Printer Friendly

Comparison of Phenotypic and Marker-Assisted Selection for Quantitative Traits in Sweet Corn.

USING LINKAGE MAPS constructed with DNA marker loci, it is possible to estimate the location and effect of individual loci influencing the expression of quantitative traits (Breto et al., 1994; Danesh et al., 1994; Xiao et al., 1996). Statistical associations between alleles at marker loci and alleles at linked quantitative trait loci (QTL) can be used to select indirectly, but with potentially high accuracy, for favorable QTL alleles, effectively increasing the heritability of traits (Soller and Beckmann, 1983; Stuber et al., 1992). While many studies have been conducted to map QTL of important traits in maize (Bernardo, 1999), very few empirical applications of marker-assisted selection (MAS) in breeding programs have been reported. Most studies employing MAS have been theoretical, using computer simulations (Lande and Thompson, 1990; Zhang and Smith, 1992; Edwards and Page, 1994; Gimelfarb and Lande, 1994), the results of which suggest that MAS can be used to improve quantitative traits in animal- and plant-breeding programs.

In theory, MAS is proposed to be more efficient than phenotypic selection (PS) when the heritability of a trait is low, where there is tight linkage between QTL and DNA markers (Dudley, 1993; Knapp, 1998), with larger population sizes (Moreau et al., 1998), and in earlier generations of selection before recombinational erosion of marker-QTL associations (Lee, 1995). Edwards and Page (1994) reported, using computer simulation, that MAS produced rapid gains early in the selection process compared with phenotypic recurrent selection (PRS); however the rate of gain diminished greatly within three to five cycles. They proposed that the distance between markers and QTL was the factor that most limited gains from MAS. Moreau et al. (1998) reported that MAS was ineffective for traits with heritabilities below 20% due to the low power of quantitative trait loci detection and the bias caused by the selection of markers. Knapp (1998) proposed that MAS would increase the probability of selecting superior genotypes and substantially decrease the resources required in breeding for a trait with low to moderate heritability.

A limited number of empirical experiments have provided equivocal results regarding the relative efficiency of MAS and PS. Using [S.sub.4:5] maize families selected by an index based on marker effects in [F.sub.2] testcrosses or phenotypic data, MAS was found as effective as PS, but neither method identified lines better than the original hybrid (Stromberg et al., 1994). Suggested as reasons for the poor selection ability of MAS were different testing environments for the [F.sub.2] and [S.sub.4:5] testcrosses, limited coverage of the genome with markers, and small population size (20 [S.sub.4:5] families). Indices combining phenotypic and marker information were successful in selecting superior [S.sub.4:5] lines for grain yield and stalk lodging in maize compared with phenotypic performance alone (Eathington et al., 1997). However, they reported that marker information was able to improve selection gain over phenotypic data only in high-yielding environments. Comparable selection gain from MAS and PS was reported in improving flowering time of Arabidopsis, which was attributed to the high heritability of the trait (Van Berloo and Stam, 1999). Zhu et al. (1999) reported significant effects of most QTL selected using MAS in a barley-breeding program for enhanced grain yield, although progress was hindered by large QTL x environment interactions. These reports suggest that further empirical investigations are required to evaluate the merits of MAS as an adjunct to PS or as an exclusive selection procedure for rapid selection during crop-breeding programs. This is particularly true when the objective is to select for several traits simultaneously, requiring the manipulation of gene frequencies at numerous loci across the genome.

In sweet corn, breeding for improved seedling emergence and eating quality is complicated because of the relationship between these traits. High kernel sugar concentration was proposed as one of the reasons for poor seedling emergence (Douglass et al., 1993). These traits are influenced by many kernel characteristics and under the control of many genes (Azanza et al., 1996b). However, since field emergence and eating quality depend on kernel chemical and physiological characteristics, selection for improved seedling emergence without concurrent attention to the fresh quality of the harvested product will be of limited value to the sweet corn industry. In this study, we compared MAS and PS for the simultaneous improvement of seedling emergence and eating-quality characteristics.

MATERIALS AND METHODS

Development of the Composite Populations ([C.sub.1])

Two sweet corn [F.sub.2:3] populations previously developed from crosses between inbreds provided the base populations and genetic information for this study. One of the populations, homozygous for sugary1 and segregating for sugary enhancer 1 (W6786su1Se1 x IL731asu1se1), consisted of 214 [F.sub.2:3] families (Azanza et al., 1996b) and the second homozygous for shrunken2 (IL451b sh2 x Ia453 sh2) of 117 [F.sub.2:3] families (Han, 1994). A near-saturated genetic linkage map of 88 RFLPs, three cDNA, and two morphological markers was constructed for the first population (Tadmor et al., 1995). In the second population, 61 polymorphic RFLP markers were used to generate a second linkage map (Han, 1994). Key alleles affecting seedling emergence, kernel eating quality, and other traits were identified and mapped in both populations (Han, 1994; Tadmor et al., 1995; Azanza et al., 1996b).

The 214 [F.sub.2:3] families in the first population were classified into three subpopulations according to segregation for se1, where families were either homozygous for Se1 (su1su1Se1Se1, sugary1), heterozygous for Se1 (su1su1Se1se1), or homozygous for se1 (su1su1se1se1, sugary enhancer 1). This classification was made to standardize the genetic background of each subpopulation regarding se1 since this gene is known to exert major epistatic influences on kernel characteristics and seedling emergence (Azanza, 1994; Wang, 1997). In this population, 38 of the [F.sub.2:3] families were homozygous for Se1 (su1su1Se1Se1) and 48 [F.sub.2:3] families homozygous for the mutant allele (su1su1se1se1). These two subpopulations were used as two base populations. In the second population, all 117 [F.sub.2:3] families were homozygous for sh2 and used as the third base population in this study. Selection using MAS and PS was applied to these three base populations separately.

Selection for single traits was applied for seedling emergence, kernel sucrose concentration at 20 d after pollination (20 DAP), and cooked kernel tenderness (20 DAP) in the su1Se1 and su1se1 populations while selection was applied only for seedling emergence in the sh2 population. Multiple trait selection using MAS and PS was applied for simultaneous improvement of (i) emergence and tenderness, (ii) emergence and hedonic rating, and (iii) emergence, sucrose, and tenderness in the su1Se1 and su1se1 populations, In the sh2 population, selection for two traits was conducted only for emergence and tenderness. The breeding scheme used to develop the [C.sub.1] composites of single and multiple traits based on PS or MAS is presented in Fig. 1.

[ILLUSTRATION OMITTED]

Phenotypic selection for single traits was based on the mean performance of replicated [F.sub.2:3] families from three and two replications for seedling emergence and kernel quality characteristics, respectively, in the Illinois environment (Han, 1994; Azanza et al., 1996a). Marker-assisted selection for single traits was based on segregation of five RFLP marker loci linked to QTL associated with significant effects. Single-factor analysis was conducted to test for the significant associations between marker-QTL and traits in the su1Se1, su1se1, and sh2 base population. Markers were selected with trait associations of P [is less than] 0.05 and located on different chromosomes or with [is greater than] 50 centimorgan separation, if on the same chromosome (Yousef, 2000). A genotypic value for each family was calculated by summing the additive effects of the five beneficial marker alleles associated with each trait. To test for repeatability of MAS, selection among families for seedling emergence in the three populations was applied using another set of markers significantly associated with this trait.

In each population, families were sorted based on their phenotypic (PS) or genotypic values (MAS). Then 20% of the families with the highest and lowest values for each trait were chosen. Those families were used to generate [C.sub.1] composites of high and low phenotypic or genotypic performance. Q-gene software (Nelson, 1996) was used to provide a list of the three genotype class means for each marker, probability values, and estimates of gene action at each associated RFLP marker in the three base populations.

For multiple-trait selection, traits were considered of equal economic value in the selection index. For each trait, phenotypic or genotypic values of each family were standardized into Z values using the formula:

Z = (family value - population mean)/ standard deviation of population.

