Molecular and Cytogenetic Characterization of Wild Musa Species

The production of bananas is threatened by rapid spreading of various diseases and adverse environmental conditions. The preservation and characterization of banana diversity is essential for the purposes of crop improvement. The world's largest banana germplasm collection maintained at the Bioversity International Transit Centre (ITC) in Belgium is continuously expanded by new accessions of edible cultivars and wild species. Detailed morphological and molecular characterization of the accessions is necessary for efficient management of the collection and utilization of banana diversity. In this work, nuclear DNA content and genomic distribution of 45S and 5S rDNA were examined in 21 diploid accessions recently added to ITC collection, representing both sections of the genus Musa. 2C DNA content in the section Musa ranged from 1.217 to 1.315 pg. Species belonging to section Callimusa had 2C DNA contents ranging from 1.390 to 1.772 pg. While the number of 45S rDNA loci was conserved in the section Musa, it was highly variable in Callimusa species. 5S rRNA gene clusters were found on two to eight chromosomes per diploid cell. The accessions were genotyped using a set of 19 microsatellite markers to establish their relationships with the remaining accessions held at ITC. Genetic diversity done by SSR genotyping platform was extended by phylogenetic analysis of ITS region. ITS sequence data supported the clustering obtained by SSR analysis for most of the accessions. High level of nucleotide diversity and presence of more than two types of ITS sequences in eight wild diploids pointed to their origin by hybridization of different genotypes. This study significantly expands the number of wild Musa species where nuclear genome size and genomic distribution of rDNA loci is known. SSR genotyping identified Musa species that are closely related to the previously characterized accessions and provided data to aid in their classification. Sequence analysis of ITS region provided further information about evolutionary relationships between individual accessions and suggested that some of analyzed accessions were interspecific hybrids and/or backcross progeny.

The production of bananas is threatened by rapid spreading of various diseases and adverse environmental conditions. The preservation and characterization of banana diversity is essential for the purposes of crop improvement. The world's largest banana germplasm collection maintained at the Bioversity International Transit Centre (ITC) in Belgium is continuously expanded by new accessions of edible cultivars and wild species. Detailed morphological and molecular characterization of the accessions is necessary for efficient management of the collection and utilization of banana diversity. In this work, nuclear DNA content and genomic distribution of 45S and 5S rDNA were examined in 21 diploid accessions recently added to ITC collection, representing both sections of the genus Musa. 2C DNA content in the section Musa ranged from 1.217 to 1.315 pg. Species belonging to section Callimusa had 2C DNA contents ranging from 1.390 to 1.772 pg. While the number of 45S rDNA loci was conserved in the section Musa, it was highly variable in Callimusa species. 5S rRNA gene clusters were found on two to eight chromosomes per diploid cell. The accessions were genotyped using a set of 19 microsatellite markers to establish their relationships with the remaining accessions held at ITC. Genetic diversity done by SSR genotyping platform was extended by phylogenetic analysis of ITS region. ITS sequence data supported the clustering obtained by SSR analysis for most of the accessions. High level of nucleotide diversity and presence of more than two types of ITS sequences in eight wild diploids pointed to their origin by hybridization of different genotypes. This study significantly expands the number of wild Musa species where nuclear genome size and genomic distribution of rDNA loci is known. SSR genotyping identified Musa species that are closely related to the previously characterized accessions and provided data to aid in their classification. Sequence analysis of ITS region provided further information about evolutionary relationships between individual accessions and suggested that some of analyzed accessions were interspecific hybrids and/or backcross progeny.

Introduction
Bananas and plantains (Musa spp.) are one of the most important food crops with the global annual production exceeding 130 Mt (faostat.fao.org). They are grown mainly by smallholder farmers for local consumption and only about 10% of the world's production is exported. Most of the cultivated bananas are parthenocarpic triploid clones (although diploids and tetraploids also occur), which originated from natural intra-and interspecific hybridizations between various subspecies of M. acuminata (A genome) and M. balbisiana (B genome) [1] within the genus Musa. The triploids are then classified as AAA, AAB and ABB and include a range of varieties of dessert bananas, cooking bananas and plantains, which provide essential nutrition for millions of people in the humid tropics. Minor group of cultivated bananas derived from the hybridization between the A, B and one of the other recognized Musa genomes, the S genome and T genome (M. schizocarpa and M. textilis, respectively) [2,3].
The production of bananas is threatened by various diseases, pests and adverse environmental conditions. This imposes the need for varietal testing and crop improvement supported by preservation of characterized banana diversity. The global Musa collection is maintained at the Bioversity International Transit Centre (ITC) hosted by the Katholieke Universiteit Leuven in Belgium [4]. The ITC collection was originally established for long term conservation of diploid, triploid and tetraploid edible cultivars, but eventually it started to be expanded by different wild species and subspecies. Currently, the collection contains more than 1,500 accessions in tissue culture. Newly introduced accessions of wild Musa species are a significant source of material to study diversity in Musa on a broader scale.
The difficulties with the classification of new cultivars and newly discovered wild Musa species using morphological descriptors call for characterization of Musa accessions using stable and reproducible characters. One of the basic characteristics of a species is its nuclear genome size. In earlier studies, the nuclear genome size was estimated mainly on edible bananas and their wild ancestors. A genome size of 600-650 Mbp was determined in M. acuminata and 550 Mbp in M. balbisiana, clearly discriminating both species [14][15][16]. Bartoš et al. [9] extended the knowledge of nuclear DNA content by representatives of former sections Rhodochlamys, Australimusa and Callimusa. Flow cytometric analysis showed that 2C nuclear DNA contents of Eumusa and Rhodochlamys accessions were overlapping, ranging from 1.130 to 1.377 pg in Eumusa and from 1.191 to 1.299 pg in Rhodochlamys. Species belonging to section Australimusa had 2C nuclear DNA content between 1.435 and 1.547 pg. M. beccarii, the only studied representative of Callimusa section, had the highest 2C nuclear DNA content (1.561 pg). However, only nineteen accessions were included in that study and the knowledge about genome size remained limited, especially in the section Callimusa.
The number and morphology of chromosomes describe a karyotype, another important characteristic of a species. In case of the genus Musa, chromosome number determines sectional classification of individual species. More detailed cytogenetic studies are, however, complicated by the small size of chromosomes, their morphological similarity and lack of chromosome-specific landmarks. To date, identification of all chromosomes within the karyotype is not possible. The only sequences cytogenetically mapped to chromosomes of a wider range of Musa accessions are ribosomal RNA (rRNA) genes. While the number of 45S rDNA loci was found conserved within individual sections, the number of 5S rDNA loci ranged from 4 to 8 per mitotic metaphase plate and varied within sections and even within different accessions of a single species [9,[16][17][18]. Similarly to genome size, genomic distribution of rRNA genes was studied mainly in M. acuminata, M. balbisiana and some cultivated varieties.
Unlike the methods mentioned above, genotyping using DNA markers provides higher resolution and is suitable for analysis of genetic diversity and establishment of phylogenetic relationships. Different types of molecular markers, such as RAPDs [19], RFLPs [20,21], AFLPs [22,23] or DArT markers [24], were used in diversity studies and molecular characterization of Musa. In 2011, a standardized microsatellite-based genotyping platform for molecular characterization of Musa germplasm was established [13]. Microsatellites (SSRs) are popular molecular markers because of their high abundance in the genome, co-dominant inheritance, extent of allelic diversity, good reproducibility and the potential for high-throughput analysis and automatisation. They proved to be useful in molecular genotyping of many economically important crops [25][26][27]. The microsatellite genotyping platform of Christelová et al. [13] is based on 19 microsatellite loci [28][29][30], which are scored using fluorescently labelled primers and high-throughput, high-resolution capillary electrophoresis. The platform enables discrimination between individual species, subspecies and subgroups of Musa accessions and is suitable for characterization of unknown accessions. Based on the results of Christelová et al. [13], the Musa Genotyping Centre was established at the Institute of Experimental Botany in Olomouc (Czech Republic) [31] under the auspices of Bioversity International. The database of reference SSR profiles they generate is enlarged with every new accession passing through the pipeline and consequently improves the grouping and enhances the probability of identifying the closest relative or an exact match for unknown accessions. Increasing the number of analysed wild species would give more accuracy to the grouping of the genus Musa.
When SSR genotyping provides incongruent results, or in case there is a need to support putative hybrid character of particular Musa accessions, the ITS sequence analysis can be used. The analysis of internal transcribed spacer region (ITS1-5.8S-ITS2) of nuclear ribosomal DNA provides additional information on genetic diversity and evolutionary relationships among individual accessions. The ITS region has been one of the most widely used markers in phylogenetic studies of flowering plants [32] and has recently been successfully employed in phylogenetic reconstruction in the family Musaceae [10,11,33]. It has been shown that ITS locus can be used for verification of genome constitution of inter-and intraspecific banana hybrids [33].
The aim of this work was to characterize a set of 21 Musa accessions which were newly introduced into the ITC collection from Helsinki University Botanic garden with a goal to increase the knowledge about the genetic diversity of wild Musa species. Most of the Musa accessions have been analyzed for the first time. Nuclear genome size, chromosome number and genomic distribution of rDNA were determined in individual accessions to shed light on their genome structure and evolution. Molecular characterization was performed using SSR genotyping platform and was extended by the analysis of ITS sequence region to verify a putative hybrid origin of some accessions. The analysis was also used to support incongruous results of SSR genotyping and to strengthen the previous knowledge on phylogenetic relationships within the Musaceae family.

