Incorporating male sterility increases hybrid maize yield in low input African farming systems

The study addresses the challenge of low maize yields in sub-Saharan Africa (SSA), where yields are the lowest globally and production increases have largely come from unsustainable area expansion. Obsolete varieties persist despite changing climates, underscoring the need for rapid-cycle breeding and faster varietal replacement. Hybrid seed production in SSA relies on manual detasseling, raising costs and quality issues; as a result, lower-yielding but cheaper three-way hybrids are common due to higher seed yields of single cross females. Seed Production Technology (SPT) offers a way to reduce cost and complexity and improve seed purity. A dominant male sterility system using the Ms44 gene produces hybrids in which half the plants are non-pollen producing (FNP), eliminating detasseling and potentially increasing yield by reallocating resources from tassel/pollen to grain. Prior US trials under managed stress showed an 8.5% yield increase in FNP hybrids, but at much higher yield levels than typical in SSA. The research question is whether FNP hybrids deliver consistent, agronomically and economically meaningful yield advantages across diverse African hybrid backgrounds, environments, and especially low-input, drought-prone smallholder conditions, and whether farmers prefer these hybrids.

The paper situates the work within efforts to modernize breeding and seed systems in SSA, noting successes with stress-tolerant varieties and their poverty-reduction impacts. It reviews hybrid seed production constraints in SSA (manual detasseling, seed purity, reliance on three-way crosses), and prior SPT systems: recessive ms45-based and dominant Ms44-based systems, the latter enabling production of FNP hybrids that increased yield under ultra-low N in US trials. An alternative dominant male sterility system using ZmMs7 has been developed, though field validation of yield benefits in FNP form had not yet been reported. The authors also discuss mixed success translating biotechnology from controlled environments to on-farm settings, and the importance of participatory, on-farm testing due to high spatial variability and low input use in SSA.

Germplasm and Ms44 incorporation: The dominant male-sterile allele Ms44 was introgressed into five inbred maize lines (four lines initially with 4–6 backcrosses; later all with ≥5 backcrosses). Ms44 must be maintained heterozygous; progeny segregate 1:1 male-sterile:fertile. Converted Ms44 inbreds were used as female parents and crossed to 3–4 male inbreds each, generating 18 single-cross hybrid pairs. Female rows segregated into pollen-producing (PP) and non-pollen-producing (NPP) plants, which were tagged separately at flowering. Ears from NPP plants produced F1 seed segregating 1:1 PP:NPP (FNP hybrids), while ears from PP plants produced 100% PP near-isogenic controls. Eight three-way crosses were produced by crossing F1 seed from NPP plants to PP inbred males, yielding segregating FNP three-way hybrids plus PP controls.

Yield testing: From 2017–2019, on-station trials (OST) and researcher-managed on-farm trials (OFT) were conducted across Kenya, South Africa, and Zimbabwe. OST environments included optimal, low-N (fields depleted of N for ≥4 seasons), heat (high temperatures at reproductive stage via delayed planting), and managed drought (dry-season planting, irrigation withheld ~2 weeks prior to mid-anthesis). Optimal, heat, and drought sites received recommended fertilization and pest/weed control; optimal sites were irrigated as needed. Rescue irrigation was applied only to avoid total crop loss where required. Trials used randomized complete block designs with split-plot restriction: hybrid background as main plot and trait (PP vs FNP) as subplot. Plots were 2–4 rows, 5 m length, 0.75 m row spacing; 4–6 reps per location. In selected OST locations, tagged PP and NPP plants within FNP plots were phenotyped: tassel traits (branch number, dried tassel weight, sampled at full shed), plant/ear height (2–3 weeks after flowering). At Zimbabwe OST sites, dehusked ears from tagged plants were photographed on a standardized setup and analyzed to estimate ear length, kernel number, 100-kernel weight, and grain weight per plant.

On-farm trials: Smallholder farmers, identified via extension agents, implemented 2–4 row plots (5 m rows, 0.75 m spacing), double-planted then thinned to 25 cm intra-row spacing, with two reps per location. Farmers were asked to manage pests and weeds but not to apply nitrogen fertilizer; target yields were <4 t ha^-1. Harvesting was by hand; ears were shelled and grain weight and moisture recorded; yields were adjusted to 155 g kg^-1 moisture.

Statistical analysis: Analyses were performed in ASReml. For grain yield, trait (PP vs FNP) and location were fixed effects; hybrid background and trait×hybrid interaction were random; blocking factors (replicates) were random. Trait effects within hybrids were predicted using BLUPs; across-hybrid trait effects were BLUEs. Differences between 100% PP and FNP were assessed with two-sided t-tests using SED from the linear mixed model, with significance at 5%.

Farmer evaluations: Conducted at eight Kenyan OFT sites in 2017 and six in 2018; evaluations were double-blind at mid-season and end-season. A total of 2697 farmers (62% women) participated. Participants rated importance of selection criteria (0–3 scale) and evaluated entries on a 5-point hedonic scale (A=5 to E=1) for traits including tassel formation, pollen shed (subset), ear size, and yield, plus overall evaluation. In 2017, participants were randomized to control or treatments including tassel/pollen criteria; in 2018, a single treatment group included tassel and pollen criteria. Scores for FNP vs PP were compared using pairwise t-tests.

Impact assessment: Maize area and production data from FAOSTAT (2018) for 50 SSA countries were combined with literature-based adoption rates of improved varieties and hybrids for the top 25 maize-producing countries (36.6 Mha total area; 12.6 Mha in hybrids). Yield benefit per country was estimated from the regression Δy = 0.006x + 180.2 (x = location mean yield). Two adoption scenarios for FNP were modeled: 10% and 25% of current hybrid area. Economic evaluation used NPV, IRR, and BCR with historical and projected annual project costs and benefits from increased production, discounting benefits to 2040.

• Across 112 locations in Kenya, South Africa, and Zimbabwe (2016–2019), FNP hybrids had a mean grain yield advantage of 202 kg ha^-1 over PP controls at equivalent grain moisture (5.2% increase; P<0.0001; N=4585 plot-level observations in Table 1). FNP out-yielded PP in 75% of locations tested.
• The yield gain was consistent across yield levels: projected +192 kg ha^-1 (+9.6%) at low-potential environments (~2000 kg ha^-1 mean) and +229 kg ha^-1 (+2.4%) at high-potential (~8000 kg ha^-1), with a relatively constant absolute gain.
• Consistency across genetic backgrounds: For 19 hybrids tested in ≥20 locations, mean FNP advantage was 178 kg ha^-1 for single crosses and 264 kg ha^-1 for three-way crosses.
• Yield component and morphology changes (N measured on tagged plants within FNP plots at OST): • Kernel number per plant: +5.9% (PP 281.3±3.7 vs FNP 297.9±3.7; P<0.001; N=464). • 100-kernel weight: +0.9% (31.7±0.34 g vs 32.0±0.34 g; P<0.01; N=469). • Ear length: +4.9% (13.2±0.18 cm vs 13.9±0.18 cm; P<0.0001; N=469). • Plant height: −3.8% (1.93±0.01 m vs 1.86±0.01 m; P<0.0001; N=469). • Ear height: −1.1% (1.01±0.01 m vs 1.00±0.01 m; P<0.01; N=471). • Tassel branches: −8.5% (16.7±0.18 vs 15.3±0.18; P<0.0001; N=475) and tassel weight: −6.7% (4.00±0.09 g vs 3.73±0.09 g; P<0.001; N=475), indicating smaller tassels.
• Farmer participatory evaluations in Kenya (2697 participants; mid- and end-season in 2017–2018): Farmers could distinguish FNP vs PP mid-season via tassel/pollen traits but consistently rated FNP higher for yield and overall evaluation at both mid- and end-season. For example, 2017 end-season yield scores: PP 3.49±0.014 (N=5839) vs FNP 3.62±0.014 (P<0.0001); overall evaluation: PP 3.49±0.014 (N=5807) vs FNP 3.61±0.014 (P<0.0001). Similar significant advantages for FNP in 2018.
• Impact assessment: Assuming 10% adoption among hybrid users (≈1.26 Mha), additional production estimated at 244,204 t yr^-1 valued at ~$40M; discounted benefits to 2040 ~$180M vs discounted costs ~$28.9M, yielding NPV ~$152M, BCR 6.25, IRR 24%. Under 25% adoption, additional production ~610,511 t yr^-1 valued at ~$100M; discounted benefits ~$452M, NPV ~$423M, BCR 16, IRR 32%.

The findings demonstrate that deploying the Ms44-based FNP trait in African hybrid germplasm and testing under farmer-managed, low-input conditions yields stable and meaningful yield gains (~200 kg ha^-1, often 10–20% under severe stress). This magnitude equates to at least six years of expected breeding progress under drought/low-N in SSA, indicating substantial agronomic significance. The consistent advantage across diverse hybrids and environments supports broad deployability across agroecologies. Mechanistically, reduced tassel development and absence of pollen in 50% of plants reallocate assimilates and nitrogen from tassel/pollen to ears, shortening the anthesis-silking interval under stress and increasing kernel number and grain weight without increasing total N uptake, thereby improving nitrogen use efficiency. Farmer participatory evaluations confirmed acceptance and preference for FNP hybrids based on perceived yield and ear traits, mitigating concerns about visible tassel differences. Economically, even conservative adoption scenarios generate benefits that substantially exceed R&D costs, with strong NPV, IRR, and BCR. Seed companies also benefit from eliminated detasseling, improved seed purity (no selfing on female rows), and anticipated higher seed yield due to increased kernel number on NPP female parents, potentially catalyzing faster replacement of outdated hybrids with climate-smart, higher-yielding products. Overall, the results support FNP as a practical, scalable technology to enhance productivity and input-use efficiency for smallholder maize systems in SSA.

This study provides robust on-farm and on-station evidence across three countries and 26 hybrid backgrounds that Ms44-based 50% non-pollen-producing hybrids deliver a consistent grain yield advantage (~200 kg ha^-1) under low-input, stress-prone smallholder conditions, driven by improved partitioning to ears and increased kernel numbers. Farmer evaluations indicate clear acceptance and preference for FNP hybrids, and economic analyses project strong returns under plausible adoption scenarios. The technology also offers operational advantages for seed companies, potentially accelerating the availability and turnover of modern hybrids in SSA. Future work should: (1) expand multi-country participatory evaluations to additional agroecologies; (2) quantify seed production gains (e.g., kernel number increases) under favorable seed production conditions in African germplasm; (3) monitor adoption dynamics and farmer perceptions post-deployment; and (4) assess complementary dominant male sterility systems (e.g., ZmMs7) through field validation.

• Adoption rates are currently unknown; projections are scenario-based and assume constant maize production.
• Kernel number increases for NPP plants were measured under low-N research conditions; effects under favorable seed production conditions in African germplasm were not yet measured within this study.
• Seven hybrid backgrounds with ≤12 locations were excluded from some consistency analyses, potentially limiting generalization to those specific hybrids.
• On-farm trials were researcher-managed with instructions to avoid N fertilizer; while representative of low-input contexts, results may differ under varying farmer management or higher-input systems.
• Field validation of similar dominant male sterility systems (e.g., ZmMs7-based) for yield benefits in FNP form was not reported here.