**4. Discussion and Conclusions**

In this study, we successfully mutagenized an exotic elite tropical maize inbred by EMS treatment on pollen. We generated a 2050-family strong M2 population and identified numerous mutants with visible phenotypes. Many mutants identified in our population sugges<sup>t</sup> the usefulness of this population as a complement to the B73 mutant collection. For example, B73 has yellow seeds while ML10 has orange seeds, and we identified several mutants that change the orange color of ML10 seeds to yellow and to white (Figure 1G–H). These mutants will be useful in identifying the natural variation that boosts provitamin A content in maize seeds [50,51]. Despite extensive mutagenesis screens, mutants in *CLE7* had not been found in any EMS populations of standard inbreds, but were identified in our population. Since our allele was created by EMS mutagenesis in an elite parent, it can directly be used in a breeding program to improve kernel row number (KRN) without transgenic regulation as compared to the CRISPR-Cas9 alleles. Our results sugges<sup>t</sup> that there was a genetic suppressor(s) of the *cle7* phenotype in BL10, the paternal parent in the elite hybrid paired with maternal ML10. Combining this suppressor and *cle7* in ML10/BL10 hybrids could lead to an improved hybrid with a higher KRN and yield without being fasciated. Mapping of this suppressor could also identify a new factor controlling meristem proliferation. The presence of a suppressor of the *cle7* mutant phenotype in the BL10 background, but not in B73 emphasizes the value of using diversified non-reference genotypes in genetic studies, allowing the study of natural suppressors and enhancers.

Our mapping-by-sequencing method is similar to MutMap in rice [27]. Like Mutmap, our mapping procedure does not require crossing to an unrelated inbred, which makes scoring of phenotypes easier and more accurate. First introduced in rice with a genome size of 370 Mb, MutMap was successfully used to map several mutants using fewer than 2300 segregating EMS-induced variants. For maize with an 8× bigger genome and many more repetitive sequences, it was questionable whether this method would work, in particular when using exotic germplasm for which no reference genome is available. To our knowledge, our study is the second example of MutMap in maize where EMS-derived variants were used for mapping without the need to cross to an unrelated inbred. The first example was from [28], where pools of nine F2 mutants and of nine wild-type siblings were sequenced, as well as the unmutagenized parent of unknown genetic background. Mutants with subtle phenotypes or ones whose phenotype is a ffected by residual heterosis, which is very common in maize, will benefit most from this approach. Our bioinformatic pipeline quickly identified the chromosomal region carrying the mutation based on only 1118 segregating variants spread largely evenly across the 10 chromosomes and pools of 69 wild-type and 72 mutant F2 plants. The number of variants used in our mapping was two-fold less than in MutMap in rice, which is due to our more stringent variant filtering to eliminate the potential false variant callings caused by the complex structure of the maize genome and the low-frequency EMS-induced marker mutations introduced only via individual F1 plants. A low number of segregating variants can be a disadvantage because it will make peak calling di fficult and limit the resolution of mapping; at the same time, it has the potential advantage of having only a few putative SNPs in the identified region to pinpoint the causal mutation. Our success in mapping by Mutmap in a non-standard maize inbred, where no prior genetic studies had been performed, will facilitate the adoption of this method to allow the rapid cloning of maize genes in diversified germplasm.

As mentioned above, our analysis used only variants between ML10 and the M5 mutant. In principle, we could have used more variants from the more heavily mutagenized E-ML10 line; however, genomic complexity in the non-reference ML10 genome, the heterozygosity of E-ML10 for many of the induced mutations and the need to collect F2 samples from multiple F1 ears prevented us from doing so. In the future, a modification to the protocol will be to self E-ML10 several times to reduce heterozygosity of the induced marker mutations before crossing to the mutant of interest and to use plants from only one single selfed F1 plant for mapping. In addition, sequencing of the very parents from which the F1 plant(s) were derived (rather than sister plants from the parental generation) would be advisable.

Many mutations (for example weak *fea* mutants a ffecting kernel row number) in maize in B73 proved to be very di fficult to map due to scoring F2 phenotypes in mixed genetic backgrounds (unpublished data from David Jackson). Based on our experience with mapping E1-9, we also generated a heavily mutagenized B73 line (the E-B73), and are now in the process of selfing it for several generations to fix the EMS-induced marker SNPs; this should result in an essentially isogenic second parent carrying a large number of SNP markers for mapping, which should be very useful to the maize research community.

In conclusion, we have developed an EMS population in an elite tropical inbred and found numerous interesting mutant phenotypes for forward genetic studies. We successfully identified a mutation in *CLE7* underlying ear fasciation by mapping-by-sequencing (MutMap). The EMS mutagenesis and rapid mapping-by-sequencing pipeline described here may encourage maize researchers to perform forward genetics in their maize germplasm of choice, which can lead to novel gene discovery due to diversified genetic backgrounds.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/11/3/281/s1, Table S1: Primers used in this study, Figure S1: Nucleotide alignment of the CLE7 locus among reference B73, unmutagenized ML10 and E1-9 mutant, Supplementary protein alignment files for FEA2, FEA3, FEA4, TD1, CT2, CLE7 and GB1.

**Author Contributions:** Conceptualization: S.L.V., M.L., C.K., T.D.K., K.H.T., and; Data analysis: C.K. and S.L.V.; Investigation and Methodology: Q.H.T., N.H.B., N.T.N.D., L.T.N., T.T.T. and S.L.V.; Writing – original draft: C.K. and S.L.V.; Writing – review & editing: S.L.V., M.L., T.D.K. and K.H.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project was funded by The National Foundation for Science and Technology Development (NAFOSTED Vietnam) under gran<sup>t</sup> number 106-NN.01-2014.48.

**Acknowledgments:** We thank Tran Thi Thu Nga, Kieu Thi Dung for their support on fieldwork. We thank David Jackson and Liu Lei for their advice on bioinformatics, mapping and comments on the manuscript. All authors have read and agreed to the published version of the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.
