Euphytica 79: 1 15-126,1994. @ 1994 Kluwer Academic Publishers. Printed in the Netherlands. Taxonomy of Portuguese Tronchuda cabbage and Galega kale landraces using morphological characters, nuclear RFLPs, and isozyme analysis: A review JoTio S. Diasl & Ant6nio A. Monteirol Instituto Superior de Agronomia, Tapada da Ajuda, 1300 Lisboa, Portugal Received 12 November 1993; accepted 15 August 1994 Key words: Brassica oleracea, evolution, landrace group, numerical taxonomy, Portuguese coles Summary Morphological characters, nuclear RFLPs, and isozyme analysis were used to study the similarity between 32 Portuguese Tronchuda cabbage and Galega kale landraces, and some cabbage cultivars traditionally grown in Portugal. Forty-six morphological characters observed in two consecutive years, RFLP data from 55 nuclear probes, detecting 291 polymorphic nuclear DNA restriction fragments, and allelic frequencies in 21 putative loci, generated by nine isozymes, were analyzed by the unweighted pair group method, using arithmetic averages (UPGMA), in order to present the results in the form of a phenogram. The three methods resulted in different clustering patterns of the 32 cole accessions. Morphological characters gave consistent clustering according to the traditional landrace definition and denomination, producing clear separation between Tronchuda cabbages and Galega kales. RFLPs were unable to separate Tronchuda cabbages from Galega kales and defined five landrace groups corresponding to their geographic origins rather than to their morphological similarities. Isozymes showed poor accession discrimination and an intermediate clustering pattern with some accessions being clustered according to their geographic origins and others according to their morphological similarities. Portuguese Tronchuda cabbages and Galega kales constitute a distinct and relatively homogenous group within Bmssica olemcea, sharing the same genetic background. It is concluded that Portuguese coles have evolved independently from a common ancestor to the present cultivated forms. Portugal should be considered as an important region of domestication of specialized leafy coles. Introduction Portuguese coles represent a unique group of vegetable crops within Brassica oleracea L. with striking morphological diversity, and they are widely grown throughout Portugal. Portuguese coles include: (i) Tronchudacabbages (B. oleracea var. costata DC. syn. B. oleracea var. tronchuda Bailey, 'Couve Tronchuda'); (ii) Galega kales (B. olemcea var. acephala DC., 'Couve Galega'); (iii) other economically less important coles such as Algarve cabbages (B. olemcea var. capitata L., 'Couve do Algarve'), and (iv) Algarve Savoy cabbage (B. oleracea var. sabauda L., 'Couve Repolho-lombarda do Algarve') (Dias, 1992). The traditional grouping of Portuguese coles considers different landraces based on their geographic origin and morphological resemblance. The landraces are sometimes distinguished by only a few morphological characters and it is not easy to classify them because of their phenotypic diversity (Dias et al., 1992). Nuclear RFLPs (Dias et al., 1992), morphological characters (Dias et al., 1993) and isozymes (Dias et al., 1994) were recently used for taxonomic studies of Portuguese coles, allowing a better understanding of variations between and within landraces. Numerical taxonomy, using 46 morphological characters, defined eight major groups among 58 accessions which clustered Tronchuda cabbage and Galega kale landraces as well as some traditional cabbage cultivars. The groupings were primarily associated with morphological differences among accessions
1 1 6 and secondarily with their geographical origin (Dias et al., 1993). Molecular taxonomy of 48 cole accessions, using nuclear RFLPs, clearly separated Portuguese Tronchuda cabbage and Galega kale landrace from other cultivated brassicas. Portuguese accessions formed a unique and closely related group which could be further divided into five subgroups corresponding to the major areas of cultivation (Dias et al., 1992). The phenetic grouping of the accessions based on RFLP data coincided more with their geographic origin than with their morphological similarities. An isozyme analysis, using two gel systems and nine enzymes revealing 21 putative loci, was carried out on a collection of 48 cole accessions including 27 Tronchuda cabbage and Galega kale landraces. Isozymes were unable to separate Portuguese cabbages and kales from the other cultivated brassicas (Dias et al., 1994). The isozyme analysis clustered the accessions into six groups but the phenogram was difficult to interpret considering the origin of the material and the morphological resemblance among the accessions. The morphological, RFLPs, and isozyme studies referred to above were not able to provide a direct comparison between the three methods since the accessions used for RFLPs and isozyme analysis were different from those used for morphological characters. Those previous taxonomical studies had resulted in inconsistent clustering patterns of the Portuguese cole landraces and raised pertinent questions about the choice and the utility of the different methods for studying allogamic varieties and for determining the genetic relationships between Portuguese cole landraces. The objective of the present study is to answer some of those questions by making direct comparison between the clustering patterns obtained by morphological characters, nuclear RFLPs, and isozymes, using the same collection of Portuguese cole landraces as far as possible. This comparison is to give further insight into the origin of Tronchuda cabbage and Galega kale landraces, as well as into their groupings and similarities. Material and methods Plant material Thirty-two accessions, including Portuguese coles representative of the main landraces and growing regions of Portugal plus some imported cabbage cultivars traditionally grown in Portugal, were used in the present study. Accession designation and their sources are listed in Table 1, and their geographical origin is shown in Fig. 1. The 32 accessions were selected by choosing the accessions common among those used in the three previously reported taxonomical studies (Dias et al., 1992 ; Dias et al., 1993 ; Dias et al., 1994). Morphological character analysis The methodology for character observation, data handling and taxonomical analysis was fully described by Dias et al. (1993). Forty-six morphological characters were observed in the 32 accessions. The selection of the characters was made by adapting the IBPGR descriptors (IBP- GR, 1989) to the method of numerical taxonomy and to the characteristics of the material, taking the work of Cerca (1946) and previous field observation of the material into account. Four characters were studied at seedling stage before transplanting 40 plants per accession which had been collected at random from multipot-trays. Fourty-two characters were recorded at maturity and at blossom stage in the field on 20 plants per accession which had been selected at random in each plot. Sub-samples of the 32 accessions under investigation were grown in the same experimental field during two consecutive years (1989/90 and 1990/91) to take the influence of climatic variation between years into consideration. The 46 characters scored for each year were analyzed, using standard numerical taxonomic techniques (cf. Sneath & Sokal, 1973). The elementary unit to be classified in each analysis was named OTU (= Operational Taxonomic Unit). The data were analyzed based on a matrix of (46 characters x 32 OTUs) on which the score for each accession (= OTU) was the mean character value for the two years. The OTUs were clustered by the unweighted pair group method using arithmetic averages (UPGMA), and the results were presented in the form of a phenogram. RFLP analysis The procedure for RFLP detection, data handling, and numerical taxonomy analysis was fully described by Dias et al. (1992). DNA of a bulk leaf sample taken from 30 plants of each accession was isolated from freeze-dried leaf tissue, and restriction endonuclease EcoRI (BRL) was used to digest the crude DNA samples. Methods for DNA isolation, restriction endonuclease digestion, electrophoresis, Southern blotting, hybridization and autoradiography were adapted from Osborn et al. (1987). Fifty-three nuclear DNA clones, from genomic DNA libraries of B. oleracea cv. Wisconsin Golden Acre and B. raga
1 1 7 Table 1. Accessions and their sources used in this study' Accession number Identification Originb Code' B. oleracea var. costata (syn. tronchuda) : 1 Penca de Chaves Loivos, Chaves, Pt ISA 10 2 Penca de Mirandela Mirandela, Mirandela, Pt ISA 2 3 Penca de Safes Safres, Alij6, Pt ISA 120 4 Penca de Mirandela da Veia Branca (Pao de Acdcar) Mirandela, Mirandela, Pt ISA 444 5 Penca da P6voa Aptilia, Esposende, Pt ISA 454 6 Penca da P6voa Agucadoura, P6voa do Varzim, Pt ISA 333 7 Coivao Belinho, Esposende, Pt ISA 129 8 CoivAo Mindelo, Vila do Conde, Pt ISA 178 9 Couve Gl6ria de Portugal Arcozelo, Gouveia, Pt ISA 84 10 Couve Gl6ria de Portugal Contengas de Baixo, Mangualde, Pt ISA 91 11 Couve Calpuda Ribeira, Condeixa, Pt ISA 284 12 Couve da Arrocha Arrocha, Condeixa, Pt ISA 265 13 Coue da Cordinha CordinhA, Mealhada, Pt ISA 318 14 Couve de Bolho Bolho, Cantanlhede, Pt ISA 330 15 Couve Tronchuda Branca de Cantanhede Cantanhede, Cantanhede, Pt ISA 494 16 Silveirinha Cabeco da Azoia, Leiria, Pt ISA 143 17 Couve Portuguesa Salemas, Loures, Pt ISA 134 18 Couve de Grelo A-dos-Caos, Loures, Pt ISA 61 19 Couve Branca Nordeste, Azores, Pt ISA 198 20 Couve de Valhascos Valhascos, Sardoal, Pt ISA 35 21 Couve de Valhascos Valhascos, Sardoal, Pt ISA 20 22 Couve Murciana Viana do Alentejo, Viana do Alentejo, Pt ISA 62 23 Couve Algarvia Verdemilho, Aveiro, Pt ISA 207 B. oleracea var. acephala: 24 Couve Galega Contengas de Baixo, Mangualde, Pt ISA 92 25 Couve Galega Sete Fontes, Cantanhede, Pt ISA 314 26 Couve Galega Tojalinho, Loures, Pt ISA 57 27 Couve Galega Sardoal, Sardoal, Pt ISA 28 B. oleracea var. capitata : 28 Couve Repolho do Algarve Alfambras, Aljezur, Pt ISA 225 29 Couve Repolho do Algarve Carrapateira, Vila do Bispo, Pt ISA 232 30 Cabbage `Coracao de Boi'?, France ISA 484 31 Cabbage 'BacalA Frisada'?, France ISA 471 B. oleracea var. sabauda: 32 Couve Repolho-lombarda do Algarve Femeiras, Albufeira, Pt ISA 67 a Botanical classification according to Williams & Hill (1986). The Portuguese accessions were tentatively ascribed to the different taxa based on their morphological characteristics. b Locality, County, Country ; Pt = Portugal ;? = unknown locality or country. I ISA = Instituto Superior de Agronomia.
1 1 8 z Q W tj 0 M INHO 7 1 5 8 POR 0 0 23 24 3 =i 13 1 0 9 15 25 12 1 1 TR A S-OS 4 RO R IVER -,M ONTES 2 &N 27 16 21 i S SPAIN 14, 180 17 26 LISBON 127 W I 22 38 N AZORES Co 29 28 ALGARVE 3 2 Fig. 1. Map of Portugal with the geographical distribution of the accessions used in the present study. The black dots refer to the collection sites, and the numbers refer to the accessions listed in Table 1.
1 1 9 cv. Michihili, were used as probes on Southern blots. Restriction fragments in all accessions hybridizing to each of the 55 probes were assigned numbers (1, 2, 3,..., n) according to decreasing molecular weights, and then each fragment was treated as a unit character and scored as present/absent in each accession. Similarities among accessions, based on a data matrix of 32 accessions x 291 characters, were computed using the simple matching coefficient. The accessions were clustered by the UPGMA method in order to present the results in the form of a phenogram. Isozyme analysis The method for isozyme staining, data handling, and taxonomy analysis was fully described by Dias et al. (1994). Forty plants per accession at cotyledonary stage were used to produce the extract for starch gel electrophoresis. Two gel systems and nine enzymes were used : 1) Tris citrate lithium hydroxide and boric acid ph 8.5 gel (May et al., 1988) with aspartate aminotransferase (AAT), phosphoglucoisomerase (PGI), gluthatione reductase (GR) and fructose bisphosphatase (FBP) ; 2) Histidine citrate ph 6.5 gel (Thorpe et al., 1987 ; May et al., 1988) with triosephosphate isomerase (TPI), leucineaminopeptidase (LAP), phosphoglucomutase (PGM) and isocitrate dehydrogenase (IDH). The superoxide dismutase enzyme (SOD) was revealed simultaneously with the FBP From the isoenzymatic patterns observed in 40 individuals of each accession, their allelic frequencies were established. The genetic similarity among accessions was estimated from the allelic frequencies based on the genetic distances of Nei (Nei, 1978) between each pair of accessions. The matrix of genetic distances obtained was then hierarchically clustered by the UPGMA method in order to obtain the respective phenogram. All computations for the analysis with morphological characters, RFLPs, and isozymes were carried out, using the NTSYS-pc (Numerical Taxonomy and Multivariate Analysis Systems) package, version 1.5 of computer programs (Rohlf, 1989). Results and discussion Similarities among cole accessions Morphological character analysis The phenogram based on the correlation coefficient among OTUs of the average results of the two observation years (Fig. 2) recovers basically the three major groups in which Por- tuguese coles are traditionally classified : (A) Tronchuda cabbages (1-13), (B) Galega kales and similar coles (16-26), and (C) common and savoy cabbages (32-30). Group A includes all the landraces usually designated as Tronchuda cabbages and suggests the existence of 4 subgroups. Subgroup Al contains eight accessions (1-8) from northern Portugal above the Douro River (cf. Fig. 1) ; Subgroup A2 includes only the Tronchuda cabbage landrace 'Couve Portuguesa' (17) from the Loures region ; Subgroup A3 includes six Tronchuda cabbage accessions (15, 18-22) from different regions of Portugal ; Subgroup A4, relatively heterogenous, contains six accessions (9-13) from Central Portugal. Group B includes Galega kales and other kale-like accessions from different geographic origins (16, 23-26). Group C constituted only by common and savoy cabbages (28-32) is well separated from the previous groups. The correlation phenogram based on the reduced number of accessions (Fig. 2) is identical to the one obtained using 58 accessions (Dias et al., 1993). Identical accession clustering is suggested in both phenograms supporting landrace grouping primarily according to phenotypic similarities and secondarily in relation to the geographic origin of an accession. RFLP analysis The phenogram based on the simple matching similarity matrix (Fig. 3) suggests the existence of two large accession groups : Group A, including 4 subgroups (Al, A2, A3, and A4), contains all Tronchuda cabbages and Galega kales ; Group B contains all common and savoy cabbages. Subgroup A l (1-4) includes accessions from northern Portugal above the Douro River (cf. Fig. 1) ; Subgroup A2 includes 13 accessions (9-18) from Central Portugal ; Subgroup A3 contains five accessions : `Couve Branca' (19), collected in the Azores islands, three Tronchuda cabbages (20, 21, 22) and one Galega kale (27) from southern inland Portugal. 'Couve Algarvia' (23) constitute the fourth subgroup (A4). Group B contains all the common and savoy cabbage accessions clustered in two subgroups : B 1 with Algarve cabbages (29-32) and B2 with cabbages `Couve Corarao de Boi' (30) and 'Couve Bacala Frisada' (31), both originally from France. The clustering pattern suggested by the phenogram produced by the reduced number of accessions (Fig. 3) is identical to the one obtained when 55 accessions
120 CORRELATION -0.30 0.00 0.30 0.60 0.90 1 t 1 t 1 3 4 2 Al 6 7 5 8 17 A2 15 21 A 20 22 18 19 9 10 11 12 14 13 16 ~ B 23 27 B 24 25 B2 26 32 C 28 Cl 29 C 31 30 I C2 Fig. 2. Phenogram of the 32 accessions based on the UPGMA method, using the matrix of morphological characters correlation coefficient. The numbers refer to the accessions listed in Table 1(cophenetic correlation r = 0.768). were used (Dias et al., 1992). The nuclear RFLPs data confirm the remarkable association between genetic affinity and geographic origin of Tronchuda cabbages and Galega kale landraces. Isozyme analysis Nei's phenogram of genetic distances (Fig. 4), despite the difficult interpretation, suggest two accession groups (Groups A and B) plus 3 isolated accessions (29, 30, 31). Group A includes Tronchuda cabbage accessions (1-8) from the North of Portugal, above the river Douro, plus 'Couve Calcuda' (11) from coastal central Portugal (cf. Fig. 1). Group B, very disperse, includes 18 Tronchuda cabbages and Galega kales collected south of the Douro river plus the common cabbage 'Couve Repolho do Algarve' (28) and the savoy cabbage 'Couve Repolholombarda do Algarve' (32). The three common cabbages 'Couve Coracao de Boi' (30), 'Couve Repolho do Algarve' (29) and 'Couve BacalA Frisada' (31) are disperse and isolated from the other common and savoy cabbages. The clustering pattern of Nei genetic distance phenogram originated by the reduced number of accessions (Fig. 4) is very similar to the clustering pattern obtained with 55 accessions (Dias et al., 1994). Isozyme data allowed a rather poor discrimination of Portuguese cole landraces. Cole accessions from northern Douro River were the only ones consistently individualized. The other accession groups were difficult
SIMILARITY 0.75 0.88 1.00 I l 1 2 1 7 6 2 3 4 9 10 24 25 11 12 13 14 15-1 16 26 17 18 19 20 21 22 27 23 29 28 32 30 31 Al A2 A3 - A4 - A B Fig. 3. Phenogram of the 32 accessions based on the UPGMA method using the matrix of nuclear RFLPs simple matching similarity. The numbers refer to the accessions listed in Table I (cophenetic correlation r = 0.964). to distinguish according to phenotypic differences or the geographic origin of the material. Taxonomy of Portuguese cole landraces The different clustering patterns of Portuguese cole accessions originated by the three taxonomic methods used in the present paper lead to an interesting reflection on the genetic affinity among Portuguese cabbage and kale landraces. The single accession group consistently revealed by the three phenograms are the five landraces 'Penca de Chaves' (1), 'Penca de Safres' (3), 'Penca de Mirandela' (2, 4), `Penca da Povoa' (5, 6), and `Coivao' (7, 8) from "Minho" and "Tras-os-Montes" provinces (northern Douro river). The congruence of the discrimination of this accession group by morphological, biochemical, and molecular markers suggests that geographic isolation strongly influenced the evolution of these landraces, leading to their stability as taxonomic entities in spite of fact that they share the same genepool. The stability of northern Douro landraces supports their uniqueness as well as the conclusion that little introgression from alien coles has taken place. Algarve cabbages (28, 29, 32) were consistently individualized from Tronchuda cabbages and Galega kales by morphological characters and by nuclear RFLPs. This individualization and the frequent clustering with other common cabbages confirms the higher genetic affinity of Algarve cabbages with common cab-
122 GENETIC DISTANCE 0.16 0.12 0.08 0.04 0.00 I I I I I 1 4 2 3 5 C7 8 11 9 24 10 23 12 19 26 27 20 21 22 28 25 13 16 14 15 32 1 7 18 30 29 31 I Al Fig. 4. Phenogram of the 32 accessions based on the UPGMA method using the matrix of the Nei's genetic distances. The numbers refer to the accessions listed in Table 1 (cophenetic correlation r = 0.885). A2 A B bages than with Tronchuda cabbages, and supports a different origin for Algarve and Tronchuda cabbages. Algarve cabbages may have resulted from foreign cabbages introduced in the region. This agrees with the assumption that Algarve cabbage might be regarded as a variety of foreign origin much cultivated in the Algarve province (Anonymous, 1896). It is, therefore, correct to include Algarve cabbages (28, 29, 32) in the capitata and sabauda varieties. Isozymes could not completely individualize Algarve cabbages from Tronchuda cabbages and Galega kales. The phenogram (Fig. 4) shows that 'Couve Repolho do Algarve' (28) clusters with 'Couve Murciana' (22) and 'Couve de Valhascos' (20, 21), and that 'Couve Repolho-lombarda do Algarve' (32) clus- ters with a heterogeneous group of Tronchuda cabbages and Galega kales (cf. Group B, Fig. 4). Contrary to morphological characters and nuclear RFLPs, isozyme data support the existence of some genetic affinity between Tronchuda cabbages and Algarve cabbages which suggests that during the local selection of Algarve cabbages some introgression of Tronchuda cabbages into imported common cabbages may have occurred. Potter (1878) confirms that Algarve cabbage was already cultivated in Portugal during the nineteenth century. A clear distinction between Tronchuda cabbage and Galega kales, and between the various Tronchuda cabbage landraces was only possible by using morphological analysis (Fig. 2). Nuclear RFLPs and isozymes did
1 2 3 not confirm the traditional grower's distinct grouping of Tronchuda cabbage and Galega kale landraces. Portuguese coles are of very high morphological diversity. A full range of morphotypes from cabbagelike plants, showing a pseudo-head, to typical kales with rosette leaves and long stems over 1 meter high before bolting can be found among Portuguese coles (Dias, 1992). The morphological taxonomy presented a clear distinction and clustering of all these various morphotypes while nuclear RFLPs grouped the various morphotypes according to their geographic origin despite the evident morphological differences. Nevertheless, in this study, Tronchuda cabbages and Galega kales were always clearly isolated from all other cultivated brassicas (Dias et al., 1992). This taxonomic pattern supports the conclusion that Tronchuda cabbages and Galega kales constitute a distinct and relatively homogenous genetic group within the B. oleracea, sharing the same genetic background. The phenotypic differences, leading to the distinction between Tronchuda cabbages and Galega kales and to the isolation of the various Tronchuda cabbage landraces within the same region, have resulted from selection pressure exercised by the growers, involving a relatively reduced number of genes. This situation may be analogous to the case of other varieties or groups and subgroups within B. oleracea. For instance, broccoli and cauliflower are genetically very similar (Song et al., 1990 ; Gray, 1989) and can only be maintained by continuous selection pressure (Gray, 1989). Therefore, it appears to be justified to classify Tronchuda cabbages (var. costata) and Galega kales (var. acephala) as two distinct varieties within B. oleracea considering their evident phenotypic differences even if they share the same genepool. It should therefore be kept in mind, when sampling accessions for screening over the full width of the genepool, that genetic differences within Portuguese cole landraces depend mainly on their geographic origin and not on phenotypic dissimilarities. The frequent crosses between Galega kales and Tronchuda cabbages, and between different Tronchuda cabbage morphotypes have increased the genetic affinity among populations within the same region. The small genetic variation between landraces from the same region suggests that it is more correct to consider groups of Tronchuda cabbage landraces as coinciding with geographic `niches' rather than to classify all existent landraces according to local terminology and slight phenotypic differences. Applying this concept to the clustering patterns of the three phenograms presented in this paper, five landrace groups of Tronchuda cabbage can be defined : Group 1 includes the landraces 'Penca de Mirandela', 'Penca de Safres', 'Penca de Mirandela da Veia Branca' or 'Penca Pao de Acdcar' and the 'Penca de Chaves' collected in Tras-os-Montes in northern inland Portugal ; Group 2 includes the landraces 'Penca da Povoa' and `Coivao' from the northern coastal area of Minho ; Group 3 includes the landrace `Gloria de Portugal' of Beira Interior province, in the highlands of central Portugal ; Group 4, relatively heterogenous, includes Tronchuda landraces collected from the large area along the coast of central Portugal, Beira Litoral and Estremadura provinces ; and Group 5 includes landraces 'Couve de Valhascos' and 'Couve Murciana' grown in the inland above the Tagus valley and in Alentejo province. When the single association between genetic affinity and geographic origin, as revealed by RFLPs, is considered, these five landrace groups can be reduced still further to three major landrace groups : one includes the cole landraces from the northern Rio Douro region (landrace groups 1 and 2) ; another the landraces from Central Portugal (Groups 3 and 4) ; and the last the cole landraces from the Tagus valley and Alentejo (Group 5). Comparison between taxonomic methods The three taxonomic methods resulted in distinct clustering patterns for Tronchuda cabbage, Galega kale, Algarve cabbage, and common cabbage accessions, as previously described. The clustering pattern in the Nei's genetic distance phenogram generated by isozyme (Fig. 4) demonstrated the poor discriminating power of isozymes and revealed some apparent contradictions with the phenograms based on morphological characters (Fig. 2), and on the RFLPs (Fig. 3). Isozymes coincide with the RFLPs clustering pattern to some extent by clustering some morphologically distinct cole accessions from the same region such as 'Couve Gloria de Portugal' (9, 10) and 'Couve Galega' (24) or 'Couve Portuguesa' (17) and 'Couve de Grelo' (18). Furthermore, isozymes show a clustering pattern similar to the morphological one by clustering some morphologically similar accessions collected from different regions e.g. 'Couve Galega' from Loures (26) and 'Couve Galega' from Sardoal (27). Other clusters resulting from isozymes are unique, e.g. Algarve cabbages (28, 32) are clustered with various Tronchuda cabbages and Galega kales in Group B (Fig. 4). In many instances,
1 24 isozyme clustering pattern seems to be a combination of the clustering patterns obtained by morphological characters and by RFLPs. Isozymes are able to estimate the sib frequencies in hybrid cultivars (Wills & Wizeman, 1980 ; Wills et al., 1979 ; Nijenhuis, 1971) or to separate relatively distant taxa (Menancio, 1987 ; Arias et al., 1987 ; Coulthart, 1979). The small number of characters not randomly distributed in the genome, as indicated by isozymes, appears to be insufficient for discriminating highly heterozygous and relatively close populations such as Portuguese cole landraces. The number of isozymes cannot be increased further to randomly cover the brassica genome due to the existence of too few enzymes and the increased cost of the utilization of several gel systems needed to obtain a higher number of alleles or putative alleles. Therefore, isozymes are not adequate for biosystematic studies within allogamic species. The differences in the clustering patterns obtained by the three taxonomic methods are partially due to the fact that the landrace material studied was subjected to a long and primitive human process of selection. The selection pressure applied to a few phenotypic traits has resulted in a distinctive evolution of morphological, biochemical, and molecular characters. These findings agree with Walters (1988) who could not distinguish Chenopodium neomexicanum from C. palmeri based on allozymes, in spite of the clear morphological differences between these two species. Walters (1988) justifies this discordance by admitting that selection forces have apparently acted on the morphologic phenotype of the two species, but not enough time has elapsed for the accumulation of different allozymic alleles. Ellstrand & Marshall (1985), in their research on the distribution of allozyme variation within and among cultivars of radish (Raphanus sativus L.), also state that the domestication and maintenance of radish cultivars has not only retained a population structure similar to that of wild populations but did also not result in any dramatic divergence in the overall composition of allozyme alleles. RFLPs generate a high number of characters representing a random sample of the full brassica genome. Morphological characters, controlled by a limited number of genes, involving a relatively small part of the genome, clearly separate individual phenotypes. Isozymes are a direct product of the genes and are active between the DNA and the expression of the phenotypic character, considering the biochemical pathway leading to the formation of a certain phenotype. The transitional situation of isozyme phenogram (Fig. 4) between RFLPs (Fig. 3) and morphological characters (Fig. 2) seems to reflect this biochemically intermediate situation. The choice of the method for generating taxonomic characters should be dependent on the objectives of the study and on the type of similarities looked for. Origin and evolution of Portuguese cole landraces The different clustering patterns, as evidence by morphological characters, nuclear RFLPs, and isozymes, provide important information for a better understanding of the evolution of the Portuguese cole landraces. It may be concluded that Tronchuda cabbage and Galega kale landraces have evolved from a common ancestor in different regions under relative geographic isolation. In each region, farmers applied selection pressures according to the utilization of their coles. Cabbages were for human consumption and kales for multiple uses, including animal feed. Morphological and RFLPs clustering patterns reflect both the two faces of this evolutionary process. Morphological characters group landraces according to growers selection pressure, while RFLPs reveal the geographic isolation of the various genepools from which the farmers have obtained their different morphotypes. The common ancestor of Portuguese coles may have been a kale-like plant. Song et al. (1990) recently confirmed, using RFLPs, that Galega kale was the accession most closely related to wild B. oleracea and to B. alboglabra, both considered as the closest ancestors of cultivated B. oleracea by these authors. This fact is also supported by E.S.D. (1875) : `The Portuguese kale or Tronchuda cabbage has such a resemblance with the wild cabbage - that grows almost hanging from the previously united surfaces of the cliffs of Dover and Cape Blanc Nez - that it is easily accepted as the lucky and rich descendent of that rough and poor ancestral'. Tronchuda cabbage landraces are in an early stage of selection from Galega kales. The two cole types differ only in a small number of genes. It has been observed that when no selection is made in Tronchuda cabbages for several generations they easily revert to kale types. This fact is supported by the rudimentary process of seed production followed by the farmers. Portuguese farmers only use plant isolation for selecting Tronchuda cabbage landraces. Galega kales are planted around vegetable gardens or down field along the irrigation furrows in order not to interfere with the other crops, and their seeds are often collected from the
125 best individuals without isolation. Frequent intercrossing between Tronchuda cabbages and Galega kales in the same area results in intermediate types between these two coles. This means that different morphotypes grown within an isolated area have intermated to some degree, causing an increase in genetic affinity between morphotypes. Similar farming systems and farmers needs in the various regions resulted in parallel selection for Tronchuda for a relatively low number of traits inside the various genepools. In addition, it may also be concluded that some farmers wanted to retain particular morphotypes differing in specific traits and have thus maintained them as specific landraces. We believe that within B. oleracea several evolutionary pathways have occurred. Therefore, it can be concluded that Portuguese coles traced an independent pathway from the common ancestor to the present cultivated forms. This evolution was analogous and parallel to that followed by cabbages or by cauliflower/broccoli landraces. Song et al. (1990) and Helm (1963), on the contrary, consider Portuguese coles as an intermediate type between wild and common cabbages. According to these authors common cabbages may have been developed from Portuguese cabbages. This hypothesis does not harmonize with the high affinity revealed in a previous study, using RFLPs, (Dias et al., 1992) between Tronchuda cabbages and Galega kales which formed a compact cluster of Portuguese coles. This cluster was clearly separated from the group of common cabbages, suggesting an only tenuous relationship between them. Based on the results and discussion presented above, Portugal should be regarded as an important region of domestication of specialized leafy coles of B. oleracea. The existing diversity and the pattern of evolution can be compared with those of broccoli and cauliflower on the Italian peninsula as described by Gray (1982) and Crisp (1983). Acknowledgements Partial funding was provided by the National Board of Scientific and Technological Research (Junta Nacional de Investigagdo Cientifica e Tecnol6gica), Lisbon, Portugal, project no. 87293/AGR. References Anonymous, 1896. Guia de Horticultura Pratica. Casa Frederico Daupias. Lisboa, 156 p. Arils, P, J.J. Balandrdn & A. Ordaz, 1987. Species identification of cultivated brassicas with isozyme electrophoresis. Crucifer Newsletter 12 : 26-27. Cerca, M.C., 1946. Subsfdios para a caracterizafao e identificagao de algumas formas cultivadas da especie Brassica oleracea L. nacionais ou de ha muito cultivadas em Portugal. Relat. Final Curso Eng. Agrbnomo. Univ. Tdcnica de Lisboa. Lisboa, 156 p. Coulthart, M.B., 1979. A comparative electrophoretic study of an allopolyploid complex in Brassica. M.Sc. Thesis, University of Alberta. Canada, 114 p. Crisp, P, 1983. The use of an evolutionary scheme for cauliflowers in the screening of genetic resources. Euphytica 31 : 725-734. Dias, J.S., 1992. Taxonomia das Couve Galaico-Portuguesas Utilizando Caracteres Morfol6gicos, Isoenzimas e RFLPs. Tese de Doutoramento, Universidade Tdcnica de Lisboa. Lisboa, 181 p. Dias, J.S., M.B. Lima, K. Song, A.A. Monteiro, PH. Williams & T.C. Osbom, 1992. Molecular taxonomy of Portuguese Tronchuda cabbage and kale landraces using nuclear RFLPs. Euphytica 58 : 221-229. Dias, J.S., A.A. Monteiro & M.B. Lima, 1993. Numerical taxonomy of Portuguese Tronchuda cabbage and Galega kale landraces using morphological characters. Euphytica 69 : 51-68. Dias, J.S., A.A. Monteiro & S. Kresovich, 1994. Genetic diversity and taxonomy of Portuguese tronchuda cabbage and Galega kale landraces using isozyme analysis. Euphytica 75 : 221-230. Ellstrand, N.C. & D.L. Marshall, 1985. The impact of domestication on distribution of allozyme variation within and among cultivars of radish, Raphanus sativus L. Theor. Appl. Genet. 69 : 393-398. E.S.D., 1875. Neglected plants. The Gardeners' Chronicle 3(67) : 462. Gray, A.R., 1982. Taxonomy and evolution of broccoli (Brassica oleracea var. italica). Econ. Bot. 36 : 397-410. Gray, A.R., 1989. Taxonomy and evolution of broccolis and cauliflowers. Baileya23(l) : 28-46. Helm, J., 1963. Morphologisch-Taxonomische Gliederung der Kultursippen von Brassica oleracea L. Die Kulturpflanze 11 : 92-210. IBPGR, 1989. Descriptors for Brassica and Raphanus. International Board for Plant Genetic Resources. Rome, 51 p. May, B., J.E. Marsden & C.G. Shenck, 1988. Electrophoretic procedures, recipes and nomenclature used in Cornell laboratory for ecological and evolutionary genetics. Cornell Univ. New York, 20 p. (Ciclost.). Menancio, D.I., 1987. Isozyme and RFLP studies in the genus glycine Willd. subgenusglycine. Ph.D. Thesis. University of Illinois. Urbana-Champaign, 119 p. Nei, M., 1978. Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 89 : 583-590. Nijenhuis, B.T.E., 1971. Estimation of the proportion of inbred seed in Brussels sprouts hybrid seed by acid phosphate isozyme analysis. Euphytica 20 : 498-507. Osborn, T.C., D.C. Alexander & J.F. Fobes, 1987. Identification of restriction fragment length polymorphisms linked to genes controlling soluble solids content in tomato fruit. Theor. Appl. Genet. 73 : 350-356. Potter, E., 1878. Culture of couve Tronchuda cabbage. The Garden 13 : 494.
126 Rohlf, F.J., 1989. NTSYS-pc. Numerical taxonomy and multivariate analysis system. Version 1.5. Exeter Publishing, Ltd. Setauket, New York, USA. Sneath,P.H.A. & R.R. Sokal, 1973. Numerical taxonomy. The principles and practice of numerical classification. W.F. Freeman. San Francisco, 573 p. Song, K., T.C. Osborn & PH. Williams, 1990. Brassica taxonomy based on nuclear restriction fragment length polymorphisms (RFLPs). 3. Genome relationships in Brassica and related genera and the origin of B. oleracea and B. rapa (syn. campestris). Theor. Appl. Genet. 79 : 497-506. Thorpe, M.L., L.H. Duke & W.D. Beversdorf, 1987. Procedures for the detection of isozymes of rapeseed (Brassica napus and Brassica campestris) by starch gel electrophoresis. Tech. Bull. OAC 887. Dept. Crop Sci. Univ. Guelph. Canada, 66 p. Walters, T.W., 1988. Relationship between isozymic and morphologic variation in the diploids Chenopodium fremontii, C. neomexicanum, C. palmeri and C. watsonii. Amer. J. Bot. 75(l) :97-105. Wills, A.B., S.K. Fyfe & E.M. Wiseman, 1979. Testing Fl hybrids of Brassica oleracea for sibs by seed isozyme analysis. Ann. Appl. Biol. 91 : 263-270. Wills, A.B. & E.M. Wiseman, 1980. Acid phosphatase isoenzymes of Brassica oleracea seedlings and their application for sib testing in Fl hybrids. Ann. Appl. Biol. 94 :137-142. Williams, PH. & C.B. Hill, 1986. Rapid-cycling populations of Brassica. Science 232 :1385-1389.