Introduction

Chaenomeles, a genus within the subfamily Maloideae (Rosaceae) comprising four diploid (2n = 34) species1, is widely planted in East Asia and has important ecological, ornamental and economic value. In China, Chaenomeles has a cultivation history of more than 3000 years. Approximately 700 years ago, people began to recognize Chaenomeles’ medicinal values: their effect on curing rheumatoid arthritis and hypopepsia2. Currently, pharmacologists find that their leaves and flowers contain substantial prunasin; their seeds have a large amount of amygdalin; and their fruits contain abundant chemical ingredients, such as oleanolic acid, malic acid, pectinic acid, Chinese wolfberry citric acid, tartaric acid, citric acid, tannin, flavonoids, saponin and proanthocyanidins, which act as antioxidant and anticancer agents3. The remaining ingredients also have different pharmacological activities4.

In the Rosaceae family, Chaenomeles has close phylogenetic relationships with the Cydonia, Malus and Pyrus genera. Lindley established the Chaenomeles genus in 1822. Subsequently, some taxonomists began to reclassify some species into this genus. In 1890, Koehne reclassified Cydonia sinensis (C. sinensis) as Chaenomeles sinensis (Thouin) Koehne (C. sinensis)5. In 1906, Schneider reclassified Cydonia cathayensis (C. cathayensis) as Chaenomeles cathayensis (Hemsl.) Schneid. (C. cathayensis)6. In 1929, Nakai also reclassified Cydonia speciosa (C. speciosa) as Chaenomeles speciosa (Sweet) Nakai (C. speciosa)7. All three of these species had been thought to be members of Cydonia before that time. However, Koehne's classification result is still controversial, and not all taxonomists are in agreement8. Some taxonomists established independent genera Pseudochaenomeles Carr. (1882) and Pseudocydonia Schneid. (1906). Phipps9,10 considered Pseudocydonia an independent genus; in addition, it may be an intermediate type between Chaenomeles and Cydonia. Due to these controversies, the number of Chaenomeles is still ambiguous. According to I. et al1, there are only four species in this genus. However, five species have been identified in China, including Chaenomeles speciosa (Sweet) Nakai (C. speciosa), Chaenomeles cathayensis (Hemsl) Schneider (C. cathayensis), Chaenomeles japonica (Thunb) Lindley (C. japonica), Chaenomeles thibetica Yu (C. thibetica) and Chaenomeles sinensis (Thouin) Koehne (C. sinensis)2,11.

Efficient methods to clarify the taxonomic status of both wild and cultivated germplasms are needed1. At present, both morphological traits and various molecular markers are used to solve the taxonomic confusion12,13,14. In the past several years, restriction fragment length polymorphisms (RFLPs), random amplified polymorphic DNAs (RAPDs), amplified fragment length polymorphisms (AFLPs), simple sequence repeats (SSRs), EST-based microsatellites (EST-SSRs), intersimple sequence repeat (ISSRs), sequence characterized amplified regions (SCARs) and single nucleotide polymorphisms (SNPs) have been integrated into the mainstream method to detect evolutionary relationships in related species at the genomic level. Many related studies were reported. In Chaenomeles, Barthish et al.15 employed RAPDs to analyse offspring families of C. cathayensis and C. speciosa. Their results showed that the RAPD-based proportion of between-family variability was much higher in the hybrid populations than in pure species. He et al.16 evaluated genetic relationships among 52 C. speciosa accessions grown in China by combining AFLPs with leaf morphology characteristics; 208 polymorphic markers were identified. Zhang et al. used EST-SSR markers of Malus to research the genetic diversity of 33 Chaenomeles germplasms and obtained exciting results17. With the development of next-generation sequencing (NGS) technologies and algorithms for data processing, RNA-seq has become an efficient and economical approach for genomic and transcriptomic resource mining and has been increasingly considered an efficient tool to identify transcript-wide molecular markers and to solve phylogenetic relationships in non-model species18,19,20,21. Although few studies have examined transcriptome datasets for individual species, increasing attention has been given to comparative transcriptomes in the genus Chaenomeles.

Here, we sequenced the transcriptomes of 10 Chaenomeles germplasms belonging to five Chaenomeles species and one D. delavayi germplasm, which is an outgroup species of Chaenomeles, by the Illumina HiSeq 2500 platform. After de novo assembly and functional annotation, common orthologous genes among the 11 germplasms were obtained. We selected single-copy genes from all common orthologous genes to construct a phylogenetic tree to confirm their evolutionary relationship. By pairwise comparison of the orthologous sequences, genes involved in speciation that were under positive selection were obtained to illustrate some key genes during the evolution of this species. Many SSR sites were identified as potential molecular markers for Chaenomeles. Common genes containing SSR sites among the 10 Chaenomeles germplasms were selected to create a gene expression model in the secondary metabolism pathway. Finally, we developed some SSR markers for these 10 Chaenomeles germplasms. This study displays the first comparative exploration of Chaenomeles transcriptomes using high-throughput RNA-seq and provides an important methodology and database resources for facilitating further studies on the phylogeny of Chaenomeles and its related genus.

Results

De novo assembly and unigene qualification results of the 11 germplasms

11 germplasms including 10 Chaenomeles and one related species were collected from 6 provinces of China (Fig. 1). After assembling the unigenes for the 11 germplasms, we obtained 42,169 unigenes in D. delavayi (DY), 42,868 unigenes in C. sinensis (GP), 48,011 unigenes in C. speciosa (ZPCA), 48,487 unigenes in C. speciosa (ZPXC), 47,160 unigenes in C. speciosa (ZPLP), 46,193 unigenes in C. speciosa (ZPLY), 43,875 unigenes in C. thibetica (XZ), 40,084 unigenes in C. thibetica (XZSS), 49,571 unigenes in C. thibetica (MYXZ), 41,091 unigenes in C. cathayensis (MY), and 47,856 unigenes in C. japonica (RB) (Fig. 2A). The N50 values were 7,584 in D. delavayi (DY), 8,069 in C. sinensis (GP), 8,273 in C. speciosa (ZPCA), 8,304 in C. speciosa (ZPXC), 8,174 in C. speciosa (ZPLP), 8,374 in C. speciosa (ZPLY), 8,077 in C. thibetica (XZ), 7,508 in C. thibetica (XZSS), 8,369 in C. thibetica (MYXZ), 7,552 in C. cathayensis (MY), and 8,338 in C. japonica (RB) (Fig. 2B).

Figure 1
figure 1

Distribution of the 11 germplasms used for the research in China. Different symbols represent different species. Each germplasm was labelled with specimen IDs as follow: C. sinensis (GP) (SWH190512), C. speciosa (ZPLY) (SWH190535), C. speciosa (ZPXC) (SWH190621), C. speciosa (ZPCA) (SWH190657), C. speciosa (ZPLP) (SWH190645), C. thibetica (XZ) (SWH190805), C. thibetica (XZSS) (SWH190823), C. thibetica (MYXZ) (SWH190837), C. cathayensis (MY) (SWH190711), C. japonica (RB) (SWH190527), and D. delavayi (DY) (SWH190726). All of the germplasms were identified by Professor J.M.J. The map was downloaded on https://www.tianditu.gov.cn/.

Full size image
Figure 2
figure 2

Assembling results and gene expression distribution in 10 germplasms of Chaenomeles and Docynia delavayi (D. delavayi). In this graph, A represents the gene numbers in the 11 materials; B represents the N50 numbers in the 11 materials; C represents the N50 length in the 11 materials; D is a violin graph of the gene expression distribution in the 11 materials. On the X-axis, DY represents D. delavayi (DY), GP represents C.sinensis (GP), MY represents C. cathayensis (MY),XZ represents C. thibetica (XZ), MYXZ represents C. thibetica (MYXZ), XZSS represents C. thibetica (XZSS), RB represents C. japonica (RB), ZPCA represents C. speciosa (ZPCA), ZPLP represents C. speciosa (ZPLP), ZPLY represents C. speciosa (ZPLY), and ZPXC represents C.speciosa (ZPXC). In violin graph, different colours represent different germplasms. The area and width of the violin graph represent the expression abundance of the expressed genes in the corresponding germplasms. The expression levels in vertical coordinates are standardized by the log10 (RPKM) algorithm. The histograms were visulized by Origin pro 2016, violin graph was visulized by R 3.6.1.

Full size image

The N50 lengths were 1,655 bp in D. delavayi (DY), 1,744 bp in C. sinensis (GP), 1,699 bp in C. speciosa (ZPCA), 1,716 bp in C. speciosa (ZPXC), 1,706 bp in C. speciosa (ZPLP), 1,696 bp in C. speciosa (ZPLY), 1,807 bp in C. thibetica (XZ), 1,857 bp in C. thibetica (XZSS), 1,724 bp in C. thibetica (MYXZ), 1,712 bp in C. cathayensis (MY), and 1,651 bp in C. japonica (RB) (Fig. 2C). These results suggested that our assembly results are good enough for further analyses. In addition, we determined the gene expression levels of the 11 germplasms (Fig. 2D). After standardization with the log10 (RPKM) algorithm, the expression levels ranged from − 0.5 to 2. Except for C. thibetica (XZ) and C. thibetica (MYXZ), the remaining genera had a higher proportion of genes expressed at 0.1–0.2 after standardization. The median gene expression level was greater in C. thibetica (XZSS) than in the remaining germplasms (Fig. 2D), indicating different gene expression distributions in C. thibetica (XZ), C. thibetica (XZSS) and C. thibetica (MYXZ).

Screening of orthologous genes and single-copy genes

To better understand the similarities and differences among the 11 germplasms, a comparative analysis of the genes among different germplasms is the most efficient method. Genes from different germplasms with the best blast hits were considered orthologues. All these orthologues were commonly used to generate orthologous pairs. We identified orthologous groups for all predicted nucleic acid and protein sequences from the 11 germplasms by using OrthoMCL software. After merging genes from the same family, we finally obtained 28,181 OrthoMCL genes in C. thibetica (XZ), 28,908 OrthoMCL genes in D. delavayi (DY), 29,882 OrthoMCL genes in C. speciosa (ZPLP), 28,725 OrthoMCL genes in C. sinensis (GP), 31,682 OrthoMCL genes in C. thibetica (MYXZ), 27,989 OrthoMCL genes in C. cathayensis (MY), 31,472 OrthoMCL genes in C. japonica (RB), 26,214 OrthoMCL genes in C. thibetica (XZSS), 30,404 OrthoMCL genes in C. speciosa (ZPCA), 30,410 OrthoMCL genes in C. speciosa (ZPXC), and 29,960 OrthoMCL genes in C. speciosa (ZPLY) (Supplementary Table S1). Then, we identified the common orthologues among the 11 germplasms by a Venn diagram. A total of 9,659 orthologous genes were obtained (Fig. 3A, Supplementary Table S2A). Single-copy genes are valuable resources for phylogenetic analysis and SSR marker selection and only have one copy in the genome of the corresponding species. We also selected single-copy genes out of 9,659 orthologues by using a Venn diagram, and 6,154 orthologous single-copy genes were obtained (Fig. 3B, Supplementary Table S2B).

Figure 3
figure 3

Venn diagrams for the selection of common genes and common single-copy genes among 10 germplasms of Chaenomeles and D. delavay. A shows the selection of common genes in the 11 materials, B shows the selection of common single-copy genes in the 11 materials. Only common genes among the 11 materials and specific genes in each germplasm are shown in this graph. The diagrams were visualized on https://www.omicshare.com/tools/.

Full size image

Phylogenetic analysis and divergence time estimation based on single-copy orthologous families

In our study, single-copy gene families were used to conduct phylogenetic analysis. According to the evolutionary tree, the 11 germplasms were mainly divided into 5 different clades (Fig. 4). All 11 germplasms shared a common ancestor 22.7 million years ago. After this time point, D. delavayi (DY) and C. sinensis (GP) began to independently diverge from the remaining nine Chaenomeles. Approximately 11.1 million years ago, C. japonica (RB) also diverged from the remaining Chaenomeles. Herein, C. japonica (RB), as a single branch, is also supported by AFLP studies22 showing interspecific differentiation. C. speciosa (ZPLY), commonly known as ‘Yizhoumugua’ in Linyi City of Shandong Province, is generally considered to originate from interspecific hybridization with other species in the genus23. Our study showed that it split from other Chaenomeles approximately 9.07 million years ago and had a relatively distant genetic relationship with the other three Chaenomeles germplasms, C. speciosa (ZPXC), C. speciosa (ZPCA) and C. speciosa (ZPLP) (Fig. 4). Approximately 8 million years ago, C. speciosa separated from C. thibetica and C. cathayensis. Three germplasms of C. thibetica, namely, XZ, XZSS and MYXZ, clustered in the same branch as C. cathayensis (MY), showing a very close phylogenetic relationship (Fig. 4). A previous study23 suggested that C. thibetica might be a hybrid species of C. cathayensis and D. delavayi (DY). However, more evidence is needed to support this result. C. speciosa is the most widely distributed species of the genus in China. Due to its outstanding pharmacological effects, it has been domesticated and cultivated for a long time and has formed three genuine producing areas, corresponding to the distribution of three local varieties, namely, ‘Cunmugua’ (ZPCA), ‘Xuanmugua’ (ZPXC) and ‘Ziqiumugua’ (ZPLP). Our phylogenetic analysis indicated their close relationship.

Figure 4
figure 4

Phylogenetic tree based on common single-copy genes of the 11 germplasms. Numbers to the left of the nodes are bootstrap values; numbers to the right of the nodes are divergence times. MYA is short for million years ago. Evolutionary tree were conducted in MEGA X (V10.02).

Full size image

GO and KEGG analysis results for single-copy genes in 11 germplasms

To understand the function of 6,154 common single-copy genes in 11 germplasms, we conducted GO and KEGG analyses (Fig. 5). According to the GO and KEGG results, these 6,154 genes shared similar metabolic processes. After merging the top 15 GO terms in each germplasm, we obtained 27 common GO terms (Fig. 5A). Common single-copy genes were significantly enriched in these 27 GO metabolic processes in all 10 Chaenomeles germplasms except D. delavayi (DY). In D. delavayi (DY), common single-copy genes were insignificantly enriched in DNA metabolic process (GO: 0006259) with a p. adjust of 0.294 and chromosome organization (GO: 0051276) with a p. adjust of 0.564, indicating its different metabolic process from Chaenomeles (Fig. 5A Supplementary Table S3). All 11 germplasms were especially enriched in nitrogen compound metabolic process (GO: 0006807), organic cyclic compound metabolic process (GO: 1901360), heterocycle metabolic process (GO: 0046483), cellular nitrogen and compound metabolic process (GO: 0,034,641), and nucleic acid metabolic process (GO: 0090304) (Fig. 5A, Supplementary Table S3).

Figure 5
figure 5

GO and KEGG analyses for common single-copy genes among 10 germplasms of Chaenomeles and D. delavayi. In this graph, the pot size indicates the number of enriched gene; the degree of significance decreases from red to green. In both the GO and KEGG analysis results, only the top15 GO terms and pathways are shown. The graphs were visualized by R 3.6.1.

