A high-quality chromosome-level genome assembly and annotation of the medicinal orchid Cremastra appendiculata (Orchidaceae)

Abstract

Cremastra appendiculata (D. Don) Makino, an endangered terrestrial orchid with significant medicinal value, lacks comprehensive genomic resources to elucidate its evolution and conservation genetics. Here, we present a chromosome-level genome assembly for C. appendiculata, generated by integrating PacBio circular consensus sequencing and Hi-C technologies. The final genome assembly totals 2.40 Gb, achieving contig and scaffold N50 values of 27.47 Mb and 98.06 Mb, respectively. Hi-C scaffolding anchored 98.66% of the assembly to 24 chromosomes, with repetitive elements comprising 83.85% of the genome and a total of 31,703 protein-coding genes predicted, of which 29,094 (91.77%) were functionally annotated. This high-quality, chromosome-level reference genome provides a foundational resource for understanding the population genetic structure, adaptive evolution and speciation mechanisms of C. appendiculata, thereby offering valuable insights into its evolutionary history and conservation.

Background & Summary

The family Orchidaceae, renowned for extraordinary species diversity and vital ecological roles, is a treasure trove of medicinal plants historically revered in traditional herbal medicine. Medicinal orchids have been integral to pharmacopeias across Asia, Europe, Australia, America, and Africa, with species such as those within Dendrobium, Gastrodia, and Bulbophyllum exhibiting bioactive compounds like polysaccharides, alkaloids, and phenolics¹. These compounds demonstrate therapeutic potential in treating inflammation, cancer, neurodegenerative disorders, and metabolic syndromes². Many studies have validated the efficacy of orchid-derived molecules, such as gastrodin from Gastrodia elata Blume in neuroprotection³ and polysaccharides from Dendrobium nobile Lindl. in immunomodulation⁴. However, these orchids face severe conservation challenges due to inherently low natural propagation rates stemming from their obligate mycorrhizal dependency. Orchid seeds, extremely small and devoid of endosperm, depend on symbiotic fungi for nutrient acquisition during germination and early seedling growth⁵. For example, the saprophytic orchid, G. elata, relies on fungal symbiosis throughout its entire growth period⁶. The conservation of medicinal orchids remains critical due to the biological bottleneck exacerbated by rampant overharvesting for commercial trade⁷. To achieve the balance between clinical demands and wild population conservation, it is essential to unravel the molecular mechanisms governing growth, development, and secondary metabolite biosynthesis. Therefore, sustainable cultivation practices and biotechnological innovations will effectively facilitate artificial propagation, enhance phytochemical yields, and ensure reliable medicinal supply chains.

Cremastra appendiculata, commonly known as “山慈菇 (shān cí gū)” in traditional Chinese medicine (Fig. 1a-b), is a shade-adapted perennial terrestrial orchid endemic to Asia, and epitomizes the intersection of ecological fragility and therapeutic promise^8,9. Its pseudobulbous rhizomes (Fig. 1c) biosynthesize and accumulate approximately 190 compounds, including phenanthrene, dihydrophenanthrene, bibenzyl, flavonoids, terpenoids, glycosides, and aromatics, which exhibit a wide range of pharmacological activities such as antitumor, antibacterial, and antioxidant⁸. It has been traditionally used for treating rheumatoid arthritis, abscesses, snake bites, and lymphatic disorders, and more recently, pharmacological studies have validated its efficacy in reducing tumor necrosis factor-alpha (TNF-α) levels and inhibiting cyclooxygenase-2 (COX-2) activity¹⁰. Despite its medicinal value, C. appendiculata suffers from low germination rates and slow vegetative growth, requiring 4–5 years to mature⁸. To enhance seed germination and seedling vigor, it requires symbiotic cultivation with specific fungal partners, primarily from the Psathyrellaceae^11–14. Recent research has focused on optimizing ex situ conservation through micropropagation techniques, including asymmetric somatic embryogenesis and temporary immersion bioreactors, to produce pathogen-free plantlets^15,16. In vitro studies reveal that mycorrhizal fungi, particularly species of Coprinellus, enhance seed germination rates and seedling vigor^13,14,17. These efforts underscore a core research strategy that integrates ecological knowledge and biotechnological innovation to balance protection and utilization. This fundamentally relies on genomic and molecular knowledge to elucidate the growth regulation, stress response, and metabolite biosynthesis in C. appendiculata.

Fig. 1.

Morphology and karyotype of a two-year-old C. appendiculata plant. (a) Plant during flowering stage. (b) Plant during seed-bearing stage. (c) Pseudobulb with roots. (d) Metaphase chromosomes from root-tip cells. (e) Karyotype analysis showing a diploid chromosome number of 2n = 2x = 48.

