Frédéric Bakry has carved for himself a unique career thanks to his readiness for taking calculated risks and exploring the roads less travelled.
Frédéric Bakry has vivid memories of the day in 1992 when he and his colleague Jean-Pierre Horry danced at the foot of a flowering banana plant. The CIRAD breeders, who were then based at the institute’s Neufchâteau field station in Guadeloupe, had used a ground-breaking, and rule-breaking, strategy to create hybrids and were curious to see what the plants would look like.
Their main innovation had been to double, using colchicine, the genome of a diploid cultivar (2 sets of chromosomes) to produce a tetraploid (4 sets). The diploid in question was a cultivar called ‘Rose’. It had inherited its genome from the wild species Musa acuminata (hence the shorthand AA to describe its genome), but had lost much of its fertility during the domestication process. Not all AA cultivars have their fertility restored from having their genome doubled, but ‘Rose’ had responded well to the procedure and the doubled diploid was used as the male parent (the pollen donor).
The choice of female parent (the ovule donor) also departed from the norm. “It’s taken as a given among breeders that only cultivars can be female parents”, says Bakry, “even though there are no genetic reason to support this view”. For this cross, the French breeders had opted for the banana’s other wild relative Musa balbisiana as the female parent.
Beyond knowing that the progeny would be triploid (3 sets of chromosomes), the breeders weren’t sure what to expect from the 4x/2x strategy, as it became known. They crossed their fingers when they saw that one of the plants was producing one hand after another. When it stopped after the 18th hand, they broke into a dance. “We couldn’t believe it” recalls Bakry. “It had worked. What we were doing was risky, but I think it’s important to try out different approaches. When CIRAD set up its breeding programme in the late 1980s we used the standard strategy to show that we were up to the task, but I wasn’t convinced that it was the way to go.”
Bakry has never been the type to settle into a routine. Once he has found the solution to a problem, he usually leaves the scaling up to others while he moves on to the next challenge, a trait that still defines him 30 years after embarking on a career on bananas.
The making of a banana breeder
After graduating in 1982 from Paris XI University, the country's incubator of plant breeders, CIRAD offered him a PhD scholarship to apply protoplast fusion to bananas as a way of bypassing the crop's sterility. Up to that point, the young Bourguignon (from France’s Burgundy region), had not envisaged a career in tropical agriculture, but he soon got hooked on the banana despite the fact that the topic he had been handed was not viable in the short term. Indeed, at the time, the tissue culture of bananas was just beginning and not enough was known about the inner workings of banana cells and tissues to start doing protoplast fusion. He accordingly changed his topic to the culture of banana tissues, an expertise that earned him a two-year stint at the Brazilian Agricultural Research Corporation Embrapa in Brasilia after finishing his PhD.
His mission was to share his know-how on callus formation, but Bakry also wanted to take advantage of his posting to learn banana breeding from one of the field’s experts, Kenneth Shepherd. Shepherd is one of the alumni of the British school that produced the likes of Ernest Cheesman, Kenneth Dodds and Norman Simmonds, Shepherd’s former boss and co-author on the seminal 1955 paper on the taxonomy and origin of cultivated bananas. The respected scientist happened to be working at Embrapa when Bakry was there, except that he was 1,300 km away in Cruz das Almas. After a couple of visits, Bakry moved northeast to work with Shepherd. When he told him that he wanted to be a banana breeder, Shepherd gave him access to his archives. “They were full of fascinating material that had never been published”, recalls Bakry. “I encouraged him to write it up. Some of it was eventually published by INIBAP as Cytogenetics of the genus Musa.”
It’s also in Brazil that Bakry developed a reputation as a tinkerer. At the time, Embrapa was not the well-resourced research institute it is today. Bakry recounts how he worked out a way to estimate the pH of a solution without a pH meter after he was told that the electrode that had just arrived after a six-month wait was already broken. As he says, “in the tropics, you learn to be resourceful”, something Shepherd was also good at. Among other useful procedures, Shepherd had taught him an easy-to-use-under-any-condition chromosome counting technique, which Bakry later published in Fruits and dedicated to the deceased scientist.
