Crop Science Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 2 October 2006
Published in Crop Sci 46:2368-2375 (2006)
© 2006 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Quero-García, J.
Right arrow Articles by Lebot, V.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Quero-García, J.
Right arrow Articles by Lebot, V.
Agricola
Right arrow Articles by Quero-García, J.
Right arrow Articles by Lebot, V.
Related Collections
Right arrow Crop Genetics
Right arrow Other Tubers

CROP BREEDING & GENETICS

Heritability of the Main Agronomic Traits of Taro

José Quero-Garcíaa,*, Anton Ivancicb, Philippe Letourmyc, Philippe Feldmannd, Tari Molisalee and Vincent Lebotf

a UPM, ETSIA, Dep. de Biotecnología, Ciudad Univ. 28040 Madrid, Spain
b Univ. of Maribor, Faculty of Agriculture, Vrbanska 30, 2000 Maribor, Slovenia
c CIRAD, UPR 13, TA 70/07, 34398 Montpellier cedex 05, France
d CIRAD, Direction of Research, TA 179/04, 34398 Montpellier, cedex 05, France
e VARTC, P.O. Box 231, Santo, Vanuatu
f CIRAD, P.O. Box 946, Port-Vila, Vanuatu

* Corresponding author (jose.quero{at}cirad.fr)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The increase of production of an important root crop such as taro [Colocasia esculenta (L.) Schott, Araceae] implies the development of new varieties through an efficient breeding scheme. The breeding of new taro cultivars is a complex process which requires experience, adequate genetic resources, and reliable data about inheritance of crucial agronomic traits. To obtain these data, 2-yr heritability taro trials were conducted in Vanuatu with 42 full-sib families of variable sizes. In the first clonal generation (C1), a fully randomized complete block design was planted with plots of 16 datum plants. During the second clonal generation (C2), parents were included with their offspring in two alpha incomplete block designs with elementary plots of 12 and 4 datum plants, respectively. Both family and narrow-sense heritabilities were higher for the number of suckers and dry matter content than for corm weight or its components (corm length and width). A high percentage of valuable hybrids were observed among a small number of families. These results, combined with the moderately high family heritability values for several traits, recommend the creation of a small number of large full-sib families when working with a narrow genetic base and the creation of numerous small full-sib families when dealing with a broad genetic base.

Abbreviations: FAO, Food and Agriculture Organization • GA3, gibberellic acid • RCB, randomized complete block • TANSAO, Taro Network for Southeast Asia and Oceania • TLB, taro leaf blight • VARTC, Vanuatu Agricultural Research and Training Centre


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
TARO RANKS FIFTH in the amount harvested among root crops, after potato (Solanum tuberosum L.), cassava (Manihot esculenta Crantz), sweet potato (Ipomoea batatas L.), and yams (Dioscorea spp.). The Food and Agriculture Organization estimates that 9.1 million Mg of corms are produced annually on a surface of 2 million ha, but this largely underestimates production as few countries keep reliable figures (http://faostat.fao.org/). Taro is a highly polymorphic species, vegetatively propagated and cultivated from New Zealand to Japan. It is a predominantly allogamous species characterized by protogyny (Purseglove, 1979). The majority of existing varieties are at least partly heterozygous, which is shown by the high variability within hybrid progenies resulting from crosses among cultivated varieties (Ivancic and Okpul, 1995) or by biochemical and molecular studies, conducted either with isozymes (Lebot and Aradhya, 1991; Ochiai et al., 2001) or with microsatellites (Mace and Godwin, 2002; Noyer et al., 2006).

Taro breeding involves three main steps: (i) creation of genetic variation, (ii) evaluation of the progenies and selection of superior individuals, and (iii) procedures associated with the release of a new variety (Wilson, 1989; Ivancic and Lebot, 2000). The strategy can be very simple (e.g., when the existing varieties are highly valuable and breeding is aimed at improving a few genetically simple traits through conservative selection) or it can be quite complex (e.g., when the breeder has to improve several traits simultaneously).

Several factors complicate taro breeding: flowering may be poor and/or not synchronized; seed germination and raising of young seedlings require special environment and care; intervals among generations are relatively long; reliable sources of genes for some traits do not exist or have not been determined yet (e.g., resistance or tolerance against alomae-bobone virus complex, taro beetle [Papuana spp.], drought, and salinity); and data regarding genetic inheritance and combining ability are scarce (Ivancic and Lebot, 2000).

The first taro breeding programs initiated in the early 1970s in Hawaii, Fiji, and Samoa included only local cultivars and aimed at improving yield and eating quality through simple mass phenotypic selection (Sivan and Tavalqia, 1984; Wilson et al., 1991). The genetic base, which was used in hybridization, was relatively narrow and after the outbreak of taro leaf blight (TLB) caused by Phytophthora colocasiae Racib., in Samoa in 1993 (Chan et al., 1998), the existing breeding programs focused on TLB resistance (Rao, 2000). The first complex breeding programs involving recurrent selection and germplasm from diverse origins, including wild materials resistant to TLB, were initiated in the Solomon Islands and Papua New Guinea (Ivancic and Okpul, 1995). The use of wild genotypes entailed the occurrence of undesired characteristics (formation of stolons, corm acridity) and, after several cycles of recurrent selection, it was difficult to find resistant hybrids with an acceptable quality (Okpul, 2002). More recently, a regional project was launched in Southeast Asia and the Pacific, to collect and exchange elite cultivars, with varying degrees of resistance to TLB (TANSAO, 2002). Once locally evaluated, the most promising of these elite cultivars will be directly released to farmers and used for introgressing resistance genes in local germplasm (Lebot et al., 2004). Along with resistance to TLB, the key agronomic traits to be improved are plant architecture (e.g., optimal number of suckers, absence of stolons, optimal number of leaves and vertical petioles), corm yield, and quality traits such as high dry matter content, corm shape, and low level of acridity.

Very few studies associated with the genetic inheritance of such traits have been conducted. Most of the reported investigations were associated with the performance testing of local or introduced varieties and/or hybrids in randomized complete block designs followed by the analysis of variance (Dwivedi and Sen, 1997) but rarely included the analysis of progenies resulting from controlled crosses. There are two main approaches for the creation of the generation materials to optimize the exploitation of taro genetic variability: (i) genetic recombination of a limited number of parents, forming large full-sib families and (ii) genetic recombination among large numbers of parents, forming small full-sib families. To compare both approaches accurately, we conducted as many crosses as possible but some of them were purposely repeated to ensure a large number of full-sibs.

The main objectives of the present study were: (i) to evaluate which of the above mentioned approaches is the most appropriate; (ii) to produce reliable information about the heritability and the correlations of some of the key agronomic traits; and (iii) to propose practical guidelines for studying quantitative inheritance of taro traits.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Plant Materials
Crosses took place from November 1999 to February 2000. The parental materials were the flowering accessions of the national taro collection on the Island of Espiritu Santo, Vanuatu (despite the use of GA3, not more than 25% of the cultivars flowered normally). Hybridization techniques employed have been described elsewhere (Ivancic et al., 2003, 2004). The total number of successful combinations was 92. Seeds were germinated and, when the plants were strong enough, 2700 F1 hybrid plants were transferred to Vanuatu Agricultural Research and Training Centre (VARTC) in March 2001. The largest 42 families (from 16 to 80 individuals) were selected for the heritability trials.

Field Experiments
The experimental site at VARTC (80 masl, 15°26'7'' S, 167°11'5'' E) has a deep and fertile soil, covering a limestone plateau, 2.8 km from the seashore. The climate of Espiritu Santo is tropical oceanic, averaging 2745 mm rainfall per year (1989–2000).

Trials were performed on two successive seasons (2002 and 2003) and involved the first and the second clonal generations (C1 and C2). In C1, four different randomized complete blocks designs (hereafter called RCB1, RCB2, RCB3, and RCB4) were planted, by choosing the 27 largest families, with respectively 5, 4, 3, and 2 replications. The families with only 16 plants were planted without replications but were also harvested and measured (Table 1). All 42 families were planted on the same plot. Only top parts (the main head sets) were used for planting. Experimental plots were squares of 16 plants. Trials were planted on 20, 21, and 22 Jan. 2002 and were harvested 9 mo after plantation (during the first 3 wk of October).


View this table:
[in this window]
[in a new window]
  		Table 1. Planting materials used for the heritability trials.{dagger}

 
In C2, sufficient quantities of vigorous parental materials were available and this made it possible to include the parents in the trials. The number of genotypes per replication being much higher, alpha (0,1) incomplete block designs were favored, to ensure a better control of the environmental variance (Patterson and Williams, 1976). Two designs were established on the same field. The first design (hereafter called {alpha}1) included four replications and five blocks of nine treatments per block. The experimental unit was composed of 12 plants (12 different full-sibs) of the progenies and four plants (four replicates of the same clone) of the parents. Overall, 19 progenies and 26 parents were used (Table 1). The second design (hereafter called {alpha}2) consisted of four replications and six blocks with 10 treatments per block. The experimental unit was composed of four plants (the same principle was used for hybrids and parents). This trial included the 42 analyzed progenies and 18 parents (Table 1). The latter belonged to the remaining parents not involved in the progenies of {alpha}1. In both alpha designs, several parents were not included in the trial because there was no adequate planting material. Trials were planted on 29 and 30 Jan. 2003 and were harvested during the period 23 Oct. to 13 Nov. 2003 (9 mo after plantation). The spacing in all trials was 0.8 by 0.8 m.

In C1, four quantitative traits were measured: number of stolons, number of suckers, corm weight, and dry matter content. Qualitative traits concerned the corm shape, by differentiating branched and unbranched corms, and several pigmentation traits. In C2, two additional quantitative traits were included: corm length and corm width. All measurements were done on individual plants.

Data Analysis
In C1, family heritabilities were computed through the variance analysis classically applied for a progeny test (Zobel and Talbert, 1991):

Formula 1[1]
where {sigma}2F represented the family variance, {sigma}2W the within-plot variance and {sigma}2FR the family x replication variance. R and P referred to the number of replications and the number of plants per family–replication plot. This formula considers the number of replications and thus provides a value of heritability relative to the experimental design. Given that four different types of RCB were planted, a modified version of this formula was used to compare more accurately the heritability values estimated for each trial:

Formula 2[2]
A mixed model was used, considering the factor replication as a fixed effect and the factors interaction family x replication and family as random effects. The model could be written as:

Formula 3[3]
where i corresponded to family, j to replication, and k to plant; µ was the overall mean, Ai was the family effect (random), ßj the replication effect (fixed), (Aß)ij the interaction family x replication (random) and {varepsilon}ijk the plant effect (random). These two last effects represented the residual error. Variance analysis was performed with the statistical package SAS 8.02 (SAS Institute, 2001) and the MIXED procedure.

In C2, although measurements were taken on individual plants, means for each experimental plot were computed and considered for the analysis. Thus, the heritability values were estimated as following:

Formula 4[4]
where {sigma}2FB represented the treatment x block variance. B referred to the number of blocks per replication.

Since trials {alpha}1 and {alpha}2 were partially balanced, the adjusted means were calculated for each treatment through the LSMEANS procedure of SAS. For this, a fixed effects model was considered and the analysis of variance was conducted through the GLM procedure. For the calculation of family heritabilities, only the genotypic variance related to progenies is informative and thus, to evaluate separately the genotypic effect due to the progenies and to the parents, a factor differentiating these two treatments was included. All progenies were given the same modality and were considered as a random effect within the treatment factor (RANDOM command), whereas parents were treated as fixed effects. The final model could be written as:

Formula 5[5]
where i corresponded to treatment (with value 1 for progenies and value 2 to n for the n – 1 parents), j to progeny, k to replication, and l to block within replication; µ was the overall mean, {alpha}i the treatment effect (fixed), Gij the progeny effect (random), ßk the replication effect (fixed), {gamma}kl the block effect (fixed), and {varepsilon}ijkl the residual effect (random).

For the calculation of narrow-sense heritabilities through the parents–offspring regression, data from {alpha}1 and {alpha}2 were pooled to include the maximum number of progenies in the calculation. This was even more necessary since several parents had not been included in the trials and midparent values could not be computed for all families. The formula could be written as (Falconer, 1991):

Formula 6[6]
where, for each family, Y represented the offspring mean, b the regression coefficient, X the midparent value, and e the residual error. The regression coefficient could be estimated as:

Formula 7[7]
where, for each progeny, Xi represented the individual parental values, Xm the midparent value (over all progenies), Yi the offspring values, and Ym the mean offspring value (over all progenies).

A fixed effects model was considered and the LSMEANS procedure was used again to estimate the adjusted means for all treatments. Since the number of plants per experimental unit was different for the progenies between {alpha}1 and {alpha}2, a weight factor (equal to the number of data finally analyzed per treatment) was introduced. Thus, we considered that the variability of each measure was inversely proportional to the number of measured plants. The model could be written as:

Formula 8[8]
where i corresponded to trial ({alpha}1 or {alpha}2), j to treatment, k to replication within trial, and l to block within replication; µ was the overall mean, {alpha}i was the trial effect (fixed), Aj the treatment effect (fixed), ({alpha}A)ij the interaction trial x treatment effect (fixed), ßik the replication within-trial effect (fixed), {gamma}ikl the block effect (within replication and within trial; fixed), and {varepsilon}ijkl the residual effect (random). LSMEANS by treatment were used to calculate regression coefficients according to Eq. [7].

Phenotypic correlation coefficients were calculated for all trials between all quantitative traits. Phenotypic correlations between qualitative traits, namely pigmentation traits, have been published elsewhere (Ivancic et al., 2003).

Due to their discrete nature, traits number of stolons and number of suckers were highly variable (CV values based on all data calculated on the first clonal generation reached 275 and 100%, respectively) and not normally distributed. The trait number of stolons was too far from a normal distribution and was not included in the heritability calculations. For the number of suckers, however, a root square transformation, as recommended for a discrete variable with a supposed Poisson distribution, allowed us to stabilize the variance and this trait could be included in the heritability calculations. To verify if a significant family effect was observed for the trait number of stolons, the FREQ procedure was applied for each of the RCBs. Chi-square tests were conducted by considering within each family two groups of plants: those that produced and those that did not produce stolons.

During the first clonal generation, agronomic criteria were applied for the selection of high-quality hybrids. Corm weight is classically considered as a complex trait with a high genotype x environment interaction. Moreover, only one plant per genotype was available at this stage. Therefore, only the traits related to plant architecture and dry matter content were considered for this selection. Selected hybrids did not form stolons, developed a minimum number of three suckers, and had a dry matter content superior to 30%, all of these characteristics being highly appreciated by local farmers (Quero-García, 2004). Due to the lack of clonal replicates for each genotype (or full-sib), when considering individual performances for each selected hybrid, it was not possible to separate experimental error (i.e., the environmental effect) from true clone (i.e., genetic) effect. To check if the environmental effects were significant when selecting hybrids within each family, a variance analysis was performed for each RCB for the percentage of selected hybrids from each family plot. Mean percentages fell outside of the 30 to 70% range and an arcsine transformation {{theta} = arcsin [rootsquare (p)], where p is the proportion} was applied, to approximate the normal distribution. The classical linear model was used, by considering the family and replication effects as fixed. The interaction family x replication represented the experimental error.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
First Clonal Generation (C1)
Family Heritabilities
Concerning the trait number of stolons, several families appeared as highly stoloniferous (more than 50% of plants produced stolons) while others did not produce a single stolon. For the four RCBs, chi-square tests were highly significant (Table 2), proving that this undesired trait had a strong family component. This was further confirmed by the observation of parental characteristics, previously evaluated (Quero-García, 2000), which showed that for all families producing a high number of stoloniferous hybrids, at least one of the parents formed stolons.


View this table:
[in this window]
[in a new window]
  		Table 2. Tests of {chi}2 for number of stolons calculated at the first clonal generation.

 
Family heritabilities were relatively high for the three analyzed traits as well as very variable among RCBs (Table 3). Traits number of suckers and dry matter content showed higher heritabilities than corm weight. For the number of suckers, the variances due to the interaction family x replication were low and similar for all trials and the low heritability value obtained for RCB4 was due to a lower family variance (Table 3). On the opposite, family variances were similar for corm weight over the four trials whereas the interaction family x replication variance was relatively high for RCB1 and RCB2, being responsible for a higher residual variance and lower heritability values (Table 3). These interactions were also observed on RCB1 and RCB2 for the dry matter content whereas the high heritability value obtained for RCB4 was mainly due to a very high family variance (Table 3).


View this table:
[in this window]
[in a new window]
  		Table 3. Family heritabilities (H = h2F; Eq. [2]) and estimations of variance for the family (F) and the family x replication (F x R) effects calculated at the first clonal generation.{dagger}

 
Evaluation of Hybrids
Following the selection criteria described, 263 (14.4%), out of the 1824 hybrids harvested, were selected as good hybrids. According to the analysis of variance, the family effect was significant for all RCBs except for RCB2 (Table 4). LSMEANS were calculated for each family, as well as a standard error of these estimations for each of the RCBs. These estimated means, along with the lower and upper limits of their confidence intervals (corresponding to ± the standard error of the transformed means), were transformed into estimated percentages of selected hybrids by converting the angle {theta} to the proportion p by the inverse transformation sin2({theta}) (Fig. 1). Despite relatively large confidence intervals, two groups of families were clearly differentiated on RCB1 (no. 2 and 7 vs. no. 1, 3, 4, 5, 6, 8, 9, and 10), RCB3 (no. 16 and 18 vs. no. 17, 19, 20, and 21) and RCB4 (no. 22, 23, and 25 vs. 24, 26, and 27). Globally, out of the 27 analyzed families, only six families presented an estimated mean percentage of selected hybrids over 30%, whereas 13 families did not reach the 10% value. Interestingly, among the latter, eight families belonged to the largest ones (within RCB1), with 80 evaluated full-sibs.


View this table:
[in this window]
[in a new window]
  		Table 4. Variance analysis of the percentage of selected hybrids per family plot.

 

Figure 1
View larger version (22K):
[in this window]
[in a new window]
  		Fig. 1. Estimated means and confidence intervals (confidence levels of 68%) of the percentage of selected hybrids for each family within RCB1, RCB2, RCB3, and RCB4.

 
Second Clonal Generation (C2)
Family Heritabilities
As for the RCBs, the trait number of stolons was not included in the variance analysis and a root-square transformation was applied to the trait number of suckers. Family heritability values were significantly lower at the second clonal generation, especially for the trait corm weight (Table 5). A strong correlation was observed between the size of the elementary plot used for each type of trial (16 plants for the RCBs, 12 for {alpha}1, and 4 for {alpha}2) and the heritability value for the corm weight. Between {alpha}1 and {alpha}2, this relationship was also verified for the corm yield components, corm length and corm width, with heritability values on {alpha}2 less than half of those observed on {alpha}1 (Table 5). This difference between {alpha}1 and {alpha}2 was not observed for traits number of suckers and dry matter content, with very similar values of heritability (Table 5).


View this table:
[in this window]
[in a new window]
  		Table 5. Family heritabilities (H = h2F; Eq. [4]) and significance of the F-test for the family effect (P) calculated at the second clonal generation.{dagger}

 
Narrow-Sense Heritabilities (Parents-Offspring Regression)
When considering altogether data from {alpha}1 and {alpha}2, by applying the model described in Eq. [8], the treatment effect was highly significant for all studied traits (results not shown). This high significance had already been observed when studying both alpha designs separately, for the calculation of family heritabilities. Therefore, combining data from {alpha}1 and {alpha}2, allowed us to use adjusted means for the parents–offspring regression. Overall, complete data for 25 progenies and their parental mean were used for regression calculations. For the traits corm length, corm width, and corm weight, several parents presented adjusted means not significantly different from 0, due to high mortality and very poor agronomic performance, and progenies involving these parents were eliminated from the analysis to avoid a strong bias in the regression calculation. This poor agronomic performance was mainly due to a low weight and quality of the initial planting materials for these parents. If included in the analysis, these progenies would artificially reduce the narrow-sense heritability since these parents performances did not correspond, due to experimental mishaps, to their true agronomic potential.

Midparent values and offspring means for all studied traits are summarized on Table 6. Mean hybrid superiority was observed for all traits excepted for dry matter content. As for the narrow-sense heritabilities, traits number of suckers and dry matter content presented the highest heritability values. Yield components, corm length and corm width, also showed higher heritability values than corm weight, although only for corm length was the regression statistically significant. Contrary to family heritabilities calculated on trials {alpha}1 and {alpha}2, narrow-sense heritability was higher for corm length than for corm width. For this last trait, progeny no. 21 (VU054 x VU379) presented a much higher difference between the midparent and the mean offspring value than the other progenies. This progeny contained a high number of branched corms, which are characterized by a large corm width (Ivancic et al., 2004). If progeny no. 21 was not considered, the narrow-sense heritability for corm width was equal to 0.46 (P = 0.04).


View this table:
[in this window]
[in a new window]
  		Table 6. Midparent values, mean offspring values, and narrow-sense heritabilities (H = B; Eq. [7]) calculated at the second clonal generation.{dagger}

 
Correlation Between Traits
Correlation coefficients calculated over two clonal generations were rather similar, with the exception of the correlation between number of suckers and dry matter content, which was negative and significant (P < 0.01) in C2 and the correlation between number of suckers and corm weight, which was significantly higher in C2 (Table 7). The negative correlation between number of stolons and number of suckers confirmed taro breeders' observations and proved that taro plants will tend to form either suckers or stolons. The undesired nature of a high stolon formation was also illustrated by the negative relationship between number of stolons and dry matter content. Sucker production was more strongly correlated with yield and its components than stolon production. Interestingly, the correlation between number of suckers and corm length was relatively high as compared with the correlations between number of suckers and corm width or corm weight. As expected, the highest correlations were observed between corm dimensions and corm yield. In comparison, the correlation between corm length and corm width was significantly lower. Again, the high number of branched corms harvested might explain this lack of correspondence between both corm dimensions. Finally, a negative relationship was observed between dry matter content and the traits corm length, corm width, and corm weight, with low values of correlation coefficients (Table 7).


View this table:
[in this window]
[in a new window]
  		Table 7. Pearson coefficients of phenotypic correlation between all studied traits at the first (C1) and second (C2) clonal generations (N = 1820 or 1817 for C1 and N comprised between 1955 and 2280 for C2).{dagger}

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
The results presented showed that significant differences for the heritabilities of different important agronomic characters for taro were observed, whether trials were conducted as RCBs or as alpha designs, and whether family or narrow-sense heritabilities were considered. Thus, the traits number of suckers and dry matter content presented significantly higher heritabilities than corm weight and its components, corm length and corm width. These differences between different types of agronomic traits have frequently been observed on other model plants, such as maize (Zea mays L.; Gallais, 1989).

Through the six trials conducted on two successive clonal generations, a high variation was observed on the family heritability values calculated. In C1, divergences were mainly due to the different genetic constitution of the families used. Such a result has often been reported on numerous crops, such as potato (Rousselle et al., 1992) or maize (Gallais, 1989). Compared with C1, family heritability values obtained in C2 were significantly lower and the same situation was verified between the two alpha (0,1) designs planted in C2. The number of plants per elementary plot proved to be an important factor for the control of the environmental variance, and squares of four by four plants represent a minimum size for this type of trials.

Narrow-sense heritabilities were slightly higher than family heritabilities estimated in C2, with the exception of corm yield, but were lower than family heritabilities estimated in C1. Several authors have proposed theoretical and practical models that account for differences between these two types of heritabilities (Falconer, 1991; Zobel and Talbert, 1991). On perennial crops such as forest trees, measurements are not conducted at the same age on parents and their progenies (Zobel and Talbert, 1991). Although taro is an annual crop, the physiological ages of parents and hybrids are different because of their clonal nature.

The family heritability values obtained, as compared to narrow-sense heritabilities, open the possibility of applying family selection in the first cycles of a taro breeding program, coupled with individual clonal selection within the best families. The importance of family selection has already been outlined on other clonal crops (Kimbeng and Cox, 2003; Simmonds, 1996; Rousselle et al., 1992). According to Kimbeng and Cox (2003), family selection is particularly useful for traits with moderate heritability, which was the case for most of the traits studied in our analysis, since, as opposed to clones, families can be more easily replicated at different places and during several years. This possibility allows a more accurate estimation of family means and indirectly, of the parental values, based on hybrids' performance.

Concerning the correlations between traits, other authors have shown significant correlations between yield and several vegetative traits, such as plant height or length of the main petiole (Simin et al., 1995; Thankamma Pillai et al., 1995; Dwivedi and Sen, 1999). Although weak and variable among progenies and years, significant negative correlations were observed between the number of stolons and two important traits, the number of suckers and the dry matter content. This result confirms that a profuse stolon production is an undesired trait since it can also provoke important competition effects between neighboring plants and a deformed corm shape (personal observations).

When looking at the quality of the harvested hybrids, only a few families presented a high percentage of good hybrids (Fig. 1) and for some of the largest families less than 10% of good hybrids were selected. This result, together with the moderately high heritability values obtained for important traits such as the number of suckers or the dry matter content, suggest that, when working with large collections, only the parents showing the most valuable agronomic characteristics should be recombined and a high frequency of good offspring genotypes should be expected.

Although the morphological variation of Vanuatu cultivars is high, the genetic base is narrow, as first demonstrated by isozymes (Lebot and Aradhya, 1991) and later confirmed by AFLP and microsatellite studies (Kreike et al., 2004; Quero-García et al., 2004; Noyer et al., 2006). When working with such a narrow genetic base, strong heterosis effects are not likely to be achieved if large numbers of different crosses are conducted, which supports the idea that only the best parents should be selected to produce large progenies. However, genetically close cultivars should not be used, to avoid inbreeding depression effects, such as a high frequency of deleterious characters.

Unexpectedly, and despite the narrow genetic base involved in the Vanuatu taro breeding program, hybrids showed a significant superiority compared to parents for all traits with the exception of the dry matter content (Table 6). However, for a vegetatively propagated crop like taro, hybrids might have an advantage only in their first generations because they result from true seeds but the rapid accumulation of viral particles and of negative mutations might induce a progressive loss of this superiority. To enlarge its genetic base, the Vanuatu breeding program has recently introduced "elite" cultivars assembled by TANSAO project that are currently being used. The second approach has therefore been favored, and a large number of crosses between Vanuatu hybrids and varieties, and Asian cultivars, have been conducted. The analysis of the resulting offspring populations should indicate which combinations are the best. In this way, breeders will be able to continue the program with the first approach and conduct large crosses only among highly valuable and genetically distinct genotypes.

On the other hand, if resources are available, more complex trials, such as half-diallels or factorial designs, could be considered to gain further insight into the genetic parameters (such as additive and dominance variance) of the main agronomic traits of taro. Moreover, although long and fastidious in the case of taro, multiplication and standardization of clonal materials should be pursued, to be able to include genotypic replicates in further trials. Outstanding hybrids could thus be selected more accurately based on their true genetic values. Further, an important aspect on clonal crops breeding, such as the comparison of genetic intrafamily and interfamily variation (Simmonds, 1996), might be studied.


   ACKNOWLEDGMENTS
 
This study was supported by the Taro Network for Southeast Asia and Oceania (TANSAO), a project funded by the INCO programme of the European Union (DG XII) (grant no. ERBIC18CT970205). The authors are thankful to the Vanuatu Agricultural Research and Training Center (VARTC).

Received for publication November 18, 2005.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 





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

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