Banana breeders screen 69 parental lines against five major pests and diseases

Banana breeders at the International Institute of Tropical Agriculture (IITA) have evaluated 69 parental genotypes from the Matooke, Mchare and Plantain breeding programs against five of the most damaging biotic constraints in East and Central Africa: the banana weevil (Cosmopolites sordidus), the burrowing nematode (Radopholus similis), banana Xanthomonas wilt, Fusarium wilt caused by Fusarium oxysporum f. sp. cubense Race 1, and black Sigatoka caused by Pseudocercospora fijiensis. The study, published in PLOS ONE, is notable less for any single resistant line it uncovered than for the way the screening was done. Rather than a conventional fully replicated field layout, the team used a partially replicated, or P-rep, experimental design, in which only a subset of genotypes is replicated while the rest appear once, allowing trials to proceed despite the chronic shortage of clonally propagated banana planting material.

The problem the design solves is specific to crops like banana. Because bananas are clonally propagated, every genotype in a trial must be multiplied in tissue culture before it can be planted, and multiplication rates vary enormously among genotypes. Some genotypes root readily and acclimatize quickly; others suffer contamination losses, root poorly, or grow so slowly that they cannot be ready when the rest of the trial is established. A randomized complete block design, the traditional standard, requires every genotype to be represented in every block simultaneously, which in practice caps the number of genotypes a banana phenotyping program can screen at any one time. With a P-rep design, genotypes can be slotted into batches as their plantlets become available, with checks and resistant controls replicated in every block to anchor the statistical comparisons.

The researchers assembled their plant material from disease-free suckers, established them in tissue culture using standard Murashige and Skoog-based protocols, and hardened plantlets for three months before trials began. Phenotyping took place at two IITA sites, in Arusha, Tanzania and Sendusu, Uganda, between 2021 and 2025, with each constraint evaluated at a single location. Four of the five constraints were assessed in screenhouse pot trials under artificial inoculation, while black Sigatoka was evaluated in the field at a disease hotspot under natural infection, with rows of a highly susceptible matooke cultivar planted as spreaders to concentrate inoculum. Trials were run in batches, five each for weevil and Xanthomonas wilt, four for nematodes, three cycles for Sigatoka and two for Fusarium wilt Race 1, with roughly 20 genotypes per batch and, on average, eight pseudo-replications per genotype per constraint.

The assays themselves were deliberately quantitative. Weevil damage was measured by infesting plants with three male and three female weevils, uprooting the plants 60 days later, and scoring corm cross-sections for the percentage of damaged cortex and inner cylinder tissue. Nematode resistance was assessed by inoculating each plant with 1,000 nematodes and, eight weeks later, measuring both root necrosis indices and extracted nematode population densities per 100 grams of fresh root. Xanthomonas wilt was scored weekly for four months after inoculation with a standardized bacterial suspension, and summarized as an area under the disease progress curve. Fusarium wilt evaluations combined external leaf symptom and pseudostem splitting scores with internal rhizome discoloration at three months. Sigatoka was tracked through standing and functional leaf counts, the youngest leaf showing streak symptoms and the youngest leaf spotted, plus severity ratings on a 0 to 6 scale across successive crop cycles.

Statistical analysis relied on a conjugate hierarchical generalized linear model, with genotype effects treated as fixed and batch and block-within-batch effects as random. Poisson distributions with log links were assumed for the weevil, nematode and Fusarium responses, while normal distributions were used for the Xanthomonas and Sigatoka traits. Genotype effects were estimated as contrasts against documented resistant controls, Calcutta 4 for most constraints and Monyet for Xanthomonas wilt, and visualized as forest plots with 95 percent confidence intervals. Genotypes whose estimates fell on or left of the reference line were classed as resistant, those with confidence intervals overlapping the line as tolerant, and those clearly to the right as susceptible. Alongside the phenotyping, the team genotyped the parental set on the DArTSeq platform, filtering to 13,306 SNP markers, and used principal component analysis and Ward’s linkage clustering to place the genotypes into four genetic clusters.

The resistance landscape that emerged was strikingly uneven. For the banana weevil, only two of 65 parental genotypes, 3.1 percent, were classified as resistant, with one further genotype tolerant. Nematode resistance was far more common: 14 of 59 genotypes, 23.7 percent, were resistant, comprising nine diploids and six tetraploids, and another 16 were tolerant. For Xanthomonas wilt, four of 64 genotypes, 6.3 percent, were resistant. Fusarium wilt Race 1 resistance appeared in 12 of 48 genotypes, 25 percent, with another 12 tolerant. Black Sigatoka proved the hardest constraint: no genotype met the resistant threshold, though two, 3.3 percent, were classified as tolerant. The susceptible landrace checks, including Mbwazirume, Mchare and Obino Lewai, consistently showed high damage and disease levels, confirming that the assays had adequate pressure.