The family Z score indicates where an individual family's performance lies in relation to the population mean in units of standard deviation. The selection index was calculated by summing the standardized phenotypic or genotypic values of families for each trait resulting in a PS and MAS index, respectively. Families in each population were then sorted on the basis of their PS or MAS index scores. Twenty percent of the families with the highest and lowest index scores were selected to constitute the [C.sub.1] composites of high and low performance for multiple traits.

The number of selected [F.sub.2:3] families (20%) was 8, 10, and 23 from the su1Se1, su1se1, and sh2 base populations, respectively, and these were used to constitute [C.sub.1] composites for single and multiple traits. On average, MAS and PS paired composites were composed of 43% of the same [F.sub.2:3] families. The percentage of composite composition of shared families ranged from 25 to 60%. From each base population, 50% of the families were randomly selected to establish a control or [C.sub.0] composite from seed grown in the same environment. A list of the created composites and the marker loci used for MAS are presented in Table 1. The average beneficial allele frequencies in suSe1, su1se1, and sh2 were 51, 51, and 54% in the base populations vs. 48, 52, and 53% in the randomly selected [C.sub.0] controls, respectively. The similarity of these allelic frequencies between the base populations and the [C.sub.0] controls suggests that drift is not a major factor in the experiment results.
Table 1. Composites created in each of su1Se1, su1se1 and sh2 base
populations by means of PS and MAS and RFLP markers associated with
the selected traits.

                                             MAS

                                                 RFLP markers
                    PS                           and [R.sup.2]

                [C.sub.1]        [C.sub.1]
                composite        composite
Trait          ID([dagger])     ID([dagger])         su1Se1

                               EMHA
Emergence     EPH (su1Se1)     (su1Se1)         bnl6.32(18%),
  (EMG)                          ([double         ([sections])
                                 dagger])
              EPL (su1Se1)     EMLA (su1Se1)    umc53(19%),
              EPH (su1se1)     EMHA (su1se1)    umc50(13%),
              EPL (su1se1)     EMLA (su1se1)    npi268(19%),
              EPH (sh2)        EMHA (sh2)       bnl7.49(21%).
              EPL (sh2 )       EMLA (sh2)

                    --         EMHB (su1Se1)    bnl632(18%),
                                 ([double
                                 dagger])
                    --         EMLB (su1Se1)    umc139(15%),
                    --         EMHB (su1se1)    umc60(13%),
                    --         EMLB (su1se1)    npi570(19%),
                    --         EMHB (sh2)       npi209(15%).
                    --         EMLB (sh2).

Sucrose       SPH (su1Se1)     SMH (su1Se1)     umc50(50%),
  (SUC)       SPL (su1Se1)     SML (su1Se1)     umc90(33%),
              SPH (su1se1)     SMH (su1Se1)     bnl5.71(17%),
              SPL (su1se1)     SML (su1se1)     npi268(28%),
                                                umc114(23%).

Tenderness    TPH (su1Se1)     TMH (su1Se1)     umc84(23%),
  (TND)       TPL (su1Se1)     TML (su1Se1)     umc139(14%),
              SPH (su1se1)     TMH (su1se1)     umc31a(23%),
              SPL (su1se1)     TML (su1se1)     npi593(17%),
                                                npi268(22%).

EMG/TND       ETPH (su1Se1)    ETMH (su1Se1)    EMG and TND
                                                  ([sections])
              ETPL (su1Se1)    ETML (su1Se1)    markers.
              ETPH (su1se1)    ETMH (su1se1)    (listed above)
              ETPL (su1se1)    ETML (su1se1)
              ETPH (sh2)       ETMH (sh2)
              ETPL (sh2)       ETML (sh2)

EMG/Hedonic   EDPH (su1Se1)    EDMH (su1Se1)    EMG markers and:
  rating      EDPL (su1Se1)    EDML (su1Se1)    umc42b(21%),
              EDPH (su1se1)    EDMH (su1se1)    umc60(26%),
              EDPL (su1se1)    EDML (su1se1)    umc90(24%),
                                                bnl16.06(32%),
                                                umc109(27%).

EMG/SUC/TND   ESTPH (su1Se1)   ESTMH (su1Se1)   EMG, SUC,
                                                  ([paragraph])
              ESTPL (su1Se1)   ESTML (su1Se1)   and TND
              ESTPH (su1se1)   ESTMH (su1se1)   markers.
              ESTPL (su1se1)   ESTML (su1se1)   (listed above)

                                               MAS

                    PS              RFLP markers and [R.sup.2]

                [C.sub.1]
                composite
Trait          ID([dagger])         su1se1              sh2

Emergence     EPH (su1Se1)     umc57c(18%),       php200689(10%),
  (EMG)

              EPL (su1Se1)     bnl5.71(11%),      umc131(11%),
              EPH (su1se1)     npi240(19%),       php200075b(7%),
              EPL (su1se1)     npi276b(22%),      umc116(10%),
              EPH (sh2)        umc114(12%).       bnl9.08(10%).
              EPL (sh2 )

                    --         umc57c(18%),       php200689(10%),
                    --         npi408(11%),       php200075b(7%),
                    --         npi240(29%),       umc116(10%),
                    --         npi585(23%),       umc160b(7%),
                    --         umc109(10%).       bnl304(7%).
                    --

Sucrose       SPH (su1Se1)     umc50(21%),               --
  (SUC)       SPL (su1Se1)     umc31a(11%),              --
              SPH (su1se1)     npi560(19%),              --
              SPL (su1se1)     php100016(14%),           --
                               npi209(17%).              --

Tenderness    TPH (su1Se1)     umc49(27%),               --
  (TND)       TPL (su1Se1)     npi420(18%),              --
              SPH (su1se1)     bnl8.15(19%),             --
              SPL (su1se1)     npi240(26%),              --
                               sh1(20%).                 --

EMG/TND       ETPH (su1Se1)    EMG and TND        EMG markers and:
              ETPL (su1Se1)    markers.           bnl6.2(8%),
              ETPH (su1se1)    (listed above)     php200075b(5%),
              ETPL (su1se1)                       umc160b(8%),
              ETPH (sh2)                          sh1(8%),
              ETPL (sh2)                          umc152(9%).

EMG/Hedonic   EDPH (su1Se1)    EMG markers and:
  rating      EDPL (su1Se1)    bnl10,38(24%),            --
              EDPH (su1se1)    umc49(26%),               --
              EDPL (su1se1)    npi434(26%),              --
                               npi263(25%),              --
                               umc114(30%).              --

EMG/SUC/TND   ESTPH (su1Se1)   EMG, SUC,                 --
              ESTPL (su1Se1)   and TND                   --
              ESTPH (su1se1)   markers.                  --
              ESTPL (su1se1)   (listed above)            --

[C.sub.0] control composites: (su1Se1), (su1se1), and (sh2)(#)

([dagger]) E = emergence, P = PS, H = high direction, L = low
direction, M = MAS, S = sucrose, T = tenderness, D = hedonic.

([double dagger]) A and B refer to the two replicates of MAS using two
different sets of markers.

([sections]) Values in parentheses are the percentages of the total
phenotypic variation explained by the specific marker locus ([R.sup.2])
in the [F.sub.2:3] population.

([paragraph]) Markers of two and three traits included EMG markers
selected in the first set.

(#) Composites created by randomly intermating 50% of families in each
base population.


