Printer Friendly

Phenotypic Recurrent Selection Methodology for Reducing Fiber Concentration in Smooth Bromegrass.

PHENOTYPIC RECURRENT SELECTION is a powerful selection tool, designed to rapidly increase the frequency of favorable alleles while maintaining genetic variation and minimizing inbreeding within populations. It is highly robust, utilizing all additive genetic variance within a population and providing observable progress even for traits with low heritability (Hallauer and Miranda, 1991, p. 169). These advantages, combined with its simplicity and ease of application, have made it the most common form of recurrent selection in forage crops. It has been widely and successfully used for improving disease resistance (Casler and Pedersen, 1996), forage nutritive value (Casler and Vogel, 1999), forage yield (Burton, 1974), and many other traits (Casler et al., 1996).

Because phenotypic recurrent selection is often applied to traits with inherently low heritability, its effectiveness, or lack thereof, can be an important issue. For example, replication of the selection unit may be necessary to increase heritability of some traits to an acceptable level (Aung et al., 1994). Even more important in monoecious species is the definition of the selection unit. The simplest form of phenotypic selection involves selection of the best plants, followed by open-pollination among all plants, including those deemed undesirable. Typically called mass selection, this method controls female gametes, but not male gametes, during the selection process. Modifying this method to eliminate pollen from undesirable plants, i.e., the unselected females, theoretically doubles the genetic gain possible with mass selection (Falconer and Mackay, 1996, p. 191; Hallauer and Miranda, 1991, p. 169). Unfortunately, this cannot be accomplished in many forage crops without lengthening cycle time, thereby reducing genetic gain per year. Most field-based phenotypic recurrent selection schemes in forage crops require 2 yr per cycle, the first for establishment of a spaced-plant selection nursery and the second for selection and recombination. Typically, to accomplish selection controlling both male and female gametes, forage breeders transplant selected plants into a polycross block, requiring a third year for recombination.

Two alternatives that would shorten cycle time have been proposed. First, Burton (1974) described a set of restrictions to phenotypic selection that he termed Recurrent Restricted Phenotypic Selection (RRPS). With some additional restrictions, Burton (1982) was able to achieve a 1-yr recurrent selection cycle time out of Pensacola bahiagrass (Paspalum notatum var. suare Parodi) by collecting data, making selections, and producing seed on selected plants during the establishment year. Furthermore, he achieved selection among both female and male gametes by excising reproductive culms, placing them in water-filled bottles in a laboratory, and shaking them daily during anthesis. Eight cycles of RRPS for forage yield resulted in an average gain of 16% [cycle.sup.-1] (Burton, 1982).

The second alternative is based on the ability to accomplish selection prior to anthesis, so that non-selected plants can be eliminated prior to pollination. This modification, based on seedling evaluations, is common for many disease resistance selection programs (Casler et al., 1996; Casler and Pederson, 1996). However, many selection programs require field evaluation of established plants, partly to ensure maintenance of selection pressure for adaptation-fitness traits or, minimally, to avoid relaxation of selection pressure for adaptation-fitness traits.

Smooth bromegrass is most commonly managed as a hay crop in which more than half of the annual dry matter is produced prior to the first harvest, typically made shortly after head emergence (Casler and Carlson, 1995). Reich and Casler (1985) proposed a more efficient selection for which plants are harvested; tissue samples are dried, ground, and analyzed; and selections are made prior to anthesis. Non-selected plants are mowed and selected plants are allowed to open-pollinate in situ. They predicted a 34% increase in selection progress if plants are harvested at an early vegetative growth stage, compared with the heading growth stage, assuming that all data can be collected and selections identified prior to anthesis (Reich and Casler, 1985).

For most cool-season forage crops, including smooth bromegrass, growth rates are greatest in the spring, during reproduction. For this reason, there is typically only about 20 to 25 d in which to complete the harvest, analysis, and selection stages of Reich and Casler's proposal. The use of traditional wet-laboratory techniques would severely reduce the number of plants that could be evaluated, reducing either effective population size or selection response. Near-infrared reflectance spectroscopy (NIRS) significantly reduces the time required to conduct forage nutritional value analyses, allowing a larger population to be analyzed in a limited time period (Reich and Casler, 1985). Therefore, for this method effectively to improve genetic gains, the selection criterion must have both a high genetic correlation between NIRS and wet-laboratory methods and a high genetic correlation between the selection and target growth stages. For smooth bromegrass, phenotypic correlations between NIRS and wet-laboratory data range from 0.85 to 0.95 (Casler et al., 1983-1997, unpublished data), while genetic correlations between growth stages are generally greater than 0.8 for forage nutritive value traits (Ehlke et al., 1986; Reich and Casler, 1985).

The concentration of neutral detergent fiber (NDF) may be the best feasible chemical selection criterion for genetically improving intake potential in forage crops (Casler and Vogel, 1999). It is rapid, amenable to prediction by NIRS, has high laboratory repeatability, and has been shown to be heritable in reed canarygrass, Phalaris arundinacea L. (Surprenant et al., 1988). The objective of this study was to measure realized gains for six methods of phenotypic recurrent selection for reduced NDF concentration in smooth bromegrass. Among the six selection methods, the comparisons of interest were as follows: selection at a vegetative vs. heading growth stage (for both open-pollination and polycrossing of selected plants), wet-laboratory vs. NIRS evaluation of selection units (for both open-pollination and polycrossing of selected plants), and open-pollination vs. polycrossing of selected plants.

MATERIALS AND METHODS

Phenotypic selection for reduced NDF concentration was practiced in the WB-[RP.sub.1] smooth bromegrass germplasm pool (Casler, 1992). Plants were raised as seedlings in the greenhouse and transplanted to a Plano silt loam (fine-silty, mixed, mesic, Typic Argiudoll) near Arlington, WI. Ten-week-old seedlings were transplanted to the field in mid-May and mowed twice during the establishment year. Weeds were controlled with herbicide (Casler, 1992). All plants were fertilized with 112 kg N [ha.sup.-1] in early spring of the year following establishment.

