Genetic diversity within and among maize populations: a comparison between isozyme and nuclear RFLP loci

 In order to compare the potential of enzyme and DNA markers to investigate genetic diversity within and among populations, ten maize populations were characterized for (1) 20 isozyme loci and (2) restriction fragment length polymorphism (RFLP) for 35 probe-enzyme combinations. Each population was represented by a sample of at least 30 individuals. The average number of alleles detected per locus was clearly higher for RFLPs (6.3) than for isozymes (2.4). Similarly, total diversity was higher for RFLPs (0.60) than for isozymes (0.23). This difference is consistent with observations on inbred-line collections and can be related to the fact that many variations at the DNA level do not change the amino-acid composition or the global charge of proteins. By contrast, the magnitude of population differentiation, relative to the total diversity, was similar for isozymes (23%) and RFLPs (22%). This suggests that the isozyme and RFLP loci considered in this study are subject to similar evolutionary forces, and that both are mostly neutral. However, RFLPs proved clearly superior to isozymes both to (1) identify the origin of a given individual and (2) reveal a relevant genetic structure among populations. The higher polymorphism observed for RFLP loci and the greater number of these loci contributed to the superior discriminative ability of the RFLP data. Received: 1 June 1997 / Accepted: 3 November 1997     

Microsatellites and RFLP probes from maize are efficient sources of molecular markers for the biomass energy crop Miscanthus

A survey of Gramineae markers was carried out with the aim of developing cost-effective methods for the molecular analysis of Miscanthus species. Ten out of twenty Gramineae RFLP probes from ”anchor” sets hybridized well to Miscanthus DNA while all 15 maize probes tested cross-hybridized successfully, showing similar patterns in both species. Cross-taxa amplification of maize microsatellite primers was then tested. This showed that 57 out of 76 (75%) give highly reproducible amplification with Miscanthus DNA. Amplification products differed in size from those in maize but there was no bias toward higher or lower molecular weights. Microsatellite polymorphism produced by 17 primer pairs was studied in detail in a panel of 11 Miscanthus clones belonging to the species Miscanthus sinensis, Miscanthus sacchariflorus, Miscanthus ×giganteus and Miscanthus condensatus. Intra- and inter-specific length polymorphisms were frequent between the tested Miscanthus clones with length polymorphisms being found for all primer pairs, detecting 3–22 alleles. Polymorphism information content (PIC) values for microsatellites ranged from 0.48 to 0.94 with an average of 0.83. Species-specific amplicons were produced by two microsatellites. Genetic similarity coefficients of the Miscanthus clones ranged from 0.35 to 0.92, with an average of 0.57. Five polymorphisms were studied in a segregating population, where they showed Mendelian inheritance. In addition, two microsatellite markers mapping 1.3-cM apart on maize chromosome 7 were linked in Miscanthus at an estimated distance of 8 cM, suggesting collinearity. The high transferability of microsatellite markers from maize will enhance the power and resolution of genome analysis in Miscanthus. Received: 14 April 2000 / Accepted: 9 June 2000     

Comparison of RFLP and morphological distances between maize Zea mays L. inbred lines. Consequences for germplasm protection purposes

 A total of 145 maize inbred lines, representative of material released in France, were differentiated using RFLP markers and a set of discriminant morphological traits in order to evaluate the use of molecular markers for large-scale germplasm diversity analysis and determination of distinctness. Several criteria are proposed with respect to choice of probes, which should give reliable results for routine studies and have a known single-locus genetic determinism to avoid redundancy. A method is proposed by which to incorporate the data from different restriction enzymes obtained with the same probe. The precision of the estimation of the genetic distance is given. The relationship between molecular and morphological distances appears to be triangular, molecular divergence behaving as a limiting factor for morphological divergence. This suggested a scheme for incorporating molecular markers in studies of distinctness. Received: 20 August 1996 / Accepted: 20 December 1996     

Parental contribution and coefficient of coancestry among maize inbreds: pedigree, RFLP, and SSR data

