Dr. Zhong Guoyue of Jiangxi University of Chinese Medicine (JUTCM) confirmed plant samples belonging to 19 different Rosaceae species that were obtained in the Tibet Autonomous Region of China (Table S1). Every voucher specimen is stored at the JUTCM Herbarium.

The Q30 values of the cp genomes of the 19 Rosaceae species ranges from 92.11% to 93.75%, indicating that the sequencing quality was reliable. The 19 cp genomes range from 153,366 bp (D. arbuscula) to 159,895 bp (S. koehneana), and all exhibit a typical quadripartite structure, consisting of LSC region (8,445–87,883 bp), SSC region (18,174–19,259 bp), and IR region (25,310–26,396 bp) (Table S2). The total GC content was 37.81%–41.1%. The GC content in LSC region, SSC region and IR region was 34.2%–35.23%, 30.25%–31.45% and 42.54%–42.9%, respectively. The total number of annotated genes in all species ranges from 132 to 138, including 87–93 PCGs, 37 tRNA genes, and 8 rRNA genes (Fig. 1, Table S3).

The relative likelihood of synonymous codons expressing a certain amino acid is known as the Relative Synonym Codon Usage (RSCU). The codon composition of 87–93 PCGs in the 19 cp genomes of Rosaceae species was analyzed, and RSCU values were calculated. The cp genomes contained 63 codons (including UAG, UAA, UGA) (Fig. 2), encoding 20 amino acids. The total number of codons encoding proteins ranged from 26,361 (A. lineata) to 27,014 (R. graciliflora).

Leucine (Leu) has the highest codon usage rate (2,776 to 2,847), while Cysteine (Cys) has the lowest codon usage rate (298 to 315). The RSCU values of the 19 cp genomes ranged from 0.3684 to 3.962. Methionine (Met), which was encoded by AUG, had the greatest RSCU value among them, whereas leucine (Leu) was encoded by CUG, which had the lowest. Furthermore, 30 codons had an RSCU greater than 1 and 33 codons had an RSCU less than 1. All of the codons with RSCU > 1 terminated in A or U, which was in line with the numerous A/T features of angiosperm cp genomes (Fig. 2).

A total of 52 (P. fruticosa) to 121 (R. biflorus) SSR loci were detected in the 19 cp genomes, comprising six types of repeats (Fig. 3). Among them, the mononucleotide repeats were the richest (38–84), followed by di- (10–20), tetra- (2–11), tri- (0–7), pentanucleotide repeats (0–4), and hexanucleotide repeats (0–3), respectively. Mononucleotide, dinucleotide, and tetranucleotide repeats existed in the cp genomes of all 19 species. Trinucleotide repeats were absent in P. trichostoma, C. adpressus, C. multiflorus, C. wardii, and S. koehneana. Pentanucleotide repeats were present in P. trichostoma, C. adpressus, C. wardii, R. sweginzowii, S. diandra, and S. koehneana, but absent in the other 13 species. Hexanucleotide repeats were only found in P. trichostoma, R. graciliflora, R. macrophylla, R. sweginzowii, S. diandra, and S. koehneana (Fig. 3).

SSRs were not evenly distributed across the cp genomes. The LSC region included 46–96 SSRs, the SSC region had 5–12 SSRs, and the IR region contained 0–15 SSRs. More specifically, there were no SSRs in the IR region of D. arbuscula’s cp genome. Future research on genetic diversity may find areas of fast evolution and suitable targets due to the large number of SSRs found in the LSC region.

There are 32–77 long repeat sequences in the cp genomes of 19 Rosaceae species. There are four different types of repeat sequences: forward (F), reverse (R), palindromic (P), and complementary (C). The large repeat sequences were between 30 and 26,396 bp long. Complementary repeat sequences were found only in nine species, i.e., C. multiflora, C. wardii, F. moupinensis, F. eriocarpa, A. lineatum, R. biflorus, R. pedunculosus, S. diandra, and S. koehneana. Forward, palindromic, and reverse repeat sequences were found in all 19 species. The number of forward repeat sequences ranged from 13 to 38, palindromic repeat sequences ranged from 15 to 27, and reverse repeat sequences ranged from 1 to 10. The most prevalent sequences were palindromic repeats, which made up 31.17% to 55.56% of all sequences, followed by forward repeat sequences, accounting for 29.17% to 49.35% (Fig. 4).

This study offers a meticulous examination of the IR boundaries and adjacent genes within 19 species of the Rosaceae family. Despite the similar lengths of the IR regions across these species—spanning from 25,310 to 26,396 base pairs—variations in their expansion and contraction were noted. The study highlights the conservation of the SSC/IRa boundary, in stark contrast to the variability observed in the LSC/IRb, IRb/SSC, and IRa/LSC boundaries, with significant differences notably present at the LSC/IRb and IRb/SSC junctures. Particularly for P. trichostoma, an anomaly in the LSC/IRb boundary positioning between the rp122 and rps19 genes was identified, diverging from the consistent placement between the rps19 and rp12 genes seen in the other 18 species.

The positioning of the rps19 maintained consistency across seven Rosa species, precisely 11 bp from the LSC/IRb boundary. Conversely, in Cotoneaster and Sorbus species, the gene straddled the LSC/IRb boundary, with portions extending into both the LSC and IRb regions. Rubus species displayed the rps19 gene merely 7 bp away from the LSC/IRb boundary, indicating a variance among different genera. The rp12 gene resided entirely within the IRb region across all species examined.

The ndhF gene, situated in the SSC region, either crossed the IRb/SSC boundary or was positioned at various distances from it in 16 species across 8 genera, excluding Cotoneaster and Sorbus. The ycf1 gene consistently spanned the transition from the SSC to IRa regions in all species. Furthermore, the trnH gene crossed the IRa/LSC boundary in seven Rosa species, diverging in its precise location across different genera.

In terms of boundaries, variations were most prominently observed at the LSC/IRb (JLB) and IRb/SSC (JSB) boundaries. Among the species examined, those in the genera Cotoneaster and Sorbus showed the greatest similarity in their IR boundaries, with sizes and distances being remarkably consistent. Interestingly, A. lineata, F. eriocarpum, and D. arbuscula exhibited significant divergences in their boundary configurations, suggesting that these differences may have contributed to their evolutionary divergence from Potentilla (Fig. 5). These findings highlight the complexity and variability in the positioning of genes relative to the IR boundaries across the Rosaceae family, providing valuable insights into the evolutionary relationships and genetic diversity within this group.

The evolutionary rate of sequences is influenced by nucleotide substitutions and selective pressures. The Ka/Ks ratio is commonly used to measure whether PCGs are under selective pressure. Only 83 PCGs in the 19 cp genomes had Ka or Ks values (Fig. 6). Since Ka or Ks = 0, which indicates that these sequences were preserved without nonsynonymous or synonymous nucleotide alterations, the Ka/Ks values for the other PCGs cannot be computed. The Ka/Ks values ranged from 0.00 to 1.06, indicating purifying selective constraints acting on the chloroplast PCGs. Genes with a Ka/Ks value of 0 included at pH, orf188, petN, psaC, psbA, psbD, psbE, psbF, psbI, psbM, psbN, rpl36, rps12, rps7, and ycf15, indicating that they were under strong purifying selection. Genes with both Ka and Ks values of 0 included orf42, psaJ, psbL, rpl23, and ycf68. Only three genes, ycf2, psaI, and infA, had a Ka/Ks value greater than 0.5. Among the 83 genes, only infA had a Ka/Ks value greater than 1, indicating positive selection acting on this gene (Fig. 6).

Higher pi values indicate greater polymorphism and divergence hotspots. By using a sliding window approach to calculate nucleotide diversity (pi) values, the average nucleotide diversity (pi) values in the LSC, SSC, and IR regions of the 19 cp genomes ranged from 0.000 to 0.215 (Fig. 7), suggesting that these species may have undergone rapid nucleotide substitutions. Among them, the SSC region exhibited the highest pi value, indicating that the variation was concentrated in this area. The pi value of IR region was the lowest, which once again verified that IR region was more conservative than LSC and SSC region. A total of nine regions with higher pi values (pi > 0.13) were identified, including 36401–37000 (0.131), 61601–62200 (0.137), 55801–56400 (0.141), 12801–13400 (0.159), 129801–130400 (0.152), 130001–130600 (0.160), 130201–130800 (0.175), 56201–56800 (0.183), and 128801–129400 (0.215). Among them, five regions were in the LSC region, and 4 regions were in the SSC region (Fig. 7).

Using the complete cp genomes of 19 Rosaceae species obtained in this study and 110 Rosaceae species downloaded from the NCBI website, the ML and BI phylogenetic trees were constructed using the Phylosuite software. The ML and BI trees have exactly the same topological structure, and the bootstarp value of most branch nodes was greater than 75 and the posterior probability was greater than 0.997, indicating that the phylogenetic relationship was reliable. The 129 species of the Rosaceae family can be divided into 8 branches, namely Clade I to Clade VIII. Clades I to III belong to Prunus L., Sorbus L., and Cotoneaster Medik., respectively, all of which are part of the Amygdaloideae subfamily. Prunus L falls within the Amygdaleae tribe, whereas Clades II and III are grouped into the Maleae tribe, forming a cohesive monophyletic cluster. These clades, representing the Maleae, stand as sister groups. Moving to Clades IV through VIII, they comprise Rubus L., Sanguissorba L., Rosa L., Argentina L., Fragariastrum L., Dasiphora L., and Fragaria L., all under the Rosoideae subfamily. Within Clade IV, Rubus L. is categorized under the tribe Rubeae and bifurcates into two branches, each supported by a 100% bootstrap value. Clade IV and Clades V to VIII share a sister group relationship. Clade V is uniquely defined by a monophyletic lineage of Sanguissorba. Clade VI features Rosa L. of the Roseae tribe, with its internal nodes primarily exceeding an 89% bootstrap value, barring one at 69%. Clades VII and VIII, belonging to the Potentilleae tribe, include Argentina L., Fragariastrum L., Dasiphora L., and Fragaria L., with Clade VI being a sister group to these clades (Fig. 8).

This study sequenced and compared the complete cp genomes of 19 Rosaceae species from Tibet. The results indicated that the 19 cp genomes exhibited a conserved gene order and gene content, and displayed a typical quadripartite structure (LSC, SSC, IRa, and IRb). These cp genomes are similar in size, structure, and gene content to previously published Rosaceae cp genomes [50–52].

The cpSSRs are characterized by maternal inheritance, abundant polymorphism, and good reproducibility in plants, as a result, they are frequently employed as molecular markers for species identification, and population genetics research [53–55]. The distribution of SSRs in the cp genomes of plants is uneven. The number of detected SSR loci in the 19 Rosaceae species ranged from 52–121, with the largest number of single nucleotide repeats. Among them, 46–96 SSRs were distributed in the LSC region, 5–12 SSRs in the SSC region, and 0–15 SSRs in the IR region, while the IR region of D. fruticosa lacked SSR distribution. In line with earlier research, there were much more SSRs in the LSC region than in the SSC and IR regions [56]. Long repetitive sequences may accelerate cp genome rearrangements and increase population genetic diversity [56]. A total of 32–77 long repetitive sequences were identified in all 19 species, including three types, i.e., forward (F), palindrome (P), and reverse (R).

Plants exhibit a preference for synonymous codon usage [57]. Codon usage bias is the result of long-term evolution and adaptation to the environment [58], and therefore is related to phylogenetic relationships [53]. The 19 Rosaceae plants’ cp genomes exhibit comparable overall codon use patterns and a propensity for A/U-end synonymous codons. The frequent A/T traits seen in the cp genomes of angiosperms are consistent with this [59,60].

Genes containing a single intron are widely present in organisms. The infA gene is a gene with a single intron. However, among the 19 species studied, the infA gene was only found in the cp genome of R. pedunculosus. The infA gene is unstable and is easily lost from the cp genome of angiosperms and transferred to the nucleus [61]. Similarly, the atpF gene is also a gene with a single intron, and it is only found in C. adpressus, C. multiflorus, C. wardii, S. koehneana, and P. trichostoma. The complete atpF gene has been discovered in some cp genomes of the Amygdaloideae subfamily, while the loss of the atpF gene has also been observed in some genera of the Rosaceae [62]. The loss of the atpF gene was widespread in the 19 cp genomes of Rosaceae family.

The cp genome may be utilized to research phylogenetic categorization and genome evolution among plant lineages since variations in the IR region’s size are frequently caused by these changes [63]. The length of IR region and LSC region was similar in most of the 19 species. However, three species in the genus Cotoneaster, as well as S. koehneana, have significantly longer IR and LSC regions than species in other genera. Additionally, these four species all possess the atpF gene, while the atpF gene is missing in six of the remaining seven genera. The loss of the atpF gene is one of the reasons for the differences in the sizes of the IR and LSC regions in the cp genomes of Rosaceae species.

The IRb/SSC (JSB) boundary in these 19 cp genomes is positioned between the ycf1 and ndhF genes, and it is consistently spanned by the ycf1 gene. However, the distance between the ndhF gene and the JSB boundary varies. For example, among the seven species of Rosa, ndhF gene is located less than 100 bp away from this boundary in six species, while in the three species of Cotoneaster, ndhF gene spans across the boundary by 12 bp. In the two species of Rubus, the ndhF gene is approximately 30 bp away from the boundary. The results above indicate that ndhF exhibits strong genus-specific characteristics and may be used to determine the correct classification of genera within the Rosaceae family.

Nine areas (pi > 0.13) with high values were identified in 19 different species. Four of them were in the SSC region and five of them were in the LSC region. As further evidence of the IR region’s greater conservation than the LSC and SSC regions, the SCS region had the greatest pi value and the IR region the lowest. The nucleotide diversity of the IR area did not contain any substantially divergent sequences, suggesting that these regions are largely conserved [64].

A common method for assessing nucleotide evolutionary rates and selection pressure locations in coding sequences is the ratio of dN/dS [65]. Due to the fact that infA is the sole gene that has experienced positive selection, as shown by a Ka/Ks value more than 1.

According to years of morphological and molecular research, the Rosaceae family can be divided into three subfamilies, namely the redefined Rosoideae, expanded Amygdaloideae, and the newly separated Dryadoideae which was formerly a part of Spiraeoideae. However, the relationships between these three subfamilies are still uncertain. Studies suggested that Dryadoideae could most likely be placed at the base of the entire Rosaceae family, or serve as the sister group of Rosoideae or Amygdaloideae [8]. Within the framework of these three subfamilies, a new system for Rosaceae family was proposed by Potter et al. [7]. In 2016, Xiang et al. [8] had further refined the classification within the Rosaceae family under the Potter system, positioning the Roseae at the forefront of the Rosoideae subfamily’s divergence, subsequently followed by the Agrimonieae. Contrary to this arrangement, the present study had proposed an alternative order. Notably, the Potter system did not delineate the Agrimonieae as a separate entity.

Prunus trichostoma belongs to the Amygdaleae tribe of Prunus genus. There have been two ways to classify Amygdaleae tribe. One is only composed of the “Genus Prunus sensu lato”, which is further divided into sub. Genus, including subg. Prunus, subg. Amygdalus, subg. Cerasus, subg. Padus, and subg. Laurocerasus. However, it is not yet clear whether Maddenia and Pygeum belong to the “Genus Prunus sensu lato”. The other way is to divide “Genus Prunus sensu lato” into smaller genera that make up the Amygdaleae tribe. The previously popular Rydberg system follows the former way. Due to technological limitations, early molecular studies usually constructed phylogenetic relationships based on a single sequence. Except that the subgenus in the Rydberg system were not monophyletic groups, and Maddenia and Pygeum were also embedded within “Genus Prunus sensu lato”, making it difficult to differentiate the various evolutionary lineages. In addition, the inconsistencies between phylogenetic trees constructed from cp and nuclear genomes have led taxonomists to prefer to maintain a “Genus Prunus sensu lato”, so that the Amygdaleae tribe has only one genus and no sub. Genus. Some researchers separate Cerasus as a genus [66]. However, APG Ⅳ systems maintain the concept of “Genus Prunus sensu lato” in the end, which is consistent with our analysis based on ML and BI trees. Besides, it is appropriate to include P. trichostoma in the “Genus Prunus sensu lato”.

The taxonomic status of Argentina Hill, Dasiphora Raf., and Fragariastrum Heist. ex Fabr. has long been a subject of debate. Initially classified within the genus Potentilla, they were later separated to form independent genera [67–69]. The species of Fragariastrum, Dasiphora, and Argentina clustered together, aligning with the taxonomic relationship in the APG IV system. A. lineata and F. eriocarpum formed Clade VII, while Dasiphora and Fragaria formed Clade VIII, indicating a closer relationship between Fragariastrum and Argentina. Therefore, the genera Argentina, Fragariastrum, and Dasiphora should be separated from Potentilla.

The investigation encompassed 19 species within the Rosaceae family. Hence, to deepen our comprehension of the phylogenetic interrelations within this family, it is imperative to extend the research to a broader array of species.