Edible banana clones are an important trade commodity in tropical and subtropical countries and a staple food crop in Eastern Africa. They have originated by natural intrasubspecific and interspecific hybridization, and polyploidization in some cases. Most edible bananas originated from crosses between two wild Musa species, M. acuminata (donor of A genome) and M. balbisiana (donor of B genome) (Simmonds and Shepherd, 1955), with the possible contribution of other Musa species (Martin et al., 2020a; Sardos et al., 2022). Both species have small genomes and contain the same number of chromosomes (2n = 22). Diversity studies showed higher variability in genome sizes as well as higher genetic variability within M. acuminata, which contains several subspecies, compared to M. balbisiana (Janssens et al., 2016; Sardos et al., 2016a, 2020; Christelová et al., 2017). Domestication of bananas began in the Holocene, around 7,000 years bp, in Southeast Asia, when human migration brought banana species and subspecies from different regions into close proximity and therefore enabled natural crosses (Perrier et al., 2011).

Hybridization and polyploidization events during banana evolution led to the formation of parthenocarpic, clonally propagated cultivars, which are characterized by a high level of heterozygosity. Seven different translocation groups among M. acuminata were described by Shepherd (1999), who studied chromosome pairing during meiosis. As was recently shown, some of these large chromosomal translocations are characteristic of individual wild diploid species or even subspecies of M. acuminata (Baurens et al., 2019; Martin et al., 2017; Šimoníková et al., 2019; Dupouy et al., 2019; Martin et al., 2020b, 2020; Liu et al., 2023). Chromosome structural heterozygosity is common and results in aberrant chromosome pairing during meiosis and reduced or even zero production of fertile gametes (Dodds and Simmonds, 1948; Fauré et al., 1993; Shepherd, 1999; Goigoux et al., 2013). The reduced fertility of edible banana cultivars, together with differences in chromosome structure among M. acuminata subspecies greatly hampers the breeding of improved banana cultivars (Batte et al., 2019; Baurens et al., 2019) with lower susceptibility to diseases and pests. Diseases in commercial plantations are currently being controlled by the frequent application of fungicides; however, this treatment has a negative effect on the environment and health of banana workers (De Bellaire et al., 2010; Cordoba and Jansen, 2014).

Nowadays, the most-grown groups of bananas are triploid cultivars, including dessert bananas such as Cavendish and Gros Michel (AAA genome), cooking East African Highland bananas (AAA genome), and plantains (AAB genome). The Cavendish subgroup of bananas itself is the world’s most exported fruit, with an annual global export quantity of about 20 million tons (FAO, 2023). Several recent studies focused on the characterization of parentage relationships of selected diploid and triploid edible banana clones, including the triploid sweet banana cultivar Cavendish, showing the contribution from Mchare bananas (Raboin et al., 2005; Perrier et al., 2009; Martin et al., 2023b). The Mchare cultivar subgroup forms a phenotypically distinct group of diploid (AA genome) cooking bananas, which are nowadays found only in East Africa and some East African Islands. The study of Martin et al. (2023b) showed that the Mchare bananas contributed the unreduced 2x gamete to the origin of Cavendish bananas and the closely related Gros Michel cultivar subgroup (Martin et al., 2023b). The identification of Mchare bananas as the direct parents of successful cultivars highlights the importance of their use in breeding programs and warrants a more detailed analysis of their genome structure in the landraces and their hybrid offspring.

Improved triploid banana cultivars can be obtained through the development of tetraploids (4x), followed by the production of secondary triploid hybrids (Bakry and Horry, 1992; Tomekpe et al., 2004; Ortiz, 2013; Ortiz and Swennen, 2014; Nyine et al., 2017). Improved diploid varieties are produced by crosses of the existing cultivars with improved diploids (2x × 2x crosses). Traditional breeding requires the production of 4x or improved 2x genotypes that produce seeds, followed by re-establishing seed-sterile end products. Unfortunately, this process is hampered by the almost complete sterility of edible cultivars and flower incompatibility when stigma development does not correlate with bract lifting (Amah et al., 2021) and fast loss of pollen viability (Bayo et al., 2024). In combination with limited information on genome structural heterozygosity, the breeding process leads to very low seed sets. For instance, four seeds per Matooke (genome AAA) and only 1.6 seeds per Mchare (genome AA) can be obtained on average from one bunch (Brown et al., 2017; Batte et al., 2019).

