**3. Results**

### *3.1. Optimization of EMS Protocol for the ML10 Inbred and Development of the Mutant Population*

ML10 is the maternal parent of a very popular hybrid, LVN10, in the South East Asian region, e.g., Vietnam, Laos and Cambodia [31]. The inbred was developed in the 1990s by phenotypic selection and compatibility testing. Due to its popularity, high adaptability, high combining ability, excellent seed germination and the ability to robustly shed non-clumpy pollen even under harsh environmental conditions, we chose ML10 as the inbred background for pollen mutagenesis.

We tested several EMS concentrations: 0.04%–0.06%–0.1%–0.185%. EMS treatment at 0.185% gave almost no seeds, while EMS treatment at 0.04, 0.06 and 0.1% gave seeds. M1 seeds from the 0.1% EMS experiment had a germination rate of approximately 50% compared to ML10 wild-type. As we wanted to make a population with as high a mutation density as possible, an EMS concentration of 0.1% was chosen to generate a large mutagenized population. Ears pollinated with 0.1% EMS-treated pollen contained on average 30 seeds/ear compared to approximately 220 seeds per ear in untreated plants (Figure 1A). This is also the concentration recommended for B73 in [5] and higher than the concentration used for B73 in [8] of 0.067%. We noticed that the weather condition during pollination is very important; in particular, partially sunny weather with temperatures from 28 to 32 ◦C and humidity from 50% to 70% seemed ideal, as otherwise the pollen go<sup>t</sup> clumpy very quickly after harvest and had reduced viability. For the tropical conditions in Ha Noi, Vietnam, the best time for pollination to ge<sup>t</sup> this optimum weather is the first two weeks of April for the Spring season and from October the 12th to November the 12th for the Autumn season (see Methods).

Using the optimized sowing time and EMS concentration of 0.1%, we mutagenized ML10 and obtained over 6000 M1 seeds. We sowed out M1 seeds, selfed the resulting M1 plants, and go<sup>t</sup> 2050 M2 families. 1185 M2 families were phenotyped to find mutants, and results are summarized in Table 2 and Figure 1B–I. Numerous mutants were found, many of which have phenotypes similar to classical mutants in maize like *sh2*, *su1*, *ramosa*, *fea*, *liguleless*, *dwarf*, etc., [6] demonstrating that our EMS mutagenesis generated a large number of induced mutations.


**Table 2.** Frequency of typical mutants observed among 1185 M2 families screened.

**Figure 1.** Typical M0 ears and mutant phenotypes observed in the M1, and later-generation mutagenized plants. (**A**) Typical M0 ears of ML10 that were pollinated with 0.1% EMS-treated pollen and a control self-pollinated ear. (**B–I**) Examples of mutant phenotypes observed. (**B**) A dominant M1 *liguleless* mutant with upright leaf angle (arrow) next to wild-type (wt) control. (**C**) Albino. (**D**) Tassel-seed (ear and tassel). (**E**) Fasciated ear. (**F**) Small ear with long shank. (**G**) White kernels; wt ears have orange kernels. (**H**) Yellow kernel mutant (mt) (kernels with asterisk \*). (**I**) Wrinkled kernels *sugary*.

### *3.2. Development of a Mapping-by-Sequencing Method via Mutmap to Map the E1-9 Mutant*

### 3.2.1. F2 Mapping Population Generated with a Heavily Mutagenized ML10 (E-ML10)

A challenge for mapping mutations in maize is that when crossed to an unrelated parent to generate an F2 mapping population, mutant phenotypes may vary due to residual heterosis and segregating genetic modifiers. To circumvent this problem, we developed a heavily mutagenized ML10, named the E-ML10 for "Evil-twin" or "Evil" ML10, which underwent two generations of EMS pollen mutagenesis, and used it as an alternative parent to generate F2 mapping populations (Figure 2B). We reasoned that this E-ML10 will contain many EMS-induced SNPs compared to the original ML10 that can be used as molecular markers, abundant enough for fine mapping. Once an F2 population was made between E-ML10, and the mutant and phenotype scoring were successful, we used a similar method for mapping-by-sequencing as MutMap in rice [27] or BSA-Seq/MutMap in maize [28].

As the first pilot experiment to test the possibility of mapping without the need to cross to an unrelated inbred, we mapped the mild fasciation mutant E1-9 identified in our screen. The E1-9 mutant had its ear tip mildly fasciated and the ear had a higher number of kernel rows than wild-type (Figure 2A). Fasciation phenotypes can be subtle and di fficult to score due to phenotypic suppression after crossing to an unrelated inbred; therefore, this phenotype was suitable to test the above mapping method. We crossed the homozygous M4 mutant to the E-ML10, and in the F2 derived from four selfed F1 plants we obtained 72 fasciated mutants from a total of 280 plants. This segregation ratio suggested that the fasciation phenotype in E1-9 was caused by a single recessive mutation. We sequenced pooled DNA from 72 fasciated mutants (F2-mutant pool) and 69 wild-type looking individuals (F2-WT pool) from this F2 mapping population. As the ML10 whole-genome sequence had not ye<sup>t</sup> been available, we also sequenced one original unmutagenized ML10 individual (ML10), and a pool of eleven M5 homozygous E1-9 mutants (E1-9) to aid variant calling (Figure 2B).