One hundred kernels of each selected [F.sub.2:3] family were planted in the greenhouse in mid-June 1996. Seedlings were transplanted into field plots at the University of Illinois' South Farm in late June. This minimized unintentional selection from direct seed planting under field conditions and to ensure that each [F.sub.2:3] family was equally represented. At anthesis, pollen was collected from 25 to 30 plants in each family within each composite, bulked, and used to randomly pollinate 8 to 10 ears per family. An equal number of mature-dry kernels from ears of each [F.sub.2:3] family in the particular composite were bulked to constitute the [C.sub.1] and [C.sub.0] composites in the three su1Se1, su1se1, and sh2 populations. This was done to generate balanced composites where each [F.sub.2:3] family was equally represented. Collectively, 27 composites (26 [C.sub.1] and 1 [C.sub.0]) in each of the su1Se1 and su1se1 base populations, and 11 composites (10 [C.sub.1] and 1 [C.sub.0]) in sh2 base population were generated and used to evaluate the efficiency of MAS and PS in improving quantitative trait performance over one cycle of selection.

Evaluation of the [C.sub.1] Composites

Seedling Emergence

Replicates of 100 kernels of each of the [C.sub.0] and [C.sub.1] composites were planted in four environments; three in Illinois (1 May, 10 June, and 15 Sept. 1997) and one at the University of Wisconsin (7 June 1997). Kernels were planted using hand planters at about 4-cm soil depth in Drummer silty clay loam soil in Illinois environments and Plano silt loan soil in Wisconsin. The average daily soil and air soil temperature was above 15 [degrees] C in all plantings, except in the May planting (9 [degrees] C). The experimental design was a randomized complete block design (RCBD) within each environment. All composites were evaluated for seedling emergence in three replications except for the Fall 1997 planting that had eight replications. Seedling emergence was determined by direct counts at 3 to 4 wk after planting in the four environments.

Eating Quality Evaluations

Eating quality traits, which included kernel sucrose concentration, cooked kernel tenderness, and hedonic rating (taste panel preference), were evaluated on ears harvested from three replicated plots on the South Farm of the University of Illinois at Urbana-Champaign in 1997 and 1998. To minimize unintentional selection due to cold weather effects on sweet corn emergence and plant vigor, plantings were direct seeded when mean daily soil and air temperature exceeded 15 [degrees] C (10 June 1997 and 15 June 1998). At anthesis, pollen was collected from 25 to 30 plants and mixed. The bulk was used to pollinate 10 random ears within the composite. Pollinated ears were harvested 20 d after pollination (DAP) for evaluation. The harvesting period did not exceed 1 wk; therefore differential accumulation of heat units was not considered an important source of variation in eating quality traits. The harvested ears were promptly placed on dry ice and transported from the field and temporarily stored in a -20 [degrees] C freezer. To facilitate removing kernels from cobs and to eliminate enzymatic activity, ears were immersed in liquid nitrogen for about 2 min until hard frozen. An equal amount (50 gm) of frozen kernels from each ear within each composite were bulked, sealed in freezer bags, and stored at -80 [degrees] C for subsequent chemical, physiological, and sensory analysis.

Kernel sucrose concentrations were estimated as milligrams of sucrose per gram freeze-dried tissue using an 80% ethanol extraction procedure described by Juvik and La Bonte (1988). Two samples of 25 g ([approximately equals] 100 kernels each) of frozen tissue per replication were analyzed since some variation was expected to occur within composites. Extracted corn samples were injected into a high-pressure liquid chromatography, where peak areas were recorded and used to calculate sucrose concentration. Sucrose concentrations of samples were calculated using a simple regression function with a range of eight sucrose standards of known concentrations.

Cooked kernel tenderness was evaluated using a Kramer Shear Press (Instron, Canton, MA), which measures sample texture (Azanza et al., 1996a). Two samples of 30 g per replication were cooked for 4 min in boiling water, air-dried for 1 min, and placed in the rectangular cavity of the machine. The force required to push the multibladed fixture through the cooked samples was estimated in kilograms by measuring the area under the curve recorded on attached chart paper.

Hedonic rating (taste panel preference) evaluation was conducted on the emergence/hedonic [C.sub.1] composites and the [C.sub.0] controls of both the su1Se1 and su1se1 composites, using test conditions and procedures described by Azanza et al. (1996b). Thirteen panelists were asked to mark a scale ranging from 1 (extreme dislike) to 9 (extreme like), indicating how much in general they preferred the cooked sample as an estimate of overall preference.

Selection Efficiencies of PS and MAS

Selection efficiencies of PS and MAS were compared for selection gain and comparative costs. Selection gain was calculated as percent increase in the [C.sub.1] composites of the high direction over the [C.sub.0] control, [([C.sub.1] - [C.sub.0])/[C.sub.0]] x 100. In the low direction, selection gain was estimated as percent decrease in the [C.sub.1] composites from the [C.sub.0] control, [([C.sub.0] - [C.sub.1)/[C.sub.0]] x 100.

Generalized cost comparisons between MAS and PS are difficult to achieve. In this study, the average costs required to evaluate and select a family based on MAS and PS were estimated in the mapping and [C.sub.1] populations. These estimates were based on the original su1se1 population size of 214 families with 94 markers for the 65 different traits that were evaluated in this population. An estimated cost of phenotyping and genotyping one family/composite for each trait was obtained and then added to the cost of evaluating the family/composite in the first cycle. In the case of PS, this included initial costs incurred for labor and field supplies and maintenance used in the base population and [C.sub.1] evaluation. Emergence evaluations included the initial field costs and actual germination counts. Eating quality evaluations (sucrose, tenderness, and hedonic rating) included, in addition to the initial field costs, labor and laboratory chemicals and supplies. Pay rates were estimated at $15/hour since chemical and sensory evaluations required experienced laborers while the assistance of a sensory technician was estimated at $25/hour. Assuming further selection cycles were to be conducted, the base population costs were divided by the number of cycles. With MAS, the costs included mapping the original population (phenotypic and marker analysis) and screening one family/composite for the number of the selected markers (five RFLP markers) in the [C.sub.1] of this study. The original mapping costs were divided among the number of projected cycles. The estimation of marker genotyping costs included greenhouse bench space and labor for sowing seeds, harvesting tissue, and labor for DNA extraction and RFLP analyses. Estimates were based on a pay rate of $15/hour as in PS evaluation since these analyses required experienced laboratory help.

Statistical Analysis

For statistical analysis, data of [C.sub.1] composites resulting from single-trait selection and the respective [C.sub.0] control were grouped and analyzed as separate data sets. When [C.sub.1] composites created based on multiple-trait selection included a trait (for example, emergence), data of that trait among [C.sub.1] composites and the respective control were grouped and analyzed as one data set. Each data set contained five composites: two from MAS (high and low), two from PS (high and low), and one [C.sub.0] control except with emergence-only selected composites. The number of emergence-only selected composites was seven where another replicate of two MAS composites (high and low) was created using a second set of RFLP markers. The statistical models used to analyze the data sets are presented in Table 2. Analysis of variance was conducted using PROC GLM procedures (SAS, 1991) with factors fixed except replications. Mean comparisons for traits were made on the basis of combined data over environments. The null hypothesis ([H.sub.0]) was tested at P [is less than] 0.05 where composite means were compared using least significant differences (protected LSD).
Table 2. Mean squares and significance of sources of variation from
ANOVA([dagger]) of seedling emergence, kernel sucrose and
tenderness, and hedonic rating in composites selected for single
or combinations of the four traits in the su1Se1, su1se1, and sh2
populations in Illinois in 1997 and 1998.

