What if the fastest way to improve maize haploid inducers is to make doubled haploid versions of the inducers themselves? This study turns that idea into a large-scale breeding pipeline—and finds offspring that outperform their parents. Haploid inducers are key components of doubled haploid technology in maize, and robust agronomic performance and better haploid induction ability are persistently sought through genetic improvement.

The study developed C1-I inducers that enable large-scale in vivo haploid induction of inducers and discovered superior inducers from the doubled haploid progenies. The haploid induction rate of C1-I inducers ranged between 5.8 percent and 12.0 percent, while the average success rate of doubled haploid production across 23 inducer crosses was 13 percent.

Doubled haploid inducers showed transgressive segregants for days to flowering, haploid induction rate, plant height, and total primary branch length, and the best haploid induction rate exceeded 23 percent. The study demonstrated that recombinant doubled haploid inducers can be accessed on a large scale and used to facilitate genetic improvement of maternal haploid inducers by in vivo doubled haploid technology.

Successful implementation of doubled haploid technology in maize breeding programs depends on the ability to produce haploids. Haploids are produced by using pollen from haploid inducers to pollinate the source germplasm from which doubled haploid lines will be developed.

The proportion of seeds with haploid embryos among the total seeds harvested on the source germplasm is called the haploid induction rate, or HIR. Over the past 15 to 20 years, advanced inducers with high haploid induction rates of 8 to 10 percent have enabled maize doubled haploid production at large scale.

The study set out to verify that maternal inducer haploids can be produced by in vivo haploid induction using a C1-I inducer. It also aimed to demonstrate that doubled haploid inducer line development is feasible and determine the relationship between traits of inducer F1s.

Figure one outlines the three-stage pipeline for developing doubled haploid inducer lines. The authors first cross genome donors with C1-I haploid inducers, then identify putative maternal haploids by the purple embryo marker R1-nj. These haploids enter a nursery, receive colchicine treatment for genome doubling to produce DH one plants, and are self-pollinated to generate DH lines.

The DHI production pipeline has three stages. First, heterozygous inducer F1s used as donor parents are induced by a C1-I inducer for haploid induction. Second, putative haploid kernels expressing purple color in the embryo are visually sorted.

Third, haploid kernels are sown in the greenhouse, and haploid seedlings are injected with colchicine solution for genome doubling at the three-leaf stage. Colchicine-treated haploids are transplanted in the field and self-pollinated to obtain DHI lines; the success rate is the percentage of transplanted haploids with seed sets leading to DHI lines.

Eight inducers were used to develop an elite-by-elite doubled haploid inducer population, and 28 inducer F1s were created from those eight inducers by a half-diallel mating design. The inducer F1s were evaluated for haploid induction ability and agronomic performance, then crossed with C1-I inducers to obtain inducer haploids.

The inducer haploids were subjected to the regular doubled haploid protocol in maize to obtain doubled haploid inducer lines. Phenotypes of inducer parents and doubled haploid inducer lines were determined in the summers of 2021 and 2022, while C1-I inducers and inducer F1s were measured only in the summer of 2020.

Plant materials were grown in 3.8-meter plots using a randomized complete block design in two planting blocks at the Iowa State University Agricultural Engineering and Agronomy Farm. A total of 537 genotypes were planted in the field for phenotypic measurement in 2021 and 2022.

Forty-two doubled haploid inducer genotypes and nine parent founders were common across the two years, with trials grown in loamy soil under rainfed conditions. Leaf tissue from three plants per entry was harvested and genotyped using DArTseq technology after the samples were shipped to CIMMYT for DNA extraction and single-nucleotide polymorphism genotyping.

A total of 88,421 unimputed SNPs per line were successfully called and reported. After imputation and filtering minor allele frequencies below 5 percent, 6,636 SNPs across the genome remained for calculating parental genome contribution and fixation index.

Table two reports the least-squares mean haploid induction rate for five C-one-I inducers, ranging from five point eight percent for genotype C to twelve point zero percent for genotype B. The ninety-five percent confidence intervals show the uncertainty around each estimate, while superscript letters indicate which means differ significantly by Tukey’s test at the five percent level.