Advanced genomic and transcriptomic analyses have unveiled genetic determinants underlying the unique biology of C. appendiculata, providing critical insights for conservation and cultivation. Chloroplast genome sequencing revealed that conserved genes, such as rbcL, ycf1, and accD, are correlated with its adaptive evolution in response to environmental factors^18,19. Transcriptomic analyses of the symbiotic interaction between C. appendiculata and its mycorrhizal fungus identified key genes related to symbiotic germination, such as the DHN gene family, Xero1, ZEP and NCED3²⁰. Differentially expressed gene (DEG) analysis suggested that IPT, CYP735A, CCD7/8, WUS, and BRC1 play critical roles in shoot branching regulation²¹. Genomic and transcriptomic data of C. appendiculata provide insights into the evolution of the biosynthetic pathway of colchicine, a pharmaceutically active compound²². Despite these advances, critical gaps still remain, as system-level regulatory networks are still fragmented, and genotype-phenotype linkages are poorly defined. Functional validation of candidate genes involved in symbiotic seed germination and clonal branch growth remains at a preliminary stage, largely due to the lack of a high-quality reference genome for C. appendiculata. Meanwhile, population genomics studies aimed at assessing its genetic diversity and local adaptation to facilitate cultivation are also limited.

To address these limitations, we present a chromosome-level genome assembly of C. appendiculata leveraging PacBio High Fidelity (HiFi) long-read sequencing and Hi-C scaffolding. The assembly achieved a contig N50 of 27.47 Mb and a scaffold N50 of 98.06 Mb, compared with 0.68 Mb and 80.16 Mb in the earlier draft²². A total of 98.66% of the assembled sequences were anchored to 24 chromosomes, compared with 87.21% previously. The assembly completeness assessment showed 1,539 complete BUSCOs (95.35%), which is significantly higher than the 86.4% completeness reported in the prior genome²². We identified 31,703 protein-coding genes and 20,439 non-coding RNAs, representing approximately a 1.51-fold and 4.89-fold increase, respectively, relative to the 20,991 protein-coding genes and 4,181 non-coding RNAs previously annotated. Repetitive sequences accounted for 83.85% of the genome, substantially higher than the 59.15% observed in the previous assembly, reflecting the improved resolution of repetitive regions in the chromosome-level assembly (Table 1). By utilizing more advanced sequencing technologies and assembly strategies, we generated a chromosome-level genome assembly for C. appendiculata of significantly higher quality than reported earlier. This enhanced genomic resource provides a robust foundation for future basic biological and application research on C. appendiculata.

Table 1.

Statistics of the genome assembly of C. appendiculata.

Genomic features		This study	Previous assembly22
Contigs	Total number	347	16,208
	Total length (bp)	2,404,494,502	2,362,341,580
	N50 length (bp)	27,470,550	677,859
	N90 length (bp)	6,998,514	60,438
	Maximum length (bp)	104,288,895	11,563,929
	GC content (%)	39.00	39.39
Scaffolds	Total number	30	8,767
	Total length (bp)	2,404,653,002	2,365,912,979
	N50 length (bp)	98,055,964	80,155,791
	N90 length (bp)	78,414,824	146,279
	Maximum length (bp)	141,062,403	117,118,055
	GC content (%)	39.00	39.39
Chromosomes		24	24
Hi-C anchored rate (%)		98.66	87.21
Assembly size (Gb)		2.40	2.36
BUSCO completeness (%)		95.35	86.4
Protein-coding genes		31,703	20,991
Non-coding RNAs		20,439	4,181
Repeat Content (%)		83.85	59.15

Methods

Plant material

The plants of C. appendiculata were sourced from cultivated populations, derived from seeds of a single maternal plant, growing under natural environmental conditions (with minimal anthropogenic intervention) on Jinfo Mountain in Chongqing, China (29°3′4″ N, 107°18′29″ E). Two individual plants were harvested on October 27, 2024, with one designated for DNA and Hi-C sequencing and the other for RNA sequencing. Immediately after collection, the samples were flash-frozen in liquid nitrogen, transported to the laboratory, and stored at −80 °C. Plants for karyotype analysis were transplanted and cultivated in the greenhouse of Beijing Forestry University (40°00′38″ N, 116°20′38″ E) with natural light and weekly watering.

Karyotype analysis

Root tips (0.5 cm) were collected in the morning and pretreated in a pre-fixation solution of saturated p-dichlorobenzene and 0.1 M 8-hydroxyquinoline (50:1) on ice for 5 min, then fixed at 4 °C for 4 h. After rinsing with distilled water, the samples were treated in 0.075 M KCl at 4 °C for 30 min, rinsed again, and fixed in Carnoy’s fluid (methanol:glacial acetic acid, 3:1) at 4 °C for 24 h. Hydrolysis was then performed in 1 M HCl at room temperature for 45 min, followed by enzymatic digestion with 6% cellulase and 4% macerozyme (2:1) at 37 °C for 5 h. The digested material was rinsed and treated in distilled water at 37 °C for 30 min.

The treated root tips were transferred onto a chilled clean slide, gently crushed with forceps to remove debris, stained with Carbol fuchsin for 2 h, and covered with a coverslip. Excess dye was removed with filter paper, and chromosomes were evenly spread by tapping the coverslip with a pencil eraser. Metaphase cells were observed with a Nikon 80i microscope, and well-dispersed chromosomes were imaged. The micrographs were analyzed using Zeiss Karyotype software for segmentation, pairing, and karyotyping. The results revealed that C. appendiculata is a diploid species with a chromosome number of 2n = 2x = 48 (Fig. 1d-e).

DNA extraction and sequencing

DNA was extracted from fresh leaves of C. appendiculata using a magnetic bead-based method. Leaves were flash-frozen in liquid nitrogen, ground to a fine powder, and lysed with 3 mL of Tissue Lysis Buffer 1 (1% PVP, 1% sodium metabisulfite, 0.5 M NaCl, 0.1 M Tris-HCl, 0.05 M EDTA, and 1.25% SDS) at 65 °C for 30 min. After cooling, the lysate was centrifuged at 4 °C and 12,000 × g for 10 min. The supernatant was mixed with 1 mL of Settling Buffer 2 (5 M potassium acetate), incubated at 4 °C for 5 min, and centrifuged. The supernatant was mixed with 900 µL lysate, 900 µL DNA Binding Buffer (20% PEG 8000 and 3 M NaCl), and 50 µL Magnetic-Bead Buffer (2% carboxylate-modified SpeedBeads, 18% PEG 8000, 1 M NaCl, 0.01 M Tris-HCl, 1 mM EDTA and 0.05% Tween-20), then briefly centrifuged, and incubated at room temperature for 5 min for DNA binding. Magnetic separation was performed until the supernatant cleared. After supernatant removal, the beads were washed twice with 75% ethanol and air-dried. DNA was eluted with 200 µL of preheated Elution Buffer (10 mM Tris-HCl, pH 8.5, 50 °C).

The extracted genomic DNA was assessed for concentration, integrity, and purity. High-quality DNA was sheared using the Megaruptor system, and 20 kb fragments were size-selected using the Sage ELF instrument. After damage and end repair, PacBio adapters containing hairpin structures were ligated to generate circularized SMRTbell libraries. Linear and damaged DNA molecules were removed by enzymatic digestion, and libraries within the target size range were further refined using the automated Sage Science instrument. A 20 kb insert size library was constructed and sequenced in circular consensus sequencing (CCS) mode on the PacBio Revio system using 2 cells, generating a total of 92.60 Gb of high-quality CCS data with an N50 of 16,787 bp (Fig. 2a). Based on an estimated genome size of 2,342 Mb by genome survey, the sequencing depth was approximately 39.54×.

Fig. 2.

Characteristics of the C. appendiculata genome assembly. (a) Length distribution of HiFi clean reads. (b) 21-mer frequency distribution spectrum from a diploid genome analysis. (c) Hi-C contact heatmap illustrating interactions across the 24 chromosomes. (d) Circos plot summarizing genomic features of the assembly. The plot from outer to inner rings depicts: chromosomes, gene density, overall transposable elements (TEs) density, Copia LTR retrotransposon density, Gypsy LTR retrotransposon density, GC content, HiFi reads depth, and collinearity blocks. Tracks II–VIII are calculated in 500 kb sliding windows.

RNA extraction and sequencing

Total RNA was extracted from fresh leaf blade, petiole, pseudobulb, root, coralloid rhizome, and scape tissues of a C. appendiculata individual using a magnetic bead-based protocol. Tissues were ground and lysed in CTAB buffer supplemented with β-mercaptoethanol, followed by incubation at 65 °C for 12 min and centrifugation. The supernatant was mixed with 800 µL of extraction solution and centrifuged again. The clarified supernatant was bound to magnetic beads and purified using an automatic nucleic acid extractor. mRNA was enriched using oligo(dT)-attached magnetic beads, fragmented, and reverse-transcribed into double-stranded cDNA. After end repair, adenylation, and adapter ligation, libraries were amplified by PCR. Circularized single-stranded DNA was then generated and subjected to DNA nanoballs (DNBs) formation followed by sequencing on the DNBSEQ platform using combinatorial Probe-Anchor Synthesis (cPAS) technology. Following data filtering with SOAPnuke²³ (v2.1.9, parameters: -n 0.01 -l 20 -q 0.4 --adaMR 0.25 --polyX 50 --minReadLen 150), a total of 43.21 Gb of clean data derived from six tissues were utilized for downstream gene annotation.