                             Seedling emergence (EMG)

                                       Composites

                                         EMG
Source of
variation                df    su1Se1      su1se1       sh2

Environment (E)           3   12 314(*)   18 162(*)   13 941(*)
Replication within
  E [R(E)]               13       30          79(*)       53
Composite (C)             6      380(*)    1 082(*)    1 249(*)
C x E interaction        18      138(*)       83(*)       46
Experimental error       78       32          31          36
Variation explained by
  model ([R.sup.2])               94%         96%         95%
Coefficient of
  variation (C.V.%)                8.0         8.6        11.4

                               Seedling emergence (EMG)

                                     Composites

                                       EMG/TND

Source of
variation                df    su1Se1     su1se1       sh2

Environment (E)           3   7 348(*)   12 986(*)   11 600(*)
Replication within
  E [R(E)]               13      54          29          64
Composite (C)             4     211(*)       38         228(*)
C x E interaction        12      29          74          48
Experimental error       52      42          39          43
Variation explained by
  model ([R.sup.2])              92%         95%         94%
Coefficient of
  variation (C.V.%)               9.2         9.4        12.0

                         Seedling emergence (EMG)

                                Composites

                                EMG/HED

Source of
variation                df   su1Se1     su1se1

Environment (E)           3   8 246(*)   12 871(*)
Replication within
  E [R(E)]               13      22          45
Composite (C)             4     126(*)      131(*)
C x E interaction        12      75(*)       48
Experimental error       52      29          32
Variation explained by
  model ([R.sup.2])              95%         96%
Coefficient of
  variation (C.V.%)               7.7         8.4

                         Seedling emergence (EMG)

                              Composites

                               EMG/SUC/TND

Source of
variation                df   su1Se1     su1se1

Environment (E)           3   7 435(*)   11 750(*)
Replication within
  E [R(E)]               13      55          19
Composite (C)             4      77         177(*)
C x E interaction        12      51          52(*)
Experimental error       52      43          20
Variation explained by
  model ([R.sup.2])              91%         97%
Coefficient of
  variation (C.V.%)               9.1         6.5

                           Kernel sucrose (SUC)

                               Composites

                        SUC                 EMG/SUC/TND

Source of
variation   df    su1Se1      su1se1      su1Se1      su1se1

E            1   10 015(*)   12 687(*)   12 250(*)   13 815(*)
R(E)         4      219         367         155       1 924(*)
C            4    5 230(*)   20 308(*)    2 336(*)    8 916(*)
C x E        4      309         756(*)      972(*)      639(*)
Error       46      128         265          92         186

[R.sup.2]            85%         89%         86%         87%
C.V.%                 7.4         7.7         6.6         6.6

                     Kernel tenderness (TND)

                             Composites

                       TND               EMG/TND

Source of
variation   df   su1Se1    su1se1    su1Se1    su1se1

E            1    417(*)   455(*)     682(*)   159(*)
R(E)         4     46       11          4        7
C            4    881(*)   383(*)     476(*)   178(*)
C x E        4     19       42(*)      68(*)    33(*)
Error       46     19       11         14        8

[R.sup.2]          83%      82%        82%      73%
C.V.%               6.4      4.9        7.3      4.6

               Kernel tenderness (TND)

                   Composites

              EMG/TND      EMG/SUC/TND

Source of
variation      sh2       su1Se1   su1se1

E              1518(*)   302(*)   429(*)
R(E)             38       28(*)     5
C               260(*)    64(*)    47(*)
C x E            63       28(*)   109(*)
Error            30        8       10

[R.sup.2]        68%      69%      70%
C.V.%             5.6      4.1      4.6

            Hedonic rating (HED)

                Composite

                 EMB/HED

Source of
variation    df   su1Se1    su1se1

E             1   38.7(*)   72.4(*)
C             4    2.1      12.0(*)
P            12   14.5(*)   23.1(*)
C x P        48    1.3       1.6
C x E         4    4.4(*)    9.5(*)
Error       320    1.7       1.9
[R.sup.2]         36%       46%
C.V.%             24.8      22.0

(*) = Significant at P < 0.05 probability level.

([dagger]) Statistical model used for ANOVA of emergence, sucrose,
and tenderness: [y.sub.ijk] = [Mu] + [E.sub.i] + R[(E).sub.(i)j]
+ [[Delta].sub.(i)j] + [C.sub.k] + [CE.sub.ik] +
[[Epsilon].sub.(ijk)], and hedonic rating: [y.sub.ijk] = [Mu] +
[E.sub.I] + [P.sub.j] + [[Delta].sub.(i)j] + [C.sub.k] +
[CE.sub.ik] + [CP.sub.jk] + [[Epsilon].sub.(ijk)] [multiplied by]
(y = response, [Mu] = overall mean, E = environment,
R = replication, C = composite, P = panelist, [Delta] = restriction
error, [Epsilon] = experimental error).


RESULTS AND DISCUSSION

Sources of Variation

Analysis of variance was conducted to examine the sources of variation associated with each trait (Table 2). In most cases, differences among environments were responsible for the greatest source of variation. The proportion of the variation that was explained by environmental effects ranged from 30 to 97% of the total variation described in the model. Excluding the environment effects, the main source of variation for emergence, kernel sucrose, and kernel tenderness was attributed to differences among composites. Variation due to composites ranged from 0.3 to 64% of the variation explained by the model. Increasing the number of selected traits resulted in reduction in the proportion of variation associated with composites compared with single-trait-selected composites. Interactions between composites and environments (C x E) were significant in 60% of data sets. In the su1Se1, su1se1, and sh2 [C.sub.1] populations, the variation attributed to the composites were 8.9, 8.1, and 12.0 times greater than the variation attributed to C x E interactions. C x E interactions were minor compared with the variation described by the composites.

Since the objective of this study was to compare the efficiency of two selection methods and not to evaluate composite performance across environments, the statistical analysis was performed on combined data across environments. Environmental variation in emergence, sucrose, and tenderness tends to agree with previous reports indicating that these traits are highly influenced by environmental conditions (Douglass et al., 1993; Azanza et al., 1996a). Variation associated with hedonic values indicated that a large amount of variation and error was not described by the model. This agrees with earlier work (Azanza et al., 1994) suggesting that sweet corn hedonic rating is a complicated trait controlled by the interaction of kernel sugar content, tenderness, aroma, and other attributes with relatively low heritability.

Comparative Evaluation of MAS and PS

Selection Gain in Single Traits

Seedling emergence of the [C.sub.1] composites (EPH, EPL, EMHA, EMLA, EMHB, EMLB) in the emergence-only selected composites and [C.sub.0] control in the three populations are presented in Table 3. In the high direction, mean emergence across environments and replications revealed that MAS enhanced seedling emergence significantly in two of three populations compared with PS. The [C.sub.1] composites (EMH) showed an increase in emergence over [C.sub.0] of 8.3 and 14.9% under MAS in the su1Se1 and su1se1 [C.sub.1] composites, respectively. In both su1Se1 and su1se1 [C.sub.1] composites, PS did not improve emergence over [C.sub.0] controls. Gains from the second set of selected markers for seedling emergence (EMHB and EMLB) showed the same pattern compared with PS in su1Se1, su1se1, and sh2 composites. This provides evidence for the repeatability of gains resulting from MAS when a second and different set of marker QTL was selected.

In the sh2 composite population, both PS and MAS enhanced emergence by 12.6 and 17.8%, respectively, compared with the [C.sub.0] control. While the MAS [C.sub.1] composite (EMH) showed higher emergence, the increase was not significantly higher than that of PS (EPH). With smaller population sizes of su1Se1 and su1se1, MAS resulted in greater gains compared with PS. However, with the relatively larger population size of sh2, PS was as effective as MAS. This suggests that by using MAS with relatively small-sized populations, important alleles can be selected resulting in rapid gain, reducing the resources needed to evaluate a larger population for effective gains using PS. This is particularly true when significant associations between marker loci and trait are detected in the population. Higher heritability for emergence in the sh2 population may partially explain the observed comparable selection gains using PS and MAS in this population compared with that observed in the su1Se1 and su1se1 populations. Realized heritability for emergence was estimated as 10, 11, and 33% in the su1Se1, su1se1, and sh2 populations, respectively (Yousef, 2000).