All selection nurseries were established with approximately 400 plants. Selection for reduced NDF concentration was practiced on 340 to 350 plants, allowing a small proportion of plants to be eliminated based on vigor or disease problems, if necessary. Nurseries were arranged into 10 blocks of 40, in a 5 by 8 arrangement. Thirty-five plants were harvested from each block and always processed together throughout the drying, grinding, and laboratory processes (Casler, 1992). Cutting height was always 5 cm. Because of the wide spacing, border and neighbor effects were assumed to be negligible.

Samples were dried at 60[degrees]C and ground through a 1-mm screen of a Wiley-type mill and then through a 1-mm screen of a cyclone mill. The concentration of NDF was determined on 0.5-g samples by the procedure of Goering and Van Soest (1970), with the exception that sodium sulfite and decahydronaphthalene were excluded from the reflux solution. All NIRS predictions of NDF concentration, during the course of selection, were made with a Pacific Scientific scanning monochromator, model 51A (Silver Spring, MD). Calibration and validation statistics are presented in Table 1. The selection criterion was a Student's t-value, computed as

[sup.t]NDF = ([X.sub.ij] - [M.sub.j])/[S.sub.j],

where [X.sub.ij] = NDF concentration of the ith plant of the jth block, [M.sub.j] is the mean of the jth block, and [s.sub.j] is the standard deviation of the jth block. Blocking enabled removal of a large proportion of field and laboratory variation in the computation of the selection criterion (Casler, 1992). Thirty-five selections were made from each nursery, giving selection intensities ranging from 10 to 10.3%.
Table 1. Calibration and validation statistics for NIRS
(near-infrared reflectance spectroscopy) prediction of NDF (neutral
detergent fiber) concentration in eight selection nurseries and
2 yr of the evaluation nursery.

Nursery               Year    Mean    SD([double dagger])
                           -- g [kg.sup.-1]

Selection nurseries
  vNB2-cycle 1        1986     431         25.6
  vNB2-cycle 2        1988     447         26.3
  vNB2-cycle 3        1990     443         23.5
  vNP3-cycle 2        1989     430         25.3
  hNU2-cycle 1        1986     585         27.1
  hNU2-cycle 2        1988     642         21.0
  hNU2-cycle 3        1990     651         19.8
  hNP3-cycle 2        1989     676         27.3

Evaluation nursery
  Vegetative          1995     395         27.4
  Vegetative          1996     399         33.6
  Heading             1995     552         41.5
  Heading             1996     630         33.4

                               Calibration([dagger])
Nursery               SEC([double dagger])   [R.sup.2]
                                -- g [kg.sup.-1]

Selection nurseries
  vNB2-cycle 1               13.4               0.73
  vNB2-cycle 2               11.2               0.82
  vNB2-cycle 3                9.5               0.84
  vNP3-cycle 2               13.4               0.72
  hNU2-cycle 1               13.4               0.75
  hNU2-cycle 2               11.4               0.80
  hNU2-cycle 3               10.8               0.90
  hNP3-cycle 2               14.1               0.73

Evaluation nursery
  Vegetative                 14.2               0.83
  Vegetative                 16.1               0.77
  Heading                    16.9               0.83
  Heading                    18.8               0.78

                                Validation([dagger])
Nursery                 SEP([doubled dagger])  [r.sup.2]
                                 --  g [kg.sup.1] --
Selection nurseries
  vNB2-cycle 1               14.1                0.84
  vNB2-cycle 2               11.9                0.81
  vNB2-cycle 3               10.4                0.79
  vNP3-cycle 2               14.1                0.75
  hNU2-cycle 1               15.4                0.86
  hNU2-cycle 2               13.1                0.83
  hNU2-cycle 3               11.7                0.87
  hNP3-cycle 2               15.0                0.70

Evaluation nursery
  Vegetative                 17.5                0.80
  Vegetative                 18.3                0.81
  Heading                    19.0                0.79
  Heading                    19.8                0.77


([dagger]) Calibration samples: n = 50; validation samples: n = 20.

([double dagger]) SD = standard deviation, SEC = standard error of calibration, SEP = standard error of prediction.

Selection Methods

Notation for the names of the nurseries and selection methods is based on a 4-character code: ijkl, where i = growth stage (v = vegetative or h = heading), j = laboratory method (N = NIRS or W = wet-lab), k = gametic selection pressure/ pollination method (B = biparental, U = uniparental, or P = replicated polycross), and 1 = number of years per cycle (Fig. 1). All nurseries and crossing blocks were established at least 100 m apart from each other and from other sources of smooth bromegrass to minimize the potential for pollen contamination.

[Figure 1 ILLUSTRATION OMITTED]

In Situ Pollination: vNB2, hNU2, and hWU2. Three nurseries of the WB-R[P.sub.1] base population were established in 1985 to begin the experiment (Fig. 1). In the vNB2 nursery, plants were harvested on 10 May 1986 when most plants were approximately 20 to 25 cm tall and consisted exclusively of leaves [Stage 21 or 22 on the Simon and Park (1983) maturity scale]. Sample size was approximately 10 g DM. A stratified random sample of 70 plants (seven plants per block) was analyzed for NDF concentration in duplicate and used to develop a calibration equation for the NIRS equipment and subsequent prediction of NDF concentration for all 350 tissue samples. Validation statistics for each equation were computed on a random sample of 20 plants and were used to assist in selection of the calibration equation (Table 1). Values of [t.sub.NDF] were computed for each plant. The 35 plants with the lowest values of [t.sub.NDF], irrespective of block, were tagged and all remaining plants were mowed prior to anthesis. Seed was harvested in July 1986, threshed, and cleaned independently from each plant, then bulked in equal quantities by mass. This was the selection method proposed by Reich and Casler (1985).

The hNU2 and hWU2 nurseries were harvested just after all plants had reached the fully headed growth stage (Stage 59 of Simon and Park, 1983). Because this population did not possess genetic variation for relative maturity, this occurred on the same day for all plants within a nursery (Casler, 1992). Samples consisted of a radial cross-section of 20 to 30 tillers. Samples from the hNU2 nursery were analyzed by NIRS as described for the vNB2 nursery (Table 1). Samples from the hWU2 nursery were analyzed for NDF by wet-laboratory methods. All plants in both nurseries were allowed to open-pollinate and set seed. Seed was harvested in July 1986, threshed, and cleaned for each plant. When the selected plants were identified following the laboratory analyses in winter 1986-1987, their seed was bulked in equal quantities by mass and seed from all other plants was discarded.

Seed of each Cycle-1 population was used to establish seedlings in the greenhouse for transplanting to the field in May 1987. Cycle 2 of the three selection methods was conducted in 1987 and 1988, and Cycle 3 in 1989 and 1990. Remnant seed of each balanced bulk seed population was kept frozen to maintain maximum viability.

Polycrossing: vNP3, hNP3, and hWP3. In late April 1987, prior to the initiation of new growth, three clonal ramets of the 35 selected plants from each of the three 1985 nurseries were removed and transplanted to a replicated crossing block in a randomized complete block design with three replicates (Fig. 1). Plants were fertilized with 50 kg N [ha.sup.-1] and watered to ensure establishment. Seed was harvested in July 1987, threshed, and cleaned for each clonal ramet and bulked in equal quantities by mass. Thus, in Cycle 1, the three polycrossing methods shared the same sample of the source population and the same individual selections as the three in situ pollination methods. Cycle 2 of the polycrossing methods was conducted in 1988-1990, independently of the in situ pollination methods.

Evaluation of Selection Progress

Twenty seedlings from each of 16 populations (CO = WBR[P.sub.1]; C1, C2, and C3 of vNB2, hNU2, and hWU2; and C1 and C2 of vNP3, hNP3, and hWP3) were transplanted to a holding nursery at Arlington, WI, in May 1993. They were fertilized and weeded as described above. In May 1994, two clonal ramets of each plant, approximately 100 [cm.sup.2] of crown area, were transplanted to a replicated field evaluation of selection progress. The experimental design was a randomized complete block with two replicates. Plots consisted of a row of 20 plants per population, with the two clonal ramets allocated to different replicates. Plants were watered, fertilized, and weeded to aid their establishment.

Plants were fertilized with 112 kg N [ha.sup.-1] in early April 1995 and 1996. In May of both years, approximately half of each plant was harvested at a vegetative growth stage, similar to that used in selection methods vNB2 and vNP3. The remaining half of the plant was allowed to continue growing until harvested at the fully headed growth stage.

Plant samples were dried and ground as previously described. All samples were analyzed for NDF concentration by two methods: wet-laboratory and NIRS. Both wet-laboratory and NIRS methods were identical to those used during the selection processes described above. The only deviation was that a stratified random sample of 120 plant samples (19% of the total number) was used to calibrate the NIRS equation. Calibrations were made separately for each harvest date in each year (Table 1).

All response variables (NDF measured as: vegetative/wetlab, vegetative/NIRS, heading/wet-lab, and heading/NIRS) were analyzed by mixed-models analysis of variance, assuming blocks, years, and their interactions to be random effects and populations (cycles and methods) to be fixed effects. The main effect of years was treated as a repeated measures factor in a univariate analysis of variance (Milliken and Johnson, 1984, p. 315-350). The linear and quadratic effects of selection cycles were tested by contrasts within analyses of variance, and selection responses were computed by linear regression.

Realized heritabilities for individual cycles were computed as the ratio of selection response to selection differential, the latter computed from the original nurseries (Casler, 1992). Realized heritabilities, pooled across cycles, were computed as the ratio of total realized gain to cumulative selection differential (Hill, 1972). Pooled realized heritabilities were subjected to an analysis of variance to test differences between selection growth stages (vegetative vs. heading), evaluation growth stages, laboratory methods (NIRS vs. wet-lab), and pollination method (in situ vs. polycross). Standard errors of realized heritability estimates were computed according to Prout (1962), pooling across cycles to correspond to pooled realized heritability estimates. Genotypic correlation coefficients between growth stages and between laboratory methods were estimated according to Falconer and Mackay (1996, p. 316). Genotypic correlations were computed for both sub-populations (cycles and methods with 14 df) and clones within sub-populations (303 df) using mean squares and mean products from a multivariate analysis of variance.

RESULTS

Differences among population means for NDF concentration were significant for all four evaluations: vegetative/wet-lab, vegetative/NIRS, heading/wet-lab, and heading/NIRS (P [is less than] 0.01). There were no significant population x year interactions for NDF concentration at either growth stage, which is consistent with results from other studies of smooth bromegrass nutritive value traits (Casler and Vogel, 1999). Therefore, all results are presented as means over 2 yr (Table 2).

Table 2. Means of neutral detergent fiber over two replicates, 2 yr, and 20 clones for each selection method, cycle, evaluation growth stage, and method of laboratory evaluation (Wet-laboratory or near-infrared reflectance spectroscopy, NIRS).
                                      Vegetative growth stage
Population/
Selection method([dagger])    Cycle     Wet-lab      NIRS
                                     -- g [kg.sup.-1] --

WB-[RP.sub.1]                   CO       406.8      403.4
vNB2                            C1       397.2      402.8
vNB2                            C2       381.8      387.2
vNB2                            C3       381.9      384.7
hNU2                            C1       401.8      403.2
hNU2                            C2       404.5      405.0
hNU2                            C3       389.8      386.0
hWU2                            C1       409.0      408.1
hWU2                            C2       399.0      405.1
hWU2                            C3       402.0      400.1
vNP3                            C1       402.5      406.6
vNP3                            C2       386.1      388.4
hNP3                            C1       404.7      398.4
hNP3                            C2       394.7      391.2
hWP3                            C1       405.5      405.0
hWP3                            C2       385.6      384.3

                                Heading growth stage
Population/
Selection method([dagger])     Wet-lab        NIRS
                                -- g [kg.sup.-1] --

WB-[RP.sub.1]                   595.3        598.7
vNB2                            592.5        588.9
vNB2                            577.6        587.0
vNB2                            583.8        582.1
hNU2                            589.8        590.1
hNU2                            606.4        596.4
hNU2                            599.2        594.1
hWU2                            603.9        602.8
hWU2                            599.9        595.0
hWU2                            582.7        593.6
vNP3                            594.4        591.4
vNP3                            585.9        581.4
hNP3                            598.1        589.1
hNP3                            586.2        582.2
hWP3                            605.2        598.9
hWP3                            579.0        573.9


([dagger]) Nomenclature: ijkl, where i = growth stage (v = vegetative or h = heading), j = laboratory method (N = NIRS or W = wet-lab), k = parental selection pressure/pollination method (B = biparental, U = uniparental, or P = replicated polycross), and 1 = number of years per cycle.

Recombination by Open-Pollination

Phenotypic recurrent selection at the vegetative growth stage (vNB2) reduced NDF concentration by -4.9 to -9.0 g [kg.sup.-1] [cycle.sup.-1] (-0.8 to -2.2% [cycle.sup.-l]). All responses were significant with P [is less than] 0.05 (Fig. 2). These responses were highly linear, with [R.sup.2] values ranging from 0.80 to 0.92. The greatest responses were measured at the vegetative growth stage, averaging 60% greater than responses measured at the heading growth stage.

[Figure 2 ILLUSTRATION OMITTED]

Selection for reduced NDF concentration at the heading growth stage resulted in less consistent responses than selection at the vegetative growth stage (Fig. 2). Responses ranged from +2.8 to -5.0 g [kg.sup.-1] [cycle.sup.-1] (+ 0.5 to - 1.2 % [cycle.sup.-1). Although quadratic responses were not significant, these responses were not highly linear, with [R.sup.2] ranging from 0.07 to 0.68. Direct selection responses at the heading growth stage were not significant, regardless of whether selection was based on NIRS (hNU2) or wet-laboratory (hWU2) methods. For selection responses measured at the vegetative growth stage, selection at the heading growth stage was effective based on NIRS, but not based on wet-laboratory methods. A large amount of transgressive segregation for reduced NDF concentration (15 of 80 = 18.8%) was observed in Cycles 1 through 3 of the vNB2 selection method (Fig. 3). The 95 % confidence interval of this percentage was (10.1, 27.5), suggesting that it is significant.

[Figure 3 ILLUSTRATION OMITTED]

Recombination by Polycrossing

Selection at the vegetative growth stage, using polycrossing for recombination of selected individuals (vNP3), was highly effective at reducing NDF concentration (Fig. 4). Linear responses ranged from -4.7 to -10.3 g [kg.sup.-1] [cycle.sup.-1] (-0.8 to -2.5% [cycle.sup.-1]), all but heading/wet-lab were significant with P [is less than] 0.05. These responses were highly linear with [R.sup.2] ranging from 0.60 to 0.99. There was a slight trend toward greater responses, on average, for the vegetative growth stage vs. the heading growth stage, as observed for the vNB2 selection method. There was little difference between estimates computed from NIRS vs. wet-laboratory evaluations, also as observed for the vNB2 method.

[Figure 4 ILLUSTRATION OMITTED]

Selection at the heading growth stage (hWP3 and hNP3) resulted in relatively consistent responses (Fig. 4). Responses to selection based on wet-laboratory analyses at heading (hWP3) were all significant and averaged 63% greater than responses to selection based on NIRS analyses at heading (hNP3), which were generally not significant. These results were consistent regardless of the growth stage or laboratory method of evaluation.

For the most effective of the three polycrossing selection methods (hWP3), the frequency of transgressive segregants was low [5 of 60 = 8.3%; 95% confidence interval of (1.2, 15.4)] and there was no indication of increased variability due to selection (Fig. 5). The cycles had similar ranges and variances, but differed primarily in their means, suggesting a shift in the entire distribution of NDF values.

[Figure 5 ILLUSTRATION OMITTED]

Open-Pollination vs. Polycrossing

Responses to selection practiced at the vegetative growth stage (vN__) were generally higher for open-pollination of selections (vNB2:-2.5 to -4.5 g [kg.sup.-1] [yr.sup.-1]) compared with polycrossing of selections (vNP3: -1.6 to -3.4 g [kg.sup.-1] [yr.sup.-1]) (Table 3). This was opposite for selection practiced at the heading growth stage: + 1.4 to -2.5 g [kg.sup.-1] [yr.sup.-1] for open-pollination of selections (h_U2) compared to -1.5 to -4.1 g [kg.sup.-1] [yr.sup.-1] for polycrossing of selections (h_P3). The vNB2 selection method had the largest mean response for a vegetative target growth stage, while the hWP3 selection method had the largest mean response for a heading target growth stage.

Table 3. Linear responses to selection for reduced NDF (neutral detergent fiber) concentration using each of six phenotypic recurrent selection methods described in Fig. 1, evaluated at two growth stages and with two laboratory methods.
                     Evaluation growth stage and method of
                             laboratory analysis
                          Vegetative growth stage
Selection          Wet-laboratory analysis   NIRS analysis
method([dagger])

vNB2                  -4.5 (-1.1) <0.01      -3.6 (-0.9) <0.01
hNU2                  -2.4 (-0.6)  0.05      -2.5 (-0.6)
hWU2                  -1.2 (-0.3)  0.32      -0.6 (-0.2)  0.60
vNP3                  -3.4 (-0.8) <0.01      -2.5 (-0.6)  0.05
hNP3                  -2.0 (-0.5)  0.12      -2.0 (-0.5)  0.12
hWP3                  -3.5 (-0.9) <0.01      -3.2 (-0.8) <0.01

                     Evaluation growth stage and method of
                             laboratory analysis
                           Heading growth stage
Selection          Wet-laboratory analysis   NIRS analysis
method([dagger])

vNB2                  -2.5 (-0.4) 0.05       -2.6 (-0.4)  0.05
hNU2                   1.4  (0.2) 0.28       -0.4 (-0.1)  0.78
hWU2                  -2.1 (-0.4) 0.11       -1.2 (-0.2)  0.38
vNP3                  -1.6 (-0.3) 0.26       -2.9 (-0.5)  0.03
hNP3                  -1.5 (-0.3) 0.28       -2.7 (-0.5)  0.04
hWP3                  -2.7 (-0.5) 0.05       -4.1 (-0.7) <0.01


([dagger]) Nomenclature: ijkl, where i = growth stage (v = vegetative or h = heading), j = laboratory method (N = NIRS or W = wet-lab), k = parental selection pressure/pollination method (B = biparental, U = uniparental, or P = replicated polycross), and I = number of years per cycle.

