A chromosome-level genome assembly of Cistanche deserticola provides insights into its evolution and molecular mechanisms of parasitism

Abstract

Cistanche deserticola (C. deserticola) is a holoparasitic plant of the Orobanchaceae family that parasitizes the roots of Haloxylon ammodendron (H.ammodendron). The absence of a high-quality genome has impeded our understanding of its parasitic mechanisms. Here, we present a chromosome-level genome assembly of C. deserticola (6.26 Gb) based on PacBio high fidelity (HiFi) and high-throughput chromosome conformation capture (Hi-C) sequencing, with a contig N50 of 81.25 Mb, 92.2% Benchmarking Universal Single-Copy Ortholog (BUSCO) completeness, and 54 640 protein-coding genes. Evolutionary analysis shows that C. deserticola diverged from related Orobanchaceae species approximately 38.23 million years ago. Among its key parasitic adaptations is the extensive loss of photosynthetic genes, which is compensated by the retention of transporters and carbon metabolic pathways for the utilization of host-derived nutrition. Bidirectional genetic exchanges include 34 H. ammodendron–derived horizontally transferred genes and 98 mobile mRNAs, as well as 14 C. deserticola–derived horizontally transferred genes and 77 mobile mRNAs targeting host defenses. Spatial transcriptomic data reveal haustorium-specific gene expression related to nutrient extraction and chemical defense, particularly the biosynthesis of phenylethanoid glycosides via dispersed-duplication-driven gene expansion. This genomic resource illuminates the evolutionary trajectory of C. deserticola and provides a foundation for conservation strategies and the biotechnological development of C. deserticola.

This study reports a chromosome-level genome assembly of the holoparasitic plant Cistanche deserticola (6.26 Gb) and indicates that it diverged from other Orobanchaceae species approximately 38.23 million years ago. The work further reveals key parasitic adaptations, including extensive loss of photosynthetic genes, bidirectional genetic exchanges with its host Haloxylon ammodendron, and haustorium-specific gene expression related to nutrient uptake and the synthesis of defense compounds.

Introduction

Cistanche deserticola (C. deserticola), a perennial holoparasitic plant endemic to desert ecosystems, belongs to the family Orobanchaceae (Wang et al., 2012). It forms an obligate parasitic relationship with its particular host, Haloxylon ammodendron (H.ammodendron), and exhibits remarkable adaptations to extreme drought stress (Song et al., 2021). This unique biology positions C. deserticola as a critical model for deciphering the molecular mechanisms of plant parasitism and environmental resilience, with implications for drought-tolerance strategies and sustainable resource utilization (Sun et al., 2020).

Research on C. deserticola is important for multiple reasons. Molecular studies are essential for accurately assessing its genetic diversity, determining its endangered status, enabling science-based conservation, and guiding resource management (Taji et al., 2002; Seki et al., 2003). As a highly valued medicinal herb, C. deserticola produces phenylethanoid glycosides (PhGs) as its principal bioactive components (Jiang and Tu, 2009; Li et al., 2016), which confer antioxidant, anti-inflammatory, antitumor, and neuroprotective effects (Fu et al., 2018; Tian et al., 2021; Wu et al., 2023). Understanding the genomic basis of PhG biosynthesis is crucial for clarifying therapeutic mechanisms, establishing robust quality standards, and guiding novel drug development (Atanasov et al., 2015; Lv et al., 2024). Characterizing the molecular interactions that govern its parasitism and ecological adaptation is fundamental for optimizing artificial cultivation systems, breeding improved varieties, sustainably meeting market demand, and reducing pressure on wild populations.

Despite the importance of C. deserticola, critical knowledge gaps persist, primarily owing to the absence of a high-quality nuclear genome assembly. Although mitochondrial, chloroplast, and plastid genomes of Cistanche species have been characterized (Li et al., 2013b; Miao et al., 2022a, 2022b), revealing structural complexity and transcriptome discrepancies (Li et al., 2015), the nuclear genome—which is essential for understanding core parasitic mechanisms—remains unsequenced. This gap hinders progress in several key areas. First, the molecular basis of its parasitism remains ambiguous. Although haustoria facilitate the exchange of nutrients and genetic material (Hettenhausen et al., 2017), including potential horizontally transferred genes and mobile mRNAs (Fan et al., 2023), the absence of a reference genome prevents accurate assessment of the scale, functional significance, and molecular consequences of these bidirectional transfers. Consequently, the core mechanisms of host manipulation remain obscure. Second, the evolutionary trajectory of holoparasitism within Orobanchaceae is poorly resolved, as insufficient genomic data have impeded robust phylogenetic comparisons, obscuring the role of whole-genome duplications (WGDs) and hindering the definition of lineage-specific adaptations such as gene loss and retention, which are critical for parasitic evolution (Westwood et al., 2010; Xu et al., 2022). Third, the metabolic compensation strategies that enable it to survive without photosynthesis remain enigmatic. As a holoparasite that spends much of its life cycle below ground, C. deserticola has undergone significant degeneration of photosynthetic pathways (Li et al., 2019). However, the molecular mechanisms for efficient import and utilization of host-derived carbon and nutrients under low-light conditions have not been explored at the genomic level. Finally, the genomic basis for the observed accumulation of defensive PhGs, predominantly at the critical host–parasite interface, is unknown.

To address these fundamental gaps and clarify the molecular underpinnings of C. deserticola’s parasitic adaptation, we generated a chromosome-scale nuclear genome assembly of C. deserticola using PacBio HiFi and Hi-C sequencing. Together with transcriptomic and comparative analyses, this genomic resource enables examination of the evolutionary history, parasitic adaptations, and medicinal compound biosynthesis of C. deserticola at unprecedented resolution, providing a foundation for conservation, cultivation, and bioprospecting.

Results

Chromosome-scale genome assembly and comprehensive annotation of C. deserticola

To characterize the nuclear genome of C. deserticola, we generated short-read sequence data using Illumina sequencing technology. K-mer (K = 19) analysis estimated the genome size at approximately 5.78 Gb with a heterozygosity rate of 1.32% (Supplemental Figure 1A; Supplemental Tables 1 and 2). We then performed an initial assembly using 177.35 Gb PacBio HiFi reads (Supplemental Figure 1B; Supplemental Table 3), yielding a preliminary assembly of 7.17 Gb in 1171 contigs. This assembly had a contig N50 of 81.25 Mb and a GC content of 38.82% (Supplemental Table 4). Further scaffolding with 619.66 Gb of high-quality Hi-C data (Supplemental Figure 1C; Supplemental Tables 5 and 6) anchored 96.97% of the genome sequence onto 20 pseudochromosomes (2n = 20) (Supplemental Table 7). The final chromosome-scale assembly comprised 137 scaffolds, with a scaffold N50 of 320.45 Mb and a total length of 6.26 Gb (Table 1; Figure 1). This assembly was used for all subsequent genomic analyses. Analysis of Benchmarking Universal Single-Copy Orthologs (BUSCOs) estimated the genome completeness at 92.2% (Supplemental Table 8), and the long terminal repeat (LTR) assembly index was 23.48 (Supplemental Figure 1D). Collectively, these metrics indicate that the C. deserticola genome assembly is highly complete and of high quality (Supplemental Figure 1E; Supplemental Table 9).

Figure 2.

Genomic evolutionary history of C. deserticola.

(A) Phylogenetic tree of C. deserticola and 12 other plant species. Hypothetical divergence times are displayed at each node. Expanded and contracted gene families on each branch are shown in green and red, respectively. The reported whole-genome duplication (WGD) event is marked by a yellow star.

