Background & Summary

Agricultural weeds are a key constraint factor in agricultural production because they can compete with crops for light, nutrients, moisture and space, leading to decreased crop quality and yield1. Among all types of crop pests, weeds are known to cause the greatest reduction in crop yield2. As reported, the potential yield reduction caused by weeds could be up to 23%, 37%, 37%, 40%, 36% and 30% in wheat, soybeans, rice, maize, cotton and potatoes, respectively, with the average yield reduction of 34%, which is far more than the impacts of animal pests and pathogens3. With the increasing world population and decreasing available resources, weed management is a particularly important and challenging task.

From the evolutionary point, weeds is an excellent example of rapid adaptation to changing environments, due to their abundant genetic variation and plasticity, weeds can evolve more rapidly than crops4,5. Weeds have shown a remarkable capacity to quickly adapt to changing environmental factors, agricultural techniques, and weed-control strategies6. Understanding the evolution and adaptation of weeds is essential for effective weeds management. However, the genetic characterization of weed systems has received comparatively little attention7.

Genomic studies is an important approach to identify the origin of weed species and study their adaptive evolution, providing a basis for establishing effective weed management strategies8. However, despite the substantial impact of weeds on agricultural production, research on weeds have not received the necessary attention in both traditional molecular biology and genomic analyses7,9. To date, 2,847 plant species have been identified as weeds10. However, only approximately 26 species have been subjected to sequencing and de novo genome assembly9.

Cyperus iria (rice flatsedge), belonging to Cyperaceae family, is an annual sedge with a fibrous root system and C4 photosynthetic system11. It can reach 60 cm in height12. C. iria is native to tropical and subtropical countries13 with the ability to multiply rapidly and readily adapts to ecological niches14. The high seed production of C. iria (3000–5000 seeds per plant) combined with its short life cycle leads to a very high rate of reproduction15,16. C. iria is an aggressive weed throughout the world and has become a major agricultural weed in rice production systems in 22 countries16,17. It has been reported that a 64% reduction in rice yield is the result of C. iria infestation throughout the crop growth period18. Competition between C. iria and rice during the first 30 days can reduce the yield of rice by up to 12.9%, and infestation during the first 40 days can the yield of transplanted rice yield by up to 43.5%19. Chemical herbicides have become the preferred choice for C. iria control because it is highly efficient, less labour intensive, and cost effective. However, as a result of extensive herbicide use, C. iria has developed resistant populations to ALS (acetolactate synthetase) inhibitors, such as pyrazosulfuron-ethyl, halosulfuron-methyl and penoxsulam20. C. iria herbicidal resistance remains largely unknown at the molecular level. Therefore, understanding the function of C. iria genes and the mechanisms by which it has evolved to become invasive is critical to the management of this weed. In addition, as a traditional Chinese medicine, C. iria has been found to show promising pharmacological effects12,13,21,22. However, despite its medical importance, genomic and genetic information on C. iria is still very limited, severely hampering molecular and genetic research into this devastating weed.

Here, we generated a chromosome-level genome of the C. iria based on Illumina, PacBio and Hi-C technology. The assembled genome size of C. iria was 479.08 Mb (92.08% of the estimated genome size) with a contig N50 of 7.02 Mb. 99.65% of the assembled genome were anchored to 68 pseudochromosomes. The number of predicted protein-coding genes is 47,395, of which 93.26% were annotated, and 37.69% repetitive sequences were identified. The high-quality chromosome-level C. iria reference genome assembly, provides a strong basis for development of new strategies for the successful management of this aggressive weed in the future.

Methods

Plant material, library construction and sequencing

Seeds of the C. iria was collected from Huzhou, Zhejiang Province (30° 87′ N, 120° 10′ E) in 2021 and stored in 4 °C until used. Plants of C. iria were grown in small pots in the Experimental Greenhouse of Zhejiang University’s Zijingang Campus (30° 30′ N, 120° 08′ E) in Hangzhou, China, with 16 hours of light and temperatures of 25 °C at day and 20 °C at night. Fresh young leaves from the same one C. iria plant were collected and immediately frozen in liquid nitrogen. The cetyltrimethylammonium bromide (CTAB) method was employed to extract high quality, high molecular weight genomic DNA from young leaves23. The genomic DNA was used for the construction of an Illumina paired-end (PE) library with ∼450 bp insert sizes, and the libraries were constructed on an Illumina NovaSeq sequencing platform using Next-Generation Sequencing (NGS) following the standard procedure. For PacBio sequencing, the genomic DNA of C. iria was used for the construction of PacBio SMRTbell libraries according to the standard SMRTbell library preparation protocol, and the PacBio Sequel II platform was then utilized to sequence the libraries. The Hi-C library constructed from fresh young leaves of same one C. iria plant was then sequenced on an Illumina NovaSeq sequencing platform.

