Molecular evolution of a reproductive barrier in maize and related species

Abstract

Three cross-incompatibility loci each control a distinct reproductive barrier in both domesticated maize (Zea mays ssp. mays) and its wild teosinte relatives. These 3 loci, Teosinte crossing barrier1 (Tcb1), Gametophytic factor1 (Ga1), and Ga2, each play a key role in preventing hybridization between incompatible populations and are proposed to maintain the barrier between domesticated and wild subspecies. Each locus encodes both a silk-active and a matching pollen-active pectin methylesterase (PMEs). To investigate the diversity and molecular evolution of these gametophytic factor loci, we identified existing and improved models of the responsible genes in a new genome assembly of maize line P8860 that contains active versions of all 3 loci. We then examined 52 assembled genomes from 17 species to classify haplotype diversity and identify sites under diversifying selection during the evolution of these genes. We show that Ga2, the oldest of these 3 loci, was duplicated to form Ga1 at least 12 million years ago. Tcb1, the youngest locus, arose as a duplicate of Ga1 before or around the time of diversification of the Zea genus. We find evidence of positive selection during evolution of the functional genes at an active site in the pollen-expressed PME and predicted surface sites in both the silk- and pollen-expressed PMEs. The most common allele at the Ga1 locus is a conserved ga1 allele (ga1-Off), which is specific haplotype containing 3 full-length PME gene copies, all of which are noncoding due to conserved stop codons and are between 610 thousand and 1.5 million years old. We show that the ga1-Off allele is associated with and likely generates 24-nt siRNAs in developing pollen-producing tissue, and these siRNAs map to functional Ga1 alleles. In previously published crosses, the ga1-Off allele was associated with reduced function of the typically dominant functional alleles for the Ga1 and Tcb1 barriers. Taken together, this seems to be an example of an allele at a reproductive barrier locus being associated with an as yet undetermined mechanism capable of silencing the reproductive barrier.

Keywords: gametophytic factors; genome assembly; maize; molecular evolution; pollen; reproductive barrier; siRNA; silencing; silk; transmission ratio distortion.

Fig. 1.

Incompatibility between GA inactive pollen and GA active silk can generate a reproductive barrier between Zea mays populations (a) when a gametophytic factor (GA) gene is active in the silk, only pollen with a matching active GA gene can grow normally down the silk toward the female gametophyte (light green arrow), which impedes the chances of fertilization by inactive GA pollen (dark red line) and generates a reproductive barrier. b) Diagram based in part on microscopy published by Lu et al. (2014): When silk and pollen both have no GA gene activity or both have matching active GAs, the pollen tube grows quickly down a transmitting tract in the silk toward the ear. Each of the 3 silk GAs impacts inactive GA pollen tube morphology differently. Tcb1 silk PMEs are shown in green, Ga1 in blue, and Ga2 in purple. Tcb1 pollen PMEs, shown in dark green, indirectly or directly interacts with silk PMEs.

Fig. 2.

GA locus, allele, and gene nomenclature. Each GA barrier is controlled by a locus containing corresponding silk- and pollen-expressed genes. GA alleles can be categorized by activity of the PMEs, or factors, encoded by the genes in each locus. Here, we propose a new ga1-O allele, which is distinct from a fully inactive ga1 allele. Weakly active Tcb1, Ga1, and Ga2 barriers have been observed, but often under the control of alleles, which could be called strong in other genetic backgrounds. Alleles like these have sometimes been called Ga1-W and Ga2-W; to date, no Tcb1-W allele has been characterized.

Fig. 3.

GA gene homologs are present in domesticated maize, 3 teosinte subspecies, and 16 other related grasses. Presence of GA genes is indicated by circles for silk-expressed (k) and squares for pollen-expressed (p) genes. Species divergence times are from (Welker et al. 2020; Chen et al. 2022), and the species tree is from (Grass Phylogeny Working Group III 2025). Divergence of Ga1/Tcb1 from Ga2 is indicated by a blue star and divergence of Tcb1 from Ga1 is indicated by a green star. Gene divergence times are based on sequence dissimilarity and Ks, see Supplementary Table 2. Trees visualized with iTOL (Letunic and Bork 2021).

Fig. 4.

Haplotype diversity at each GA locus. Detailed view of haplotypes at the Ga1 (blue) and Tcb1 (green) loci on maize chromosome 4 and the Ga2 (purple) locus on maize chromosome 5. Each row of shapes represents a sequence of full-length gene copies on a haplotype from 5′ to 3′, where circles are silk genes and squares are pollen genes. Shared gene copy color represents shared gene copy sequence. Gene copies with stop codons are gray, and gene copies found in only 1 haplotype are white. Hashed lines between gene copies represent more than a kilobase of distance between copies. Location of the distal nonfunctional silk gene copy found in some Ga2 haplotypes is also marked on chromosome 5 in gray. Gene copies on a haplotype are all in the same direction, except where indicated by arrows. Known barrier activity is marked for inbred lines phenotyped for both silk (k) and pollen (p) activity, where bolded, colored, capital letters represent an active barrier (K) or ability to overcome the barrier (P). Haplotypes which seem similar to fully active haplotypes are grouped by brackets.

Fig. 5.

Signals of positive selection during the evolution of functional GA genes. GA genes display signals indicative of episodic positive selection (ω > 1) on gene tree branches subtending functional silk (a) and pollen (b) genes. Maximum likelihood trees were built in RAxML (Stamatakis 2014). Branches are labeled with lengths above in black and bootstrap values below in gray. Branches with significant evidence for positive selection are red, with additional P-value labels also in red on the (a) silk gene tree and (b) pollen gene tree. Trees visualized with iTOL (Letunic and Bork 2021).

Fig. 6.

Sites likely under positive selection mapped onto 3D model of protein structure models of Tcb1 silk and pollen proteins colors represent inferred active sites without evidence of positive selection (blue), inferred active site with evidence of positive selection (purple), other sites with evidence of positive selection (orange), and location of predicted PME–PMEI interaction surface (pink). Sites displaying signals of positive selection are all on the surface and are not residues predicted to participate in PME inhibition via PME–PMEI binding. Active site residues (blue) were inferred via alignment to validated residues from (Johansson et al. 2002). PME–PMEI interaction residues are inferred based on alignments to validated residues from (Di Matteo et al. 2005). Structures predicted with AlphaFold2 (Jumper et al. 2021) and visualized in Pymol (Schrödinger n.d.).

Fig. 7.

Zea mays premeiotic 0.4 mm anthers from maize inbred line plants homozygous for the ga1-O allele produce more unique 24-nt siRNA sequences targeting the GA silk gene sequences than inbred line plants homozygous for Ga1-S (dark blue) or Ga1-M (light blue) in anther RNAseq libraries. We observed no genotype-associated difference in the number of 24-nt siRNAs targeting GA pollen genes. Likewise, we observed no genotype-associated differences in nonanther tissues.