Chloroplast capture, a phenomenon that can occur through interspecific hybridization and introgression, has been frequently suggested as an explanation for cytonuclear discordance in plants. In theory, the captured donor chloroplasts may not cooperate with the recipient nuclear genome, especially chloroplast-targeted nuclear genes. However, relatively few studies have documented the mechanisms of cytonuclear coevolution and its potential species differentiation and possible functional differences in the face of chloroplast capture. To explore this crucial question, we chose the Aquilegia genus, which is known for having minimal sterility among the species, and we inferred that A. amurensis captured the plastome of A. parviflora based on cytonuclear discordance and gene flow between these two species. We focused on the introgression region and its differentiation with closely related species, especially its composition in a chloroplast capture scenario. We found that nuclear genes encoding cytonuclear enzyme complexes or organelle localized (CECs) were significantly enriched in the introgression regions, indicating that the CEC genes of chloroplast donor species were selectively retained and displaced the original CEC genes in chloroplast receptor species due to cytonuclear interactions during introgression. Notably, the intrinsic factor of cytonuclear compatibility may have a higher degree of evolutionary distance for the introgressed CEC genes between A. amurensis and A. parviflora. Introgression from A. parviflora promotes the differentiation of A. amurensis and A. japonica. Furthermore, we found that one of the overrepresented gene ontology terms in these introgressed genes was terpene synthase activity (GO: 0010333) in which more than one-third of the genes were CEC genes, showing that A. amurensis had similar release patterns for terpenes in flowers of A. parviflora when compared with A. japonica. Altogether, this study helps to clarify the mechanisms of cytonuclear coevolution, species differentiation and functional differences in face of chloroplast capture and highlights a critical role of chloroplast capture in adaptation.

The raw reads were quality-checked using FASTQC (Andrews, 2010). The available chloroplast genome (MH142265) of Semiaquilegia adoxoides and the nuclear reference genome of Aquilegia coerulae (v3.1) were downloaded from GenBank and Phytozome, respectively. Then, we used Burrows-Wheeler Aligner (BWA) (Li & Durbin, 2009) to map the genomic reads of each individual against the chloroplast and nuclear reference at the default setting. Then, the single nucleotide polymorphisms (SNPs) were detected using SAMtools’ mpileup command (Li et al., 2009; Li, 2011) at the default setting appropriate for diploid organisms. Raw SNPs were filtered using VCFtools, requiring a minimum depth (DP) of 3, minimum mapping quality (MQ) of 30, minor allele frequency (MAF) of 0.04, and maximum missing rate of 0.5, and all non-biallelic SNPs were filtered, which may be caused by sequencing. Additionally, each sample of Paraquilegia microphylla and Semiaquilegia adoxoides (SRX127267) was used as an outgroup for downstream analyses, and the approach of variant calling was the same as that above.