The authors note that genotypes A, B, and D exceeded eight percent, a level supporting large-scale doubled-haploid induction. The least-squares means of haploid induction rates for the five C1-I inducers ranged from 5.8 percent to 12.0 percent, and haploid induction rates differed significantly between C1-I inducers.

The inducer parents used in the study carried the red root phenotypic marker for haploid sorting and included BHI306, Mo17-derived, A637-derived, and B84-derived inducers. Kernel pigmentation after haploid induction by C1-I inducers depended on the inducer F1 genotypes.

Figure two links kernel-level haploid identification with field performance. Panels A through E show anthocyanin pigmentation after induction: light-blue arrows mark clearly pigmented haploids, while orange arrows identify suspicious purple-red kernels, with pigmentation patterns varying among the inducer crosses and checks.

Panels F and G then show a large inducer nursery and DHI lines growing in field trials, illustrating how this screening pipeline supports doubled-haploid production and agronomic evaluation. Applying doubled haploid technology for inducers within two seasons produced 100 percent homozygous doubled haploid inducer populations with uniform agronomic performance.

The haploid induction rates of C1-I inducers A, B, and D exceeded 8 percent, enabling large-scale doubled haploid inducer production. Among inducer haploid seedlings treated with colchicine, 22 percent were shedding pollen, and the average success rate based on initially transplanted inducer haploids was 13 percent.

The anthesis–silking interval was negatively correlated with success rate, while days to flowering was positively correlated with success rate. There was no evidence that haploid induction rates and primary branch length were correlated with success rate in doubled haploid inducer production.

Figure three relates DH production success rate to four inducer F1 traits: ASI, days to flowering, HIR, and total primary branch length. The fitted relationships are summarized by correlation coefficients: negative for ASI, and positive for the other three traits, although the reported p-values indicate varying statistical support.

This matters because the authors report success rates from three point eight to twenty-one point five percent, with thirteen point three percent on average, helping identify traits associated with producing DHI lines. The ranges of maximum and minimum performance for the four traits were wider in doubled haploid inducer progenies than in their parents, indicating transgressive segregation.

The mean differences between inferior and superior groups of doubled haploid inducers were significant for all four traits. Figure four shows the distributions of four traits across the doubled haploid inducer population: days to flowering, haploid induction rate, plant height, and total primary branch length.

Green histograms show the progeny values, while dashed lines mark the eight elite inducer parents and solid lines mark the traditional inducer. The wider progeny ranges, including values beyond the parental performances, illustrate transgressive segregation and support the authors’ conclusion that elite-by-elite crosses can generate improved inducer lines.

The best doubled haploid inducer progenies from the PHG83-derived by BHI306 and PHG83-derived by Mo17-derived crosses had average haploid induction rates above 13.9 percent and outperformed their parental lines with induction rates exceeding 20 percent. The majority of parental genome contributions to doubled haploid inducers ranged from 0.40 to 0.55.

Elite-by-elite inducer crosses therefore appear promising for deriving even better-performing doubled haploid inducers. Figure five shows a histogram of parental genome contribution in doubled haploid inducers derived from twenty-three biparental crosses. Most values cluster around roughly zero point four to zero point six, consistent with the reported first quantile of zero point four, mean and median of zero point four eight, and third quantile of zero point five five.

This matters because it indicates that most DHIs received approximately equal genomic contributions from both parents, supporting elite-by-elite inducer crosses as a way to generate improved lines. Figure six maps genome-wide SNP Fst values across chromosomes for four traits, comparing superior and inferior doubled haploid inducers in three nested association mapping populations.

The dashed line marks the ninety-ninth-quantile threshold, while crosses identify outlier SNPs; patterns differ among the B73-, BHI306-, and PHG83-derived populations. The authors use this population-specific variation to show that extraordinary inducer individuals carried different SNP genotypes associated with phenotypic differences.

The study shows that recombinant doubled haploid inducer lines can be produced at scale and can reveal superior haploid inducers, including lines with induction rates above twenty percent. That creates a practical route for improving the technology that makes doubled haploids.