Plant Material and Genomic DNA Extraction
In vitro rooted plants of 21 Musa accessions (S1 Table; basic description of accessions available at [5] were obtained from the International Transit Centre (ITC, Katholieke Universiteit, Leuven, Belgium). The plants were transferred to soil and maintained in a greenhouse. Fresh unopened (cigar) leaves were harvested, their segments lyophilized and maintained at room temperature until use. Genomic DNA of all Musa species used for the SSR genotyping and ITS analysis was isolated from lyophilized leaves using the NucleoSpin PlantII kit (Macherey-Nagel GmbH & Co. KG, Düren, Germany) following the manufacturer's recommendations.

Estimation of Genome Size
Nuclear genome size was estimated according to Bartoš et al. [9]. Approximately 50 mg of a young Musa leaf and 10 mg of a soybean leaf (Glycine max L. cv. Polanka, 2C = 2.5 pg DNA) which served as internal standard [14] were used for sample preparation. Suspensions of cell nuclei were prepared by simultaneous chopping of leaf tissues of Musa and Glycine in a glass Petri dish containing 500 μl Otto I solution (0.1 M citric acid, 0.5% v/v Tween 20). Crude homogenate was filtered through a 50 μm nylon mesh. Nuclei were then pelleted (300 g, 5 min) and resuspended in 300 μl Otto I solution. After 1 hour incubation at room temperature, 900 μl Otto II solution (0.4 M Na 2 HPO 4 ) [34] supplemented with 50 μg/ml RNase, 50 μg/ml propidium iodide and 3 μl/ml 2-mercaptoethanol, were added. Samples were analyzed using Partec PAS flow cytometer (Partec GmbH, Münster, Germany) equipped with 488-nm argon laser. At least 5,000 nuclei were analyzed per sample. Three individuals were analyzed in each accession, and each individual was measured three times on three different days. Nuclear DNA content was then calculated from individual measurements following the formula: 2C nuclear DNA content ½pg ¼ 2:5 x G 1 peak mean of Musa=G 1 peak mean of Glycine Mean nuclear DNA content (2C) was then calculated for each plant. Genome size (1C value) was then determined considering 1 pg DNA equal to 0.978×10 9 bp [35].
Statistical analysis was performed using NCSS 97 statistical software (Statistical Solutions Ltd., Cork, Ireland). One-way ANOVA and a Bonferoni's (All-Pairwise) multiple comparison test were used for analysis of variation in nuclear DNA content. The significance level α = 0.01 was used. Spearman's correlation analysis was used to test the relationship between chromosome number and nuclear DNA content of studied accessions.

