The primary outputs of this intended project would be a seed-based collection of banana progenitors and a validated new approach to breeding
that will be much more efficient than current breeding schemes. These outputs should lead to the desired primary outcome, adoption of a much
more efficient banana breeding system for small-scale producers (SSPs) in Sub-Saharan Africa. A possible spillover outcome could be the
revamping of other major banana breeding programs that target other regions if desired by BRIN.
Banana is unique among major food crops (#7 in global production) in its dependence on a few genotypes for global production. Global banana
production is largely characterized by a few dominant ancient genotypes propagated clonally for centuries or millennia that have undergone
diversification by somaclonal variation. The #1 clone set is Cavendish, which accounts for >55% of global production. The second clone set,
African plantains represent nearly 15% and the third, East African Highland Bananas, contribute s >10%. A handful of other significant clone sets
account for most of the balance of global production.
Banana breeding for East and Central Africa has been successful over the last 20 years. The first varieties released ~10 years ago had large yield
increases over local cultivars (50-70%) but lacked important preferred traits for end-users. More recent v.2 releases have further improved yield
marginally but have much improved end-user attributes. More gain in yield and quality is expected in the coming decade.
However, classical banana breeding that begins with nearly-sterile cultivars is inherently limited and inefficient. Fertility is reduced about 99.9%
compared to wild bananas, which means banana breeding requires very large crossing blocks that produce relatively few seeds. Furthermore,
embryo viability is also reduced and only a variable minority (2-40%) of seeds, depending on genetic background, are viable even with embryo
rescue. These effects likely are caused by the fact that different banana progenitor subspecies or species vary in genome structure by large
reciprocal translocations. To date, at least 7 of the 11 banana chromosomes within Musa acuminata are known to be involved in large reciprocal
translocations. While the above effects reduce breeding efficiency, an equally important breeding hindrance is that the genome structural
differences result in large blocks of “linkage drag” in which affected chromosomal sections aren’t subject to normal crossing over during meiosis.
This means it is nearly impossible to separate desirable from undesirable alleles within those large blocks.
Provided access to the appropriate wild Musa germplasm is arranged, it should be possible to redesign banana breeding to remain within highly
fertile breeding pools based on genome structure. Inter-pool hybrids would then lead to sterile seedless bananas with hybrid vigor. Since
triploids are more sterile than diploids, ideally a modern program would have 3 breeding pools (A1, A2, A3) to recreate domestication to use
unreduced gametes from A1A2 hybrids crossed to the A3 pool to make triploid hybrids with A1A2A3 structure. The most productive ancient
banana cultivars: Cavendish and Gros Michel have this A1A2A3 structure. Similarly, the most productive AAB variety, Pome, has A1A2B
structure. Less productive traditional African cultivars have A1A1A2 or A1A1B structure with only 2 sub-genomes per triploid cultivar. Moving
from diploid parents to triploid progeny could happen via unreduced gametes, which are naturally more common in hybrids with heterogeneous
genome structure, or by chromosome-doubling initial A1A2 hybrids. Proof-of-concept could be achieved using heterozygous individuals
throughout the breeding process, but a longer-term effort might be to push pools toward inbreeds to eliminate genetic load and further
improve the breeding process by enabling use of DH technology and backcrossing to improve otherwise excellent parents that lack one or a few
essential traits.
As implied above, another limitation is that the global banana gene bank is deficient in the wild banana accessions needed to pilot the new
breeding scheme. Its focus is banana cultivars, including many clonal variants of important cultivar sets (=clone sets). Since cultivars’ seedless
state is largely caused by being heterozygous for different sub-genomes, the global gene bank is of little use to breeding. Of the 7 recognized
wild subspecies of M. acuminata, the gene bank includes only 3 subspecies with more than 3 accessions; 4 have from 0 to 3 accessions. Of
course, one clonal accession represents only a fraction of the gene pool of the population from which it was collected, and a small fraction of
what a seed-based accession from the same inflorescence would contain. A seed-based accession would represent the entire genome of the
mother and usually those of multiple male parents. Besides the recognized subspecies, an Indonesian taxonomist identified 15 botanical
varieties, of which only 3 corresponded to recognized subspecies and 12 were newly reported taxa. As sequencing information grows, we now
know that some portions of the genomes of important cultivars derive from subspecies not yet represented in the sequence database.
The BBTV Mitigation project has laid the foundation for collecting for a future seed-based Musa gene bank through studies on seed biology and
population biology. That work has confirmed that banana has orthodox seeds that can be stored long-term in cold (5 C), freezer (-20 C) or
ultracold (-196 C) storage. Storage success depends on fruit maturity and proper drying of seeds. An issue for collecting teams is that they rarely
arrive at just the right time to harvest bunches at optimum maturity. To address this, innovative methods to collect quality seed need to be
established and validated. Options might be to work with locals and geo-tag plant locations so that bunches could be harvested at the right time
for extracting quality seeds, or possibly to move to full crowd-sourcing for collection and extraction according to well-developed guidelines, with
plant selection first validated by digital imagery.
Population biology studies focused on M. acuminata subspecies that vary in degree of selfing, which results in large differences in the level of
heterozygosity. Genotyping seed lots allowed the researchers to model the number of female parents that need to be sampled to represent the
population to represent the gene pool. Since recently-discovered botanical varieties have yet to be characterized for degree of selfing, average
level of heterozygosity can serve as a good proxy for degree of outcrossing to determine the number of samples to represent a population.