Published in Agron J 99:1137-1142 (2007)
DOI: 10.2134/agronj2006.0291
© 2007 American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA
Roots
Genetic Variation and Genotype x Environment Interaction for Yield and Other Agronomic Traits in Cassava in Nigeria
C. N. Egesia,*,
P. Ilonab,
F. O. Ogbea,b,
M. Akorodab and
A. Dixonb
a National Root Crops Res. Inst. (NRCRI), Umudike, PMB 7006, Umuahia, Abia State, Nigeria
b Int. Inst. for Tropical Agric. (IITA), Oyo Rd., PMB 5320, Ibadan, Oyo State, Nigeria
* Corresponding author (cegesi{at}yahoo.com)
Received for publication October 19, 2006.
 |
ABSTRACT
|
|---|
The identification of superior genotypes and mega-environments on the basis of multiple traits is a key objective of multi-environment trials in cassava (Manihot esculenta Crantz). The objective of this study was to examine the genetic variation and genotype x environment interaction (GEI) effects for fresh root yield, six other agronomic traits, and severity ratings for cassava mosaic disease (CMD) and cassava green mite (CGM) in 40 genotypes of cassava. Experiments were conducted using a randomized complete-block design with four replications for 2 yr in three representative agro-ecological zones in Nigeria. Site regression (SREG) analysis revealed that GEI was a major source of fresh root yield variation and the different testing sites discriminated among the genotypes. Genotypes TMS 98/0581, TMS 97/4763, TMS 98/0002, TMS 99/3073, and M98/0068 were highest yielding at Otobi and Umudike, whereas TMS 98/0510, TMS 97/4779, and TMS 92B/00068 yielded the most at Ishiagu. TMS 98/2226, TMS 92/0325, and M98/0028 had the poorest performance across all locations. Genotypes with the highest yield showed the lowest CMD scores, whereas very tall (well above 2 m) plants had low harvest index on the basis of multiple trait analysis. We identified optimally adapted genotypes for commercial cassava production in different areas in Nigeria.
Abbreviations: CGM, cassava green mite severity • CMD, cassava mosaic disease severity • E, environment main effect • G, genotype main effect • GEI, genotype x environment interaction • GGE, genotype plus genotype x environment interaction • GL, genotype x location interaction • GT, genotype x trait interaction • GY, genotype x year • GYL, genotype x year x location interaction • L, location main effect • MET, multi-environment trials • PC, principal component • SREG, site regression • Y, year main effect • YL, year x location interaction
|
INTRODUCTION
|
|---|
CASSAVA is a most important food staple in Africa. For instance, 50% of Nigeria's population (ca. 70 million) eats cassava at least once a day (Phillip et al., 2005). In the rest of sub-Saharan Africa, the same scenario applies to as many as 30 to 70% (ca. 400 million people) of the region's inhabitants (FAOSTAT, 2005). Cassava is also assuming a new role not only as an important animal feed and industrial raw material, but also in the emerging biofuel economy.
The success of cassava in Africa, as a food security crop, is largely because of its ability and capacity to yield well in drought-prone, marginal wastelands under poor management where other crops would fail. However, large differential genotypic responses occur under varying environmental conditions despite cassava's ability to grow in marginal areas (Bokanga et al., 1994; Mkumbira et al., 2003). This phenomenon is referred to as genotype x environment interactions (GEI), which is a routine occurrence in plant breeding programs (Kang, 1998). The main objective of most multi-environment trials (MET) is to identify superior genotypes and locations that best represent production environments. This approach leads to higher heritability of traits because of increased efficiency of predictable gains of traits under selection, which ensures rapid progress in breeding programs. Traditionally, plant breeders tend to select genotypes that show stable performance as defined by minimal GEI effects across a number of locations and/or years. Several statistical tools to explore GEI data generated from MET exist (Kang, 1998), but among the most suitable approaches are those based on multiplicative and factorial regression models (Crossa and Cornelius, 1997). Genotypic evaluation based on multiple traits is used sparingly in plant breeding due to the challenge of complex relationships among traits. However, breeders would like to identify genotypes as packages of desirable agronomic and economic traits (Yan and Kang, 2003). The objective of this study was to examine the genetic variation and genotype x environment interaction (GEI) effects for fresh root yield, six other agronomic traits, and severity ratings for cassava mosaic disease (CMD) and cassava green mite (CGM) in 40 genotypes of cassava grown in three diverse agro-ecologies in Nigeria in 2 yr via SREG and multiple-trait data analysis.
|
MATERIALS AND METHODS
|
|---|
Experimental Data
The study was conducted in 2003 and 2004 at three sites: Umudike (with annual rainfall of 2200 mm; altitude 120 m; mean annual temperature of 22 to 31°C; coordinates 7°24' E, 5°29' N; Dystric Luvisol soils; humid forest); Ishiagu (with annual rainfall of 1800 mm; altitude 150 m; mean annual temperature of 24 to 32°C; coordinates 7°31' E, 5°56' N; Dystric Luvisol soils; forest–savanna transition); and Otobi (with annual rainfall of 1500 mm; altitude 319 m; mean annual temperature of 24 to 35°C; coordinates 7°20' E, 8°41' N; Ferric Luvisol soils; southern Guinea savanna) in Nigeria. These sites represent the major cassava-growing agro-ecological zones in the country. Thirty-seven cassava genotypes at advanced stages of breeding and three checks were used in the study. The checks, TMS 30572, TMS 4(2)1425, and TMS 82/00058, were improved cultivars and widely grown in Nigeria and other cassava-growing areas of Africa because of their outstanding agronomic performance and moderate resistance to major pests and diseases. The experimental design was a randomized complete block design with four replications at each site under rainfed conditions. Each plot consisted of 36 plants in six 6-plant rows. The ridges were 1 m apart, 30 cm high, and 6 m long. Spacing between plants was 1 m, giving a total plant population of 10 000 plants ha–1. Fertilizer (N–P–K, 15–15–15) was applied at each site at 2 mo after planting (MAP) at the rate of 600 kg ha–1. Pre-emergence herbicide as recommended and hand weeding were used to control weeds.
Data were collected from the inner 24 plants within a plot 12 MAP on fresh root yield (Mg ha–1), number of roots per plot, fresh top biomass (Mg ha–1), number of 1-m long stems per plot, number of main stems per plot, and plant height (m). Harvest index was measured as the ratio of fresh root weight to total biomass. Other parameters included severity ratings of CMD taken at 1, 3, and 5 MAP and CGM taken twice at the peak of dry season between January and March of each year [scale of 1–5; where 1 = no apparent symptoms, and 5 = very severe symptoms according to IITA (1990)]. The data for CMD and CGM were converted to a severity index based on the cumulative area under the disease or pest progress curve as outlined by Shaner and Finney (1977).
Genotype x Environment Interaction Analysis
The SREG model was used to analyze the GEI patterns using the SAS software (SAS Institute, 2001; Burgueño et al., 2001). The program was used to generate a genotype (G) plus GE (GGE) biplot that was constructed from the first two principal components (PC1 and PC2) derived by subjecting the environment-centered yield data (which contains G and GE) to singular value decomposition (SVD) (Yan et al., 2000). The SREG linear-bilinear model is represented by:
 | [1] |
where 
ij is the mean of the ith genotype in the jth environment for g genotypes and e environment (i = 1, 2, ..., g and j = 1, 2, ..., e); µ is the overall mean; 
j is the environment effect; 
k (1 
2 ... t) are scaling constants (singular values) that allow the imposition of orthonormality constraints on the singular vectors for genotypes, 
ik (ik, ..., gk) and environments, 
jk (1k, ..., ek); ik and jk for k = 1, 2, 3, ... are called "primary," "secondary," "tertiary," effects of the ith genotype and jth environment, respectively; 
ij. is the residual error assumed to be normally and independently distributed (0, 
2/r) (where 2 is the pooled error variance and r is the number of replicates).
Genotype x Trait Analysis
The genotype x trait (GT) biplot method outlined in Yan and Rajcan (2002) was used to display the genotype x trait data in a biplot. The model for the GT biplot is as follows:
 | [2] |
where Tij is the average value of genotype i for trait j; 
j is the average value of all genotypes for trait j, and Sj is the standard deviation of trait j among the genotype averages; 
in and 
jn are scores for genotype i and trait j on PCn (n = 2), respectively, and ij is the residual associated with genotype i for trait j. Because different traits had different units, standardization was needed to remove the units. In this biplot constructed by plotting PC1 against PC2, vectors were drawn from the origin of the biplot to each trait to visualize relationships among traits. Coefficient of correlation (r) between any two traits is approximated by the cosine of the angle between their vectors (Yan and Rajcan, 2002) (e.g., r = cos180° = –1, cos0° = 1, and cos90° = 0). The GT biplot was used to compare genotypes on the basis of multiple traits and to identify genotypes that possessed several desirable traits.
|
RESULTS
|
|---|
Analyses of Variance and Agronomic Performance
Combined analysis of variance of fresh root yield of the 40 genotypes tested at the three locations across 2 yr showed significant interaction (P < 0.0001) for all the sources of variation except the year (Y) effect (Table 1). The analysis indicated that 15.8% of the total sum of squares was attributable to location (L), 19.58% to year x location (YL), 20% to G, 14.83% to genotype x location (GL) interaction, and 9.1% to genotype x year (GY) interaction. Thus, there were no substantial differences in climatic variables during the study years. The GL effect contributed 14.93% to the total variation, indicating fluctuations in genotypic responses to the different environments.
View this table:
[in this window]
[in a new window]
|
Table 1. The combined analysis of variance for mean fresh root yield (Mg ha–1) in 40 genotypes of cassava grown at three locations in Nigeria in 2003 and 2004.
|
|
The by-year analyses of variance revealed that in each year, the L, G, and GL effects were statistically significant (P < 0.001) (Table 2). The GL effect was quite substantial, capturing a third of the total variation in each year. The variation caused by the GL effect was of similar magnitude as that of the G effect. Though the variation among locations was relatively low in 2003 (22.28%), it was 45% of the total variation in 2004.
View this table:
[in this window]
[in a new window]
|
Table 2. Genotype (G), location (L), and genotype x location (GL) variance terms for mean fresh root yield (Mg ha–1) of 40 genotypes of cassava grown at three locations in Nigeria in 2003 and 2004.
|
|
Significant differences (P < 0.0001) among locations were found for all the agronomic parameters measured (Table 3), which showed a very wide range of mean values. The average fresh root yield (Mg ha–1) was 17.28, 22.24, and 16.49 at Ishiagu, Otobi, and Umudike, respectively. Top biomass yield was significantly higher at Otobi (13.75 Mg ha–1) than at Ishiagu (8.88 Mg ha–1) and Umudike (9.99 Mg ha–1). Harvest index was 0.67 at Ishiagu and 0.62 at Otobi and Umudike. The CMD severity was higher at Umudike (10.35) than at Ishiagu (7.65) and Otobi (8.05). The CGM severity was 11.75 at Ishiagu, whereas it was 13.5 at Otobi and 13.25% at Umudike.
View this table:
[in this window]
[in a new window]
|
Table 3. Mean, standard error (SE), range and coefficient of variation (CV) for mean fresh root yield (Mg ha–1) and other agronomic characters in 40 cassava genotypes grown at three locations in Nigeria in 2003 and 2004.
|
|
Genotype plus Genotype x Environment (GGE) Biplot Analysis
The first two principal components (PC1 and PC2) obtained by SVD of the location-centered data (SREG model) explained 84.86% of the total variability attributable to G + GEI for both years (Fig. 1). There was a relatively high correlation of PC1 (r = 0.74, P < 0.0001) and no relationship of PC2 (r = 0.01, P = 0.932) with fresh root yield.
View larger version (20K):
[in this window]
[in a new window]
|
Fig. 1. Genotype plus genotype x environment (GGE) biplot based on the fresh root yield (Mg ha–1) data of 40 genotypes of cassava grown at three locations in Nigeria in 2003 and 2004. The vertex genotype markers located away from the plot origin were connected to form a polygon. Test sites are given in block letters.
|
|
The biplot enabled visual comparison of the locations and genotypes studied and their interrelationships. Genotypes TMS 98/0510, TMS 97/4779, and TMS 92B/00068 were identified as the highest yielding genotypes at Ishiagu, as they were grouped together in one sector of the polygon. The "winning" genotypes at Umudike and Otobi were TMS 98/0581, TMS 97/4763, TMS 98/0002, TMS 99/3073, and M98/0068. The genotypes TMS 98/2226, TMS 92/0325, and M98/0028 were the poorest yielding, as they were located farthest from all location markers.
As indicated by the relative length of its vector, the Otobi location had the highest PC1 scores. This indicated that the site was best for identifying high yielding genotypes. Umudike had the lowest PC2 score, as indicated by the short vector and represented an average tester. The Ishiagu location had the highest PC2 score, indicating it was not highly representative of other locations.
Genotype x Trait (GT) Biplot and Trait Relations Analyses
The GT biplot (Fig. 2) was used to compare genotypes on the basis of multiple traits and to identify genotypes that were particularly desirable relative to several traits. The proportion of total variation explained by the first two PC axes was 55.4%. The most outstanding relationships revealed by the biplot were: (i) a strong negative association between CMD and fresh root yield, as indicated by the large obtuse angles between their vectors, and between CMD and number of roots; (ii) a strong negative association between plant height and CGM and between plant height and harvest index; (iii) a positive association between fresh root yield and number of roots per plot, and between fresh root yield and number of main stems, as indicated by the acute angle between them; and (iv) a near-zero correlation between plant height and fresh root yield, as indicated by the near-perpendicular vectors. Similar to the GGE biplot, a scatter-plot was drawn for the GT biplot. Genotypes TMS 98/0581, TMS 97/4763, and M98/0068, among others, had the largest PC1 scores and were placed very close to root yield, number of roots per plot, and number of main stems per plot. The same genotypes were shown to have low CMD scores. This is in contrast to genotypes TMS 4(2)1425, TMS 94/0561, TMS 99/6012, and TMS 92/0325, which were very closely associated with CMD disease and had the lowest yields. Also, TMS 99/3073 and TMS 4(2)1425 had high severity scores for CGM and large harvest index but relatively shorter plants. The correlation coefficients among the agronomic traits showed that the GT biplot displayed a relatively reliable relationship between the traits with large PC1 and PC2 scores (Table 4). However, while the GT biplot described the general pattern of trait interrelationships, the correlation analysis showed relationship between pairs of traits and as such might not show the exact associations as the biplot explained only 55% of the total variation.
View larger version (26K):
[in this window]
[in a new window]
|
Fig. 2. Genotype x trait (GT) biplot based on forty genotypes of cassava evaluated for mean fresh root yield (RTYLD), mean number of roots per plot (NOROOT), agronomic traits (harvest index [HARVESTI], plant height [PLTHT], number of main stems per plot [NOSTM], number of 1 m long stems [LONGSTM], fresh top biomass yield [BIOMYLD]), and biotic stresses (severity of cassava mosaic disease [CMD] and severity of cassava green mite [CGM]) at three locations in Nigeria in 2003 and 2004.
|
|
View this table:
[in this window]
[in a new window]
|
Table 4. Correlation coefficients (n = 40) among agronomic traits of 40 genotypes of cassava measured at three locations in Nigeria in 2003 and 2004.
|
|
|
DISCUSSION
|
|---|
Different genotypes produced highest yields in different locations. The substantial G x Y x L interaction effect indicated differential genotypic responses to the different environments. This phenomenon reduced the usefulness of cultivar means, as cultivar performance was confounded by noise rather than the pattern of the G main effect (Pham and Kang, 1988). This necessitated a thorough investigation of yield, adaptation, and stability potentials in MET, which should lead to improved repeatability and heritability of important traits (Becker and Leon, 1988). The partitioning of variance components showed that GEI was mainly due to predictable environmental factors (locations), as indicated by its significance as opposed to the unpredictable factors (years). When GEI is caused by predictable environmental factors, the ultimate goal should be to develop specific cultivars for specific environments (Adugna and Labuschagne, 2002), rather than breeding for stable cultivars that will perform well across all target environments. These location-dependent differences may be attributed to environmental variables, such as temperature, rainfall, altitude, and soil characteristics of the test sites. The high significant correlation between PC1 and the fresh root yield indicated that the PC1 axis explained most of the variation observed in the data set and thus PC1 scores could effectively represent the G main effect (Yan et al., 2001; Crossa et al., 2002).
The biplot enabled a visual comparison of the locations and genotypes, and their interrelationships. The vertex genotypes were the most responsive to the closest environment(s) and represented either the best or the poorest performers at some or all locations, having appeared farthest from the biplot origin (Yan et al., 2000). However, different genotypes emerged as "winners" in different locations, suggesting two mega-environments. The genotypes TMS 98/0510, TMS 97/4779, and TMS 92B/00068 were identified as "winners" at Ishiagu—the first mega-environment. The winning genotypes in the second mega-environment (Umudike and Otobi) were TMS 98/0581, TMS 97/4763, TMS 98/0002, TMS 99/3073, and M98/0068. The genotypes TMS 98/2226, TMS 92/0325, and M98/0028 were the poorest yielding in all environments.
The GT biplot allowed comparative evaluation of genotypes for multiple traits and helped identify genotypes that were desirable relative to several traits. The relatively low proportion of the first two PC axes of 55.4% reflects the complexity of the interrelationships of the traits measured (Kroonenberg, 1995). The biplot provided good insight into the pattern of associations of the traits. The very strong negative association between CMD and fresh root yield agreed with many reports of the devastating effects of the disease on productivity (Thresh et al., 1994, 1997; Owor et al., 2004). The differences in productivity from one genotype to another could arise principally during plant development due to differential resistance to diseases, for example, a negative correlation of CMD with yield. Fresh root yield in cassava is computed by multiplying biomass with harvest index (Kawano et al., 1998). However, the harvest index is considered to be a more reliable tool for indirect selection for yield in advanced stages of cassava cultivar selection than biomass (Cock, 1983; Kawano et al., 1998). Similar to the GGE, the GT biplot identified genotypes TMS 98/0581, TMS 97/4763, and M98/0068 among those that efficiently combined high yield and related agronomic traits. The GT biplot was useful because it allowed genotype comparisons on the basis of multiple traits, such that genotypes particularly good in some traits can be used as parents in a breeding program (Yan and Rajcan, 2002).
The biplots displayed pattern of variability of the genotypes, the locations, and their interactions. Interrelationships among agronomic characteristics allowed identification of optimal genotypes for several traits. The high yielding genotypes with good agronomic attributes have potential applications in commercial cassava production in Africa.
|
ACKNOWLEDGMENTS
|
|---|
The authors thank the Federal Government of Nigeria, the State Governments of the Niger Delta areas of Nigeria, the Niger Delta Development Commission (NDDC), Nigerian National Petroleum Corporation (NNPC), Shell Petroleum Development Company (SPDC), and the U.S. Agency for International Development (USAID) for co-funding the Integrated Cassava Project.
|
REFERENCES
|
|---|
- Adugna, W., and M.T. Labuschagne. 2002. Genotype–environment interactions and phenotypic stability analyses of linseed in Ethiopia. Plant Breed. 121:66–71.
- Becker, H.C., and J. Leon. 1988. Stability analysis in plant breeding. Plant Breed. 101:1–23.
- Bokanga, M., I.J. Ekanayake, A.G.O. Dixon, and M.C.M. Porto. 1994. Genotype–environment interactions for cyanogenic potential in cassava. Acta Hortic. 375:131–139.
- Burgueño, J., J. Crossa, and M. Vargas. 2001. SAS programs for graphing GE and GGE biplots. SAS Inst., Cary, NC.
- Cock, J.H. 1983. Cassava. p. 341–360. In Potential productivity of field crops under different environments. IRRI, Los Banos, Philippines.
- Crossa, J., and P.L. Cornelius. 1997. Sites regression and shifted multiplicative model clustering of cultivar trial sites under heterogeneity of error variances. Crop Sci. 37:405–415.
- Crossa, J., P.L. Cornelius, and W. Yan. 2002. Biplots of linear-bilinear models for studying crossover genotype–environment interaction. Crop Sci. 42:619–633.[Abstract/Free Full Text]
- FAOSTAT. 2005. Statistical database of the Food and Agricultural Organization of the United Nations. Available at http://faostat.fao.org/ (accessed 14 Aug. 2006; verified 2 Apr. 2007). FAO, Rome.
- IITA. 1990. Cassava in tropical Africa. A reference manual. Chayce Publication Services, UK.
- Kang, M.S. 1998. Using genotype-by-environment interaction for crop cultivar development. Adv. Agron. 62:199–252.
- Kawano, K., K. Narintaraporn, P. Narintaraporn, S. Sarakarn, A. Limsila, J. Limsila, D. Suparhan, V. Sarawat, and W. Watananonta. 1998. Yield improvement in a multistage breeding program for cassava. Crop Sci. 38:325–332.[Abstract/Free Full Text]
- Kroonenberg, P.M. 1995. Introduction to biplots for G x E tables. Res. Rep. 51. Dep. of Mathematics, Univ. of Queensland, Australia.
- Mkumbira, J., N.M. Mahungu, and U. Gullberg. 2003. Grouping locations for efficient cassava evaluation in Malawi. Exp. Agric. 39:167–179.
- Owor, B., J.P. Legg, G. Okao-Okuja, R. Obonyo, and M.W. Ogenga-Latigo. 2004. The effect of cassava mosaic geminiviruses on symptom severity, growth and root yield of a cassava mosaic virus disease-susceptible cultivar in Uganda. Ann. Appl. Biol. 145:331–337.[Web of Science]
- Pham, H.N., and M.S. Kang. 1988. Interrelationships among repeatability of several stability statistics estimated from international maize trials. Crop Sci. 28:925–928.[Abstract/Free Full Text]
- Phillip, T.A., D.P. Taylor, L.O. Sanni, M. Akoroda, C. Ezedinma, R. Okechukwu, A. Dixon, E. Okoro, and G. Tarawali. 2005. The Nigerian cassava industry statistical handbook. IITA, Nigeria. Int. Inst. of Tropical Agric., Ibadan, Nigeria.
- SAS Institute. 2001. SAS system for Windows. Version 8.1. SAS Inst., Cary, NC.
- Shaner, G., and R.E. Finney. 1977. The effect of nitrogen fertilization on the expression of slow-mildewing resistance in Knox wheat. Phytopathology 67:1051–1056.[Web of Science]
- Thresh, J.M., D. Fargette, and G.W. Otim Nape. 1994. Effects of African cassava mosaic geminiviruese on the yield of cassava. Trop. Sci. 34:26–42.
- Thresh, J.M., G.W. Otim-Nape, J.P. Legg, and D. Fargette. 1997. African cassava mosaic virus disease: The magnitude of the problem. Afr. J. Root Tuber Crops 2:13–19.
- Yan, W., P.L. Cornelius, J. Crossa, and L.A. Hunt. 2001. Two types of GGE biplots for analyzing multi-environment trial data. Crop Sci. 41:656–663.[Abstract/Free Full Text]
- Yan, W., L.A. Hunt, Q. Sheng, and Z. Szlavnics. 2000. Cultivar evaluation and mega-environment investigation based on the GGE biplot. Crop Sci. 40:597–605.[Abstract/Free Full Text]
- Yan, W., and M.S. Kang. 2003. GGE biplot analysis: A graphical tool for breeders, geneticists, and agronomists. CRC Press, Boca Raton, FL.
- Yan, W., and I. Rajcan. 2002. Biplot analysis of test sites and trait relations of soybean in Ontario. Crop Sci. 42:11–20.[Abstract/Free Full Text]
TOP
ABSTRACT
INTRODUCTION