Kernel sucrose concentrations at 20 DAP in the sucrose-selected composites are presented in Table 3. Results show that both MAS and PS were effective in significantly enhancing kernel sucrose concentration in both composite populations. However, in the high direction of selection, mean sucrose comparisons among the su1Se1 and su1se1 [C.sub.1] composites revealed that MAS was superior to PS. In the su1Se1 [C.sub.1] composites, MAS improved sucrose concentration by 24.8% (SMH) while with PS gain was 9.7% (SPH). Kernel sucrose concentration was also enhanced in the su1se1 [C.sub.1] population by 26.4 and 18.2% over [C.sub.0] using MAS and PS, respectively.
Table 3. Response to single trait selection for seedling emergence and
kernel sucrose and tenderness in the su1Se1, su1se1, and sh2 [C.sub.1]
composites after one cycle of MAS and PS in Illinois in 1997 and 1998.

                            Seedling emergence (EMG)

                           su1Se1          su1se1          sh2

            Comp.
            ([da-       EMG       %      EMG     %      EMG     %
Selection   gger])       %       gain     %     gain     %     gain

[C.sub.0]            66.1b        --    58.9b          49.3b
                       ([para-
                       graph])

[C.sub.1]    EPH     67.1b        1.5   60.7b    3.1   55.5a   12.6
  (PS)       EPL     62.0c        6.2   48.3d   18.0   41.2c   16.4

[C.sub.1]    EMHA    71.6a        8.3   67.7a   14.9   58.1a   17.8
  (MAS)      EMLA    61.0c        7.7   45.8d   22.2   35.8d   27.4

[C.sub.1]    EMHB    75.4a       14.1   67.7a   14.9   54.9a   11.4
  (MAS)      EMLB    64.3bc       2.7   53.7c    8.8   36.8d   25.4

             LSD      3.8                3.8            4.1

                 Kernel sucrose concentration (SUC)

                            su1Se1          su1se1

             Comp.
            ([double     SUC      %       SUC      %
Selection   dagger])   (mg/gm)   gain   (mg/gm)   gain

[C.sub.0]              147.1c           207.1c     --

[C.sub.1]     SPH      161.4b     9.7   244.7b    18.2
  (PS)        SPL      134.0d     8.9   177.0d    14.5

[C.sub.1]     SMH      183.6a    24.8   261.8a    26.4
  (MAS)       SML      134.4d     8.6   167.6d    19.1

[C.sub.1]     --
  (MAS)       --

              LSD        9.3             13.4

                      Kernel tenderness (TND)

                              su1Se1          su1se1

               Comp.        TND     %      TND     %
Selection   ([sections])   (kg)    gain   (kg)    gain

[C.sub.0]                  66.1b    --    66.8b    --

[C.sub.1]       TPH        65.9b    0.3   64.8b    3.0
  (PS)          TPL        74.8c   13.2   71.1c    6.4

[C.sub.1]       TMH        58.3a   11.8   57.7a   13.6
  (MAS)         TML        80.3d   21.5   71.6c    7.2

[C.sub.1]       --
  (MAS)         --

                LSD         3.6            2.7

([dagger]) Composite. EPH and EPL, [C.sub.1] composites created in the
high and low direction using PS. EMHA and EMLA, [C.sub.1] composites
created in the high and low direction using additive gene effect. EMHB
and EMLB, [C.sub.1] high and low composites using a second set of RFLP
markers.

([double dagger]) SPH and SPL, high and low PS sucrose composites. SMH
and SML, high and low MAS sucrose composites.

([sections]) TPH and TPL, high and low PS tenderness composites. TMH
and TML, high and low MAS tenderness composites.

([paragraph]) Means with different letters within the column are
significantly different at P < 5%.


Tenderness, measured as the force required to crush cooked kernels, is another important sweet corn eating quality. The less force required, the more tender and preferred the sweet corn is to the consumer (Azanza et al., 1996a). Therefore selection for tenderness was conducted using alleles associated with lower tenderness values. Composites subjected to MAS were significantly more tender (lower values) compared with those based on PS (Table 3). Phenotypic selection for tenderness was relatively ineffective in both the su1Se1 and su1se1 [C.sub.1] composites. Selection gain from MAS was 11.8 and 13.6% in su1Se1 and su1se1 in tenderness-selected composites (TMH), respectively.

Selection Gain in Multiple Traits

Selection for multiple traits is much more typical in plant-breeding programs than selection for single traits due to economic returns. In sweet corn breeding, selection for eating quality without concurrent attention to seedling emergence will be of limited commercial value. In this study, multiple-trait selection was practiced for seedling emergence and eating-quality traits to develop germplasm superior for both. Selection gains achieved with PS and MAS when incorporating more than one trait in the selection indices were not consistent across traits or composites. Negative selection gains, in which performance was in a direction opposite to expectation, were observed in a few composites.

Emergence and Tenderness

Interaction among genes in the su1Se1, su1se1, and sh2 [C.sub.1] composites, which was reported previously (Wang, 1997), and indirect selection due to a negative correlation between emergence and eating quality, appeared to affect selection gains for both PS and MAS (Table 4). In the sh2 population, with a larger population size, PS for emergence/tenderness did not enhance emergence while MAS increased emergence by 14.2% (ETMH) over the [C.sub.0]. The [C.sub.1] composite (ETMH, sh2), developed through MAS and possessing improved emergence and tenderness, represents a potential source of germplasm for breeding programs designed to simultaneously improve seed and eating quality in sweet corn. In a recent study, three of the marker loci linked to beneficial alleles that enhance emergence in sh2 population were introgressed into three distinct sweet corn genetic backgrounds (Yousef, 2000). Comparable effects in emergence as observed in the [F.sub.2:3] population was observed for these alleles in this study, suggesting their effects are not background-specific.
Table 4. Response to multiple trait selection for seedling emergence
and kernel tenderness in the su1Se1, su1se1, and sh2 [C.sub.1]
composites after one cycle of simultaneous selection for both traits
using MAS and PS in Illinois in 1997 and 1998.

                                 Seedling emergence (EMG)

                            su1Se1           su1se1          sh2

            Comp.
            ([da-       EMG        %      EMG      %     EMG      %
Selection   gger])      (%)       gain    (%)    gain    (%)    gain

[C.sub.0]            66.1b               58.9a          49.3b
                       ([double
                       dagger])

[C.sub.1]    ETPH    71.5a         8.2   61.7a    4.8   47.1b   -4.5
  (PS)       ETPL    64.3b         2.7   59.8a   -1.5   46.7b    5.3

[C.sub.1]    ETMH    71.1a         7.6   62.9a    6.8   56.3a   14.2
  (MAS)      ETML    63.2b         4.4   59.7a   -1.4   47.2b    4.3

             LSD      4.4                 4.3            4.5

                                 Kernel tenderness (TND)

                         su1Se1        su1se1            sh2

            Comp.
            ([da-     TND      %     TND     %       TND      %
Selection   gger])   (kg)    gain   (kg)    gain     (kg)    gain

[C.sub.0]            66.1a          66.8c           99.8b

[C.sub.1]    ETPH    63.2a    4.4   59.0a   11.7    95.2a     4.6
  (PS)       ETPL    76.3b   15.4   63.4b   -5.1   104.6c     4.8

[C.sub.1]    ETMH    65.6a    0.8   58.9a   11.8    92.8a     7.0
  (MAS)      ETML    76.2b   15.3   66.5c   -0.4   100.8bc    1.0

             LSD      2.2            2.3             4.5

([dagger]) ETPH and ETPL, PS composites created based on selection for
high and low index scores of both emergence and tenderness. ETMH and
ETML, MAS composites created based on selection for high and low index
scores of both emergence and tenderness.

([double dagger]) Means with different letters within the column are
significantly different at P < 5%.


Emergence and Hedonic Rating

Significant enhancement in both seedling emergence and hedonic value was observed with MAS over PS in the high direction in su1se1 population (EDMH) (Table 5). This increase was not significant compared with [C.sub.0] controls. No selection gain in the positive direction was observed for both traits in the su1Se1 and su1se1 populations with MAS and PS except for emergence in MAS. Including the hedonic trait in the selection index made it difficult to achieve gain comparable with that of the other selected traits because the complex nature of this trait led to larger experimental error. Hedonic rating is a difficult trait to evaluate since it is controlled by attributes that human subjects perceive and prefer differentially (Azanza et al., 1994).
Table 5. Response to multiple trait selection for seedling emergence
and hedonic in the su1Se1 and su1se1 [C.sub.1] composites after one
cycle of simultaneous selection for both traits using MAS and PS
in Illinois in 1997 and 1998.

                                     Seedling emergence (EMG)

                                      su1Se1           su1se1

                    Comp.        EMG         %      EMG      %
Selection         ([dagger])     (%)        gain    (%)     gain

[C.sub.0]                      66.1b               58.9b
                                 ([double
                                 dagger])

[C.sub.1] (PS)    EDPH         62.9b        -4.8   56.6b    -3.9
                  EDPL         63.5b         3.9   62.9a    -6.8

[C.sub.1] (MAS)   EDMH         65.6b        -0.8   64.2a     9.0
                  EDML         70.4a        -6.5   61.0ab   -3.6

                  LSD           3.6                 3.8

                                  Hedonic (HED, 1-9 score)

                                  su1Se1         su1se1

                    Comp.               %               %
Selection         ([dagger])    HED    gain    HED     gain

[C.sub.0]                      5.33a          6.38ab

[C.sub.1] (PS)    EDPH         5.28a   -0.9   6.08bc   -4.7
                  EDPL         5.14a    3.6   5.83c     8.6

[C.sub.1] (MAS)   EDMH         5.33a    0.0   6.78a     6.3
                  EDML         4.95a    7.1   5.90c     7.5

                  LSD          0.41           0.42

([dagger]) EDPH and EDPL, PS composites created based on selection for
high and low index scores of both emergence and hedonic rating. EDMH
and EDML, MAS composites created based on selection for high and low
index scores of both emergence and hedonic rating.

([double dagger]) Means with different letters within the column are
significantly different at P < 5%.


Emergence, Sucrose, and Tenderness

In the su1se1 population, MAS enhanced emergence significantly over PS (ESTPH vs. ESTMH) (Table 6). Selection gain from PS was 9.8%; with MAS it was 15.4% over [C.sub.0]. In the su1Se1 population, PS resulted in higher gain in emergence but this was associated with a reduction in kernel sucrose concentration compared with MAS. Due to reduction in kernel sucrose concentration in the low direction composites, emergence was indirectly enhanced. This could explain the higher mean emergence in the low-direction composites with both PS and MAS compared with [C.sub.0]. For sucrose, PS showed an increase of only 7.3 and 3% while MAS resulted in increases of 12.8 and 19.9% in su1Se1 and su1se1 composites in the high direction, respectively. Incorporating three traits in the PS selection index has led to reduced gain among traits, as was proposed by Jones et al. (1993). However, simultaneous improvement in emergence and sucrose, but not tenderness, was observed when MAS was applied to the su1se1 population in the high direction ([C.sub.1] ESTMH composite). Although there is a negative correlation between sucrose and emergence, some marker QTL alleles such as the one linked to umc50 were found to be associated with high emergence as well as sucrose (Azanza et al., 1996b). Other markers are associated with both high sucrose and tenderness, or emergence and tenderness, such as npi268, npi240, and umc114. The su1se1 [C.sub.1] composite (ESTMH) provides germplasm that is of particular use in breeding programs to develop sweet corn with improved emergence and eating quality.
Table 6. Response to multiple trait selection for seedling emergence,
kernel sucrose, and kernel tenderness in the su1Se1 and su1se1
[C.sub.1] composites after one cycle of simultaneous selection for all
three traits using MAS and PS in Illinois in 1997 and 1998.

                                     Seedling emergence (EMG)

                                      su1Se1           su1se1

                    Comp.         EMG        %      EMG      %
Selection         ([dagger])      (%)       gain    (%)     gain

[C.sub.0]                      66.1b               58.9d
                                 ([double
                                 dagger])

[C.sub.1] (PS)    ESTPH        72.0a         8.9   64.7b     9.8
                  ESTPL        69.9ab       -5.7   60.7cd   -3.1

[C.sub.1] (MAS)   ESTMH        68.9ab        4.2   68.0a    15.4
                  ESTML        67.1b        -1.5   62.8bc   -6.6

                  LSD           4.4                 3.0

                               Kernel sucrose concentration (SUC)

                                    su1Se1          su1se1

                    Comp.        SUC      %       SUC      %
Selection         ([dagger])   (mg/gm)   gain   (mg/gm)   gain

[C.sub.0]                      147.1c           207.1b

[C.sub.1] (PS)    ESTPH        157.8b     7.3   213.4b     3.0
                  ESTPL        135.9d     7.6   185.7c    10.3

[C.sub.1] (MAS)   ESTMH        165.9a    12.8   248.3a    19.9
                  ESTML        133.2d     9.4   179.3c    13.4

                  LSD            8.0             11.2

                                   Kernel tenderness (TND)

                                   su1Se1          su1se1

                    Comp.       TND      %      TND      %
Selection         ([dagger])    (kg)    gain    (kg)    gain

[C.sub.0]                      66.1ab          66.8ab

[C.sub.1] (PS)    ESTPH        67.0ab   -1.4   64.3a     3.7
                  ESTPL        67.3b     1.8   66.0a    -1.2

[C.sub.1] (MAS)   ESTMH        65.0a     1.7   64.8a     3.0
                  ESTML        71.1c     7.6   69.3b     3.7

                  LSD           2.2             2.6

([dagger]) ESTPH and ESTPL, PS composites created based on selection
for high and low index scores of emergence, sucrose, and tenderness.
ESTMH and ESTML, MAS composites created based on selection for high
and low index scores of emergence, sucrose, and tenderness.

([double dagger]) Means with different letters within the column are
significantly different at P < 5%.


Asymmetrical Response to Selection

When we practiced selection in the high and low directions, the data displayed asymmetrical selection response where the low direction composites showed different significant gains compared with selection in the high direction for both PS and MAS (Table 7). The MAS-low-direction composites contained some QTL alleles associated with high trait performance since it was impossible to find families without all of the five selected marker loci. The presence of these marker alleles in the low-direction MAS composites may have been partially responsible for the comparable gains observed between MAS and PS in the low direction. Families with missing marker information are another factor that might affect MAS efficiency in either direction. Wang (1997) reported that missing marker information hinders the efficiency of MAS in breeding programs. In addition, unintentional selection in the creation of [C.sub.1] composites from some of the selected families with very poor vigor could interfere with the random mating scheme. For example, selecting families for low emergence would result in indirect selection for high kernel sucrose concentration since a negative correlation has been observed between these two traits in sweet corn (Azanza et al., 1996a). Selection for high sucrose would result in reduced seedling vigor and the possibility of these families being unequally represented in the [C.sub.1] composite seed lots. This could then result in lower means in the high-sucrose composites.
Table 7. Summary of the significant paired comparisons made between MAS
and PS and mean selection gains across the three [C.sub.1] composite
populations.

                                   MAS >
                                 PS([double   PS >     MAS >
                   n([dagger])    dagger])    MAS    [C.sub.0]

Single-trait
    selections
  High                 10           80%         0%     100%
  Low                  10           30%        10%      90%
Multiple-trait
    selections
  High                 16           31%         0%      50%
  Low                  16           19%         6%      31%
Overall
    selections
  High                 26           54%         0%      69%
  Low                  26           23%         8%      58%

Overall average:       52           38%         4%      63%

                               Selection gains
                                ([sections])

                     PS >
                   [C.sub.0]    MAS     PS

Single-trait
    selections
  High                43%      17.3%    7.1%
  Low                100%      14.7%   11.7%
Multiple-trait
    selections
  High                38%       7.0%    3.4%
  Low                 43%       3.6%    2.4%
Overall
    selections
  High                39%      12.0%    5.2%
  Low                 61%       9.1%    7.0%

Overall average:      50%      10.9%    6.1%

([dagger]) Number of the paired comparisons made between MAS and PS and
[C.sub.0] (except in PS > [C.sub.0] for single traits only was 7 paired
comparisons).

