Background & Summary

Rosaceae, an economically significant family of angiosperms, is renowned for its abundant edible and ornamental plants1. While high-quality genome sequences are accessible for several edible genera such as Malus, Prunus, and Fragaria, genomic resources for many ornamental plants remain limited2. Whole genome sequences represent a fundamental resource that underpins a broad range of molecular and genetic studies. Deciphering the complete DNA sequence of ornamental plants through genome sequencing facilitates the identification and characterization of genes associated with key traits relevant to landscaping, as well as the genetic basis of environmental adaptability and disease resistance3. Such insights are helpful to the molecular studies associated to the landscape applications, while enhancing our understanding of genetic diversity for conservation purposes.

Dasiphora, commonly known as shrubby cinquefoils, is a small shrub genus belonging to the tribe Potentilleae within the subfamily Rosoideae4. This genus comprises two morphologically distinct species: the yellow-petaled Dasiphora fruticosa (Fig. 1) and the white-petaled Dasiphora glabra. Both species were previously classified under Potentilla but have been recognized as an independent genus based on molecular phylogeny5,6,7,8. The presence of woody plants distinguishes Dasiphora from other genera of Potentilleae, facilitating its wide range of landscape applications.

Fig. 1
figure 1

Photographs illustrating the morphology and habitats of Dasiphora fruticosa. Photo credit: Xu Zhang and Juntong Chen.

Full size image

D. fruticosa is widely distributed across cool temperate and subarctic regions, spanning altitudes from 400–5000 m in the Qinghai-Tibet Plateau (QTP)4. High-altitude environments in the QTP are characterized by freezing temperatures, high ultraviolet (UV) radiation, and hypoxia. Genome-wide studies of alpine plants revealed genomic signatures of high‐altitude adaptation often involving pathways related to abiotic stress response and biosynthesis of secondary metabolites9,10,11,12. However, due to the limited availability of alpine plant genome sequences and the restricted distribution of most species to high-altitude regions only9,10,11,12, our understanding of adaptive evolution and speciation along altitudinal gradients remains constrained. Given its extensive elevation range, D. fruticosa presents an ideal model for investigating genetic adaptations along altitudinal gradients13,14.

In this study, we generated a high-quality chromosome-scale genome assembly of D. fruticosa by integrating Oxford Nanopore, Illumina, and high-throughput chromosome conformation capture (Hi-C) technologies (Table 1). The genome size of D. fruticosa was estimated to be 264.18 Mb using flow cytometry (Fig. 2a,Table S1) and 264.14 Mb via 19 k-mer analysis (Fig. 2b, Table S2). The ploidy level was determined to be diploidy (Fig. 2c). We obtained 52 high-quality contigs (contig N50 = 14.01 Mb), with a total assembly size of 250.21 Mb and anchored 249.23 Mb onto seven pseudochromosomes using Hi-C data (Fig. 2d). The D. fruticosa genome contains 35.73% repetitive sequences (89.04 Mb), of which long terminal repeats (LTRs) and DNA transposons account for 13.14% and 8.51%, respectively (Table S3). We then predicted 32,613 protein-coding gene (PCG) models (Fig. 3), with functional annotation available for 30,643 PCGs (93.96%) by aligning against public protein databases (Table 1). Our genome assembly provides a valuable genomic resource for molecular studies of shrubby cinquefoils, and genetic basis of high-altitude adaptation in future studies.

Table 1 Statistics and comparison of genome assembly and annotation of Dasiphora fruticosa.
Full size table
Fig. 2
figure 2

Genome size estimation and chromosome assembly of Dasiphora fruticosa. (a) Flow cytometry results using Solanum lycopersicum (~900 Mb) as an internal reference for genome size estimation. (b) Distribution of k-mer depth and frequency at 19 k-mer for D. fruticosa genome. (c) Visualizations of k-mer counts output from Smudgeplot: the AB (diploid) k-mer pairs were most prominent (0.65) in the genome. (d) A Hi‐C interaction heatmap of D. fruticosa genome showing the interactions among seven pseudochromosomes.

Full size image
Fig. 3
figure 3

The genomic landscape of seven pseudochromosomes of Dasiphora fruticosa. Circles from outside to inside are chromosome length (A), gene density (B), GC content (C), total LTR density (D), LTR-Gypsy density (E), LTR-Copia density (F), and syntenic blocks across chromosomes (G).