Large genome structural changes can be identified by new sequencing technologies that span large sequence regions, e.g., mate-pairs Illumina sequences or long-read sequences produced by PacBio or Oxford Nanopore Technologies (Baurens et al., 2019; Dumschott et al., 2020; Pucker et al., 2022). Molecular cytogenetic techniques represent another option for the identification of large genome rearrangements. Localization of whole chromosome painting probes onto mitotic chromosomes in situ provides a powerful tool to identify large chromosomal translocations and to perform comparative analysis of chromosome structure in plants (e.g., Braz et al., 2018; Hou et al., 2018; Yu et al., 2022). Oligo-painting fluorescence in situ hybridization (FISH) is based on the in silico identification of large sets of unique oligomers specific to genome/chromosomal regions of interest (reviewed in Jiang, 2019) that can serve as probes. In bananas, chromosome-arm-specific sets of 45-nt-long sequences have already been identified and used as probes for in situ localization (Šimoníková et al., 2019). It has been shown that banana-specific painting probes designed based on the reference genome sequence of M. acuminata ssp. malaccensis ‘DH Pahang’, can be used to study genome rearrangements in other banana species and subspecies, including those that play an important role in the evolution of most edible cultivars (Šimoníková et al., 2020). Their use permitted the identification of general chromosome structures in different M. acuminata (A genome) subspecies, M. balbisiana (B genome) and M. schizocarpa (S genome), and most importantly, in edible banana clones (Šimoníková et al., 2020). These results showed the presence of specific chromosome structures in different species and subspecies. Furthermore, the presence of translocation events detected only in one chromosome set in some wild diploid species (e.g., M. acuminata ssp. siamea ‘Pa Rayong’, and M. acuminata ssp. burmannica ‘Tavoy’) was observed and indicated their hybrid origin (Šimoníková et al., 2020).

Our present study is focused on the characterization of large chromosomal translocations in a set of wild M. acuminata subspecies and edible banana cultivars that originated from natural intrasubspecific crosses. One of the main aims was to provide insight into the chromosomal evolution of diploid banana species and their cultivars and to identify variability and mode of the genome rearrangements. In addition to natural cultivars, we analyzed the chromosome structure of F1 progeny that resulted from crosses between Mchare bananas (the ancestor of Cavendish and Gros Michel) and the wild M. acuminata ssp. burmannicoides ‘Calcutta 4’ genotype. As Mchare bananas were found to be of hybrid origin, we wanted to analyze if some of the chromosome structures were preferentially transmitted to the progeny. Comparative chromosome painting revealed large variations in the genome structure within M. acuminata cultivars and the presence of translocation events, which were not observed in wild species analyzed so far. Our findings support previous assumptions about a more complex mode of Musa acuminata evolution and also show that the origin of edible banana clones was most probably accompanied by repeated introgressions and backcrosses (De Langhe et al., 2010; Martin et al., 2023a).

To enlarge the knowledge of general chromosome structure in Musa acuminata and to shed light on the evolution of M. acuminata, we provided cytogenetic analysis in 25 diploid M. acuminata accessions, including wild species and natural cultivars. We also analyzed chromosome structure in F1 hybrids obtained from ‘Mchare’ × M. acuminata ssp. burmannicoides ‘Calcutta 4’ crosses, to shed light on the transfer of Mchare chromosomes differing in their structures in comparison to ‘Calcutta 4’—chromosomes 1, 3 and 8—and corresponding reshuffled chromosome structures: reciprocal translocations between long arms of chromosomes 1 and 4 (1L/4L, 4L/1L) and Robertsonian translocation between chromosome 3 and 8 (3S/8L and 8S/3L).

Complex phylogenetic relationships within Musa and edible banana cultivars shown by Neighbor-Net analysis correspond to previous studies (e.g., Perrier et al., 2011; Janssens et al., 2016; Christelová et al., 2017; Sardos et al., 2022; Martin et al., 2023a). Some of the analyzed accessions, mostly those with hybrid origins, which were further confirmed by oligo-painting, showed unexpected/conflicting positions in the Neighbor-Net graph. The high level of admixture and mosaic genome structure of cultivated banana clones originating from hybridizations between subspecies of M. acuminata was proposed recently by Martin et al. (2020a; 2023a, b). Reticulate evolution, caused by polyploidy and hybridization (i.e., allopolyploidy), can promote rapid diversification of numerous plant lineages (reviewed in Stull et al., 2023).

Ancient polyploidization and hybridization events are known to be the major drivers of plant diversification and speciation. These processes lead to the multiplication of chromosome sets in genomes or “genome upsizing” (Ibarra-Laclette et al., 2013; Wendel, 2000), followed by the postpolyploid diploidization processes, which are thought to be associated with extensive loss of DNA or “genome downsizing”, and structural chromosomal changes (Schubert and Vu, 2016; Mandáková et al., 2017; Wang et al., 2021; Farhat et al., 2023). Until now, this phenomenon has been studied in detail mainly in the Poaceae and Brassicaceae families (e.g., Salse, 2016; Mandáková et al., 2017; Mandáková and Lysak, 2018). Recent developments in oligo-painting FISH and long sequencing technologies that enable the production of chromosome-scale genome sequences even in nonmodel species permit the study of this phenomenon in other plant species.

In the current study, we analyzed and compared the chromosome structure in M. acuminata and their natural and artificial hybrids by chromosome-arm-specific oligo-painting FISH, which facilitates revealing large chromosome translocations (Šimoníková et al., 2019, 2020). Genetic linkage between different chromosome-specific markers suggested the presence of five different reciprocal chromosome translocations (between chromosomes: 1 and 4; 2 and 8; 1 and 9; 1 and 7; and 3 and 8) in M. acuminata subspecies and their hybrids (Martin et al., 2017; Dupouy et al., 2019; Martin et al., 2020b). These and additional translocations were detected by chromosome-arm-specific oligo-painting FISH (Šimoníková et al., 2019, 2020). Eight additional chromosome rearrangements in M. acuminata were revealed in our present work ( Supplementary Table S1 ).

Even though the chromosome painting of the additional three banksii genotypes did not reveal any presence of large chromosomal translocations, we detected a short segment of chromosome 9S inserted into the peri-centromeric region of chromosome 5 in the ‘Higa’ accession ( Supplementary Figure S2F ). Unfortunately, based on the oligo-painting method, we are not able to distinguish whether this short segment represents duplication or a translocation event.

Despite the very close evolutionary relationships of burmannica, burmannicoides, and siamea subspecies (e.g., Perrier et al., 2009; Christelová et al., 2017; Dupouy et al., 2019), their representatives contained a large number of variable chromosome structures in their genomes. All species contained already described large translocations between chromosomes 1 and 9, and 2 and 8, which seem to be specific for this evolutionary group (Šimoníková et al., 2020; Martin et al., 2020b). On the other hand, using chromosome painting, additional translocations in the genomes of burmannica–burmannicoides–siamea representatives were revealed. Šimoníková et al. (2020) described the presence of translocations between chromosomes 3 and 4 in ‘Pa Rayong’ (ssp. siamea), and reciprocal translocations between chromosomes 7 and 8 and 3 and 8 in the genome of ‘Tavoy’ (ssp. burmannica), all presented in the homozygous state. The presence of reciprocal translocation between chromosomes 7 and 8 in the homozygous state was suggested based on genetic linkage patterns in ‘Khae Phrae’ and confirmed cytogenetically by BAC-FISH in ‘Long Tavoy’ in the work of Martin et al. (2020b). The translocation breakpoints were located between 21.8 and 26.3 Mb of ‘Pahang’ reference chromosome 7 and between 22.6 and 32.1 Mb of ‘Pahang’ reference chromosome 8 (Martin et al., 2020b). In comparison, chromosome oligo-painting of ‘Khae Phrae’ performed in our current work revealed the presence of specific chromosome structures containing chromosomes 7, 8, and 10 ( Figure 2M ; Supplementary Figure 2D ). These reciprocally recombined chromosomes, which contained a short arm of chromosome 7, a short segment of 7L, the long arm of chromosome 8, a segment of chromosome 10L, the long arm of chromosome 7, a short segment of 7S, and the long arm of chromosome 8, were found in the heterozygous state ( Figures 2H–J ; Supplementary Figure 2D ). Unfortunately, the chromosome oligo-painting method did not enable to localize translocation breakpoints, so we were not able to find out if the short segments of chromosomes 7L or 7S in the chromosome structures mentioned above are consequences of translocation or duplication events. Additional work utilizing long-read sequencing technologies and further whole genome assembly have to be used to unambiguously answer these questions.

In our current study, the burmannica–burmannicoides–siamea phylogenetic clade contained also one accession (Pisang Karok391) originally described as representative of ssp. malaccensis, which thus seems to be mislabeled. On the other hand, the malaccensis group of subspecies with the inclusion of ‘Pisang lilin’ and accession described as Zebrina (M. acuminata ssp. zebrina) represents a sister cluster to burmannica–burmannicoides–siamea, indicating their common evolutionary history. As it was mentioned earlier, the polyploidization and hybridization events, which were accompanied by large chromosomal rearrangements (Schubert and Vu, 2016; Mandáková et al., 2017; Wang et al., 2021; Farhat et al., 2023), are known to be the major drivers of plant diversification and speciation. We can therefore speculate that the presence of specific translocations events in the burmannica–burmannicoides–siamea group of subspecies might have had a direct effect on their own diversification and separation from the malaccensis subspecies (Janssens et al., 2016). The close relationship between malaccensis accessions and the Pisang lilin cultivar is supported by the presence of reciprocal translocation between chromosomes 1 and 4 in their genomes.

Phylogenetic analysis showed that the only representative of ssp. microcarpa ‘Borneo’ is closely related to geographically close accession ‘Buitenzorg’ (described as zebrina ssp.) and one accession described as ‘Pisang Mas Ayer’ (Sucrier), a sister group of the other two accessions of the zebrina subspecies ( Figure 1 ; Supplementary Figure 1 ). Unfortunately, ‘Buitenzorg’ and ‘Pisang Mas Ayer’ were not available during our study for the oligo-painting FISH to analyze their chromosome structure. We can thus only speculate that these two accessions can be of complex/hybrid origin, explaining their unexpected/conflicting position in the Neighbor-Net graph. This speculation can also be supported by previous results of chromosome painting that did not reveal the presence of any specific translocation event in the genome of the ‘Borneo’ accession, which could indicate a close relationship to the zebrina subspecies or Sucrier cultivars that were analyzed so far (Šimoníková et al., 2020). High level of molecular heterogeneity in ssp. microcarpa was mentioned in the study of Perrier et al. (2009). Previous studies showed that ‘Borneo’ accession shared alleles with banksii and always clustered together with banksii accessions (Perrier et al., 2009; Martin et al., 2023a). Perrier et al. (2009) showed that other representatives of microcarpa were closer to zebrina. Unfortunately, in all the studies mentioned, only a small number (at the most two representatives) of microcarpa accessions were analyzed; thus, we cannot speculate about the speciation processes that led to the origin and diversification of this subspecies. Nevertheless, the high level of molecular heterogeneity mentioned by Perrier et al. (2009) can indicate that hybridization events could play a major role in the speciation of the microcarpa subspecies. To shed more light on the evolution and genome structure of ssp. microcarpa, a larger number of microcarpa accessions from different geographic areas have to be collected and analyzed.

The zebrina genotypes did not form a distinct phylogenetic clade, and most of them clustered in close proximity to Sucrier cultivars. The genomes of ssp. zebrina and Sucrier cultivars contain reciprocal Robertsonian translocation between chromosomes 3 and 8 in the homozygous and heterozygous state, respectively. The chromosome painting showed that the Sucrier cultivar also contained additional translocation events, all in the heterozygous state. Reciprocal translocation between chromosomes 1 and 4, specific to the malaccensis group, and translocation between chromosomes 1 and 7 ( Figures 2B, L ; Supplementary Figure S2C ) indicate the involvement of a geographically close subspecies zebrina and malaccensis in the origin of Sucrier cultivars. Studies by Martin et al. (2020a, 2023a) indicated that at least four different progenitors were involved in the origin of Sucrier cultivars—malaccensis, zebrina, banksii, and an unknown M_2 progenitor. The same results of chromosome painting were obtained for Zebrina ITC1139. Even though the genomes of Zebrina ITC1139 and Sucrier cultivar ‘Ma’i hapai’ shared the chromosome painting structures, an unexpected phylogenetic position of ‘Zebrina’ ITC1139 was found. ‘Zebrina’ ITC1139 clustered together with ‘Pisang lilin’. This result can indicate the hybrid origin of the ‘Zebrina’ ITC1139, whose genome can arise by several rounds of hybridization, similar to Sucrier, but contains a large proportion of malaccensis-like genome regions.

The genomic constitution of edible banana cultivars showed a high level of admixture (Martin et al., 2020a; Martin et al., 2023a). The Robertsonian translocation between chromosomes 3 and 8 was also observed in the genomes of two additional banana hybrid clones, ‘Himone’ and ‘Maleb’. Neighbor-Net inference showed that these clones, together with the other two cultivars, ‘Vudu Beo’ and ‘Marakudu’, are closely related to Mchare. Similarly to other studies (Christelová et al., 2017; Perrier et al., 2019; Martin et al., 2020a), Mchare representatives formed a distinct evolutionary clade, which indicates their exceptional position in banana evolution. Previous phylogenetic studies were not able to unambiguously identify the mode of evolution and origin of this important group of edible banana clones, which represents a unique genetic source within Musa. It was shown that two subspecies of M. acuminata, ssp. zebrina and banksii shared some alleles with the Mchare clones (Perrier et al., 2009; Hippolyte et al., 2012; Perrier et al., 2019). The involvement of zebrina in the origin of Mchare bananas is supported by oligo-painting FISH, which confirmed the presence of a Robertsonian translocation between chromosomes 3 and 8 in the genomes of all Mchare clones analyzed in the present study and previously by Šimoníková et al. (2020). On the other hand, no Mchare banana cultivar was found in the region, where banksii and zebrina subspecies occur and could have contributed to the origin of these cultivars (Perrier et al., 2009; Hippolyte et al., 2012).

Chromosome painting was used to analyze the genome structure of several F1 Mchare hybrids resulting from different crosses between Mchare and M. acuminata ssp. burmannicoides ‘Calcutta 4’. Regarding the fact that oligo-painting FISH can only reveal those chromosomes consisting of specific structures (the presence of translocations), it can only provide partial information on the whole genome composition of the F1 progenies. As expected for intersubspecific crosses, all progenies, except NM237–8, contained one chromosome set from both parents. This was confirmed by Illumina resequencing, which identified the contribution of parental subgenomes to the hybrid clones. The identification of parental-specific SNPs and their distribution along the chromosomes of reference genome sequence M. acuminata ‘DH Pahang’ revealed a balanced presence of both parental genomes in all F1 hybrids. The only exception was the clone NM237–8, which contained only Mchare-specific SNPs, suggesting that the clone might have originated from an unreduced Mchare gamete. Genome composition analysis using Illumina resequencing did not reveal the presence of any aneuploid genome regions in F1 hybrids.

This analysis was done with the reference genome sequence of M. acuminata ssp. malaccensis (version 4), which is distinct from Mchare and ‘Calcutta 4’. Therefore, we cannot exclude the presence of genome regions that do not contain a proportional representation of parental subgenomes. Future high-quality chromosome-scale assembly of Mchare and/or other parental genomes used in the crosses will provide detailed information on the genome composition of the hybrid progenies.

Illumina sequencing data are available on NCBI sequence Read Archive (SRA) under the BioProject ID: 1033816 (SRA experiments: SRX22339926 - SRX22339938). The SNP files which were used for in silico painting of F1 hybrid clones are available in the Dryad repository (https://doi.org/doi:10.5061/dryad.44j0zpcnq).

DB: Investigation, Methodology, Project administration, Validation, Writing – original draft, Writing – review & editing. JČ: Investigation, Validation, Visualization, Writing – review & editing. GM: Formal analysis, Investigation, Software, Writing – review & editing. AD: Investigation, Writing – review & editing. HM: Resources, Writing – review & editing. AB: Resources, Writing – review & editing. RS: Resources, Writing – review & editing. EH: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Visualization, Writing – original draft, Writing – review & editing.