([double dagger]) Values following parentheses are P-values from F-tests for linear responses to selection.

Genetic gain per cycle of selection was slightly greater for vNP3 than for vNB2, averaging 18% greater across the four evaluations. However, because vNP3 required an extra year per cycle, its overall efficiency was lower than that of vNB2 (Table 3). Gain per year was always greater for vNB2 compared to vNP3. Selection response was similar for uniparental and polycross matings when NIRS was used to measure NDF at the heading growth stage.

Realized Heritability

Estimates of realized heritability were highly variable among the six selection methods (Table 4). The vegetative growth stage selection methods had approximately double the realized heritability of the heading growth stage selection methods, except when evaluated at heading with wet-laboratory methods. Methods of laboratory analysis during selection could only be compared for selection at the heading growth stage to avoid confounding with vegetative vs. heading growth stage differences. Selection based on NIRS had approximately double the realized heritability of selection based on wet-laboratory methods, but only when evaluated by NIRS. However, this difference was not large enough to be significant (P [is greater than] 0.05), nor was it consistent for the wet-laboratory evaluation. Realized heritability was consistently higher for the polycrossing method than for the open-pollination method, although this difference was significant (P [is less than] 0.05) only for the heading/NIRS evaluation and for the mean over the four evaluations.

Table 4. Estimates of realized heritability for each of six selection methods following two or three cycles of selection for reduced NDF (neutral detergent fiber) concentration, computed from evaluations at two growth stages and using two laboratory evaluation methods.
                             Evaluation growth stage and method
                                   of laboratory analysis
                             Vegetative growth stage
Selection method                    Wet-lab

vNB2                          0.262 [+ or -] 0.108
hNU2                          0.166 [+ or -] 0.113
hWU2                          0.037 [+ or -] 0.052
vNP3                          0.381 [+ or -] 0.173
hNP3                          0.171 [+ or -] 0.118
hWP3                          0.240 [+ or -] 0.083
Mean                                0.209
Selection growth stage
  Vegetative                        0.322
  Heading                           0.153(*)
Laboratory analysis method
  Wet-laboratory (hW-)              0.138
  NIRS (hN-)                        0.168 ns
Crossing method
  Open-pollination                  0.155
  Polycrossing                      0.264 ns

                             Evaluation growth stage and method
                                   of laboratory analysis
                             Vegetative growth stage
Selection method                       NIRS

vNB2                           0.373 [+ or -] 0.087
hNU2                           0.156 [+ or -] 0.093
hWU2                          -0.025 [+ or -] 0.049
vNP3                           0.217 [+ or -] 0.136
hNP3                           0.243 [+ or -] 0.110
hWP3                           0.197 [+ or -] 0.083
Mean                                 0.193
Selection growth stage
  Vegetative                         0.295
  Heading                            0.143(*)
Laboratory analysis method
  Wet-laboratory (hW-)               0.086
  NIRS (hN-)                         0.199 ns
Crossing method
  Open-pollination                   0.168
  Polycrossing                       0.219 ns

                             Evaluation growth stage and method
                                   of laboratory analysis
                             Heading growth stage
Selection method                    Wet-lab

vNB2                          0.121 [+ or -] 0.118
hNU2                         -0.038 [+ or -] 0.103
hWU2                          0.097 [+ or -] 0.064
vNP3                          0.172 [+ or -] 0.222
hNP3                          0.128 [+ or -] 0.143
hWP3                          0.184 [+ or -] 0.083
Mean                                0.110
Selection growth stage
  Vegetative                        0.146
  Heading                           0.093 ns
Laboratory analysis method
  Wet-laboratory (hW-)              0.140
  NIRS (hN-)                        0.045 ns
Crossing method
  Open-pollination                  0.060
  Polycrossing                      0.161 ns

                             Evaluation growth stage and method
                                   of laboratory analysis
                             Heading growth stage
Selection method                      NIRS            Mean

vNB2                          0.401 [+ or -] 0.106   0.289
hNU2                          0.152 [+ or -] 0.107   0.109
hWU2                          0.036 [+ or -] 0.096   0.036
vNP3                          0.454 [+ or -] 0.180   0.306
hNP3                          0.370 [+ or -] 0.150   0.228
hWP3                          0.278 [+ or -] 0.114   0.225
Mean                                0.282            0.199
Selection growth stage
  Vegetative                        0.428            0.279
  Heading                           0.209(**)        0.150(**)
Laboratory analysis method
  Wet-laboratory (hW-)              0.157            0.195
  NIRS (hN-)                        0.261 ns         0.233 ns
Crossing method
  Open-pollination                  0.196            0.160
  Polycrossing                      0.367(*)         0.268(**)


ns, (*), (**) Members of a pair within a column are not significantly different, or different at P < 0.05 or 0.01, respectively.

DISCUSSION

Calibration and validation statistics for the NIRS prediction equations for NDF concentration were remarkably consistent across years and nurseries. Values of [R.sup.2] were consistently high (0.72-0.90), but not as high as most people might desire for NDF calibration equations. This may be a result of the older model scanning monochromator (model 51A) and its inability to develop calibrations as precisely as can be obtained with more current equipment and software. For example, software developments made during the course of the selection experiment now allow the researcher to choose a calibration set using a stratified random sample based on wavelength characteristics. Such a sample will be much more robust with regard to prediction than the older method of stratified random sampling based on experimental design. This points out a potential irony inherent in long-term selection experiments. Methods, equipment, and software used during selection must be kept constant to maintain the integrity of hypothesis tests, but may become outdated during the course of the selection experiment itself.

