phenotypic selecton and genetic analysis of yield components in cucumber. PhD dissertation, University of Wisconsin Madison
Fazio G, Chung SM, Staub JE (2003a) Comparative analysis of response to phenotypic and marker-assisted selection for multiple lateral branching in cucumber (Cucumis sativus L.). Theor Appl Genet 107:875–883
Fazio G, Staub JE, Stevens MR (2003b) Genetic mapping and QTL analysis of horticultural traits in cucumber (Cucumis sativus L.) using recombinant inbred lines. Theor Appl Genet 107:864–874
Flint-Garcia SA, Darrah LL, McMullen MD, Hibbard BE (2003) Phenotypic versus marker-assisted selection for stalk strength and second-generation European corn borer resistance in maize. Theor Appl Genet 107:1331–1336
Francia E, Tacconi G, Crosatti C, Barabaschi D, Bulgarelli D, Dall’Aglio E, Vale G (2005) Marker assisted selection in crop plants. Plant Cell Tissue Organ Cult 82:317–342
Fredrick LR, Staub JE (1989) Combining ability analyses of fruit yield and quality in near-homozygous lines derived from cucumber. J Am Soc Hort Sci 114:332–338
Gimelfarb A, Lande R (1994a) Simulation of marker assisted selection for nonadditive traits. Genet Res 64:127–136
Gimelfarb A, Lande R (1994b) Simulation of marker assisted selection in hybrid populations. Genet Res 63:39–47
Gupta PK, Roy JK, Prasad M (2001) Single nucleotide polymorphisms: a new paradigm for molecular marker technology and DNA polymorphism detection with emphasis on their use in plants. Curr Sci 80:524–535
Hoeck JA, Fehr WR, Shoemaker RC, Welke GA, Johnson SL, Cianzio SR (2003) Molecular marker analysis of seed size in soybean. Crop Sci 43:68–74
Huang S, Du Y, Wang X, Gu X, Xie B, Zhang Z, Wang J, Li R, Li S, Ren Y, Wang J, Yang H, Jin W, Fei Z, Kilian A, Staub JE, van der Vossen E, Li G (2008) The cucumber genome initiative—an international effort to unlock the genetic potential of an orphan crop using novel genomic technology. In: Abstracts of the plant and animal genomes XVI conference, San Diego, California, 12–16 January 2008
Kuchel H, Fox R, Reinheimer J, Mosionek L, Willey N, Bariana H, Jefferies S (2007) The successful application of a marker-assisted wheat breeding strategy. Mol Breed 20:295–308
Kupper RS, Staub JE (1988) Combining ability between lines of Cucumis sativus L. and Cucumis sativus var. hardwickii (R.) Alef. Euphytica 38:197–210
Lande R, Thompson R (1990) Efficiency of marker-assisted selection in the improvement of quantitative traits. Genetics 124:743–756
Lower RL, Edwards MD (1986) Cucumber breeding. In: Bassett MJ (ed) Breeding vegetable crops. AVI, Westport, pp 173–207
Mohring S, Salamini F, Schneider K (2004) Multiplexed, linkage group-specific SNP marker sets for rapid genetic mapping and fingerprinting of sugar beet (Beta vulgaris L.). Mol Breed 14:475–488
Moreau L, Charcosset A, Gallais A (2004) Experimental evaluation of several cycles of marker-assisted selection in maize. Euphytica 137:111–118
Nam YW, Lee JR, Song KH, Lee MK, Robbins MD, Chung SM, Staub JE, Zhang HB (2005) Construction of two BAC libraries from cucumber (Cucumis sativus L.) and identification of clones linked to yield component quantitative trait loci. Theor Appl Genet 111:150–161
Nijs APMD, Visser DL (1980) Induction of male flowering in gynoecious cucumbers (Cucumis sativus L.) by silver ions. Euphytica 29:273–280
Polashock JJ, Vorsa N (2002) Development of SCAR markers for DNA fingerprinting and germplasm analysis of American cranberry. J Am Soc Hort Sci 127:677–684
Robbins MD (2006) Molecular marker development, QTL pyramiding, and comparative analysis of phenotypic and marker-assisted selection in cucumber. Dissertation, University of Wisconsin Madison
Robbins MD, Staub JE (2004) Strategies for selection of multiple, quantitatively inherited yield components in cucumber. In: Lebeda A, Paris HS (eds) Progress in cucurbit genetics and breeding research. proceedings of Cucurbitaceae 2004, the 8th EUCARPIA meeting on cucurbit genetics and breeding. Palacky University, Olomouc, pp 401–408
Robbins MD, Staub JE, Fazio G (2002) Deployment of molecular markers for multi-trait selection in cucumber. In: Cucurbitaceae 2002. American Society of Horticultural Science Press, Alexandria, pp 41–47
Romagosa I, Han F, Ullrich SE, Hayes PM, Wesenberg DM (1999) Verification of yield QTL through realized molecular marker-assisted selection responses in a barley cross. Mol Breed 5:143–152
SAS (2003) SAS software, Version 9.1 for Windows. Copyright© 2002–2003 by SAS Institute Inc., Cary, NC
Schneider KA, Brothers ME, Kelly JD (1997) Marker-assisted selection to improve drought resistance in common bean. Crop Sci 37:51–60
Serquen FC, Bacher J, Staub JE (1997a) Genetic analysis of yield components in cucumber at low plant density. J Am Soc Hort Sci 122:522–528
Serquen FC, Bacher J, Staub JE (1997b) Mapping and QTL analysis of horticultural traits in a narrow cross in cucumber (Cucumis sativus L.) using random-amplified polymorphic DNA markers. Mol Breed 3:257–268
Shetty NV, Wehner TC (2002) Screening the cucumber germplasm collection for fruit yield and quality. Crop Sci 42:2174–2183
Staub JE, Crubaugh LK (2001) Cucumber inbred line USDA 6632E. Cucurbit Genet Coop Rpt 24:6–7
Staub JE, Knerr LD, Hopen HJ (1992) Plant density and herbicides affect cucumber productivity. J Am Soc Hort Sci 117:48–53
Staub JE, Crubaugh LK, Fazio G (2002) Cucumber recombinant inbred lines. Cucurbit Genet Coop Rpt 25:1–2
Staub JE, Robbins MD, Chung S, López-Sesé AI (2004) Molecular methodologies for improved genetic diversity assessment in cucumber and melon. Acta Hort 637:41–47
Steele RGD, Torrie JH, Dickey DA (1996) Principles and procedure in statistics, 3rd edn. McGraw-Hill, New York
Steele KA, Price AH, Shashidhar HE, Witcombe JR (2006) Marker-assisted selection to introgress rice QTLs controlling root traits into an Indian upland rice variety. Theor Appl Genet 112:208–222
Stromberg LD, Dudley JW, Rufener GK (1994) Comparing conventional early generation selection with molecular marker assisted selection in maize. Crop Sci 34:1221–1225
Tang SX, Kishore VK, Knapp SJ (2003) PCR-multiplexes for a genome-wide framework of simple sequence repeat marker loci in cultivated sunflower. Theor Appl Genet 107:6–19
Van Berloo R, Stam P (1999) Comparison between marker-assisted selection and phenotypical selection in a set of Arabidopsis thaliana recombinant inbred lines. Theor Appl Genet 98:113–118
van Berloo R, Stam P (2001) Simultaneous marker-assisted selection for multiple traits in autogamous crops. Theor Appl Genet 102:1107–1112
Wehner TC (1989) Breeding for improved yield in cucumber. Plant Breed Rev 6:323–359
Wehner TC, Lower RL, Staub JE, Tolla GE (1989) Convergent-divergent selection for cucumber fruit yield. HortScience 24:667–669
Willcox MC, Khairallah MM, Bergvinson D, Crossa J, Deutsch JA, Edmeades GO, Gonalez-de-Leon D, Jiang C, Jewell DC, Mihm JA, Williams WP, Hoisington D (2002) Selection for resistance to southwestern corn borer using marker-assisted and conventional backcrossing. Crop Sci 42:1516–1528
Xu YB, Crouch JH (2008) Marker-assisted selection in plant breeding: from publications to practice. Crop Sci 48:391–407
Yousef GG, Juvik JA (2001) Comparison of phenotypic and marker-assisted selection for quantitative traits in sweet corn. Crop Sci 41:645–655
Yu K, Park SJ, Poysa V (2000) Marker-assisted selection of common beans for resistance to common bacterial blight: efficacy and economics. Plant Breed 119:411–415
Zhang W, Smith C (1992) Computer simulation of marker-assisted selection utilizing linkage disequilibrium. Theor Appl Genet 83:813–820
Zhang C, Tar’an B, Warkentin T, Tullu A, Bett KE, Vandenberg B, Somers DJ (2006) Selection for lodging resistance in early generations of field pea by molecular markers. Crop Sci 46:321–329
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by I. Paran.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Rights and permissions
About this article
Cite this article
Robbins, M.D., Staub, J.E. Comparative analysis of marker-assisted and phenotypic selection for yield components in cucumber. Theor Appl Genet 119, 621–634 (2009). https://doi.org/10.1007/s00122-009-1072-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00122-009-1072-8