Hi-C library preparation and sequencing

Hi-C libraries were constructed from cross-linked leaf tissue of C. appendiculata using formaldehyde fixation followed by cell lysis and chromatin digestion with MboI. The digested fragments were biotin-labeled, proximity-ligated, and de-crosslinked, then purified and treated with T4 DNA polymerase to remove unligated biotin marks. The purified DNA was fragmented via Covaris, and biotinylated fragments were captured with streptavidin-coated magnetic beads. Libraries were prepared through end repair, adenylation, adapter ligation, and PCR amplification, followed by circularization to generate single-stranded DNA templates. Single-stranded circular DNA was amplified via rolling circle replication to produce DNBs, each containing multiple copies of the original template. High-quality DNBs were then deposited into patterned nanoarrays using a high-intensity DNA nanochip technique and sequenced by cPAS on the DNBSEQ platform, and a total of 296.42 Gb of raw data was obtained. After filtering out adapter sequences and low-quality reads using SOAPnuke (parameters: -n 0.02 -l 20 -q 0.4 -i -G 2 --polyX 50 -Q 2 --seqType 0), 293.66 Gb of high-quality clean data were retained.

Genome assembly

The genome size of C. appendiculata was estimated through k-mer spectrum analysis (k = 21) using DNBSEQ short reads. K-mer frequency distribution was generated with Jellyfish²⁴ (v2.2.6) and analyzed through GenomeScope²⁵ (v2.0), yielding an estimated genome size of 2.34 Gb with 0.988% heterozygosity (Fig. 2b). For genome assembly, PacBio HiFi reads were assembled using Hifiasm²⁶ (v0.25.0-r726) with default parameters, producing a 2.56 Gb draft genome consisting of 721 contigs, with a contig N50 of 27.47 Mb, indicating a high-quality assembly (Table 2). Assembly quality was assessed with BUSCO²⁷ (v5.4.7) using embryophyta_odb10 dataset. The contig-level assembly showed 95.50% complete BUSCOs, demonstrating a high genome completeness.

Table 2.

Statistics for the initial assembly of HiFi reads.

Type	Length (bp)	Number
N10	83,372,276	3
N20	59,646,780	7
N30	49,484,795	12
N40	38,395,531	17
N50	27,470,550	25
N60	22,661,324	35
N70	18,802,091	48
N80	12,386,643	65
N90	6,050,625	93
Max Length	105,405,287	—
Total Number	—	721
Total Length	2,562,488,398	—
GC Ratio	39%	—

Hi-C data evaluation and scaffolding

Hi-C paired-end reads were aligned to the assembled genome using Juicer²⁸ (v1.6). The quality of the Hi-C library was assessed based on the proportion of valid Hi-C contacts. The results showed that 56.47% of the sequenced read pairs were valid Hi-C contacts, accompanied by high-quality metrics including an 88.72% ligation motif rate, 3.87% PCR duplication rate, and 80.52% alignable reads (Table 3). The remaining read pairs consisted of 19.04% chimeric ambiguous and 0.44% unmapped reads, which are typical for plant Hi-C experiments. Before scaffolding, PCR duplicates were removed using Juicer with default parameters. Chromosome-scale scaffolding was performed using 3D-DNA²⁹ (parameters: -m haploid --early-exit) for the first round, avoiding the error-prone iterative refinement that often introduces mis-joins and excessive fragmentation. The assembly was then manually adjusted using Juicebox³⁰ (v2.20.00) to correct errors and fix mis-joins. After removing 157,993,896 bp of redundant sequences, a total of 2,404,494,502 bp remained in the assembled genome. Of these, 2,372,317,629 bp were successfully anchored to 24 chromosomes, yielding a chromosome-level assembly with a high Hi-C anchoring rate of 98.66% and a scaffold N50 of 98.06 Mb (Table 4).

Table 3.

Statistics of Hi-C data.

Type	Number of reads pair	Percentage (%)
Sequenced read pairs	978,883,141	—
Normal paired	376,797,425	38.49
Chimeric paired	411,442,276	42.03
Chimeric ambiguous	186,371,921	19.04
Unmapped	4,271,519	0.44
Ligation motif present	868,435,162	88.72
Alignable (normal + chimeric paired)	788,239,701	80.52
Unique reads	750,381,094	76.66
PCR duplicates	37,858,607	3.87
Optical duplicates	0	0.00
Hi-C contacts	552,792,944	56.47

Table 4.

Statistics for the chromosome-level assembly.

Type	Contig		Scaffold
	Size (bp)	Number	Size (bp)	Number
N10	77,617,613	3	135,709,658	2
N20	50,696,069	7	123,746,081	4
N30	46,320,845	12	111,658,371	6
N40	35,086,371	18	105,500,870	8
N50	27,470,550	25	98,055,964	11
N60	22,534,736	35	92,744,121	13
N70	16,879,318	47	89,574,049	16
N80	12,529,757	63	84,163,752	19
N90	6,998,514	88	78,414,824	22
Max Length	104,288,895	—	141,062,403	—
Total Number	—	347	—	30
Total Length	2,404,494,502	—	2,404,653,002	—
GC Ratio	39%	—	39%	—

Assembly quality was evaluated by dividing the genome into non-overlapping 600 kb bins and quantifying Hi-C interaction intensity as the number of read pairs between any two bins. These values were visualized in a heatmap, which showed strong interaction signals along the diagonal of the Hi-C contact matrix and weaker off-diagonal signals. The diagonal pattern revealed 24 distinct chromosomal scaffolds with frequent local interactions, consistent with Hi-C-based scaffolding and confirming the high accuracy and continuity of the chromosome-level assembly (Fig. 2c-d, Table 5).

Table 5.

Summary of 24 chromosomes of C. appendiculata.

ID	Length (bp)	N content (%)	GC content (%)	Gap number
chr1	141,062,403	0.00	38.86	9
chr2	135,709,658	0.00	38.49	12
chr3	130,665,096	0.00	39.16	9
chr4	123,746,081	0.00	38.53	2
chr5	117,369,186	0.00	37.74	7
chr6	111,658,371	0.01	38.64	12
chr7	111,328,060	0.00	38.91	3
chr8	105,500,870	0.00	38.42	4
chr9	98,813,456	0.00	38.17	6
chr10	98,258,163	0.01	38.19	12
chr11	98,055,964	0.00	39.27	9
chr12	95,890,182	0.00	38.44	3
chr13	92,744,121	0.00	38.50	5
chr14	92,647,880	0.00	38.76	4
chr15	89,737,884	0.00	38.78	5
chr16	89,574,049	0.01	39.04	9
chr17	88,668,927	0.00	38.06	4
chr18	87,964,890	0.01	38.99	9
chr19	84,163,752	0.01	38.20	11
chr20	83,684,254	0.00	39.41	4
chr21	83,362,299	0.00	38.83	8
chr22	78,414,824	0.00	38.35	4
chr23	76,936,972	0.00	38.70	6
chr24	56,360,287	0.00	40.27	1

Repeat annotation

Tandem repeats were identified with Tandem Repeats Finder³¹ (v4.09, parameters: 2 7 7 80 10 50 2000 -d -h), accounting for approximately 8.67% of the genome. Transposable elements (TEs) were annotated using RepeatProteinMask³² with default parameters and RepeatMasker³² (v4.1.5, parameters: -nolow -no_is -norna -engine ncbi) based on the RepBase library³³, accounting for approximately 23.02% and 19.47%, respectively. De novo prediction using LTRharvest³⁴ (v1.6.5, parameters: -mintsd 4 -maxtsd 6 -motif tgca -similar 80 -longoutput -motifmis 1 -minlenltr 150 -maxlenltr 15000) identified LTR retrotransposons covering 22.93% of the genome. Using the de novo TEs predicted by LTRharvest and RepeatModeler³⁵ (v2.0.5, parameters: -engine ncbi) as a custom library, the genome was re-annotated with RepeatMasker, revealing 78.24% de novo repetitive sequences. After removing overlaps, the final non-redundant repeat sequences (including simple repeats, satellites, TEs and unclassified repeats) accounted for 83.85% of the genome (Table 6). The classification results of the repetitive sequences were presented in Table 7.

Table 6.

Statistics results of repetitive sequence.

Type	Repeat length (bp)	% of genome
TRF	208,438,218	8.66812
RepeatProteinMask	553,605,129	23.02225
RepeatMasker	468,071,229	19.46523
Denovo	1,881,411,880	78.24047
LTRharvest	551,335,944	22.92788
Total	2,016,220,372	83.84662

Table 7.

The classification results of the repetitive sequences.

Type		Length (bp)	% of genome
Retro	LTR/Copia	700,212,684	29.11907
	LTR/Gypsy	973,512,552	40.48453
	LTR/Ngaro	255,907	0.01064
	LTR/TATE	0	0.00000
	LTR/Pao	207,179	0.00862
	LTR/Viper	1,302	0.00005
	LTR/ERV	1,164,795	0.04844
	LTR/ERVL	23,614	0.00098
	LTR/ERVK	403,942	0.01680
	LTR/Delta	0	0.00000
	LTR/Lenti	0	0.00000
	LTR/Foamy	0	0.00000
	LTR/Caulimoviru	66,431	0.00276
	LTR/Other	9,767,886	0.40621
	SINE	2,174,066	0.09041
	LINE	77,034,377	3.20355
	Other	0	0.00000
DNA	Academ	123,831	0.00515
	CACTA	22,149,693	0.92112
	Crypton	1,491,867	0.06204
	Dada	196,085	0.00815
	Ginger	1,164,529	0.04843
	hAT	19,209,855	0.79886
	Helitron	4,972,280	0.20678
	Kolobok	924,428	0.03844
	Mutator-like	9,057,334	0.37666
	P_Element	1,482,525	0.06165
	PIF-Harbinger	1,649,819	0.06861
	PiggyBac	354,730	0.01475
	Sola	406,816	0.01692
	TcMar	1,155,458	0.04805
	Other	2,976,975	0.12380
Satellite		151,994	0.00632
Simple_repeat		0	0.00000
Other		611,152	0.02542
Unknown		123,448,292	5.13373
Total		1,956,352,398	81.35695

Gene prediction and functional annotation

An integrated approach, which involved de novo prediction, homology search, and transcriptome-based prediction, was employed to annotate protein-coding genes in the genome. Homology-based prediction was performed utilizing annotated information from eight related species, Bletilla striata (Thunb.) Rchb. f.³⁶, C. appendiculata²², Cymbidium goeringii (Rchb. f.) Rchb. f.³⁷, Cymbidium mannii Rchb. f.³⁸, Dendrobium loddigesii Rolfe³⁹, D. nobile⁴⁰, G. elata⁴¹, and Papilionanthe Miss Joaquim ‘Agnes’⁴². Initial homology prediction was conducted using GeMoMa⁴³ (v1.9, parameters: AnnotationFinalizer.r = NO AnnotationFinalizer.u = YES p = false o = true). Structurally complete genes were selected from these predictions to train the de novo gene prediction tools Augustus⁴⁴ (v3.5.0) and SNAP⁴⁵ (v2006-07-28) with default parameters. The transcriptome-based prediction was performed by mapping transcriptomic reads to the assembled genome using HISAT2⁴⁶ (v2.2.1, parameters: --max-intronlen 50000 --sensitive --dta --dta-cufflinks --phred33 --no-discordant --no-mixed), transcripts were assembled with StringTie⁴⁷ (v2.2.1), and open reading frames (ORF) were predicted using TransDecoder (v5.7.1, https://github.com/TransDecoder/TransDecoder, parameters: --single_best_only). Gene models predicted by the above tools were integrated using EVidenceModeler⁴⁸ (EVM, v1.1.1). The EVM-integrated gene models were refined and analyzed for alternative splicing using PASA⁴⁹ with transcriptomic evidence. This integrated annotation pipeline generated 31,703 genes and 42,258 transcripts, with summary statistics presented in Table 8.

Table 8.

Basic structural statistics of gene prediction.

Type	Gene set	Number	Average gene length (bp)	Average CDS length (bp)	Average exon per gene	Average exon length (bp)	Average intron length (bp)
De novo	Augustus	343,798	4,018.55	1,549.48	5.94	260.88	499.87
	SNAP	344,620	7,041.08	1,014.83	4.84	209.66	1,569.20
Homolog	B. striata	11,576	10,747.09	1,118.60	4.45	251.61	2,794.24
	C. appendiculata	91,610	4,143.94	1,556.78	2.42	642.30	1,817.11
	C. goeringii	32,062	8,269.31	897.17	3.52	254.65	2,921.77
	C. mannii	29,111	11,217.15	1,035.89	4.33	239.04	3,054.15
	D. loddigesii	66,927	11,614.01	2,371.17	2.79	851.24	5,176.49
	D. nobile	51,276	12,279.67	1,310.13	3.29	397.95	4,785.63
	G. elata	28,262	9,115.09	1,234.68	3.54	348.39	3,097.76
	Papilionanthe	43,431	7,386.97	857.70	3.17	270.79	3,012.40
RNA-seq	Transcript	37,619	17,577.67	1,115.07	5.31	210.10	3,822.03
Final	—	31,703	10,806.29	1,199.21	4.60	260.83	2,670.30

Subsequently, the completeness of the gene set was assessed using BUSCO with embryophyta_odb10 database. The assessment revealed that the final gene set contained 1,520 conserved orthologs (94.2%), including 1,474 single-copy (91.3%), and 46 duplicated (2.9%) genes (Table 9). Among the predicted genes, a total of 29,094 (91.77%) were functionally annotated in at least one database, 13,795 genes were annotated across all five databases of Non-Redundant Protein Database (NR, https://ftp.ncbi.nlm.nih.gov/blast/db/FASTA/), InterPro⁵⁰, Kyoto Encyclopedia of Genes and Genomes (KEGG)⁵¹, Swiss-Prot⁵², and euKaryotic Orthologous Groups (KOG, https://ftp.ncbi.nih.gov/pub/COG/KOG/) (Fig. 3a). Specifically, 17,942 genes (56.59%) were annotated in the KOG database (Fig. 3b), 15,754 genes (49.69%) were annotated in the Gene Ontology (GO) database^53,54 (Fig. 3c), and 18,117 genes (57.14%) were mapped to KEGG pathways (Fig. 3d). The 17,908 genes (56.5%) lacking cross-database annotation were not enriched for repetitive sequences or species-specific loci, with incomplete annotation likely reflecting limited functional annotation coverage for non-model plants. Although they tend to have shorter CDS and simpler structures, 64.1% of these genes possessed identifiable protein domains, supporting their status as genuine protein-coding genes rather than annotation artifacts.

Table 9.

Evaluation results of gene set completeness.

Type	Number	Percentage (%)
Complete BUSCOs (C)	1,520	94.2
Complete and single-copy BUSCOs (S)	1,474	91.3
Complete and duplicated BUSCOs (D)	46	2.9
Fragmented BUSCOs (F)	28	1.7
Missing BUSCOs (M)	66	4.1
Total BUSCO groups searched	1,614	—

Fig. 3.

Gene functional annotation of C. appendiculata genome. (a) Venn diagram of gene annotations against five databases: NR, InterPro, KEGG, Swiss-Prot, and KOG. (b) KOG functional classification. The y-axis represents KOG categories, and the x-axis shows the number of genes. (c) GO classification at two levels. The bar chart shows the distribution of genes across various GO terms. (d) KEGG pathway annotation. The y-axis lists KEGG pathways, and the x-axis indicates the number of genes assigned to each.

Compared with the previous draft assembly, 7,435 newly identified genes were defined as those lacking any BLASTP match or showing both <30% identity and <50% coverage against the earlier gene set. Functional enrichment analysis of these newly recovered genes using GO, KEGG, KOG, and InterPro databases revealed prominent categories including ribosome biogenesis (GO: U3 snoRNA binding, K14774, KOG: nuclear structure), primary metabolism and stress signaling (K00860, K12261, K22312), and transposable element-associated domains (PF03004) (Fig. 4). Complementary keyword-based annotation further identified 28 genes associated with secondary metabolism, 48 with disease resistance, and 24 with nutrient transport potentially relevant to mycorrhizal symbiosis. These findings underscore the value of this high-quality genome for future functional and applied research on C. appendiculata.

Fig. 4.

Functional enrichment analysis of 7,435 newly identified genes compared with the previous draft assembly. Bubble charts showing enriched functional categories in the new gene set based on (a) GO, (b) KEGG, (c) KOG, and (d) InterPro annotations. The x-axis represents the gene ratio (proportion of new genes assigned to each term). Bubble size indicates the number of newly recovered genes in each category, and color intensity reflects -log10(p-value) from hypergeometric tests using the full gene set as background.

Non-coding RNA annotation

Non-coding RNAs (ncRNAs), including miRNAs, tRNAs, rRNAs, and snRNAs, perform important biological functions despite not being translated into proteins. Using covariance models from the Rfam database⁵⁵ and the Infernal software package (v1.1.5)⁵⁶, 115 miRNAs and 1,010 snRNAs were identified in the genome. The tRNAscan-SE 2.0 tool⁵⁷ was employed to identify tRNA sequences based on their structural characteristics, and 523 tRNAs were identified. For rRNAs, homologs from related species were used as references for BLASTN alignment, through which 18,791 rRNAs were identified. In total, ncRNA sequences accounted for approximately 0.20% of the genome (Table 10).

Table 10.

Statistics of ncRNAs annotation results.

Type		Copy	Average length (bp)	Total length (bp)	% of genome
miRNA		115	122.86087	14,129	0.000588
tRNA		523	77.0038241	40,273	0.001675
rRNA	18S	2,125	1,213.57882	2,578,855	0.107244
	28S	5,644	142.061481	801,795	0.033343
	5.8S	1,410	155.45461	219,191	0.009115
	5S	9,612	108.341136	1,041,375	0.043307
snRNA	CD-box	523	108.026769	56,498	0.00235
	HACA-box	47	127.93617	6,013	0.00025
	Splicing	440	107.731818	47,402	0.001971
Total		20,439	—	—	0.199843

Data Records

All sequencing data, including PacBio HiFi reads, Hi-C reads, and RNA-seq data from six tissues, have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under accession SRP691102⁵⁸, and the chromosomal-level genome assembly has been deposited in NCBI GenBank under accession GCA_058369495.1⁵⁹. The same data have also been deposited in the Genome Sequence Archive (GSA)⁶⁰ in the National Genomics Data Center (NGDC)⁶¹ of the China National Center for Bioinformation (CNCB) under BioProject accession PRJCA046810, with GSA accession number CRA031683⁶². The assembly and annotations are available in the NGDC Genome Warehouse (GWH)⁶³ under the accession number GWHGWEK00000000.1⁶⁴.

Technical Validation

BUSCO assessment

Genome assembly quality was evaluated using BUSCO with default parameters. A lineage-specific set of conserved single-copy orthologous genes from the embryophyta_odb10 database served as the reference. The assembled genome was assessed for alignment coverage, integrity, and accuracy against the reference. The assembly achieved 95.35% complete BUSCOs, indicating high genome completeness.

Merqury assessment

The reference-free quality assessment tool Merqury⁶⁵ (v1.3) was used to evaluate genome accuracy and completeness. Using PacBio HiFi reads, Merqury reported a phred-scaled quality value (QV) of 69.76, corresponding to an error rate of approximately 1.06 × 10⁻⁷, with a completeness of 86.58%. Using DNBSEQ short reads, the QV was 52.24 with an error rate of approximately 5.96 × 10⁻⁶ and a completeness of 87.00%. The HiFi-based QV was notably higher, reflecting the superior accuracy of the long-read data and providing a more reliable estimate of base-level consensus quality. Together, these results confirmed the exceptionally high accuracy and completeness of the assembly.

CRAQ assessment

The quality and consensus reliability of the genome were further assessed using CRAQ⁶⁶ (v1.10), which evaluates read-assembly congruence through multidimensional metrics. The genome showed excellent coverage, with 99.37% of bases covered by reads and only 0.91% identified as low-confidence regions. The clip-based regional heterozygosity (Avg.CRH = 7.71) and clip-based structural heterozygosity (Avg.CSH = 1.35) indicated that the genome exhibits a certain degree of natural heterozygosity, which is not unusual for orchids and does not compromise assembly accuracy. Most significantly, the clip-based regional errors (Avg.CRE = 0.21; R-AQI = 97.94) and clip-based structural errors (Avg.CSE = 0.03; S-AQI = 96.73) both substantially exceeded the threshold for high-quality assemblies, demonstrating that this genome of C. appendiculata achieved reference-grade quality.

QUAST assessment

QUAST⁶⁷ (v5.3.0) assessment was performed to evaluate the contiguity of the assembled genome. The assembly demonstrated exceptional contiguity, with N50 and N90 values of 98.06 Mb and 78.41 Mb, respectively, and corresponding L50 and L90 values of 11 and 22. These results showed that 11 scaffolds cover 50% of the whole genome and 22 scaffolds cover 90%, confirming that the assembly represents a high-quality, contiguous, and biologically representative reference genome for C. appendiculata.

Author contributions

J. C. and Y. L. designed the study. Y. T. and B. X. performed the experiments. Y. T. and R. Y. analyzed the data and drafted the manuscript. S. D., X. Q. and X. Z. helped in sample collection and revised the manuscript. All authors revised and approved the final version of the manuscript.

Funding

J. C. discloses support for the research of this work from the Special Funds for Biosafety and Genetic Resource Management of the National Forestry and Grassland Administration [KJZXSA202504]. Y. T. discloses support for the research of this work from the China Postdoctoral Science Foundation [2025M772745].

Data Availability

The PacBio HiFi, Hi-C, and RNA-seq raw reads generated in this study are available in the NCBI Sequence Read Archive under accession SRP691102 and in the Genome Sequence Archive of the NGDC under accession CRA031683. The chromosome-level genome assembly is available in NCBI GenBank under accession GCA_058369495.1 and in the Genome Warehouse of the NGDC under accession GWHGWEK00000000.1.

Code availability

All bioinformatics analyses in this research were performed using recognized community-standard software following protocols and guidelines provided by the respective software developers. The specific software versions and corresponding parameters have been exhaustively delineated in the Methods section. No custom code was developed or implemented for this research.

Competing interests

The authors declare no competing interests.