Despite the age of the equipment and software used in this experiment, the results should be useful in predicting the value of these selection methods. As expected on the basis of [R.sup.2] values throughout the duration of the experiment (Table 1), simple correlations between wet-laboratory and NIRS measurements of NDF concentration were only moderately high: 0.79 and 0.81 for vegetative and heading growth stages, respectively (df = 1278). However, phenotypic correlations, with environmental main effects removed, were generally higher: 0.92 and 0.82 for vegetative and heading growth stages, respectively, for populations and 0.86 and 0.81 for vegetative and heading growth stages, respectively, for clones within populations. Finally, genotypic correlations, with environmental and genotype x environment interaction effects removed, were higher still: 0.96 [+ or -] 0.02 and 0.88 [+ or -] 0.07 for vegetative and heading growth stages, respectively, for populations (df = 14), and 0.94 m 0.01 and 1.00 m 0.00 for vegetative and heading growth stages, respectively, for clones within populations (df = 303). Thus, environmental and genotype x environment interaction effects within the calibration and validation data sets were a large part of the reason for low [R.sup.2] values.

The high genotypic correlation coefficients support the general observation that gains from selection and differences among the six selection methods were similar whether estimated by wet-laboratory or NIRS methods. Most of the variation in selection responses between wet-laboratory and NIRS measurements of progress appeared to be random in nature. Selection based on wet-laboratory or NIRS measurements of NDF did not lead to preferential gains for the corresponding laboratory method employed in the evaluation experiment (Fig. 2 and 4, Table 3). Although there was some evidence for differential heritability associated with wet-laboratory vs. NIRS analysis for the heading growth stage (Table 4), this was compensated by opposite differences in selection differentials between wet-laboratory and NIRS selection methods. Thus, for a given method of selection, the choice of wet-laboratory vs. NIRS evaluation of selection responses does not appear to be important. For large experiments, NIRS will likely be more efficient and appears to be more commonly employed than strictly wet-laboratory analysis in contemporary breeding programs.

Reich and Casler (1985) predicted that the vNB2 selection method would result in a 34% increase in selection response for low NDF concentration compared to the hWU2 or hNU2 selection methods. Their prediction was based on quantitative genetic theory and the use of smooth bromegrass as a hay crop, harvested at the heading growth stage. Three cycles of empirical selection to validate these predictions led to a 339% greater average response for the vNB2 selection method compared to the hNU2 and hWU2 selection methods for NDF measured at the heading growth stage. Thus, the empirical advantage of vNB2 was 10 times greater than predicted on the basis of quantitative genetic theory. This was despite genotypic correlation coefficients between vegetative and heading growth stages that were considerably less than 1 (0.64 [+ or -] 0.07 for NIRS and 0.77 [+ or -] 0.05 for wet laboratory evaluations). This may indicate that quantitative genetic parameters estimated by Reich and Casler (1985) lacked precision because of small sample size. It may also be due to differences between Reich and Casler's population WB8HD and the population used in the selection experiment WB[RP.sub.1]. The predicted selection response for NDF at the heading growth stage in WBHD was -3.8 g [kg.sup.-1] [yr.sup.-1] (Reich and Casler, 1985), while the average response in WB-[RP.sub.1] was -2.3 g [kg.sup.-1] [yr.sup.-1] (Table 3). This was similar to results of Hopkins et al. (1993) who showed that predicted gains were greater than realized gains from selection for in vitro dry matter digestibility (IVDMD) in switchgrass (Panicum virgatum L.).

In addition to its integral use in the vNB2 selection method, the vegetative growth stage would be representative of early-spring smooth bromegrass pasture. The vNB2 selection method had an empirical selection response 141% greater than the hNU2 and hWU2 selection methods when evaluated at the vegetative growth stage. The observed gain at this growth stage from vNB2 selection was -4.5 g [kg.sup.-1] [yr.sup.-1], compared with -2.5 g [kg.sup.-1] [yr.sup.-1] estimated by Reich and Casler (1985). The superiority of the empirical response compared with the predicted response was opposite of that observed at the heading growth stage and suggests that heritability of NDF concentration at the vegetative growth stage may be higher than expected on the basis of results of Reich and Casler (1985). Because selection intensities were held constant in this study, differences in selection responses between growth stages were due to differences in heritability and selection differential between those stages (Falconer and Mackay, 1996, p. 232). The consistently superior heritability from selection at the vegetative growth stage compared with the heading stage (Table 4) was likely due to greater within-plant uniformity at the vegetative stage. Vegetative plants consisted entirely of leaf blades and sheaths, whereas plants at heading contained varying proportions of leaf blades and sheaths, stems, and heads. In addition, stems within a plant varied widely in height, diameter, and leaf number, while some stems were not reproductive. All of these factors combined to make the process of random sampling extremely difficult at the heading growth stage. This sampling variation likely caused serious reductions in heritability. Thus, for the open-pollination methods of recombination, selection at the vegetative growth stage was most effective regardless of the target growth stage.

Although phenotypic recurrent selection using the vNB2 method produced a high frequency of transgressive segregants for low NDF concentration, it was not effective at eliminating high-NDF plants, as noted by transgressive segregants for high NDF concentration. Because of the nature of the vNB2 method, there were large distances between "neighboring" plants in the in situ crossing blocks, and some plants were as far apart as 40 m. Smooth bromegrass is wind pollinated, but most pollen does not travel great distances. Almost no pollen is dispersed upwind and downwind pollen frequency decreases rapidly with distance (Jones and Newell, 1946). Furthermore, pollination among smooth bromegrass plants is non-random (Hittle, 1954; Knowles, 1969), due partly to differential cross-fertility among clones (Adams, 1953) and distance between clones (Jones and Newell, 1946). Thus, the vNB2 selection method most likely favors non-random pollination among clones. Greater distances between clones likely led to a higher frequency of full-sibs, and possibly selfed progeny, within seed lots harvested from a single plant. Increasing the frequency of full-sibs and selfed progeny will eventually cause greater inbreeding in the vNB2 selection method compared with hNU2 and hWU2 or in a polycrossing method. The residual transgressive segregants for high NDF concentration may be a result of this mating system.