Full size image

In the KEGG analysis, after merging the top 15 pathways in each germplasm, we finally obtained 31 common pathways (Fig. 5B). The common single-copy genes were significantly enriched in nonhomologous end-joining (ko03450), ribosome biogenesis in eukaryotes (ko03008) and purine metabolism (ko00230) in all 11 germplasms (Fig. 5B Supplementary Table S4). In addition to these two pathways, D. delavayi (DY) was also enriched in biosynthesis of amino acids (ko01230) with a gene number of 97 out of 297 (Q value 0.032). C. sinensis (GP) was also enriched in spliceosome (ko03040) and biosynthesis of amino acids (ko01230) with gene numbers of 89 out of 263 (Q value 0.008) and 96 out of 296 (Q value 0.0158), respectively. C. cathayensis (MY) was enriched in aminoacyl-tRNA biosynthesis (ko00970) with a gene number of 31 out of 70 (0.006). C. thibetica (MYXZ) was enriched in base excision repair (ko03410) with a gene number of 24 out of 52 (Q value 0.0006). C. japonica (RB) was also enriched in N-glycan biosynthesis (ko00510) with a gene number of 24 out of 57(Q value 0.0257). C. thibetica (XZ) was also significantly enriched in biosynthesis of amino acids (ko01230) and nucleotide excision repair (ko03420) with gene numbers of 97 out of 290 (Q value 0.0097) and 36 out of 87 (Q value 0.0097), respectively. C. thibetica (XZSS) was enriched in base excision repair (ko03410) and spliceosome (ko03040) with gene numbers of 24 out of 48(Q value 0.0204) and 88 out of 247(Q value 0.0204), respectively. C. speciosa (ZPCA) was significantly enriched in nucleotide excision repair (ko03420) with a gene number of 36 out of 89 (Q value 0.0014). Both C. speciosa (ZPLP) and C. speciosa (ZPLY) were enriched in base excision repair (ko03410) with gene numbers of 25 out of 54 (Q value 0.0016) and 25 out of 46 (Q value 0.00096), respectively. C. speciosa (ZPXC) was enriched in aminoacyl-tRNA biosynthesis (ko00970) and spliceosome (ko03040) with gene numbers of 32 out of 70 (Q value 0.0004) and 88 out of 290 (Q value 0.0034), respectively (Fig. 5B, Supplementary Table S4).

Detecting genes under selective pressure

The nonsynonymous substitution rate (Ka) and synonymous substitution rate (Ks) were used to estimate the change in coding protein sequences. The magnitude of the Ka/Ks ratio provides evidence of genes under strong functional constraints (Ka/Ks < 1) or undergoing adaptive evolution (Ka/Ks > 1)24,25,26. The ratio of the nonsynonymous substitution rate (Ka) and synonymous substitution rate (Ks) were used to assess whether genes were under selection. After filtering, a total of 6057 genes remained for further analysis. According to the BLASTn results, a total of 6057 single-copy genes underwent selective processing (Fig. 6A, Supplementary Table S5A). Among them, 467 out of 6057 genes underwent purification selection, indicating that these genes are disadvantageous during the evolutionary process. A total of 693 out of 6,057 genes experienced weak positive selection, and 45 out of 6,057 genes experienced strong positive selection, indicating that those genes are domain genes that determine the direction of evolution (Fig. 6B, Supplementary Table S5B-D). Furthermore, we blasted these 45 positively selected genes to Arabidopsis (Supplementary Table S5D). Nineteen out of 45 genes had homologous genes in Arabidopsis. ORTHOMCL10413 (RGA2) mainly acts as a repressor of the gibberellin (GA) signalling pathway through a transcription coactivator of the zinc finger transcription factors GAF1/IDD2 and ENY/IDD1 to regulate gibberellin homeostasis and signalling27. ORTHOMCL9348 (RPL5) participates in cell proliferation and plays a role in translation in leaf dorsoventral patterning to specify leaf adaxial identity28. ORTHOMCL8960 (HEL) has a modular structure consisting of an N-terminal hevein-like domain (CB-HEL) and a C-terminal domain (CD-HEL) that is post-translationally processed. Both domains show strong antifungal activity, indicating that this gene is responsible for defending plants against pathogen attack29. ORTHOMCL9718 (FIP2) can delay flowering by repressing the expression of FLOWERING LOCUS C30. ORTHOMCL4564 (CSA1) also plays an important role in disease defence31. ORTHOMCL9773 (MYB27) is a member of the MYB family, and many studies have indicated that MYB family members can regulate secondary metabolism, control cell shape, defend against disease and increase hormone responses32. ORTHOMCL8338 (BZIP60) is involved in controlling endoplasmic reticulum pressure to enhance the immune response in both animals and plants33. In addition, although there are still no reports on the function of ORTHOMCL5656 (CYS6), ORTHOMCL6947 (HPR3) and ORTHOMCL6257 (F11P17.9) in plants, according to the GO analysis results, these three genes mainly participated in the disease defence process (Supplementary Table S3). All of these results indicate that most of the positively selected genes are involved in building the plant disease defence system. Thus, these plants can survive long-term evolution and selection processes.

Figure 6
figure 6

Selection pressure analyses for single-copy genes in common genes of 10 germplasms of Chaenomeles and D. delavayi. In graph A, each coloured dot represents one single-copy gene family. The top diagonal line represents Ka/Ks = 1; the first dotted line represents Ka/Ks = 0.5; and the second dotted line represents Ka/Ks = 0.1. In graph B, PS is short for purification selection; WPS is short for weak positive selection; and SPS is short for strong positive selection.Scatter diagram was visualized by R 3.6.1. Histogram was visulized by Origin pro 2016.

Full size image

Model of gene expression for genes containing simple sequence repeats (SSRs) in the biosynthesis of secondary metabolic pathways of 10 Chaenomeles germplasms

We predicted the SSR sites in all 6,154 common single-copy genes by using MISA software. A total of 1,210, 1,151, 1,128, 1,253, 1,186, 1,225, 1,178, 1,181, 1,170, and 1,165 genes contained SSR sites in C. sinensis (GP), C. japonica (RB), C. cathayensis (MY), C. thibetica (XZ), C. thibetica (XZSS), C. thibetica (MYXZ), C. speciosa (ZPCA), C. speciosa (ZPLP), C. speciosa (ZPXC), and C. speciosa (ZPLY), respectively (Supplementary Table S6). Five SSR types, di-, tri-, tetra-, penta- and hexanucleotides, were counted. Among them, di- and trinucleotides were the main repeat units, ranging from 60.43% to 63.69% and 28.72% to 30.89%, respectively (Supplementary Table S7A-J). Then, we selected the common genes that contained SSR markers for 10 germplasms. In total, 292 common single-copy genes were selected (Fig. 7, Supplementary Table S8). Furthermore, we conducted KEGG analyses for these 292 common single-copy genes. All 10 germplasms were significantly enriched in the biosynthesis of secondary metabolites (Supplementary Table S9). Therefore, we examined this pathway to determine the gene expression model for each germplasm. Although all 10 germplasms had genes participating in the same process in the pathway, there still existed differences in the gene expression level. For examples, from phosphatidylethanolamine to 1,2-diacyl-sn-glycerol and 1,2-diacyl-sn-glycerol to phosphatidylcholine, NPC6 expression was lower in C. thibetica (MYXZ) than in the rest of the germplasms. From magnesium protoporphyrin to protoporphyrin, CHLD expression was lower in C. thibetica (XZ) than in the remaining germplasms. During transformation from 3-oxoacyl-CoA to acyl-CoA and malonyl-CoA, KCSLL expression was higher in C. sinensis (GP), C. japonica (RB), C. thibetica (XZSS) and C. speciosa (ZPXC) than in the other germplasms (Fig. 8, Supplementary Table S10). However, there were similarities in the gene expression levels in the 10 germplasms. For instances, during the transformation of 6-geranylgerany-2-methylbenzene-1,4-diol to 6-geranylgeranyl-2,3-dimetheyl benzene-1,4-diol, shikimate 3-phosphate to O5-(1-Carboxyvinyl) -3- phosphoshikimate, (3S)-3-methyl-2-oxopentanoate to L-isoleucine, 3-methyl-2-oxobutanoate to L-valine and L-leucine to 2-oxoisocaproate, we detected that VTE3, EPSPS-1 and BACT2 had a high expression level in all of 10 Chaenomeles germplasms (Fig. 8, Supplementary Table S10).

Figure 7
figure 7

Venn diagram and KEGG analysis for common single copy genes in 10 germplasms of Chaenomeles containing SSR sites. In graph B, the dot size indicates the number of enriched gene; the degree of significance decreases from red to green.Venn diagram was visualized on https://www.omicshare.com/tools/. KEGG result was visulized by R3.6.1.

Full size image
Figure 8
figure 8

Gene expression models in biosynthesis of secondary metabolites pathway. This graph displays part of the biosynthesis of secondary metabolites. Dotted lines describe indirect metabolism flow; solid lines describe direct metabolism flow. The coloured bar describes the gene expression level in the corresponding position. RPKMs in heatmaps are standardized by the log2 (RPKM) algorithm; green indicates a low expression level, and red indicates a high expression level. Metabolites pathway was drawn by Adobe Illustrator CS5 software, heatmaps were visualized by TBtools 0.6669.

Full size image

Primer screening, polymorphism study and reconstruction of an evolutionary tree based on 10 selected SSRs

Then, we designed primers for genes containing SSR sites with Primer 3 following the rules described in the methods. In total, we obtained 373, 357, 362, 366, 355, 354, 352, 365, 370 and 372 primers for C. sinensis (GP), C. cathayensis (MY), C. thibetica (MYXZ), C. japonica (RB), C. thibetica (XZ), C. thibetica (XZSS), C. speciosa (ZPCA), C. speciosa (ZPLP), C. speciosa (ZPLY) and C. speciosa (ZPXC), respectively (Supplementary Table S11A). The number of common genes used for primer design in 10 Chaenomeles germplasms was 110 (Supplementary Table S11B).

A total of 110 pairs of SSR primers were synthesized, and the 10 Chaenomeles germplasms were used to verify the PCR amplification ability. Based on this result, a total of 45 pairs of primers were screened out (Supplementary Table S12). Then, capillary electrophoresis was used to detect the levels of polymorphism and specificity of the 45 primers. The results showed that only 10 pairs of primers had good polymorphism levels. Primers from ORTHOMCL6247, ORTHOMCL8473, ORTHOMCL7263 ORTHOMCL4541, ORTHOMCL9476, ORTHOMCL9834 and ORTHOMCL8700 had the best polymorphism levels (Table 1, Supplementary Fig. S1). The primers from ORTHOMCL4735, ORTHOMCL8947 and ORTHOMCL6535 also had relatively good polymorphism levels but had poor amplification results for individual germplasms (Table 1, Supplementary Fig. S1). Table 1 displayed all 10 selected primers, including tandem repeat units, annealing temperature and detailed forward and reverse primers (Table 1).

Table 1 Primer and fragment length for selected genes.
Full size table

Polymorphism analysis of these SSR loci showed that the average allele number was 6.6, ranging from 5 to 9. The average observed heterozygosity was 0.243, ranging from 0 to 0.455. The average expected heterozygosity was 0.823, ranging from 0.723 to 0.935 (Table 2). Lower expected heterozygosity than observed heterozygosity suggests a departure from Hardy–Weinberg equilibrium and a high probability of inbreeding34. All 10 selected SSRs had lower expected heterozygosity than observed heterozygosity, indicating that inbreeding processes occurred among Chaenomeles. Traditional studies have divided the polymorphism information content into three classes: high polymorphism (PIC > 0.5), moderate polymorphism (0.25 < PIC < 0.5) and low polymorphism (PIC < 0.25)35. In our research, the PIC values of the 10 selected SSRs were more than 0.5, suggesting high polymorphism and many effective genetic markers from our screened primers.

Table 2 Polymorphic information of the 10 SSR loci in the 10 germplasms.
Full size table

Then, we used the genes amplified by these 10 primers to build an evolutionary tree. The results showed that the 11 germplasms were mainly divided into two clades (Supplementary Fig. S2). D. delavayi (DY), C. sinensis (GP), C. japonica (RB), C. speciosa (ZPLY), C. speciosa (ZPLP) and C. cathayensis (MY) were clustered in the same clade and diverged sharply from the remaining germplasms. Among them, D. delavayi (DY), C. sinensis (GP), C. japonica (RB) and C. speciosa (ZPLY) had a similar evolutionary relationship based on all 6,154 single-copy genes (Fig. 4, Supplementary Fig. S2). Three germplasms from Tibet, C. thibetica (MYXZ), C. thibetica (XZ) and C. thibetica (XZSS), diverged from the above six germplasms, suggesting genetic differentiation between them. The four C. speciosa germplasms distributed in different clades may be caused by the influence of different geographical distributions and artificial domestication on these ten screened genes. In total, the evolutionary relationship based on 10 screened genes did not completely correspond with that based on 6,154 single-copy genes.

Discussion

Several approaches, including morphological markers, cytological markers, and biochemical markers, have been explored to classify plant resources22. However, each technique is imperfect. For example, great phenotypic differences between individuals from geographically isolated populations create obstacles for taxonomists. Subjective factors can lead to misclassification. The limited number of isoenzymes and low resolution of cytology prevent taxonomists from recognizing many plants. Due to these flaws, molecular markers have become a prevalent tool in evolutionary studies in Chaenomeles and related species36,37,38,39. Various molecular markers have provided reliable and convincing evidence for related studies, promoting phytotaxonomic research, whereas some technologies are still unaffordable because of their high cost. Currently, RNA-seq has become an economical and efficient tool for accessing molecular information.

The phylogeny and classification of Chaenomeles have been controversial for several centuries, especially whether C. sinensis is a member of Chaenomeles. Robertson et al.40 examined the phylogenetic relationships among 88 genera of Rosaceae, and the results show that it is far from the Chaenomeles group, indicating that C. sinensis is not a member of Chaenomeles. In our research, we found that C. sinensis had a closer evolutionary relationship with D. delavayi, confirming Robertson’s result. There was obvious differentiation between C. sinensis (GP) and all other species of Chaenomeles. In terms of phenotype, C. sinensis (GP), which has the characteristics of branches unarmed, flowers solitary, coetaneous, sepals reflexed, stipules ovate-lanceolate and margin glandular serrate, is also significant different from the other Chaenomeles species, which have the characteristics of branches armed, flowers fascicled, precocious or coetaneous, sepals erect, rarely reflexed, stipules herbaceous, reinform or auriculate and margin serrate23. Pollen morphology also shows that C. sinensis (GP) is distant from other species41. Therefore, based on the phylogenetic results and phenotypic characteristics, we support that C. sinensis should not belong to Chaenomeles but should be an independent genus Pseudocydonia. This view was also supported by our other study on the phylogeny of Chaenomeles based on chloroplast genome sequencing42.

The genetic characteristics of populations should be dictated by the interplay of genetic drift, gene flow and natural selection. These processes may be strongly influenced by the demography and spatial distribution of populations43. In our study, the same species from different regions exhibited not only differences in appearance but also differences in their main chemical components; for this reason, the term genuine regional drug appeared in traditional Chinese medicine. From a molecular marker perspective, although these germplasms are the same Chaenomeles, they have formed genetically diverse offspring leading to different types (Fig. 8), as has been reported in Bartish’s research, which discriminated 42 Chaenomeles by the RAPD method44.

For closely related taxa, the advantages of homologous single-copy genes for phylogenetic and phylogeographic analysis are clear because of their rapid evolutionary rates and clear avoidance of paralogy45,46,47,48. According to past studies, many single-copy genes have been identified in many plants, including Euasterids, Rosaceae, Poaceae, Cycades and 29 other angiosperms49,50,51,52,53. For example, Duarte et al. identified a series of 959 single-copy genes in A. thaliana, P. trichocarpa, V. vinifera and O. sativa and used 18 single-copy genes to improve the resolution of Brassicaceae phylogeny45. In this study, we identified 6,154 single-copy genes according to the strict filtering criterion. All these detected single-copy genes proved highly effective in the phylogenetic reconstruction of Chaenomeles species, which provided a rough phylogenetic framework for the whole genus (Fig. 4). Therefore, this result indicated that all these candidate single-copy genes from the transcriptomes of Chaenomeles species have good application in molecular phylogeny studies of the Chaenomeles genus.

In our results, 10 out of 45 SSR markers were finally confirmed to distinguish the 10 Chaenomeles germplasms. They have good polymorphism and amplification levels. The efficiency of the primer development rate was 22.2%, which was slightly lower than that reported by Wang et al. (28.6%), who developed SSR markers in pineapple. For the development rate, many factors can influence the results, such as the selected length of the primers, the genetic diversity of the tested species, and development methods. For instance, Wang54 compared four methods, including ISSR, SSR, SCoT and RAPD, by using 47 germplasms of Ananas comosus (L.) Merr. Among them, the development rate of RAPD reached 47.06%, which was significantly higher than that of the other methods. However, the SCoT and SSR methods had good comprehensive performance compared to the two other methods. Taken together, our results have proven that SSR marker development based on the transcriptome is promising.

Although classification studies in Chaenomeles have made great progress, nomination and classification systems still have some problems. The results of this study support the idea of C. sinensis as an independent genus, Pseudochaenomeles. With increasing exploration of its edible and medicinal value, it is important to distinguish the various Chaenomeles to avoid misuse. Thus, the determination of authoritative criteria is urgent. Due to the delay in the sequencing of the Chaenomeles genome, it is still difficult to explore SSR markers efficiently. Because of the low cost of transcriptome sequencing, this method will be the best choice; thus, our research provides a new methodology and reference for future related research in Chaenomeles. The high genetic diversity in the genus Chaenomeles, as inferred from molecular markers, is also useful for improvement, breeding and selection projects in related species.

Methods

Plant material, total RNA extraction and Illumina sequencing

All five species of Chaenomeles and one related species (D. delavayi) were collected (Fig. 1). C. sinensis (GP) was collected in Anji County of Zhejiang Province. According to the local common name, C. speciosa was divided into four types, including C. speciosa ‘Chunmugua’ (ZPCA) from Chunan County of Zhejiang Province, C. speciosa ‘Ziqiumugua’ (ZPLP) from Changyang County of Hubei Province, C. speciosa ‘Xuanmugua’ (ZPXC) from Xuancheng City of Anhui Province and C. speciosa ‘Yizhoumugua’ (ZPLY) from Linyi City of Shandong Province. Three local types of C. thibetica were also collected and distributed in Lhasa (XZ), Bomi (XZSS) and Motuo (MYXZ) of Tibet. C. cathayensis (MY) was collected from Shiping County, Yunnan Province. C. japonica (RB) was introduced from Japan and collected from the Germplasm Resource Bank of Chaenomeles in Anji County of Zhejiang Province. D. delavayi (DY) was collected from Lancang County of Yunnan Province. The voucher specimens were deposited in the Herbarium of Research of Institute of Subtropical Forestry, Chinese Academy of Forestry with the following specimen IDs: C. sinensis (GP) (SWH190512), C. speciosa (ZPLY) (SWH190535), C. speciosa (ZPXC) (SWH190621), C. speciosa (ZPCA) (SWH190657), C. speciosa (ZPLP) (SWH190645), C. thibetica (XZ) (SWH190805), C. thibetica (XZSS) (SWH190823), C. thibetica (MYXZ) (SWH190837), C. cathayensis (MY) (SWH190711), C. japonica (RB) (SWH190527), and D. delavayi (DY) (SWH190726). All of the germplasms were identified by Professor J.M.J.

A total of 15 duplicate leaves were collected from each germplasm resource. According to the manufacturer’s instructions, total RNA was isolated from mature leaf samples as described in Owczarek et al.55. After testing the qualification of extracted RNA, including concentration, RIN value, ratio of 28S to 18S, and OD260 to OD280 by using an Agilent 2100 Bioanalyser and a NanoDrop, the RNAs from the same germplasm were pooled to prepare a cDNA library by using the cDNA Synthesis Kit (Illumina Inc., San Diego, CA) following the manufacturer’s recommendations56. Finally, the constructed cDNA libraries were sequenced using an Illumina HiSeq 2500 by a paired-end 150 bp strategy.

Filtering and assembly of the reads, quantitative calculation and functional annotation for unigenes

First, raw reads were filtered with fastp software (V0.19.4) by removing adaptors and low-quality reads57. Then, the cleaned reads were de novo assembled using Trinity with the default parameters for each individual sample58. The expression level of unigenes was calculated by RSEM59 software (V1.3.0) and standardized by RPKM. Unigenes were queried against the nonredundant protein (Nr) database, the Swiss-Prot protein database, the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database and the Cluster of Orthologous Groups (COG) database using BLASTx with an E-value cut-off of 1e−560. According to the BLAST results, the coding region sequences (CDSs) of unigenes were extracted and translated into peptides. For the unigenes without matches in the above four databases, we predicted the coding region sequences by using TransDecoder. Gene Ontology (GO) annotation of the unigenes was obtained using Blast2GO61. Enrichment analyses of GO terms and KEGG pathways were conducted as described in Li et al.62.

Identification of orthologues, single-copy genes and multiple-copy genes

Diamond (V3.1)63 and OrthoMCL (V2.0.9)64 software were used to identify the orthologues among the 11 germplasms. In the first step, multiple sequence comparisons among different germplasms were conducted with cut-offs of E-value ≤ 1e−5 and query cover ≥ 30% by using Diamond. In the second step, the orthologues were merged into different families by using OrthoMCL. These obtained orthologous genes can be divided into single-copy genes and multiple-copy genes according to their copy numbers.

Phylogenetic analysis and selective pressure analysis

Based on the above analyses, we first constructed sequence alignment by using MUSCLE (multiple sequence alignment with high accuracy and high throughput, V3.8.31) for single-copy genes. Then, a timetree was estimated by the RelTime method65,66 with the maximum likelihood method67. Evolutionary analyses were conducted in MEGA X68. To detect whether genes were affected by natural conditions, Paml-codeml was used to compute the nonsynonymous substitution rate (Ka), synonymous substitution rate (Ks) and ratio of Ka to Ks. To avoid influences from nonsignificant or excessive variation, these single-copy genes were discarded as described by Goodman et al.69. We aligned orthologous pairs70. Orthologous genes with a Ka/Ks > 1 were considered genes under strong positive selection; orthologous genes with 0.5 < Ka/Ks < 0.1 were considered affected by weak positive selection; and orthologous genes with a Ka/Ks < 0.1 were considered affected by negative selection. The significance tests were measured with p values, which were corrected via the false discovery rate (FDR).

Expression model of single-copy genes in the secondary metabolite pathway

After performing KEGG analysis, a significantly enriched pathway, the secondary metabolite pathway, was selected for detecting a gene expression model among the 10 Chaenomeles germplasms. Genes in key steps of the secondary metabolite pathway were obtained according to web-version map01110. The pathway was depicted by Adobe Illustrator CS5 software71, and heatmaps were visualized by TBtools 0.666972.

Identification, SSR screening, polymorphism study of SSR loci and reconstruction of phylogenetic tree among 10 Chaenomeles germplasms

We identified SSR motifs in all unigenes using MISA software73. Based on tandem repeats, including AAC, ACA, CAA, GTT, TGT and TTG, all these SSR motifs were divided into di-, tri-, tetra-, penta- and hexanucleotide types. PCR primers were designed for the SSRs using the program Primer3; they were designed to have more than 150 bp flanking sequences74. All of the primers used for SSR development are shown in Supplementary Table S12.

DNA was extracted from the leaves of the 11 germplasms as described in Bartish et al.75. Fluorescently labelled primers were synthesized for amplifying gene fragments based on the following rules: the primer should have 2, 3, 4, or 5 tandem repeat units; the length of the PCR products should range from 150 to 300 bp; the position of genes should not focus on one site; the polymorphic sites should be chosen first; different combinations of tandem repeat units should be chosen on average; the repeat base in primer should be less than four; the length of the primers should be approximately 20–23 bp; the Tm value of the primers should be ~ 60 ℃; the number of consecutive As or Ts at the 5′ or 3′ region of the primer should be less than two; and repeat sequences within primer should be forbidden. The PCR products were detected by capillary electrophoresis.

Polymorphism indexes for the SSR loci, such as the number of alleles, observed heterozygosity, expected heterozygosity and polymorphism information content, were calculated by Cervus 3.0.776. We used neighbor-joining methods in NTSYS (V2.10e)77 software to rebuild a phylogenetic tree.

Ethics approval and consent to participate

The 11 plant materials in this paper were collected with the permission of the local authorities. The research on plants and collection of plant material complied with institutional, national or international guidelines. We also complied with the IUCN Policy Statement on Research Involving Species at Risk of Extinction and the Convention on the Trade in Endangered Species of Wild Fauna and Flora.