Of the 42 genotypes phenotyped against all five constraints, eight were susceptible to every one of them, while the remaining 34 showed mixed profiles. Several lines stood out as multi-constraint donors. The wild diploid Calcutta 4 was resistant to weevil, nematode, Fusarium Race 1 and Sigatoka, though susceptible to Xanthomonas wilt. TMB2x 6142 was resistant to Xanthomonas wilt and tolerant to both Fusarium and Sigatoka. Five genotypes, 1297-3, 201087K-3, 5610S-1, 401K-1 and TMB2x 9172, combined resistance to both nematodes and Fusarium Race 1. Others paired resistance in one constraint with tolerance in another, such as TMB2x 8075-7, tolerant to weevil and resistant to nematodes, and cv. Rose, resistant to Xanthomonas wilt and tolerant to Fusarium. No genotype, however, was resistant or tolerant to all five constraints, underscoring how difficult broad-spectrum resistance remains in this crop.

The genetic clustering added a layer of interpretation that matters for crossing decisions. Resistant genotypes for the same constraint appeared both within the same cluster, suggesting they may share resistance alleles inherited from common backgrounds, and across different clusters, hinting at genetically distinct sources of resistance that could be pyramided. For nematode resistance, resistant lines turned up in clusters 1, 3 and 4; for Fusarium Race 1, resistance was spread across all four clusters, including PITA 5, 201071K, 29275S-4, 1438K-1, 201087K-3, TMB2x 9172 and 1297-3. The authors are careful to note that these patterns do not confirm distinct genetic mechanisms, but they give breeders a rational basis for selecting genetically diverse parents and broadening the genetic base of resistance in banana improvement.

The precision argument for the P-rep design was borne out by a direct comparison with earlier randomized complete block studies of weevil damage in overlapping genotypes. Standard errors in the P-rep trials ranged from 0.3 to 0.6, against 1.0 to 9.8 reported in the RCBD studies, while genotype classifications were broadly consistent, with Mbwazirume susceptible, Calcutta 4 and TMB2x 8075-7 resistant and SH 3142 tolerant in both. A theoretical calculation drove the practical point home: with the same total number of plants and a fully balanced three-block design requiring 12 plants per genotype, the trials could have accommodated only 75, 60, 22, 61 and 61 genotypes for the five constraints respectively, and only if all plants of each genotype had been available simultaneously. The P-rep design instead let the team screen genotypes as material became available, over 36 months of batched screening, with density plots showing broadly consistent distributions across batches.

For banana breeding programs working under the resource constraints typical of East and Central Africa, where banana is both a staple and a cash crop and smallholder farmers have limited access to chemical or cultural controls, the message is twofold. First, the resistance profiles generated here give the Matooke, Mchare and Plantain pipelines a vetted shortlist of parents for resistance pyramiding and pre-breeding, with several diploids flagged as promising donors for introgressing resistance into elite backgrounds. Second, the P-rep design offers a template for phenotyping clonally propagated crops where uniform planting material is the bottleneck rather than an afterthought. The raw datasets and genotype lists have been released openly on Figshare under a CC-BY 4.0 license, and the work was funded by the Gates Foundation, with the authors noting that screenhouse results will still need to be followed by multi-location field evaluation before any of these lines reaches farmers’ fields.

Subject of Research: Phenotyping banana germplasm for resistance to five biotic constraints using a partially replicated experimental design

Article Title: Efficient phenotyping of banana ( Musa spp.) germplasm for resistance to five biotic constraints using Partially replicated (P-rep) design

Article References: Kanaabi, R., Mwanje, G., Serebe, C., Mpanda, H., Jomanga, K., Uwimana, B., Vuylsteke, M., Brown, A., Swennen, R., & Nakato, G. V. (2026). Efficient phenotyping of banana (Musa spp.) germplasm for resistance to five biotic constraints using Partially replicated (P-rep) design. PLOS One, 21(10), e0361182. https://doi.org/10.1371/journal.pone.0361182

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0361182

Keywords: banana breeding, Musa spp, P-rep design, plant resistance, banana weevil, Radopholus similis, banana Xanthomonas wilt, Fusarium wilt, black Sigatoka, phenotyping, experimental design, IITA