Concern over this issue led to the inclusion of the three polycrossing methods of intercrossing selected individuals. These methods were patterned after the three open-pollination methods, with the exception of requiring an extra year per cycle for polycrossing. Polycrossing increased the heritability and selection response per cycle for nearly all combinations of selection method with evaluation growth stage and method of laboratory analysis (Fig. 2 vs. 4, Tables 3 and 4). Compared with their open-pollination counterparts, the greatest improvements in heritability and selection response due to polycrossing were for the hWP3 selection method (compared with hWU2). Polycrossing improved this selection method sufficiently that it ranked highest in selection response per year at the heading growth stage. It resulted in a selection response at the heading growth stage that averaged 33 % greater than that observed with the vNB2 method and 106% greater than that observed with the hWU2 method, its open-pollination counterpart. Its superiority over the vNB2 method may be attributed to genotypic correlation coefficients between growth stages significantly less than 1. The most likely explanation for the difference between hWP3 and hWU2 is that polycrossing resulted in significantly better interpollination of selected individuals, reducing non-random pollination, chain-crossing, and possible selfing. The improvement in random pollination and intercrossing most likely led to greater recombination between a larger number of selected individuals than was possible in the vNB2 method. This recombination did not lead to as much transgressive segregation as observed in the vNB2 selection method (Fig. 3 vs. 5), but it resulted in relatively constant variance and range among individuals as the population mean decreased due to selection.

In contrast to the hWP3 method, the hNP3 method did not respond consistently to the polycrossing option, i.e., hNP3 was not consistently superior to hNU2. Realized heritabilities were higher for hWP3 than for hWU2, but selection differentials were all lower in the original nurseries and phenotypic variances were lower for hWP3 than hWU2 in the evaluation nursery. Because selection response is a function of the product of heritability and selection differential, there was no advantage conferred to the hNP3 method resulting from the polycrossing option. The lower phenotypic variance in the original nurseries and evaluation plots for hNP3 compared with hWP3 was a general phenomenon observed in this study for wet-laboratory vs. NIRS measurements. The NIRS calibration reduced the amount of phenotypic variance within populations during both the selection and evaluation phases of this experiment, by an amount equal to 1- [r.sup.2], where r = the simple correlation coefficient between wet-laboratory and NIRS NDF values. Thus, to be more effective than a wet-laboratory selection method, a NIRS selection method must have sufficiently higher heritability to offset this reduction in phenotypic variance. This did not occur for the two heading-stage selection methods that used polycrossing. In fact, it only occurred for hNU2 vs. hWU2 when evaluated at the vegetative growth stage, causing reduced gains for NIRS vs. wet-laboratory analysis for all other hN_vs. hW_comparisons (Table 4). Thus, NIRS analysis may be more sensitive than wet-laboratory analysis to the extreme sampling variation at the heading growth stage.

On the basis of quantitative genetic theory (Falconer and Mackay, 1996, p. 191), the hNP3 and hWP3 selection methods should have had twice the selection response per cycle and 50% greater responses per year than their open-pollination counterparts because of selection of both sexes. Empirically, these two polycrossing methods had an average 3.6 times greater response per cycle and 2.4 times greater response per year than their open-pollination counterparts. The unexpected improvement due to polycrossing suggested that polycrossing of selected individuals provided more of a benefit than simply allowing selection among both male and female gametes. Polycrossing may allow more random pollination than is possible within an open-pollination nursery without selection among males. Because of the dramatic reduction in cross-pollination between smooth bromegrass plants as their pairwise distance increases, large open-pollination nurseries are likely subject to non-random pollination.

On the basis of this logic, there should have been a similar positive effect of polycrossing for the vegetative growth stage selection methods. However, the lack of improvement for vNP3 vs. vNB2 suggests that the selection response for vNB2 was near the genetic maximum for this population. For a given selection intensity (i), the upper limit of selection response (R) is fixed by the phenotypic standard deviation ([Sigma].sub.P]) and additive genetic variance ([MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]), i.e., [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (Falconer and Mackay, 1996, p. 316). While additive genetic variance may be increased by additional and more complete recombination among selected individuals, it is also reduced by selection of the extreme lowest individuals (Falconer and Mackay, 1996, p. 201), limiting its possible role in increasing selection response due to polycrossing. Conversely, decreases to the phenotypic standard deviation can only arise by reduction to environmental or genotype x environment interaction effects or an increase in replication of the selection units. Having been based on highly uniform plants made up only of leaf blades and sheaths, the vNB2 method may have reduced environmental and/or genotype x environment interaction effects by the maximum amount for unreplicated selection units.

CONCLUSIONS

The concentration of NDF in WB-[RP.sub.1] smooth bromegrass is a moderately heritable trait. The progress that can be achieved by plant breeding in a relatively short time period has a potentially significant impact on the feeding value of smooth bromegrass forage. This research tested the effectiveness of the Reich and Casler (1985) proposal that selection efficiency can be improved by selection at a vegetative growth stage, with selection of both male and female plants. Heritability was considerably higher when selection was conducted at a vegetative growth stage than at the heading growth stage, likely because of highly uniform plant morphology. However, because of moderate genetic correlations between harvest growth stages, the optimal selection system for each growth stage was based on selection at that growth stage. Polycrossing of selected individuals increased the time required by 50%, a loss of efficiency that was rarely compensated by the increased gains associated with polycrossing. Nevertheless, for long-term selection programs that aim to reduce NDF concentration by more than the average of 13.6 g [kg.sup.-1] in the 6 yr of this selection study, polycrossing of selected individuals is likely the better choice. While in situ open-pollination combined with both female and male selection (i.e., the Reich and Casler proposal) resulted in the greatest average gains, it appears susceptible to significant inbreeding due to non-random pollination.