The genetic relationship between inbreds i and j can be estimated from pedigree or from molecular marker data. The objectives of this study were to: (1) determine whether pedigree, restriction fragment length polymorphism (RFLP), and simple sequence repeat (SSR) data give similar estimates of parental contribution and coefficient of coancestry (f _ij ) among a set of maize (Zea mays L.) inbreds, and (2) compare the usefulness of RFLP and SSR markers for estimating genetic relationship. We studied 13 maize inbreds with known pedigrees. The inbreds were genotyped using 124 RFLP and 195 SSR markers. For each type of marker, parental contributions were estimated from marker similarity among an inbred and both of its parents, and were subsequently used to estimate f _ij . Estimates of parental contribution differed significantly (α<0.05) between pedigree data and either type of marker, but not between the marker systems. The RFLP estimates of parental contribution failed to sum to 1.0, reflecting a higher frequency of non-parental bands with RFLP than with SSR markers. The f _ij estimated from pedigree, RFLP, and SSR data were highly correlated (r=0.87–0.97), although significant differences were found among the three sets of f _ij estimates. We concluded that pedigree and marker data often lead to different estimates of parental contribution and f _ij , and that SSR markers are superior to RFLP markers for estimating genetic relationship. A relevant question is whether or not the inbreds previously genotyped with an older marker system (e.g., RFLP) need to be re-analyzed with a newer marker system (e.g., SSR) for the purpose of estimating genetic relationship. Such re-analysis seems unnecessary if data for the same type of marker are available for a given inbred and both of its parents. Received: 2 June 1999 / Accepted: 30 July 1999     

RFLP mapping of the centromere of chromosome 4 in maize using isochromosomes for 4S

 The centromere of maize chromosome 4 was previously localized to a 26-cM interval using molecular markers and B-A translocations. The objective of the present study was to refine the placement of the centromere using secondary trisomics. Two independently isolated secondary trisomics (having an isochromosome plus two normal homologs) for 4S were recovered. RFLP analysis of populations segregating for them placed the centromere of chromosome 4 between bnl15.45 and bnl7.20, two RFLP loci that are 5.4-cM apart on the UMC map and 11.5-cM apart on the BNL map. Received: 19 May 1997 / Accepted: 6 October 1997     

Genetic distance of inbred lines and prediction of maize single-cross performance using RAPD markers

 To evaluate the genetic diversity of 18 maize inbred lines, and to determine the correlation between genetic distance and single-cross hybrid performance, we have used random amplified polymorphic DNA (RAPD), a PCR-based technique. Eight of these lines came from a Thai synthetic population (BR-105), and the others derived from a Brazilian composite population (BR-106). Thirty two different primers were used giving a total of 325 reproducible amplification products, 262 of them being polymorphic. Genetic divergence was determinated using Jaccard’s similarity coefficient, and a final dendrogram was constructed using an unweighted pair-group method with arithmetical averages (UPGMA). Cluster analysis divided the samples into three distinct groups (GI, GII and GIII) that were confirmed by principal-coordinate analysis. The genetic distances (GD) were correlated with important agronomic traits for single-cross hybrids and heterosis. No correlation was found when group division was not considered, but significant correlations were detected between GI×GII and GI×GIII GDs with their respective single-cross hybrid grain-yield values. Three groups were identified; that is, the BR-106 population was divided in two different groups and the BR-105 population remained mostly as one group. The results indicated that RAPD can be used as a tool for determining the extent of genetic diversity among tropical maize inbred lines, for allocating genotypes into different groups, and also to aid in the choice of the superior crosses to be made among maize inbred lines, so reducing the number of crosses required under field evaluation. Received: 24 May 1996 / Accepted: 22 November 1996     

Evaluation of the extent of genetic variability among Theobroma cacao accessions using RAPD and RFLP markers

 Random amplified polymorphic DNA (RAPD) and restriction fragment length polymorphism (RFLP) markers were used to evaluate genetic relationships within the Theobroma cacao species and to assess the organization of its genetic diversity. Genetic variability was estimated with 18 primers and 43 RFLP probes on 155 cocoa trees belonging to different morphological groups and coming from various geographic origins. The majority of the RFLP probes issued from low-copy DNA sequences. On the basis of on the genetic distance matrices, the two molecular methods gave related estimates of the genetic relationship between genotypes. Although an influence of cocoa morphological groups and geographical origins of trees was observed, a lack of gene differentiation characterized the T. cacao accessions studied. The continuous RFLP variability observed within the species may reflect the hybridization and introgressions between trees of different origins. Nevertheless, the Nacional type was detected to be genetically specific and different from well-known types such as Forastero, Criollo and Trinitario. Some of those genotypes were characterized by a low heterozygosity rate and may constitute the original Nacional pool. These results also provide information for the constitution of a cocoa tree core collection. Received: 10 June 1996/Accepted: 11 October 1996     

Towards a saturated sorghum map using RFLP and AFLP markers

 A near-saturated sorghum genetic linkage map was produced using RFLP, AFLP and morphological markers. First a composite, essentially RFLP-based genetic linkage map was obtained from analyses of two recombinant inbred populations. This map includes 343 loci for 11 linkage groups spanning 1352 cM. Since this map was constructed with many previously mapped heterologous probes, it offers a good basis for synteny studies. Separately, an AFLP map was obtained from the analysis of 168 bands revealed from 12 primer pair combinations. It includes 137 loci for 11 linkage groups spanning 849 cM. Taking into account the different data sets, we constructed a combined genetic linkage map including 443 loci spanning 1899 cM. Two main features are to be noted: (1) the distribution of AFLPs along the genome is not uniform; (2) an important stretching of the former core map is induced after adding the AFLPs. Received: 10 May 1998 / Accepted: 13 July 1998     

RFLP markers and predicted testcross performance of maize sister inbreds

 Inbreds selfed from the same F₂ or backcross population are referred to as sister inbreds. In some situations, maize (Zea mays L.) sister inbreds may not have testcross data available for best linear unbiased prediction (BLUP) of single-cross performance. This study evaluated the usefulness of BLUP and restriction fragment length polymorphism (RFLP)-based coefficients of coancestry ( f ) in predicting the testcross performance of sister inbreds. Parental contributions (p) were estimated from 70 RFLP loci for 15 inbreds that comprised three sister inbreds selfed from each of five F₂ populations. Estimates of p were subsequently used to calculate RFLP-based f. Grain yield, moisture, and stalk lodging data were obtained for 2265 single crosses tested by Limagrain Genetics in multilocation trials from 1990 to 1995. Performance of the sister inbreds when crossed to several inbred testers was predicted from the performance of the tested single crosses and RFLP-based f. Correlations between predicted and observed performance, obtained with a delete-one cross-validation procedure, were erratic and mostly low for all three traits. Correspondence was poor between ranks for predicted and observed general combining ability of the sister inbreds. The results suggested that the proportion of the genome derived by a sister inbred from a given parental inbred does not solely determine its testcross performance. The failure of BLUP and RFLP-based f to consistently predict testcross performance indicated that actual field testing will continue to be necessary for preliminary evaluation of sister inbreds. Received : 17 March 1997 / Accepted : 18 April 1997     

A maize bacterial artificial chromosome (BAC) library from the European flint inbred line F2

We report here the construction and characterisation of a BAC library from the maize flint inbred line F2, widely used in European maize breeding programs. The library contains 86,858 clones with an average insert size of approximately 90 kb, giving approximately 3.2-times genome coverage. High-efficiency BAC cloning was achieved through the use of a single size selection for the high-molecular-weight genomic DNA, and co-transformation of the ligation with yeast tRNA to optimise transformation efficiency. Characterisation of the library showed that less than 0.5% of the clones contained no inserts, while 5.52% of clones consisted of chloroplast DNA. The library was gridded onto 29 nylon filters in a double-spotted 8 × 8 array, and screened by hybridisation with a number of single-copy and gene-family probes. A 3-dimensional DNA pooling scheme was used to allow rapid PCR screening of the library based on primer pairs from simple sequence repeat (SSR) and expressed sequence tag (EST) markers. Positive clones were obtained in all hybridisation and PCR screens carried out so far. Six BAC clones, which hybridised to a portion of the cloned Rp1-D rust resistance gene, were further characterised and found to form contigs covering most of this complex resistance locus. Received: 30 August 2000 / Accepted: 6 December 2000     

Comparative mapping in loblolly and radiata pine using RFLP and microsatellite markers

Genetic linkage maps were constructed for loblolly pine (Pinus taeda L.) and radiata pine (P. radiata D. Don) using a common set of RFLP and microsatellite markers. The map for loblolly pine combined data from two full-sib families and consisted of 20 linkage groups covering 1281 cM. The map for radiata pine had 14 linkage groups and covered 1223 cM. All of the RFLP probes readily hybridise between loblolly and radiata pine often producing similar hybridisation patterns. There were in total 60 homologous RFLP loci mapped in both species which could be used for comparative purposes. A set of 20 microsatellite markers derived from radiata pine were also assayed; however, only 9 amplified and revealed polymorphic loci in both species. Single-locus RFLP and microsatellite markers were used to match up linkage groups and compare order between species. Twelve syntenic groups were obtained each consisting of from 3 to 9 homologous loci. The order of homologous loci was colinear in most cases, suggesting no major chromosomal rearrangements in the evolution of these species. Comparative mapping between loblolly and radiata pine should facilitate genetic research in both species and provide a framework for mapping in other pine species. Received: 25 November 1998 / /Accepted: 19 December 1998     

Comparative QTL mapping of resistance to Ustilago maydis across four populations of European flint-maize

 We mapped and characterized quantitative trait loci (QTLs) for resistance to Ustilago maydis and investigated their consistency across different flint-maize populations. Four independent populations, comprising 280 F₃ lines (A×B^I), 120 F₅ lines (A×B^II), 131 F₄ lines (A×C) and 133 F₄ lines (C×D), were produced from four European elite flint inbreds (A, B, C, D) and genotyped at 89, 151, 104, and 122 RFLP marker loci, respectively. All F_n lines were evaluated in field trials with two replications in five German environments. Genotypic variances were highly significant for the percentage of U. maydis infected plants (UST) in all populations, and heritabilities exceeded 0.69. Between five and ten QTLs were detected in individual populations by composite interval mapping, explaining between 39% and 58% of the phenotypic variance. These 19 different QTLs were distributed over all ten chromosomes without any clustering on certain chromosomes. In most cases, gene action was dominant or overdominant. Fourteen pairs of the detected QTLs for UST displayed significant digenic epistatic interactions, but only two of them did so after arcsin √UST/100 transformation. Significant QTL× environment interactions occurred frequently. Between two to four QTLs were common between pairs of populations. Population C×D was also grown in Chartres, a location with a high U. maydis incidence. Two out of six QTLs identified for Chartres were in common with QTLs detected across five German environments for C×D. Consequently, marker-assisted or phenotypic selection based on results from natural infection seem to be suitable breeding strategies for improving the resistance of maize to U. maydis. Received: 3 July 1998 / Accepted: 24 July 1998     

Mapping quantitative trait loci controlling silking date in a diallel cross among four lines of maize

 We describe and apply an interval mapping method for quantitative trait locus (QTL) detection using F₃ and testcross progenies derived from F₂ populations obtained from a diallel cross among four elite lines of maize. Linear model-based procedures were used for the test and estimation of putative QTL effects together with genetic interactions including epistasis. We mapped QTL associated with silking date and explored their genetic effects. Ten QTL were detected, and these explained more than 40% of the phenotypic variance. Most of these QTL had consistent and stable effects among genetic backgrounds and did not show significant epistasis. QTL-by-environment interaction was important for four QTL and was essentially due to changes in magnitude of allelic effects. These results show the efficiency of our method in several genetic situations as well as the power of the diallel design in detecting QTL simultaneously over several populations. Received: 2 September 1996 / Accepted: 20 December 1996     

Molecular diversity in pineapple assessed by RFLP markers

Pineapple, Ananas comosus (L.) Merr, is the third most important tropical fruit cultivated in all tropical and subtropical countries. Pineapple germplasm includes all seven species of the genus Ananas and the unique species of the related genus Pseudananas. A knowledge of its diversity structure is needed to develop new breeding programs. Restriction fragment length polymorphism (RFLP) was used to study molecular diversity in a set of 301 accessions, most of which were recently collected. This sample was analysed using 18 homologous genomic probes. Dissimilarities were calculated by a Dice index and submitted to Factorial Analysis. The same data were represented as a diversity tree constructed with the score method. Pseudananas sagenarius displayed a high polymorphism and shares 58.7% of its bands with Ananas. Within Ananas, variation appears continuous and was found mostly at the intraspecific level, particularly in the wild species Ananas ananassoides and Ananas parguazensis. As for the cultivated species, Ananas comosus appears relatively homogeneous despite its wide morphological variation and Ananas bracteatus, which is grown as a fence and for fruit, appears still much less variable. By contrast Ananas lucidus, cultivated by the Amerindians for fiber, displays a high polymorphism. This tree displayed a loose assemblage of numerous clusters separated by short distances. Most species were scattered in various clusters, a few of these being monospecific. Some accessions which had not been classified, as they shared morphological traits typical of different species, re-group with one or the other, and sometimes with both species in mixed clusters. No reproductive barrier exists in this germplasm and these data indicate the existence of gene flow, enhancing the role of effective sexual reproduction in a species with largely predominant vegetative mutiplication. Received: 8 March 2000 / Accepted: 14 April 2000     

Comparing AFLP, RAPD and RFLP markers for measuring genetic diversity in melon

Three different types of molecular markers, RAPD, AFLP and RFLP were used to measure genetic diversity among six genotypes of Cucumis melo L. Each line represented a different melon genotype: Piel de Sapo, Ogen, PI161375, PI414723, Agrestis and C105. A number of polymorphic RAPD, AFLP and RFLP bands were scored on all materials and the genetic similarity measured. Clustering analysis performed with the three types of markers separated the genotypes into two main groups: (1) the sweet type, cultivated melons and (2) the exotic type, not cultivated melons. While the data obtained suggest that all three types of markers are equally informative, AFLPs showed the highest efficiency in detecting polymorphism. Received: 30 December 1999 / Accepted: 24 January 2000     

Molecular marker diversity among current and historical maize inbreds

Advanced-cycle pedigree breeding has caused maize (Zea mays L.) inbreds to become more-elite but more-narrow genetically. Our objectives were to evaluate the genetic distance among current and historical maize inbreds, and to estimate how much genetic diversity has been lost among current inbreds. We selected eight maize inbreds (B14, B37, B73, B84, Mo17, C103, Oh43 and H99) that largely represented the genetic background of current elite inbreds in the U.S. seed industry. A total of 32 other inbreds represented historical inbreds that were once important in maize breeding. Cluster analysis of the inbreds, using data for 83 SSR marker loci, agreed well with pedigree information. Inbreds from Iowa Stiff Stalk Synthetic (BSSS), Reid Yellow Dent, and Lancaster clustered into separate groups with only few exceptions. The average number of alleles per locus was 4.9 among all 40 inbreds and 3.2 among the eight current inbreds. The reduction in the number of alleles per locus was not solely due to sample size. The average genetic distance (D _ij ) was 0.65 among the eight current inbreds, 0.67 among the 32 historical inbreds, and 0.67 among all 40 inbreds. These differences were statistically insignificant. We conclude that genetic diversity among current inbreds has been reduced at the gene level but not at the population level. Hybrid breeding in maize maintained, rather than decreased, genetic diversity, at least during the initial subdivision of inbreds into BSSS and non-BSSS heterotic groups. We speculate, however, that exploiting other germplasm sources is necessary for sustaining long-term breeding progress in maize. Received: 21 August 2000 / Accepted: 5 January 2001     

Mapping of the Rf-3 nuclear fertility-restoring gene for WA cytoplasmic male sterility in rice using RAPD and RFLP markers

The cytoplasmic male sterility (CMS) of wild-abortive (WA) cytoplasm has been widely used for breeding hybrid rice. Two restorer genes for the CMS have been found by traditional genetic analysis. To tag the restorer genes we used a set of near-isogenic lines (NILs) of Zhenshan 97 carrying different genotypes for fertility restoration from IR24, to perform RAPD analysis. From the survey of 720 random primers, six RAPD markers were identified to be associated with Rf-3. Three of these OPK05-800, OPU10-1100 and OPW01-350, were mapped on chromosome 1. Two populations from the crosses between Zhenshan 97 A and a near-isogenic restorer line ZSR21 and between Zhenshan 97 A and IR24 were used for mapping Rf-3. The three RAPD markers and three RFLP markers, RG532, RG140 and RG458, were found to be closely linked to Rf-3 in the two populations. The same location of Rf-3 was also found in a population from the cross of IR58025 A//IR36/IR58025 B. At the RG532 locus, different alleles were found between two CMS lines, Zhenshan 97 A and IR58025 A, and between two restorer lines, IR24 and IR36. The use of these molecular markers closely linked to Rf-3 in facilitating the development of hybrid rice is discussed. Received: 3 January 1996 / Accepted: 17 May 1996     

Mapping of three self-fertility mutations in rye (Secale cereale L.) using RFLP, isozyme and morphological markers

 Three mutations determining self-fertility at the S, Z and S5 self-incompatibility loci on chromosomes 1R, 2R and 5R of rye, respectively, were mapped using three different F₂ populations. There was a close linkage of one isozyme and four RFLP markers, and no recombinant plants were detected. These markers are Prx7, Xiag249 and Xpsr634 for the S locus (1R), Xbcd266 for the Z locus (2R) and Xpsr100 for the S5 locus (5R). Linkage data for markers associated to the self-fertility mutations at the S, Z and S5 loci were calculated and compared with genetic maps computed by MAPMAKER multipoint analysis. Received: 8 October 1997 / Acepted: 26 November 1997     

Inheritance of resistance to southern corn rust in tropical-by-corn-belt maize populations

 The inheritance of resistance to southern rust (caused by Puccinia polysora Underw.) was investigated in two F_2:3 populations derived from crossing two temperate-adapted, 100% tropical maize (Zea mays L.) inbred lines (1416-1 and 1497-2) to a susceptible Corn Belt Dent hybrid, B73Ht×Mo17Ht. The inbred lines possess high levels of resistance to southern rust and may be unique sources of resistance genes. Heritability for resistance was estimated as 30% and 50% in the two populations from regression of F_2:3 family mean scores on F₂ parent scores, and as 65% and 75% from variances among F_2:3 families on a single-plot basis. RFLP loci on three chromosomal regions previously known to possess genes for resistance to either southern rust or common rust (P. sorghi Schw.) were used to localize genes affecting resistance to southern rust in selected genotypes of both populations, and to estimate their genetic effects. A single locus on 10S, bnl3.04, was associated with 82–83% of the variation among field resistance scores of selected F_2:3 families in the two populations. Loci on chromosomes 3 (umc26) and 4 (umc31) were significantly associated with resistance in the 1497-2 population, each accounting for 13–15% of the phenotypic variation for F_2:3 field scores. Multiple-marker locus models, including loci from chromosomes 3, 4, and 10 and their epistatic interactions, accounted for 96–99% of the variation in F_2:3 field scores. Similar results were obtained for resistance measured by counting pustules on juvenile plants in the greenhouse. An attempt was made to determine if the major gene for resistance from 1416-1 was allelic to Rpp9, which is also located on 10S. Testcross families from the cross (1416-1×B37Rpp9)×B14AHt were evaluated for resistance to southern rust in Mexico. Neither source of resistance was completely effective in this environment, preventing determination of allelism of the two genes; however, both sources of resistance had better partial resistance to southern rust than did B14AHt. Received: 6 May 1997/Accepted: 19 September 1997     

Genetic analysis of adaptation differences between highland and lowland tropical maize using molecular markers

Molecular-marker loci were used to investigate the adaptation differences between highland and lowland tropical maize. An F₂ population from the cross of two inbred lines independently derived from highland and lowland maize germplasm was developed, and extracted F_3:4 lines were phenotype in replicated field trials at four thermally diverse tropical testing sites, ranging from lowland to extreme highland (mean growing season temperature range 13.2–24.6°C). Traits closely related with adaptation, such as biomass and grain yield, yield components, days from sowing to male and female flowering, total leaf number, plant height and number of primary tassel branches (TBN), were analyzed. A large line × environment interaction was observed for most traits. The genetic basis of this interaction was reflected by significant, but systematic, changes from lowland to highland sites in the correlation between the trait value and genomic composition (designated by the proportion of marker alleles with the same origin). Joint analysis of quantitative trait loci (QTLs) over sites detected 5–8 QTLs for each trait (except disease scores, with data only from one site). With the exception of one QTL for TBN, none of these accounted for more than 15% of the total phenotypic variation. In total, detected QTLs accounted for 24–61% of the variation at each site on average. For yield, yield components and disease scores, alleles generally favored the site of origin. Highland-derived alleles had little effect at lowland sites, while lowland-derived alleles showed relatively broader adaptation. Gradual changes in the estimated QTL effects with increasing mean site temperature were observed, and paralleled the observed patterns of adaptation in highland and lowland germplasm. Several clusters of QTLs for different traits reflected the relative importance in the adaptation differences between the two germplasm types, and pleiotropy is suggested as the main cause for the clustering. Breeding for broad thermal adaptation should be possible by pooling genes showing adaptation to specific thermal regimes, though perhaps at the expense of reduced progress for adaptation to a specific site. Molecular marker-assisted selection would be an ideal tool for this task, since it could greatly reduce the linkage drag caused by the unintentional transfer of undesirable traits. Received: 10 October 1998 / Accepted: 9 April 1999