Preliminary Phenotypic and SNP-Based Molecular Characterization of Maize (Zea mays L.)-Mexicana (Zea mays SSP. Mexicana) Introgression Lines under Inbred Background of 48-2

Wild relatives possess potential genetic diversity 
for maize (Zea mays L.) improvement. 
Characterization of maize-mexicana introgression lines (ILs) is of great value to diversify the genetic base and 
improve the maize germplasm. Four maize-mexicana IL generations, i.e. BC1, BC2, BC3, 
and RIL, were constructed under the elite inbred background of 48-2, elite 
inbred line that is widely used in maize breeding in Southwestern China, and 
were phenotyped in different years and genotyped with 56110 SNPs. The results 
indicated that 48-2 had higher phenotypic performances than all the 
characterized ILs on most of the agronomic traits. Compared with other ILs, BC2 
individuals exhibited more similar performance to 48-2 on most traits and 
possessed the highest kernel ratio (66.5%). Population structure and principal 
component analysis indicated that BC3 individuals gathered closer to 48-2 and 
exhibited the lowest mexicana-introgression 
frequency (0.50%), while BC2 (29.06%) and RIL (18.52%) showed higher 
introgression frequency. The high level of genetic diversity observed in the 
maize-mexicana ILs demonstrated that Z. mays ssp. mexicana can serve as a 
potential source for the enrichment of maize germplasm.


Introduction
Besides being one of the most important cereals worldwide, maize (Zea mays L.) is also known as a model in plant breeding and the studies of genetics, evolution, and domestication [1]. Maize was originally domesticated from teosinte (Z. mays ssp. parviglumis) about 9000 years ago, and its genetic diversity has been greatly reduced because of natural and artificial selections during the domestication and the following breeding and improvement processes [2]. To meet the challenges of increasing demands including faster yields increase, better edible and commercial quality, and improved resistance to biotic and abiotic stresses, maize breeders and researchers tend to extend crosses to the wild relatives to introduce novel and potential alleles and diversify the genetic base of elite breeding materials [3].
In rice and barley, many studies have demonstrated that wild species are useful gene reservoirs for germplasm improvement and candidate allele mining [4] [5]. In the case of maize, alien introgression has been accomplished for improvement of kernel composition, yield and yield-related traits [6] [7] [8] [9].
The wild relatives of maize have also long been recognized for their remarkable potential for resistance to biotic and abiotic stresses [10] [11] [12] [13]. Most of these valuable traits such as strong growth vigor, high protein content in the kernel and notable immune ability or resistance to multiple fungal diseases can be found in highland teosinte, including Zea mays ssp. mexicana (mexicana) [7]. In addition, the same chromosome number of mexicana and maize makes transferring useful genes from mexicana into elite inbred lines to create excellent breeding materials an easy to practice, interesting, and fruitful strategy to improve the quality and tolerance of maize. Introgressive hybridization is considered a successful method to transfer such genes to the chromosomes of maize with minimal amounts of accompanying foreign chromatin [14]. This type of gene transfer is now facilitated by high through-put single nucleotide polymorphism (SNP) markers available in maize [15]. Indeed, the use of SNPs is a powerful approach to analyze the genetic diversity among introgression lines and the distribution of alien alleles on maize chromosomes. Systematic and integrated phenotypic and DNA-based molecular characterization can provide practical guidance for the selection and mining of potential individuals and alleles from the ILs [15] [16]. Up to now, comprehensive characterizations of maize-mexicana ILs were seldom documented. In the present study, a systematic phenotypic and intensive SNP-based fingerprinting of maize-mexicana ILs were performed to reveal the characteristics of these ILs, provide practical guidance for their application, and allele mining of these lines.

Plant Materials and Field Trials
In

Phenotypic Data Collection and Statistical Analysis
The morphological traits of plant height (cm) and ear height (cm) were collected from five continuous individuals beginning from the 3 rd plant of all individuals of each line. Ears were harvested at physiological maturity and naturally dried atabout 37˚C for ≥5 days as done by Liu et al. [8]. We then collected ear and kernel-related traits including ear length (cm), ear diameter (cm), cob diameter (cm), kernel row number, ear weight (g), cob weight (g), ear-kernel weight (g), 10-kernel length (cm), 10-kernel width (cm), 10-kernel thickness (cm), and 100-kernel weight (g). The kernel ratio (%) was calculated from ear weight and In order to reveal the overall phenotypic performances and statistical differ- Each SNP was re-checked manually to identify any errors in known homozygote and heterozygote genotypes [15]. A total of 38,751 (69.06%) SNPs showing less than 20% of missing data, less than 20% heterozygosity and minor allele frequency (MAF) greater than 5% was selected for further analysis [16].

Genetic Diversity and Population Structure
The prepared dataset was imported into Power Marker V3.25 for counting the no. of SNPs, and calculating the diversity parameters, including polymorphism information content (PIC), major allele frequency (MAF), gene diversity, and heterozygosity [17]. Population structure of these ILs was evaluated using the software STRUCTURE 2.2 [18]. A subset of 16

PCA and Alleles Detection
In addition, the same subset of SNPs used in STRUCTURE has been imported into Tassel 5 in order to conduct a principal component analysis (PCA) and visualize the genetic relationships between the ILs and the inbred 48-2 [20]. Finally, these 16,330 polymorphism SNPs were used to detect the alleles shared or unique between 48-2 and the ILs via the online tool of Venn (http://bioinformatics.psb.ugent.be/webtools/Venn/).  three years, some ILs of different generations, exhibited the maximum performances on some of the traits ( Figure 2). These results confirmed the phenomenon of linkage drag that wild relative introgression through pollination usually leads to negative phenotypic influences to agronomic traits. Besides, some individuals showing improved performances can also be screened out from the descendants of ILs. And among the ILs descendants from different generations including BC1, BC2, BC3, and RIL, ILs of BC1-BC3 exhibited relatively higher phenotypic performance on most of the traits during the years 2017 to 2019, while ILs of RIL presented a wider range on these traits.

Pairwise Comparisons of 48-2 and IL Generations
Pairwise T-test comparison procedure was used to reveal the differences between 48-2 and the IL generations for 14 traits (

Population Structure and Principal Component Analysis
Bayesian model-based clustering algorithm implemented in STRUCTURE was used to infer population structure of all tested ILs and it was run for the number of fixed subgroups k from 1 to 9. The ΔK value was calculated for each k. The analysis of ΔK line plot (Figure 4(a)) indicated that the optimal number of  subgroups is two and three as there are peaks at these k values. When k = 2, structure analysis revealed 100% of membership of reference individuals 48-2 ( Figure 4(b)). As expected, The ILs had a mixture of genetic ancestry from the inbred line 48-2 and mexicana (Figure 4(b)). The lowest frequency of introgression was observed on BC3 individuals with 0.50% of genetic ancestry from mexicana while BC2 and RIL individuals showed very high frequency of introgression with 29.06% and 18.52% genomic region donated by mexicana respectively (Figure 4(b)). Assignment at higher k value (k = 3) continued to indicate strong membership of reference individuals, low frequency of introgression on BC3 individuals and high frequency of introgression on BC2 and RIL (Figure 4(c)).
The result of the PCA showed clear separation of the ILs and good agreement with the result from Structure analysis ( Figure 5). The first two principal components (PC1 and PC2) clearly indicated that individuals from BC3 generation were the closest to the inbred line 48-2 ( Figure 5). The BC3 individuals were followed by individuals from BC1 which were relatively close to the inbred 48-2.
PC1 and PC2 accounted respectively 12.4% and 6.8% of the genetic variation in American Journal of Plant Sciences

Allele Dissection
The subset of polymorphic SNPs used for structure and PCA has been used to extendedly compare the background 48-2 with the ILs. The number of shared alleles and unique alleles separately belonged to the ILs is presented in Figure 6.   Among our ILs with mexicana as donor, the overall phenotypic performance of BC1, BC2, and BC3 were intermediate between mexicana and 48-2. This observation is in line with previous researches by Briggs et al. and Wang et al. [7] [26]. In addition, some ILs of RIL showed mexicana-like traits such as reduced kernel row numbers and size of kernels. This is consistent with previous reports suggesting that hybrids plants produced by pollinating mexicana with maize had some traits of the wild parent [7]. Besides the characterized traits, we also observed some distinctive phenotypes differed from 48-2, including fragile cob, premature senescence, kernels with twin embryos, tween ears, kernels encased by fruitcases, very small kernels, stay green, shortened internodes, and so on.

Discussion
In crop breeding activities, introgressing alleles from wild relatives to improve the parental lines mainly focused on quality and tolerance related traits, and the phenomenon of linkage drag usually causes negative effects to agronomic traits of introgressive descendants [27] [28]. Though wide range of phenotypic changes were observed among all the characterized agronomic traits of four IL generations in the present study, most of these changes exhibited decreasing trends ( Figure 2, Figure S1), similar to those reported in both maize and other crops [29] [30].
These negative changes, or the effects caused by linkage drag, of all the IL generations might not influence the mining of target alleles for interest traits, while for breeding utilization of these ILs, more and intensive efforts should be laid to tolerance or quality related traits, such as the content of protein, oil, starch, and potential tolerance to limited water and fertilizer supply, higher temperature, or insects attack or pathogens infection, rather than the agronomic traits. Additionally, by integrating the characterization of agronomic traits and intensive quality or tolerance identification of these ILs, candidate NILs with improved quality or tolerance and non-significantly changed agronomic traits might be screened from all these ILs, and serve as enhanced alternatives of 48-2 in the maize breeding in Southwest China.
From the intensive SNPs genotyping, we have found credible evidence of mexicana introgression into maize ILs. Our structure analysis clearly indicated that mexicana has made genomic contributions to maize ILs, suggesting that the introgressed segments in these lines maybe contains genes conferring some traits of agricultural importance. A similar suggestion was made by a recent study conducted by Yang et al. who estimated that about 10.7% of maize genomic regions which may have contributed to genetic improvement were introgressed from mexicana [24]. It was reported that crossing between maize and its wild relatives, i.e. Zea mays ssp. mexicana, provided potential alleles for resources improvement [24]. In the present study, 42 most rooted ILs between the cross of maize and mexicana were characterized. This sample size is somewhat small, while some distinctive traits and alleles or genetic variations were identified (Figure 2, Figure   6), which suggested that the maize-mexicana ILs in the present study possessed potential variations for the further screening. On the other hand, whether these introduced alleles from mexicana linked to those changed traits, or the loci conferring these traits remain ambiguous. Genome-wide association studies (GWAS) and linkage analysis, i.e. QTL mapping with the genotypic and trait data from inbred 48-2 and the ILs will be necessary to identify loci responsible for the distinctive traits and the phenotypic differences between maize inbred 48-2 and the ILs.

Conclusion
Introducing genetic variation from wild relatives is a potential way to widen the genetic bases of breeding lines in maize. in the present study, we characterized an introgression population between the cross of maize and mexicana, and the phenotypic performance of the maize-mexicana ILs indicated that these ILs inherited some distinctive alleles from mexicana. This was confirmed by the SNP-based molecular characterization of the generations by self-and backcross-pollinating.
Our results suggested that introgressing mexicana into maize germplasm to produce ILs might provide the breeder with a broad source of variation which is necessary for maize germplasm improvement.