(B) Synonymous substitution rate (Ks) curves of C. deserticola, O. cumana, and Coffea canephora. Peak 1 (x = 0.73) represents the WGD event identified in C. deserticola that occurred around 70.4 mya. Peak 2 (x = 0.80) represents the WGD event identified in O. cumana that occurred around 76.1 mya. Peak 3 (x = 0.34) represents the speciation of C. deserticola and O. cumana that occurred around 16.4 mya.

(C) Ka/Ks ratios for duplicated genes in P. aegyptiaca (Paeg), C. deserticola (Cdes), and O. cumana (Ocum). The Ka/Ks ratio represents the ratio of non-synonymous to synonymous substitutions. Boxplot elements: center line, median; box limits, first and third quartiles (25th and 75th percentiles); whiskers, minimum and maximum values within 1.5× interquartile range (IQR) of the respective quartile; points, outliers. Red stars indicate statistically significant differences (P < 0.05) in pairwise comparisons between species for the individual duplication modes.

(D) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of duplicated genes. KEGG pathways with significant enrichment (P < 0.05) are shown. Circle color represents the statistical significance of the enriched KEGG pathway, and circle size represents the number of genes in the pathway.

Figure 6.

Biosynthesis of phenylethanol glycosides (PhGs) in C. deserticola.

(A) Tissue-specific expression of genes related to phenylethanol glycoside (PhG) biosynthesis. The schematic depicts PhG biosynthetic pathway intermediates (black) and catalyzing enzymes (orange). Heatmaps show relative gene expression levels across C. deserticola tissues.

(B) Tissue-specific expression of key PhG biosynthesis gene families. Heatmap shows expression profiles of the AADH, CSE, C3H, and HCT gene families in haustorium (Hau) and stem tissues.

(C) RT–qPCR validation of the expression of key PhG genes. Independent validation of AADH1, CSE2, and CSE3 transcript levels in host roots (Ham), the haustorium (Hau), and the C. deserticola stem is shown. Data are presented as mean ± SD (n = 6).

(D) Co-expression network of PhG genes and transcription factors (TFs). Red/gray nodes: TFs; yellow/orange/blue/dark blue/purple: PhG biosynthetic genes. Edges connect nodes with an absolute correlation coefficient > 0.99. Node size reflects correlation strength. Enzyme abbreviations: PAL, phenylalanine ammonia-lyase; C4H, cinnamic acid 4-hydroxylase/trans-cinnamate 4-monooxygenase; CYP98A2, 5-O-(4-coumaroyl)-D-quinate 2′-monooxygenase; C3H, coumarate 3-hydroxylase; CSE, caffeoyl shikimate esterase; HCT, shikimate O-hydroxycinnamoyltransferase; 4CL, 4-coumarate-CoA ligase; AADC, aromatic-L-amino-acid/L-histidine decarboxylase; TYRDC, tyrosine/DOPA decarboxylase; CuAO, copper amine oxidase; PPO, polyphenol oxidase/catechol oxidase; AADH, aryl-alcohol dehydrogenase/alcohol dehydrogenase; UGT, UDP-glycosyltransferase/UDP-glucosyl transferase. Red stars indicate DSD-derived genes. Tissue abbreviations: Ham, roots of H. ammodendron (host); Hau, C. deserticola haustorium (host–parasite interface); Bottom, C. deserticola stem segment (1 cm from haustorium); Middle, middle segment of C. deserticola stem; Top, apical segment of C. deserticola stem, including inflorescence.

Table 1.

Assembly statistics for the C. deserticola genome.

Item	Value
Total length (bp)	6 263 570 418
Scaffold number	137
N50 (bp)	320 445 392
N90 (bp)	233 036 329
Average (bp)	45 719 492.10
Median (bp)	1 475 953.00
Min (bp)	10 000
Max (bp)	434 435 716
GC content (%)	38.63

Figure 1.

Genomic features of C. deserticola.

Chromosome-level landscape of the C. deserticola genome. Tracks, from outside to inside, depict the 20 assembled pseudochromosomes, gene densities, repeat regions, and GC contents at a resolution of 50 kb. Each linking line in the center of the circle connects a pair of homologous genes. Horizontally transferred genes are marked with red circles.

Repetitive sequences (93.34%, 5.85 Gb) dominate the C. deserticola genome, with LTRs (Gypsy, 42.82%; Copia, 29.69%) constituting 81.71% of the repeats (Supplemental Figure 1F; Supplemental Table 10). Gene prediction, integrating homology-based annotation, de novo annotation, and transcriptome evidence, identified 54 640 protein-coding genes in the C. deserticola genome. The average gene length was 12 521.23 bp, with an average coding sequence (CDS) length of 1030.18 bp, an average exon length of 683.15 bp, and an average intron length of 2568.80 bp (Supplemental Figure 1G; Supplemental Tables 11–13). Comparison with databases of known non-coding RNAs enabled the annotation of 8140 rRNAs, 2184 tRNAs, 8067 small nucleolar RNAs, and 156 microRNAs within the C. deserticola genome (Supplemental Table 14).

Absence of recent species-specific whole-genome duplication (WGD) events in C. deserticola

WGD events play a significant role in plant evolution, particularly for angiosperms. Characterizing the occurrence of WGD events in individual species is crucial for understanding their evolutionary trajectories. To investigate potential independent WGD events in C. deserticola, we performed comparative genomic analyses. We clarified its phylogenetic position by analyzing it with six other Orobanchaceae species, including Lindenbergia luchunensis (L. luchunensis), Orobanche cumana (O. cumana), Phelipanche aegyptiaca (P. aegyptiaca), Striga hermonthica (S. hermonthica), Striga asiatica (S. asiatica), and Phelipanche japonicum (P. japonicum), and six outgroup species, including H. ammodendron, Angelica sinensis (A. sinensis), Panax notoginseng (P. notoginseng), Populus euphratica (P. euphratica), Cleistogenes songorica (C. songorica), and Welwitschia mirabilis (W. mirabilis).

Phylogenetic reconstruction based on 143 single-copy genes resolved C. deserticola as a sister group to O. cumana and P. aegyptiaca (Supplemental Figure 2A; Supplemental Table 15). Molecular dating, calibrated using the W. mirabilis fossil record, estimated the divergence time between C. deserticola and O. cumana at approximately 38.23 million years ago (mya) (95% highest posterior density [HPD], 29.4–48.8 mya) (Figure 2A). Analysis of synonymous substitution rates (Ks) for collinear gene pairs revealed a prominent peak at 0.73 (peak 1) in C. deserticola paralogs, which closely matched a corresponding peak in O. cumana (Ks = 0.80, peak 2) (Figure 2B; Supplemental Table 16). Combined with a 2:2 syntenic depth ratio between the two species (Supplemental Figure 2B and 2C), these findings indicate that both share an ancient WGD event dated to approximately 70.3 ± 2 mya.

A second Ks peak at 0.34 (peak 3) in both C. deserticola and O. cumana corresponds to a divergence time of 33.3 ± 9.5 mya (Figure 2B). This Ks-derived estimate is concordant with the molecular dating result (38.23 mya), confirming that lineage divergence occurred after the shared ancient WGD event. Importantly, no 1:1 syntenic blocks were detected within the C. deserticola pseudochromosomes (Supplemental Figure 2D), providing strong evidence against the occurrence of a recent, lineage-specific WGD event in C. deserticola.

Differential selection on ancient WGD genes has driven lineage-specific expansion in C. deserticola

Gene duplication is the primary driver of gene family expansion, and understanding the modes of gene duplication and their functional contributions is crucial for deciphering evolutionary processes and adaptive diversification.

We identified five distinct gene duplication modes in C. deserticola: WGD (3369 genes, 5.39%), tandem duplication (1642 genes, 2.63%), proximal duplication (2572 genes, 4.12%), transposed duplication (11 739 genes, 18.78%), and dispersed duplication (DSD; 43 180 genes, 69.08%) (Supplemental Table 17). Comparative genomic analysis across Orobanchaceae revealed conserved proportions of these duplication modes, with the proportion of WGD-derived genes showing remarkable stability (C. deserticola: 5.39%, O. cumana: 4.94%, and P. aegyptiaca: 5.17%; ANOVA, P = 0.9971) (Supplemental Table 17). These findings suggest that despite divergent parasitic traits that evolved independently over long-term evolution, genes derived from a shared ancient WGD event are highly conserved across the three species.

Gene duplication mechanisms play a fundamental role in gene family expansion during evolution. In C. deserticola, 1875 gene families have undergone expansion (7300 genes), and 89.37% (6524) of these genes originated from gene duplication events. Comparative analysis showed no significant difference in the proportional contributions of the five duplication modes to expanded genes across the three species (ANOVA, P = 0.8232), and all five duplicated gene types exhibited Ka/Ks ratios < 1 (Figure 2C), indicating purifying selection. Notably, C. deserticola displayed a significantly higher contribution of WGD to expanded gene families (32.05%, 2091/6524) compared with O. cumana (23.64%, 532/2250) and P. aegyptiaca (20.76%, 1126/5423) (Supplemental Table 17), alongside elevated Ka/Ks ratios for WGD-derived genes (P < 0.0001) (Figure 2C). This indicates intensified positive selection on ancestral WGD genes, driving lineage-specific expansion. Functional enrichment analysis confirmed that these expanded WGD genes were predominantly involved in defense and specialized metabolism (e.g., terpenoid/polyketide metabolism, phenylalanine/tyrosine/tryptophan biosynthesis, and terpenoid backbone biosynthesis) (Figure 2D; Supplemental Dataset 1), supporting their role in adaptive evolution rather than recent polyploidization events.

Bidirectional transfer of genetic material between H. ammodendron and C.deserticola facilitates parasitism and suppresses host defense

Building on established evidence of molecular exchange between parasitic plants and their hosts (Kim et al., 2014), we investigated the bidirectional transfer of genetic material in the H. ammodendron–C. deserticola system.

At the DNA level, 34 horizontally transferred genes from H. ammodendron to C. deserticola were identified (Supplemental Table 18; Supplemental Figure 3A and 3B; Supplemental Dataset 2), dispersed across the 20 pseudochromosomes of C. deserticola without clustering (Supplemental Figure 3C). These horizontally transferred genes exhibited significantly longer gene and intron lengths than native genes in both species (P < 0.05) (Supplemental Figure 3D), with intron length increasing with increasing divergence time (Figure 3A). Horizontally transferred genes with an intron length (2–3 kb) comparable to that of native C. deserticola genes showed elevated expression levels (Supplemental Figure 3E), indicating intron-mediated genomic adaptation. Notably, 18 host-derived horizontally transferred genes contained jasmonic acid (JA)-responsive cis-elements (G-box and GCC-box) and displayed haustorium-specific high expression (frgments per kilobase per million mapped reads [FPKM] = 11.09 ± 7.2; stem: 7.11 ± 4.8, P < 0.05), suggesting that C. deserticola may use horizontally acquired host genes to subvert JA-mediated defense (Figure 3B). Concurrently, 14 genes horizontally transferred from C. deserticola to H. ammodendron were detected, confirming bidirectional DNA transfer. Functional annotation revealed that these parasite-derived genes encoded proteins involved in host metabolic manipulation, predominantly transporters and kinases (Supplemental Dataset 3).

Figure 3.

mRNA and DNA exchanges between C. deserticola and its host.

(A) Temporal patterns of structural changes in horizontally transferred genes relative to inferred divergence time. Changes in gene length, CDS length, and intron length are plotted against relative divergence times estimated using the RelTime (relative time framework) method in MEGA7 on the basis of the maximum likelihood (ML) tree.

(B) Left: frequency of hormone-responsive cis-elements in promoters of horizontally transferred genes. GA, gibberellic acid; SA, salicylic acid; ABA, abscisic acid; JA, jasmonic acid. Right: tissue-specific expression heatmap of horizontally transferred genes in C. deserticola.

(C) Co-expression modules of host-derived mobile mRNAs. Left: co-expression patterns of mobile mRNAs clustered into four modules. Red and blue boxes indicate mobile mRNAs with increased and decreased abundance, respectively. Right: KEGG functional enrichment of mobile mRNAs within each module. Tissue abbreviations: Ham, roots of H. ammodendron (host); Hau, C. deserticola haustorium (host–parasite interface); Bottom, C. deserticola stem segment (1 cm from haustorium); Middle, middle segment of C. deserticola stem; Top, apical segment of C. deserticola stem, including inflorescence.

At the RNA level, 98 mobile mRNAs that were transferred from H. ammodendron to C. deserticola (Supplemental Table 19; Supplemental Figure 4A) exhibited distinct spatial expression patterns: cluster 1 (16 mobile mRNAs) was highly expressed in H. ammodendron roots and functionally associated with GTP-binding protein activity; cluster 2 (49 mobile mRNAs) was enriched in C. deserticola haustoria and linked to energy metabolism processes; cluster 3 (27 mobile mRNAs) accumulated in the basal stem of C. deserticola and was related to signaling proteins and glycan metabolism; and cluster 4 (6 mobile mRNAs) was concentrated in the mid-stem of C. deserticola and involved in cofactor metabolism (Figure 3C; Supplemental Dataset 4). This spatial–functional partitioning indicates that host-derived mRNAs are selectively used in specific parasite tissues, implying that they may have specialized roles in establishing and maintaining the parasitic interface. Significantly, 77 mobile mRNAs that were transferred from C. deserticola to H. ammodendron (Supplemental Table 19) were enriched in core host metabolic pathways (terpenoid biosynthesis, the tricarboxylic acid cycle, oxidative phosphorylation, and carbon fixation), with 53.24% (41/77) highly expressed in haustoria, suggesting that these mobile RNAs reprogram host metabolism to enhance nutrient flux (Supplemental Dataset 5).

This bidirectional transfer, particularly of parasite-derived horizontally transferred genes that encode regulatory/metabolic proteins and mobile mRNAs that target core host physiology, suggests sophisticated mechanisms for resource appropriation and host manipulation by C. deserticola.

Host-dependent transcriptional plasticity optimizes C. deserticola parasitism

To clarify the molecular underpinnings of parasitism, we investigated spatial gene expression dynamics in C. deserticola. Principal-component analysis of 86 169 non-redundant genes revealed striking transcriptional heterogeneity driven by host genetic divergence (Figure 4A and 4B): parasites on the same host root exhibited relatively low differential gene expression (1697 differentially expressed genes [DEGs]) and near-identical transcriptomes (r = 0.94 ± 0.02, P < 0.05; Figure 4C), whereas those on different roots of the same host showed much greater differences (12 130 DEGs).

Figure 4.

Transcriptional heterogeneity among individual C. deserticola plants.

(A) Spatial sampling strategy. Diagram illustrating C. deserticola tissue collection for comparative transcriptomics. Haustorium (Hau), stem segments (Bottom: 1 cm from haustorium; Middle: mid-stem; Top: apical stem), and host H. ammodendron roots (Ham). Comparisons include C2 vs. C1 (parasites on the same host root), C3 vs. C1 (parasites on different roots of the same host), and C4 vs. C1 (parasites on different hosts).

(B) Principal-component analysis (PCA) of transcriptomes. Three-dimensional PCA plot (axes: PC1, PC2, and PC3) showing global gene expression variation, with each point representing a biological sample.

(C) Hierarchically clustered heatmaps of differentially expressed genes (DEGs) across individual C. deserticola plants.

(D) Venn diagram showing shared and unique DEGs from two C. deserticola plant comparisons.

(E) Venn network diagram showing enriched KEGG pathways in DEGs from two C. deserticola plant comparisons.

Remarkably, parasites exploiting genetically distinct hosts exhibited dramatic transcriptomic divergence, with 15 198 DEGs (17.64% of the transcriptome) (Supplemental Figure 5A and 5B) and significantly reduced expression similarity (r = 0.49 ± 0.02, P < 0.05). Functional convergence analysis of 7536 common DEGs (Figure 4D)—shared between parasites on different roots of the same host and parasites on different hosts—identified core adaptive pathways enriched in nutrient processing (carbohydrate, lipid, and amino acid metabolism), energy infrastructure (mitochondrial biogenesis), and host interface mechanisms (transmembrane transporters and signal transduction) (Figure 4E; Supplemental Dataset 6). This host-dependent transcriptional plasticity reveals a co-adaptive relationship in which parasite gene expression dynamically adjusts to individual host genotypes.

Spatial profiling identified 206 tissue-specific genes (TSGs) (Supplemental Dataset 7), 64.73% of which (134 genes) were expressed exclusively in haustoria (Supplemental Figure 6A and 6B). These haustorium-specific TSGs were functionally enriched in host resource extraction (transfer RNA biogenesis, messenger RNA biogenesis, and nucleocytoplasmic transport) and chemical defense (including the biosynthesis of phenylpropanoids, precursors of PhGs) (Supplemental Dataset 8). By contrast, TSGs specific to the top of the stem (n = 60) were associated with nutrient utilization pathways (carbohydrate metabolism, lipid metabolism, amino acid metabolism, fatty acid degradation, and lysine, arginine, and proline metabolism) (Supplemental Dataset 8), reflecting a division of labor between the parasitic interface (haustoria) and storage tissues (stem). Importantly, 478 constitutively expressed genes (housekeeping genes) maintain aspects of essential parasitic infrastructure across all tissues (Supplemental Dataset 9), including cellular maintenance (transfer RNA biosynthesis and translation factors) and energy metabolism (mitochondrial biogenesis and signal transduction) (Supplemental Figure 6C; Supplemental Dataset 10).

Metabolic compensation for photosynthetic loss in C. deserticola

Gene loss is a defining feature of parasitic plants (Frailey et al., 2018), profoundly altering their physiology and metabolism. BUSCO analysis revealed extensive gene loss in holoparasites (Figure 5A): compared with autotrophic plants, which contained 6626 conserved orthologous groups (OGs), C. deserticola exhibited a loss of 10.62% (704/6626), higher than that of P. aegyptiaca (8.96%, 594/6626) and O. cumana (6.91%, 458/6626) and much higher than that of hemiparasites such as S. asiatica (3.4%, 25/6626) and P. japonicum (2.26%, 150/6626). Comparative analysis further highlighted niche-driven convergent evolution: 111 conserved OGs were lost in both root-parasitic C. deserticola and stem-parasitic Cuscuta australis (C. australis) and were predominantly associated with dispensable plastid functions (Supplemental Figure 7A and 7B; Supplemental Dataset 11). Whereas C. australis retains the photoreception module adaptive to its aerial lifestyle (Supplemental Figure 7C), C. deserticola has completely lost light-nutrition-related genes, underscoring adaptive differentiation shaped by ecological niches. Notably, 223 conserved OGs have been lost across all Orobanchaceae holoparasites (Figure 5B). Gene Ontology analysis showed that 60.68% (71/117) of these losses affect chloroplast/plastid functions, and 22.43% (24/107) disrupt photosynthesis or chlorophyll biosynthesis (Figure 5C; Supplemental Dataset 12), defining a lineage-specific pattern of photosynthetic degeneration. In C. deserticola, this decay is exacerbated by lineage-specific losses: 50% (38/76) of genes involved in light-dependent reactions are absent, with the remaining genes transcriptionally silenced (Supplemental Figure 8A; Supplemental Dataset 13); chlorophyll and carotenoid biosynthesis pathways are nearly eliminated (Figure 5D; Supplemental Datasets 14 and 15); and 39.66% (69/174) of Calvin cycle genes—including ribulose-1,5-bisphosphate carboxylase—are lost (Supplemental Dataset 16), disrupting 3-phosphoglycerate synthesis essential for sucrose and starch production.

Figure 5.

Patterns of gene loss in the parasitic plants C. deserticola, P. aegyptiaca, O. cumana, and C. australis.

(A) Simplified phylogeny of the analyzed parasitic species. Benchmarking Universal Single-Copy Ortholog (BUSCO) assessments revealed substantially more gene loss in C. deserticola relative to O. cumana, P. aegyptiaca, C. australis, and non-parasitic L. luchunensis, as indicated by the larger fraction of missing BUSCOs (gray segments).

(B) Shared orthogroup loss among Orobanchaceae holoparasites. The Venn diagram shows shared and unique orthogroup losses among C. deserticola, O. cumana, and P. aegyptiaca.

(C) Gene Ontology (GO) enrichment analysis of orthogroups lost in all three Orobanchaceae holoparasites (C. deserticola, O. cumana, and P. aegyptiaca). Significantly enriched GO terms (P < 0.05) are shown. Circle color indicates enrichment significance, and circle size represents the number of genes per term.

(D) Photosynthesis-related gene loss across autotrophic and parasitic plants. Percentages of intact genes (relative to A. thaliana) in key photosynthetic pathways are shown for four species: non-parasitic L. luchunensis and holoparasites C. deserticola, O. cumana, and P. aegyptiaca. Pathways include PS II (photosystem II complex), PS I (photosystem Ⅰ complex), F-type ATPase, cytochrome b6f complex (Cytb6f), chlorophyll biosynthesis, carotenoid biosynthesis, and carbon fixation.

Interestingly, C. deserticola has retained genes encoding certain transport proteins, such as the phosphate translocator (GPT), the pyruvate cotransporter (BASS), ATP/ADP transporters (NTT), the NAD transporter (NDT), and the plastid glucose transporter (pGlc) (Supplemental Dataset 17). These proteins play a vital role in facilitating the transport of photosynthetic products into the cytoplasm for carbohydrate synthesis. This implies that C. deserticola might exploit the photosynthetic products derived from H. ammodendron for glucose and ATP synthesis. Compared with Arabidopsis thaliana (A. thaliana), C. deserticola has retained 50% (34/68) of glycolysis/gluconeogenesis genes, 37.46% (124/331) of sucrose/starch metabolism genes, and 85.05% (74/87) of fatty acid synthesis genes (Supplemental Datasets 18–20). Moreover, among the mobile mRNAs transferred from H. ammodendron to C. deserticola, the two genes CdesChr3G00084160 and CdesChr3G00075650 encode beta-glucan phosphorylase 2 (PHS2) and hexokinase 3 (HXK), respectively (Supplemental Figures 4B and 4C). These proteins are of utmost importance for carbohydrate metabolism, suggesting that C. deserticola may directly use mobile mRNAs from H. ammodendron for carbohydrate biosynthesis.

Dispersed duplications (DSDs) drive haustorial specialization in C. deserticola’s chemical defense

As specialized defense compounds, PhGs accumulate predominantly in the C. deserticola haustorium, which is the critical host–parasite interface (Feng et al., 2023). Genomic analysis identified all 13 enzyme families (62 genes) involved in PhG biosynthesis (Figure 6A). Comparative genomics analysis revealed C. deserticola–specific expansions (two- to three-fold) in key families (AADH, CSE, C3H, and HCT) compared with its parasitic relatives P. aegyptiaca and O. cumana. Notably, gene duplication analysis demonstrated that DSDs have driven this expansion, contributing 51.6% (32/62) of the genes in the PhG pathway.

Spatial transcriptomics demonstrated that DSD-derived genes dominate haustorium-specific expression: two CSE genes (CdesChr18G00471050 and CdesChr1G00010780) and one AADH gene (CdesChr2G00035130) exhibited 2.5-fold higher FPKM in haustorium vs. stem (P < 0.05; Figure 6B), and these expression patterns were confirmed by RT–qPCR across tissues (Figure 6C). Targeted metabolomics confirmed that PhGs were undetectable in H. ammodendron roots, whereas haustorial accumulation was 1.13 ± 0.16-fold higher than that in stems (P > 0.05), with distinct monomer profiles (Supplemental Figure 9). Integrated analysis established that DSD-mediated expansion of biosynthetic genes directly enables haustorium-specific expression, further regulated by DSD-enriched transcription factors (C3H, bHLH, and C2H2) that form core regulatory networks (Figure 6D). In summary, this cascade of gene expansion, tissue-specific expression, and transcriptional regulation culminates in localized PhG production, representing C. deserticola’s evolutionary innovation in chemical defense—achieved through DSD-driven neofunctionalization to fortify the parasitic interface.

Discussion

The obligate holoparasite C. deserticola thrives in arid deserts by completely exploiting H. ammodendron, representing an extreme paradigm of adaptive evolution (He et al., 2021). Our chromosome-scale genome assembly (6.26 Gb, encoding 54 640 protein-coding genes), generated using high-coverage PacBio and Hi-C sequencing technologies, provides the first comprehensive nuclear genomic insights into its evolutionary trajectory (Table 1; Figure 1). Molecular dating analyses resolved C. deserticola as a sister to O. cumana, with divergence occurring 38.23 Mya. This places C. deserticola as potentially the earliest diverging holoparasitic lineage within the Orobanchaceae family (Xu et al., 2022). Its exceptional genome size, currently the largest in Orobanchaceae (Xu et al., 2022), is driven primarily by massive retrotransposon proliferation (LTRs constitute 81.71% of its 93.34% repetitive content), rather than by a recent lineage-specific WGD event.

Gene duplication is an important source of novel genes and drives the evolution of genetic diversity (Panchy et al., 2016). It fuels adaptive innovation through two synergistic mechanisms: first, the legacy of ancient WGD events contributes via lineage-specific gene retention. C. deserticola has retained a higher proportion of WGD-derived genes in expanded families (32.05%) compared with its relatives (O. cumana: 23.64%; P. aegyptiaca: 20.76%; P < 0.05), and this enrichment is particularly pronounced in genes related to defense and specialized metabolism. Notably, the proportion of WGD-derived genes is conserved across Orobanchaceae (C. deserticola: 5.39%; O. cumana: 4.94%; P. aegyptiaca: 5.17%; ANOVA, P = 0.9971). All duplication types are subject to purifying selection (Ka/Ks < 1), but C. deserticola exhibits significantly elevated Ka/Ks ratios for WGD-derived genes (P < 0.0001), indicating intensified positive selection acting on these ancient duplicated genes (Zhang et al., 2013; Xu et al., 2022).

Holoparasitism has shaped massive gene loss in C. deserticola, as evidenced by significant deficiencies in photosynthesis and chlorophyll biosynthesis pathways—a phenomenon that reflects coordinated nuclear–plastid streamlining (Li et al., 2013b; Guo et al., 2023). To achieve metabolic compensation, functional adaptations have occurred within key pathways: retained plastid transporters (GPT and BASS) import host photosynthates (Furumoto et al., 2011; Bockwoldt et al., 2019) (Supplemental Dataset 17), and central metabolic pathways use these acquired photosynthates directly in glycolysis/gluconeogenesis (50% retained), sucrose/starch metabolism (37.46% retained), and fatty acid synthesis (85.05% retained). This metabolic adaptation parallels patterns observed in other parasitic lineages (Xu et al., 2021). Interestingly, host mobile mRNAs (e.g., PHS2 and HXK) further facilitate the integration of hexoses into central metabolism. Convergent loss of dispensable plastid genes in C. australis highlights niche-driven evolution (Sun et al., 2018), whereas the unique erosion of light-sensing modules in C. deserticola aligns with its obligate subterranean lifestyle (Conde et al., 2011). This integrated strategy—combining transporter networks, metabolic retention, and molecular piracy—epitomizes root-parasitic adaptation.

Bidirectional genetic transfer underpins host manipulation through stable horizontally transferred genes and transient mobile mRNA exchange (Yang et al., 2019). At the DNA level, we identified 14 genes horizontally transferred from C. deserticola to H. ammodendron (Supplemental Dataset 3). Functional annotation revealed that their encoded proteins are involved in key functions such as kinase regulation, signal transduction, and ubiquitination modification, suggesting that they may interfere with host cell-signaling pathways or regulate host metabolism, creating favorable conditions for C. deserticola parasitism. At the RNA level, we detected 77 mobile mRNAs transferred from C. deserticola to H. ammodendron (Supplemental Table 19). These mobile mRNAs were significantly enriched in the host’s core metabolic pathways, including terpene biosynthesis, the tricarboxylic acid cycle, oxidative phosphorylation, and carbon fixation. A large fraction of these mobile mRNAs (53.24%, 41/77) were highly expressed in haustoria, suggesting that they may regulate host biosynthesis and energy metabolism, promoting the transport of nutrients from H. ammodendron to C. deserticola to compensate for the defect in energy acquisition caused by loss of photosynthesis. Together, these transfers form a synergistic system: DNA-transferred regulators may suppress host defenses and alter signaling, while the mRNA-transferred metabolic enzymes directly exploit host resources, enabling C. deserticola to systematically acquire nutrients from its host.

We identified 120 core immune regulatory genes (Supplemental Dataset 22) in the genome of H. ammodendron on the basis of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway annotations, 28 of which were involved in JA synthesis and signaling pathways in H. ammodendron (Supplemental Dataset 22). Transcriptome analysis revealed that 85.7% (24/28) of these genes showed low to moderate expression in the H. ammodendron root and were not significantly upregulated during parasitism (Supplemental Figure 8B). This indicates that although H. ammodendron retains a functional JA pathway, it does not strongly activate this pathway during C. deserticola parasitism—likely linked to active defense-interference strategies of the parasite (Bao et al., 2014; Hua, 2021; Qi et al., 2022). Interestingly, we found that C. deserticola expresses 23 JA pathway–related genes, including the JA signaling inhibitor JAZ1. This gene is highly expressed in C. deserticola haustoria (FPKM > 1000) (Johnson et al., 2023), suggesting that it may play a key role in weakening the host defense response by interfering with host JA signal transduction (Sanchez-Puerta et al., 2017; Ma et al., 2022).

Host genotype–dependent transcriptional plasticity further optimizes host exploitation, with parasites infecting genetically distinct hosts exhibiting dramatic transcriptomic divergence (15 198 DEGs; 17.64% of the transcriptome). This divergence centers on 7536 core DEGs enriched in parasitism-related functions, refined through spatial specialization: 134 haustorium-specific genes are focused on resource extraction (tRNA/mRNA biogenesis and transport) and chemical defense (phenylpropanoid/PhG biosynthesis), whereas stem-top genes regulate internal nutrient utilization. This dynamically tuned division of labor underscores exquisite co-adaptation to the host environment (Sasaki and Nagano, 2004; Granot et al., 2014).

Defense at the haustorial interface is achieved through DSD-driven innovation (Zhang et al., 2022; Cao et al., 2024): 51.6% of the genes in the PhG pathway arose through DSD (Supplemental Dataset 21), enabling 2–3-fold higher expression of key enzyme family genes (e.g., AADH, CSE, C3H, and HCT) in C. deserticola compared with related parasites (Supplemental Dataset 21). DSD-derived genes (e.g., CSE and AADH) show 2.5-fold higher expression in haustoria (P < 0.05), which is correlated with greater PhG accumulation. This expression is regulated by DSD-enriched transcription factors (e.g., C3H, bHLH, and C2H2) (Figure 6D) (Chen and Murata, 2002; Li et al., 2013a; Kim and Hwang, 2014). Targeted metabolomics analysis did not detect PhGs in the roots of H. ammodendron (Supplemental Figure 9), a finding consistent with the genomic data: the H. ammodendron genome lacks genes encoding key enzymes of PhG biosynthesis (e.g., CYP98A3, HCT, and 4CL; Supplemental Dataset 21), indicating that it cannot autonomously synthesize these compounds. Thus, PhGs produced by C. deserticola are unique to the parasite and absent from host tissues. Given the known antioxidant and anti-inflammatory properties of PhGs, their enrichment at the host–parasite interface likely shields haustorial cells from host-derived oxidative stress, thereby sustaining nutrient translocation channels (Georgieva et al., 2017).

The present research has several limitations. First, the genome assembly is not yet fully gapless: with the exception of chromosome 20, all the chromosomes still contain artificially inserted small gaps, which should be resolved in future work using telomere-to-telomere sequencing. Second, although expression differences were observed in JA pathway genes and JAZ1 was highlighted, the functional relevance of these differences remains to be determined experimentally. Third, owing to sampling constraints, the dynamics of mobile gene accumulation over the course of parasitism remain unclear. Finally, the mechanisms underlying bidirectional gene transfer still lack direct evidence, and a systematic interaction network has yet to be established. To address these questions, future work will focus on (1) improving genome continuity through telomere-to-telomere sequencing; (2) functional validation of candidate genes through comparative experiments, gene silencing, and knockout approaches; (3) construction of protein–protein interaction networks and the combination of fluorescence in situ hybridization with gene editing to dissect mechanisms of bidirectional gene transfer; and (4) metabolite intervention assays, dynamic sampling, and field studies to clarify the ecological and molecular dynamics of host–parasite interactions.

In summary, the genome of the desert parasite C. deserticola reveals its comprehensive strategy for adaptation to desert environments: the selective retention of genes derived from ancient WGD provides an evolutionary basis, and the massive loss of photosynthetic genes forces C. deserticola to compensate through metabolic reconfiguration. Lineage-specific duplications synergize with bidirectional nucleic acid exchange to achieve physiological control, and host genotype–dependent transcriptional plasticity optimizes resource extraction. This multi-layered adaptation enables C.deserticolato use resources from H. ammodendron in extreme arid environments. Future research is expected to reveal the developmental trajectory of molecular exchanges and the dynamic mechanisms of the parasitic relationship.

Methods

Biological materials

Whole-genome sequencing sample collection: on June 20, 2022, samples of C. deserticola were collected in Bayannur City (Inner Mongolia, China). To ensure sample quality and reduce potential host contamination, samples were specifically obtained from the middle section of the fleshy stem.

Transcriptome sequencing sample collection: on April 14, 2023, samples of C. deserticola were collected in Bayannur City (Inner Mongolia, China). The following groups were designed for transcriptome samples: two C. deserticola from the same root of H. ammodendron (C1 vs. C2), two C. deserticola from different roots of H. ammodendron (C1 vs. C3), and two C. deserticola from distinct H. ammodendron plants (C1 vs. C3). Samples included several key parts of C. deserticola: the haustorium, the bottom of the fleshy stem (1 cm from the haustorium connection, labeled as bottom), the middle section of the fleshy stem (labeled as middle), and the top of the fleshy stem (including the inflorescence stem, labeled as top). Roots of H. ammodendron were also collected. To ensure accuracy and reliability, three replicates of each tissue were collected, totaling 53 samples. After collection, the samples were washed in 75% ethanol for 10 min to prevent contamination before being prepared for transcriptome sequencing.

DNA and RNA extraction

DNA and RNA extraction: high-quality genomic DNA was extracted from the stem of C. deserticola using the cetyltrimethylammonium bromide method, consistent with established protocols (Ahmed et al., 2009). RNA was extracted from C. deserticola and H. ammodendron using the R6827 Plant RNA Kit (Omega Bio-Tek, CT, USA).

Sample quality assessment: the integrity of the extracted DNA and RNA was evaluated by 0.75% agarose gel electrophoresis. The integrity value of DNA was required to be greater than eight and that of RNA greater than six. The purity of DNA and RNA was assessed using a NanoDrop One spectrophotometer (Thermo Fisher Scientific, MA, USA) by measuring absorbance at specific wavelengths to evaluate the presence of impurities. DNA and RNA concentrations were determined using a Qubit 3.0 fluorometer (Life Technologies, CA, USA). High-quality DNA and RNA samples, verified through quality assessments, were preserved in TE buffer to ensure stability and prevent degradation.

PacBio HiFi library construction and sequencing

The experimental procedures followed the standard protocol established by PacBio. (1) Single-molecule real-time (SMRT) library construction: the qualified DNA was randomly fragmented into approximately 15-kb lengths using a g-TUBE (Covaris, MA, USA). The SMRTbell Express Template Prep Kit 2.0 (Pacific Biosciences, CA, USA) was used to construct a SMRTbell HiFi library with 15 μg of DNA. The library was purified using 1× AMPure PB magnetic beads (Pacific Biosciences), resulting in a DNA library with an insert size of 15 kb after removal of SMRTbells smaller than 15 kb. Finally, the library was purified using magnetic beads to create a PCR-free SMRTbell library. (2) SMRT library quality detection: quality control involved preliminary quantification with a Qubit 2.0 fluorometer (Life Technologies) and an Agilent 2100 instrument (Agilent, CA, USA). When the insert size met the expected criteria, qPCR was performed for precise quantification of library concentration. (3) Sequencing procedure: the library was sequenced at a concentration of 120 pM on the PacBio Sequel II platform. The resulting BAM files were converted into raw reads using bam2fastx, and the results were stored in FASTQ format. Raw reads were filtered using SMRTlink (v8.0), resulting in clean reads processed with CCS (v9.3) with the command “-minPasses 3,” yielding HiFi reads with a total length of 177 354 495 417 bp for subsequent assembly.

Hi-C library construction and sequencing

The experimental procedure followed the standard Illumina protocol. (1) Hi-C library construction (Belton et al., 2012): cells from fresh tissue were fixed with formaldehyde to preserve chromatin conformation, then digested with the restriction endonuclease Dpn II to generate sticky ends. Biotinylated bases were introduced for sticky-end repair, and DNA ligase was used to cyclize the interacting DNA fragments. Proteins cross-linked with DNA were removed using protease, after which the DNA was purified and fragmented to 300–700 bp. Biotinylated DNA was then captured using streptavidin magnetic beads for Hi-C library construction. (2) Hi-C library quality detection: quality control involved preliminary quantification with a Qubit 2.0 fluorometer (Life Technologies) and Agilent 2100 instrument (Agilent). When the insert size met the expected criteria, qPCR was performed for precise quantification of library concentration. (3) Sequencing procedure: high-throughput sequencing was performed on the Illumina NovaSeq 6000 platform (Illumina, CA, USA), yielding 631.08 Gb of raw reads. Raw reads were filtered to produce 619.66 Gb of clean data for subsequent assembly.

Genome assembly and chromosome construction

We used Hifiasm (v0.14.2) (https://github.com/chhylp123/hifiasm) (Cheng et al., 2021) with the overlap layout consensus algorithm to assemble all HiFi reads (177 354 495 417 bp). First, all reads underwent all-vs.-all overlap alignment, followed by three rounds of error correction on overlapping reads. Subsequently, an all-vs.-all comparison was performed to construct a string graph, from which haplotypes were established based on the reads’ bubbles. Finally, Purge_dups (v1.2.5) (https://github.com/dfguan/purge_dups) (Roach et al., 2018) was used to remove heterozygosity, resulting in the draft genome.

Using HICUP (v0.8.0) (https://www.bioinformatics.babraham.ac.uk/projects/hicup/) (Wingett et al., 2015), the clean Hi-C sequencing data (619.66 Gb) were aligned to the preliminary assembly of the C. deserticola genome (7.17 Gb), resulting in 154 298 995 uniquely mapped paired-end reads. After filtering out 33 470 213 invalid interaction pairs, 120 828 782 valid interaction pairs remained, 61.07% (94 237 694) of which were unique valid paired-end pairs used for auxiliary assembly.

ALLHIC (v0.9.8) (https://github.com/tangerzhang/ALLHIC) (Zhang et al., 2019) was used to apply an agglomerative hierarchical clustering algorithm for ordering and orienting the contigs. Interactions between contigs were converted into Hi-C files using 3D-DNA (v180419) (https://github.com/theaidenlab/3d-dna) (Dudchenko et al., 2017) and Juicer (v1.6) (https://github.com/aidenlab/juicer) (Durand et al., 2016b). Juicebox (v1.11.08) (https://github.com/aidenlab/Juicebox) (Durand et al., 2016a) was then used for manual correction and redundancy removal of the oriented contigs. Ultimately, a total genome sequence of 6 074 066 816 bp was assigned to 20 pseudochromosomes, achieving a mapping rate of 96.97%. The Hi-C interaction heatmap was generated using HiCExplorer (v3.6) (https://hicexplorer.readthedocs.io/en/latest/) (Wolff et al., 2020) on the basis of contig interaction strengths and positional relationships to assess the effectiveness of the Hi-C-assisted genome assembly.

Genome evaluation

BWA (v0.7.17) (Li and Durbin, 2009) was used to map quality-controlled clean reads from second-generation sequencing to the draft genome assembly, achieving a DNA read alignment rate of 99.85% (> 90%) and a coverage of 99.24% (> 95%). BUSCO (v5.3.0; parameters: –evalue 1e-05) (Simão et al., 2015) was used to construct a set of single-copy genes from closely related species in the evolutionary branch of C. deserticola, using the OrthoDB database. The genome sequences of C. deserticola were compared with the single-copy homolog gene set using hmmsearch, yielding a complete BUSCO score of 92.2%. An LTR assembly index of 23.48 (> 20) was calculated using LTR_retriever (v2.9.0) (Ou and Jiang, 2018). The assembled scaffolds were segmented into 10-kb bins, and both GC content and sequencing depth (GC-depth) distributions were plotted to assess the uniformity of sequencing depth. The results showed that GC content was concentrated with no contamination from non-target species, indicating a relatively consistent GC content (average: 38.62%) and sequencing depth distribution (average: 63.57×), thus demonstrating good sequencing depth uniformity.

Repetitive sequence analysis

We identified the repeat sequences in the C. deserticola genome through de novo and homology-based annotation. (1) De novo annotation: we used RepeatModeler (v1.0.11; parameters: BuildDatabase-name mydb; RepeatModeler-database mydb-pa 10) (Rodriguez and Makałowski, 2022) in conjunction with LTR_FINDER (v1.07; parameters: -threads 16 -harvest_out -size 1000000 -time 300) (Xu and Wang, 2007) to predict LTR sequences. We then used LTR_retriever (v2.9.0; parameters: -threads 16) to deduplicate the sequences predicted by LTR_FINDER and merge the de novo sequences. (2) Homology-based annotation: the de novo sequences were combined with RepBase (v20181026) and subjected to homology annotation using RepeatMasker (v4.0.9; parameters: -nolow -no_is -norna -parallel 2), resulting in the de novo + RepBase output. Subsequently, we used RepeatProteinMask (v4.0.9; parameters: -noLowSimple -pvalue 0.0001) (Tarailo-Graovac and Chen, 2009) to predict transposable element (TE) protein-type repeat sequences. Finally, we merged all repeat prediction results to remove duplicates, yielding a comprehensive set of genomic repeat sequences referred to as combined TEs.

Gene structure prediction and function annotation

We identified protein-coding genes in the C. deserticola genome through homology-based, de novo, and RNA sequencing–based annotation. (1) Homology-based annotation: we aligned protein sequences from closely related species (L. luchunensis, O. cumana, P. aegyptiaca, S. asiatica, and S. hermonthica) in the UniProt SProt database (release-2020_05) to the C. deserticola genome using tblastn (v2.7.1; parameters: -t 16 -q 7) (Camacho et al., 2009). Exonerate (v2.4.0; parameters: -model protein2genome –showtargetgff 1) was then used to predict transcripts and coding regions on the basis of the alignment results (Slater and Birney, 2005). (2) De novo annotation: we performed de novo annotation using Augustus (v3.3.2; parameters: –uniqueGeneId=true –noInFrameStop=true –gff3=on –strand=both) (Stanke et al., 2008), GlimmerHMM (v3.0.4; parameters: -f -g) (Delcher et al., 2007), and Genscan (v1.0; parameters: HumanIso.smat) (Shah et al., 2003). (3) RNA sequencing–based annotation: we filtered raw reads from second-generation transcriptome data using Fastp (v0.21.0; parameters: -j out.json -h out.html) (Chen et al., 2018) and aligned clean reads to the reference genome using HISAT2 (v2.1.0; parameters: -p 15) (Kim et al., 2019). BAM files were then analyzed using StringTie (v2.1.4; parameters: -p 15) (Kovaka et al., 2019) to predict transcripts, and TransDecoder (v5.1.0) was used to predict coding frames, resulting in predicted protein-coding genes. Finally, we integrated all prediction results using MAKER (v2.31.10; parameters: maker_exe.ctl maker_opts.ctl maker_bopts.ctl –ignore_nfs_tmp -fix_nucleotides) (Holt and Yandell, 2011) and evaluated the annotation results using BUSCO (v5.2.2; parameters: -c 20 -m genome –augustus –long; input genome sequence) (Manni et al., 2021), which yielded a complete BUSCO score of 93.3% (> 90%). A frequency distribution plot of gene lengths, CDS lengths, intron lengths, and exon lengths was created for C. deserticola and its closely related species using a window size of 10 bp.

Predicted protein sequences were aligned to public protein databases (UniProt Consortium, 2023), NR (Pruitt et al., 2005), and KEGG (Kanehisa and Goto, 2000) using Diamond Blastp (v2.0.11.149; parameters: –evalue 1e-5) (Buchfink et al., 2015) for functional annotation. KOBAS (v3.0) (Xie et al., 2011) was then used to search the KEGG gene database, assigning KEGG Orthology and inferring KEGG pathway associations. Subsequently, InterProScan (v5.52-86.0; parameters: -goterms -pa -dp -verbose -cpu 20) (Blum et al., 2021) was used to predict conserved sequences, motifs, and structural domains in the proteins by comparison with the CDD, Gene3D, Hamap, Panther, Pfam, Phobius, Pirsf, Pirsr, Prints, Prosite, Sfld, Smart, Superfamily, and Tigrfam databases. In addition, hmmscan (v3.3.2; parameters: -E 0.01) (Mistry et al., 2021) was used for structural domain prediction. In total, 48 438 (88.65%) of the protein-coding genes were annotated in at least one of these databases, indicating a good result for gene functional annotation.

Non-coding RNA annotation

To identify tRNA and rRNA sequences in the C. deserticola genome, we used tRNAscan-SE (v1.23; parameters: -q) (Chan et al., 2021) and RNAmmer (v1.2) (Lagesen et al., 2007), respectively. Non-coding RNA sequences were predicted using INFERNAL (v1.1.2; parameters: –cut_ga –rfam –nohmmonly –cpu 15) (Nawrocki and Eddy, 2013) on the basis of the Rfam database.

Gene-family clustering and phylogenetic analysis

To clarify the phylogenetic relationships between C. deserticola and the major clades of Orobanchaceae, we collected reference genomic sequences from seven Orobanchaceae species (C. deserticola, L. luchunensis, O. cumana, P. aegyptiaca, S. hermonthica, S. asiatica, and Phtheirospermum japonicum) and six outgroup species (H. ammodendron, A. sinensis, P. notoginseng, P. euphratica, C. songorica, and W. mirabilis). Using OrthoFinder (v2.3.12; parameters: -M msa) (Emms and Kelly, 2019), we identified 82 774 orthologous gene families, including 143 single-copy orthologs. A supergene was then constructed by aligning the 143 single-copy orthologs using MUSCLE (v3.8.31) (Edgar, 2004) and trimAI (v1.2rev; parameters: -gt 0.2) (amino acid length ≥ 100) (Capella-Gutiérrez et al., 2009). Phylogenetic analysis was performed by the maximum likelihood method using RAxML (v8.2.10) (Stamatakis, 2014) with 1000 bootstrap replicates. Species divergence times were estimated using the MCMCTree subroutine from the PAML package (v4.9; parameters: nsample=3000000; burnin=8000000; seqtype=0; model=4) (Yang, 2007), calibrated with fossil time nodes from TimeTree (http://timetree.org/).

Gene-family expansion and contraction analysis

Using CAFE (v3.1; parameter: --filter) (De Bie et al., 2006), we estimated the numbers of ancestral gene family members for each branch on the basis of the species phylogenetic tree and the gene-family clustering results. This approach enabled us to predict the contraction and expansion of gene families in C. deserticola relative to its ancestors. Gene families with a family-wide P value of less than 0.05 were considered to have undergone significant expansion or contraction. We performed enrichment analyses based on the Gene Ontology (Ashburner et al., 2000) and KEGG databases using clusterProfiler (v3.14.0) (Wu et al., 2021).

WGD analysis

We compared the protein sequences of C. deserticola and O. cumana using BLAST (v2.6.0+) (Altschul et al., 1990) and MCScanX, identifying syntenic genome blocks and calculating synonymous mutation frequencies (Ks) of syntenic gene pairs using PAML (v4.9) (Tang et al., 2008). Ks density plots were visualized using ggplot2 (v2.2.1), revealing WGD events. The timing of WGDs was estimated using the formula T = Ks/2r (r = 5.26 × 10⁻⁹) (Song et al., 2020).

Collinearity analysis

We aligned gene sequences of C. deserticola and O. cumana using LAST (v1170) (Frith et al., 2010) to identify similar gene pairs, then used JCVI (v0.9.13) (Tang et al., 2024) to extract syntenic genes from the annotation files (GFF). A map illustrating the genomic synteny between the two species was generated using TBtools (v2.069) (Chen et al., 2023).

Genome duplication analysis

Using DupGen_finder (v1170) (Qiao et al., 2019) with default parameters, we identified various gene duplication patterns, including WGD, tandem duplication, proximal duplication, transposed duplication, and DSD. Subsequently, we estimated Ka (the number of non-synonymous substitutions per non-synonymous site), Ks (the number of synonymous substitutions per synonymous site), and Ka/Ks values for gene pairs generated by different duplication modes using KaKs_Calculator (v2.0) (Wang et al., 2010) under the YN model.

Analysis of selection pressure

Protein sequences of identified single-copy gene families were aligned using MAFFT (parameters: –localpair –maxiterate 1000) (Katoh and Standley, 2014) and then converted to codon alignments using PAL2NAL (v14) (Suyama et al., 2006). The CODEML (Bielawski et al., 2016) subroutine of PAML (v4.9) was used to detect genes under positive selection (P < 0.05) via likelihood ratio tests between model A (foreground ω > 1) and the null model, which were performed with the chi2 program in PAML.

Identification of pathway-specific genes

We used a three-step approach to identify candidate genes in pathways of interest. Amino acid sequences from C. deserticola were used as queries in a BLASTP search (< 1 × 10⁻⁵), followed by HMMER (Johnson et al., 2010) searches using domain structure models (v3.1b2). The results were validated using the CDD (https://www.ncbi.nlm.nih.gov/cdd/) and SMART (http://smart.embl.de/) databases and manually curated to retain complete domain sequences. Transcription factors for candidate genes were predicted using PlantTFDB (Jin et al., 2017).

Statistical analysis

All quantitative and statistical analyses were performed using the R computational environment and the packages mentioned above. The significance of differences in means between two groups was evaluated using an unpaired two-tailed Student’s t-test. Differential gene expression was assessed using the Wald test, followed by adjustment for multiple testing through the Benjamini–Hochberg procedure, as implemented in DESeq2. Pearson’s correlation coefficient, together with its significance, was calculated using the R package “performance analytics” (v1.5.3).

Data and code availability

The genome sequences, genome assembly, and transcriptome sequencing data have been deposited in the Chinese National Genomics Data Center (https://ngdc.cncb.ac.cn/) under BioProject accession number PRJCA035066.

Funding

This study was supported by the Science and Technology Innovation Plan of the Shanghai Science and Technology Commission (grant number: 22YF1420500), the Fundamental Research Funds for the Central Universities (grant numbers: KLSB2022QN-01 and KLSB2024KF-06), Shanghai Jiao Tong University Scientific and Technological Innovation Funds (grant numbers: YG2022QN070 and 19X190020005), the National Natural Science Foundation of China (grant number: 81872274), and the Initiation Program for New Teachers of Shanghai Jiao Tong University (grant number: 23X010502168).

Acknowledgments

No conflict of interest is declared.

Author contributions

R.Z. and J.H. analyzed the data and wrote the manuscript; J.H. and W.H. conceived and designed the study and provided research funding. R.Z., J.H., and W.H. designed the research. J.H., R.Z., H.X., J.W., J.S., Y.Y., J.X., and Y.D. participated in sample collection. All authors take responsibility for the entire content of this manuscript and have approved its submission.

Published: October 30, 2025