Introduction

Genome sequences are frequently used for elucidating evolutionary processes, genetic diversity, interest-trait mapping, and phylogenic relationships1,2,3. In plants, the chloroplast, mitochondrion, and nucleus contribute genetic information to offspring4. The chloroplast genome is small in size, has single maternal inheritance, has a low nucleotide substitution rate, and has a highly conserved gene order and gene content in comparison to the nuclear genome5. Therefore, the chloroplast genome is considered an ideal model for investigating diversity and evolution.

Chloroplasts, which are unique to algae and plants, are cell organelles that are involved in photosynthesis, the process of converting light to energy6,7. Shinozaki et al. 8 first assembled the complete tobacco chloroplast genome. To date, more than 6000 complete chloroplast genomes have been recorded by the GenBank database (https://www.ncbi.nlm.nih.gov/genbank/). Most chloroplast genomes have a double-stranded, circular, typical quadripartite structure with a pair of inverted repeats (IRs) separated by a large single copy (LSC) region and a small single-copy (SCC) region9. Generally, chloroplast genomes have a DNA length of 120–160 Kb, about 79 protein-coding genes, 30 transfer RNAs (tRNAs), and four ribosomal RNAs (rRNAs)10. The variation in size of chloroplast genomes mainly originates from IR expansion/contraction or loss, which has played a vital role in evolution11,12. For example, the chloroplast genomes of some algae and legumes do not contain an IR region, resulting in a shorter chloroplast genome length4.

Spinach (Spinacia oleracea L.) is mostly a dioecious species belonging to the family Amaranthaceae13,14,15. This family is composed of approximately 175 genera and more than 2500 species and is distributed nearly worldwide. A number of species, including spinach, beets, and quinoa, are important food crops. Two wild spinach, species, S. turkestanica Iljin and S. tetrandra Stev., are distributed in different regions—S. turkestanica is located in Central and Southern Asia, while S. tetrandra occurs in the Middle East and Transcaucasia16,17. Previous investigations demonstrated that S. turkestanica was more closely related to the cultivated S. oleracea than S. tetrandra13,16,18. These wild species have been proven to be a valuable genetic resource for improving the quality of spinach varieties, including regarding disease and pest resistance19,20. However, the genetic resource collection of wild species is limited thus far, and consequently, the genetic structure of spinach germplasms is still largely unclear. Therefore, increasing the collection of valuable genetic sources and exploring the genetic diversity and phylogenetic relationships of spinach germplasms should help inform elite germplasm utilization and improve spinach breeding programs. Furthermore, chloroplast genome sequences in Spinacia have remained limited to date, as only the S. oleracea chloroplast genome has been reported21.

In this study, we reported three complete chloroplast genomes of S. tetrandra, S. turkestanica, and S. oleracea using next-generation sequencing technology. By comparing the three chloroplast genomes to each other and to previously reported chloroplast genomes from Amaranthaceae, we specifically aimed to (1) elucidate the genetic diversity and conservation of the chloroplast genomes in Spinacia; (2) identify and develop optimized markers for discriminating different spinach species; and (3) assess the phylogenetic relationships within Amaranthaceae using chloroplast genome sequences.

Results

Characterization of chloroplast genomes in Spinacia species

We generated a total of 400,673,634, 377,024,012, and 335,160,130 paired-end (150 bp) clean reads for S. oleracea, S. turkestanica, and S. tetrandra, respectively. Among these clean reads, 0.93% (3,763,621), 1.24% (4,712,787), and 1.51% (5,078,425) of the clean reads mapped on the reference chloroplast genome sequence of S. oleracea (GenBank Accession Number: NC_002202.1) were used for the de novo assembly of S. oleracea, S. turkestanica, and S. tetrandra, respectively. Finally, we obtained three complete chloroplast genomes for these Spinacia species. Detailed information on the chloroplast genomes of the three Spinacia species is provided in Table 1.

Table 1 Summary of the three Spinacia species chloroplast genomes.
Full size table

The chloroplast genomes of S. oleracea, S. turkestanica, and S. tetrandra shared a typical quadripartite structure containing two copies of an IR region that separate the LSC and SSC regions (Table 1 and Fig. 1). Specifically, the LSC region was from 82,525 to 82,733 bp, which was longer than the SSC (17,868–17,959 bp) and IR regions (25,073–25,098 bp). Furthermore, the overall GC content of the three Spinacia chloroplast genomes was approximately 36%. Interestingly, the IR region exhibited a higher GC content (~ 42%) than the LSC (~ 34%) and SSC regions (29%).

Figure 1
figure 1

The chloroplast genome maps of the three Spinacia species. Genes shown inside the circle are transcribed clockwise, and those outside the circle are transcribed counterclockwise. Genes belonging to different functional groups are color-coded.

Full size image

The three chloroplast genomes encompassed an identical set of 143 genes, including 98 protein-coding genes, 37 tRNA genes, and eight rRNA genes (Table 1). The gene content, order, and orientation in the three Spinacia species were similar (Fig. 1). All eight rRNA were located in the IRs, 23 tRNA genes existed in the single-copy region, and the others were found in the IRs. Among the 12 intron-containing genes, a total of nine genes (rps16, atpF, rpoC1, petB, petD, rpl1p6, ndhB, tRNA-Ile, and ndhA) had one intron while three genes (ycf3, clpP, and rps12) possessed two introns (Table S1). Significantly, rps12 was a trans-spliced gene with one of its exons (5’_end) located in the LSC regions while the other (3’_end) existed in the IR regions. The chloroplast genome sequences were deposited in GenBank (accession numbers: MZ516907 for S. oleracea, MZ516906 for S. turkestanica, and MZ569012 for S. tetrandra).

Shrinkage and expansion of the IR regions might account for the different sizes of the chloroplast genomes. Thus, we compared the IR border position and its adjacent genes between the three Spinacia chloroplast genomes (Fig. S1). The rps19, ndhF, ycf1, rpl2, and trnH-GUG genes were positioned at the junctions of the LSC/IRB, IRB/SSC, SSC/IRA, and IRA/LLC regions, respectively. Interestingly, the border position in S. oleracea was the same as that in S. turkestanica, which implied no IR expansion or contraction. Additionally, only two genes, ycf1 and rpl2, exhibited different boundary regions among the three Spinacia chloroplast genomes. Specifically, ycf1 was 4057 bp in the SSC regions of S. oleracea and S. turkestanica but 4063 bp in the SSC region of S. tetrandra. The gene rpl2 was 819 bp size in S. oleracea and S. turkestanica and 825 bp in S. tetrandra. Altogether, no significant shrinkage/expansion of the IR regions, especially between S. oleracea and S. turkestanica, was detected among the Spinacia species. Therefore, we propose that the shrinkage/expansion of the IR regions is not the main reason for the different sizes of the chloroplast genomes in the study.

Phylogenetic analysis and Spinacia chloroplast genome evolution

To explore the phylogenetic position of Spinacia and further clarify its evolutionary relationships with other species from the Amaranthaceae family, 14 Amaranthaceae species were selected for phylogenetic tree construction with Fagopyrum tataricum (Polygonaceae) as an outgroup (Table S2). We utilized different data, including the complete chloroplast genome, LSC, IR, SSC, and protein sequences to construct the phylogenetic trees (Fig. 2). All of the phylogenetic trees had the same topology with high bootstrap support. Specifically, the 14 species from the Amaranthaceae family grouped into one cluster, which was further divided into seven clusters corresponding to different genera. Furthermore, Amaranthus was placed in the most basal clade among the Amaranthaceae species, and then Beta, Chenopodium and Spinacia formed a sister clade to Haloxylon, Bienertia and Salicornia. Additionally, the three Spinacia species were more closely related to Chenopodium than B. vulgaris, with 100% bootstrap support values, which is consistent with a previous study using single-copy genes from the genome13,22. In the resulting phylogenies, we confirmed, based on chloroplast genome that S. turkestanica exhibited a closer relationship with S. oleracea than with S. tetrandra, and the chloroplast genomes were conserved, as the phylogenetic trees of the quadripartite structures shared the same topology.

Figure 2
figure 2

Phylogenetic trees of the 14 Amaranthaceae plants based on chloroplast genomes and the variation in their chloroplast genome size. The phylogenetic tree was constructed using the (a) complete chloroplast genome data, (b) LSC region, (c) IR region, (d) SSC region, and (e) protein sequences, with F. tataricum as an outgroup. Numbers near branches are bootstrap values.

Full size image

Variant and population genetic analyses across Spinacia species

To infer the variation between the cultivated and two wild progenitors, we aligned the complete chloroplast genome sequences of S. turkestanica and S. tetrandra against the S. oleracea chloroplast genome sequence. A total of three and 690 variants were identified in this comparison (Table 2). The three variants within S. oleracea and S. turkestanica were also detected between S. oleracea and S. tetrandra, and thus a total of 690 non-redundant variants, including 559 single nucleotide polymorphisms (SNPs) and 131 insertion-deletions (Indels), were obtained among S. oleracea, S. turkestanica, and S. tetrandra. Among the 690 non-redundant variants, the average variant density in the LSC and SSC regions was 6.3 variants per kb, whereas this value was only 0.9 variants per kb in the IR regions (Fig. 3a and Table 2), revealing that the IR regions were more conserved than the single-copy regions, which was further supported by the genetic distance between S. oleracea and its two wild relatives (Fig. 3b).

Table 2 Summary of the variation between S. oleracea and the two wild progenitors.
Full size table
Figure 3
figure 3

Variability of the three Spinacia species based on the chloroplast genome. (a) The purple lines in the innermost circle correspond to variants between S. turkestanica and S. oleracea. The green region in the outermost circle corresponds to variants between S. tetrandra and S. oleracea with a 1-Kb window and 100-bp steps. Genes represented by blue rectangles are on the positive strand while genes represented by red rectangles are on the negative strand. (b) Genetic distance between S. oleracea and the wild relative chloroplast assemblies of S. tetrandra (red) and S. turkestanica (blue). (c) The phylogenetic tree was constructed based on three variants between S. turkestanica and S. oleracea using 85 Spinacia accessions, including eight S. tetrandra, 16 S. turkestanica, and 61 S. oleracea individuals. Yellow rectangles indicate that the homozygous genotype in the accession was consistent with the reference chloroplast genome, whereas blue rectangles indicate that the homozygous genotype differed from the reference chloroplast genome.

Full size image

Furthermore, analysis of the distribution of the 690 variants revealed that 342 (49%) of them were situated in intergenic regions, 108 (15%) were located in intron regions, and 240 (34%) were positioned in coding regions (Table 2). Two hundred and thirty SNPs out of the 240 variants within the coding regions were further divided into 92 nonsynonymous and 138 synonymous SNPs, which were located in 42 genes (Fig. S2). Among these synonymous SNPs, 61 (44%) SNPs were enriched on the ndhF, psaA, rpoB, ropC2, rbcL, and ycf1 genes. Only three out of the 10 Indels within the coding regions were predicted to result in frameshifts of the matK and rpl22 genes (Fig. S2).

Interestingly, only three variants were identified between S. oleracea and S. turkestanica. To validate that the three variants specifically existed in S. oleracea, we constructed a phylogenetic tree based on the three variants using the 85 spinach accessions including eight S. tetrandra, 16 S. turkestanica, and 61 S. oleracea individuals13 (Fig. 3c). The result showed that almost all of S. oleracea and the two wild relatives could be separated into two distinct clades with different genotypes of the three variants, whereas there were still 13 S. oleracea individuals in the same clade as the wild progenitors (Fig. 3c). Furthermore, we detected 503 SNPs and 83 indels using the 85 spinach accessions. A total of 566 (96.58%) of these variants were shared in the 690 variants obtained by global sequence alignment of the three chloroplast genomes (Fig. 4a). Similarly, a neighbor-joining phylogeny based on these variants still showed that the 13 S. oleracea and S. turkestanica accessions clustered together (Fig. 4b). However, all of the S. oleracea accessions clustered together based on the 1,084,637 variants identified from spinach nuclear genome23 (Fig. 4c). Meanwhile, we noticed that the 13 S. oleracea accessions exactly include 29 variants from S. turkestanica (Fig. 4d). Of these 29 variants, more than 86% variants were existed in Sp155, Sp39, Sp40, and Sp61, while Sp35 encompassed the lowest number of variants from S. turkestanica (Fig. 4e), which was consistent with the phylogenetic tree (Fig. 4b).

Figure 4
figure 4

Population genetic analysis of the 85 Spinacia species. (a) Comparison of the variant data detected by the 82 resequencing accessions and the three complete chloroplast genomes. ‘Both’ indicates that the SNPs detected by 82 resequencing accessions were present in the SNP dataset from the three chloroplast genomes, whereas ‘Single’ indicates that the SNPs were specifically detected in the 82 resequencing accessions. A neighbor-joining tree of 85 Spinacia accession using the spinach variants’ (b) chloroplast genome and (c) nuclear genome. The individuals with asterisks are used to assembled in the study. (d) Variants in S. oleracea, S. turkestancia, and S. tetrandra. Gray lines indicate reference (S. oleracea) homozygous genotype, while red and black lines represent alternative homozygous and heterozygous genotype, respectively. The samples surrounded with blue dotted lines are the 13 S. oleracea clustered with S. turkestanica in (b). The red arrows mean the three conserved variants between S. oleracea and wild progenitors in Fig. 2c. Tet, S. tetrandra; Tuk, S. turkestanica; Ole, S. oleracea. (e) The percentage of variants (from wild or cultivar) in 29 variants existing in 13 S. oleracea clustered with S. turkestanica in Fig. D. Wild-V indicates the variants from wild species, while Cult-V represents the variants from cultivar species. (f) The inheritance patterns of the 13 S. oleracea.

Full size image

Importantly, chloroplasts of most species are inherited from the female parent24. Therefore, we propose that the 13 S. oleracea have cross-pollinated with S. turkestanica, with the latter having been the female parent, followed by continuously cross or/and backcross with paternal parents of S. oleracea (Fig. 4f). Consequently, the 13 S. oleracea accessions bear a similar chloroplast genome to S. turkestanica and a similar nuclear genome to S. oleracea. This provided an effective approach for us to create new spinach germplasm.

Repeat sequence analysis

A total of 46, 46, and 49 pairs of repeats (≥ 20 bp), termed long repeats in this study, were identified for S. oleracea, S. turkestanica, and S. tetrandra, respectively, using the program REPuter25 (Table S3). Specifically, 19, 19, and 20 forward repeats, two, two, and one reverse repeats, 25, 25, and 28 palindromic repeats, and no complementary repeats were obtained in S. oleracea, S. turkestanica, and S. tetrandra, respectively (Fig. 5a). The length distribution of the long repeat sequences was mainly 25–29 bp and rarely 40–45 bp among the three Spinacia species (Fig. 5b). Significantly, the long repeat sequences of S. oleracea and S. turkestanica were identical, with only three variants found between them (Fig. 3a). Importantly, none of the long repeats located in the three variants were identified. The majority of long repeats in the three Spinacia species were situated in intergenic regions, particularly, between rrn4.5 and rrn5 (Table S3). The ycf3 intron and ycf2 coding region also exhibited multiple nested long repeats (Table S3). Eleven pairs of polymorphic long-repeat sequences were identified, seven of which were S. tetrandra-specific repeats, and the remaining were S. oleracea/S. turkestanica-specific repeats (Fig. S3a).

Figure 5
figure 5

Repeat sequences in the chloroplast genomes of three Spinacia species. (a) Number of different long-repeat types and (b) length distribution in the Spinacia species. (c) Distribution of SSRs in the LSC, SSC, and IR regions, as well as (d) intron, coding, and intergenic regions. (e) Analysis of SSRs for different lengths of repeat (bp) and repeated sequences. (f) Comparison of variant (SNPs and Indels) numbers between the repeat regions and equivalently sized regions randomly selected from the chloroplast genome. The repeat regions indicate upstream and downstream 50-bp repeats. The genome represents the 50 regions randomly sampled from the chloroplast genome (including upstream and downstream 50-bp repeats). The P-repeat refers to polymorphic long-repeat regions, whereas N-repeat refers to non-polymorphic long-repeat regions. The different letters on the top of the boxplot indicate significant differences (Kruskal–Wallis test, P < 1.4e−15).

Full size image

Additionally, we also identified 59, 59, and 50 simple sequence repeats (SSRs) in S. oleracea, S. turkestanica, and S. tetrandra, respectively (Table S4). Similar to the observations in the long repeat sequences, S. oleracea shared the same SSRs as S. turkestanica, supporting the closer relationship between them. Among these SSRs, the vast majority were enriched in the LSC region, and S. tetrandra shared less SSRs than S. oleracea/S. turkestanica in the LSC and IR regions (Fig. 5c). These SSRs from the three Spinacia species were mainly located in intergenic regions (56–62%), followed by coding regions (25–30%), and only five and four SSRs were detected in the intron regions for S. oleracea/S. turkestanica and S. tetrandra, respectively (Fig. 5d). Single-nucleotide SSRs accounted 62.7%, 62.7%, and 54% of the total number of SSRs in S. oleracea, S. turkestanica, and S. tetrandra, respectively (Fig. 5e). There were multiple polymorphic SSRs between S. oleracea/S. turkestanica and S. tetrandra. For example, (A)13, (TAAA)3, and (TACA)4 SSRs were shared in S. tetrandra, but not in S. oleracea and S. turkestanica. In contrast, (AATT)3, (AAGTTT)3, and (TTAAAC)3 were specific to S. oleracea and S. turkestanica (Fig. 5e).

To explore the relationship between variants (SNPs and Indels) and these repeats among the three Spinacia chloroplast genomes, we compared the variation levels within the polymorphic, non-polymorphic long repeats, SSRs, and their flanking regions with equivalently sized regions randomly selected from the chloroplast genome, which revealed that the polymorphic long repeats, SSRs, and their flanking regions exhibited multiple variants, whereas the non-polymorphic long repeats and their flanking regions did not (Fig. 5f). Taken together, these findings suggested that the polymorphic long repeats and SSRs were highly variable and have been essential in spinach chloroplast genome evolution. Furthermore, variants could be regarded as important factors for generating polymorphic repeats. For example, the 11 polymorphic long repeats were generated due to the repeat sequences being destroyed by SNPs or Indels (Figs. S3b, c). A phylogenetic tree was constructed using 85 Spinacia accessions including eight S. tetrandra and 77 S. oleracea/S. turkestanica accessions based on the five variants identified from the polymorphic repeats (Fig. S3d). The phylogenetic tree showed that seven S. tetrandra and 75 S. oleracea/S. turkestanica accessions could obviously be divided into two clades based on the five variants, thus revealing that these variants or polymorphic repeats have excellent value for molecular marker investigation and spinach breeding programs.

Rates of synonymous and non-synonymous substitutions

Non-synonymous (KA), synonymous (KS) nucleotide substitution rates, and the ratio of KA/KS are widely defined as indicators of the selective pressures on genes during evolution. To examine the molecular evolution of chloroplast protein-coding genes in the Spinacia species, we calculated the KA, KS, and KA/KS values of 98 chloroplast protein-coding genes from S. turkestanica and S. tetrandra, with S. oleracea as a reference (Fig. 6). The KA and KS values of S. turkestanica were zero, strongly suggesting a closer relationship between S. turkestanica and S. oleracea, as only three variants (not in the coding region) existed (Table S5). However, the KA values of S. tetrandra ranged from 0.0000 to 0.0229 (rpl22), and the KS values ranged from 0.0000 to 0.0426 (psbT) (Figs. 6a, b). Otherwise, 64 (65%) and 55 (56%) genes did not show KA and KS rate changes in S. tetrandra, respectively. Among these genes, 43 genes shared zero values for both KA and KS, suggesting that these genes were highly conserved during Spinacia evolution (Fig. 6c and Table S5). Twenty-three out of 98 chloroplast protein-coding genes exhibited KA/KS values greater than 0, and only the rpl22 and ndhA genes, located in LSC and SSC, respectively, were under positive selection (Fig. 6d). The rpl22 and ndhA genes are related to photosynthesis and transcription or translation, respectively.

Figure 6
figure 6

Non-synonymous (KA), synonymous (KS), and KA/KS substitution values of S. tetrandra. (a) KA, (b) KS, and (d) KA/KS substitution values. (c) Venn diagram of protein-coding genes with KA = 0 and KS = 0. The genes highlighted in red were under positive selection during the Spinacia evolution.

Full size image

Discussion

Chloroplast genomes are known to be highly conserved in both gene order and gene content26. The three Spinacia chloroplast genomes shared good collinearity and the same numbers of genes (Table 1 and Fig. 1). Due to their conserved characteristic, chloroplast genomes are valuable for evolutionary analysis5,27. Earlier phylogenetic analyses utilized partial chloroplast sequences, such as LSC, SSC, IRs, or CDS sequences. However, these sequences might not provide complete information for each species, and thus the use of complete chloroplast genome sequences is regarded as a more effective approach for deciphering phylogenetic relationships, especially for closely related taxa6,7. In this study, we constructed phylogenetic trees of 14 Amaranthaceae species based on complete chloroplast genome sequences, LSC regions, IR regions, SSC regions, and protein sequences (Fig. 2 and Table S2). Remarkably, the phylogenetic trees based on the partial chloroplast sequences and complete chloroplast genome exhibited the same topology with high bootstrap support, strongly indicating the conserved nature of the chloroplast genomes of the 14 Amaranthaceae species and further demonstrating that the phylogenetic tree was highly reliable (Fig. 2). Furthermore, both the phylogenetic tree and genetic distances between S. oleracea and the two wild relatives revealed that S. turkestanica shared a closer relationship with S. oleracea than S. tetrandra did, which is consistent with previous studies13,16,23. To the best of our knowledge, this is the first comprehensive phylogenetic analysis of Amaranthaceae species using different information and is thus a valuable resource for deciphering the evolutionary relationships of Amaranthaceae.

In general, the IR regions are believed to be the most conserved of the chloroplast genome28,29,30. The distribution of the variants on the chloroplast genome and the genetic distances between S. oleracea and S. tetrandra confirmed that the IR regions were more conserved than the LSC/SSC regions (Figs. 3a, b and Table 2). In other respects, the expansion, shrinkage, or difference in boundary region size of LSC/IRs and SSC/IRs is a primary cause of differences in chloroplast genome size31,32,33. Expansions/shrinkage has been reported in many plants, such as in Chinese bayberry29, geranium34, and green alga35. Here, however, no significant IR length variation was detected among the Spinacia chloroplast genomes (Fig. S1), which is consistent with C. quinoa and C. album, which also belong to the family Amaranthaceae7. As a large number of variants, including Indels, were located in the LSC/SSC regions (Figs. 3a, b), we thus propose that the length of LSC/SSC was the main contributor to the change in Spinacia chloroplast genome size (Fig. S1).

Variations in repeat sequences in the chloroplast are considered important molecular markers and are widely used in plant breeding programs, population genetics, and the identification of commercial cultivars4,6. For example, Huang et al. (2015) used chloroplast microsatellite (cpSSR) markers to investigate the genetic relationships between domesticated jujube cultivars and wild relative populations36. We identified 46, 46, and 49 repeat sequences (≥ 20 bp) for S. oleracea, S. turkestanica, and S. tetrandra (Table S3). Only 11 polymorphic repeats were obtained between S. tetrandra and S. turkestanica/S. oleracea (both of which shared identical repeats) (Fig. S3a). Similar to the long repeats, both S. oleracea and S. turkestanica also exhibited identical SSRs (Table S4). Both the long repeats and SSRs were enriched in the intergenic region, which is consistent with previous investigations4,37. Gao et al.38 found that different types of repeats and their flanking regions could exhibit a high level of variation. Indeed, our study demonstrated that polymorphic long repeats and SSRs could accumulate more variants (Fig. 5f). Actually, variants including indels and SNPs accounted for the polymorphic long repeats or SSRs (Figs. S3b, c), and the variants within the polymorphic repeats could perfectly distinguish S. oleracea/S. turkestanica and S. tetrandra (Fig. S3d). Therefore, the repeats obtained in the study could be used as a valuable resource for studying differences in chloroplast genomes.

A large number of accessions are typically used to reveal the domestication history and evolutionary relationships in nuclear genome analyses, whereas this is seldom the case for the chloroplast genome2,39. Here, we used 85 Spinacia accessions, including S. tetrandra, S. turkestanica, and S. oleracea individuals, to obtain a comprehensive variant dataset containing 503 SNPs and 83 Indels. On the basis of the variants, remarkably, we found that 13 S. oleracea individuals might have crossed with S. turkestanica, with the latter acting as the maternal parent (Fig. 4f). Introgression of beneficial traits from wild relatives has contributed to improvement of cultivated crops40. For example, quite a few disease resistance loci in cultivated crops were originated from wild relatives, such as powdery mildew resistance loci in watermelon41, downy mildew resistance loci in sunflower42, and lettuce43. Thus, the 13 cultivated spinach varieties that originated through introgression from S. turkestanica (Fig. 4f), which is a valuable resource for introducing specific advantageous traits from wild progenitors into cultivated crops. We believed that, apart from the 13 S. oleracea accessions, there were still S. oleracea accessions crossed with S. turkestanica that we could not detect, as S. turkestanica served as the paternal parent (Fig. S4a). Importantly, this reminds us that breeders could contribute Spinacia genetic exchange and further increase spinach germplasm via reciprocal crossing between cultivated and two wild progenitors (Fig. S4b, c). These results will have a significant meaning for spinach breeding programs, and even provide a breeding reference for other crops.

Methods

Plant materials, DNA extraction and sequencing

We used S. tetrandra, S. turkestanica, and S. oleracea from 17S24, 17S31, and inbred line 10S15 for genome resequencing, respectively. Besides the three accessions above, 82 Spinacia accessions including seven S. tetrandra, 15 S. turkestanica, and 60 S. oleracea were used for population analysis13. The wild relatives were collected from the USA Department of Agriculture (https://npgsweb.ars-grin.gov/), while S. oleracea relatives came from the Institute of Vegetables and Flowers (IVF) of the Chinese Academy of Agricultural Sciences (CAAS), Beijing, China (Table S6). The materials used in the study were planted in the field at the IVF of the CAAS in spring 2019. The young leaves from each individual were collected and frozen in liquid nitrogen prior to DNA extraction. High-quality genomic DNA was extracted using a modified cetyltrimethylammonium bromide (CTAB) method44. The DNA quality and concentration were measured by electrophoresis on 1% agarose gels and a Thermo Scientific NanoDrop. Then, the Illumina genomic library was constructed and sequenced using a HiSeq 2500 Instrument (Illumina, San Diego, CA, USA) by BioMarker (Beijing, China). Finally, approximately 50 Gb of raw reads were generated for each individual with 150 bp paired-end read lengths. in other respects, the plant materials procured and used in the study comply with China’s guidelines and legislation. All the experiments were carried in accordance with national and international guidelines.

Chloroplast genome assembly and annotation

Low-quality reads and adapters were filtered using fastp (v0.20.0; parameters ‘-q 20’; https://githup.com/OpenGene/fastp#get-fastp). To identify paired-end reads belonging to the chloroplast genome, we mapped the clean reads against the reference chloroplast genome sequence of S. oleracea (GenBank accession number: NC_002202.1)21 using bowtie2 (v2.3.5.1; parameters ‘-q’)45. The de novo assembly was conducted on mapped paired-end reads using unicycler (v0.4.8) with the default parameters (–min_fasta_length 100–keep 1)46. Finally, the chloroplast sequence contigs were ordered and oriented based on the reference chloroplast genome sequences.

The online program GeSeq47 (https://chlorobox.mpimp-golm.mpg.de/geseq.html) was used to annotate the S. tetrandra, S. turkestanica, and S. oleracea genomes. Based on a comparison of homologs from previously reported chloroplast genomes of S. oleracea in the database and the chloroplast assembly in the study, the three annotation results, specifically the start and stop codons, were further corrected manually. The chloroplast genome maps were constructed using the OGDRAW program48.

Repeat structure and sequence analysis

REPuter25 (http://bibiserv.techfak.uni-bielefeld.de/reputer/) was used to find and analyze the sizes and locations of forward, reverse, palindromic, and complementary repeats with a minimal length of 20 bp and a sequence identity greater than 90%25.

SSRs (mono-, di-, tri-, tetra-, penta-, and hexanu-cleotide repeats) were identified using MISA (https://webblast.ipk-gatersleben.de/misa/), with thresholds for mono-, di-, tri-, tetra-, penta-, and hexane-nucleotide SSRs of 10, five, five, three, three, and three repeat units, respectively49.

Genetic distance analysis among the Spinacia species

To infer the genetic contributions of the two wild progenitors to the cultivated spinach, the genetic divergence between each S. tetrandra, S. turkestanica and S. oleracea was estimated using distmat (http://www.bioinformatics.nl/cgi-bin/emboss/distmat) with the Jukes–Cantor correction for each non-overlapping 1-kb window on the S. oleracea chloroplast genome.

Synonymous (KS) and non-synonymous (KA) substitution rate analysis

Based on the synteny alignments using D-GENIES, homolog genes were identified among the three Spinacia species. The yn00 program of PAML v4.9j50 was used to estimate KS, KA, and KA/KS values between S. oleracea and the two wild relatives. Boxplots were generated with the R software (v3.6.1). The Venn diagram was performed using BMK Cloud platform (http://www.biocloud.net/).

Phylogenetic analysis

The chloroplast genomes of 14 Amaranthaceae plants were selected for phylogenetic analysis (Table S2), and the Fagopyrum tataricum (Polygonaceae) chloroplast genome was used as the outgroup. The IR, LSC, SSC, protein, and complete chloroplast genome sequences were used for phylogenetic analysis using maximum likelihood (ML). First, multiple alignments were performed using MAFFT v7.158b software51. As for the protein sequences, 67 genes shared by the 15 species were identified using a custom Python script and then aligned using MAFFT v7.158b. The conserved aligned regions were extracted using Gblock (v0.91b) with parameters ‘–t = p, -b4 = 5, -b5 = h’ to obtain concatenated protein52. Finally, each alignment was used to build a maximum likelihood phylogeny using IQ-TREE (v2.1.2; parameters ‘-bb 1000’) with 1000 bootstrap replicates determined by IQ-TREE53.

Variant calling and population genetic analyses

First, two wild relative chloroplast genomes were aligned to the S. oleracea chloroplast genome using the Nucmer program54 with parameter “-c 80,” followed by the identification of one-to-one alignment blocks using delta-filter with the parameters “-r -q.” Finally, the SNPs and Indels were obtained using show-snp with the parameter “-Clr.” Furthermore, SNPs and Indels within the chloroplast genome and nuclear genome were also detected using large-scale resequencing accessions13. Illumina paired-end reads were processed to remove adapters and low-quality sequences using fastp v0.20.1 with the default parameters 55. Cleaned reads were mapped to the S. oleracea chloroplast genome21 using BWA-MEM v0.7.17-r1188 with default parameters56, and mapped reads were obtained using SAMtools v1.6–3 with the parameters “-F 12.” As for the variants identified from the nuclear genome, we mapped cleaned reads against the spinach genome23 using BWA-MEM v0.7.17-r118856. Variants were called using BCFtools v1.857 and filtered using VCFtools v0.1.1658 with the parameters “-maf 0.05 –minQ 30 –max-missing 0.9.” ANNOVAR59 was used to annotate the effects of the SNPs and Indels. A neighbor-joining phylogeny was constructed based on the P distance matrix calculated by VCF2Dis v1.43 (https://github.com/BGI-shenzhen/VCF2Dis).