My banana breeding project

Row of small potted banana plants.

Many people who attended the gathering at my place last year would have noticed that among other things I have a large and varied collection of bananas. With a marginal climate and only a small commercial industry, W.A. has only a small number of varieties and there are also identification issues with the privately traded plants here. My collection can be broken into two groups. The first group of plants are of tissue culture origins that have been imported from the Australian arm of the world Musa gene bank. These plants are effectively cloned from identifiable certified disease free stock to meet quarantine requirements (group 1). Group 2 are locally sourced and only tentatively named; a couple currently have no real name at all. I am hoping to firm up the identity of these when they flower.

I have over twenty varieties of edible banana in my collection. This sounds like a lot but there are over three thousand edible cultivars in the central Musa database on Promusa, even after allowing for multiple naming, and that's ignoring the Fe'i which are another group altogether with upright fruit. Then there are also all the wild species. Domestic bananas are a varied group in both features and origin. They belong in the genus Musa, Section Musa (now including those formerly in Rhodochlamys) with a diploid chromosome count of x = 11, 2n=22.

Barring a few genetic oddities edible bananas are either pure Musa acuminata, which has a number of sub-species (donor of the A genome and designated AA), or almost exclusively interspecific hybrids between M. acuminata and M. balbisiana (donor of the B genome with their hybrids denoted as AB). The A genome usually contributes sweetness and the B starch that usually requires cooking. Additionally their ploidy count can vary; the largest group of edible cultivars being almost sterile triploids, followed by the diploids; tetraploids are usually of synthetic origin and are rarely found in nature. Any plants that end up with a ploidy higher than this tend to be lacking in vigour and short-lived. Common genetic combinations are: AA, AAA, AAAA, AB, AAB, AAAB, ABB and ABBB. I have the following varieties in my collection:

Group one:
Gros Michel AAADwarf Cavendish AAADwarf Kalapua ABB
Inarnibal AAMangaroa Toroka AABVunamami AS
Silver Bluggoe ABBDwarf Red Dacca AAABlue Java ABB
Pisang Ceylan AABPacific Plantain AABSucrier AA
Lady Finger AABWilliams Cavendish AAALacatan AA
Dwarf French Plantain AABDwarf Ducasse ABBGoldfinger AAAB
Group two:
Monkey FingerGreen RedTall Plantain
CanarvonBluggoeLady Finger
Banana tissue cultured clones in their laboratory tubes.

I have always been curious why there are no real cold hardy edible bananas, but it's actually very simple. As with most crops, research into breeding is focused on disease resistance and improved performance. Breeding bananas is a difficult, time consuming and expensive exercise. It's largely funded by banana growing countries; the majority of these countries are in the developing world with finite research capabilities and being generally tropical in climate they have no need for cold hardiness. More temperate climate countries have evolved their own cash crops to grow and aren't going to spend money developing a crop that can be easily purchased and shipped, especially when the reality would be that they couldn't produce them cheaper than their competitors.

Economics aside, is it possible to breed e.g. cold hardy and or short cycle bananas? Can banana plants that will tolerate the cooler weather and ripen fruit in the shorter summers of the higher latitudes be bred? Theoretically yes - all one has to do is introduce the appropriate genes into the gene pool of domestic lines, but implementation is the hard bit.

There are quite a few cold hardy wild diploid bananas within the Musa section, the toughest being the Japanese fibre banana Musa basjoo. Growing happily down to -2°C, the underground corm will take -10°C and short spans of -12°C. Other slightly less impressive but still hardy species of Musa are yunnanensis, sikkimensis, griersonii, formosana, iterans, thomsonii, plus an M. acuminata found at high elevation in China. Furthermore there are the smaller short cycle bananas formerly in the subsection known as Rhodochlamys with species such as laterita, velutina, ornata and rubinea known for their speed to fruit in areas with shorter summers.

There are also other cold hardy species placed in the second Musa section (Callimusa) using the latest molecular taxonomic classifications. With a chromosome count of 2n = 14, 18 and 20, this is where Musa maclayi which is the ancestor of the Fe'i bananas (M troglodytarum) resides. There are quite a few cold hardy species in this group and even a few hybrids between the two sections (Tat'ia, Umbubu, Sar) but essentially I'll be ignoring them due to a lack of available material.

I've been fortunate enough to obtain a number of plant specimens of cold hardy bananas. This has led me to consider trying my hand at breeding some unusual hybrids. As stated, most edible bananas are pure acuminata or acuminata x balbisiana. Currently having found no wild pure balbisiana plants with parthenocarpy genes there is uncertainty whether purely balbisiana edible bananas exist, and any BB or BBB references you see may well be in error.

The seedlessness of bananas is due to two separate mechanisms. The first is sterility - bananas can be male or female, sterile, or both, and this property varies depending on ploidy level as well as parentage. The second is parthenocarpy, and currently three genes (probably amongst many others as the A and B genomes each have about 36,000 genes) are known that control this. However these two separate mechanisms don't always go hand in hand; this means you can have bananas that are fertile males or females with the parthenocarpy genes producing seedless fruit. It is this property I intend to exploit in an effort to cross some domestic cultivars with wild species.

The variety Vunamami in my collection is a good example of the possibility. One of a few cultivars of an unusual type from PNG, it's an edible banana with the genome code AS. It's a diploid hybrid of acuminata and schizocarpa. Whilst a wild selection, it demonstrates that edible hybrids can be made between acuminata and other wild species beyond balbisiana.

For my wild donors I will be utilizing two species in Musa, section Musa, namely Musa yunnanensis, M. thomsonii and M. griersoni with the possibility of using M. formosana at a later date. One cold tolerant banana I do have but am reluctant to use is Musa sikkimensis. This species has many seedling selections such as Red tiger, Red flame, Darjeeling Giant, Plain Sikkimensis, Helen's Hybrid etc. However there is considerable uncertainty whether these varieties are pure M. sikkimensis or wild hybrids, and as such I'm unsure if my specimen is of pure parentage. I have also been unable to locate anyone with Musa basjoo locally, and doubt it's in the country let alone the state. This is not a huge loss as it is only very distantly related to acuminata and in previous studies they were shown to not cross very successfully.

On the domesticated side it gets a little harder. With all the different hybrids and ploidy levels the combinations seem endless. However there are a few considerations to take into account, the main one being the limitations of home breeding and identifying the genetics of the progeny produced. For example, if I start with hybrid acuminata (A) x balbisiana (B) species (i.e. diploid AB) and cross them with Musa thomsonii, the offspring could be a combination of all three species or in fact hybrids of either i.e. AT, ABT or BT. If the offspring only pick up the balbisiana genome then they may be of little use because according to current knowledge balbisiana has no parthenocarpy genes. Also hybrids with B genomes may contain banana streak virus incorporated in the genome (endogenous BSV) and this can form free infectious particles passed on with vegetative propagation, only showing up when plants are stressed. Whilst not limited to the B genome with rare documented occurrences in pure A groups, it can cause plant health and quarantine issues.

So after consideration, as much as I would like to use my B type bananas (they're a big chunk of my collection) I'm going to concentrate on pollinating my wild Musa species with AA or AAA edibles and vice versa. It will eliminate the risk of eBSV as well as simplify the genetics of my breeding program, with the added benefit that I can focus on growing a bigger number of fewer types to increase my chances of synchronous flowering.

To the current way of thinking most of the wild non-acuminata/balbisiana species are diploid in nature only. Triploids and above generally only occur in two cases; firstly, wild Musa acuminata and its hybrids with balbisiana and/or other Musa species; and secondly, domesticated hybrids also usually contain the A gene, and as stated, there are currently no domesticated cultivars listed as BB, BBB etc. By pollinating wild species with pollen from pure (A) groups my F1 progeny could be as follows: using Inarnibal (AA) and yunnanensis I could theoretically get AY or AAY from the diploid pollen and if I used Red Dacca (AAA) I could get AY, AAY or AAAY from the triploids. It is extremely unlikely I will get any AYY or higher levels of wild ploidy in the first generation. The fact that my donor wild species are all known diploids should limit the F1 ploidy variation to just the domesticated side.

Another possible outcome to be tested is that with only pure A type domestic bananas in my crosses it could possibly simplify identification of the offspring. I intend to use the classic scoring system for banana features given by Simmonds and Sheppard (see Nomenclature of cultivated bananas ¦ News, knowledge and information on bananas). Points are given depending on whether certain traits are representative more of acuminata or balbisiana. Although this is heavily dependent on subjective assessment of morphological features and more fundamentally has been shown to be complicated through considerable gene exchange (meiosis) between A and B, it nevertheless is a good first approximation. If I were to use AB, AAB or ABB cultivars in my program, this would be of little use. But as I intend to use pure AA species I might be able to copy their system as long as: (i) my wild species (yunnanensis, thomsonii etc ) show uniformity at the measured points and match taxonomic descriptions, and (ii) the points of difference used match in with the differences between my wild species and acuminata.

Given the knowledge that my wild species are all diploid, plus I know the ploidy level and consistent A genomes of my cultivated stock, there should be a reasonable chance of knowing what my F1 offspring are. I'm not sure if anyone has bothered to do this before so once my pure wild species start to flower I'll start taking pictures as reference. It will help me confirm my wild species are what I was told and give me a starting point to compare future hybrids.

In theory I'm going to breed for cold hardiness, parthenocarpy and taste, and then back cross to the wild species for cold hardiness, Cycle time will also be an important factor. Given the possible gene exchange (meiosis) with each new crossing, the A and Y genes could become mixed with the 2nd and subsequent generations (it is a point of conjecture whether meiosis will occur in the first generation hybrids moving into the second generation and onwards).

How will this translate from theory into practice? Due to the possibility of a location change mid-programme my plan is to grow my parent stock in cut-down ibc's (these are industry standard modular bulk fluid containers of one thousand litres) with three hundred litres of potting mix. This way I can move different plants into close proximity to facilitate pollination, and also move them in or out of my poly-tunnel if needed to try and synchronise flowering. Bananas have been fruited in as little as fifty-litre pots so mine should be fine. One program in Africa recommended at minimum five plants of each cultivar intended for crossing to get synchronous flowering. Once they are of flowering age I'd begin by crossing a diploid like Inarnibal or Sucrier and triploids like Red Dacca, Cavendish or Gros Michel with either - thomsonii, yunnanensis, griersonii or formosana. I'll pollinate both ways if possible to get offspring both ways as anecdotal evidence suggests the wild species will produce more seeds than the domestic hybrids. Chloroplast and mitochondrial DNA (maternal and paternal contributions resp.) are known to contribute to plant properties, so it will be interesting to see if the forward and reverse crosses fare differently in my case.

Once I get first generation seed to germinate I'm thinking of dividing the plants into groups based on vigour. Both low and high vigour ones I'd grow out, then rate their appearance relative to their parents to gauge which genomes are expressed. When they flower I'll remove the male flowers and test for parthenocarpy. Then I can split the groups depending on whether they display parthenocarpy or not. This would give me four groups.

The parthenocarpic ones (if any) could be taste-tested, followed by an assessment of cold hardiness. It's fanciful to think I'll get a good tasting, cold hardy, parthenocarpic banana first go (if I do I'll buy 10 Lotto tickets!). Those parthenocarpic plants with good taste will be back-crossed to the wild species to increase the content of wild DNA, and the inferior tasting and non-parthenocarpic ones will be back-crossed to the domestic parent to improve the traits I'm after. A decision will be made for some poor-tasting, non-parthenocarpic and low-vigour plants on whether to discard to save room and labour.

One unknown in all this is whether subsequent breeding of the F1 hybrids will increase the possibility of higher wild species ploidy? I'm not sure I can naturally breed triploids with more than one copy of wild species DNA in the 2nd, 3rd and subsequent generations. Can I pollinate an F1 hybrid designated AY with Y pollen and get an AYY for example? There seems no certainty about this. It has happened and is done with the ABB, AAB, ABT and ABS triploids, so I'm hopeful. If not, I'm limited to A+, AA+, AAA+ progeny. Depending on the answer to this question, meiosis may become important to spread the genomes between A and Y.

I'm unsure if meiosis will take place in my first gen hybrids going into the second generation but for those unfamiliar with the term, it's a form of crossing over between the two pairs of chromosomes as part of sexual reproduction. If you imagine looking at the picture below the green as an A chromosome and the yellow as a Y chromosome, you can see that there is a trading of genes in the reproductive organs. In theory this means the A would be unique as would the Y as each incorporates properties of the other. Whether this in fact will happen is not certain.

Spacer. Diagram of chromosomes trading genes.

Earlier, I briefly mentioned the old Rhodochlamys section bananas now included in Musa subspecies Musa. This taxonomic section was originally created for the smaller, shorter-cycling bananas; these species can grow and fruit very quickly. In ideal conditions M. velutina can grow from dormant corms and be finished fruiting in as little as twenty weeks. Some are also cold hardy, but I already have four species and once I start growing enough to try and get flowering at the same time, it represents a lot of banana plants. I have considered rubinea, velutina, laterita and ornata, as I have these currently. But I am undecided whether to add them to my plans and how to integrate them so if I'm unhappy with the performance of one of my initial wild species I may swap it with one of these.

Another method I considered for increasing the wild species genes into the domesticated bananas is a little more difficult but may be worth considering if I can't create natural triploids. I could start with diploid plants and maybe even create some wild diploid crosses then turn them into synthetic tetraploids. I'm hoping to purchase a tissue culture kit (kitchen culture maybe) to eventually clone my offspring if things go as planned. I could use this in conjunction with colchicine to up the ploidy level of the wild species to tetraploid level. The only downside is without sending samples to a research lab I have no way of verifying success afterwards; synthetic tetraploids do have a habit of reverting to diploids.

However, I should explain the advantages this may hold. I could cross and grow yunnanensis x rubinea hybrids YR (this has been done before) to give a cold hardy, short-cycle diploid. I could then tissue culture the progeny and treat with colchicine to create YRYR tetraploids (colchicine has been used to increase ploidy in plants for a long time). Theoretically these could be pollinated with pollen from AA diploids to give possibly AYR or AAYR. It would also be useful to incorporate more than one copy of single wild species DNA, ie if I treat thomsonii with colchicine I could get TTTT, back-crossed with acuminata I could get ATT and AATT. These tetraploids could have benefits, as more copies of the wild genomes should mean better expression of their traits and the hybrids would give me more spread of beneficial qualities. An AATT or AAYR would almost have to be parthenocarpic. For interest I have included some draft breeding pathways that may give me the desired results. Please note there will probably be many crosses and more generations than what is shown. But it gives an idea of how the crosses will work.

Breeding Plan for diploid parents

Breeding Plan for diploid parents. Spacer.

Breeding Map for diploid parents with a third species pollen crossed in

Breeding map for diploid parents with third species pollen crossed in.

Breeding Map for triploid domestic and diploid wild parents

Breeding Map for triploid domestic and diploid wild parents, showing tetraploid cultivars Spacer. A large potted banana plant.

I intend to do this over the next many years as a labour of love. It will be at least two years before my first plants flower, and there's also no guarantee I will get synchronous flowering, so I may have to note the dates and modify the climate accordingly. If I get sufficient seeds it will be a further 2-3 years before I can get fruit from them and then see what the first generation produces. People doing this sort of breeding work expect it could take twenty years for good results, and only in the last several years has any real headway been made. It's an incredibly tough problem and it's most unlikely I'll get an end result any time sooner, but if I can produce a variety with a good mix of being dwarf size, vigorous, short cycle, high yield, frost tolerant and seedless with good flavour it'll be worth it.

Troy Bailey

Images/Stars Spacer. LitchiLogo