Full size image

Methods

Plant material, DNA extraction and sequencing

Fresh young leaves were collected from a mature D. fruticosa plant at the Kunming Botanical Garden (Kunming, China) (102.748° E, 25.144°) and were sent to Wuhan Benagen Technology Company Limited (Wuhan, China) for genome sequencing. The voucher specimen (accession number: KUN1604678) was collected and preserved in the Herbarium of Kunming Institute of Botany (KUN). Genomic DNA was extracted using the Qiagen DNeasy Plant Mini Kit. Short-read sequencing libraries with an insert size of 500 bp were constructed using the Illumina TruSeq PCR-free HT (Illumina, San Diego, CA, USA) and were sequenced for paired-end 150 bp reads using an Illumina HiSeq. 4000 platform. For Oxford Nanopore sequencing, high-quality genomic DNA was extracted from fresh leaves using the CTAB method15. DNA quality and concentration were assessed using the NanoDrop One spectrophotometer (Thermo Fisher Scientific) and the Qubit 3.0 Fluorometer (Life Technologies, Carlsbad, CA, USA). The libraries were then constructed using the SQK-LSK109 ligation kit following the standard protocol. The purified library was loaded onto primed R9.4 Spot-On Flow Cells and was sequenced using a PromethION sequencer (Oxford Nanopore Technologies, Oxford, UK). For Hi-C sequencing, extracted DNA was first crosslinked with 40 ml of 2% formaldehyde solution to capture interacting DNA segments. Subsequently, the crosslinked DNA was digested with the DpnII restriction enzyme, and libraries were constructed and sequenced using an Illumina HiSeq. 4000 instrument with paired-end 150 bp reads.

For transcriptome sequencing, fresh tissue samples including stem, leaf, and flower were collected with three copies from the same D. fruticosa plant and immediately frozen in liquid nitrogen. Total RNA was subsequently extracted using the TRIzol® Reagent (Invitrogen, Shanghai, China). The concentration and quality of RNA were assessed using a NanoDrop 2000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and a Bioanalyzer 2100 system (Agilent Technologies, CA, USA). Subsequently, paired-end cDNA libraries were prepared from mRNA enriched with anti-polyA magnetic beads, fragmented, circularized, and then subjected to PE150 sequencing on the Illumina HiSeq. 4000 platform.

Genome size estimation

The genome size D. fruticosa was estimated using two methods. First, flow cytometry was performed using Solanum lycopersicum (~900 Mb) as an internal reference, following a step-by-step protocol described by Pellicer and Leitch (2014)16. Briefly, fresh leaves of D. fruticosa and S. lycopersicum were chopped together in LB01 buffer. Nuclei were isolated, passed through a 30 mm nylon mesh, and stained with 50 mg/ml propidium iodide (PI) and 50 mg/ml ribonuclease. Samples were then analyzed on a BD FACScalibur flow cytometer, with approximately 10,000 nuclei per sample collected for three repetitions. Modifit v3.0 software was used to calculate the ratio and plotting the histogram. The genome size was estimated to be 264.18 Mb based on flow cytometry (Fig. 2a, Table S1).

Secondly, a k-mer based analysis was conducted using GCE v1.0.017. The 17 and 19 k-mer frequency-depth distribution was calculated using Jellyfish v2.2.1018 with Illumina short reads. The analysis indicated a genome size of 264.14–264.34 Mb (Fig. 2b, Table S2), consistent with the flow cytometry result. In addition, considering observed ploidy level variation in wild populations of D. fruticosa19, we estimated ploidy level of our sequenced material using Smudgeplot v0.2.520 based on the k-mer frequency output from Jellyfish. The analysis showed that AB k-mer pairs were predominant (0.65) in the genome, indicating the sequenced individual is diploid (Fig. 2c).