The use of NIRS in selecting for reduced NDF concentration can significantly reduce the cost and time required for selection, allowing for a potential increase in selection intensity and considerable utility to forage breeding programs. However, the NIRS technique appeared to be more sensitive than the wet-laboratory technique to the extreme sampling variation at the heading growth stage, generally showing reduced progress. Thus, to achieve the same results, the use of NIRS for evaluation of NDF at reproductive growth stages may require considerably more care and/or replication than required at vegetative growth stages.

Finally, the magnitude of realized gains was lower in this study than predicted from a previous study, but this may be attributed to differences in the reference population. More important, the relative differences among selection methods in predicted gains agreed with the relative differences in realized gains. Thus, genetic parameter estimates may be useful in developing efficient selection methods for forage nutritive value traits.

ACKNOWLEDGMENTS

I thank several research technicians (Margaret Straub, Chuck Gedye, Debbie Schneider, and Kim Darling) for their dedication, perseverence, patience, and faith, working for years on a long-term selection experiment without tangible evidence of progress. I also thank Jodi Scheffler for her dedication and willingness to provide interim assistance with this experiment and to help train a new research technician.

REFERENCES

Adams, M.W. 1953. Cross- and self-incompatability in relation to seed setting in Bromus inermis. Bot. Gaz. (Chicago) 115:95-105.

Aung, M., D.E. Rowe, and R.G. Pratt. 1994. Necessity of replicated measurements for selection of alfalfa plants resistant or susceptible to stem inoculation by Sclerotinia trifoliorum. Plant Dis. 78:14-17.

Burton, G.W. 1974. Recurrent restricted phenotypic selection increases forage yields of pensacola bahiagrass. Crop Sci. 14:831-834.

Burton, G.W. 1982. Improved recurrent restricted phenotypic selection increases bahiagrass forage yields. Crop Sci. 22:1058-1061.

Casler, M.D. 1992. Usefulness of the grid system in phenotypic selection for smooth bromegrass fiber concentration. Euphytica 63:239-243.

Casler, M.D., and I.T. Carlson. 1995. Smooth bromegrass, p. 313-324. In R.F Barnes et al. (ed.) Forages. Vol. I. An introduction to grassland agriculture. Iowa State Univ. Press, Ames, IA.

Casler, M.D., J.F. Pedersen, G.C. Eizenga, and S.D. Stratton. 1996. Germplasm and cultivar development, p. 413-469. In L.E. Moser et al. (ed.) Cool-season forage grasses. ASA, Madison, WI.

Casler, M.D., and G.A. Pederson. 1996. Host resistance and tolerance and its deployment, p. 475-507. In Chakraborty et al. (ed.) Pasture and forage crop pathology. ASA, Madison, WI.

Casler, M.D., and K.P. Vogel. 1999. Accomplishments and impact from breeding for increased forage nutritive value. Crop Sci. 39:12-20.

Ehlke, N.J., M.D. Casler, P.N. Drolsom, and J.S. Shenk. 1986. Divergent selection for in vitro dry matter digestibility in smooth bromegrass. Crop Sci. 26:1123-1126.

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

Goering, H.K., and P.J. Van Soest. 1970. Forage fiber analysis (apparatus, reagents, procedures, and some applications). USDA Agric. Handb. 379. U.S. Gov. Print. Office, Washington, DC.

Hallauer, A.R., and J.B. Miranda Fo. 1991. Quantitative genetics in maize breeding. 2nd ed. Iowa State Univ. Press, Ames, IA.

Hill, W.G. 1972. Estimation of realized heritabilities from selection experiments. I. Selection in one direction. Biometrics 28:767.

Hittle, C.N. 1954. A study of the polycross progeny testing technique as used in the breeding of smooth bromegrass. Agron. J. 46:521-523.

Hopkins, A.A., K.P. Vogel, and K.J. Moore. 1993. Predicted and realized gains from selection for in vitro dry matter digestibility and forage yield in switchgrass. Crop Sci. 33:253-258.

Jones, M.D., and L.C. Newell. 1946. Pollination cycles and pollen dispersal in relation to grass improvement. Nebraska Agric. Exp. Stn. Res. Bull. 148.

Knowles, R.P. 1969. Nonrandom pollination in polycrosses of smooth bromegrass. Agron. J. 47:15-19.

Milliken, G.A., and D.E. Johnson. 1984. Analysis of messy data. I. Designed experiments. Van Nostrand Reinhold Co., New York.

Prout, T. 1962. The error variance of the heritability estimate obtained from selection response. Biometrics 18:404-407.

Reich, J.M., and M.D. Casler. 1985. Effect of maturity and alfalfa competition on expected selection response for smooth bromegrass forage quality traits. Crop Sci. 25:635-640.

Simon, U., and B.H. Park. 1983. A descriptive scheme for stages of development in perennial forage grasses, p. 416-418. In J.A. Smith and V.W. Hays (ed.) Proc. XIV Intl. Grassl. Congr., Lexington, KY. 15-24 June 1981. Westview Press, Boulder, CO.

Surprenant, J., D.K. Barnes, R.H. Busch, and G.C. Marten. 1988. Bidirectional selection for neutral detergent fiber and yield in reed canarygrass. Can. J. Plant Sci. 68:705-712.

Dep. of Agronomy, Univ. of Wisconsin-Madison, Madison, WI 537061597. Research supported by Hatch formula funds. Received 6 March 1998. M.D. Casler, Corresponding author (mdcasler@facstaff.wisc.edu).

Published in Crop Sci. 39:381-390 (1999).
COPYRIGHT 1999 Crop Science Society of America
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1999 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Casler, M.D.
Publication:Crop Science
Date:Mar 1, 1999
Words:7917
Previous Article:Field Performance of Cell Suspension-Derived Tall Fescue Regenerants and Their Half-Sib Families.
Next Article:Two Cycles of Recurrent Selection for Seed Yield in Common Bean.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters