Molecular characterisation of indigenous Swedish apple cultivars based on SSR and S-allele analysis

Trees of 68 apple cultivars, aimed for preservation by the ‘National Program for diversity of cultivated plants’ as mandate cultivars, were analysed using a set of 10 SSR (simple sequence repeat) primer pairs and the self-incompatibility (S )locus to evaluate genetic diversity and reveal inter-cultivar relationships. The 12 polymorphic SSR loci exhibited 2 to 15 alleles, with expected heterozygozity (He) ranging from 0.36 to 0.88 and a mean of 0.74. Numerous alleles were classified as rare or unique (35% and 18% respectively). For the S-locus, a total of 14 alleles were identified in this study. Five alleles, S1 S3, S5 and S7 had frequencies ranging from 11 to 18%, whereas the remaining 9 alleles were below 6%. All sexually obtained cultivars could be distinguished with the set of SSR loci. Sports were identical with their progenitors in two cases, but differed in one SSR allele in a third case. An SSR-based dendrogram, based on Roger’s genetic distances, did not reveal any clear pattern of clustering. The genetic distances were, however, correlated with a corresponding matrix obtained in a previously conducted RAPD-based study of the same cultivars. Non-mandate parents of Swedish mandate cultivars together with some other reference cultivars were included in this study to check the accuracy of allele scoring, verify parentage and compare the results of this study with those presented in previously published studies. Some discrepancies in allele sizing were revealed and the possibilities of avoiding this problem are discussed.

Accurate and permanent identification of plant material within a germplasm collection is of utmost importance, especially for vegetatively propagated cultivars which are expensive to maintain, and should consist of a single genotype in the whole distribution area. In Sweden, the preservation of clonally propagated plants is presently managed by a governmentally appointed unit, the 'National Program for Diversity of Cultivated Plants'. Mandate status has been granted to indigenous cultivars which have been named, bred, propagated and marketed in Sweden. Some foreign cultivars with a long history of being grown in Sweden are also included (HJALMARSSON and WALLACE 2004). In apple (Malus )domestica Borkh.), about 220 mandate cultivars have been appointed and are now conserved mainly in 11 clone archives (each with 10Á50 mandate cultivars) located at outdoor museums or other public places. The largest apple collection in Sweden, with about 1000 cultivars in total, is however found at Balsgård in the southernmost province Skåne, where both research and applied breeding of new cultivars is undertaken. Of the 100 mandate cultivars in the germplasm collection at Balsgård, 68 have apparently been developed in Sweden, and are regarded as having a Swedish origin although one or both parents may have been developed elsewhere.
Unfortunately, many of the older (local) cultivars have not been properly identified, and there are problems with e.g. synonyms and mis-labellings in these plant collections. In order to ensure that our research data, obtained from rather costly analyses of e.g. chemical contents in apple cultivars (NYBOM et al. in press), can be unambiguously connected with well-identified genotypes, we need to provide these cultivars with proper and lasting identification. Morphological characters have historically been, and still are, used for identification of apple cultivars (NILSSON 1986;SVENSSON and KASTMAN 2005), but environmentally induced variation makes correct identification difficult. Therefore, molecular protein-based markers like isozymes (WEEDEN and LAMB 1985), and DNA-based markers have become increasingly popular in the characterisation of apple genetic resources.
Several molecular DNA techniques have been employed for characterisation of apple germplasm collections, as well as in apple genetics and breeding, i.e. RFLP (NYBOM and SCHAAL 1990), RAPD (KOLLER et al. 1993), ISSR (GOULAO and OLIVEIRA 2001), AFLP (XU and KORBAN 2000) and SSR (LIEBHARD et al. 2002). Each of these methods has its strengths and weaknesses. During the past 10Á15 years, microsatellites or simple sequence repeats have become the markers of choice for verification of cultivar identity and for diversity studies due to their abundance in plant genomes, large number of alleles per locus and high informativeness, codominant inheritance, and suitability for automatization (WEISING et al. 2005). SSR loci are relatively easy to score, and alleles are inherited in a Mendelian manner, which allows the verification and reconstruction of cultivar pedigrees. In addition, triploid and tetraploid individuals are revealed by the appearance of more than two alleles per locus. In apple, SSR markers have been used for assessment of genetic diversity in germplasm collections (GUARINO et al. 2006;PEREIRA-LORENZO et al. 2007), cultivar identification (GUILFORD et al. 1997;GALLI et al. 2005), fingerprinting of apple rootstocks (ORAQUZIE et al. 2005), construction of genetic linkage maps (KENNIS and KEULEMANS 2005;SILFVERBERG-DILWORTH et al. 2006) and for parent identification (KITAHARA et al. 2005).
Another, highly polymorphic locus is the S-locus, which encodes for the different S-RNases that determine the S-alleles of different genotypes and causes self-incompatibility. Apple cultivars that share one S-allele have reduced compatibility and do not achieve their potential yield capacity if planted together with no other cultivars close-by (SCHNEIDER et al. 2005). If both S-alleles are shared, the cultivars are usually incompatible and do not yield at all except for occasional fruits obtained by bypassing the selfincompatibility system through pollination between genetically incompatible genotypes or through selfing. In apple, allele-specific primer pairs have been developed and used in a number of cultivar screenings (JANSSENS et al. 1995; VAN NERUM et al. 2001;BROOTHAERTS 2003;BROOTHAERTS et al. 2004;MELOUNOVA et al. 2005;MATSUMOTO et al. 2007). Information about S-locus allele composition can be used for identification of apple cultivars and determination of parentage (KITAHARA et al. 2005) if complemented with other marker systems.
The purpose of this study was 1) to fingerprint 68 mandate apple cultivars with SSR and S-locus markers and 2) to determine genetic diversity and examine genetic relationships within the set of mandate cultivars.

DNA-isolation
Newly expanded leaves were collected in AprilÁMay and stored at (808C until use. Leaves were powdered with liquid nitrogen in pre-cooled mortars. Genomic DNA was isolated from approximately 100 mg of leaf powder using the Qiagen Dneasy TM Plant Mini Kit and following the manufacturer's protocol.

SSR analysis
Ten SSR primer pairs were used in this study (Table 2). All primer pairs were obtained from the list of 140 (LIEBHARD et al. 2002): CH01d03, CH01h02, CH02b10, CH02c06, CH02c09, CH02c11, Ch02d08, CH04c06, CH04e05, COL. All but one, CH02b10, belong to the ''standard set'' defined by the European working group on apple genetic resources (LAURENS et al. 2004;GUARINO et al. 2006). The primer pair CH02b10 had previously been used successfully at our laboratory (MATTISSON and NYBOM 2005) and was therefore included in this study.
All PCR amplifications were performed in a 15 ml volume according to GIANFRANCESCHI et al. (1998). A Px2 Thermal Cycler (Thermo Hybaid) was programmed according to the original protocol (GIANFRANCESCHI et al. 1998) for all primer pairs, except CH02c06. The cycling profile for Ch02c06 was modified according to K. Antonius, MTT, Finland (pers. comm.) and consisted of an initial denaturation at 948C for 2 min 30 s followed by 34 cycles of 948C for 30 s, 588C for 1 min, 728C for 1 min. A final extention at 728C for 5 min was included. PCR products were preliminary checked by running on 2% agarose gel in a 1 )TBE buffert, stained with ethidium bromide and visualised by UV light.
The forward primers were fluorescently labelled at the 5?-end with either FAM (Ch01d03, CH04c06, Ch04e05) or HEX (CH01H02, CH02b10, CH02c06,  Ch02c09, CH02c11, CH02d08, COL). The PCR products were separated and analysed on a 3730 DNA Analyzer (Applied Biosystems). The size of the amplified products was calculated based on an internal standard (500ROX TM Size Standard (Applied Biosystems)) with GeneMapper † Software ver. 3.0 (Applied Biosystems). A manual binning step was included to assign all detected alleles to their repeat units equivalent.
To monitor the reproducibility between runs, we included the same three cultivars, namely 'Discovery', 'Golden Delicious' and 'Guldborg' in every run. In addition, two cultivars (different in each run) were amplified twice.
Allele-specific primers (MWG) were employed for detection of alleles S 1 ÁS 5 , S 7 , S 9 , S 10 , S 16 , S 20 , S 22 , S 24 , S 25 and S 28 using the nucleotide sequences published by BROOTHAERTS (2003), except for S 5 and S 25 which were detected with primers published by MATSUMOTO et al. (1999) and by KITAHARA and MATSUMOTO (2002), respectively.
PCR conditions used for all primers except for S 5 and S 25 , included 1xPCR buffer (ABgene), 1.75 mM of MgCl 2 (1.25 mM for S 20 ), 200 mM dNTPs, 1 mM of each primer and 0.6 U of Taq DNA polymerase in 20 ml total volume. Approximately 40 ng of genomic DNA template were used per reaction. For primers S 5 and S 25 , a 35 ml total volume was used including 1) buffer, 1.07 mM MgCl 2, , 200 mM dNTPs, 1 mM of each primer and 2.5 U of Taq DNA polymerase. Five different PCR amplification protocols (Table 3) were run on a Px2 Thermal Cycler.
A protocol published by BROOTHAERTS (2003) for the discrimination of five different alleles, S 6 , S 14 , S 17 , S 20 and S 21 using the single primer pair S 6/14/17/20/21 , did not work properly in our laboratory. By lowering the amount of MgCl 2 to 1.25 mM, a clear band was however produced in controls reported to have S 20 and not in any other. This procedure was therefore used to detect S 20 in the investigated cultivars.
Detection of the three alleles S 4 , S 16 and S 22 was carried out using a single primer pair, S 4/16/22 , and subsequently digesting 20 ml of the amplification product with 10 U restriction endonuclease Taq1  Table 2. Genetic diversity estimators of 11 polymorphic SSR loci. (Sigma) in 3 ml 10) restriction endonuclease buffer and a total volume of 25 ml for 2 h at 658C. The amplified and digested fragments were separated on a 2.5% agarose gel stained with ethidium bromide. The gels were run for approximately 2Á3 h, for the digested fragments 4 h, at 120Á150 V, together with a DNA molecular weight marker, and then analyzed on a UV-light table.

Flow cytometry
Cultivars that amplified more than two alleles in an SSR locus and/or the S-locus were analysed for ploidy level. Leaf samples were subjected to flow cytometry analysis by Plant Cytometry Services (JG Schijndel, the Netherlands). The leaves were chopped in icecold buffer with DAPI. Flow cytometry was performed on a CyFlow ML (Partec GmbH, Münster, Germany) using Lactuca sativa as internal standard.

Data analysis
The SSR fragments were scored in terms of loci and alleles and thus allelic composition of each cultivar was determined. To evaluate the genetic diversity within our set of cultivars, we used the following parameters: alleles per locus (A), the effective number of alleles per locus (A e ), expected heterozygosity (H e ) and observed heterozygosity (H o ).
Effective number of alleles per locus (A e ) was calculated as in ARANZANA et al. (2003): where p is the frequency of the ith allele at a locus.
Expected heterozygosity (H e ), i.e. the probability that two alleles from the same locus would be different when chosen at random, was calculated for each SSR locus according to NEI (1973): Observed heterozygosity (H o ) was calculated by dividing the number of heterozygous individuals by the number of individuals scored.
Power of discrimination (PD), which estimates the probability that two randomly sampled individuals are discriminated, was calculated for each locus according to KLOOSTERMAN et al. (1993): where P i is the frequency of the ith genotype.
Confusion probability, which is the probability that any two cultivars are identical in their SSR genotypes at all loci by chance alone, was calculated based on PD values as in ARANZANA et al. (2003): where PD i is the PD value at the ith locus.
Genetic distances using Rogers' dissimilarity coefficient (ROGERS 1972) were calculated and a matrix was produced. Rogers' coefficient is linearly related to the coefficient of coancestry, and is regarded as appropriate for the uncovering of pedigree relationships among operational taxonomic units such as the detection of essentially derived varieties in plant breeding or the identification of duplicates and collection gaps in seed banks (REIF et al. 2005). The matrix was used in an UPGMA analysis performed using the NTSYS-pc statistical package ver. 2.2 (ROHLF 2005) and a dendrogram, illustrating the genetic relationships among cultivars, was produced.
In addition, the SSR fragments were also scored phenotypically, with 1 for presence of a fragment and 0 for absence. Jaccard's coefficient of similarity was used to produce a similarity matrix.
Correlation between the two matrices obtained with Rogers' genetic distances and with Jaccard's similarity coefficients, respectively, was investigated with a Mantel test (MXCOMP in NTSYS-pc, using 9999 permutations to compute the significance of a given correlation). In addition, Mantel tests were also carried out to compare each of the two SSR-based matrices with a RAPD-based Jaccard's similarity matrix produced from data on the same trees in a previous study (GARKAVA-GUSTAVSSON and NYBOM 2007).

SSR polymorphism
All but one of the primer pairs amplified clear and easily scored fragments. The accuracy of scoring was checked by comparing SSR profiles of cultivars with the profiles of their known or putative parents when available, and with the reference cultivars. We did, however, experience difficulties when scoring fragments amplified by primer pair CH01d03. According to LIEBHARD et al. (2002), this primer pair amplifies more than one locus and it was difficult to assign some of the alleles to the correct locus. Therefore, fragments amplified by this primer pair were not considered in the statistical analyses, although they can still be used for identification purposes. Primer pairs CH01h02, CH02c11 and CH04c06 amplified two loci each. The main locus, i.e. the locus described in LIEBHARD et al. (2002), is denoted as locus 2 (L2) in the present study. The other locus, denoted as locus 1 (L1), had shorter fragments and was less polymorphic than L2 for CH01h02 and CH04c06, and was monomorphic for CH02c11. The monomorphic locus CH02c11-L1 was not included in the statistical analyses.
Thus, 11 polymorphic loci amplified by 9 primer pairs were analysed statistically, yielding a total of 113 polymorphic alleles in the set of 68 cultivars ( Table 2) In general, allele frequencies were distributed unevenly within the investigated loci (Fig. 1). Each locus had 1 to 4 more common alleles, which in some cases (CH01h02-L1, CH01h02-L2, CH02d08 and CH04e05) had considerably higher frequencies than the other alleles in these loci. All but one (CH01h02-L1) had rare alleles, here defined as having a frequency below 0.05. All but two loci, L1 and L2, amplified by primer CH01h02, had unique alleles, i.e. present in one cultivar but not in any other. Loci CH02c09, CH02c11 and CH04c06-L2 had 1 unique allele each, and 2, 4 and 5 rare alleles respectively. Loci CH04e05 and CH02b10 had 3 unique alleles each, and 7 and 5 rare alleles respectively. Finally, loci CH02c06 and CH02d08 had 5 unique alleles each, and 5 and 4 rare alleles respectively. Thus, in the whole set of cultivars, only 15 alleles (approx. 13%) were found at frequencies above 0.20, 38 alleles (approx. 34%) were found at frequencies between 0.05 and 0.20, 39 alleles (35%) were rare and, finally, 21 alleles (18%) were unique.

SSR-based cultivar identification
Allelic compositions for each analysed apple cultivar are presented in Table 1. Diploid cultivars amplified one or two fragments in each locus. When only one fragment was detectable in a diploid cultivar, we considered the cultivar to be homozygous. Still, presence of null alleles cannot be excluded, and therefore our heterozygosity values can be underestimated. A putative homozygous null allele was detected in cultivar 'Granatäpple, Kungsbacka' at L1 amplified by primer pair CH01h02. The amplification reaction was repeated four times using two different 96-well plates but no PCR-product was obtained.
Three distinct alleles were revealed at 3 to 8 loci, indicating triploidy, in one of the reference cultivars ('Boskoop') and in 7 of the Swedish cultivars ('Frösåker', 'Holländaräpple', 'Kalmar Glasäpple', 'Kinnekulle Kantäpple', 'Norrstack', 'Villands Glasäpple'and 'Vrams Järnäpple') ( Table 1). The triploid status of these cultivars was confirmed by flow cytometry analyses. We also checked one cultivar that was already known to be tetraploid, 'Alfa 68'. By comparing the SSR profiles of this cultivar with the profiles of its two parents, 'Boskoop' and 'Filippa', we concluded that the maternal parent, 'Boskoop', had contributed an unreduced triploid egg cell while' Filippa' had contributed a normal, haploid pollen. In total, 'Alfa 68' amplified 4 different alleles at 6 loci, 3 alleles at 4 loci and 2 alleles in 2 loci. In those cases where only two alleles were amplified in a  Table 1. Hereditas 145 (2008) heterozygous triploid cultivar, we usually managed to determine which allele occurred in two copies by comparing allele peak areas. However, the two alleles COL-230 or COL-232 had similar peak areas in our reference triploid cultivar 'Boskoop' and in its offspring, 'Alfa 68'.
Analysis of parents for some of the cultivars in our material contributed to the accuracy of allele scoring. All alleles found in 'Alfa 68', 'Eva-Lotta', 'Katja', 'Kim', 'Mio' and 'Sylvia' were thus found also in their parents except for two loci (CH01h02-L1 and COL). At locus CH01h02-L1, one of the parents of 'Eva-Lotta', 'Kim' and 'Sylvia' showed only one allele and this was not found in the offspring, which also had only one allele. Similar cases were revealed for 'Mio' and 'Sylvia' at locus COL. Although presence of null alleles can be suspected, segregating populations should be analysed before drawing any definite conclusions.
Comparison of SSR data for some of the presently analysed cultivars with data from previously published studies revealed a few discrepancies in allelic composition, and several differences in allele sizes (Table 4). Since these studies have been based on different trees, minor somatic mutations cannot be ruled out, but experimental artefacts are also quite likely, especially concerning allele sizes.

SSR-based cluster analysis
A dendrogram based on Rogers' genetic distances did not produce any major clustering (Fig. 2). Clear grouping was also missing in the plot produced by a principal coordinate analysis (PCO) (data not shown). The majority of bifurcations in the dendrogram occurred at genetic distances above 0.30, indicating that the indigenous Swedish mandate cultivars are generally not closely related to one another. A similar lack of well-defined clusters was observed in a previous study based on RAPD (GARKAVA-GUSTAVSSON and NYBOM 2007).
Two pairs of cultivars, which could not be distinguished in the previous RAPD-based study, namely 'Spässerud' and 'Särsö ', and 'Å kerö' and 'Å kerö, Gripsholm', remained identical in the present study as well. However, 'Grågylling' and its red-coloured mutant 'Fagerö ' were identical according to their RAPD band patterns but differed in one SSR allele; 'Grågylling' had allele composition 220:241 in locus COL, whereas 'Fagerö ' did not have allele 241 (the most common allele in this locus) but instead the unique allele 228. The PCR reactions were repeated four times using two different 96-well plates to confirm this fact.
Cultivars 'Hanaskog' and 'Oranie' clustered at a distance of 0.17 in the dendrogram. Thus a close relationship between these two cultivars can be suspected. We compared the SSR allelic profiles of 'Hanaskog' and 'Oranie' and found that these cultivars shared 2 alleles in 7 loci, and 1 allele in 4 loci. Thus, 'Hanaskog' may well be an offspring of the presumably much older 'Oranie'.
We found a highly significant negative correlation between the two SSR-based matrices, calculated with Rogers' genetic distance and with Jaccard's similarity coefficient, respectively (r 0(0.863, P B0.001). We also found a significant negative correlation (r 0 (0.310, P B0.001) between Rogers' genetic distances, obtained in this study and Jaccard's coefficient of similarity, obtained in the RAPD-based study (GARKAVA-GUSTAVSSON and NYBOM 2007). Finally, we found a positive correlation between the two Jaccard's coefficient of similarity matrices, for RAPD and for phenotypically scored SSR fragments, respectively, r00.377, PB0.001.

S-allele configuration
A total of 14 different S-alleles were analysed in the mandate cultivars and non-mandate parents ( Table 1). The S-alleles found in 11 reference cultivars were identical with those previously reported (BROOTHAERTS et al. 2004;MATSUMOTO et al. 2007) except that we did not screen for S 13 (present in 'Gravensteiner') or S 21 (present in 'Ribston'). One of the mandate cultivars ('Katja') and three of the non-mandate parents ('Ingrid Marie', 'Cortland' and 'Worcester Pearmain' had also been investigated before (BROOTHAERTS et al. 2004;MATSUMOTO et al. 2007) and found to have the same S-allele configurations as in the present study.
In general, two (or fewer) S-alleles were found in diploids and three (or fewer) in triploids, with one exception; the putatively diploid 'Ö kna Vita Vintergylling' had three alleles possibly resulting from the ancient amphidiploidy of domestic apple.
Complete S-allele constitution was determined for 42 of the 67 analysed mandate cultivars. Another 23 cultivars lacked one allele, and 2 cultivars (both triploid), lacked two alleles. In total, this means that there were 19% missing alleles provided that all the cultivars are heterozygous in the S-locus.

Identification of SSR alleles
Proper choice of primer pairs is a very important component in SSR-based analyses, especially when results from several laboratories are pooled together in a larger database. For Theobroma cacao, 15 SSR primers were chosen as an international molecular standard because they had the highest reproducibility and consistency within a common genotype, while still allowing good differentiation of separate genotypes (SAUNDERS et al. 2004). When 46 grape cultivars were analysed by ten partners at six SSR loci without any efforts to standardize equipment or protocols, results obtained by some partners were very similar but different allele sizes were obtained in other cases . A strategy for data comparison by means of reference to the alleles detected in wellknown cultivars was therefore proposed. In apple, different sets of primers have been used, mainly from the list of LIEBHARD et al. (2002). The primers used in this study have previously been applied by GIANFRANCESCHI et al. (1998), LIEBHARD et al. (2002, GUARINO et al. (2006) and RAMOS-CABRER et al. (2007). Silfverberg et al. (2006) proposed to include two or three reference cultivars, namely 'Fiesta', 'Discovery' and 'Prima', which have been tested with most apple SSR markers and have been involved in many genetic studies, being parental cultivars of various mapping populations in Europe. We included 'Discovery' and 'Prima' and several other reference cultivars in the present study (Table 1), and could therefore compare our results directly with previously published data concerning genetic diversity estimators, allele sizes and allele compositions for some internationally well-known cultivars. Allelic composition was usually similar, but allele size differences of 1 to 4 bp were found in all but 3 locus/study combinations (Table 4). Moreover, pairwise comparisons of data from different studies revealed a serious lack of consistency between loci and sometimes even within the same locus. The differences are probably due mainly to the use of different internal size standards and to differences in binning procedures.
Since we can expect true alleles to differ by only 1 bp due to mutations in repeat flanking regions (GUARINO et al. 2006), questionable allele sizing must be avoided. Most apple SSR primers contain dinucleotide repeats, which are characterized by intensive stuttering, sometimes leading to misscoring of homozygous versus heterozygous alleles. By contrast, for SSR loci containing trinucleotide repeats, is not always possible to distinquish SSR amplicons from other PCR products due to lack of stutter bands. In such cases, all amplicons are usually reported, which may lead to an overestimation of the level of polymorphism of these SSRs (SILFVERBERG-DILWORTH et al. 2006).
Several studies have advocated the merits of registering relative instead of absolute size of alleles since many factors can influence the migration of PCR fragments through the gel (GALLI et al. 2005;AMOS et al. 2007). Two methods of standardisation have been applied to SSR fingerprinting in Theobroma cacao: 1) the use of a partial allelic ladder through the production of cloned and sequenced allelic standards and 2) the use of standard genotypes selected to display a diverse allelic range (CRYER et al. 2006). The use of cloned sequence-specific internal DNA standards to calibrate the size of microsatellite alleles could eliminate the problem with allele sizing in plants in international projects and is already widely used in human clinic and forensic studies (BRULAND et al. 1999;LECLAIR et al. 2004).

Identification of S-alleles
Fourteen S-alleles were identified in our study, with allele frequencies ranging from 1 to 18%. Possibly, there are many, albeit rather rare, S-alleles in apple that have not yet been detected. BROOTHAERTS et al. (2004) report allelic configurations that include 3% unidentified alleles and MATSUMOTO et al. (2007) have 4% unidentified alleles. In our material, number of unidentified alleles is much higher, 19%. One reason is probably that two very popular cultivars in Scandinavia, 'Ingrid Marie' and 'James Grieve', each have one allele for which there is no DNA marker yet (BROOTHAERTS et al. 2004). In addition, S 13 and S 21 which are present in another two important cultivars in this geographic region, 'Gravensteiner' and 'Ribston', respectively, were not investigated in our analyses.
In comparison with previously published data for DNA-derived identification of S-allele composition in apple (BROOTHAERTS et al. 2004;MELOUNOVA et al. 2005;MATSUMOTO et al. 2007) some geographically based discrepancies in allele frequencies can be noted. In our study, S 5 and S 7 are more common and S 9 more rare than in the previous studies. One major reason for this difference is probably the occurrence of S 5 and S 7 in several cultivars that are well-known in the Scandinavian countries (e.g. 'Aroma', 'Astrakan red', 'Ingrid Marie', 'James Grieve' and 'Katja') and are likely to have some relatives in the analysed material. By contrast, S 9 occurs in e.g. 'Cox's Orange Pippin' and 'Red Delicious' which have had a large influence on the development of cultivars in North America, central and south Europe and Japan, but probably have much fewer relatives in the Scandinavian countries with their harsher climate.

Genetic diversity among cultivars
A set of apple mandate cultivars had been appointed and approved for preservation already before the initiation of this study. We decided to analyse all those mandate cultivars that had a Swedish origin and were present in the Balsgård apple collections, including also those that were suspected to be sports only. The SSR-based values of expected heterozygosity (H e ), obtained in this study, are generally in good agreement with those published by LIEBHARD et al. (2002), GUARINO et al. (2006) andGALLI et al. (2005). When the less polymorphic loci in multilocus SSR markers were omitted (as in GUARINO et al. 2006), the average H e values were almost identical in all three studies based on the same markers: 0.80 (in this study), 0.81 in GUARINO et al. (2006) and 0.80 in LIEBHARD et al. (2002), in spite of the different numbers of cultivars analysed. The average PIC value, which is basically the same as expected heterozygosity, was 0.72 in GALLI et al. (2005), and 0.74 in the present study when all loci (also the less polymorphic loci in the bilocus systems) were included.
Number of SSR alleles per locus in our study was generally in agreement with those reported by GUARINO et al. (2006) but in some cases we found considerably more allele variants. Thus we found 15 alleles for primer CH02c06, while GUARINO et al. (2006) report only 10 alleles, and for primer CH04e05 we found 14 alleles, as compared to 8 in the GUARINO et al. (2006) study. These primers amplified mainly unique or rare (frequency at or below 0.05) alleles, namely 10 out of 15 (67%) in the case of CH02c06 and 10 out of 14 (71%) in the case of CH04e05. Consequently, the higher number of analysed individual genotypes, 66 in this study compared to 27 in the study by GUARINO et al. (2006), may explain the differences in allele number.
The expected heterozygosity (H e ) is influenced by the number of alleles and by the distribution of allele frequencies. Thus, the most polymorphic SSR locus in this study, CH02b10, (H e 00.88) revealed 15 rather evenly distributed alleles. Interestingly, locus CH02c09 with only 10 evenly distributed alleles had the same heterozygosity as locus CH02c06 (H e 00.85), which amplified 15 alleles, 10 of which were found at a frequency below 0.05. Even though loci CH02d08 and CH04e05 amplified 13 and 14 alleles, respectively, they had one very common allele each (frequency approx. 0.50) and, consequently, lower expected heterozygosity values, 0.70 and 0.71 respectively. The pattern of allele distribution seems to have a stronger impact on expected heterozygosity values than the number of alleles does. GALLI et al. (2005) pointed out the importance of an even distribution of alleles, and our study confirms this fact.
The large number of relatively abundant alleles in the S-locus makes this an important tool in identification of cultivars. In addition, valuable information about compatibility relationships is obtained. S-alleles have also been applied for paternity assignment in studies on fruit tree pollination and fruit set (SCHNEIDER et al. 2001a(SCHNEIDER et al. , 2001b(SCHNEIDER et al. , 2005ZISOVICH et al. 2005). Still, the discriminatory power of S-allele analysis is much smaller than for e.g. SSR analysis (where more loci are available) and this analysis can therefore be regarded only as a complement to other methods.

Genetic relatedness among cultivars
For analysis of closer relationships, both SSR alleles and S-locus alleles are very useful. Sports have almost identical allele composition and can easily be distinguished from closely related (parentÁoffspring) cultivars derived by sexual recombination. For assessment of phenetic similarities among cultivars, high multiplex ratio markers, like AFLP or ISSR would probably produce better results than SSR (GOULAO and OLIVEIRA 2001). In the case of highly heterozygous and diverse apple cultivars, it is, however, doubtful whether any other type of molecular marker would produce well-defined clusters.
The indigenous Swedish mandate cultivars probably represent a broad gene pool due to e.g. numerous cross-pollinations between highly heterozygous genotypes during the long period of apple cultivar development, frequent exchange of plant material between countries and plant breeding programs, and a slow turnover of popular cultivars, many of which are more than a hundred years old. Indigenous Swedish mandate cultivars may therefore originate from foreign progenitors, and represent several centuries of apple cultivation even in the most recent generations. Not surprisingly, the SSR-based UPGMA dendrogram failed to produce a clear group structure in our study, thus confirming the previous results obtained with RAPD markers (GARKAVA-GUSTAVSSON and NYBOM 2007).
Still, the discrepancies in S-allele frequency distribution when Swedish cultivars are compared to mainly North American, central and south European and Japanese cultivars (an over-representation of S 5 and S 7 in Sweden, and an under-representation of S 9 ) suggest that allele frequencies can change considerably due to genetic drift.
Other SSR-based studies in apple have also failed to produce a clear grouping structure except in a Spanish study of locally derived cider apple cultivars ). In the analysis of 66 apple rootstock clones, two broad groups but no distinct clusters were thus observed (ORAQUZIE et al. 2005). A genetic distance-based dendrogram, obtained in GUARINO et al. (2006) in a study of local Italian apple cultivars, showed no relationships with the morphological traits. Furthermore, eight reference cultivars from other countries were spread across the whole dendrogram. This distribution was explained by the fact that all the plant material probably came from the same genetic pool of apple germplasm, collected, bred and exchanged across Europe and Central Asia over many centuries.