Genome assembly and quality control

NextDenovo v2.3.1 (https://github.com/Nextomics/NextDenovo) was employed to correct and assemble the Nanopore long reads with the following parameters: “read_type = ont, read_cutoff = 1k, genome_size = 250 m, sort_options = -m 20 g -t 15, and minimap2_options_cns = -t 8”. The initial assembly was then polished using NextPolish v1.4.121, incorporating both Illumina short reads and Nanopore long reads. After polished, the contig assembly had a total size of ~ 250.21 Mb, with a contig N50 value of 14.01 Mb (Table 1). Next, Hi-C data were used for chromosome mounting and assembly of contigs based on the 3D-DNA pipeline22 with default settings. Manual error correction was executed using Juicerbox v2.2022. A total of 249.23 Mb of the genome sequences was successfully anchored onto seven pseudochromosomes (Fig. 2d). The completeness of the assemblies was subsequently assessed by BUSCO (Benchmarking Universal Single-Copy Orthologs) assessments based on the eukaryota_odb10 database23. A total of 98.9% (94.5% single-copy BUSCOs) completeness was revealed by the analysis (Table 1). To evaluate the assembly quality, the Illumina short-reads were mapped to the assembled genome sequence using BWA v0.7.1724 with default settings, and the mapping quality score was calculated using Qualimap2 v2.2.125. A mapping rate of 94.24% and a mapping quality score of 51.15 were estimated by the analysis (Table 1).

Genome annotation

Repetitive sequences were annotated based on de novo and homology-based strategies. RepeatModeler2 v2.0.1026 was used to generate a de novo repeat library using default parameters. RepeatMasker v4.0.727 was used to run a homology search for known repeat sequences against the Repbase database. Then, RepeatMasker was employed to merge the library files from both methods and to identify the repeat contents with the parameter “-e rmblast”. The result showed that D. fruticosa genome comprises 35.73% repetitive sequences totaling 89.04 Mb, with LTRs and DNA transposons constituting 13.14% and 8.51%, respectively (Table 1 and S3).

For gene structure annotation, a combination of de novo-, homology-, and transcript-based methods were applied. RNA-seq reads were assembled using Trinity v2.1.128 with the parameter “–genome_guided_max_intron 10000”, after which coding DNA sequences (CDS) were predicted using TransDecoder v5.7.129 with default settings. De novo gene prediction was carried out with Braker2 v2.1.530. In the Braker2 run, paired-end RNA-seq reads were aligned to the D. fruticosa genome sequence using HISAT231 with default settings, and the assembled transcripts were used for training gene models in Braker2. For the homology-based method, GeMoMa v1.6.132 was employed to predict homologous sequences by mapping protein sequences of Fragaria vesca, Prunus persica, Rosa chinensis, Rubus occidentalis, Arabidopsis thaliana, and Vitis vinifera. The parameters of GeMoMa analysis were as follows: “tblastn = true, GeMoMa.Score = ReAlign, and AnnotationFinalizer.r = NO”. Finally, predicted gene models from the three methods were merged to produce consensus models using EVidenceModeler v1.1.133 following the manual provided by the authors (https://github.com/EVidenceModeler/EVidenceModeler/wiki). A total of 32,613 predicted gene modules were obtained. To validate the predicted genes, RNA-seq reads were aligned to the genome sequence of D. fruticosa using HISAT2. The reads mapped to gene regions were then quantified using featureCounts v2.0.634. The BUSCO completeness of predicted gene models was evaluated against embryophyta_odb10 datasets under the protein mode. The result showed 97.1% (92.9% single-copy BUSCOs) BUSCOs complete matched (Table 1). The Circos tool (http://www.circos.ca) was utilized to visualize gene density, GC content, and repeat content on each pseudochromosome (Fig. 3).

For the functional annotation, predicted PCGs were aligned to the UniProt and NCBI nonredundant (NR) protein databases using BLASTP (BLAST + v2.12.035) with e-value ≤ 1e-5. The functional domains of protein sequences were subsequently identified by InterProScan v5.51 based on the Pfam database36 and KEGG and GO terms of PCGs were obtained using eggNOG-mapper v2.0.137 with default settings. A total of 30,643 predicted genes (93.96%) were functionally annotated by aligning against public protein databases (Table 1).

Comparative genomic analyses

The OrthoVenn338 platform was used to identify orthologous clusters for subsequent phylogenetic and gene family contraction and expansion analyses. Nine Rosaceae species (Malus domestica, Pyrus bretschneideri, P. persica, Argentina anserina, Potentilla micrantha, F. vesca, D. fruticosa, R. chinensis and R. occidentalis), along with three outgroup species (Cannabis sativa, A. thaliana, and V. vinifera) were included. In the OrthoVenn3 analysis, highly conserved single-copy genes were aligned using MAFFT v7.52039 and trimmed by Trimal v1.240. FastTree v2.141 was used to infer the phylogenetic tree using maximum likelihood method. The SH test method was used to assess the credibility of each phylogenetic node. Divergence time was estimated based on the TimeTree5 database (http://www.timetree.org/) using MCMCTree in PAML v4.942. Subsequently, CAFÉ543 was used to estimate the expansion or contraction of gene families. Gene families with contractions and expansions were subjected to GO annotation and enrichment analysis in OrthoVenn3, matched against the UniProt database. All above analyses were implemented in the OrthoVenn3 platform and the parameters were set to defaults.

Phylogenetic results showed that Dasiphora exhibits a closer relationship to Fragaria than to Potentilla, with an estimated divergence time of approximately 15.8 million years ago (Fig. 4). Subsequently, our analysis identified 202 gene families that had contracted and 120 that had expanded (Fig. 4). Gene Ontology (GO) enrichment analyses revealed that these expanded gene families are primarily associated with response to oxidative stress, photosynthesis light reaction, response to UV-B, response to cold, and protein phosphorylation (Table S4). These pathways are primarily associated with abiotic stress response and likely play crucial roles in the adaptation of D. fruticosa to extremely cold, high UV radiation, and hypoxic environments of high altitudes.

Fig. 4
figure 4

Phylogenomic results among nine Rosaceae species. Gene family expansion (orange) and contraction (blue) information is labeled on the tree.

Full size image

Data Records

Oxford Nanopore, Hi-C sequencing, Illumina, and RNA-seq data generated in this study have been deposited at the NCBI Sequence Read Archive (SRA) under Bioproject PRJNA113109844 and the China National GeneBank Database (CNGBd, https://db.cngb.org/) with accession number CNX0946916-CNX094691945,46,47,48. The genome assembly has been deposited at GenBank under the accession JBEWQC00000000049. The annotation files of the genome are available at the Figshare database: https://doi.org/10.6084/m9.figshare.25272223.v150.

Technical Validation

Genome assembly validation

The quality of the D. fruticosa genome assembly was evaluated using three methods. First, both flow cytometry and a k-mer based estimation of genome size were very similar. A final genome assembly size also revealed a close match to the estimated genome size. Secondly, the BUSCO assessment of the genome assembly indicated a high level of completeness, with 98.9% (94.5% single-copy BUSCOs) complete matches to the embryophyta_odb10 dataset. Finally, Illumina data mapping was employed to assess assembly quality. The results showed a mapping rate of 94.24% and a mapping quality score of 51.15 (corresponding to a base accuracy of 99.999%), suggesting that the genome is of high quality.

Genome annotation validation

Genome annotation was conducted using transcript-based, de novo, and homology-based prediction methods. These methods led to the prediction of 32,613 gene models. RNA-seq read mapping revealed that 29,280 genes (95.55% of all predicted genes) were detected with mapped reads in at least one of the three tissues. Functional annotation was available for 30,643 genes (93.96%) through alignment with public protein databases. The BUSCO analysis showed that the annotated protein coding genes of the D. fruticosa genome had a complete match rate of 97.1% (92.9% single-copy BUSCOs) with the embryophyta_odb10 dataset. In addition, our phylogenetic result based on PCGs was consistent with previous studies supporting a closer relationship between Dasiphora and Fragaria than with Potentilla5,6,7,8. Functional analysis of expanded gene families revealed genomic signatures of high-altitude adaptation involving pathways related abiotic stress response, in line with previous findings9,10,11,12.

Comparison with a published D. fruticosa genome

We further compared the statistics of the D. fruticosa genome assembly and annotation from our present study with those of a recent study14 (Table 1). Overall, both assemblies demonstrated similar quality, with our genome assembly showing higher completeness and continuity. Specifically, our genome assembly exhibited longer contig length and anchored genome length, along with higher BUSCO completeness compared to the published D. fruticosa genome. Moreover, our study provided flow cytometry data to validate the genome size and estimates ploidy level of our sequenced material. These results are crucial for ensuring the reliability of our genome assembly, particularly in light of observed ploidy level variation in wild populations of D. fruticosa19.