SSR Genotyping
The standardized platform for molecular characterization of Musa germplasm [12] was used to genotype all 21 Musa accessions. The system is based on 19 microsatellite loci that are scored after the PCR with fluorescently labeled primers and capillary electrophoretic separation with internal standard (GeneScan 500 LIZ size standard, Applied Biosystems). The PCR products were multiplexed prior to the separation and loaded onto the automatic 96-capillary ABI 3730xl DNA Analyzer. Electrophoretic separation and signal detection was carried out with default module settings. The resulting data were analyzed and called for alleles using GeneMarker v1.75 (Softgenetics), manually checked and implemented into the marker panels [12]. The SSR profiles of newly analyzed accessions were integrated to the binary table of SSR profiles obtained in the work of Christelová et al. [12] and analyzed together. The genetic similarity matrices based on Nei´s genetic distance coefficient [36] were calculated using PowerMarker v3.25 [37] and the unweighted pair-group method with arithmetic mean (UPGMA) [38] was used to assess the relationship between individual genotypes. The results of UPGMA cluster analysis were visualized in a form of a tree using FigTree v1.4.0 [39].
PCR products were purified using ExoSAP-IT (USB, Cleveland, OH, USA) according to the manufacturer's instructions, cloned into TOPO vector, and transformed into E. coli electrocompetent cells (Invitrogen Life Technologies, Carlsbad, USA). For each accession, at least 28 cloned PCR products were sequenced. Sequencing was carried out using the BigDye Terminator v3.1 Cycle Sequencing kit (Applied Biosystems, Foster City, USA) according to the manufacturer's instructions and run on ABI 3730xl DNA analyzer (Applied Biosystems, Foster City, USA). Nucleotide sequences were edited using Staden Package [41]. Sequence boundaries of the spacers were assessed and phylogenetic relationships analysis was performed according to Hřibová et al. [33]. Sequence diversity was identified using DnaSAM program [42] with 1000 simulations. SplitsTree4 v4.1.11 [43] was used to construct phylograms based on the Jukes-Cantor model. Non-parametric bootstrapping with 1000 pseudoreplicates was performed to assess the nodal support. Phylogenetic trees were drawn and edited using FigTree program [39].

Chromosome Preparations and Chromosome Counting
Mitotic metaphase spreads were prepared according to Doleželová et al. [17]. Actively growing root tips were pre-treated in 0.05% (w/v) 8-hydroxyquinoline for 3 hrs at room temperature and then fixed in 3:1 ethanol: acetic acid overnight. Fixed roots were washed in a solution of 75 mM KCl and 7.5 mM EDTA (pH 4) and meristem tips were digested in a mixture of 2% (w/v) pectinase and 2% (w/v) cellulase in 75 mM KCl and 7.5 mM EDTA (pH 4) for 90 min at 30°C. Protoplast suspension was then filtered through a 150 μm nylon mesh and pelleted. The pellet was resuspended in 75 mM KCl and 7.5 mM EDTA (pH 4) and incubated for 5 min at room temperature. After pelleting, the protoplasts were washed three times with 70% ethanol. Five μl of the suspension were dropped onto a slide and shortly before drying out, 5 μl of 3:1 fixative were added to the drop to induce protoplast bursting. Finally, the slide was rinsed in 100% ethanol and air-dried.
For chromosome counting, the preparations were stained with DAPI (Vectashield Mounting Medium with DAPI; Vector laboratories). Slides were examined with Olympus AX70 fluorescence microscope. Images were captured using a cooled high-resolution black and white camera and processed using MicroImage software (Olympus, Tokyo, Japan). In each plant, two slides were observed, each with at least five metaphase plates.

Fluorescence In Situ Hybridization (FISH)
Probes for 45S rDNA and 5S rDNA were prepared by labeling Radka1 DNA clone (45S rDNA) and Radka2 DNA clone (5S rDNA) [44] with digoxigenin-11-dUTP or biotin-16-dUTP (Roche Applied Science). Both probes were labeled by PCR using M13 forward and reverse primers (Invitrogen). Hybridization mixture consisting of 50% formamide, 10% dextran sulfate in 1×SSC and 1 μg/ml of each labeled probe was added onto slides and denatured at 80°C for 3 min. The hybridization was carried out at 37°C overnight. The sites of probe hybridization were detected using anti-digoxigenin-FITC (Roche Applied Science) and streptavidin-Cy3 (Vector Laboratories, Burlingame, USA), and the chromosomes were counterstained with DAPI. The slides were examined with Olympus AX70 fluorescence microscope and the images of DAPI, FITC and Cy-3 fluorescence were acquired separately with a cooled high-resolution black and white CCD camera. The camera was interfaced to a PC running the MicroImage software (Olympus, Tokyo, Japan). At least ten complete metaphases were examined for every accession.