After two years in Brazil, Bakry was ready to move on. In 1987, he was sent to Guadeloupe, where CIRAD had recently set up a banana breeding programme at its Neufchâteau station. Curious to try with bananas the techniques he had learned at Paris XI, he convinced his colleagues that they needed doubled haploids to produce F1 hybrids. “Huge mistake”, he says in retrospect. “Too much homozygosity is not good for bananas. The plants were weak and would have been useless for that purpose.” But even his failures have a way of coming back as successes. The DH Pahang accession Bakry and his PhD student produced when they were developing the anther culture technique for generating doubled haploids was later selected to sequence the banana genome.
Bakry was considering his next move when he received the visit of Ivan Buddenhagen, another larger-than-life personality known for his inquisitive mind. “Besides my CIRAD colleagues, Ivan is along with Ken one of the banana scientists I admire the most. I owe him a lot. When he visited, I showed him the doubled haploids. That’s when he told me about a paper by E. T. Bingham, Maximizing Heterozygosity in Autopolyploids. We had gotten it all wrong. Instead of maximizing homozygosity, we should have been maximizing heterozygosity. Ivan’s visit confirmed to us that we were on the right track with the 4x/2x strategy, which he had himself recommended in a 1986 article, and on which we had also been working.”
The high point of that proof-of-concept experiment was the 18-hand hybrid, but the celebratory mood dampened when the plants started showing symptoms of a disease caused by the Banana streak virus (BSV). The strange thing was that half of the plants were sick while the other half were healthy, as if the disease was following Mendel’s genetic laws. After considering all possibilities, Bakry deduced that the virus was integrated in the DNA of Musa balbisiana and immediately shared his findings with his CIRAD colleagues. Once again, what looked like a failure turned out to be a scientific breakthrough. From a breeding point of view, however, it meant placing a moratorium on using Musa balbisiana and B genome cultivars in crosses.
The making of edible bananas
After a period of working with cultivars derived from Musa acuminata, which don’t have activable BSV sequences integrated in their genome, Bakry and his colleague Christophe Jenny wanted to start using the B genome again. As luck would have it, the Indian cultivar ‘Kunnan’ (AB genome) they had set their sights on turned out to be free of activable BSV sequences, most likely as a result of farmers unwittingly selecting plants with degenerated sequences, reckons Bakry. But what really blew him away was finding hybrids that could pass as Mysore bananas (AAB genome), when the doubled ‘Kunnan’ (AABB) was crossed with Musa acuminata ssp. malaccensis, and Pisang awak look-alikes (ABB genome), when it was crossed with Musa balbisiana. The CIRAD breeders had not set out to produce either type of banana, but the fact that both types showed up among the progenies raised the tantalizing prospect of recreating familiar triploids.
Bakry is not suggesting starting from scratch with wild species. In fact, one of the reasons he and his fellow breeder wanted to work with ‘Kunnan’ was to capture the genetic heritage of generations of discriminating farmers polishing off nature’s handiwork. The first stirrings of edibility most likely appeared in parthenocarpic bananas, i.e. fruit that can develop without being pollinated. There had to be humans around to propagate these plants vegetatively, but there’s more to the domestication of bananas than parthenocarpy. These edible, but still fertile, diploid bananas continued mating with nearby plants, some of which would have been more or less distant relatives, a process that produced increasingly sterile and seedless cultivars. Triploid bananas materialized when one of the parents normally passed on half of its genome, while the other contributed an unreduced genome (a phenomenon called meiotic restitution which, in triploids at least, is more common in ovules than in pollen). From that point on, the only way to produce diversity depended on farmers selecting and vegetatively propagating the natural mutants that appealed to them.
To explain the genetic composition of certain triploid cultivars, some scientists have speculated that an extra step, during which the triploid cultivar was fertilized by a diploid parent (a process called backcrossing), was also involved. Bakry is not convinced. Not only does he disagree that backcrossing needs to be invoked to explain the structure of triploids, he also doesn’t see how it could have happened. “It’s possible to fertilize a triploid cultivar that has some residual female fertility with the pollen of a wild or edible diploid. That’s what breeders do all the time in standard breeding schemes. But not only do they obtain very few seeds, the seeds also have a very low germination rate. Breeders get around the problem by ‘rescuing’ the embryo from the seed and culturing it, but farmers depend on seeds germinating normally. The probability that the seed of a cross between a triploid cultivar and a wild or edible diploid germinated and that there was a farmer around to propagate the resulting seedling is infinitesimal, even more so if it has to be a triploid, let alone one that is better than the original.” Indeed, because triploid cultivars cannot undergo normal meiosis, the few reproductive cells (gametes) they manage to produce are either, haploid, diploid or triploid. Only the diploid gametes would have produced a triploid when fertilized by pollen. The other seeds would have been diploid or tetraploid.
Even though it’s highly unlikely that the mating of a triploid with a diploid contributed to the domestication of triploid cultivars, the 3x/2x breeding scheme was how breeders finally managed to produce synthetic banana hybrids, except that these hybrids are not triploids but tetraploids. The rationale for selecting the tetraploid progeny over the triploid one was to avoid breaking up the cultivar’s genome. For breeders like Shepherd, it was important to preserve the genetic integrity of the cultivar.
On that point, Bakry is in total agreement with his former mentor. "Like the 3x/2x breeding scheme, the 4x/2x strategy is all about capturing the thousands of years of domestication that went into producing a cultivar, except that it does so at a lower ploidy level." A 2x/2x scheme could achieve the same thing, but only if meiotic restitution kicks in for one of the parents. A fertile doubled diploid (4x) undergoing normal meiosis is a more reliable producer of diploid gametes than a more or less fertile diploid cultivar.
Where the two strategies part ways is in the energy they each invest in generating seeds, as opposed to evaluating hybrids in the field. One admittedly extreme example of the 3x/2x strategy was presented at the 2011 ISHS-ProMusa symposium. The breeders obtained 200 seeds, but only after hand-pollinating 20,000 plants. By contrast, Bakry and his colleagues harvested as many as 500 seeds from a single plant. Their main problem is lack of space to evaluate all the hybrids they could potentially produce. To be sure, not all of them are gems. For example, 6 of the 38 hybrids from a cross between Kunnan 4x and Musa acuminata spp. malaccensis were “non-parthenocarpic plants bearing seedless stunted fruits”. On the other hand, half of the progeny had a bunch weight superior to the one of either parent (see graph).
Not surprisingly, Bakry doesn’t see any reasons for using triploid cultivars in breeding schemes, except when BSV is an issue. He also doesn’t see any reasons why breeders should only work with the banana’s two best known wild ancestors. Other wild species would produce hitherto unknown bananas, but they might have traits that could give the beleaguered cultivated bananas a much needed boost. For now, however, what Bakry wants to find out is whether the 4x/2x strategy can deliver the most sought-after triploid banana of all: a disease-resistant stand-in for the Cavendish cultivars that dominate the export trade.
Several years ago, an analysis of genebank accessions using RLFP markers had revealed an unsuspected kinship between diploid Mlali bananas collected in the Comoros at the end of the 1980s, and Cavendish and Gros Michel bananas. Bakry had seen the Mlali accessions in the Guadeloupe field collection and had not been impressed. He found it hard to believe that they were related to the current export banana as well as to the previous holder of the title. When a later study using SSR markers gave similar results, he reckoned there was something to explore, except that he didn’t want to be constrained by the Mlali accessions in the CIRAD collection. He also wanted to see what would happen if they used better specimens. So off he went in 2010 to collect more samples. To make it easier to bring back material, he went to neighbouring Mayotte, a French overseas department. Thanks to his local assistant, he was able to collect plants that produce 25 kilo plus bunches. The accessions have since had their genome doubled in Montpellier, where Bakry is now based, and were sent back to Guadeloupe where they are currently flowering.
In the meantime, the breeders in Guadeloupe had started collecting data on the accessions from the Comoros, some of which Bakry presented at the 2011 ISHS-ProMusa symposium. At the time, Bakry’s hypothesis to explain the adaptability of Cavendish bananas was their high level of heterozygosity. Now he would say that it is probably because they had a great parent, a Mlali cultivar. He is so impressed by these diploids that he would gladly spend the rest of his career trying to find out what makes them so special, if only to honour Yvette Dattée, an emeritus professor of plant breeding and president of the Société National d’Horticulture de France. In 1995, she visited him in Guadeloupe and made a comment that has been haunting him ever since. She said, "Monsieur Bakry, the day you will understand why a single genotype accounts for nearly half of world production, is the day you will be able to run a good banana improvement programme".