crop breeding, genetics & cytology - PubAg - USDA

CROP BREEDING, GENETICS & CYTOLOGY Heterosis in a Broad Range of Alfalfa Germplasm Heathcliffe Riday* and E. Charles Brummer

Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved.

ABSTRACT

terial (Riday and Brummer, 2002a, 2002b; Riday et al., 2002). Few improved falcata breeding populations exist, making the implementation of a sativa–falcata semihybrid breeding system for cultivar development difficult at this time. Thus, a program to select improved falcata populations appears to have merit. Unfortunately, little is known about which falcata germplasm shows the best heterosis with elite sativa breeding material and could serve as the basis of a falcata population improvement program. A further problem is defining the purity of sativa or falcata germplasm; elite midwestern sativa germplasm obviously contains falcata alleles from repeated introgression events (Barnes et al., 1977). This study takes a practical approach by assaying falcata populations (defined solely on the basis of flower color) for the production of heterosis with elite midwestern sativa germplasm, which we readily admit represents an amalgamation of Eurasian sativa germplasm together with falcata introductions. The actual amount of falcata germplasm in modern sativa cultivars is difficult to determine given the repeated rounds of selection and extensive mixing of germplasm that has occurred over the past century. Falcata is yellow flowered and, compared with sativa, tends to be more winterhardy, to have more prostrate growth, to regrow slower, and to yield less in the late summer and early autumn (Lesins and Lesins, 1979; Riday and Brummer, 2002b). Geographically, falcata is distributed in the colder areas of Russia, Mongolia, Scandinavia, and China, while sativa grows naturally in the Middle East, southern Europe and northern Africa (Hansen, 1909; Lesins and Lesins, 1979). Wild falcata and sativa germplasm overlap in some European regions and in Central Asia, where their natural hybrid, M. sativa subsp. varia (Martyn) Arcang., is found (Hansen, 1909; Lesins and Lesins, 1979). Currently, 470 falcata accessions, including both diploid (2n ⫽ 2x ⫽ 16) and tetraploids (2n ⫽ 4x ⫽ 32), are listed in the USDA National Plant Germplasm System’s Germplasm Resource Information Network (USDA, GRIN, 2003). In addition to GRIN accessions, various germplasm centers throughout the world, as well as a few semiimproved North American populations, have been collected. We hypothesize (i) that all falcata germplasm does not express heterosis with elite sativa germplasm equally well, but (ii) that some combination of ecogeographic, molecular genetic, and/or morphological characteristics of falcata germplasm are predictive of superior sativa– falcata hybrids. The objective of this study was to evaluate falcata genotypes from a broad range of wild and semi-improved populations for performance per se and

Crosses between Medicago sativa L. subsp. sativa and subsp. falcata (L.) Arcang. produce progeny exhibiting heterosis for biomass yield. The purpose of this study was to determine which subsp. falcata germplasm produced the best hybrids in testcrosses with elite subsp. sativa material to guide future breeding efforts. Over 100 falcata genotypes from 40 populations were test crossed to four elite sativa populations. Testcross progeny and parental clones were grown for 2 yr in two locations and harvested three times per year to determine biomass yield. A broad range of testcross performance was observed, with mean heterosis values approximately zero. The highest yielding sativa–falcata hybrids were derived from European falcata germplasm. North American semi-improved falcata germplasm performed well in hybrid testcrosses. Preselection of parental falcata genotypes for autumn growth was associated with higher yielding testcross progeny. Positive heterosis was seen during the first harvest, but negative heterosis was often observed during second and, to a smaller extent, third harvests. Superior sativa–falcata hybrids were observed that showed good biomass yield and heterosis during all three harvests. Parental yield was least predictive of hybrid progeny yield during first harvest (h2 ⫽ 0.12). Heritability increased during second and third harvest to 0.31 and 0.33, respectively. Expected genetic gain per selection cycle is greater from progeny testing compared with simple recurrent phenotypic selection.

A

lfalfa represents about 2.5% of the total agricultural hectarage and 6000 million dollars of annual production in the USA (USDA, NASS, 2003). Primary traits of interest in alfalfa are yield, nutritive value, disease resistance, persistence, and winter hardiness. Current alfalfa breeding methods are almost exclusively based on recurrent phenotypic selection, which involves intercrossing selected parents to produce synthetic varieties (Hill et al., 1988). Hybrid or semihybrid cultivars could be used to express hybrid vigor in farmer’s fields (Brummer, 1999). Effective implementation of any hybrid breeding system requires the improvement of at least two independent and complementary populations, which combine well to produce heterosis. M. sativa subsp. falcata (hereafter “falcata”) has been identified as a subspecies that shows heterosis in crosses with elite M. sativa subsp. sativa (hereafter “sativa”) breeding maH. Riday, USDA-ARS, U.S. Dairy Forage Research Center, Madison, WI 53706; E.C. Brummer, Raymond F. Baker Center for Plant Breeding, Dep. of Agronomy, Iowa State Univ., Ames, IA 50011. This journal paper of the Iowa Agric. Home Econ. Exp. Stn., Ames, IA, Project No. 6631, was supported by Hatch Act and State of Iowa funds. Received 27 May 2003. *Corresponding author (riday@wisc. edu). Published in Crop Sci. 45:8–17 (2005). © Crop Science Society of America 677 S. Segoe Rd., Madison, WI 53711 USA

8

9

RIDAY & BRUMMER: FALCATA GERMPLASM HETEROSIS

in testcrosses with elite sativa tester genotypes. Our goal was to characterize the distribution of sativa–falcata hybrids for biomass yield heterosis on a whole year and on a harvest basis to guide the development of improved falcata populations.


MATERIALS AND METHODS Plant Materials A total of 125 genotypes was used as parents in this experiment (Table 1). Sixteen were elite sativa genotypes from four populations, designated as testers: five genotypes from Pioneer Hi-bred International (Des Moines, IA), three genotypes from Forage Genetics (West Salem, WI), three elite genotypes derived from Hungarian germplasm also from Forage Genetics, and five ‘Innovator ⫹Z’ genotypes derived from one generation of inbreeding. Three wild sativa genotypes from different populations were also included. The remaining 106 genotypes were falcata, derived from 37 wild or semi-improved populations from throughout the native range of falcata. A number of populations included genotypes with variegated flower color, suggesting past sativa-falcata introgression, even though the population was listed as falcata and in one case, sativa, in the GRIN system (Table 1). The genotypes used for crossing were yellow flowered, however, and unless specifically noted otherwise, genotypes from variegated populations were designated as falcata. In several populations during the autumn of 1999, vigorous genotypes were visually selected from a space planted germplasm evaluation trial at Ames, IA (Brummer et al., 1997). The trial was in its second postestablishment year and had been harvested twice during 1999. At the time of selection, most plants were dormant with minimal regrowth, so our selections was for genotypes with less dormancy than the overall populations. The most vigorous

two to three individuals from within in the given population were selected. All 125 genotypes were crossed to the four tester populations in the greenhouse during the autumn–winter 1999 to 2000, producing a total of 500 cross entries, 76 of which were sativa by tester and 424 falcata by tester. Single plant-to-plant crosses were made between the 125 parental genotypes and individual plants in the tester populations; florets were not emasculated. Usually the tester was used as the female, because of its higher seed production. Several falcata genotypes exhibited low pollen production, in which case, they were used as the female. For most testcross entries, no self-fertilization was observed on the basis of flower color segregation, although a few entries expressed up to about 10% self-fertilization. The self-fertilization always occurred when the falcata parent used as male was agronomically weak. For field measurements other than biomass yield, obvious self-fertilized individuals were avoided when possible. In Spring 2000, seed from the 500 cross entries, the three wild sativa and 37 wild or semi-improved falcata populations from which the genotypes used for crossing were derived, and two check cultivars (Vernal and 5454) were planted in the greenhouse. Stem cuttings of the 125 parental genotypes were made at the same time. A total of 667 entries was included in this experiment (500 crosses, 125 parental clones, 40 populations, and two checks).

Experimental Design Seedlings and cuttings were hand transplanted at the Agronomy and Agricultural Engineering Research Farm west of Ames, IA, in a Nicollet loam soil (fine-loamy, mixed, superactive, mesic Aquic Hapludolls) on 1 Aug. 2000 and at the Northeast Research Farm south of Nashua, IA, in a Readlyn loam (fine-loamy, mixed, mesic Aquic Hapludolls) on 8 Aug.

Table 1. Origin of the 44 populations evaluated and number of genotypes selected from 44 populations used in this study. Population Wild germplasm 1. PI 631591 2. PI 251836§ 3. PI 631579¶ 4. PI 253451¶ 5. PI 631796 6. 5291/88†† 7. 5299/88¶†† 8. PI 631857 9. PI 494661¶# 10. PI 494658§ 11. PI 502441# 12. PI 384890§ 13. PI 538985# 14. PI 440539¶ 15. PI 631608 Improved Germplasm 30. PI 631620¶ 31. PI 631597# 32. PI 631596¶ 33. PI 502453# 34. PI 631797¶ 35. PI 631799¶ Elite Sativa Germplasm 41. Pioneer Hi-Bred 42. Forage Genetics

Origin† 9ⴗ38ⴕ, 11ⴗ19ⴕ, 13ⴗ29ⴕ, 16ⴗ38ⴕ, 16ⴗ38ⴕ, 17ⴗ52ⴕ, 17ⴗ55ⴕ, 18ⴗ37ⴕ, 23ⴗ25ⴕ, 26ⴗ54ⴕ, 43ⴗ53ⴕ, 55ⴗ01ⴕ, 56ⴗ19ⴕ, 76ⴗ57ⴕ, 84ⴗ00ⴕ,

No.‡

Population

Origin

No.

47ⴗ15ⴕ 44ⴗ10ⴕ 45ⴗ52ⴕ 46ⴗ22ⴕ 49ⴗ12ⴕ 47ⴗ12ⴕ 47ⴗ13ⴕ 57ⴗ17ⴕ 46ⴗ55ⴕ 46ⴗ34ⴕ 46ⴗ11ⴕ 36ⴗ25ⴕ 50ⴗ50ⴕ 43ⴗ15ⴕ 29ⴗ00ⴕ

2 1 3 1 5 1 1 4 5 1 5 1 5 1 3

16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.

PI 631612 PI 631601 PI 538993# PI 577561 PI 499661 PI 631645 PI 631639 PI 499548 W6 16608 PI 214218¶ PI 631806# PI 314092¶# PI 631811‡‡ PI 573175#

84ⴗ00ⴕ, 29ⴗ00ⴕ 84ⴗ38ⴕ, 44ⴗ09ⴕ 85ⴗ53ⴕ, 50ⴗ50ⴕ 85ⴗ59ⴕ, 52ⴗ00ⴕ 88ⴗ31ⴕ, 44ⴗ05ⴕ 102ⴗ34ⴕ, 47ⴗ23ⴕ 112ⴗ00ⴕ, 47ⴗ30ⴕ 116ⴗ02ⴕ, 43ⴗ58ⴕ 118ⴗ06ⴕ, 48ⴗ00ⴕ Denmark North Russia South Russia S. Kazakhstan Pro. China

3 4 3 4 2 5 3 1 1 1 4 4 1 3

77ⴗ00ⴕ, 34ⴗ00ⴕ 45ⴗ55ⴕ, 51ⴗ30ⴕ 91ⴗ48ⴕ, 53ⴗ27ⴕ Russia Russia Latvia

2 5 2 4 3 2

36. 37. 38. 39. 40.

IA-3018# Lodgeland¶ SD 201‡‡ PI560333 #§§ PI 468015¶

Ames, IA Brookings, SD Brookings, SD Madison, WI ⫺107ⴗ48ⴕ, 51ⴗ25ⴕ

5 2 1 5 1

Johnston, IA West Salem, WI

5 3

43. Hungarian 44. ABI Alfalfa

West Salem, WI Napier, IA

3 5

† Longitude, Latitude. ‡ Number of genotypes per population used for crossing. § Wild sativa population. ¶ Populations that contain some genotypes with variegated flowers. # Populations from which genotypes were both visually and randomly selected. †† Accession from Institute of Agrobotany, Hungary. ‡‡ Colchicine doubled diploid. §§ WISFAL (Bingham, 1993).


10

CROP SCIENCE, VOL. 45, JANUARY–FEBRUARY 2005

2000. At each location, the field experiment was arranged in an augmented plot design consisting of 20 incomplete blocks of 40 plots each, for a total of 800 plots. All 665 entries were present once, except for 55 randomly selected entries that were replicated twice. These 720 entries were then distributed among the incomplete blocks. In addition, each incomplete block contained the two check cultivars and was completed by the addition of replications of two or three entries (randomly chosen out of the 720 total) not already present in that incomplete block. Each plot consisted of sixteen plants that were planted in a two by eight plant grid, with plants separated by 30 cm within a plot and plots separated by 75 cm on all sides. Harvests for biomass yield were taken on 3 June 2001, 24 July 2001, 11 Sep 2001, 30 May 2002, 13 July 2002, and 30 Aug. 2002 at Ames and on 10 June 2001, 18 July 2001, 30 Aug. 2001, 13 June 2002, 18 July 2002, and 12 Sep. 2002 at Nashua. Plant counts were taken on each plot concurrent with every harvest. Whole plots were harvested and weighed wet with a flail harvester during each harvest. At each harvest– location combination, several subsamples were taken from throughout the experiment, but not on each individual plot, weighed wet, dried for 5 d at 60⬚C, and then reweighed to determine an average dry matter percentage for that particular harvest. Plot wet weights were adjusted for plant counts and average dry matter percentage to determine biomass yield on a per plant dry matter basis.

Data Analysis Entry and Class Mean and Variance Analysis Analyses were conducted on dry matter yield at each individual harvest and on total yearly dry matter yield. Entry least squares means (LSmeans) on a location ⫻ year ⫻ harvest basis were calculated by adjusting plot values for incomplete blocks by the MIXED procedure of the SAS statistical software package (Littell et al., 1996). Homogeneous variances were assumed among different entries. The 500 cross entry means and 125 clonal entry means on a location ⫻ year ⫻ harvest basis were combined into a data set for further analysis. Total yearly biomass yield was combined into a data set on a location ⫻ year basis. The entries were grouped into four classes: the 500 cross entries were categorized as being either sativa ⫻ sativa crosses (SSC ⫽ 76 entries) or sativa ⫻ falcata crosses (SFC ⫽ 424 entries); and the 125 clonal entries were categorized as being either sativa clonal (SC ⫽ 19 entries) or falcata clonal (FC ⫽ 106 entries). For experiment-wide mean and yearly mean analysis on a total yearly or individual harvest basis, entries, years, and harvests were considered to be fixed effects and locations were considered random. For location mean analysis on a total yearly or individual harvest basis, a completely fixed model was used. Environment (i.e., location, year, location ⫻ year, and harvest) ⫻ entry variances for each category (i.e., SSC, SFC, SC, FC) on a total yearly and individual harvest basis were calculated with environmental variables being fixed and entries being random. Environment ⫻ entry variances were compared among entry class by a test of equality of two variances (Snedecor and Cochran, 1967). Because no differences were found among the four testers in crosses with the 125 genotypes and no tester by environment interaction was noted (data not shown), the 500 cross entries were reduced to 125 by calculating a single genotype by average tester progeny halfsib value. For each of the four tester populations (i.e., Pioneer Hi-bred, Forage Genetics, Hungarian, Innovator ⫹Z), the population mean was estimated as

the average of the within population genotypes (Table 1) testcrossed to the whole population. The four tester population means were averaged to estimate the average tester mean.

Heterosis Analysis The 125 clonal means and the 500 cross entry means were used to calculate all heterosis measures. Heterosis measures were calculated on a total yearly and individual harvest yield basis. Midparent heterosis was calculated as the deviation of a genotype’s testcross progeny performance from the average of the genotype’s clonal and the tester’s performance (i.e., the “midparent” value). High-parent heterosis was calculated as the deviation of a genotype’s testcross performance from the higher of the genotype’s clonal or the tester’s performance. Because both of these heterosis measures were dependent on parental performance, we also calculated a third heterosis measure as the residual values from the regression of progeny yield on midparent yield. The midparent-offspring regression residuals for each of the 125 genotypes tested were calculated for each of the four testers for each year ⫻ location combination on a yearly total and individual harvest biomass yield basis. For heterosis analyses, tested genotypes, years, testers, and harvests were considered fixed effects and location effects were considered random. Heritability Estimates To ascertain expected genetic gain per selection cycle, we calculated narrow sense heritability estimates. Because we only had a few genotypes (two to five) for each of the 30 populations, we estimated narrow sense heritability on the basis of the mean population genetic variance component estimates from the 30 populations (i.e., random genotypes nested in fixed populations). The among halfsib family variance, adjusting for fixed populations, was estimated as ␴ 2hs ⫽ 1/4␴ 2A ⫹ 1/36␴ 2D, where ␴ 2hs ⫽ among halfsib family variance, ␴ 2A ⫽ additive genetic variance, and ␴ 2D ⫽ dominance genetic variance (Levings and Dudley, 1963). Testers, locations, and years were considered fixed effects; genotypes (i.e., ␴ 2hs) and genotype interactions (i.e., ␴ 2hsT, ␴ 2hsL, ␴ 2hsY, ␴ 2hsTL, ␴ 2hsTY, ␴ 2hsLY, and ␴ 2hsTLY, where T ⫽ testers, L ⫽ locations, and Y ⫽ years) were considered as random effects. All random effects were estimated directly by PROC MIXED. A variance among parental clones, adjusting for fixed populations was estimated as ␴ 2c ⫽ ␴ 2G, where ␴ 2c ⫽ among clones and ␴ 2G ⫽ total genetic variance (Levings and Dudley, 1963). Locations and years were fixed effects and genotypes (i.e., ␴ 2c) and genotype interactions (i.e., ␴ 2cL, ␴ 2cY, and ␴ 2cLY) were random effects. Finally, parent-offspring covariances (␴po ⫽ 1/2␴ 2A ⫹ 1/6␴ 2G) were estimated, based on testcross progeny means and genotypic clonal performance, adjusting for fixed populations (Levings and Dudley, 1963). An analysis of covariance (Nguyen and Sleper, 1983) was accomplished with locations and years as fixed effects and genotypes (i.e., ␴po) and genotype interactions (i.e., ␴poL, ␴poY, and ␴poLY) as random effects using an adapted PROC MIXED program described by Zamudio and Wolfinger (2002). On the basis of variances and covariances, ␴ 2A and ␴ 2G were estimated. Narrow sense heritability on a plot-mean basis was calculated as

h2 ⫽ ␴ 2A

冒冪

␴ 2GLY ␴ 2GL ␴ 2GY ⫹ ⫹ ⫹ ␴ 2G. ly l y

All estimates were done assuming no epistatic variance, no double reduction, random mating within populations, linkage

11



equilibrium, and no linkage (Cockerham, 1956; Kempthorne, 1969). In cases where ␴ 2D was negative, this variance was assumed to be zero and the weighted average of ␴ 2hs and 2␴ 2po was used as the ␴ 2A estimate (Dudley et al., 1969). The same calculations were conducted for genotype interactions. Heritability standard errors were calculated using the asymptotic covariance matrix from PROC MIXED and approximation formulas described in Hallauer and Miranda (1988). Estimated genetic gain per cycle (Fehr, 1991) was calculated for clonal selection, on the basis of gridded mass selection, as:

⌬Gc ⫽ kc␴ 2A/√␴ 2w ⫹ ␴ 2 ⫹ ␴ 2GY ⫹ ␴ 2GL ⫹ ␴ 2G, with c (parental control) ⫽ 1, k (selection intensity of 20%) ⫽ 1.40, ␴ 2 ⫽ between plot variance, ␴ 2w ⫽ within plot variance, and for halfsib progeny testing as:

冒冪

2 2 2 ⌬Gc ⫽ kc␴ 2A 4 ␴ ALY ⫹ ␴ AL ⫹ ␴ AY ⫹ 1 ␴ 2A, 4ly 4l 4y 4

with c ⫽ 2, k(20%) ⫽ 1.40. No within the 16 clonal-plant-plot (␴ 2w) variance estimates were measured; therefore, we reported the maximum range of gain per cycle by assuming that either ␴2 ⫽ 0 or ␴ 2w ⫽ 0 in the following equation:

␴ 2cLY ⫽ (␴ 2w/n) ⫹ ␴ 2 (Fehr, 1991). Variance Proportions and Population Mean Comparisons The proportions of biomass yield and heterosis variance explained by within and between population variation were calculated by PROC MIXED. A Wald’s test, in PROC MIXED, was used to test whether variances deviated from zero. PROC MIXED was used to calculate population mean testcross biomass yield and biomass yield heterosis on a total yearly and harvest basis using experiment wide genotypic testcross and heterosis values. Because of unequal number of genotypes within populations, approximate least significant differences were estimated by averaging all pairwise standard error values generated from the DIFF option of the LSMEANS statement of PROC MIXED and multiplying this value by the appropriate t value.

RESULTS AND DISCUSSION Class Mean Comparisons Experiment-wide comparisons for total yearly biomass yield showed that FC yielded less than SSC, SFC, and SC, all of which yielded the same (Table 2). Similar patterns were observed across locations during 2001 and 2002, as well as across years at Ames and Nashua. An

exception was noted during 2001, when SFC outyielded SC (Table 2). Our earlier experiment, which considered a restricted set of generally desirable falcata genotypes, had shown that SFC performance was similar or superior to SSC and both were better than falcata ⫻ falcata crosses (Riday and Brummer, 2002a). The relationship among classes was more variable on an individual harvest basis. At first harvest, SFC produced more biomass yield than any other class, SSC and SC were intermediate, and FC produced poorly (Table 2). Second harvest biomass yield was greater for SSC and SC than SFC, but all were still superior to FC. Sativa clonal, SSC, SFC biomass yields were similar to each other and all were greater than FC during third harvest. Riday and Brummer (2002a) observed a similar pattern of superior SFC compared with SSC performance during the first harvest, but a reversal in second harvest. These results confirm the sativa–falcata heterotic pattern that we have proposed previously (Brummer, 1999; Riday and Brummer, 2002a). Under an additive genetic model, no heterosis would be expressed and SFC should fall midway between the SC and FC class means. A major focus of this experiment was to examine a variety of falcata germplasm to determine which would be most useful in producing high yielding sativa–falcata hybrids. Sativa ⫻ falcata crosses had greater total yearly biomass yield variance, as well as greater individual harvest variance, during each harvest, than SSC. For total yearly yield, a proportion of the SFC distribution extended beyond the SSC distribution, showing a number of SFC outyielding the best SSC (Fig. 1). During first harvest, biomass yield distributions show a proportion of SFC outperforming the best SSC (Fig. 1), demonstrating the yield advantage of SFC compared with SSC during the first harvest. For the second harvest, some SFC yielded at the same levels as the best SSC, indicating the existence of falcata germplasm that produces SFC yields that match the best SSC. Both SSC and SFC have similar third harvest biomass yield on the upper end of their distributions (Fig. 1), but a proportion of SFC underperform the worst SSC during the second and third harvest (Fig. 1). Clearly, there exists falcata germplasm that shows no heterotic advantage in crosses with elite sativa germplasm and in some cases presents a clear disadvantage.

Environmental Interaction We observed entry ⫻ environment interactions for the three environmental factors of location, years, and

Table 2. Biomass yield means comparisons for sativa ⫻ sativa crosses, sativa ⫻ falcate crosses, sativa clones, and falcata clones on an experiment wide, yearly, location, and harvest mean basis. Entry type

Experiment wide

Year 2001

Location 2002

Ames

Harvests Nashua

1st

2nd

3rd

210a 215a 212a 149b

90b 97a 88b 70c

75a 70b 76a 56c

55a 52a 50a 40b

plant⫺1

sativa ⫻ sativa sativa ⫻ falcata sativa clones falcata clones

219a 219a 213a 166b

† Indicated differences significant at P ⫽ 0.05.

157ab† 160a 138bc 119c

280a 279a 289a 214b

g 229a 223a 214a 184b


12


Across environmental levels, the falcata clones and crosses often had greater interaction variances than the two sativa categories (Table 3). No differences were noted between SC and FC entry ⫻ environment interaction variances for first harvest (Table 3). Entry ⫻ location interactions were only observed for third harvest. Differences between SSC and SFC entry ⫻ environment variances were seen for all situations except entry ⫻ location interactions for total yearly yield and for second harvest. No entry ⫻ year interaction differences were noted for second harvest.

Heterosis

Fig. 1. Distribution of alfalfa intra- and intersubspecific crosses for total biomass yield and for biomass yield of three individual harvests based on data across 2 yr and two locations in Iowa, USA.

harvests. Many of the entry ⫻ location interactions were due to rank changes between locations, but interactions with harvest and year were primarily magnitude changes. No rank change interactions of entry class (i.e., SC, FC, SSC, and SFC) with location, year, and harvest were observed except for the case of SFC being superior to SSC during the first harvest but worse during the second harvest (Table 2). Although few entry ⫻ environment interactions were observed, the presence of rank changes of entries across locations, suggests that testing in multiple locations is necessary to select for environmentally stable yield. We used a test of equal variance to compare entry ⫻ environment variances of SSC with SFC and of SC with FC. A lower variance would suggest that that subspecies has more environmental stability than the other. Although we did not directly compare crosses and clones, the two clonal categories appear to have greater environmental interaction variances than the two testcross categories (Table 3).

Heterosis, the converse of inbreeding depression (Falconer and MacKay, 1996), is defined as the superior performance of progeny compared with some measure of parental performance (Hallauer and Miranda, 1988). The initial intent of this study was to use high parent heterosis (the superiority of progeny relative to the best performing parent) because from a breeding standpoint, this is the most important type of heterosis. High parent heterosis showed an inverse quadratic relationship (R2 ⫽ 0.37) with testcross performance (Fig. 2). Falcata parents that had performance slightly below tester yields tended to be more likely to produce progeny exhibiting high parent heterosis than those either substantially below or above the tester. Second, we correlated midparent heterosis with testcross mean yield (Fig. 2). This correlation is strongly negative (r 2 ⫽ 0.57), suggesting that midparent heterosis is strongest when one of the parents is very weak; in this case, very poor performing falcata produce high heterosis when crossed with elite testers. In summary, heterosis would be expected to decline as midparent performance increased, simply because the absolute amount of biomass increase in the progeny would need to become increasingly large as parental performance improved. Thus, more interesting to us than the relative heterosis is the residual value of progeny performance from the midparent-offspring yield regression. In other words, examining the relationship between progeny performance per se and the midparental performance level would be more instructive. Midparent-offspring regression is usually used in random mating populations to estimate heritability, where h2 ⫽ slope, with an expectation that the slope regresses toward the population mean (i.e., the slope crosses the 1:1 xy-line at the population mean) (Galton, 1888). The residual value from the midparent-offspring regression is the measure of heterosis that we use in the discussion below (Fig. 2). In a sense, this measure is similar to specific combining ability (general combining ability being the regression line itself).

Heritability In this study, we calculated narrow sense heritability on a plot mean basis (h2) across all populations on the basis of average within population genetic variances. The low heritability for total yearly biomass yield (h2 ⫽ 0.31) indicated that parental performance has a limited

13


Table 3. Variances for alfalfa crosses and clones interacting with years, locations, and harvests based on data collected over 2 yr and two locations in Iowa, USA. Harvest Entry Interaction Variance

Experiment wide sativa†

falcata

1st sativa

2nd falcata g

3rd

sativa

falcata

sativa

falcata

92 121 101

224* 247NS 417**

36 53 92

218*** 138* 189NS

62 99 82

88NS 117NS 131*

23 36 36

38** 56* 63**

plant⫺1


Clones ␴2entry⫻year ␴2entry⫻location ␴2entry⫻year⫻location ␴2entry⫻harvest

675 967 1685 120

1933* 824NS 1586NS 172NS

349 267 951

␴2entry⫻year ␴2entry⫻location ␴2entry⫻year⫻location ␴2entry⫻harvest

262 449 371 39

485* 648NS 663** 57**

93 87 132

746NS 255NS 664NS Crosses 159** 170*** 238**

Sativa different from falcata. * Significantly different from zero at the 0.05 probability level. ** Significantly different from zero at the 0.0 probability level. *** Significantly different from zero at the 0.001 probability level. NS not significant. † Sativa refers to either sativa clones or sativa by sativa crosses; similarly, falcata refers to either the clone per se or the falcata ⫻ sativa cross.

influence on progeny performance (Table 4). Thus, progeny testing potential parental genotypes before selection is necessary. First harvest yield heritability was the lowest of the three harvests (h2 ⫽ 0.12), but second (h2 ⫽ 0.27) and third (h2 ⫽ 0.34) harvests were somewhat better. First harvest biomass yield accounted for 41% of the biomass yield per year in this population, yet parental performance during the first harvest has almost no predictive value of testcross performance. Therefore, selection based on parental performance could be conducted during second preferably third harvests, although this would still not be a strong predictor of hybrid performance. Remarkably few studies have estimated broad or narrow sense heritability for biomass yield in alfalfa. Two studies have estimated the ratio of additive to genetic variance for total yearly biomass yield, both using parent-offspring regression and polycross families, as 0.37 for ‘Cherokee’ (Dudley et al., 1969) and 0.30 for ‘Ranger’ (Kehr and Gardner, 1960). These values compare favorably to the ratio we calculated (0.45) from the parentoffspring regression, despite the fact that our estimate was based on the average across a number of small populations (Table 4). Like Dudley et al. (1969), we observed negative ␴ 2D estimates because 4␴ 2hs ⬎ 2␴po. To our knowledge, this is the first report of the heritability of biomass yield at different harvests during the year. We calculated expected genetic gain per selection cycle to determine how much of an advantage to expect from progeny testing versus simple recurrent phenotypic selection (Table 4). For total yield and individual harvest yields, selecting parental clones for intercrossing based on halfsib progeny testcrosses increased genetic gains two to nine fold over simple phenotypic recurrent selection depending on the mount of assumed within plot variation between clones. Of the three types of “replications” (i.e., locations, years, and tester populations) tester ⫻ genotype and higher order interactions

containing both tester and genotype, had the greatest variances (data not shown). This indicates that breeders will achieve greater genetic gains by using an increased number of tester populations over increasing the number of progeny test environments (i.e., years and locations).

Progeny Biomass and Heterosis Comparisons Because progeny testing is difficult and resource intensive, we wanted to determine if we could identify falcata germplasm that had a higher incidence of producing superior sativa–falcata hybrids. Although the NPGS has 470 falcata accessions listed in GRIN, many of the accessions or populations contain varying proportions of genotypes with flower colors ranging from cream or greenish-yellow to blues, browns, and dark greens, suggesting they should be reclassified as M. sativa subsp. varia (or in some cases, subsp. sativa). Of the 37 falcata populations used in this study, 13 had incidence of variegated flowers, although we preferentially selected genotypes with yellow or near yellow flowers from these sources for crosses. If the sativafalcata heterotic pattern were predicated on underlying genetic distinctions between the falcata and sativa subspecies, then variegated populations, representing natural hybrids between sativa and falcata, should show less heterosis. Unexpectedly, the sativa-variegated crosses (SVC) had higher total yearly yield than SFC (in this analysis, genotypes derived from putatively variegated populations were not included in the SFC mean). During first harvest, SVC and SFC produced equivalent yield, and both exceeded the SSC. The SVC were intermediate to SSC and SFC during second harvest but were equivalent to SSC and greater than SFC during the third harvest. In terms of heterosis (based on the residual of the midparent-offspring regression), SVC and SFC were equiv-

14


Table 4. Ratio of additive to total genetic variance, narrow sense heritabilities on a plot mean basis (h2), and expected genetic gains per selection cycle (Gc) using simple phenotypic recurrent selection (PRS) and halfsib progeny testing recurrent selection based on halfsib progeny performance (HSRS), for total and harvest mean biomass yield in alfalfa. G c§

Biomass yield

␴ /␴ 2 A

2 G

h

2


Ratio Year total Harvest

1st 2nd 3rd

0.44† 0.27 0.52 0.63

0.30‡ 0.13 0.27 0.34

PRS

HSRS

g plant⫺1 5–15¶ 3–4 1–4 2–5

29 7 9 10

† SE about 0.15 for total, 1st, 2nd, and 3rd harvest. ‡ SE about 0.07 for total, 1st, 2nd, and 3rd harvest. § Selection intensity (k) ⫽ 20%. 2 ⫽ ¶ Max. range; min. assumes ␴ 2 ⫽ 0, max. assumes ␴ w2 ⫽ 0, where ␴ GLY (␴ 2w /n) ⫹ ␴ 2.

Fig. 2. Three measures of heterosis in the progeny of diverse alfalfa germplasm crossed to elite genotypes based on data across 2 yr and two locations in Iowa, USA versus the total yearly biomass yield of the average performance of the parents involved in each cross. In all three graphs, the vertical line represents the average biomass yield of the elite tester populations. (i) High-parent heterosis (R2 ⫽ 0.37). The horizontal line represents zero heterosis. (ii) Midparent heterosis (R2 ⫽ 0.57). The horizontal line represents zero heterosis. (iii) Halfsib testcross progeny yield (R2 ⫽ 0.20). The diagonal line represents 1:1 ratio between axes.

alent for total yearly heterosis and first harvest heterosis (Table 5). The SVC were the only cross group that had consistently positive heterosis throughout the year. Thirteen of the 36 genotypes we selected from variegated populations had variegated flowers. We compared variegated genotypes with yellow flowered genotypes for populations that had both types of genotypes. No differences were detected between flower color types for biomass yield or for biomass yield heterosis during any harvest (Table 5). However, most of the variegated populations were also improved populations, so the superior variegated population performance could be the result of human selection. Improved populations, including both variegated and falcata, had superior total yearly yield and third harvest yield and third harvest heterosis compared with wild populations (Table 5). Improved germplasm appears to have better autumn performance in crosses with sativa than does wild germplasm. The superior performance of the improved germplasm mirrors the variegated population performance per se. Variegated germplasm may also have superior performance because it might contain more allelic variability than pure falcata germplasm given that varia represents the natural hybrid between sativa and falcata. The additional variation from the sativa parents of the varia accessions may be distinct from that present in the elite sativa material we used. If this were the case, our varia ⫻ sativa hybrids would have more allelic variation than falcata ⫻ sativa hybrids, and consequently, would be expected to yield more on the basis of the theory of maximum heterozygosity (Bingham et al., 1994). To isolate a possible effect due to the potential hybrid ancestry of accessions with variegated flowers, we compared variegated and falcata performance from genotypes taken only from wild germplasm populations. In this comparison, the variegated populations had superior biomass yield performance for total yearly yield and second and third harvests (Table 5), but no differences were seen for heterosis (possibly because of small sample sizes). This latter comparison would seem to suggest that although variegated germplasm offers no heterotic advantage, they offer superior yield for all harvests except the first. Visual observations in the field led us to suspect that

15


Table 5. Testcross mean comparisons of alfalfa genotypes grouped into different classes for yearly and individual harvest biomass yield and heterosis based on data measured across two Iowa locations in 2001 and 2002. Testcross yield


Testcross Grouping

No.†

Testcross heterosis Harvest

Year total

1st

2nd

3rd

Harvest

Year total

1st

2nd

3rd

⫺7.4b ⫺0.1a 2.0a

0.7a 1.1a ⫺0.8a

0.2a 0.7a ⫺0.4a

sativa populations Variegated populations falcata populations

19 36 70

219ab† 224a 217b

90b 96a 97a

75a 73a 69b

g plant⫺1 55a ⫺7.0b 55a 1.0a 51b 1.3a

Variegated genotypes Yellow genotypes

13 23

226a 222a

97a 96a

73a 72a

56a 54a

2.4a 0.0a

0.9a 0.3a

1.4a 0.4a

1.4a ⫺0.5a

“Improved” populations Wild populations

32 74

224a 217b

98a 96a

71a 70a

55a 51b

1.4a 1.1a

1.2a 1.3a

⫺0.5a 0.0a

1.6a ⫺0.8b

Wild variegated populations Wild falcata populations

19 55

224a 215b

96a 96a

73a 69b

56a 50b

1.5a 1.0a

⫺0.1a 1.7a

1.6a ⫺0.5a

1.2a ⫺0.6a

European populations Asian populations

40 34

229a 204b

100a 91b

74a 65b

55a 47b

9.2a ⫺8.8b

4.6a ⫺2.7b

3.3a ⫺4.0b

1.4a ⫺3.4b

European variegated populations European falcata populations

15 25

227a 229a

97b 102a

75a 74a

57b 54a

3.7b 12.7a

0.9b 7.0a

2.5a 3.8a

1.2a 1.6a

Selected genotypes Unselected genotypes

20 27

233a 217b

101a 96b

75a 70b

57a 51b

10.8a ⫺2.5b

5.0a 0.7b

3.1a ⫺1.5b

3.1a ⫺1.8b

† Indicated differences significant at P ⫽ 0.05. ‡ Number of genotypes per grouping.

European falcata or variegated germplasm was more robust and created superior hybrids in crosses with sativa than did Asian germplasm. We contrasted all wild falcata and variegated populations west of 60⬚E longitude against all wild falcata and variegated populations east of this longitude. Both biomass yield and biomass yield heterosis across harvests and the total biomass yield, western germplasm was clearly superior to eastern germplasm (Table 5). Although this sounds vaguely imperialistic, we suspect that it merely indicates that the western germplasm is the better adapted to the Iowa environment. A comparison of wild falcata populations with wild variegated populations from the European region reveals that for total yearly yield, falcata and variegated germplasm are equivalent, although in terms of biomass heterosis, falcata germplasm has a strong advantage (Table 5). During the first harvest, wild European SFC not only expresses superior heterosis but also produces superior biomass yield compared with wild European SVC. Second harvest biomass yield and heterosis were equivalent for SFC and SVC. During the third

harvest in autumn, the SVC outyielded SFC, but both classes had equivalent heterosis levels. For several populations, we had included genotypes that had been randomly selected as well as those visually selected in the field for autumn vigor. The testcross performance of the selected genotypes was superior to their randomly chosen counterparts in all aspects of biomass yield and yield heterosis. Of all various subgroupings of the 125 genotypes, the selected individuals had the highest progeny mean testcross yields (Table 5). The selected individuals also had among the highest testcross biomass yields for first and third harvest, even superior to SSC. During second harvest, the selected genotypes had testcross yields equivalent to the SSC, when on average falcata testcross performance was clearly inferior to SSC. Excluding the European classification, the selected group had the highest levels of heterosis among any subgroup for total yearly yield heterosis and yield heterosis during each harvest. Unlike any other grouping, the progeny of the selected individuals exhibited good heterosis during the third harvest (Table 5).

Table 6. Partitioning of total variance for alfalfa biomass yield and yield heterosis into proportions for population, genotype within population, and geographic separation based on data from 2 yr in two Iowa locations. Testcross yield

Testcross heterosis

Harvest Population

Total

1st

2nd

Harvest 3rd

Total

1st

2nd

3rd

proportion All populations ␴ 2population ␴ 2genotype(population) Wild falcata and varia populations ␴ 2east–west ␴ 2population(east–west) ␴ 2genotype(east–west⫻population)

0.42* 0.58***

0.45* 0.55***

0.29* 0.71***

0.48** 0.52***

0.32* 0.68***

0.43* 0.57***

0.24NS 0.76***

0.29* 0.71***

0.40NS 0.15NS 0.45***

0.30NS 0.24* 0.46***

0.35NS 0.08NS 0.49***

0.37NS 0.22* 0.41***

0.27NS 0.15NS 0.58***

0.22NS 0.27* 0.51***

0.32NS 0.05NS 0.63***

0.18NS 0.17NS 0.65***

* Significantly different from zero at the 0.05 probability level. ** Significantly different from zero at the 0.0 probability level. *** significantly different from zero at the 0.001 probability level. NS not significantly different from zero.

16



Biomass Yield and Yield Heterosis Variances Because we had included multiple genotypes from some populations, we partitioned total genotypic variation into components for population and for genotypes within populations (Table 6). This partitioning is tentative, given that we only sampled one to five genotypes per population. Populations accounted for 42 to 48% of the variance for total yearly yield, and first and third harvest testcross yield and 29% for the second harvest yield. Populations also accounted for a lower proportion of the biomass heterosis variation, ranging from 29% to 43%, with the lowest value again at second harvest. Examining only genotypes from wild falcata and variegated populations, we partitioned the biomass yield and yield heterosis variation among genotypes, populations, and Asian-European classification (Table 6). For biomass yield, the variation among genotypes ranged from 41% to 49% with the remaining variation being

split between populations and east-west classification. During second harvest, population classification accounted for only 8% of the total biomass yield variation. Biomass yield heterosis had a higher proportion of the total variation accounted for by genotypes (range 51– 65%) than was the case for biomass yield variation. The remaining heterosis variation was divided about equally between population and east-west classification, the obvious exception was the second harvest, in which population classification accounted for only 5% of the biomass yield heterosis variation.

Population Mean Biomass Yield and Biomass Yield Heterosis Since population classification of genotypes accounted for a large proportion of both total biomass yield and biomass yield heterosis variation, we calculated population mean testcross yields (Table 7). Populations that

Table 7. Yield and heterosis of testcross progenies for each of the 44 alfalfa populations evaluated across 2 yr and two locations in Iowa, USA. Testcross yield

Population

Year total

Testcross heterosis

Harvest 1st

2nd

3rd

Harvest

Year Total

1st

2nd

3rd

⫺21 ⫺18 7 2 15 26 4 36 12 . 17 20 ⫺5 ⫺31 ⫺12 ⫺24 ⫺24 ⫺3 ⫺11 ⫺8 8 5 7 ⫺16 2 9 ⫺9 ⫺5 ⫺1 ⫺14 ⫺25 5 20 ⫺20 ⫺3 8 17 16 12 22 ⫺5 ⫺13 ⫺14 ⫺5 0 21

⫺14 ⫺11 3 14 8 13 1 15 2 . 9 2 2 ⫺15 ⫺4 ⫺15 ⫺11 2 ⫺2 3 4 ⫺4 1 ⫺7 2 6 ⫺5 ⫺1 7 ⫺5 ⫺10 3 10 ⫺7 0 1 9 8 8 4 ⫺6 ⫺10 ⫺11 ⫺6 0 10

⫺2 ⫺3 3 ⫺5 5 12 3 9 6 . 4 ⫺1 ⫺4 ⫺11 ⫺7 ⫺9 ⫺7 ⫺1 ⫺6 ⫺6 3 5 1 ⫺8 1 5 0 ⫺7 ⫺3 ⫺3 ⫺10 3 5 ⫺7 ⫺3 3 6 0 0 10 3 0 ⫺1 1 0 10

⫺7 ⫺4 3 ⫺4 4 5 2 11 4 . 2 16 ⫺4 ⫺8 ⫺2 ⫺4 ⫺8 ⫺4 ⫺4 ⫺4 ⫺2 3 4 ⫺5 1 ⫺3 ⫺3 2 ⫺5 ⫺4 ⫺4 0 5 ⫺5 ⫺3 6 3 8 6 7 1 ⫺1 ⫺2 0 0 6

g plant⫺1 1. PI 631591 2. PI 251836 3. PI 631579 4. PI 253451 5. PI 631796 6. 5291/88 7. 5299/88 8. PI 631857 9. PI 494661 10. PI 494658 11. PI 502441 12. PI 384890 13. PI 538985 14. PI 440539 15. PI 631608 16. PI 631612 17. PI 631601 18. PI 538993 19. PI 577561 20. PI 499661 21. PI 631645 22. PI 631639 23. PI 499548 24. W6 16608 25. PI 214218 26. PI 631806 27. PI 314092 28. PI 631811 29. PI 573175 30. PI 631620 31. PI 631597 32. PI 631596 33. PI 502453 34. PI 631797 35. PI 631799 36. IA-3018 37. Lodgeland 38. SD 201 39. WISFAL 40. PI 468015 41. Pioneer Hi-Bred 42. Forage Genetics 43. Hungarian 44. ABI Alfalfa Mean LSD†

199 200 235 227 244 264 232 247 237 241 233 227 210 175 196 178 188 213 207 208 218 210 220 191 221 219 211 206 217 208 204 228 238 202 207 239 235 231 237 234 223 218 208 220 218 22

† Least significant differences at P ⫽ 0.05 level.

88 84 100 110 106 110 98 109 98 99 103 96 97 78 90 78 84 97 93 98 97 89 95 85 97 100 90 93 101 92 87 99 106 89 94 99 105 103 105 98 91 87 84 91 95 11

67 69 76 67 78 90 75 77 78 83 73 68 66 55 60 56 62 68 66 64 72 71 72 61 71 73 71 62 68 66 66 74 76 65 65 77 77 70 72 79 77 77 72 75 71 10

45 48 59 53 60 68 60 62 61 63 55 64 47 41 48 44 43 48 47 46 48 50 54 44 53 46 50 50 47 49 51 54 56 48 48 63 54 58 60 57 56 54 51 55 53 6



had testcross yields equal to or above 223 g plant⫺1, the value of the highest yielding SSC population (Pioneer Hi-bred), include PI631796, 5291/88, and PI631857. Populations that were especially impressive visually were PI631796 and PI494661; although the parental genotypes were agronomically undesirable, their testcross progeny combined the robust growth form of the sativa parent with the extreme vegetative density and creeping rooted features of the falcata parentage to create impressive hybrids. Other populations such as PI631857 and WISFAL had growth forms approaching sativa and when testcrossed to sativa, produced strong hybrids with very “sativa like” growth forms. This and previous studies (Riday and Brummer, 2002a, 2002b) have demonstrated the strength of sativa–falcata hybrids during the first harvest. In this study, we demonstrated that numerous populations exist that have competitive second and third harvest yields (Table 7). Finally, it is readily apparent that many populations have negative heterosis values and perform poorly in sativa–falcata hybrids. Thus, the important factor in generating heterosis is not genetic diversity per se, but usable variation, preferably complimentary between parents, at the various salient, but unknown, loci. Genetic diversity alone is not sufficient to ensure heterosis.

SUMMARY On the basis of the results from this study, we conclude that the best falcata germplasm for use in sativa– falcata hybrids in Iowa, and probably the upper midwestern USA, comes from European populations and improved germplasm sources already available. Pure falcata populations—i.e., those without genotypes having variegated flowers—have higher levels of heterosis during the first harvest, but their productivity, both per se and in testcrosses, diminishes throughout the year. Simple visual selection for superior late season growth greatly improves the ability of falcata or varia populations to complement sativa and produce highly heterotic progeny. The low ratio of additive to total genetic variance for yield shows the value of considering a hybrid cultivar development scheme to capture nonadditive variance. This study demonstrates the clear advantage of progeny testing. The progeny testing described in this experiment not only ensures the maintenance of the heterotic pattern between the two breeding populations, but also allows much more rapid genetic gain per selection cycle (i.e., gain based on additive variance) within each of the two breeding populations. REFERENCES Barnes, D.K., E.T. Bingham, R.P. Murphy, O.J. Hunt, D.F. Beard, W.H. Skrdla, and L.R. Teuber. 1977. Alfalfa germplasm in the

17

United States: Genetic vulnerability, use, and maintenance. USDAARS Tech. Bull. 1571. USDA-ARS, Hyattsville, MD. Bingham, E.T. 1993. Registration of WISFAL alfalfa (Medicago sativa subsp. falcata) tetraploid germplasm derived from diploids. Crop Sci. 33:217–218. Bingham, E.T., R.W. Groose, D.R. Woodfield, and K.K. Kidwell. 1994. Complementary gene interactions in alfalfa are greater in autotetraploids than diploids. Crop Sci. 34:823–829. Brummer, E.C. 1999. Capturing heterosis in forage crop cultivar development. Crop Sci. 39:943–954. Brummer, E.C., J.D. Berdahl, A.R. Boe, R. Michaud, S.R. Smith, and P.C.S. Amand. 1997. Evaluation of tetraploid Medicago sativa ssp. falcata accessions. In Proc. 25th Central Alfalfa Impr. Conf., La Crosse, WI. 16–18 July 1997. Univ. of Wisconsin, Madison. Cockerham, C.C. 1956. Effects of linkage on the covariances between relatives. Genetics 41:138–141. Dudley, J.W., T.H. Busbice, and C.S. Levings, III. 1969. Estimates of genetic variance in ‘Cherokee’ alfalfa (Medicago sativa, L.). Crop Sci. 9:228–231. Falconer, D.S. and T.F.C MacKay. 1996. Introduction to quantitative genetics. Longman, Essex, England. Fehr, W.R. 1991. Principles of cultivar development. Vol. 1. Iowa State Univ., Ames, IA. Galton, F. 1888. Co-relations and their measurement, chiefly from anthropometric data. Proc. Roy. Soc. London. 45:135–145. Hallauer, A.R., and J.B. Miranda. 1988. Quantitative genetics in maize breeding. Iowa State Univ., Ames, IA. Hansen, N.E. 1909. The wild alfalfas and clovers of Siberia, with a perspective view of the alfalfas of the world. USDA Bull. 150. U.S. Gov. Print. Office, Washington, DC. Hill, R.R., Jr., and J.S. Shenk, and R.F Barnes. 1988. Breeding for yield and quality. p. 809–825. In A.A. Hanson et al. (ed.) Alfalfa and alfalfa improvement. ASA, CSSA, and SSSA, Madison, WI. Kehr, W.R., and C.O. Gardner. 1960. Genetic variability in Ranger alfalfa. Agron. J. 52:41–44. Kempthorne, O. 1969. An introduction to genetic statistics. Iowa State Univ., Ames, IA. Lesins, K., and I. Lesins. 1979. Genus Medicago (Leguminasae): A taxogenetic study. Kluwer, Dordrecht, the Netherlands. Levings, C.S., and J.W. Dudley. 1963. Evaluation of certain mating designs for estimation of genetic variance in autotetraploid alfalfa. Crop Sci. 3:532–535. Littell, R.C., G.A. Milliken, W.W. Stroup, and R.D. Wolfinger. 1996. SAS system for Mixed Models. SAS Institute Inc., Cary, NC. Nguyen, H.T., and D.A. Sleper. 1983. Theory and application of halfsib matings in forage grass breeding. Theor. Appl. Genet. 64:187– 196. Riday, H., and E.C. Brummer. 2002a. Forage yield heterosis in alfalfa. Crop Sci. 42:716–723. Riday, H., and E.C. Brummer. 2002b. Heterosis of agronomic traits in alfalfa. Crop Sci. 42:1081–1087. Riday, H., E.C. Brummer, and K.J. Moore. 2002. Heterosis of forage quality in alfalfa. Crop Sci. 42:1088–1093. SAS Institute. 2000. SAS language and procedure: Usage. Version 8, 1st ed. SAS Inst., Cary, NC. Snedecor, G.W., and W.G. Cochran. 1967. Statistical methods. Iowa State Univ., Ames, IA. USDA. ARS, National Genetic Resources Program. Germplasm Resources Information Network-(GRIN). [Online Database] National Germplasm Resources Laboratory, Beltsville, MD. Available: Http://www.ars-grin.gov/cgi-bin/npgs/html/tax_acc.pl?104918; verified 6 August 2004. USDA. National Agricultural Statistical Service. State level data for field crops: Hay http://www.nass.usda.gov:81/ipedb/main.htm; verified 6 August 2004. Zamudio, F., and R.D. Wolfinger. 2002. Growth increments and stability over time in fast-growing forest tree species. Can. J. For. Res. 32:942–953.