Estimation of Genome Size
The amount of nuclear DNA was estimated after flow cytometric analysis of propidium iodidestained nuclei. All analyses resulted in histograms of relative DNA content with two dominant peaks corresponding to G 1 nuclei of Musa and Glycine, the latter serving as internal reference standard (Fig 1). The 2C nuclear DNA content determined based on the ratio of G 1 peaks positions ranged from 1.217 to 1.772 pg ( Table 1). The differences between the accessions from section Musa were statistically significant with 2C DNA content ranging between 1.217 and 1.315 pg. Much higher interspecific variation (27.5%) of 2C DNA content was observed within the section Callimusa (2C = 1.390-1.772 pg).
Bonferoni's (All-Pairwise) multiple comparison test revealed twelve groups based on the differences in the nuclear DNA content ( Table 1) Spearman's correlation coefficient showed a strong negative correlation (r = -0.88) between 2n chromosome number and 2C nuclear DNA content among the Musa accessions. The same trend (r = -0.85) was observed when the results of Bartoš et al. [9] were included in the analysis (Fig 2). On the other hand, after including representatives of two other genera of the family Musaceae, Ensete gilletii [9] and Musella lasiocarpa (not published) the strong negative correlation between nuclear DNA content and chromosome number was no longer observed (r = -0.63).

Chromosome Counting and Cytogenetic Mapping of rRNA Genes
Protoplast dropping technique was used to prepare metaphase spreads for chromosome counting and localization of rRNA genes. All accessions were confirmed to be diploid and their chromosome numbers corresponded to section-specific classification ( Table 1) FISH with probes for 45S rDNA revealed distinct hybridization sites on one chromosome pair in the secondary constriction in all accessions of the section Musa (Table 1, Fig 3).  Table 1). Two of these signals were observed in the secondary constriction of one chromosome pair, while the other sites of probe hybridization were localized in interstitial chromosome regions.
A significantly higher variation was detected in the number of 5S rDNA loci (Table 1). In a majority of accessions from the Musa section, 5S rDNA sites were localized on two or three chromosome pairs. Three signals were observed on chromosomes of M. mannii (ITC.1574, Fig 3B) and seven signals per metaphase plate were detected in M. rosea x siamensis (ITC.1592, Fig 3C). Callimusa species contained 5S rDNA gene clusters on two to eight chromosomes per metaphase plate. In M. beccarii var. beccarii (ITC.1516), M. beccarii var. hottana (ITC.1529) and M. borneensis (ITC.1531), two of the 5S rDNA sites co-localized with interstitially localized 45S rDNA genes.

SSR Genotyping
Molecular characterization of the newly introduced ITC accessions based on SSR markers included inspection of their clustering pattern within the reference set of diploid entries used in the study of Christelová et al. [13]. The UPGMA cluster analysis based on the Nei [37] genetic distance resulted in relatively clear grouping of genotype groups and subgroups (Fig 4). Inclusion of the new accessions did not change the overall grouping of the species as compared to the results of Christelová et al. [13]. Most of the entries under this study clustered with the reference accessions as shown in our previous study [12] in accordance to their expected classification based on plant morphology. Newly introduced accessions belonging to the Musa section were grouped within the cluster A comprising other species from section Musa. The two accessions described as M. itinerans (ITC.1526 and ITC.1571) and M. rubinea (ITC.1518) formed a separate clade within the cluster A. Accessions described as Callimusa species were grouped within the cluster B together with reference accessions from this section. The allele sizes of analyzed accessions which were converted into binary code and added to the previous set of analyzed diploid Musa accessions [12] are available on the web pages of the Musa Genotyping Centre [31].  (Table 2, S2 Table). GC content of ITS1 varied from 57.41 to 65.02% and was slightly

Identification of rDNA Pseudogenes and Phylogenetic Analysis
The secondary structure of ITS2 and 5.8S rDNA sequence regions was reconstructed for all accessions under study. ITS2 sequences formed specific four-helices structure with typical pyrimidine-pyrimidine bulge in helix II and the most conserved primary sequence included the TGGT in the helix III. Secondary structure of 5.8S rDNA sequence was reconstructed under specific settings for base pairing as described by Hřibová et al. [33]. Moreover, 5.8S rDNA sequences were checked for the presence of three conserved motives [45]. The nucleotide changes in conserved motives of 5.8S rDNA of analyzed accessions, the information on GC content, presence of conserved motives in the 5.8S rDNA sequence and ability of ITS2 and 5.8S rDNA sequence to fold into a conserved secondary structure enabled identification of putative pseudogenes ( Table 2). As shown in Table 2, more than half of the analyzed accessions contained at least two types of ITS sequences. On the contrary, M. rosea x siamensis ITC.1598 which has been described as hybrid clone had only one type of ITS sequence region.
With the aim to support clustering based on SSR markers, phylogenetic analysis of the ITS sequence region was done. Dataset for this analysis comprised ITS sequences of the Musa spp. previously described by Hřibová et al. [33] and ITS types of the 21 studied accessions, excluding the putative pseudogenic ITS types (S1 Fig). Inclusion of the new accessions did not change the overall grouping of the species as compared to the results of Hřibová et al. [33]. Phylogenetic analysis of ITS region supported the clustering obtained by SSR analysis for majority of the studied accessions. Close relationships of M. itinerans and M. rubinea accessions was supported by SSR as well as ITS analysis. On the other hand, ITS analysis indicated close relationships of these accessions to M. balbisiana, but this observation was not supported by the SSR analysis. Similarly, close relationships of M. yunnanensis (ITC.1573) and M. rosea hybrid (ITC.1598) to M. ornata and M. velutina obtained after SSR analysis was not supported by ITS analysis. Finally, different clustering was observed for accession Musa x fennicae (Musa siamensis (male) x Musa rosea (female) (ITC.1522) which is found in the cluster containing ornamental bananas (former Rhodochlamys section) in the SSR cladogram while using ITS analysis, all ITS sequence types including putative pseudogenes are clustered together with AA genotypes.
In our previous study [33], we have shown that, with certain caution, pseudogenic rDNA sequences can be used to identify parental genotypes of hybrid clones. Considering the fact that more than a half of the analyzed accessions contained at least two types of ITS sequences, phylogenetic reconstruction with ITS sequences including putative pseudogenes was done (S2 Fig). As it is visible from the ITS clustering, the genome of Musa yunnanensis (ITC.1573) contains divergent ITS sequence types which clustered together with ornamental bananas as well as with AA genotypes. Similar observation was obtained also for genotype described as hybrid clones of ornamental banana genotypes-M. rosea x siamensis (ITC.1592), while hybrid clone Musa x fennicae (Musa siamensis (male) x Musa rosea (female) (ITC.1522) contained only ITS sequence types which clustered together with the AA genotypes (see above).

Nuclear DNA Content and Distribution of rDNA
Most of the knowledge on nuclear genome size and genomic distribution of rDNA loci in banana comes from the analysis of triploid cultivars and their wild ancestors [14-16, 17, 18, 46]. Only the study of Bartoš et al. [9] included a wider range of Musa species and provided the first complex picture of the whole genus. Our study significantly expands the number of wild Musa species where these key characteristics of nuclear genome are described.
The diversity in genome size observed in species belonging to the section Musa is in line with previous studies [9,[14][15][16]46] which estimated 2C nuclear DNA content ranging from 1.108 pg in M. balbisiana to 1.377 pg in M. schizocarpa. The genome size of M. laterita and M. mannii were determined by Bartoš et al. [9]. However, in both cases, different ITC accessions (ITC.0627 and ITC.1411, respectively) were used. While the difference in 2C nuclear DNA content estimates between both accessions (ITC.1411 and ITC.1574) of M. mannii is negligible (1.0%), there is much higher difference (7.7%) between M. laterita accessions (ITC.0627 and ITC.1575). This observation suggests that accessions ITC.0627 and ITC.1575 could represent different varieties of M. laterita.
Genomic distribution of 45S and 5S rDNA clusters in Musa accessions as observed in this study is similar to the results obtained by Bartoš et al. [9] and together with overlapping genome size values brings further evidence about the close relationships between representatives of former sections Eumusa and Rhodochlamys. The traditional division into sections Eumusa and Rhodochlamys was mainly based on morphological characters. The separation of these sections was however doubtful, because morphological differences were difficult to detect  in some species and interspecific hybridization was frequent [6,47]. Our findings, as well as previous analyses based on different molecular markers [8,10,21,23], support the recent merger of sections Eumusa and Rhodochlamys into the section Musa. The presence of odd numbers of 5S rDNA loci has been previously observed in some edible cultivars and wild species [9,17] and could be explained by structural chromosome heterozygosity and/or as a result of hybridization between two genotypes bearing different 5S rDNA sites. However, it is also possible that one of 5S rDNA loci contained a lower number of repeat units and the resulting signal was below the detection limit of FISH.
The representatives of the section Callimusa showed high variation in nuclear genome size as well as in the number of rDNA loci and were clearly discriminated from the accessions of the section Musa. For M. borneensis (ITC.1531), the highest nuclear DNA content (2C = 1.772 pg) known for diploid Musa species was determined. The increased genome size was accompanied by a higher number of 45S and 5S rDNA loci in comparison with other Callimusa accessions. Until now, the largest genome size was detected in M. beccarii [9,48], which has also been characterized by more abundant distribution of rRNA genes. The present estimates of genome sizes as well as chromosome number and distribution of rDNA loci in both M. beccarii varieties are in line with previous investigations [9,48].
An interesting outcome of this study is the observation of a negative correlation between the basic chromosome number (x) and nuclear genome size (1C), which became evident in this study after a wider range of Musa species was analyzed. This trend could be explained by the evolution from a common ancestor by chromosome fission accompanied by DNA loss. This hypothesis is in line with Simmonds [49], who speculated that the lower chromosome number  [50] suggested that Ensete glaucum and M. beccarii are ancestral forms of Ensete and Musa with a common ancestor that is yet to be established and the ancestral chromosomal number of both genera is x = 9. On the other hand, if this hypothesis was true, representatives of the genera Ensete and Musella would be expected to have a genome size comparable to that of and other Callimusa species. A more extensive study of the karyotype evolution in the family Musaceae will be necessary to explain this "the less chromosomes, the larger genome" trend emerging in the genus Musa, but not observed in other genera of the family.

SSR and ITS Analysis
SSR analysis provided important information about the genomes of 21 accessions of wild Musa species, which were genotyped using molecular markers for the first time. The grouping of the accessions within the diploid reference set was revealed by the UPGMA cluster analysis. In most cases, the results of this grouping (Fig 3) were consistent with the characterization based on traditional morpho-taxonomic classification. Apart from better characterization of the wild Musa species, the SSR profiles obtained in this work expand the database of reference SSR profiles and thus improves the grouping and enhances the probability of identifying the closest relative or an exact match for unknown accessions. Short length, utilization of highly conserved primers and relatively fast evolution of internal transcribed spacers ITS1 and ITS2 as compared to rRNA genes, made the ITS region one of the most popular markers in phylogenetic studies. The first detailed information about the structure and diversity of ITS region in the genus Musa was provided by Hřibová et al. [33], while the work of Christelová et al. [12] showed the usefulness of the ITS for further analysis of accessions which give incongruous results with SSRs. High level of nucleotide diversity and presence of more than two types of ITS1-5. The close relationship of M. itinerans to B genome representatives observed after ITS analysis is in agreement with the study of Liu et al. [11]. This relationship was not supported by the SSR analysis. It is, however, possible that the grouping of some species will change after more representatives from both sections pass through the SSR genotyping pipeline. In a similar way, we have observed discrepancies in the position of M. yunnanensis (ITC.1573) and M. rosea hybrid (ITC.1598), as well as Musa x fennicae (Musa siamensis (male) x Musa rosea (female) (ITC.1522) in the tree constructed from SSR and ITS data (see above ; Fig 4, and S1 and S2 Figs). The ITS analysis indicates that the M. yunnanensis accessions included in our study originated from hybridization of different genotypes (subspecies?) ( Table 2), while the presence of only one ITS type sequence in the accession labeled as M. rosea hybrid (ITC.1598) does not support its hybrid character. In Musa x fennicae (Musa siamensis (male) x Musa rosea (female) (ITC.1522), divergent types of ITS sequences were identified but all of them were related to AA genotypes, while SSR analysis showed close relationships of this clone with ornamental bananas (genotypes of former Rhodochlamys section). These observed discrepancies could be a consequence of dominance of one type of rDNA sequence in hybrids, which may lead to complete homogenization of rDNA locus originating from the second parent [51][52][53].
Our results confirm the usefulness of SSR markers for molecular characterization of unknown accessions of Musa [12]. As the SSR genotyping is based on scoring alleles, it may not be appropriate for diversity and phylogenetic inference estimation in case of inter-specific hybrids and their backcross progenies [33]. Thus, we performed a detailed analysis of the nucleotide composition and structure of the ITS region and showed that some of the wild diploids contained polymorphic ITS regions. Moreover, in silico analysis of the ITS sequences indicated the presence of putative pseudogenic ITS types. These observations indicate that some of the analyzed accessions of wild Musa species originated from hybridization between different genotypes within a species or between putative subspecies [33,54].