([double dagger]) The > refers to significantly higher for the selected
trait(s) at P < 5%.

([sections]) The number of [C.sub.1] composites used to calculate the
mean gains was the same as the number of paired comparisons shown in
the n column.


In previous work with the base populations, certain QTL exerted large effects on the selected traits in this study (Han, 1994; Tadmor et al., 1995). Selection for these major QTL could also explain some of the observed asymmetrical responses (Falconer and MacKay, 1996). The presence of interaction among alleles at different loci (epistasis) could modify the selection responses in the high or low directions. Significant and substantial epistasis was reported in the original populations used in his study (Wang, 1997). Therefore, when combining [F.sub.2:3] families to constitute [C.sub.1] composites, interaction among the QTL could cause the [C.sub.1] means to be higher or lower than anticipated. In addition, natural selection may aid artificial selection in one direction or hinder it in the other. The fertility or plant fitness may change so that a higher intensity of selection is achieved in one direction than in the other. Variation in selection differential influences the response per generation and the agreement between observed and predicted responses (Falconer and Mackay, 1996). The phenotypic distribution of the base population resulted in different gains from PS in the high vs. low direction. The distribution was skewed to the low direction in emergence (S = -0.19) and tenderness (S = -0.54), and skewed in the high direction in kernel sucrose concentrations (S = 0.68) with unequal selection deferential among the high and low composites.

Relative Efficiency and Cost of MAS and PS

A total of 52 paired comparisons were made between MAS and PS [C.sub.1] composites in enhancing economic traits in sweet corn (Table 7). In 38% of the paired comparisons, MAS resulted in significantly higher gains than PS across the three [C.sub.1] composite populations, while PS was significantly greater than MAS in only 4% of the comparisons. Cycle 1 MAS and PS were significantly greater than [C.sub.0] performance in 63 and 50% of the paired comparisons, respectively. MAS showed an average gain nearly twice of that of PS across all three populations. The average selection gain across the composite populations and selected traits, calculated as percent increase or decrease from the [C.sub.0], was 10.9 vs. 6.1% with MAS and PS, respectively.

The estimated cost required for phenotypic evaluation differed among the selected traits. For seedling emergence, the cost in [C.sub.1] was $56/family since only the initial field costs and germination counts were required. With eating quality traits (sucrose and tenderness), the [C.sub.1] costs were higher since they required laboratory analysis in addition to field costs. The cost to evaluate one family for sucrose or tenderness were $178/family or $158/family, respectively. The cost of hedonic evaluation was much greater ($370/family), requiring a much larger input of labor and time for taste panel preference evaluation. In contrast, with MAS the cost/family in [C.sub.1] ranged from $103 for emergence to $260 for hedonic rating. The estimated costs using MAS and PS for the selected traits in the second ([C.sub.2]) and third ([C.sub.3]) cycles are presented in Table 8. Phenotypic selection appeared more costly than MAS for quality traits. These costs will vary depending on the selected trait, number of evaluated traits, population size, labor cost, number of environments and replications, selection intensity, number of polymorphic markers detected, and type of DNA markers used in the breeding program. Marker systems such as RFLP are considered expensive and laborious compared with PCR-based markers. Advances in DNA technology have reduced costs associated with the application of molecular markers in breeding programs (Gu et al., 1995). It appears that making an accurate comparison between MAS and PS costs is very difficult and may not be applicable to every breeding scheme. These estimates reflect costs incurred in our program and are presented to provide a case study for comparing costs associated with PS and MAS for quantitative traits with different phenotypic evaluation costs.
Table 8. Estimates of the average evaluation costs (US$) for the
selected traits using PS and MAS([dagger]).

                 [C.sub.1]   [C.sub.2]   [C.sub.3]
                 ([double     ([sec-      ([sec-
                 dagger])     tions])     tions])

Trait            PS    MAS   PS    MAS   PS    MAS

Emergence         56   103    42    78    37    70
Sucrose          178   164   134   109   119    90
Tenderness       158   154   119   104   105    87
Hedonic rating   370   260   278   157   247   122

([dagger]) Average costs of selecting and evaluating one family for
each trait in the first cycle ([C.sub.1]) and in subsequent cycles
([C.sub.2] to [C.sub.3]).

([double dagger]) Estimated costs based on actual expenses.

([sections]) Projected costs based on costs associated with PS and MAS
in the first cycle of selection.


Selection gains over cycles of selections will probably decrease from one cycle to the next using MAS or PS. Simulation results on annual crop species such as corn reported by Edwards and Page (1994) indicated that MAS provides rapid responses compared with phenotypic recurrent selection (PRS) early in the selection process; however, the differences diminished greatly within three to five cycles. They suggested that MAS could offer a primary advantage of enabling two selection cycles/year vs. the 2 yr/cycle required by most PRS for the evaluation of testcross progeny. Marker-assisted selection would be of more advantage in the first 2 to 3 yr of PRS, after which time conventional methods might replace MAS to achieve further responses.

Selection is a process that increases favorable allelic frequencies. In our study, the first cycle of selection for single traits increased favorable allelic frequencies from about 50% in the suSe1, su1se1, and sh2 base populations to 73, 75, and 70% vs. 62, 66, and 56% in MAS and PS [C.sub.1] composites, respectively. The favorable allelic frequencies across all traits in the su1Se1, su1se1, and sh2 [C.sub.1] populations were increased by 45, 48, and 32% and by 24, 31, and 6% using MAS and PS, respectively. Change in allelic frequencies from [C.sub.0] to [C.sub.1] was highly correlated with selection gains (r = 0.70, P [is less than] 0.01). After one cycle of MAS, none of the beneficial alleles were fixed in any of the composites. The increase in beneficial allelic frequencies should result in most of the desirable alleles being fixed after two cycles of MAS with complete fixation by [C.sub.3]. The change in allelic frequencies with PS was lower than those observed in MAS.

Comparing average selection gains of MAS (10.9%) with that of PS (6.1%) and costs associated with MAS and PS, we conclude that once beneficial alleles are identified, MAS can provide higher economic return in breeding programs. This advantage includes seedling emergence with low phenotypic evaluation costs ($56/ family) compared with RFLP genotyping costs ($103/ family) since the selection gain from MAS was approximately twice that of PS. However in the long run, MAS would be more advantageous, not only based on costs but also in accelerating the progress of breeding programs. Results generated from this study suggest that with traits requiring laboratory analysis and labor, MAS can provide higher economic returns in a shorter time while decreasing the probability of losing the desirable alleles. This agrees with the theoretical expectations of many authors (Dudley, 1993; Stuber, 1995; Knapp, 1998) and computer simulations (Zhang and Smith, 1992; Edwards and Page, 1994).

CONCLUSIONS

The data obtained in the high direction, which is the aim of breeding programs, from this study of MAS is encouraging. Significant gains in the MAS [C.sub.1] composites compared with PS [C.sub.1] were obtained from selection based on DNA markers linked to QTL. These results are of particular importance since the population sizes used were relatively small, indicating that one may not need a large population size to achieve progress from MAS. In most cases, MAS provided simultaneous improvements for multiple traits. Many of these traits require laboratory evaluation and are difficult and expensive to characterize. Although MAS is relatively expensive, so is phenotypic-based selection. Marker-assisted selection can economically compete with PS, particularly with the advances in DNA technology and the gains resulting from the reduced size and duration of breeding programs. Therefore, incorporating DNA markers into traditional breeding programs can reduce the time and money needed to achieve breeding goals.

ACKNOWLEDGMENTS

The authors acknowledge funding from grant numbers US-1709-89 and US-2242-92C from the US-Israel Binational Agricultural Research and Development, Minia University of Egypt, and the Egyptian Cultural and Educational Bureau in Washington, D.C. for support to the senior author during his study in the USA. Thanks to B. Klein, Dept. of Food Sciences, Univ. of Illinois for her guidance in the taste panel evaluation. The authors also thank William Tracy for seedling emergence evaluation. This study is a portion of a Ph.D. dissertation by the senior author at the University of Illinois, Urbana-Champaign.

REFERENCES

Azanza, F. 1994. Genetics, physiology and sensory perception of sweet corn eating quality. Ph.D. diss. (Diss. Abstr. 94-03134). University of Illinois, Urbana-Champaign, IL.

Azanza, F., A. Barzur, and J.A. Juvik. 1996a. Variation in sweet corn kernel characteristics associated with stand establishment and eating quality. Euphytica 87:7-18.

Azanza, F., J.A. Juvik, and B.P. Klein. 1994. Relationships between sensory quality attributes and kernel chemical composition of fresh-frozen sweet corn. J. Food Qual. 17:159-172.

Azanza, F., Y. Tadmor, B.P. Klien, T.R. Rocheford, and J.A. Juvik. 1996b. Quantitative trait loci influencing chemical and sensory characteristics of eating quality in sweet corn. Genome 39:40-50.

Bernardo, R. 1999. Selection response with marker-based assortative mating. Crop Sci. 39:69-73.

Breto, M.P., M.J. Asins, and E.A. Carbonell. 1994. Salt tolerance in Lycopersicon species: III. Detection of quantitative trait loci by means of molecular markers. Theor. Appl. Genet. 88:395-401.

Danesh, D., S. Aaron, G.E. McGilland, and N.D. Young. 1994. Genetic dissection of oligogenic resistance to bacterial wilt in tomato. Mol. Plant-Micro. Interaction 4:464-471.

Douglass, S.K., J.A. Juvik, and W.E. Splittstoesser. 1993. Sweet corn seedling emergence and variation in kernel carbohydrate reserves. Seed Sci. Tech. 21:433-445.

Dudley, J.W. 1993. Molecular markers in plant improvement: Manipulation of genes affecting quantitative traits. Crop Sci. 33:660-668.

Eathington, S.R., J.W. Dudley, and G.K. Rufener II. 1997. Usefulness of marker-QTL association in early generation selection. Crop Sci. 37:1686-1693.

Edwards, M.D., and N.J. Page. 1994. Evaluation of marker-assisted selection through computer simulation. Theor. Appl. Genet. 88: 376-382.

Falconer, D.S., and T.F.C. Mackay. 1996. Introduction to quantitative genetics. 4th ed. Longman, Essex, England.

Gimelfarb, A., and R. Lande. 1994. Simulation of marker assisted selection in hybrid populations. Genet. Res. (Cambridge) 63:39-47.

Gu, W.K., N.F. Weeden, J. Yu, and D.H. Wallace. 1995. Large-scale, cost-effective screening of PCR products in marker-assisted selection applications. Theor. Appl. Genet. 91:465-470.

Han, T. 1994. Investigations into the physiology and genetics of sweet corn seedling emergence and vigor. M.S. thesis. University of Illinois, Urbana-Champaign, IL.

Jones, M.W., R.C. Pratt, W.R. Frindly Jr., S.K. St. Martin, and W.D. Guthrie. 1993. Multiple trait improvement in ohs3 using a "rank and replace" S1 recurrent selection method. Maydica 38:283-290.

Juvik J.A., and D.R. LaBonte. 1988. Single kernel analysis for the presence of the sugary enhancer: se1 gene in sweet corn. Hort-Science 23:384-386.

Knapp, S.J. 1998. Marker-assisted selection as a strategy for increasing the probability of selecting superior genotypes. Crop Sci. 38: 1164-1174.

Lande, R., and R. Thompson. 1990. Efficiency of marker-assisted selection in the improvement of quantitative traits. Genetics 124: 743-756.

Lee, M. 1995. DNA markers and plant breeding programs. Adv. Agron. 55:265-344.

Moreau, L., A. Charcosset, F. Hospital, and A. Gallais. 1998. Marker-assisted selection efficiency in population of finite size. Genetics 148:1353-1365.

Nelson, C. 1996. Q-gene v2.20. Cornell University, Ithaca, NY.

SAS Institute. 1991. SAS user's guide: Statistics. Version 6, 1st ed. SAS Inst., Cary, NC.

Soller, M., and J.S. Beckmann. 1983. Genetic polymorphism in varietal identification and genetic improvement. Theor. Appl. Genet. 67: 25-33.

Stromberg, L.D., J.W. Dudley, and G.K. Rufener. 1994. Comparing conventional early generation selection with molecular marker assisted selection in maize. Crop Sci. 34:1221-1225.

Stuber, C.W. 1995. Mapping and manipulating quantitative traits in maize. Trends Gen. (Cambridge) 11:12 477-481.

Stuber, C.W., S.E. Lincoln, D.W. Wolf, T. Helentjaris, and E.S. Lander. 1992. Identification of genetic factors contributing to heterosis in a hybrid from two elite maize inbred lines using molecular markers. Genetics 132:823-839.

Tadmor, Y., F. Azanza, T. Han, T.R. Rocheford, and J.A. Juvik. 1995. RFLP mapping of sugary enhancer 1 gene in maize. Theor. Appl. Genet. 91:489-494.

Van Berloo, R., and P. Stam. 1999. Comparison between marker-assisted selection and phenotypical selection in a set of Arabidopsis thaliana recombinant inbred lines. Theor. Appl. Genet. 98:113-118.

Wang, S.-S. 1997. Epistasis and other factors influencing the detection of marker-QTL associations. Ph.D. thesis (Diss. Abstr. 98-12802). University of Illinois, Urbana-Champaign, IL.

Xiao, J., J. Li, L. Yuan, and S.D. Tanksley. 1996. Identification of QTLs affecting traits of agronomic importance in a recombinant inbred population derived from a subspecific rice cross. Theor. Appl. Genet. 92:230-244.

Yousef, G.G. 2000. Empirical evaluation marker-assisted selection for improving quantitative traits in sweet corn. Ph.D. Thesis (Accession # 45742859) University of Illinois, Urbana-Champaign, IL.

Zhang, W., and C. Smith. 1992. Computer simulation of marker-assisted selection utilizing linkage disequilibrium. Theor. Appl. Genet. 83:813-820.

Zhu, H., G. Briceno, R. Dovel, P.M. Hayes, B.H. Liu, and S.E. Ullrich. 1999. Molecular breeding for grain yield in barley: An evaluation of QTL effects in a spring barley cross. Theor. Appl. Gent. 98:772-779.

Abbreviations: DAP, days after pollination; [C.sub.0], base population; [C.sub.1], composites created by intermating the selected [F.sub.2:3] families using PS or MAS; MAS, marker-assisted selection; PRS, phenotypic recurrent selection; PS, phenotypic selection; QTL, quantitative trait locus/loci; su1, se1, and sh2 are sugary1, sugary enhancer 1, and shrunken2 endosperm mutations in maize (Zea mays L.); RFLP, restriction fragment length polymorphism.

Gad G. Yousef and John a. Juvik(*)

Dep. of Natural Resources and Environmental Sciences, Univ. of Illinois at Urbana-Champaign, 307 ERML, 1201 W. Gregory Dr., Urbana, IL 61801. Received 28 Jan. 2000. (*) Corresponding author (j-juvik@ uiuc.edu).
COPYRIGHT 2001 Crop Science Society of America
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2001 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Yousef, Gad G.; Juvik, John A.
Publication:Crop Science
Geographic Code:1USA
Date:May 1, 2001
Words:9979
Previous Article:Discovery and Deployment of Molecular Markers Linked to Fusarium Head Blight Resistance: An Integrated System for Wheat and Barley.
Next Article:Two Types of GGE Biplots for Analyzing Multi-Environment Trial Data.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters