Switchgrass is a native warm-season grass of the North American

Transcription

1 Published January 21, 2013 RESEARCH Space-Plant versus Sward-Plot Evaluation of Half-Sib Families to Select Parents for Synthetic Cultivars with Superior Biomass Yield in Lowland Switchgrass H.S. Bhandari,* V. A. Fasoula, and J.H. Bouton Abstract Switchgrass (Panicum virgatum L.) needs significant biomass yield improvement for its viable use as bioenergy feedstock. Conventional family evaluation under sward plot is very effective in cultivar improvement, but it is time and resource consuming. Family evaluation under space planting using honeycomb design can reduce time and inputs. The objective of this study was to compare conventional sward-plot and modified space-plant evaluation of half-sib families in their biomass yield gain and resource use. Using same initial set of genotypes selected from Alamo, two experimental synthetics were developed by the following methods: (i) summer field polycrossing of selected genotypes to produce half-sibs, half-sib family evaluation under sward plot conditions (HSSward) at two locations for 2 yr, and polycrossing the five genotypes showing superior half-sibs to produce the experimental and (ii) greenhouse polycrossing genotypes during the winter to produce half-sibs, half-sib family evaluation under space-plant conditions using honeycomb design (HSSpace) for 2 yr at one location, and polycrossing the five best progeny within the five superior half-sibs to produce the experimental. The 2 to 3 yr biomass yield performance of two experimentals across five locations demonstrated that both HSSward and HSSpace led to a significant biomass yield gain over Alamo in three and two of the five locations, respectively. The HSSward produced a superior and more stable synthetic across a wide region than HSSpace. However, the HSSpace produced its synthetic a year sooner and was more resource efficient and, under finite resources, it could be a method of choice for cultivar development especially for a narrow geographic region. H.S. Bhandari, Dep. of Plant Sciences, Univ. of Tennessee, 2431 Joe Johnson Drive, Knoxville, 37996; V.A. Fasoula, Institute of Plant Breeding, Genetics & Genomics University of Georgia, 111 Riverbend Road, Athens, GA 30602; J.H. Bouton, Forage Improvement Division, The Samuel Roberts Noble Foundation, Inc., 2510 Sam Noble Parkway, Ardmore, OK Received 15 June *Corresponding author (hsbhandari@utk.edu). Abbreviations: AWF-HS, among and within half-sib family selection; c1syn-1, cycle-1 synthetic generation 1; c1syn-2, cycle-1 synthetic generation 2; HSF, among half-sib family selection; HSPT, half-sib progeny testing; HSSpace, half-sib family evaluation under spaceplant conditions using honeycomb design; HSSward, half-sib family evaluation under sward-plot conditions; PLS, pure live seed; RCBD, randomized complete block design; UGAPSF, University of Georgia Plant Sciences Farm. Switchgrass is a native warm-season grass of the North American tallgrass prairie. It is a perennial grass adapted to a wide geographic range and has high biomass yield potential, including in marginal lands with minimum inputs and water use. Its roots grow several meters deep in the soil profile and are capable of sequestering a large quantity of atmospheric CO 2 into organic C each year (Anderson-Teixeira et al., 2009; Blanco-Canqui, 2010; Liebig et al., 2008). Because of these and several other favorable attributes, switchgrass was chosen as an important herbaceous species for use in lignocellulosic feedstock production in the United States (McLaughlin and Kszos, 2005). Current breeding efforts are directed at improving the biomass yield potential of switchgrass to meet the cellulosic feedstock demand of future bioenergy industries. Published in Crop Sci. 53: (2013). doi: /cropsci Freely available online through the author-supported open-access option. Crop Science Society of America 5585 Guilford Rd., Madison, WI USA 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 crop science, vol. 53, march april 2013

2 Switchgrass possesses a tremendous amount of genetic diversity (Hultquist et al., 1996; Missaoui et al., 2006; Narasimhamoorthy et al., 2008). Native populations of switchgrass are broadly classified into two distinct cytotypes, L and U, that are associated with lowland and upland ecotypes, respectively (Hultquist et al., 1997). Several ploidy levels have been reported (Hopkins et al., 1996; Lu et al., 1998; McMillan and Weller, 1959; Riley and Vogel, 1982). Lowlands are exclusively tetraploids, capable of producing high biomass yield, and are adapted to southern regions with mild temperatures. Uplands are mostly octaploids and less frequently tetraploids, lower yielding, and more winter-hardy than lowlands and, as such, are more adapted to northern environments. Polyploidy, outcrossing, and selfincompatibility systems have maintained a high degree of diversity within populations offering tremendous potential for genetic improvement through selection. Conventional methods of cultivar development through recurrent phenotypic or genotypic selection have been successful in improvement of biomass yields and other agronomic traits in grass species (Burton, 1989). Half-sib and full-sib family evaluation procedures are widely used as conventional genotypic selection methods of population improvement in most allogamous species. In half-sib family selection, one can practice either among half-sib family selection (HSF) and use selected families as recombination units or practice half-sib progeny testing (HSPT) and use parental clones as recombination units or adopt both among and within half-sib family selection (AWF-HS) and use superior individual plants within selected families as recombination units. The expected gain from selection following these methods depends on the plant heritability of the trait and selection intensity (Casler and Brummer, 2008). One of the major constraints in breeding perennial grasses such as switchgrass is that it involves a long selection cycle. A typical cycle involves selection of parental clones (random or selected from a previous breeding cycle), a year of polycrossing of selected clones in the field to generate half-sibs and/or full-sibs, 2 yr of family evaluation in a replicated sward plot at one or more locations and selection of clones following HSF, HSPT, or AWF-HS based on family and/or genotypes within family performance, and a year for polycrossing in the field to produce experimental synthetics or the base population for the next cycle. Because of these steps, conventional genotypic selection is resource and time intensive. However, procedures that involve family evaluation in sward-type plots have been very effective in improving switchgrass biomass yield (Taliaferro et al., 1999; Vogel, 2004). Evaluation of families in a space-plant nursery can be accomplished with fewer resources. The space-plant evaluation requires fewer plants of each family, for which the seed can be produced by polycrossing among selected clones in the greenhouse during the winter, thus saving a year per selection cycle. In addition, use of a honeycomb design in space-plant evaluation is reported to be effective in selection among and within families, thereby improving gain from selection (Abraham and Fasoulas, 2001; Fasoula and Fasoula, 1997, 2000, 2002; Fasoulas and Fasoula, 1995; Missaoui et al., 2005). Using Alamo as a base population, two methods of half-sib family evaluation were followed to select parents for new experimental synthetics: (i) selection half-sib family evaluation under sward plot conditions, the method hereafter referred to as HSSward, leading to development of the experimental hereafter referred to as Swardvar, and (ii) selection half-sib family evaluation under space-plant conditions using honeycomb design, the method hereafter referred to as HSSpace that led to the experimental hereafter referred to as Spacevar. The objective of this study was to compare the performance of Spacevar and Swardvar relative to the performance of Alamo and elucidate the breeding efficiency of modified space-plant evaluation using a honeycomb design compared to the conventional swardplot method of family evaluation. Materials and Methods Development of Experimental Synthetics An initial population of 966 genotypes of Alamo, a southern lowland switchgrass variety, was established in the field at the University of Georgia Plant Sciences Farm (UGAPSF) near Watkinsville, GA ( ² N, ² W), in 1996 using a commercial seed lot. The nursery was space planted in a 69-row by 14-plant plot following an unreplicated honeycomb design with 1.2 m plant spacing (Fasoulas and Fasoula, 1995). Honeycomb designs are sets of selection designs that evaluate individual plants by reducing the confounding effects of interplant competition and soil heterogeneity (Fasoula and Fasoula, 1997). In these designs, plants are arranged in staggered field rows in a hexagonal pattern and the plant-to-plant spacing is always the same. Figure 1 illustrates the unreplicated honeycomb design that evaluated the 966 genotypes of Alamo. In honeycomb layout, each plant holds a central position and is encircled by a random sample of plants in concentric rings of different sizes, from very small (6 plants) to very large (>42 plants) as shown in Fig. 1 for two random plants (Fasoulas and Fasoula, 1995). These rings are called moving rings because they are moved from plant to plant in the field (Fasoula and Fasoula, 2000). The yield of each plant is expressed relative to the mean yield of its neighboring plants within the moving ring of given size. The purpose of the moving-ring selection is to reduce effectively the masking effects of field or environmental variation on single-plant yields. Selection in the unreplicated honeycomb design of the 966 Alamo plants was performed using a moving ring of 30 plants (shown in Fig. 1) and the plant yield potential, that is, as measured by (x/ x r ) 100, in which x is the yield of each plant and x r is the mean plant yield of neighboring 30 plants within the moving ring. Moving-ring selection objectively ranked the plants and identified 45 superior genotypes (about 5% selection pressure) for use in polycrossing to generate half-sib families. Two methods of half-sib family evaluation, (i) replicated sward-plot evaluation crop science, vol. 53, march april

3 Figure 1. The unreplicated honeycomb design. The black circles represent the position of the plants in the field. Each plant is encircled by a random number of plants allocated in the periphery of concentric rings. As shown in the figure, these rings can have different sizes and are called moving rings. Moving rings of 6, 12, 18, 30, 36, and 42 plants are illustrated here for two random plants, but larger rings can also be obtained (Fasoulas and Fasoula, 1995). In the unreplicated switchgrass trial, the confounding effect of soil and/or environmental heterogeneity on single-plant yields was reduced by adjusting the yield of each plant to the mean yield of its 30 neighboring plants, thus allowing selection of plants in high- as well as in low-yielding field spots. (HSSward) and (ii) modified space-plant evaluation using honeycomb design (HSSpace), were followed to develop two experimental synthetics. In the HSSward, the selected 45 genotypes were planted in the field at UGAPSF and polycrossed in the late summer through fall of Four clonal replicates of each genotype were planted in randomized complete block design (RCBD). Plants were spaced 1 m apart. The seed from all the replicates of a given genotype was pooled to constitute a half-sib family seed. Only 14 genotypes were able to produce enough seeds needed for replicated sward-plot evaluation of half-sib families. These 14 families along with Alamo check were evaluated in replicated sward plots at two locations in Georgia, Athens and Tifton, during 1998 and 1999 using RCBD with three replications. Based on mean performance of half-sib families across location and years, five (~36%) superior biomass yielding families were selected. The parental clones of selected families were established in a polycross nursery in fall Six clonal replicates of each of five genotypes were planted in RCBD and intercrossed. All the seeds were pooled by genotypes and a small quantity of balanced bulk was kept as remnant seed, and all remaining seed was bulked to form cycle-1 synthetic generation 1 (c1syn-1) seed of experimental synthetic Swardvar. In 2001, the cycle-1 synthetic generation 2 (c1syn-2) seed was produced from the random c1syn-1 seed sample. This seed increase was established at the UGAPSF in a 400 m 2 area as drilled rows spaced 0.9 m apart at a seeding rate of 1.5 kg pure live seed (PLS) ha -1. In the HSSpace, the 45 genotypes selected from the initial population were clonally multiplied to make four copies each and planted in pots in the University of Georgia Crop and Soil Sciences greenhouse. The pots were placed in random arrangement, and the plants were polycrossed during the winter through spring of 1996/1997. Seed produced by all the four clones of each genotype was pooled to constitute a half-sib family seed. All the genotypes produced enough seeds for family evaluation using space-plant nursery following honeycomb design. The 45 families along with four Alamo check entries were evaluated at the UGAPFS in a replicated 49 honeycomb design (Fasoulas and Fasoula, 1995) in 26-row 38-plant plot with 1.2 m plant spacing (Fig. 2). There were 20 genotypes planted for each of the families and checks. Plants of a certain family form a triangular grid pattern and are surrounded by a set of neighboring plants belonging to the same set of families (Fig. 2). This unique feature allows expressing the yield of a given family relative to the mean of a common multiplant moving ring of neighboring plants. The moving ring of 30 plants is illustrated in Fig. 2 for three random plants of family 15 (gray circles) and the same applies for all the plants of family 15 as well as for the plants of each family in the honeycomb design. Based on 2 yr biomass yield performance, five superior halfsib families (i.e., superior 11% families) were identified using two criteria: (i) high mean biomass yield per plant to select for high yield potential and (ii) low family CV to select for high buffering (Fasoula and Fasoula, 1997). Within-family selection was performed based on plant yield potential obtained by using equation (x/ x r ) 100, in which, x is the yield of a given plant and x r is the mean plant yield of its 30 neighboring plants (as shown in Fig. 2; gray circles). Five superior genotypes were selected from within each of the five selected families (25% within family). The selected 25 genotypes were removed from the field, clonally multiplied to make six copies each, and planted in a replicated polycross nursery in fall 1999 at the UGAPSF. The seed harvested from all the replicates of a given genotype was pooled and a small quantity of balanced bulk was kept as remnant seed, and all remaining seed was bulked to compose c1syn-1 seed of the experimental synthetic Spacevar. The c1syn-2 seed was produced from the random c1syn-1 seed sample. This seed increase was established at the UGAPSF in a 400 m 2 area as drilled rows spaced 0.9 m apart at a seeding rate of 1.5 kg PLS ha -1. Testing Experimentals in Performance Trails The two experimentals, Swardvar and Spacevar, the Alamo check, and depending on the trial location, other lowland and upland cultivars, were evaluated at various locations: Tifton, GA (31.45 N, W), Starkville, MS (33.25 N, W), Raymond, MS (32.13 N, W), Overton, TX (32.18 N, W), and Ardmore, OK (34.11 N, W). The Tifton test site had clay loam soil, with soil ph = 5.8, available P = kg ha -1, and K = 47.1 kg ha -1. The soil at the Starkville site was Marietta fine sandy loam, with ph = 5.4, P = kg ha -1, and K = kg ha -1. The Raymond site had Loring silt loam soil, crop science, vol. 53, march april 2013

4 Figure 2. The replicated 49 honeycomb design that evaluated the 49 entries (45 families and 4 check entries) of the switchgrass trial in 26 rows and 38 plants per row. Plants are arranged in horizontal rows in an ascending order and the number set is repeated regularly. The starting number is different in each row and is given by a specific algorithm. The numbers in the figure represent the position of the plants in the field. Plants of any family are positioned in a triangular grid pattern that spreads evenly across the field sampling effectively for soil heterogeneity. The triangular grid is shown here for the plants of family 15. In addition, plants of any family are always surrounded by neighboring plants that belong to the same families and this is used to express the yield of each plant in percentage of the mean yield of its neighboring plants within a certain ring. Single-plant selection within the selected families was performed using a moving-ring average of 30 plants, illustrated here in gray rings for three random plants of entry 15. with ph = 5.7, P = 44.8 kg ha -1, and K = 80.7 kg ha -1. The Overton site had sandy loam type soil, with ph = 6.0, P = kg ha -1, and K = kg ha -1. The soil at the Ardmore site was Normangee clay, with ph = 6.7, P = 51.6 kg ha -1, and K = kg ha -1. The precipitation ranged from 110 to 135 cm at Tifton, 100 to 110 cm at Starkville and Raymond, and 75 to 90 cm at the Ardmore and Overton sites. Each trial was planted in sward plots (3.1 by 4.6 m) using a RCBD with six replications, except in Raymond where only four replications were planted. The trial at Tifton was conducted using c1syn-1 seed of Swardvar and c1syn-2 seed of Spacevar. Different synthetic generations were used at Tifton because c1syn-1 seed of Spacevar was not sufficient for testing under replicated sward plots while c1syn-2 seed of Swardvar was yet to be produced. The trials in all other sites were conducted using c1syn-2 seeds of both experimentals. Seeding rate was 5 kg PLS ha -1. At Tifton, the trial was planted on 2 May 2001, and a single end-season biomass yield was recorded for the establishment year. In 2002 and 2003, biomass was clipped twice in midsummer (i.e., 23 July 2002 and 24 June 2003) and at the end of the fall (26 Nov and 20 Nov. 2003) and totaled to obtain seasonal biomass yield. At Starkville, the trial was planted on 5 May 2006, and biomass was clipped on 11 Dec. 2007, 27 Jan (for 2008 crop), and 5 Jan (for 2009 crop). At Raymond, the trial was planted on 2 May Two adjacent plots for each entry were planted in each replicate, one for normal harvest at the end of fall and the other for the delayed harvest in the spring. Fall harvest was performed on 6 Dec. 2007, 12 Dec. 2008, and 8 Nov The spring harvests were performed on 28 Mar. 2008, 19 Mar. 2009, and 18 Mar (i.e., for the 2007, 2008, and 2009 crop seasons, respectively). At Overton, the trial was planted on 26 May Biomass was clipped on 11 and 12 October in all 3 yr. At Ardmore, the trial was planted on 8 June The biomass was clipped on 20 Dec. 2007, 30 Dec. 2008, and 22 Jan (i.e., for 2009 season). The biomass was harvested using a flail-type carter forage harvester at Tifton, Starkville, and Ardmore while a behind-the-wheel sickle bar mower was used at the Overton and Raymond sites. The moisture content at harvest was estimated for each plot using approximately 200 g of chopped biomass sampled during harvesting, oven dried for 72 h, and used to compute dry matter yields of corresponding plots. Fertilizers were applied to the experimental plots in early May of each year according to common practice adapted in each test location. At Tifton, 85 kg ha -1 each of N, P 2 O 5, and K 2 O was applied each year using complete fertilizer (19:19:19). At Starkville, 85 kg ha -1 N was applied each year using NH 4 NO 3 (34:0:0). At Raymond, 60 kg ha -1 each of N, P 2 O 5, and K 2 O was applied in crop science, vol. 53, march april

5 2007 using complete fertilizers (15:15:15), and 80 kg ha -1 N was added in the subsequent years using urea (40:0:0). At Overton, complete fertilizers (16:6:12) were applied each year to provide 54:10:20 kg ha -1 of N, P 2 O 5, and K 2 0, respectively. At Ardmore, 110 kg ha -1 N were applied using urea (40:0:0) in each year. Data Analysis Biomass dry matter yield data were analyzed using mixed model in SAS (SAS Institute, 2010), and the least square means for varieties were obtained. The establishment year data were excluded from analysis. The data were analyzed separately for each location because of differences in year of evaluation and harvesting systems, that is, a single cut at the end of the fall, a single cut at the end of the spring, or two cuts, one during midsummer and the other at the end of the fall. Cultivar and year effects were assumed to be fixed, and replication effects were considered random. Cultivar was considered main plot, and year effects were considered split plots. Mixed analysis included Swardvar, Spacevar, Alamo check, and two other experimentals of lowland switchgrass that were evaluated in all the locations. This discussion, however, is restricted to comparison between the two experimentals, Swardvar and Spacevar, relative to their source population, Alamo. Biomass yields of these three varieties from established stands (i.e., second and third year harvest) were also subjected to stability analysis using Tai s procedure (Tai, 1971; Thillainathan and Fernandez, 2001). In this procedure, the variety environment (i.e., year in our case) interaction term is partitioned into a linear response to environmental effects measured by a and the deviation from the linear response measured by l. The cultivars with (a,l) = (0,1) are perfectly stable, and cultivars with (a,l) = (-1,1) are average in stability. Since the experiment at Tifton was performed in different years, data from the Tifton location were excluded from this analysis. The result from Tai s stability analysis is presented in a three-dimensional plot (Fig. 3). Results At the Tifton, GA, location, the two experimentals, Spacevar and Swardvar, performed similarly, and both Figure 3. Plot of mean biomass yield by stability across different locations of two experimental synthetics of lowland switchgrass: Swardvar, developed by using conventional sward-plot evaluation, and Spacevar, developed using space-plant evaluation using honeycomb design, compared with Alamo, the parent population. The a (ALPHA) measures the linear response to the environment, and l (lambda) measures the deviation from the linear response. In the figure, two experimentals are shown by the blue club for which l is significant, and Alamo is shown by green spade for which both a and l are not significant. The varieties with (a, l) = (0,1) are perfectly stable, and varieties (a, l) = (-1,1) are average in stability. 3-D, three-dimensional; LSMEAN, least square mean. outperformed Alamo, the base population, across the 2 yr (Table 1). This showed that HSSpace could be as effective as HSSward in achieving biomass yield gain over a limited and more targeted geographic region. The average yield gain over Alamo was 29%. The yield advantage of these two experimentals was 46% in 2002 and 18% in The yield gap between the two experimentals and Alamo tended to close over the years as the stands mature. Biomass yield in Table 1. Biomass yields of experimentals of lowland switchgrass: Swardvar, with selection based on half-sib family evaluation under sward plot conditions (HSSward), and Spacevar, with selection based on half-sib family evaluation under space-plant conditions using honeycomb design (HSSpace) (Tifton, GA). Biomass yield, Mg ha Overall Cultivar Summer Fall Total Summer Fall Total mean Swardvar (c1syn-1) 14.25a 3.26a 17.51a 16.65a 2.63a 19.28a 18.40a Spacevar (c1syn-2) 12.62a 3.68a 16.30a 15.55ab 2.83a 18.38ab 17.34a Alamo 9.80b 1.81b 11.61b 13.43bc 2.59a 16.01b 13.81b Mean LSD CV% Te s to ffi x e de f f e c t s ( F t e s t ) Cultivar (<0.0001) Year 3.26 (<0.1016) Cultivar year 0.87(0.4911) c1syn-1, cycle-1 synthetic generation 1; c1syn-2, cycle-1 synthetic generation 2. The figures with same letters are not significantly different at p crop science, vol. 53, march april 2013

6 Table 2. Biomass yields of experimentals of lowland switchgrass: Swardvar, with selection based on half-sib family evaluation under sward plot conditions (HSSward), and Spacevar, with selection based on half-sib family evaluation under space-plant conditions using honeycomb design (HSSpace) (Starkville, MS). Biomass yield, Mg ha -1 Cultivar Mean Swardvar (c1syn-2) 9.36a 19.88ab 21.84a 17.03a Spacevar (c1syn-2) 8.34a 23.78a 23.36a 18.49a Alamo 7.37a 16.67b 20.23a 14.76b Mean LSD CV% Te s to ffi x e de f f e c t s Cultivar 3.0 (<0.0261) Year 74.6 (<0.0001) Cultivar year 1.8 (<0.106) c1syn-2, cycle-1 synthetic generation 2. The figures with same letters are not significantly different at p The test of fixed effects are F values and the figures in the parentheses are their probabilities (P > F). summer clippings in 2002 and 2003, respectively, was 81 and 85% of the corresponding year s seasonal biomass yields. The effects of cultivar year interaction were not evident. Results from the Starkville, MS, site followed a similar trend where the two methods of selection led to similar but significant biomass yield gain over Alamo. However, the yield gain following either method of selection was smaller at Starkville compared with the yield gain at Tifton, GA (Table 2). On average, the two experimentals produced 20% higher biomass yield when compared with Alamo (14.76 Mg ha -1 ). At Raymond, MS, the 3-yr average fall biomass yield of Swardvar (19.36 Mg ha -1 ) was 14% higher when compared with Alamo (17.0 Mg ha -1 ) (Table 3). In contrast, the average fall biomass yield of Spacevar was 17% less compared with Alamo. Biomass yield was found to be influenced by cultivar year interaction. In 2007, both Spacevar and Swardvar perform similarly to Alamo. However, in 2008, Swardvar produced more biomass than both Alamo and Spacevar (Table 3). In 2009, Swardvar and Alamo produced significantly more biomass than Spacevar. This trend was similar across years when the biomass harvest was delayed until spring except in 2007 season, when Spacevar was superior to both Swardvar and Alamo. Delayed harvesting led to loss of biomass yield as much as 43% in mature stands (Table 3), which could be attributable to weight loss of tillers after maturity and the biomass that could not be picked up by harvester due to lodging (Adler et al., 2006). Alamo lost more biomass than Swardvar and Spacevar from fall to spring in 2007 and This difference could be due to differences in maturity. Alamo, which was a week earlier compared to the two experimentals (results not shown), could have lost more biomass due to its early start of nutrient recycling. At Ardmore, both Swardvar and Spacevar performed similarly to Alamo (Table 4). The mean biomass yield of the experiment was Mg ha -1 in 2008, which dropped by 47% in There was no evidence of cultivar year interaction. The reduction in yield from 2008 to 2009 could be due to freeze damage in the spring of 2009 that occurred during the initial green-up phase of the plots breaking dormancy. The larger decline in yield in Spacevar compared to Swardvar and Alamo could indicate the low plasticity of the Spacevar to environmental fluctuations. At Overton, biomass yields of both Swardvar and Spacevar were similar to Alamo (Table 5). There was no evidence of cultivar year interaction for biomass yield. The mean biomass yield of the experiment was 8.21 Mg ha -1, and the average biomass yields between 2008 and 2009 were Table 3. Biomass yields of experimentals of lowland switchgrass: Swardvar, with selection based on half-sib family evaluation under sward plot conditions (HSSward), and Spacevar, with selection based on half-sib family evaluation under space-plant conditions using honeycomb design (HSSpace) (Raymond, MS). Fall harvest Biomass yield, Mg ha -1 Spring harvest Cultivar Mean Mean Swardvar (c1syn-2) 5.80a 24.07a 29.50a 19.36a 5.43b 22.81a 22.48a 16.28a Spacevar (c1syn-2) 6.63a 14.14b 21.38b 14.05b 7.52a 15.61b 14.97b 12.79b Alamo 5.82a 17.13b 28.04a 17.00a 4.38b 13.10b 21.14a 12.76b Mean LSD CV% Te s to ffi x e de f f e c t s Cultivar 4.84 (<0.0014) 4.09 (<0.0026) Year (<0.0001) (<0.0001) Cultivar year 4.85 (<0.0014) 4.09 (<0.0026) c1syn-2, cycle-1 synthetic generation 2. The figures with same letters are not significantly different at p The test of fixed effects are F values and the figures in the parentheses are their probabilities (P > F). crop science, vol. 53, march april

7 Table 4. Biomass yields of experimentals of lowland switchgrass: Swardvar, with selection based on half-sib family evaluation under sward plot conditions (HSSward), and Spacevar, with selection based on half-sib family evaluation under space-plant conditions using honeycomb design (HSSpace) (Ardmore, OK). Biomass yield, Mg ha -1 Cultivar Mean Swardvar (c1syn-2) 15.20a 10.87a 13.02a Spacevar (c1syn-2) 18.64a 10.42a 14.53a Alamo 16.76a 12.77a 14.76a Mean LSD CV% Te s to ffi x e de f f e c t s Cultivar 1.16 (>0.3450) Year (<0.0017) Cultivar year 0.95 (>0.4463) c1syn-2, cycle-1 synthetic generation 2. The figures with same letters are not significantly different at p The test of fixed effects are F values and the figures in the parentheses are their probabilities (P > F). Table 5. Biomass yields of experimentals of lowland switchgrass: Swardvar, with selection based on half-sib family evaluation under sward plot conditions (HSSward), and Spacevar, with selection based on half-sib family evaluation under space-plant conditions using honeycomb design (HSSpace) (Overton, TX). Biomass yield, Mg ha -1 Cultivar Mean Swardvar (c1syn-2) 8.49a 8.00a 8.24a Spacevar (c1syn-2) 7.97a 7.21a 7.59a Alamo 7.66a 8.32a 7.99a Mean LSD CV% Te s to ffi x e de f f e c t s Cultivar 2.03 (<0.1062) Year 0.07 (< ) Cultivar year 0.99 (0.4229) c1syn-2, cycle-1 synthetic generation 2. The figures with same letters are not significantly different at p The test of fixed effects are F values and the figures in the parentheses are their probabilities (P > F). not different. This is expected because, unlike at Ardmore, freezing was not an issue at Overton. It was apparent that HSSward produced a relatively more stable variety compared with HSSpace. Tai s stability analysis (Tai, 1971; Thillainathan and Fernandez, 2001) across four locations (Tifton site excluded) showed that Spacevar, developed using HSSpace, was less stable compared to Swardvar, developed using HSSward, as well as Alamo, the source population from which these two experimentals were developed (Fig. 3). It was not clear if such results were attributable to the use of single location during the selection of Spacevar as opposed to two locations used during the selection of Swardvar. Discussion The genetic diversity available within native strains of switchgrass offers tremendous potential for biomass yield improvement through selection. The two experimental synthetics, Swardvar and Spacevar, reported here were derived from Alamo, a widely grown southern lowland variety, following two methods of selection among half-sib families using the same set of initially selected genotypes. Development of Swardvar involved selection among half-sib families based on sward-plot evaluation and use of parental clones as recombination units (HSSward). Development of Spacevar involved selection based on space-plant evaluation following honeycomb design (HSSpace) and among- and within-family selection. The relative advantage of amongand within-family selection over among family selection using parental clones as recombination units has been discussed by Casler and Brummer (2008). The results demonstrated that the overall mean yield gain of Spacevar was as high as that obtained for Swardvar only at Tifton and Starkville; only Swardvar demonstrated significant yield gain at Raymond, and the two experimentals did not produce any yield gain over the Alamo parent in Ardmore and Overton. Casler and Brummer (2008) demonstrated higher theoretical yield gain with AWF-HS selection compared with HSPT selection. In this current study, somewhat different experimental settings were used such as two different selection nurseries, two different plant densities, and two versus one evaluation locations for HSSward and HSSpace evaluation, respectively. However, our expectation was that HSSpace would have produced a higher yield gain due to ability of honeycomb design to select superior genotypes within selected families (i.e., among- and within-family selection) while correcting for the influence of soil heterogeneity on plant performance. Given the fact that Spacevar was developed using both among-family selection (i.e., superior 5 out of 45 families or 11% selection intensity) and within-family selection (25% within family selection intensity) as opposed to Swardvar that was developed from parental clones of five superior performing families out of 14 evaluated (i.e., among-family selection or 36% selection intensity), HSSpace did not show any advantage over HSSward in biomass yield improvement. Surprisingly none of the selected families was common between the two methods. One of the 14 families evaluated in HSSward was among the superior five families selected under HSSpace method, but that family was not one of five selected under HSSward performance. This suggests a likely contrast between the two methods in ranking half-sib family performance. The similar or low biomass yield gain of HSSpace method compared to HSSward method could be partly due crop science, vol. 53, march april 2013

8 to low plant heritability for biomass yield in switchgrass. Although the estimate of plant heritability for the population used to derive these two varieties is not available, the results from studies in switchgrass showed low plant heritability for biomass yield (Bhandari et al., 2010; Godshalk et al., 1986; Talbert et al., 1983). According to Casler and Brummer (2008), higher plant heritability is required to realize a greater response to selection using AWF-HS compared with that using HSPT. With low plant heritability, within-family selection (i.e., phenotypic selection) based on individual plant performance in space-plant nursery can be influenced by environment impeding selection gain. Alternatively, the additional gain that could have been achieved, if any, following among- and within-family selection using HSSpace method may not have been realized when such cultivars were evaluated under sward-plot conditions. It was reported that expression of heterosis in hybrids between upland Summer lowland Kanlow variety of switchgrass was observed when such hybrids were evaluated in sward plot, which was not found under space-plant evaluation (Martinez-Reyna and Vogel, 2008; Vogel and Mitchell, 2008). In tall fescue (Festuca arundinacea Schreb.), indirect selection using space-plant biomass yield performance was found less predictive of the biomass yield under sward (Waldron et al., 2008). In cereal rye (Secale cereale L.), the forage yield gain made over several cycles of recurrent phenotypic selection under space-plant evaluation could not be realized when the variety was evaluated under sward condition (Bruckner et al., 1991). In maize (Zea mays L.), Kesornkeaw et al. (2009) reported that the response to selection for ear number was 60% greater when selected under high compared to low plant density. Differences in yield gain of two experimentals between Tifton and other locations could also have arisen due to the use of different seed lots. The trial at Tifton was conducted with c1syn-2 seed of Spacevar and c1syn-1 seed of Swardvar while the trials elsewhere were conducted with c1syn-2 seeds. It is not understood if the low yield of Spacevar compared to Swardvar at Tifton site was associated with use of c1syn-2 in place of c1syn-1 that could have resulted in loss of heterosis. Use of different seed lots of Alamo as check could also have led to differences in yield gain between locations. Tifton used the original seed lot of Alamo from which these two experimentals were derived, thus making direct comparison more sensible. The trials conducted elsewhere used Alamo seeds from commercial sources. It was not clear if these Alamo seed lots have undergone some level of natural and/or artificial selection. The current results demonstrated that a single cycle of selection could produce biomass yield gain as high as 30% when selection is performed in environments proximal to the production environments. This demonstrated the tremendous potential for biomass yield improvement in switchgrass. Such results could reflect the shift in population genetic structure with respect to favorable and/or unfavorable genes. Such yield gain, however, could be reduced dramatically as these varieties were evaluated farther from the selection environment (Tables 1, 2, 3, 4, and 5). In fact, the latitude among test locations varied less (31 34 N) than their longitudinal differences (83 91 W) so that the locations were generally in the same latitudinal hardiness zones but created an east to west gradient that moved from the area of selection to geographies farther away from the selection area. Differences in yield gain among locations that used a common seed lot of Alamo (i.e., locations other than Tifton) indicated the presence of cultivar location interaction. Several studies in the past have demonstrated the presence of significant genotype environment interaction for biomass yield in switchgrass (Hopkins et al., 1995). This suggests the need for multilocation evaluation of selectionnursery- or plant-adaptation-region-based breeding programs (Casler et al., 2007b) while establishment of regional gene pools and breeding according to plant adaptation regions could help achieve the higher biomass yield gain in the target regions (Casler et al., 2007a, 2007b). The HSSward method appeared to produce a relatively more stable synthetic cultivar compared to HSSpace. The family evaluation to develop Spacevar was performed in a single location at Athens, GA, which could have favored selection of an environment-specific variety, thereby losing the plasticity to adaptation across broader geographic areas. Development of Swardvar involved family evaluation in two environments in Athens and Tifton, GA, that could have favored the development of a variety with greater plasticity to environmental adaptation. In a different study, in which 150 half-sib families of lowland switchgrass were evaluated for biomass yield, there was no correlation (r < 0.1) between the Tifton and Athens sites (results not shown). The above discussion indicates that HSSward is still the superior method compared to HSSpace in terms of producing relatively stable variety with similar or better selection gain. However, the HSSpace method completes the selection cycle a year earlier than the HSSward method. This was possible because field evaluation of half-sib families following HSSpace required smaller seed quantities that were produced by polycrossing initially selected 45 genotypes in the greenhouse in the winter through spring 1996/1997; thus, the field evaluation of half-sib families could be initiated in In HSSward method, polycrossing to generate half-sib family seeds was performed in the field in fall 1997 because replicated sward-plot evaluation with HSSward would need large amount of seeds, which could not be achieved in greenhouse space meaning the first family evaluations could not commence until Consequently, per year yield gain could actually be equal or higher when following the HSSpace method compared with the conventional HSSward method. Moreover, HSSpace is more resource efficient compared to HSSward (Table 6). Further crop science, vol. 53, march april

9 Table 6. Amount of selection effort (i.e., number of space plants or sward plots evaluated) used for conventional swardplot versus space-plant evaluation of half-sib families in development of experimental synthetic of lowland switchgrass. Selection method Number of space plants Evaluation efforts Number of sward plots Swardvar Spacevar study is needed to evaluate if HSSpace selection under multiple environments could produce varieties that are as high yielding and stable as varieties produced following HSSward. If the breeding objective is to speed up the selection cycle under finite resources, HSSpace evaluation could be the method of choice for switchgrass cultivar development especially when targeting a specific geographic region. Acknowledgments We are thankful to Dr. Basoondat Macoon, Central Mississippi Research Station, Mississippi State University, Raymond, MS; Mr. Jimmy Ray Parish, Department of Plant and Soil Sciences, Mississippi State University, Starkville, MS; Dr. Monte Roquette and Mr. Joel Kerby, Texas Agrilife Research and Extension Center, Overton, TX; Mr. Donald Wood, University of Georgia, Athens, GA; and Mr. Brian Motes, Forage Improvement Division, the Samuel Roberts Noble Foundation, Inc., Ardmore, OK, for their invaluable contributions in conducting the field trials at their respective locations. We also like to thank Dr. Mark Newell, Forage Improvement Division, the Samuel Roberts Noble Foundation, Inc., for his critical review of this manuscript. This study was supported in its initial stages by funding from the Oak Ridge National Lab and UT-Battelle for the Department of Energy (SUB-02-19XSV810C/01) and in its later stages via a collaborative work research agreement between the Samuel Roberts Noble Foundation, Inc. and Ceres, Inc. References Abraham, E.M., and A.C. Fasoulas Comparative efficiency of three selection methods in Dactylis glomerata L. and Agropyron cristatum L. J. Agric. Sci. 137: doi: / S Adler, P.R., M.A. Sanderson, A.A. Boateng, P.J. Weimer, and H.G. Jung Biomass yield and biofuel quality of switchgrass harvested in fall or spring. Agron. J. 98: doi: / agronj Anderson-Teixeira, K.J., S.C. Davis, M.D. Masters, and E.H. Delucia Changes in soil organic carbon under biofuel crops. Glob. Change Biol. Bioenergy 1: doi: /j x Bhandari, H.S., M.C. Saha, P.N. Mascia, V.A. Fasoula, and J.H. Bouton Variation among half-sib families and heritability for biomass yield and other traits in lowland switchgrass (Panicum virgatum L.). Crop Sci. 50: doi: / cropsci Blanco-Canqui, H Energy crops and their implications on soil and environment. Agron. J. 102: doi: / agronj Bruckner, P.L., P.L. Raymer, and G.W. Burton Recurrent phenotypic selection for forage yield in rye. Euphytica 54: doi: /bf Burton, G.W Registration of Tifton 9 Pensacola bahiagrass. Crop Sci. 29:1326. doi: /cropsci x x Casler, M.D., and E.C. Brummer Theoretical expected genetic gains for among-and-within-family selection methods in perennial forage crops. Crop Sci. 48: doi: / cropsci Casler, M.D., C.A. Stendal, L. Kapich, and K.P. Vogel. 2007a. Genetic diversity, plant adaptation regions, and gene pools for switchgrass. Crop Sci. 47: doi: /cropsci Casler, M.D., K.P. Vogel, C.M. Taliaferro, N.J. Ehlke, J.D. Berdahl, E.C. Brummer, R.L. Kallenbach, C.P. West, and R.B. Mitchell. 2007b. Latitudinal and longitudinal adaptation of switchgrass populations. Crop Sci. 47: doi: / cropsci Fasoula, D.A., and V.A. Fasoula Competitive ability and plant breeding. Plant Breed. Rev. 14: Fasoula, V.A., and D.A. Fasoula Honeycomb breeding: Principles and applications. Plant Breed. Rev. 18: Fasoula, V.A., and D.A. Fasoula Principles underlying genetic improvement for high and stable crop yield potential. Field Crops Res. 75: doi: /s (02) Fasoulas, A.C., and V.A. Fasoula Honeycomb selection design. In: J. Janick, editor, Plant breeding reviews, Vol. 13. John Wiley & Sons, Hoboken, NJ. p Godshalk, E.B., J.C. Burns, and D.H. Timothy Selection for in-vitro dry matter disappearance in switchgrass panicum-virgatum regrowth. Crop Sci. 26: doi: /cropsci X x Hopkins, A.A., C.M. Taliaferro, C.D. Murphy, and D. Christian Chromosome number and nuclear DNA content of several switchgrass populations. Crop Sci. 36: doi: /cropsci x x Hopkins, A.A., K.P. Vogel, K.J. Moore, K.D. Johnson, and I.T. Carlson Genotype effects and genotype by environment interactions for traits of elite switchgrass populations. Crop Sci. 35: doi: /cropsci x x Hultquist, S.J., K.P. Vogel, D.J. Lee, K. Arumuganathan, and S. Kaeppler Chloroplast DNA and nuclear DNA content variations among cultivars of switchgrass, Panicum virgatum L. Crop Sci. 36: doi: /cropsci x x Hultquist, S.J., K.P. Vogel, D.J. Lee, K. Arumuganathan, and S. Kaeppler DNA content and chloroplast DNA polymorphisms among switchgrasses from remnant midwestern prairies. Crop Sci. 37: doi: /cropsci x x Kesornkeaw, P., K. Lertrat, and B. Suriharn Response to four cycles of mass selection for prolificacy at low and high population densities in small ear waxy corn. Asian J. Plant Sci. 8: doi: /ajps Liebig, M.A., M.R. Schmer, K.P. Vogel, and R.B. Mitchell Soil carbon storage by switchgrass grown for bioenergy. Bioenergy Res. 1: doi: /s crop science, vol. 53, march april 2013

10 Lu, K., S.W. Kaeppler, K.P. Vogel, K. Arumuganathan, and D.J. Lee Nuclear DNA content and chromosome numbers in switchgrass. Great Plains Research: A journal of natural and social sciences. University of Nebraska, Lincoln 22: Martinez-Reyna, J.M., and K.P. Vogel Heterosis in switchgrass: Spaced plants. Crop Sci. 48: doi: / cropsci McLaughlin, S.B., and L.A. Kszos Development of switchgrass (Panicum virgatum) as a bioenergy feedstock in the United States. Biomass Bioen. 28: doi: /j.biombioe McMillan, C., and J. Weller Cytogeography of Panicum virgatum in central North America. Am. J. Bot. 46: doi: / Missaoui, A., V. Fasoula, and J. Bouton The effect of low plant density on response to selection for biomass production in switchgrass. Euphytica 142:1 12. doi: /s y Missaoui, A.M., A.H. Paterson, and J.H. Bouton Molecular markers for the classification of switchgrass (Panicum virgatum L.) germplasm and to assess genetic diversity in three synthetic switchgrass populations. Genet. Resour. Crop Evol. 53: doi: /s Narasimhamoorthy, B., M.C. Saha, B. Swaller, and J.H. Bouton Genetic diversity in switchgrass collections assessed by EST-SSR markers. Bioenergy Res. 1: Riley, R.D., and K.P. Vogel Chromosome numbers of released cultivars of switchgrass, indiangrass, big bluestem, and sand bluestem. Crop Sci. 22: SAS Institute The SAS system version 9.3. SAS Inst., Cary, NC. Tai, G.C.C Genotypic stability analysis and its application to potato regional trials. Crop Sci. 11: doi: /crops ci x x Talbert, L.E., D.H. Timothy, J.C. Burns, J.O. Rawlings, and R.H. Moll Estimates of genetic parameters in switchgrass. Crop Sci. 23: doi: /cropsci x x Taliaferro, C.M., K.P. Vogel, J.H. Bouton, S.B. McLaughlin, and G.A. Tuskan Reproductive characteristics and breeding improvement potential of switchgrass. In: R.P. Overend and E. Chornet, editors, Biomass A growth opportunity in green energy and value added products. Proceedings of the Fourth Biomass Conference of the Americas, Oakland, CA. 29 Aug. 2 Sept Elsevier Science Ltd., Kidlington, Oxford, UK. p Thillainathan, M., and G.C.J. Fernandez SAS applications for Tai s stability analysis and AMMI model in genotype environmental interaction (GEI) effects. J. Hered. 92: doi: /jhered/ Vogel, K.P Switchgrass. In: L.E. Moser, B.L. Burson, and L.E. Sollenberger, editors, Warm-season (C4) grasses. ASA, CSSA, and SSSA, Madison, WI. p Vogel, K.P., and R.B. Mitchell Heterosis in switchgrass: Biomass yield in swards. Crop Sci. 48: doi: / cropsci Waldron, B.L., J.G. Robins, M.D. Peel, and K.B. Jensen Predicted efficiency of spaced-plant selection to indirectly improve tall fescue sward yield and quality. Crop Sci. 48: doi: /cropsci crop science, vol. 53, march april