**1. Introduction**

Maize is an important staple crop worldwide used for food, feed and industrial products including biofuels [1]. The demand for maize has steadily risen in past decades and is expected to continue to increase, with maize demand predicted to double by 2050 [2,3]. Maize is not only of agronomic importance; it has also been a very important model organism for genetic studies.

Forward genetics in maize has been very fruitful and there are several mutant populations in maize; however, most of these are in standard inbred backgrounds. Most induced mutants in maize were created by two methods: pollen mutagenesis with Ethyl Methanesulfonate (EMS) or crossing with maize lines that carry highly active transposons [4]. As a mutagen, EMS has the unique advantage that it can potentially create allelic series ranging from null, hypomorphic, hypermorphic to gain-of-function

alleles, which can be very useful for studying gene function and applied breeding. For introducing point mutations, treating pollen with EMS has been the preferred method, because it is much more efficient than treating maize seeds or the use of radiation [4,5]. Using pollen mutagenesis, Prof. Gerald Neuffer has created a large number of mutants that have been used by researchers globally till now [6]. Several additional mutagenesis populations have been created: a TILLING population in B73 comprising 750 M1 plants [7], an indexed population in B73 that has been exome-sequenced comprising 1086 M1 plants [8], and other populations in W22 [7] and in the Iodent inbred PH207 [9]. By transposon mutagenesis, several large mutation populations have been created for reverse genetics (reviewed in [10]): the rescueMu created in A188 × B73 (HiII hybrid) [11]; the MTM Mu in B73 [12]; TUSC (Trait Utility System for Corn) in mixed background developed by the Pioneer Hi-Breed Company, the Mu population developed by Biogemma, UniformMu in W22 [13,14], the Photosynthetic Mutant Hunt population [15], and the ChinaMu in B73 [16]. A γ-irradiation mutant population in B73 was also created [17]. Despite the tremendous usefulness of these populations, in the South Asia tropical area, for example in Vietnam, the temperate standard maize inbreds (e.g., B73, W22, Mo17) grow very poorly, are highly susceptible to leaf disease, and have large anthesis-silking intervals making them very difficult to conduct genetics with.

Diversifying the genetic resources can help identify novel genes, whose function is masked by the genetic background of the standard inbreds, and is essential for studying adaptive, abiotic and biotic-stress related traits. Tropical maize inbreds have proved to be very valuable germplasm resources for several important traits. For examples, two drought-tolerant genes were cloned from the tropical inbred CIMBL55 [18,19], a head smut resistance and other disease resistance genes were identified from a Thailand Suwan and other tropical maize varieties [20–23]. Many of the studies on exotic germplasm in maize exploit natural variation, for example using GWAS [24]. So far, no artificial mutant population has been made in a non-standard maize inbred to allow forward genetic studies directly on the exotic germplasms.

The lack of materials is not the only issue; to successfully conduct forward genetic studies, one must consider standing challenges in mutant mapping/map-based cloning. First, maize has a large and complex genome (~2300 Mb) with lots of repetitive and hypervariable regions [25,26]. As such, the maize genome is several times larger than that of the two model plant species *Arabidopsis thaliana* (~135 Mb) and rice (~430 Mb). Second, heterosis is so strong and prevalent in maize that many mutant phenotypes cannot be scored accurately in F2 mapping populations of the original mutant outcrossed with an unrelated inbred. Subtle, quantitative, and physiology-based phenotypes such as kernel row number, plant height, root length, flowering time, nitrogen use efficiency, abiotic and biotic-stress related phenotypes, etc. require an isogenic background to correctly classify phenotypes. Third, background modifiers from the second parent, which are common in maize, can interfere with phenotype scoring. Therefore, in many cases several rounds of backcrossing, each of which can show different modifying effects, and careful pedigree/segregation analysis are required to create good mapping populations. This process is time-consuming and sometimes impossible for complex traits and genetics.

To solve these difficulties, several approaches combining next generation sequencing and mapping populations generated in the same genetic background have been developed. These include crossing to the unmutagenized parent in the Mutmap method in rice [27] and in maize [28], or crossing to a normal looking plant of a different M2 family (the so called "evil twin" in Sorghum [29]) or even not crossing at all, but only sequencing M3 plants as in Mutmap + (rice) [30], or sequencing individual M2 plants combined with zygosity analysis via progeny in maize [9]. These approaches utilized only the segregating mutagen-induced variants between the two parents or siblings to map, and hence they all shared the limitation of relying on the low number of mutagen-induced variants. It is unclear in maize whether these so called modified versions of the Mutmap approach can robustly identify causal mutations given the predicted low number of mutagen-induced variants for such a big, complex genome, especially when using non-reference germplasm.

In this study, we sought to create a new mutant population resource and establish a simple and rapid mutant mapping approach to facilitate the forward genetic study in a non-standard maize variety. Using the EMS pollen-mutagenesis method, we have created a high-quality mutant population in an elite tropical inbred ML10, one of the inbred parents for a very popular maize hybrid in tropical South Asia. We have used a modified Mutmap pipeline in maize that is simple, cost-effective and does not require crossing to an unrelated inbred. Instead the mutant in question was crossed to a ML10 line that was mutagenized for two consecutive generations with the aim to introduce more segregating variants for mapping. Despite this, the mutation was mapped based solely on the EMS-induced mutations in the focal mutant relative to the unmutagenized background, suggesting that even simple backcrossing to the unmutagenized background should be sufficient. We have illustrated the usefulness of this population and our mapping-by-sequencing strategy by cloning a fasciation mutant and identifying a promoter deletion in *ZmCLE7*.
