Assessment of Morphological Variation in Irish Brassica oleracea Species

This study aimed at assessing the morphological variation in 25 accessions of an Irish collection of Brassica oleracea using 44 morphological traits. The morphological traits showed a pronounced variation among accessions. Vegetative preflowering and maturity traits revealed higher variation than seedling traits. Of the 44 morphological descriptors used, 28 were significantly different and proved useful in examining diversity and differentiating among accessions studied. Principal component analysis (PCA) showed variation among accessions, where 87.24% of the total variation in the morphological data was separated on the first five components, indicating a high degree of correlation among traits studied. Cluster analysis grouped the 25 accessions into 4 main groups, corresponding to kales, cauliflowers, cabbages and sprouts, and also showed a sort of correlation between clustering pattern and eco-geographical distribution of the accessions. Furthermore, this study identified 9 highly diverse accessions, providing opportunities for optimising parental sources in future breeding programs to develop new or more productive Brassica varieties. Morphological traits were deemed useful for assessing the diversity and relationships in Irish Brassica oleracea species.


Introduction
The genus Brassica L. is one of the most economically important genera in the tribe Brassiceae, which in turn belongs to the family Brassicaceae (Rakow, 2004).This genus comprises a diverse group of species including major vegetable and oilseed crops with a wide range of agronomic traits (Rich, 1991;Christopher et al., 2005).It contains six economically important species with great genetic and morphological diversity which are cultivated worldwide (Saha et al., 2008).Three of these species are diploid (Brassica oleracea,2n = 18;Brassica rapa,2n = 20;Brassica nigra,2n = 16), and three are amphidiploid (Brassica napus,2n = 38;Brassica juncea,2n = 36;Brassica carinata,2n = 34).Brassica species play an important role in agriculture and horticulture, as well as contributing both to the economy and health of populations around the world (King, 2005;Zhao, 2007).
Assessment of diversity is an essential component in germplasm characterisation and conservation, which in turn are essential to the continued maintenance and improvement of agricultural production and, thus, to sustainable development and poverty alleviation (Karp et al., 1997).Morphological variation in plant species has been described for traits controlled by a single or multiple gene systems.However, the variation patterns in the morphological traits are considered to be the result of both genetic and environmental attributes (Rohlf et al., 1990;Rohlf, 2000).Morphological traits have been used to assess the genetic variation and relationships among populations of different plant species, for example, mungbean (Arshad et al., 2009;Tantasawat et al., 2010), soybean (Iqbal et al., 2008), rice (Bibi et al., 2009), eggplant (Tümbilen et al., 2011) and Brassica (Rabbani et al., 1999;Kop et al., 2003;Balkaya et al., 2005).Rabbani et al. (1999) used the morphological traits to assess the variation in oilseed mustard.35 phenotypic traits were recorded for all the 52 accessions studied from seedling emergence up to crop harvest.The results revealed a considerable level of diversity among all tested accessions for various traits.Seedling traits showed less variation, while the largest variation was found for flowering and maturity stage traits.Generally, a low correlation was observed among different traits, however, some of the related characters were significantly correlated with each other.The results also indicated that oilseed mustard in Pakistan has narrow genetic base and experiences a high level of genetic erosion.Balkaya et al. (2005) determined the differences regarding morphological variation of white head cabbage (Brassica oleracea var.capitata subvar.alba) genetic resources collected from different regions of Turkey.Their cluster analysis based on 12 quantitative and 10 qualitative variables identified 10 groups.Morphological variability was high among the white head cabbage genotypes of Turkey.The genotypes also displayed a wide diversity of characters for selection and combination of interesting genotypes for important characters to obtain improved varieties.
Although significant progress has been made in the past decade, Brassica biodiversity loss has not been halted in the island of Ireland.The status of many of Irish habitats and species is judged to be poor (NBP, 2011).Moreover, information on the genetic diversity and relationships in Irish Brassica oleracea species is currently very limited.Therefore, the main aim of this study was to assess the morphological variation and character association in Irish Brassica oleracea accessions based on morphological traits, thus providing a correct taxonomic ranging and a precise characterisation of these accessions.

Plant Material
Twenty-Five accessions of Irish Brasssica oleracea were obtained from the germplasm collection maintained at the Horticultural Research Institute (HRI), Wellsbourne, United Kingdom (Table 1).These accessions were chosen based on their sampling site covering a broad and diverse geographic range of the island of Ireland.However, the selected accessions represented 4 subspecies within Brassica oleracea species (Brassica oleracea capitata, Brassica oleracea acephala, Brassica oleracea botrytis and Brassica oleracea gemmifera).

Experimental Design and Plant Growth
Seeds were sown in PPE modules of 12 x 12 cells filled with seedling compost (Shamrock) in the greenhouse facilities at the Teagasc Horticultural Research centre in Kinsealy, Dublin, Ireland.Four weeks after germination, the seedlings were transplanted to larger PPE pots containing compost in the greenhouse.After a further 5 weeks, the modules were removed and placed on a bench outside the greenhouse in order to harden the seedlings off for 2 weeks.Land plots were prepared within the field trial site at Teagasc Kinsealy.Prior to transplanting, the seedlings were treated with Dursan (an insecticide for the control of caterpillars, cabbage root fly and cutworms).The seedlings were then transferred into 5 blocks in the field, with 5 accessions per block in a randomized scheme (Figure 1).Sowing and growing the accessions at different time points are shown in Figure 2.

Evaluation of Morphological Traits
Forty-four discriminating descriptors were selected from the list of morphological descriptors for Brassica and Raphanus published by the International Board for Plant Genetic Resources (IBPGR, 1990).These descriptors were considered to be suitable for Brassica oleracea, and were used in this study to characterize 10 plants of each of the 25 accessions studied.The descriptors selected included 5 descriptors at the seedling stage, 32 descriptors at the mature pre-flowering stage and 7 descriptors at the flowering stage (Table 2).

Data Analysis
The morphological data were standardised and converted to z-scores to avoid bias in the analysis due to scale differences in the variables (Legendre and Legendre, 1998).The standardised dataset was then used to estimate the genetic diversity and relationships among the accessions which were evaluated by Jaccard similarity index, cluster analysis, factor analysis, analysis of variance and correlation coefficients.These multivariate analyses were made using the software package SYSTAT for Windows, version 7.0 copyright (C) 1997, SPSS INC.A dendrogram was constructed through the average linkage-joining rule.

Cluster Analysis
Morphological data differentiated all the accessions of Brassica oleracea studied.The phenogram constructed using each accession as an OTU (operational taxonomic unit) was performed on the basis of the collected morphological data of the 25 accessions of Brassica oleracea (Figure 3).This phenogram exhibited that all the genetic distance among the examined accessions was around 65%, and showed 4 major groups.The first group contained the 2 accessions of brussels sprout.The second group contained all accessions of kale and fodder kale.
The third group was homogenous and contained all accessions of winter cauliflower.The fourth group split into 4 subgroups.Furthermore, the accessions of most cabbage types were distributed among different clusters within these subgroups.None of the studied cabbage types formed a distinct cluster except the 2 accessions of common cabbage which were separated in one cluster within the third subgroup.The accessions of cabbages HRIGRU5915 and HRIGRU4588 were very similar to the accessions of cabbages HRIGRU4497 and HRIGRU4498.

Jaccard Binary Similarity Coefficients (Sj)
The overall mean similarity index for Brassica oleracea accessions calculated by the Jaccard similarity index (Sj) ranged from 0.051 to 0.976 with an average of 0.503 (Table 4).The highest similarity indices (0.976) were between the accessions of the same crop variety and geographical region, e.g.cabbage HRIGRU4497 and HRIGRU4498 from Roscommon.The second indices (0.952) were also between the accessions of the same group and geographical region, e.g.spring cabbages HRIGRU4566 and HRIGRU4564 from Cork, and spring cabbages HRIGRU4564 and HRIGRU4571 from Cork.The Sj similarity between the accessions of fodder kales HRIGRU4502 and HRIGRU4503 from Kildare was about 0.93.The accession of common cabbage HRIGRU4585 from Donegal had a high similarity index (0.907) with the accession of common cabbage HRIGRU4586 from Mayo.The Sj similarity between the accessions of common cabbage HRIGRU4586 and cabbage HRIGRU5915 was about 0.64.The lowest similarity indices and the highest variation were found between the accessions of kale HRIGRU7227 from Sligo and brussels sprout HRIGRU4494 from Dublin.The highest genetic distance in the cabbage types was found between the accessions of cabbage HRIGRU12532 and cattle cabbages HRIGRU4579, HRIGRU4561 and HRIGRU4508.

Principal Component Analysis (PCA)
Principal component analysis (PCA) showed variations among the accessions studied (Table 5).The first 5 varieties of the principal component analysis accounted for 87.24% of the total variation among the 25 accessions.
The first component (PC1) explained 37.88% of the total variation, while PC2, PC3, PC4 and PC5 contributed 27.97%, 11.89%, 5.13% and 4.38% of the total variation, respectively.The values of the principal component analysis for all morphological traits were shown in Table 5.The traits, which showed the greatest variability on the first component, were leaf apex shape (0.708), heading habit (0.954), head shape in longitudinal section (0.954), head-forming leaf overlap at terminal region (0.972), head cover from subtending leaves (0.996), primary colour of outer head leaves (0.866), head size in relation to plant size (0.976), head solidity (0.954), head length (0.945), head diameter (0.938), stem length in head estimated (0.954), tendency of head to split (0.796) and head holding ability (0.954).
However, these traits exhibited some degree of variation among the accessions of Brassica oleracea.

Analysis of Variance (ANOVA)
Of the 44 morphological traits examined in this study, 28 were significantly different among the accessions of Brassica oleracea (Table 6).These 28 significant descriptors were deemed useful and can be used for further characterisation of the Brassica accessions.The regression coefficients (r 2 ) values ranged from 0.041 to 0.55 with an average of 0.31.

Correlation Coefficients among the Morphological Traits
Pearson correlation coefficients were computed for all pairs of traits (data not shown).The highest correlation was between leaf margin and leaf bloom (R= 1), and flower stalk length estimated and flower stalk internode length (R=1).There was a high correlation between flowering plant hairiness trait and some of enlarged buds traits, e.g.firmness, size and number of enlarged buds (R=1).There was also a high correlation among head traits (heading habit, head shape in longitudinal section, head solidity, stem length in head and head holding ability) (R=1).The head holding ability (bolting resistance) was high in most of the accessions studied.On the contrary, leaf margin and leaf bloom were negatively correlated with flowering plant hairiness and most enlarged buds traits, e.g.number, size and firmness (R= -1).Days to flowering trait was also negatively correlated to most traits, e.g.number of leaves estimated, leaf apex shape, leaf retention on stem, stem length in head, head length, size and number of enlarged buds, and all head and enlarged buds traits.

Discussion
To conserve and utilise germplasm efficiently, it is essential to investigate the extent of diversity available.Morphological characterisation is an essential step in the characterisation and classification of crop germplasm because a breeding program mainly relies on the magnitude of morpho-phenological variability (Koffi et al., 2008).The morphological traits used in this study showed a pronounced variation among accessions.Seedling traits such as juvenile development and seedling leaf colour showed the lowest variation.The hypocotyl colour and seedling leaf (marginal incisions) traits had no taxonomic value among accessions of Brassica oleracea because these traits showed the same patterns within all of the accessions studied.However, they could be used to assess the variation among wild and cultivated species of genus Brassica.Vegetative preflowering and maturity traits such as petiole section, petiole and midvein colour, leaf colour, primary colour of outer head leaves, leaf division (incision), etc. exhibited a high level of variation among the accessions studied.These results were consistent with that reported by Rabbani et al. (1999).
The cluster analysis showed that the cauliflowers were more closely related to the cabbages than the kales were.However, the kales were more closely related to cauliflowers and cabbages than the brussels sprouts were.The accessions of most cabbage types were distributed among different clusters within cabbage subgroups, and this agreed with Balkaya et al. (2005), who reported that there was some degree of variation among Turkish white head cabbages.However, the cluster analysis from this current study is more reliable, as most of the accessions of each group and subgroup were collected from closely related geographical regions.Only a few accessions were scattered among the different subgroups.This showed a sort of correlation between clustering pattern and eco-geographical distribution of the accessions.This result agreed with that reported by Faltusová et al. ( 2011), but was not consistent with that reported by Balkaya et al. (2005).The difference in these data could be attributed to the differences in accessions or the traits used which could in turn be influenced by environmental factors such as geographical area, elevation of temperature, and soil fertility.
The output of the Jaccard binary similarity coefficient from our morphological data revealed that the highest similarity indices were between the accessions of the same crop variety and geographical region.This confirmed the correlation between grouping on morphological data and eco-geographical distribution of the accessions.However, this high similarity could be attributed to the low gene flow within the same crop variety.This study also suggested that the accessions of kale HRIGRU7227, common cabbage HRIGRU4586, spring cabbage HRIGRU5914, cattle cabbage HRIGRU4561, HRIGRU4508 and HRIGRU4579 and cabbage HRIGRU12532, HRIGRU5915 and HRIGRU4588, due to the highest variation found among them, could provide the opportunities for selecting as parental source in future breeding programs to develop new or more productive Brassica varieties.
In the analysis made to estimate the relative contributions of the different morphological traits studied towards the overall phenotypic variation among the 25 accessions, a total of 5 principal components (PCs) explained 87.24% of the total variation in all the 44 traits studied.This data indicated a high degree of correlation among traits of the accessions studied.About 38% of the total variability explained by the first PC alone was due chiefly to variations in enlarged apical buds (head) traits, indicating a very high degree of correlation among them.Furthermore, all enlarged vegetative buds and most of enlarged inflorescences (floral branching pattern) traits were separated on the second component (explained about 28% of the total variability), indicating a high degree of correlation among them.We could thus conclude that the traits of enlarged apical buds, enlarged vegetative buds and enlarged inflorescences proved to be useful markers for separating Brassica oleracea varieties.These observations agreed with the classification criteria of Brassica oleracea varieties reported by Paterson et al. (2001) and Hong et al. (2008), that are mainly based on extreme morphological characteristics such as the enlarged single apical bud of cabbage, the several enlarged vegetative buds of brussels sprout and the enlarged inflorescences (floral branching pattern) of cauliflower.
One-way analysis of variance (ANOVA) of the morphological traits used showed that 28 traits were significantly different among the accessions studied, and can be used for further characterisation of Brassica accessions.However, the low percent of contribution of the morphological traits (31%) could be attributed to the low number of flowering traits used.Furthermore, most of accessions were of orbicular leaf blade shape, dentate leaf margin, rounded leaf apex, green leaf colour, intermediate leaf bloom, rounded petiole section, spheric head shape, strong head forming leaf overlap, covered head, green primary colour of outer head leaves, yellow green flower stalk, short flower stalk length and pale yellow petals.These results were relatively consistent with that reported by Katz (2003).
The correlation is a measure of the degree to which variables differ or a measure of the intensity of association (Rabbani et al., 1999).The correlations revealed by Pearson correlation matrix in this study suggest that when the flower stalk length estimated is used as a distinguishing character, the flower stalk internode length could be redundant for characterizing the accessions.Most of the highly correlated traits (e.g.heading habit, head solidity, stem length in head firmness, size and number of enlarged buds and flower stalk length estimated) are of economic importance and are identified as useful for Brassica breeding programs and taxonomic purposes.

Conclusion
This study differentiated all the accessions of Brassica oleracea studied, and identified 28 morphological traits which were significantly different among the accessions and can be used for further characterisation of Brassica accessions.It also identified 9 highly diverse accessions, providing opportunities for optimising parental sources in future breeding programmes to develop new or more productive Brassica varieties.Morphological traits proved useful in assessing the diversity and relationships of Irish Brassica oleracea germplasm.

Figure 1 .Figure 2 .
Figure 1.Randomised planting scheme used for transplanting Brassica oleracea accessions (the numbers in the blocks refers to the accessions numbers)

Figure 3 .
Figure 3. Dendrogram showing the relationship among 25 accessions of Brassica oleracea based on morphological data

Table 1 .
Accession numbers, crop names, and collection sites of the accessions of Brassica oleracea studied

Table 3 .
Statistical parameters for quantitative and qualitative traits of Brassica oleracea accessions (see Table1for identification of accessions numbers)

Table 4 .
Sj Jaccard binary similarity coefficients calculated from the morphological data for 25 accessions of Brassica oleracea (see Table1for identification of accessions numbers)

Table 5 .
Matrix of eigenvectors and values of the principal components for the morphological traits of 25 accessions of Brassica oleracea

Table 6 .
One-way ANOVA for the morphological traits among 25 accessions of Brassica oleracea