Collecting missions

The collecting missions were conducted in two phases. The first phase, in 2018, took place in North Vietnam and was the most significant in terms of collected material. It spanned over 400 km from Ha Long Bay in the East to the borders with Laos and China in the Northwest (Fig 1). The second phase, in 2019, occurred in Laos, extending from Vientiane up north to the borders of Vietnam and Myanmar, and then into China for exploration in Yunnan Province (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Geographical distribution of the 220 studied specimens.

Colours by species, according to the legend. Shifting of overlapping points using QGIS internal displacement option. Terrain data sourced from OpenStreetMap. Map projection: WGS 84.

https://doi.org/10.1371/journal.pone.0307592.g001

The collecting missions were carried out by travelling along roads and collecting all Musa material of interest as identified by the team or indicated by local farmers. This material was found in forest, along roadsides, in farmers’ fields or in gardens near houses. For each site, the team proposed a taxonomic classification, recorded the geolocation and administrative reference, and noted the morphological characteristics and context of discovery. Additionally, the team gathered available elements of local knowledge, such as vernacular names, uses, and origins, and took photographs of the environment, the whole plant and its characteristic details. A portion of the leaf blade (about 10 cm²) was placed in a paper bag, dried in silica gel for 48 hours, and then kept in dry conditions until it was stored at -80°C for genomic analysis. When available and with the agreement of the farmers, a plant sucker was also collected. These suckers were replanted for preservation in existing ex situ collections in Vietnam (Nomafsi Institute) and China (Kunming Institute of Botany), and a new field collection was created for this purpose at the Lao Academy of Science and Technology in Vientiane.

Although the three countries surveyed have ratified the Nagoya Protocol, the rules on access to genetic resources and associated traditional knowledge are specific to each country. The necessary procedures were completed in advance in each case, authorizing the collection of dried leaf samples, their transfer to France for genomic analysis, and their study by the various project partners.

For the collecting mission in Vietnam, an application for access to genetic resources was signed in 2018 between one partner from Vietnam (the Institute of Tropical Biology–ITB), the French Agricultural Research Centre for International Development (CIRAD), and the National Museum of Natural History of Paris (MNHN). This document was signed by the three directors of these scientific institutions. ITB was responsible for establishing contracts with the parks and reserves where applicants collected banana samples. CIRAD was responsible for the molecular characterisation of the banana samples, and MNHN was responsible for the long-term storage of the banana samples. Additionally, a collecting permit (No. 1961/TCLN-DDPH) was issued by the Vietnamese Ministry of Agriculture and Rural Development on 9 November 2018.

For the collecting mission in Laos, an application for access to genetic resources was signed in 2018 between the Institute of Tropical Biology (ITB) for Vietnam, the French Agricultural Research Centre for International Development (CIRAD), the National Museum of Natural History of Paris (MNHN), and the Cabinet of the Lao Academy of Science and Technology (CLAST). CLAST is a public establishment responsible for national scientific research projects and the promotion of cultural and scientific activities. Therefore, it was responsible for establishing contracts with the parks and reserves where applicants collected banana samples. The application for access to genetic resources was signed by the four directors of these scientific institutions. Additionally, a Material Transfer Agreement was also signed between the directors of CLAST and CIRAD in March 2019.

For the collecting mission in China, a Material Transfer Agreement was signed in 2019 between the directors of the Kunming Institute of Botany (KIB), CIRAD, and MNHN. Finally, a consortium agreement for the implementation of each of the two research projects (the Sud Expert Plantes Développement Durable programme through the DivBa SEP2D project (AAP3-97) and the BforBB open science project supported by the Agropolis Fondation under the reference ID 1605–011 through the “Investissements d’avenir” programme (Labex Agro:ANR-10-LABX-0001-01)) that financed the research described in our article was signed by ITB, MNHN, CIRAD, and the Northern Mountainous Agriculture and Forestry Science Institute (NOMAFSI) from Vietnam.

Data organization

A significant amount of data was generated from field collections, genotyping processes and bioinformatics and statistical analysis procedures, as described below. A specific interactive mapping tool was developed in R-Shiny to represent the geographically structured data. This tool allows for interactive multi-criteria selection of the samples to be displayed and their graphical illustration using sets of colours to depict the modalities of several variables of interest (geographical information, taxonomic group, and genomic group). The geographical representations shown in this article were refined using QGIS 3.28.0 with the Outdoor layer provided by MapTiler & OpenStreetMap. The QGIS internal displacement option was used to display overlapping points.

Genotyping by sequencing

The genomic characterisation of the samples was performed using a high-throughput genotyping method, specifically resequencing targeted portions of the genome through the GBS (Genotyping By Sequencing) technique [51].

DNA from each sample was extracted following a modified CTAB protocol [52]. Genotyping was conducted according to published procedures [51]. The GBS libraries based on StdGBS methods were constructed following standard protocols with the PstI and MseI restriction enzymes. The 300–500 bp short-insert libraries were sequenced in 150 bp paired-end reads using Illumina HiSeq2500 (Illumina, San Diego, CA, USA) by Genewiz, Azenta Life Sciences, USA.

The first GBS analysis (batch 1) was performed on the 182 samples collected in Vietnam in 2018. A set of 34 references representative of the genetic diversity in the genus Musa was added to insert the observed diversity within the global diversity. A second GBS analysis (batch 2) was carried out on the 153 samples collected during the 2019 survey in Laos and China. This analysis included 32 references (27 of which were identical to batch 1) and seven samples from batch 1, whose DNA quality had been insufficient in the previous analysis. In each batch, all samples were analysed twice to increase the sequencing depth, except for some genetically more distant samples (Ensete, Callimusa and Musa from the former Rhodochlamys section). Five samples were repeated on each plate to serve as internal repeatability controls.

Bioinformatics procedures

After demultiplexing with GBSX [53], each of the ’paired end’ sequencing files (fastq_R1 and fastq_R2) was checked using FASTQC software [54] available at https://www.bioinformatics.babraham.ac.uk/projects/fastqc, for quality metrics, presence of adapters, and GC content. The files were then cleaned to remove Illumina adapter sequences and low-quality ends (Phred score > 20) using Cutadapt [55]. Reads shorter than 30 bp after trimming were discarded. The cleaned reads were aligned against the v2 genome of Pahang-HD, a doubled haploid from M. acuminata malaccensis ‘Pahang’ [9] obtained from the Banana Genome Hub [13, 14] using BWA-MEM [56]. Re-alignment was performed using the IndelRealigner module from GATK v4.1 [57]. The recommended GATK pipeline for non-model organisms was followed, including a recalibration step. This involved an initial round of SNP calling on the original uncalibrated data, selecting the SNPs with the highest confidence, and then executing a round of base recalibration on the original mapped read files. For duplicate specimens, a script using Sambamba software was employed to merge the recalibrated bam alignment files. SNP and indel calling were performed using the GATK module HaplotypeCaller v4.1. Subsequently, a script (gVCF2vcf -gz.pl) was written to combine the individual gVCF files into a single VCF file. The GenomicDB procedure from GATK was used to build the gVCF SNP database and the GenotypeGVCF procedure was then used to extract only the variant positions. The snpcluster exclusion procedure was applied to process SNP clusters, set with a threshold of three or more SNPs per 10 bp window. These procedures have been compiled in the pipeline Musa_NGS_Suite available at https://github.com/CathyBreton/Genomic_Evolution [58].

Data analysis

The Musa_NGS_Suite was utilised to process the Vietnam and Laos/China specimen datasets separately, as they were obtained from two distinct analytical batches, producing two raw VCF files.

Using the GBS-Div software (for a detailed description, see S1 File), a filtered VCF file was extracted for each batch, retaining only bi-allelic sites with depths between 20 and 300. For any given site, only alleles representing at least 5% of the total depth at the site were kept. The two files were subsequently merged into a final VCF file keeping only the SNPs found to be polymorphic in both batches simultaneously.

Data analysis of the merged VCF file was also performed using GBS-Div. Statistical indicators for missing data, allele frequencies, depth distributions and ploidy ratios were applied to clean and validate the data. The ploidy ratios were also used to confirm the diploid status of all specimens. A dissimilarity matrix was calculated between pairs of specimens as the ratio of the number of identical alleles (0, 1 or 2) at a SNP to the total number of SNPs present for both specimens. A principal coordinate analysis (PCoA) was conducted on the dissimilarity matrix to provide an overall representation of the collected diversity. A Neighbour-Joining tree [59] was inferred to describe the relationships at the specimen level. This method was chosen because it is free of the assumptions usually required about the underlying evolutionary process, which are often unmet, or only partially met in plants that propagate clonally by budding and whose seeds and suckers may have been selectively disseminated by humans. An a posteriori least-squares re-estimation of edge length was then performed. The high number of markers limiting the efficiency of the bootstrap method, instead, other indicators were used to assess the reliability of the inferred tree. First, a Jackknife resampling procedure was applied to the specimens to detect influential units, measured by the impact of their removal on the tree’s topology. The tree model, which assumes no exchanges between different branches, was a priori considered valid since applied here at the species level. However, reticulations between tree nodes that significantly improved the fit to observed dissimilarities were sought as indicators of potential horizontal exchanges. All analyses were performed using DARwin software [60].

A standard genomic profile for the main pools (species or subspecies) was constructed as the majority consensus of the specimens considered reliable representatives of the pool.

Diagnostic SNPs that discriminate these pools were then determined as the loci that were homozygous for one allele in the pool consensus and homozygous for the alternate allele in all other pool consensuses. Finally, the genotypes at these diagnostic positions were used for chromosome painting of all specimens not involved in the construction of pool consensuses, to detect potential contributions of these major pools to those unassigned specimens. Both DARwin and GBS-Div software are freely available at: https://darwin.cirad.fr/.

Musa balbisiana

The species M. balbisiana, which is abundant throughout the surveyed area, was the most frequently sampled, with 106 specimens collected (S1 Fig). The highly distinctive morphological characteristics of this species made identification straightforward in the field. These features include: petiole canal varying from "margins curved inward" to "margins overlapping", black petiole margins, waxy bracts without internal discolouration, absence of bract rolling, bracts often persistent on the male rachis, and fertile fruits with seeds (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Morphologic diversity of Musa balbisiana.

(A-D) male bud. (E) the basis of the leaf. (F) petiole canal. (G) bracts. (H-I) male flowers. (J-K) fruits and seeds. (L-P) pseudostem. (Q-T) a complete bunch. Pictures by Matthieu Chabannes—CIRAD.

https://doi.org/10.1371/journal.pone.0307592.g003

Almost all these specific characteristics were observed in the collected plants, although some variations were noted such as differences in fruit size or the level of persistence of male flowers and bracts on the rachis. Additionally, pseudostem colours varied from light green, without macules, to black, and with varying abundant wax, a diversity already recognised among the reference specimens in field collections.

Of the 106 specimens collected, 68 were from Vietnam, with two-thirds found in lowland areas, nearly always below 300 m, around Hanoi. The other specimens were observed at elevations over 600 m in western Vietnam, near Laos and Yunnan. In Vietnam, M. balbisiana was rarely found in wild populations, but was more commonly cultivated in gardens for its pseudostem used as animal fodder, fruits, and seeds, which are medicinally used in alcoholic tinctures. These cultivated plants are clonally propagated, and although their fruits are seeded, the seeds often fail to fully develop and remain soft at maturity [62]. The 26 specimens from Laos and the 12 from Yunnan were collected at medium elevations, around 600–700 m. Wild populations were more abundant towards the West, but the species was also commonly found cultivated in gardens.

The GBS analysis was conducted on 101 collected plants and eight references (two of which are analysed twice). It revealed a quite structured diversity within M. balbisiana, which was unexpected for a species reputed to be very homogeneous (Fig 4A). Several clusters were identifiable, the most homogeneous was cluster 1a, which included 38 of the 101 specimens and several references: ‘Pisang Batu’, ‘Pisang Klutuk Wulung’, ‘Pisang Klutuk’, and ‘Klue Tani’. Cluster 1b, which included nine specimens, was related to cluster 1a. Cluster 2 had 38 specimens and showed a wider diversity than 1a (S2 Fig). Notably, none of the M. balbisiana references belonged to this cluster. Lastly, several specimens could not be assigned to the above-mentioned clusters (shown in black in the tree), with some, such as #304, being very distant. It is also worth noting that the references of Indian origin (‘Lal Velchi’, ‘Singapuri’), or unknown origin (‘Balbisiana HND’ and ‘Balbisiana CMR’) were not linked to any of the clusters described here, in intermediate position between clusters 1a/1b and cluster 2, indicating a broader genetic diversity in M. balbisiana species than what was collected in this study area. Therefore, these reference specimens that do not cluster, probably derive from missing wild populations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Hierarchical subtrees by species.

Subtrees are extracted from the whole tree (Fig 2C), split by species. (A) Musa balbisiana, (B) Musa itinerans, (C) Musa acuminata, (D) Musa yunnanensis.

https://doi.org/10.1371/journal.pone.0307592.g004

This genetic diversity was clearly structured geographically (S1 Fig). Clusters 1a and 1b were concentrated in the lowland areas of Vietnam, at a mean elevation of about 100 meters and a median elevation of less than 40 m. In contrast, specimens belonging to cluster 2 were found growing between 200 and 1000 m (with median and mean elevations of more than 600 m). They were mainly found in northern Laos, southern China and northwestern Vietnam, and more rarely farther north in Yunnan.

The heterozygosities of M. balbisiana specimens were among the highest when compared to the other collected species. Among Musa species, many are hermaphrodite with bisexual basal flowers followed by functionally male distal flowers over time. However, several species have basal flowers that are functionally female only, with reduced male organs. These species are therefore mainly cross-pollinated by time shifting. This is the case for M. balbisiana [63], which is consistent with the observed heterozygosities. However, no incompatibility factors have been found in section Musa, so even in the case of unisexual basal flowers, self-pollination is possible by geitonogamy between two inflorescences of the same clump.

Heterozygosities were higher in clusters 1a and 1b (4.0% on average) than in cluster 2 (2.3% on average) (Fig 5). Consequently, a “pole 1” with higher heterozygosities could be identified, including clusters 1a, 1b, and the closely related unassigned specimens, as well as most of the references. Conversely, “pole 2” with low heterozygosity, would include cluster 2, the unclassified samples from China and Laos as well as the two references of unknown origin (‘Balbisiana HND’ and ‘Balbisiana CMR’).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Histograms of heterozygosities by species.

Distributed over genomic clusters, unassigned specimens (other), and references (ref.) (Colours according to the legend).

https://doi.org/10.1371/journal.pone.0307592.g005

It should be noted that the distinction between the identified clusters, discriminated by only a small number of SNP markers, is not sufficient to propose a subspecies-type botanical division, as is usually agreed for M. acuminata. This is consistent with the scarce morphotaxonomic differences observed.

The natural range of M. balbisiana is extensive, covering the entire range of section Musa, from India to the Philippines and the islands of nearby Oceania [64]. However, the range of spontaneous forms is much narrower, covering mainland Southeast Asia, from Myanmar, parts of northeast India [65] and the Himalayan foothills, to the provinces of southern China [66]. In the Indian peninsula, the absence of macroglossine bats, the main pollinators of Musa, suggests that M. balbisiana is necessarily cultivated there [67]. Similarly, elsewhere, M. balbisiana would have been introduced by humans and the observed natural populations would be naturalised forms: Java [64], Sulawesi, Maluku, Lesser Sunda islands [68], Taiwan [69], the Philippines [66], and Papua New Guinea [40, 70]. Indeed, M. balbisiana is still widely cultivated in gardens, or on small plots, for uses other than fruit [71]. It is particularly resistant to several leaf spots, so its leaves are everywhere favoured and sold for wrapping food or used as plates [72]. The male bud is often used as a vegetable, the pseudostem fibres are used in some regions to make cloth, and various plant parts are known to have medicinal properties [73]. The fruits can be eaten before the seeds ripen and harden. Additionally, some varieties bear fewer seeds, remaining soft at maturity, and can therefore be eaten. In the Philippines, for instance, the variety Pacol displays such partial seed sterility and is cultivated by subsistence farmers [43]. These soft-seeded banana plants may have been selected, leading to a form of domestication where different varieties, vegetatively propagated, are recognized and named [72]. M. balbisiana is particularly robust and adaptable to conditions quite different from its area of origin. When moved by humans, it can spread over large areas, where crop escapes can then form spontaneous populations.

The study area covers part of the primary endemic area of M. balbisiana and of the areas of secondary spread. In Vietnam, it was found in compound gardens, in small cultivated plots and much more rarely in natural populations. As previously mentioned, the only wild populations were observed in the northwest of the country [74]. Lê et al. [62] identified two types of M. balbisiana populations in Vietnam: (1) wild, with many seeds, often locally named ‘Chuoi hot rung’; (2) low and soft-seeded, named ‘Chuoi Hot’ or ‘Chuoi Hot Qua Lep’.

We propose that genetic cluster 2 and the few other linked high-elevation specimens (collectively referred to as “pole 2” above) represent the native “wild” compartment of M. balbisiana. Clusters 1a and 1b, along with the related non-clustered linked specimens, (collectively referred to as “pole 1”) represent a compartment of M. balbisiana, considered here as "domesticated", in contrast to the previous “wild” compartment. This domesticated pole shows lower overall intra-diversity, which is expected for a plant, at least partially, vegetatively propagated by men. It also displays higher heterozygosity compared to the heterozygosity of pole 2 which is similar in magnitude to that of other wild species (Fig 5). This higher heterozygosity can be attributed to the recruitment of alleles from a broader area with significant geographical mixing, coupled with the fixation of rare alleles through vegetative propagation. Spatially, the wild pole 2 is mainly located in the more remote areas of Yunnan and Laos, whereas the domesticated pole 1 is present in the more open areas of Vietnam. Similarly, Mertens et al. [75] concluded from comparison based on SSR markers that populations sampled in central and northern Vietnam belong to the native range of M. balbisiana, while those in southern Vietnam appear to be imported and non-native. It is also worth noting that all the reference specimens, except for ‘Balbisiana CMR’ and ‘Balbisiana HND’, of unknown origin, fall within pole 1. When the origin of these references is known, they are from regions outside the native area of M. balbisiana and can be considered as references for the domesticated and ubiquitous pole 1. This is also the case for the different M. balbisiana accessions conserved at the International Musa Germplasm Transit Centre (ITC) ex situ collection. According to Mertens et al. [76], using SSR marker data, except for one specimen (ITC1527, from Xishuangbanna in southern Yunnan), 27 of the 28 M. balbisiana accessions at the ITC do not group with native M. balbisiana populations, which questions their ‘wild’ status.

It is therefore important to enrich the ex-situ, in vivo, or in vitro collections with specimens from the primary zones of the species. Such specimens would serve as true references for the wild pole 2 of M. balbisiana. The higher intra-diversity observed in pole 2 suggests that these populations could retain some original traits of interest that may have been lost in pole 1 and that could be crucial for future conservation and breeding programmes.

Musa itinerans

Wild populations of M. itinerans were frequently found in wild areas, along roadsides and in forests. A total of 35 plants were collected in North Vietnam, 11 in Laos, and 10 in Yunnan (S3 Fig).

All observed plants displayed the distinctive characteristics of the species: a bud with mixed yellow and reddish-brown colours, small fruits with a long pedicel, and suckers branching at a great distance from the mother plant, hence the species name proposed by Cheesman [77] (S6 Fig). However, phenotypic variations were noted among the collected specimens. These differences did not challenge the botanical classification as M. itinerans, which is known to be highly polymorphic [47]. The main variations included the shape and colour of the male bud (Fig 6A–6C and 6N), which was more or less ovoid. The fruits were sometimes empty, indicating that the plant is not hermaphrodite (Fig 6F and 6N) as previously noted [47, 77]. When pollinated, the fruits were filled with well-formed and complete seeds (Fig 6H). Fruits could vary in length but were all located at the end of a characteristic long pedicel. At maturity, the colour of the fruits ranged from light green to brown, pink-red, and purple (Fig 6G, 6J and 6K). The bunch position was also variable, often sub-horizontal, sometimes horizontal, or angled (Fig 6N, 6Q and 6R). Additionally, some specimens had a black-purple pseudostem, while others had green or yellow pseudostems (Fig 6O and 6P).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Morphologic diversity of Musa itinerans.

(A–C) male bud. (D) male flowers. (E, I, L & M) leaves and midrib shape and colour. (F, G, H, J & K) fruits and seeds. (O-P) pseudostem. (N, Q & R) a complete bunch. (S) rhizome. Pictures by Matthieu Chabannes—CIRAD.

https://doi.org/10.1371/journal.pone.0307592.g006

These populations were found from the lowlands of Vietnam to elevations up to 1600 m in the mountainous areas of North Vietnam and Yunnan, in quite different ecological zones. M. itinerans is known to form large populations easily, thanks to its spreading rhizomatous root system, and is often a common pioneer species in degraded secondary tropical forests [47]. The species is dispersed across continental Southeast Asia including Yunnan, upper Burma (Myanmar) [77, 78], Laos, Vietnam [62, 74], and North Thailand [64]. In India, the species is confined to the northeast states: Manipur, Mizoram, and Arunachal Pradesh [72], as well as Nagaland [65]. M. itinerans is also distributed in southern China, including Hainan Island, Guangxi, and Guangdong provinces [79, 80], and extends to Taiwan where several varieties have been described [47]. Thus, the species appears to be widely geographically distributed. The area covered in this study is central to this distribution, and the collected specimens can therefore be considered well representative of the species.

The diversity tree of the 41 M. itinerans SNP genotypes (plus one reference from the CIRAD collection in Guadeloupe: PT-BA-00223) showed two main clusters, each grouping a core of closely related specimens, designated as iti.1 and iti.2, along with a few other slightly more distant specimens that were (shown in black on the tree, Fig 4B). A smaller cluster, iti.3, was still closely related to the first two. Cluster iti.4 was more distinct, indicating intraspecific diversity within M. itinerans. The high reticulations within the Itinerans group (Fig 2C) suggested that clusters iti.1 and iti.2 were partly intermixed. In contrast, the reticulation involving iti.4 showed a slight relationship with the Acuminata pole. Similarly, on the PCoA plan 1–2, the iti.4 specimens appear as the four lowest points of the Itinerans cloud, attracted by the Acuminata pole (Fig 2B). In these specimens, the loci specific to M. itinerans predominate across all the chromosomes, but some loci specific to M. acuminata are also present. These loci are scattered among the Itinerans loci rather than grouped into continuous segments ruling out recent interspecific hybridisation. All specimens in cluster iti.1 were found in the lowland areas of Vietnam (or in Yunnan, but just north of the Vietnamese border) (S3 Fig). The reference from the Guadeloupe collection logically fitted into this group according to its origin to the northwest of Hanoi. The second cluster iti.2 was geographically quite distinct, it included all plants collected in Laos and Yunnan at higher elevations, ranging from 500 to 1000 m, and was absent from Vietnam. The specimens in iti.3 were located in an intermediate area between the previous two, northeast of Dien Bien Phu, near the Chinese border. The specimens in iti.4, genetically more distant, were found in the same region as iti.1, at low elevations around Hanoi in Vietnam. However, it is noteworthy that two iti.1 specimens, were collected from high elevations in western Vietnam and in the north, on the Chinese border. Conversely, two iti.2 specimens were collected at low elevations in the Vientiane region of Laos (S3 Fig). This geographical distribution of the iti.1 and iti.2 genomic clusters should thus be interpreted as differentiation by partial isolation rather than adaptive selection to different ecological conditions. Importantly, none of the observed morphological variations correlated with the distribution of individuals across the different genomic clusters. This suggests that the varieties that have been proposed to date, within M. itinerans, could be just morphological variants without necessarily genetic support.

The distribution of dissimilarities (S2 Fig) indicated nearly equal intra-group diversity for iti.1 and iti.2, but both at higher values than those observed within clusters of M. balbisiana, within M. acuminata subsp. burmannica, or within M. yunnanensis. These higher diversities might reflect the particularly effective vegetative propagation of this species, which ensures the persistence of various genotypes over time. The observed heterozygosities were on average slightly lower than those for M. balbisiana, M. itinerans also having only female basal flowers [47]. Within the species, heterozygosities were higher for iti.1 than for iti.2 (Fig 5), displaying a bimodal distribution similar to that of M. balbisiana.

Although M. itinerans was not observed in cultivation in the surveyed area, it has been reported that, throughout the area, flower buds, inner pseudostem, and tender inner flower stem are commonly eaten as vegetables, the fruits are used medicinally, and the pseudostem is used as fodder for pigs, buffaloes, or cows [74]. It was also pointed out that among the various sympatric wild Musa species that are also gathered, M. itinerans is particularly valued as a delicate vegetable [46]. In Southwest China, young flowers and young pseudostems form a popular dish offered in local restaurants [81]. Furthermore, Uma [72] describes cultural practices in Northeast India, where local Adi tribes maintain this species at the periphery of their fields or plant stoloniferous suckers in their backyards. The interest of local people in this species for various purposes and its easy vegetative propagation suggest a human contribution to the observed allelic admixture and geographical distribution. Cluster iti.2 likely represents local endemic wild forms of M. itinerans in remote highland areas, while cluster iti.1, with higher heterozygosities, includes more intermixed forms resulting from contacts and local movements within the broad distribution area of M. itinerans. It thus encompasses a broader diversity than that present in iti.1, as evidenced by specimens in cluster iti.4.

Like M. balbisiana, the GBS genotypes show two subsets that differ in their levels of heterozygosity and are geographically disjoint. However, unlike M. balbisiana, M. itinerans is primarily a wild-gathered plant, with cultural practices limited to sucker transplantation, which nevertheless implies possible human selection and spread. We may therefore be in an in-between situation where human exploitation of natural populations would already impact the species’ diversity. As with M. balbisiana, it would be beneficial to enrich reference collections with representatives from the native cluster iti.2.

Musa acuminata

Among the specimens collected, 22 samples were assigned to M. acuminata based on their morphological characteristics. The position of the bunch, often subhorizontal and strongly compact (Fig 7B, 7I–7M), along with the colour and shape of the male bud with purple bracts (Fig 7A and 7D) and the brown colour of the macules on the pseudostem (Fig 7G and 7H), were all indicative of the subspecies burmannica or siamea. Other notable features included the frequent white discolouration on the inner side of the bract (Fig 7A), the general white colour of the male flowers (Fig 7A and 7D), and the open shape of the petiole canal (Fig 7F).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Morphologic diversity of Musa acuminata.

(A&D) male bud, male flowers. (B-C) fruits and seeds. (E) the basis of the leaf. (F) petiole canal. (G-H) pseudostem. (I-M) a complete bunch. Pictures by Matthieu Chabannes—CIRAD.

https://doi.org/10.1371/journal.pone.0307592.g007

All plants were collected in the wild, with observation sites concentrated in Laos, particularly in the northern regions, and extending to the adjacent border areas of northeast Vietnam and southern Yunnan (S4 Fig). They were almost absent from the entire eastern part of northern Vietnam, as well as from the northernmost surveyed areas in Yunnan. The plants were found at elevations ranging from 200 to 1200 m, mostly around 400 to 500 m.

Among the subspecies previously proposed for M. acuminata, three have been identified in this geographical area: subsp. burmannica [64, 82], subsp. siamea N.W. Simmonds [64] and subsp. burmannicoides E. De Langhe [83], with the latter being later proposed to be reduced under burmannica. The morphological characteristics observed did not allow for a clear assignment of the collected specimens to any one of these subspecies.

Of the 22 specimens collected, 19 were genotyped, along with 18 reference specimens representing the various acuminata subspecies. The genotypes of these 19 specimens were quite similar and could be regarded as belonging to the same genomic type (Fig 4C). The reference specimens burmannicoides ’Calcutta 4’ and siamea ’Pa Rayong’ were included in this cluster which was distinct from other acuminata subspecies: banksii, zebrina, truncata, and malaccensis (Fig 4C). The divergences within this cluster were much lower than the divergences with the other subspecies (S2 Fig), and the cluster appeared genetically homogeneous, without obvious internal structuring. Heterozygosity was low (1.72% on average) (Fig 5). Typically, subspecies of M. acuminata have female basal flowers [82, 84], however, most of the burmannica specimens observed during the Vietnam/Laos/China expedition had hermaphroditic basal flowers. Cheesman [82] also mentioned a specimen from upper Burma, morphologically assignable to M. acuminata, but with hermaphroditic basal flowers. Nevertheless, some collected specimens, although morphologically identical, had only female basal flowers. Similarly, Joe et al. [85] reported female basal flowers in specimens from southern India, but sometimes also hermaphroditic flowers with five functional stamens. This suggests that this trait could easily shift within a population in response, for instance, to ecological conditions.

This hermaphrodite character is in agreement with the observed low heterozygosities. These were much higher in the ‘Calcutta 4’ and ‘Pa Rayong’ references (more than 5%), which raises questions about their status as reference specimens. The heterozygosity of subspecies banksii was also low (1.6% on average), a subspecies known to have hermaphroditic basal flowers [82]. Higher heterozygosities were observed in zebrina (3%) and reached 7 to 8% in microcarpa ’Microcarpa’ and malaccensis ’Pahang’.

Failing to distinguish subspecies, we adopted the notion of a subspecies complex, referred to as the Burmannica complex, which groups the three described subspecies: siamea, burmannica, and burmannicoides. These results call for reconsidering the relevance of maintaining three distinct subspecies versus reducing them to three morphological variants of the same subspecies.

Previous genetic studies using molecular markers [31, 86–88], and more recent high-throughput SNP genotyping [30], also highlighted the close proximity of the nuclear, as well as cytoplasmic, genomes of these three subspecies compared to other acuminata subspecies.

Additionally, two reciprocal translocations (chromosomes 2/8 and 1/9) provide a specific genetic signature common to the burmannica, burmannicoides, and siamea accessions [89]. Shepherd predicted these translocations using cytogenetic approaches [90], concluding similarly about the similarity of these three subspecies. However, structural chromosome heterozygosity was detected in ‘Pa Rayong’ and ‘Long Tavoy’, indicating a more complicated evolution within the Burmannica complex [91].

The distance of the references ’Calcutta 4’ and ’Pa Rayong’ from the collected specimens (Fig 4C) suggests a greater intraspecific diversity within this Burmannica complex, not sampled in this study. This also highlights that the references available in main germplasm collections lack representatives of this group from North Vietnam, North Laos, and Yunnan, and, therefore, do not encompass the overall diversity of this subspecies.

The geographical range covered in this study also extends to Myanmar (then Burma): the area around Myitkyina in the north and further south near Dawei (formerly Tavoy), as well as the coastal islands in the Andaman Sea [64, 82, 83]. In particular, the references ‘Calcutta 4’ and ‘Long Tavoy’ from the Calcutta Botanical Garden are of Burmese origin. The subspecies siamea is documented in Thailand but the great variability reported questions about the validity of its subspecies status [46].

This range does not appear to extend westward into the northeastern Indian subcontinent, where, however, several other Musa species are attested [92, 93]. Only a few isolated instances of the Burmannica complex are reported in Assam and Meghalaya [72]. However, occurrences of the Burmannica complex are attested on the southwest coast, in the moist evergreen mountain forests of the Western Ghats [94, 95]. It should also be noted that the specimens from the Madras province described by Jacob [96] as Musa kattuvazhana (‘forest banana’ in Malayalam), and recently redescribed [97], are native to the Western Ghats. The synonymy of M. kattuvazhana K. Jacob with M. acuminata subsp. burmannica has been established [85].

Further south, in a small mountainous region in southwest Sri Lanka, a seminiferous banana known locally as ’Unel’ or ’Unakehel’ has been identified as M. acuminata subsp. burmannica [98]. Additionally, specimens of M. acuminata described in the Andaman and Nicobar Islands [99] as M. kattuvazhana, have been genomically confirmed as part of the Burmannica complex of M. acuminata [100].

The spatial discontinuity and distance from the central distribution area, alongside sporadic distribution, as well as the absence of other Musa species, suggest local naturalization of introduced plants in the Western Ghats and Sri Lanka. The presence in the Andaman and Nicobar Islands, off the Burmese coast, could imply a transfer to Sri Lanka via the Indian Ocean and a subsequent transfer to the Indian coast. These sporadic and discontinuous occurrences do not support natural dispersal mechanisms, such as bird-mediated seed spread, prompting questions about the people, and their motivations, behind these transfers.

Complete sequencing of various subspecies of M. acuminata suggests rapid radiation shortly after divergence from M. balbisiana [29]. The Burmannica complex, positioned at the base of the evolutionary tree, represents a primary form. Later, the dispersal of M. acuminata subspecies from northwest to southeast Asia led to differentiation through geographic isolation. This differentiation however, is only partial, allowing for intersubspecific hybridisations. For instance, hybrid forms involving subspecies malaccensis and the Burmannica complex have been detected by high-density genotyping [30], these wild hybrids originated in northern Thailand, an area where the distributions of the two subspecies overlap.

In summary, M. acuminata is well-represented in the study area, by forms belonging to the Burmannica complex. Unlike M. balbisiana or M. itinerans, the observed range does not cover the entire study area but is restricted to northern Laos and neighbouring regions of Vietnam and Yunnan. Publications of earlier collecting missions extend this range to eastern Myanmar and northern Thailand, but scarcely further. The region centred on upper Laos thus appears to be the botanical core of this complex and the genome common to the collected specimens can be seen as the original genome of this complex. These genotypes are notably absent from reference collections, highlighting the need for their introduction, firstly in the in vitro collection of the International Transit Center (ITC). This is especially important as the Burmannica complex is distinct from other acuminata subspecies and represents a primitive acuminata form ensuring the connection with the other species of Musa.

Musa yunnanensis

The wild species M. yunnanensis was commonly found as a wild plant, growing abundantly along roadsides, and occasionally cultivated for animal feed and non-fruit uses. We collected 22 specimens in Yunnan Province of China and three specimens in northern Laos, near the Chinese border (S5 Fig). The morphological type was easily identifiable according to the morphotype description [101], with some variations in male bud persistence and the presence of withered bracts.

As described by Häkkinen and Wang [102], the species is highly cold-tolerant and occurs exclusively in high-altitude forests up to 2100 m in Yunnan and likely in neighbouring Myanmar. This species thrives in wild populations along roadsides, ravines, and steep slopes, as well as in cultivated populations as animal fodder. Four varieties have been defined within M. yunnanensis [102]. The yunnanensis variety is the common type, about 5 m high with 8-handed bunches. The caii variety is smaller, not exceeding 4 m, with 3-handed bunches. The yongpingensis variety is the largest, reaching up to 6 m in height, with10-handed bunches. The jingdongensis variety is about 4 m high, with 10-handed bunches.

Genome polymorphisms resulting from GBS genotyping of 22 of the 25 analysed specimens clearly distinguished the genome of M. yunnanensis from that of M. itinerans, M. balbisiana and M. acuminata, with much lower intraspecific diversity than these other species. In the diversity tree as on the factorial plan (Fig 2B and 2C), the proximity to specimens of the former Rhodochlamys section was notable. Li and colleagues [3] had already shown, based on nuclear ITS markers, the close relationship between M. yunnanensis and M. ornata. Despite the low intraspecific variability of M. yunnanensis, the tree showed three distinct clusters (Fig 4D). Cluster yun.1 was found in mid-elevation areas, around 1100 m on average, from Laos to central Yunnan. Cluster yun.2 was related to yun.1 but with slightly wider internal variability, at higher elevations between 1600 and 1800 m. The last cluster, yun.3, included specimens observed only at high elevations, around 2000 m. The heterozygosity of clusters yun.1 and yun.3 was similarly low (1.1 and 0.9% on average, respectively), whereas it was much higher for yun.2 (2.3%) (Fig 5). Musa yunnanensis has hermaphroditic basal flowers and self-pollination is the main mating system [101]. This would support the low heterozygosity of clusters yun.1 and yun.3 but raises questions about yun.2.

Unlike M. balbisiana and M. itinerans, the genomic clusters of M. yunnanensis correlate with morphological types. For yun.1 specimens, the fruits, filled with seeds, are small and swollen at maturity. The bud is tapered, sometimes slightly discoloured, and the bracts are strongly revolute, revealing a white/yellow internal side. The trunk is sometimes entirely black (Fig 8A–8H). This type corresponds to the classic morphotype M. yunnanensis var. yunnanensis [102].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Morphologic diversity of Musa yunnanensis.

(A–H) Musa yunnanensis classical type. (L-Q) caii variety. (R-Y) hybrid type. Pictures by Matthieu Chabannes—CIRAD.

https://doi.org/10.1371/journal.pone.0307592.g008

The yun.3 cluster included specimens with massive, obtuse buds and globose fruits at maturity (Fig 8I–8Q). These specimens can be morphologically assigned to M. yunnanensis var. caii, described by Häkkinen and Wang [102] from specimens typically found up to 2000 m elevation, in mountainous areas where no other Musa species grows.

The specimens in the yun.2 cluster displayed combinations of characters assigning them to var. yongpingensis [102]. The bunch position is intermediate (horizontal in var. yunnanensis and almost vertical in var. Caii). The petioles are waxy, with dry papery wings at the base. The male bud has bracts with blue-purple external faces and cream internal faces. The male flowers are cream-yellow with an orange apex of the compound tepal. The male buds often degenerate before fruit maturity. The seeds are tuberculate and large, up to 8-10mm across. The bunch is compact, bearing short fruits with a width roughly half their length and a short pedicel about 1cm long (Fig 8R–8Y). Some of the mentioned characters, in addition to the pale purple colour of the dorsal face of young leaves (Fig 8U and 8Y), are reminiscent of M. sikkimensis. This species was initially described in the highlands of Sikkim, on the slopes of the Himalayas in northern India, neighbouring Bhutan [64], Tibet, and possibly Myanmar [103]. The higher heterozygosities of this yun.2 cluster could suggest a more allogamous mating system for this variety.

Chromosome paintings of the three unassigned specimens display a genome combining one haplotype from M. yunnanensis and one haplotype from M. acuminata (S6 Fig), indicating interspecific F1 hybrid between these species. Their very high heterozygosity, averaging 8.4%, confirms this hybrid nature.

The phenotype of these interspecific hybrids was reminiscent of any seminiferous Musa section species, with a subhorizontal bunch and numerous hands with small fruits (Fig 9), but with characteristics specific to both M. yunnanensis and M. acuminata, which led to the assumption of a possible hybrid form when collected in the field. A few rare traces of pollen were found on the flowers and vestiges of seeds were observed inside the fruits, but these seemed aborted and non-functional. The hybrid was therefore most likely sterile or had very low male fertility. The fruits showed no indication of parthenocarpy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Morphologic characteristics of Musa acuminata × yunnanensis hybrids.

(A–B) Male bud, (C) pseudostem, (D) whole plant in a natural context, (E, F & H) bunch and fruits, (G) petiole canal, (J) the basis of the leaf. Pictures by Matthieu Chabannes—CIRAD & Gabe Sachter-Smith.

https://doi.org/10.1371/journal.pone.0307592.g009

These specimens were found in wild conditions, with two specimens in Laos, at 1250 m elevation, and the third, at 550 m elevation, in China. While the first two were close enough to potentially represent the same population, the third was over 80 km away, suggesting at least two independent events. This indicates that natural interspecific hybridisation is not exceptional. These hybrids were collected in areas where wild populations of the two parent species coexist, specifically at the southern edge of M. yunnanensis range, corresponding to yun.1 area. Thus, it is likely that the M. yunnanensis genome contributing to these hybrids comes from the yun.1 cluster.

Species of the former section Rhodochlamys

Several representatives of the ornamental species from the former section Rhodochlamys were observed, in Laos and China (S7 Fig). The classical types, M. ornata and M. velutina, were cultivated in gardens. In contrast, M. rubinea was found growing spontaneously in the forest. Musa rubra (synonym M. laterita) was observed only once, in cultivation, in southern Laos, near Thailand, where it was said to originate. However, a farmer in Laos indicated that he saw this plant, known as ’Kuay Ka Chung’, growing wild deep in the forest, far from his farm, but the collecting team could not locate any wild populations.

Note that M. rubra is often referred to as M. laterita. Cheesman [104] published M. laterita as a new species from seeds sent from Myanmar while recognising its proximity to M. rubra. This morphological and genetic similarity was confirmed when Joe and colleagues [5] reduced M. laterita to be conspecific with M. rubra.

Genomic characterisation showed that the specimens from the former section Rhodochlamys do not form a homogeneous group. In the diversity tree (Fig 2C), a first group, which included the species M. ornata, M. velutina and M. rubinea, was distinct from M. acuminata, M. balbisiana and M. itinerans. This group was compact showing much lower intraspecific diversity compared to the three previously mentioned species. As noted earlier, M. yunnanensis is closely related to this group.

In contrast, the species M. rubra (both the field specimen and the collection reference), which is morphologically similar to Rhodochlamys species in its erect inflorescence, was linked to M. acuminata (Fig 2B and 2C). Within M. acuminata, M. rubra was closer to subsp. burmannica (Fig 4C) than to the other acuminata subspecies. Previous studies emphasised the morphological and genomic proximity of M. laterita to M. acuminata [3, 86, 105]. Based on the chloroplast genome, M. laterita was proposed to be a hybrid between a species from section Rhodochlamys and M. acuminata, with the latter providing the maternal material [12]. Our GBS data on the nuclear genome showed highly homozygous chromosomes for M. rubra, with a significant number of alleles specific to M. acuminata subsp. burmannica, interspersed among alleles specific to other primary pools. For example, on chromosome 6, (S8A Fig), M. rubra showed a pattern distinct from a pure burmannica specimen (S8B Fig). Therefore, M. rubra cannot be interpreted as a direct hybrid with burmannica but rather as a closely related form, sharing a common ancestor that provided their similar plastome. If burmannica is a primitive form within M. acuminata [29], it makes sense that M. rubra, as a close relative to M. acuminata, shares more specific alleles with burmannica than with other acuminata subspecies.

M. rubra/M. laterita is primarily known worldwide as an ornamental plant [106]. However, this species is also found naturally in South-western Myanmar [107] and in northern and western Thailand [46]. Additionally, it has been reported as wild only in Manipur and Mizoram in North-East India [5]. The range of M. rubra thus overlaps with that of M. acuminata burmannica; their joint occurrence in the same geographical area suggests that this region is likely the origin of their common ancestral forms.