Crop Science
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 19 March 2008
Published in Crop Sci 48:434-442 (2008)
© 2008 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Casler, M. D.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Casler, M. D.
Agricola
Right arrow Articles by Casler, M. D.
Related Collections
Right arrow Other Forage Crops
Right arrow Crop Genetics

Among-and-within-Family Selection in Eight Forage Grass Populations

M. D. Casler*

USDA-ARS, U.S. Dairy Forage Research Center, Madison, WI 53706-1108. Research supported by the College of Agricultural and Life Sciences, Univ. of Wisconsin, Madison, WI 53706-1197

* Corresponding author (mdcasler{at}wisc.edu).


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Forage yields increased little during the twentieth century, despite intensive breeding efforts in many species. Half-sib family (HSF) and/or among-and-within-family (AWF) selection methods may overcome this problem. The objective of this study was to compare HSF and AWF selection using forage yield of sward plots as the among-family selection criterion and natural selection of survivors within swards as the within-family selection criterion. Selection for increased forage yield was practiced in eight populations of orchardgrass (Dactylis glomerata L.), smooth bromegrass (Bromus inermis Leyss.), and hybrid wheatgrass [Elytrigia x muctonata (Opiz ex Bercht.) Prokud.]. Two methods of recombination were used: random plants from remnant seed (HSF) or sod cores from selected families four years after establishment (AWF). There were no consistent differences between HSF and AWF selection for orchardgrass, a bunch grass. For the other two species, both highly rhizomatous, AWF selection was two to four times more effective than HSF selection for increasing forage yield. Natural selection within families of the two rhizomatous grasses favored genotypes capable of filling in open spaces left by plant and/or tiller mortality. The consistent differences in selection responses between orchardgrass and the two rhizomatous species suggested that natural selection acted on some characteristic of the rhizomatous trait of these two grasses.

Abbreviations: AWF, among-and-within-family • HSF, half-sib family • HSPT, half-sib progeny-test • NNA, nearest neighbor analysis • PLS, pure live seed • SIR, sets-in-replicate


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
FORAGE YIELDS INCREASED little during the twentieth century, despite intensive breeding efforts in a large number of species. In Europe, forage yield increased by 4 to 5% per decade while grain yield of cereal crops increased by 14% per decade (Humphreys, 1997). In North America, the gap between forage and cereal crops is greater due to significantly lower gains in forage yield (Casler et al., 2000a, 2000b, 2001; Lamb et al., 2006). Reduced progress in increasing forage yield in North America vs. Europe is most likely due to the vast majority of European breeding efforts located in the private sector, while nearly all forage breeding in North America occurs in the public sector, where applied plant breeding is often diluted by the need to design and publish scientific research findings. This difference leads to greater emphasis on forage yield per se, greater emphasis on cultivar development relative to scientific advancement, higher selection intensities, and longer term stability and uniformity of selection pressures for the private European programs. Only recently and only in alfalfa (Medicago sativa L.) has there been much demand for increased forage yields in North America.

The lag in yields of harvestable biomass of forage crops relative to cereal crops can be attributed to exploitation of heterosis in some cereal crops, huge improvements in harvest index of most cereal crops, the shorter breeding cycle of annual cereal crops compared to perennial forages, and the overwhelming importance of other traits in forage crops (Humphreys, 1997). Particularly in North America, there has been very little emphasis on direct selection for forage yield and many of these efforts have been misplaced and misdirected, failing to accurately measure forage yield breeding values. Populations are often synthesized from a widely diverse array of germplasm, creating potential heterotic effects and linkage disequilibria that disrupt the breeder's ability to accurately measure breeding values of parental genotypes (Brummer, 2005). Selection intensities are often low, due to the limited number of families evaluated in genotypic selection programs (Casler et al., 2000b). Selection is often conducted on spaced plantings, which are frequently unable to provide accurate predictions of sward-plot forage yield (Casler et al., 1996).

Because of low correlations between spaced-plant and sward-plot forage yield, many attempts to improve sward-plot forage yield by selecting on the basis of spaced plants have been unsuccessful (Carpenter and Casler, 1990; Hayward and Vivero, 1984; Surprenant et al., 1988). Notable successes include two perennials (Pensacola bahiagrass, Paspalum notatum var. saure Parodi [Burton, 1982] and Italian ryegrass, Lolium multiflorum Lam. [Fujimoto and Suzuki, 1975]) and one annual (rye, Secale cereale L. [Bruckner et al., 1991]). In Pensacola bahiagrass, average genetic gains for forage yield of 6.3% yr–1 over eight years (Burton, 1982) have far surpassed long-term gains typically observed in perennial forage crops (Humphreys, 1997).

Genotypic selection methods received wide attention by forage grass breeders in the 1940s and 1950s, but fell out of favor in the 1960s when breeders began to document the rapid progress achievable via phenotypic selection for important traits other than forage yield per se. The convincing argument of Breese and Hayward (1972), that genotypic selection squanders additive genetic variation of populations (due to inability to utilize genetic variances within families), was probably an additional important factor. While methods such as half-sib family (HSF) selection are highly inefficient with regard to additive genetic variance of the entire population, among- and-within-family (AWF) selection methods provide a mechanism to utilize all the additive genetic variance within a population, provided that meaningful selection pressure can be placed on plants within populations. This might involve measurement of the same trait on all plants, either as spaced plants (Vogel and Pedersen, 1993) or as spaced plants in swards (van Dijk and Winkelhorst, 1978), or by natural selection of survivors within swards. The objective of this study was to compare HSF and AWF selection using forage yield of sward plots as the among-family selection criterion and natural selection of survivors within swards as the within-family selection criterion.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Germplasm
Eight populations were selected for this experiment: two of smooth bromegrass, two of hybrid wheatgrass, and four of orchardgrass (Casler, 1998). Each population had gone through a minimum of one generation of random mating at Arlington, WI, following the selection of source materials. Random seeds were germinated in the glasshouse and seedlings were split into three clonal ramets. Clonal propagules were transplanted to the field at Arlington in May 1991 into polycross blocks. The design of each polycross block was a randomized complete block with three replicates. Plants were spaced on 0.9-m centers. Plants were clipped twice during the establishment year and fertilized once with 56 kg N ha–1.

In April 1992, plants were fertilized again with 56 kg N ha–1. Annual weeds were controlled by herbicide as described by Casler (1998). Seeds were harvested on each clone in July, threshed and cleaned, and equal amounts of seed from the three ramets of each clone were bulked to create each polycross family. Each polycross block was initiated with 200 clones for smooth bromegrass, 350 for hybrid wheatgrass, or 150 for orchardgrass. Some clones or families were discarded due to poor transplant vigor or establishment, lack of persistence over the winter of 1991–1992, insufficient seed production in 1992, or at random so that the number of families per population was a multiple of seven.

Polycross Family Evaluation
Polycross plots, measuring 0.9 by 1.5 m were established at Arlington in April 1993 using a five-row drill. The soil type was a Plano silt loam (fine-silty, mixed, mesic Typic Argiudoll). Seeding rates were: 650 pure live seed (PLS) m–2 (~22.4 kg ha–1) for smooth bromegrass and hybrid wheatgrass and 1600 PLS m–2 (~11.2 kg ha–1) for orchardgrass. Each family was tested for germination and seeding rates were adjusted to a PLS basis for each family. The experimental design was a sets-in-replicates design (Schutz and Cockerham, 1966) with 50 sets for orchardgrass and smooth bromegrass or 60 for hybrid wheatgrass, seven polycross families per set, and two replicates. Set size was determined partly by data from an orchardgrass uniformity trial (Casler and Tageldin, 1996) and partly by previous experiences with block sizes ranging from 4 to 35 in numerous cultivar tests. Each set contained polycross families from only one population. Plots were clipped twice during the establishment year and fertilized once with 56 kg N ha–1.

Plots were harvested with a flail-type harvester three times each (for orchardgrass) or twice each (for the other species) in 1994 and 1995. Samples of approximately 500 g were taken for dry matter determination. First harvest was timed for late boot of smooth bromegrass and hybrid wheatgrass and anthesis of the earliest orchardgrass families. Second and third harvests of orchardgrass occurred in late July and late October, respectively. Second harvest of the other species occurred in late October. Plots were fertilized with 90 kg N ha–1 in early spring and after first and second harvests in 1994 and 1995. Levels of P and K were adequate according to soil test results. Forage yield data was expressed as the sum over harvests within each year.

Two methods of controlling intrablock error were investigated: the sets-in-replicates design per se (SIR) and nearest neighbor analysis (NNA) (Brownie et al., 1993). Nearest neighbor analysis was applied by computing moving means of the residuals from the model

Formula 1[1]
where Xij = the season total forage yield of the ijth plot, m = the overall mean, ti = the effect of the ith family, and eij = the residual of the jth plot of the ith family. These residuals represent estimates of environmental yield potential for each plot (Tamura et al., 1988) and were computed separately for each year. Two NNA covariates were computed: the mean of the two north–south adjacent plots and the mean of the two east–west adjacent plots (Pearce and Moore, 1976). The nearest neighbor covariates were computed separately for each year.

Adjusted family means were computed for two methods of error control: the SIR design and NNA. Adjusted family means for the SIR design were computed as

Formula 2[2]
where Xijk = the yield of the ith family in the jth set in the kth replicate, M.jk = the mean of the jth set in the kth replicate, and M... = the overall mean, and then averaging values of X'ijk for the two replicates. Selection on the basis of the SIR adjustment was equivalent to the within-block method of selection of Schutz and Cockerham (1966) and the homogeneous variances method of Bos (1983). Adjusted family means for NNA were computed as least-squares means from the NNA of covariance. The relative efficiency of spatial adjustment ranged from 265 to 371% for SIR and 236 to 295% for NNA (Casler, 1998). The rank correlation of family means between adjusted family means for SIR and NNA ranged from 0.50 to 0.59, indicating a significant difference between the two adjustment methods (Casler, 1998).

Selection and Recombination
Four selection and recombination methods were used to create selected populations. Half-sib family selection (HSF) was used to select the best eight families, which were recombined in crossing blocks consisting of 50 seedlings per family in a completely randomized design (400 plants per crossing block). Seedlings were generated from remnant seed stored at –5°C. Among- and-within half-sib family selection (AWF) was used to select within the same eight families, based on circular 10-cm sod cores dug at random from 3-yr-old plots. Mean ground cover after 3 yr was 62% for smooth bromegrass, 73% for hybrid wheatgrass, and 81% for orchardgrass. Although plants were selected at random from these plots, the key to success of this breeding method is that these survivors have higher breeding value than strictly random plants taken from a bag of remnant seed. Twenty-four cores from each replicate of each family were planted in a crossing block for recombination (384 sod cores, which may represent a larger number of actual genotypes). Within each selection method, the eight families with the highest adjusted forage yield within each population were chosen based on the SIR or NNA adjustment methods, resulting in four selection and recombination methods: HSF-SIR, HSF-NNA, AWF-SIR, and AWF-NNA. Thus, selection generated a total of 32 selected populations (eight base populations x two recombination methods x two adjustment methods).

Selected sod cores or seedlings were established in crossing blocks in May 1996. Seedlings were established in the greenhouse 10 wk before transplanting to the field. No attempt was made to separate multiple potential genotypes within sod cores. Crossing blocks were grouped by the four selection-recombination methods, which were isolated by a minimum distance of 0.5 km. A fifth group of eight crossing blocks was also created with 400 unselected (random) seedlings of each base population. These eight crossing blocks represented random seed increases of each base population. Within each selection-recombination method, crossing blocks for the eight base populations were staggered by species (smooth bromegrass–orchardgrass–hybrid wheatgrass–orchardgrass...) giving a minimum same-species isolation distance of 40 m for orchardgrass or 80 m for the other two species. Plants within each of the 40 crossing blocks were fertilized with 56 kg N ha–1 in spring 1997. Annual weeds were controlled as described by Casler (1998). In July 1997, seed was harvested from each plant, threshed, cleaned, and bulked in equal quantities across all plants within each crossing block.

Evaluation of Selection Progress
Germination of the 40 1997 populations was tested by standard procedures in March 1998 (Association of Official Seed Analysts, 1998). The 40 populations were planted in 0.9- by 1.5-m plots at three locations in May 1998. The locations and soil types were Arlington, WI (43°20' N, 89°23' W, Plano silt loam [fine-silty, mixed, mesic, Typic Argiudoll]); Marshfield, WI (44°39' N, 90°08' W, Withee silt loam [fine-loamy, mixed, superactive, frigid Aquic Glossudalf]), and Ashland, WI (46°35' N, 90°54' W, Portwing silt loam [fine, mixed, superactive, frigid Oxyaquic Glossudalf]). Populations of each species were planted in separate experiments at each location. The experimental design was a randomized complete block with 16 blocks for smooth bromegrass and hybrid wheatgrass or four blocks for orchardgrass (due to limited seed production). Within each block, the random seed increases of each base population (Cycle 0) were replicated four times. Base populations (and all populations derived from them) were arranged as whole plots within each complete block.

Seeding-year management, fertilization, harvest management, and harvest methods were identical to those used on the original polycross family evaluation (1993–1995), with the exception that data were collected for 3 yr (1999–2001). Ground cover percentage was visually rated immediately following first harvest in 2001. Because all plots had excellent establishment, ground cover represented persistence of each population over the 3-yr life of the stand. Maturity of orchardgrass plots was visually rated immediately before first harvest of each year, using the 1-to-8 scale of Casler (1988). Leafspot of orchardgrass (caused by Drechslera spp., most likely D. dactylidis Shoem.) was visually rated immediately before second and third harvests of each year, using a scale of 0 to 10, where 0 = no symptoms and 10 = leaves completely covered by lesions.

All data were analyzed by analysis of variance with spatial adjustment for within-block variation (Casler, 1998, 1999). Spatial adjustment was conducted on total forage yields for each location-year combination using three spatial adjustment methods: trend analysis, NNA with one covariate, or NNA with two covariates (Casler, 1998, 1999). The best spatial adjustment method was chosen for each location-year combination based on the lowest average variance of an adjusted population mean. Then, for each location-year combination, the following three-step procedure was used to create spatial-adjusted plot values: (i) run an ANOVA without blocks or populations, including only the spatial covariate adjustment terms; (ii) output the residuals from this ANOVA, which contain all variability that could not be described by the spatial covariance adjustment; and (iii) add the grand mean to each residual value to restore the original scale of measurement.

Spatial-adjusted plot values were used in combined analyses of variance across locations and years, subtracting the appropriate number of degrees of freedom from error for the total number of covariate terms (Smith and Casler, 2004). All effects were assumed to be random, except populations, which were fixed. Orthogonal contrasts were used to test the five selection effects: overall change (C1 mean – C0 mean), the effect of HSF selection, the effect of AWF selection, the effect of SIR adjustment before selection, and the effect of NNA adjustment before selection.


   RESULTS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Population x location, population x year, and population x location x year were significant in most analyses of variance, accounting for less than 40% of the variance of a population mean in most cases (data not shown). Individual contrasts of selection effects generally did not show interactions with locations or years or, if they did show such an interaction, it was nearly always manifested as a change in magnitude of selection effects across locations or years, seldom as a change in direction of response. Furthermore any such crossover interactions of selection effects with locations or years were not consistent, often manifested as a single unusual location-year combination for one of the eight base populations. Therefore, all results are presented as means over three locations and three years. Data are not presented for ground cover of orchardgrass, because there were no significant differences among populations.

Orchardgrass
Selection for increased forage yield was highly effective in WO88S-Alt orchardgrass, the most highly variable of the four base populations and the base population with the lowest mean forage yield (Table 1 ). Selection for increased forage yield was ineffective in WOPI-C3 and WOPI-C16 and moderately effective in WOPI-C19. The four selection and recombination methods were nearly identical in their effects on all four orchardgrass populations, with an average improvement in forage yield of 2.4% in WOPI-C19 and 16.3% in WO88S-Alt. The marginal and inconsistent significance levels of selection effects for the three WOPI populations were likely due to the effect of poor seed production at Arlington in 1997 and its ultimate effect on reducing the number of replicates for the orchardgrass selection evaluation compared to the other two species.


View this table:
[in this window]
[in a new window]

  		Table 1. Mean forage yield of original and selected orchardgrass populations. Means are over four replicates, three locations, and three years.

 
Mean maturity stage at the time of first harvest was significantly altered by selection for increased forage yield in all four populations (Table 2 ). The greatest responses for maturity occurred in WOPI-C19 and WO88S-Alt, the two populations with significant responses for increased forage yield. The selection response for maturity was toward later maturity in all three WOPI populations, which tended to be medium to early maturity (Casler, 1998). Conversely, the selection response was toward earlier maturity in WO88S-Alt, the late-maturity base population. There were no obvious differences or trends for maturity selection responses among the four selection and recombination methods.


View this table:
[in this window]
[in a new window]

  		Table 2. Mean maturity score of original and selected orchardgrass populations. Means are over four replicates, three locations, and three years.

 
Selection responses for Drechslera reaction were highly inconsistent and generally reflected a low apparent genetic correlation with forage yield (Table 3 ). Three of four significant responses indicated a shift toward reduced reaction to Drechslera, but the majority of responses were not significant. Selection responses for Drechslera reaction tended to be greater for AWF than for HSF selection and for SIR than for NNA adjustment, measured in WOPI-C19 and WO88S-Alt, but the opposite direction of response in these two populations suggested that such a conclusion should be viewed as tentative at best.


View this table:
[in this window]
[in a new window]

  		Table 3. Mean Drechslera leafspot reaction of original and selected orchardgrass populations. Means are over four replicates, three locations, three years, and two harvests.

 
Smooth Bromegrass
Selection for increased forage yield was effective in both smooth bromegrass populations, based on at least one of the four selection and recombination methods (Table 4 ). Averaged over the two populations, AWF selection (3.0% gain) was 2.1 times more effective than HSF selection (1.4% gain) and SIR adjustment (3.1% gain) was 2.5 times more effective than NNA adjustment (1.3% gain). Average gains were greatest for WB88S, the population with the lowest mean forage yield.


View this table:
[in this window]
[in a new window]

  		Table 4. Mean forage yield and ground cover of selected smooth bromegrass populations deriving from two base populations (WB19e and WB88S). Forage yield means are over three locations, three years, and 16 replicates. Ground cover means are over three locations and 16 replicates.

 
Ground cover responded to selection in only one of the four selection and recombination methods, AWF selection with SIR adjustment (Table 4). The response was similar for the two populations and resulted in a 14-fold greater selection response for AWF selection over HSF selection (average gain of 4.2 vs. 0.3%) and a 23-fold greater selection response for SIR over NNA adjustment (average gain of 4.3 vs. 0.2%), averaged over the two populations.

Hybrid Wheatgrass
Two of the four selection and recombination methods resulted in increased forage yield of both hybrid wheatgrass populations, averaging 3.9% improvement over the base populations (Table 5 ). The two spatial adjustment methods had nearly equal effects on selection gains, but were both ineffective within the HSF selection method. For hybrid wheatgrass, AWF selection was effective and HSF selection was ineffective, while the two adjustment methods had no effect on selection gains. Selection tended to produce greater gains in forage yield in WEm-3LR, the base population with lower forage yield, but this effect was not as dramatic as observed for orchardgrass and smooth bromegrass.


View this table:
[in this window]
[in a new window]

  		Table 5. Mean forage yield and ground cover of selected hybrid wheatgrass populations deriving from two base populations (WEm-1LR and WEm-3LR). Forage yield means are over three locations, three years, and 16 replicates. Ground cover means are over three locations and 16 replicates.

 
Selection responses for ground cover were opposite for the two populations (Table 5). For WEm-1LR, the population with the lowest ground cover, both AWF selection criteria led to significant gains in ground cover. Conversely, WEm-3LR had nearly perfect ground cover, so the only direction that selection could move the population mean would be in the negative direction. Both HSF selection criteria resulted in significant reductions in ground cover, while both AWF selection criteria resulted in no change for WEm-3LR.

Observed vs. Expected Gains
Across the eight populations, observed gains in forage yield were not related to expected gains for forage yield (Fig. 1 ). However, there were some notable trends within species. Within hybrid wheatgrass, an expected difference in gains was not realized, as both populations had similar observed gains. Within smooth bromegrass, expected gains were predictive of the difference in observed gains for forage yield. Within orchardgrass, observed and expected gains were not related to each other across the four populations, but there was a very close relationship between observed and expected gains for the three WOPI populations, taken together, compared to WO88S-Alt, which has a different origin than WOPI (Casler, 1998). For the three WOPI populations taken together vs. WO88S-Alt, expected gains predicted that WO88S-Alt would have approximately double the gains of WOPI, whereas realized gains were more than 15-fold greater for WO88S-Alt vs. WOPI.


Figure 1
View larger version (14K):
[in this window]
[in a new window]

  		Figure 1. Relationship of observed gains for half-sib family (HSF) selection (left) and among-and-within-family (AWF) selection (right) vs. expected gains for HSF selection within eight populations of perennial grasses, the latter data derived from Casler (1998).

 

   DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Half-sib family selection and AWF selection differ in several fundamental and important characteristics. Half-sib family selection utilizes one selection criterion, forage yield, adjusted for spatial variation by either one of two methods. Genetic variation utilized by HSF selection is that present among HSF means, theoretically one-quarter of the additive genetic variance within the population, based on a disomic inheritance model (Falconer and Mackay, 1996; Hallauer and Miranda, 1988). Because random seeds from selected families were used to contribute gametes to the next generation, and C1 populations were created as balanced bulks of all plants within their respective crossing block, HSF selection did not utilize any within-family genetic variation.

Utilization of within-family genetic variation in forage grass breeding requires some method of measuring a meaningful trait on individual plants within each family. Spaced plantings are extremely common in forage grass breeding and provide a basis to evaluate and select superior families and superior plants within those families (Vogel and Pedersen, 1993). For spaced-plant traits, AWF selection is expected to be 3.6 times more effective than HSF selection, assuming that among-family and within-family selection are based on the same trait and that the among-family and within-family phenotypic variances are similar in magnitude (Vogel and Pedersen, 1993). The key to success of an AWF selection scheme is to ensure that meaningful and heritable traits are measured both on a family-mean and individual-plant basis. With spaced plantings, this is often impossible (Casler et al., 1996).

Accurate measurement of forage yield on spaced plants is often an impediment to use of spaced plantings for yield improvement. There are a few successes in which selection for spaced plant yield has led to genetic gains, realized from forage yield tests of sward plots in rye (Bruckner et al., 1991), Italian ryegrass (Fujimoto and Suzuki, 1975), and Pensacola bahiagrass (Burton, 1982). However, there is an equal or greater number of examples in which sward-plot forage yield could not be improved by selection on spaced plants, including smooth bromegrass (Carpenter and Casler, 1990); reed canarygrass, Phalaris arundinacea L. (Casler and Hovin, 1985; Surprenant et al., 1988); perennial ryegrass, Lolium perenne L. (Hayward and Vivero, 1984); and tall fescue, L. arundinaceum (Schreb.) Darbysh. (Rotili et al., 1976). Spaced plantings can be highly deceiving, because dry-matter yield of spaced plants has a higher heritability than forage yield of sward plots (Oliveira, 1992), but the frequent lack of genetic correlation between these two traits negates any apparent advantage of spaced plantings (Hayward and Vivero, 1984; Casler et al., 1996; Wilkins and Humphreys, 2003).

Van Dijk and Winkelhorst (1978) and van Dijk (1983) developed an ingenious method to solve two problems in forage grass breeding, the application of meaningful selection pressure to spaced plants within half-sib families and the need to measure performance or persistence of forage grasses in mixed swards with perennial legumes or other grass species. Using the spaced-plants-in-swards method, in which spaced plants of the target species are interspersed on a regular grid within a sward of a contrasting competitor species, these authors showed that AWF selection had 4 to 25% greater gains in forage yield of perennial ryegrass compared to HSF selection. These gains were considerably lower than the theoretical efficiency of AWF selection compared to HSF selection when one trait is the focus of selection both among and within families (Vogel and Pedersen, 1993). This suggests a lack of validity to the assumption of equal among-family and within-family phenotypic variances. In reality, phenotypic variance within families is likely to be much larger than phenotypic variance among families due to (i) three times more genetic variability within families compared to among families and (ii) replication of families in most experimental designs vs. typical lack of replication of individual plants. These factors are likely responsible for a greatly reduced narrow sense heritability within families compared to among families. The ryegrass results (van Dijk and Winkelhorst, 1978; van Dijk, 1983) and the results reported herein suggest that the true advantage of AWF over HSF selection is likely to be something less than that predicted by Vogel and Pedersen (1993).

It should also be recognized that the theoretical efficiency of AWF selection, as described herein, may be less than the theoretical efficiency of half-sib progeny-test (HSPT) selection in which parents are saved and used as the recombination unit, doubling the theoretical gain compared to HSF selection used in this study. As used in the current study, expected gain in forage yield (Y) from AWF selection ({Delta}GY) can be expressed as

Formula 3[3]
where k1 = the standardized selection differential among families, c = the parental control factor (0.5 for AWF), {sigma}2A = additive genetic variance, {sigma}PF = phenotypic variance among families, k2 = the standardized selection differential within families, rg = the genetic correlation between forage yield and the within-family selection criterion (X = survivorship), hX and hY = square root of narrow sense heritabilities for traits X and Y, {sigma}PW = phenotypic variance within families, and y = the number of years per cycle (Falconer and Mackay, 1996; Hallauer and Miranda, 1988). Saving parental genotypes gives an advantage to HSPT by increasing c from 0.5 to 1 (also resulting in k2 = 0). Therefore, the latter within-family term in Eq. [3] is key to the efficiency of AWF selection. Clearly, the success of AWF selection compared to HSPT selection would be favored by (i) a high genetic correlation between forage yield and the within-family selection criterion (X), (ii) a higher heritability for trait X, compared to forage yield, and (iii) k2 > k1, which may not be consistently true. The proportion of families selected in this study ranged from p1 = 0.060 to 0.104 for orchardgrass and 0.025 to 0.076 for smooth bromegrass and hybrid wheatgrass, compared to p2 (within families) = 0.011 for orchardgrass and 0.027 for smooth bromegrass and hybrid wheatgrass, values of p2 computed as the ratio of the number of 3-yr-old sod cores to the number of pure-live-seeds planted per square meter. One further advantage in favor of AWF is that HSPT will often require an extra two years per cycle, due to the need for an additional recombination event to generate families for testing in the next cycle, particularly for grasses that require vernalization, increasing y from 4 to 6 or from 5 to 7 for HSPT selection. Without any further theoretical framework or empirical comparisons, it is difficult to predict the relative advantage of AWF vs. HSPT selection methods.

The results of the current study demonstrate an alternative method to target and utilize selection pressure within half-sib families of forage grasses. Forage grasses are planted at seeding rates that result in intense competition almost immediately after seedling emergence. Competition results in high mortality rates during the establishment year (Charles, 1961), likely continuing in subsequent years as swards continue to evolve under stressful or fluctuating environmental conditions (Snaydon, 1978). Mortality results from natural selection for traits that relate to plant survival and persistence (Linhart and Grant, 1996), with differential adaptive responses occurring as rapidly as two years and over distances as short as 10 m (Snaydon, 1978). The consistency of observed selection gains across the three years of this study (lack of crossover population x year interactions) implies that these effects can be observed early in the life history of these grass stands and likely occurred rapidly in the original stands, possibly within the establishment year and/or the first production year.

Natural selection within swards of smooth bromegrass and hybrid wheatgrass, two rhizomatous grasses, has the potential to greatly improve the gains that can be made over HSF selection for forage yield. Plants that survived three years under hay management in seeded swards had greater breeding value for both ground cover and forage yield compared to random plants from selected families. Because improvements in ground cover occurred only as a result of AWF selection in these two species, it appears that natural selection of plants within these swards acted to improve family survivorship (on a plot basis) and the ability of surviving individuals to fill in open spaces left by plant and/or tiller mortality. The lack of response for all orchardgrass populations implies that the success of AWF was linked to the rhizomatous growth habit of smooth bromegrass and hybrid wheatgrass. Genetic marker studies are currently underway to quantify changes in the number and size of survivors in sward plots of these species.

Natural selection for survivorship under grazing or frequent defoliation leads to an increase in the frequency of plants that are shorter, more prostrate, and higher in tiller density than random plants from the original population (Breese, 1983; Casler et al., 1996; Falkner and Casler, 1998). Shorter or more prostrate plants are generally expected to have reduced forage yield potential (Casler, 2004; Casler and Brummer, 2005). Forage yield and ground cover in this study were uncorrelated (r = 0.12 to 0.27). Thus, selection pressures were more or less independent for these two traits, resulting in simultaneous, but not necessarily cause and effect, improvements. This may be a direct result of natural selection under relatively infrequent defoliation.

Because HSF selection had no effect on ground cover by itself, the improvements in ground cover were due to the effects of natural selection within three-year-old swards of smooth bromegrass and hybrid wheatgrass. There were two mechanisms for improvement of forage yield per se. In hybrid wheatgrass, the forage yield results were similar to those for ground cover in which within-family selection provided the only improvement in forage yield. This result indicated that surviving plants had higher breeding value for both ground cover and forage yield. In smooth bromegrass, both populations were improved by HSF selection, but to a lesser extent than for AWF selection (1.4 vs. 3.0%, on average). Thus, about half the gain for AWF was due to selection among families for forage yield per se and half due to selection within families for survivorship. This result suggests that three-year survivors in smooth bromegrass plots have greater forage yield potential than random plants from these populations. Phenotypic selection of surviving plants within Lolium swards has also led to concomitant increases in both survival and forage yield (Charles, 1972; Novy et al., 1995).

Wilkins and Humphreys (2003) reported gains in forage yield of nearly 10% per decade from a long-term commitment to intensive selection using HSF selection within an elite population. These gains are approximately double those observed from long-term breeding efforts in Europe, and have resulted in multiple cultivar releases to date (Wilkins and Humphreys, 2003). Despite their successes, Wilkins and Humphreys (2003) suggest that the number of families that can be evaluated and the number of plots that can be effectively harvested in one day are factors severely limiting the success of family selection methods. The results of the current study point out two potential solutions to this problem. First, incomplete block designs and/or spatial analyses can be used to adjust for spatial variation caused by numerous factors, including soil variation and day to day variation caused by an inability to harvest all plots on one day. The latter problem can be ameliorated by any number of different blocking designs. Second, reductions in plot size can improve throughput without needlessly sacrificing precision. Lin and Binns (1986) provide a mechanism to predict the effect of changes in plot size on statistical precision. Clearly, in the current study, the use of 1.35-m2 plots to evaluate half-sib families was sufficient to accurately estimate forage yields in three different perennial grasses, providing sufficient heritability to support realized selection gains. For plots of this size, approximately 100 plots h–1 were harvested with a three-person crew, allowing each HSF trial of up to 840 plots to be harvested in one day.


   ACKNOWLEDGMENTS
 
I thank Charlie Brummer, University of Georgia, for many stimulating conversations and suggestions about breeding and selection methodology, particularly as related to improvement of forage yield per se.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.

Received for publication May 11, 2007.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 




This article has been cited by other articles:


Home page
 Crop Sci.Home page
M. D. Casler and E. C. Brummer
Theoretical Expected Genetic Gains for Among-and-Within-Family Selection Methods in Perennial Forage Crops
Crop Sci., May 1, 2008; 48(3): 890 - 902.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Casler, M. D.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Casler, M. D.
Agricola
Right arrow Articles by Casler, M. D.
Related Collections
Right arrow Other Forage Crops
Right arrow Crop Genetics

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome