Molecular Parallelism Underlies Convergent Highland Adaptation of Maize Landraces

Abstract

Convergent phenotypic evolution provides some of the strongest evidence for adaptation. However, the extent to which recurrent phenotypic adaptation has arisen via parallelism at the molecular level remains unresolved, as does the evolutionary origin of alleles underlying such adaptation. Here, we investigate genetic mechanisms of convergent highland adaptation in maize landrace populations and evaluate the genetic sources of recurrently selected alleles. Population branch excess statistics reveal substantial evidence of parallel adaptation at the level of individual single-nucleotide polymorphism (SNPs), genes, and pathways in four independent highland maize populations. The majority of convergently selected SNPs originated via migration from a single population, most likely in the Mesoamerican highlands, while standing variation introduced by ancient gene flow was also a contributor. Polygenic adaptation analyses of quantitative traits reveal that alleles affecting flowering time are significantly associated with elevation, indicating the flowering time pathway was targeted by highland adaptation. In addition, repeatedly selected genes were significantly enriched in the flowering time pathway, indicating their significance in adapting to highland conditions. Overall, our study system represents a promising model to study convergent evolution in plants with potential applications to crop adaptation across environmental gradients.

Keywords: convergent adaptation; flowering time; polygenic adaptation; population branch statistic.

Fig. 1.

Sampling locations and expansion route of maize landraces, modified from Ross-Ibarra and Piperno (2020). Domestication and expansion times for maize populations are from published articles (Piperno and Flannery 2001; Dickau et al. 2007; Piperno et al. 2009; Hall 2010; van Heerwaarden et al. 2011; Grobman et al. 2012; Hufford et al. 2012; da Fonseca et al. 2015; Bush et al. 2016; Stephen Athens et al. 2016; Kistler et al. 2018; Wang et al. 2020). The exact location of our samples is indicated in supplementary fig. S1, Supplementary Material online.

Fig. 2.

Patterns of parallel adaptation at two candidate loci. (A) Distribution of PBE values in a 2-Mb region around the gene PIF3.1. The branch length of inset trees was based on PBE values and indicates the difference between a selected and unselected SNP. (B) Distribution of PBE values in a 0.5-Mb region around the gene GRMZM2G078118 involved in jasmonic acid biosynthesis. (C) Barplot of the reference allele frequency of one SNP located in PIF3.1. (D) Barplot of the nonreference allele frequency of one SNP located in GRMZM2G078118.

Fig. 3.

Patterns of parallel adaptation across loci. (A) Distribution of codirectional and antidirectional SNPs in neutral and outlier SNP sets. n: neutral SNPs; s: common outliers; s1: a subset of common outliers with MAF between $0.3\hspace{0.17em}\text{and}\hspace{0.17em}0.5$. (B) Number of best fit models for genetic source of repeatedly selected SNPs. ind: independent de novo mutation; mig: migration; sv: standing variation; sv_source: standing variant with a source population. Abbreviations for populations: AN, Andes; GH, Guatemalan highlands; MH, Mexican highlands; US, Southwestern US highlands.

Fig. 4.

Polygenic adaptation along elevation in the landraces. (A) A linear regression between polygenic score for all 29 trait-environment combinations tested and elevation of origin. The lines for the seven traits that showed significant signals of polygenic adaptation are colored and all other traits are shown in gray. (B) Polygenic score for days to silk in Florida (2006) for all landraces is negatively correlated with elevation of origin.

Fig. 5.

Convergent adaptation in the flowering time pathway. (A) Intersection of flowering time outlier genes among four highland populations. (B) Distribution of mean PBE values of SNPs located within and in the 10-kb flanking regions of core flowering time genes (Dong et al. 2012) (the red line) against a background of all genes (the black bars). (C) Diagram showing selected genes in the flowering time network. Colored dots indicate the population(s) in which selection was detected.