Total RNAs were extracted from four tissues (root, stem, leaf, and flower) of the same C. iria plant using TRIzol reagent. RNA from these four tissues was mixed equally for the construction of a PacBio Iso-Seq library. The cDNA was generated from mixed RNA using the SMARTer PCR cDNA Synthesis Kit (Clontech). The purified cDNA was then used to construct a Iso-Seq SMRTbell library using the SMRTbell Express Template Prep kit 1.0, which was subsequently sequenced on the PacBio Sequel II platform.

Genome survey

After removing low-quality reads, 10,000 high-quality data pairs were randomly selected and mapped to the NCBI nucleotide (NT) database, revealing the top five matched species. Jellyfish software (version 2.3.0)24 was used for k-mer analysis of all high-quality data. Based on the k-mer frequency analysis (k-mer = 19), genome characteristics including genome size, heterozygosity and repeat rate, were estimated by using GenomeScope25. The genome size was calculated as following Eq. (1):

$$G=\frac{N\left(L-K+1\right)-B}{D}$$
(1)

wherein G represents the genome size, N means the total number of reads, L means the average length of reads, K is the k-mer length, and D is the peak depth that is estimated from the k-mer distribution (pkdepth). Low frequency k-mers before the first valley were discarded to minimise the influence of sequencing errors. The k-mer analysis (k-mer = 19) analysis determined the genome size to be 520.28 Mb, with the heterozygosity rate and repetitive fraction were 0.08% and 47.23%, respectively (Table S1 and Fig. 1a). These findings suggest that the genome of C. iria belongs to the simple genome.

Fig. 1
figure 1

Genome size estimation and ploidy levels. (a) Estimation of genome size and heterozygosity based on k-mers (k = 19). (b) Ploidy inference based on Smudgeplot analysis.

Full size image

Smudgeplot26 was used to estimate ploidy levels from corrected reads with a k-mer value of 19. The Smudgeplot performs gymnastics with heterozygous k-mer pairs extracted from k-mer count databases. It can disentangle genome structure by comparing the sum of coverages of k-mer pair (CovA + CovB) to their relative coverage (CovB/(CovA + CovB))26. Smudgeplot analysis indicates this genome assembly was a diploid (AB) with with a confidence of 0.91 (Fig. 1b).

Genome assembly

To obtain a contig-level genome, PacBio High Fidelity (HiFi) reads were assembled using Hifiasm27. Totally 18.51 Gb Pacbio HiFi reads were generated, resulting in a genome assembly of 479.08 Mb with a contig N50 of 7.02 Mb and a GC content of 35.64% (Tables 1, 2). Purge_dups (version 1.2.5)28 was used to remove redundant sequences in the polished assembly. After that, only 80 contigs (462.83 Mb) remained for further analysis (Table 3). BWA (version 0.7.12)29 was used to compare the corrected high-quality Illumina reads sequence with the assembled genome sequence. The results showed that the read sequence comparison rate reached 99.25%, and the average sequencing depth was 215×. The sequence depth of 20× or more accounted for 99.78% (Table 4).

Table 1 Statistics of PacBio sequencing data of C. iria genome.
Full size table
Table 2 Statistics of pre-assembly of C. iria genome.
Full size table
Table 3 Statistics of pre-assembly of C. iria genome after redundancy removal.
Full size table
Table 4 Statistics on the coverage of reads of C. iria genome.
Full size table

To obtain a chromosome-level genome assembly of the C. iria, HiC-Pro software (version 3.1.0)30 was used to align clean Hi-C reads to the draft genome sequences in comparison mode. All invalid read pairs were removed using HiC-Pro software except unique mapped paired-ends, which were retained for further analysis. In total, Hi-C sequencing produced 65.73 Gb clean reads (Table 5), and 99.15% of Hi-C reads mapped to assembled contigs, including 65.30% unique mapped read pairs (Table 5). PacBio sequencing assemblies were organized into chromosome-level scaffolds by integrating valid interaction pairs from the unique mapped read pairs (Table 5). The ensuing Hi-C reads were then used to correct misjoins, order, orient, and anchor in the draft genome assembly with the 3D de novo assembly (3D-DNA) (version 201008)31. We used the Hi-C data to attach the draft genome to the chromosome level, 68 pseudochromosomes were constructed, whereby 99.65% of the assembled sequences were anchored. The heatmap of Hi-C interaction revealed a higher intensity of interactions in diagonals is higher than that in nondiagonal positions in each group, indicating the chromosome-level genome assembly was complete and robust (Fig. 2a). Chromosome lengths varied from 3,585,826 bp (Chr54) to 10,521,918 bp (Chr66) (Table S2 and Fig. 2b). Subsequently, the assembled results was polished using Racon (version 1.4.20)32. Benchmark for Universal Single Copy Orthologues (BUSCO, version 5.4.3)33 with the plant dataset (embryophyte) was used to assess the genome completeness and continuity. BUSCO analysis showed that 95.11% (0.50% fragmented and 4.40% missing BUSCOs) of the BUSCO genes in the C. iria genome were successfully identified as complete BUSCO, which indicates the high completeness of the genome assembly (Table S3). Furthermore, the contiguity of the genome was evaluated by calculating LTR Assembly Index (LAI) using LTR_retriever (version 2.9.9)34 with default parameters. The LAI value of the genome assembly was 9.77.

Table 5 Statistics of Hi-C data.
Full size table
Fig. 2
figure 2

Overview of the C. iria genome assembly. (a) Hi–C interaction heatmap for C. iria. (b) Chromosomal features of C. iria. The circle diagram from the outside to the inside represents: Pseudochromosome, Gene density, GC content, LTR/Gopia transposon density, LTR/Gypsy transposable factor density. The connection lines in the circle are synteny lines of paralogous sequences in the genome.

Full size image

Genome annotation

To identify repeat sequences, a combination of the homology-based prediction and de novo-based prediction was performed. For de novo-based prediction, RECON (version 1.0.8)35, RepeatModeler (version 2.0.4)36, and RepeatScout (version 1.0.6)37 were used to identified repetitive elements. For homology-based prediction, RepeatMasker (version 4.1.4)38 was used to identify repetitive elements by comparing to consensus sequences in the Repbase library39. The analysis of homology-based comparisons and de novo-based methods revealed that 37.69% of the C. iria genome consists of repetitive sequences, with LTRs representing 8.95% of the total. Gypsy was identified as the predominant LTR type (5.85%), followed by Copia (2.69%) (Table 6).

Table 6 Transposable elements in C. iria genome.
Full size table

Non-coding RNA (ncRNA) species, such as rRNA, tRNA, miRNA, and snRNA, were annotated with RNAmmer (version 1.2) (for rRNAs prediction)40, tRNAscan-SE (version 1.3.1) (for tRNAs prediction)41, and Perl program Rfam (version 1.0) (for other two non-coding RNAs prediction)42. Totally 2,355 ncRNA genes were also identified, which included 599 rRNA, 985 tRNA and 771 other ncRNA (Table 7).

Table 7 The statistical results of non-coding RNA of C. iria.
Full size table

To predict protein-coding genes, a combination of homology-based, de novo-based, and transcript-based prediction approaches was used. The de novo-based prediction was conducted with Augustus (version 3.3.2)43, GlimmerHMM (version 3.0.4)44, GeneMark (version 4.35)45, and GeneID (version 1.4) with default parameters. The homology-based prediction was performed using Exonerate (version 2.2.0)46 together with protein sequences from Carex littledalei, Arabidopsis thaliana and Oryza sativa. For transcript-based prediction, the high-quality full-length transcript data were compared and spliced using PASA (version 2.5.2)47 to obtain the corresponding gene prediction results. Finally, EvidenceModeler (version r2012-06-25)48 was used to integrate homology-based, de novo-based, and transcript-based prediction results. The prediction results revealed that C. iria genome contained 47,395 protein-coding genes, and the average gene length is 2,762.5 bp (Table 8). The BUSCO analysis evaluated the gene set completeness, revealing that 96.28% (1,554 genes) of the BUSCO genes were present in the C. iria gene set. This finding confirms the high quality of gene prediction. BLASTP (version 2.0.14.152) (E-value < 1 × 10−5) searches against the NR and SwissProt databases were performed for the final protein-coding, functionally annotated genes. Functional domains were then obtained by searching publicly available databases using InterProScan (version 5.61–93.0)49. GO annotation of protein-coding genes was performed using interproscan. The KEGG pathway annotation of protein-coding genes was mainly performed by KEGG Automatic Annotation Server (KAAS, version 2.1)50. Approximately 93.26% of the protein-coding genes were functionally annotated (Table 9). To further validate the accuracy of our gene annotations, we compared the gene sequences with transcript sequences using geneBody_coverage (version 5.0.1)51. This analysis assessed the sequence coverage across the entire gene length, from the 5′ to the 3′ end, to determine if the sequencing reads were evenly distributed. Figure S1 illustrates the alignment results, demonstrating that the sequencing reads were uniformly distributed without bias towards the 5′ or 3′ ends. The mapping rate of transcript sequences to the gene sequences was 93%, further supporting the accuracy of the gene annotations.

Table 8 The statistical results of gene prediction of C. iria.
Full size table
Table 9 The statistical results of gene function annotation of C. iria.
Full size table

Data Records

Raw Illumina, PacBio HiFi and Hi-C of C. iria genome sequencing data were deposited in the NCBI BioProject database under project accession number PRJNA115799452, with accession numbers SRR3058810953 for Illumina sequencing data, SRR3058810854 for Pacbio sequencing data, SRR3058810755 for Hi-C sequencing data, SRR3058810656 for Iso-Seq data. The genome assembly has been deposited at GenBank under the accession JBHOFI00000000057. The annotation of the C. iria genome has been submitted to the online open-access repository Figshare58 database.

Technical Validation

BWA (version 0.7.12)29 was used to compare the corrected high-quality Illumina reads sequence with the assembled genome sequence. The read sequence comparison rate reached 99.25%, and the average sequencing depth was 215×. The sequence depth of 20× or more accounted for 99.78% (Table 4). Benchmark for Universal Single Copy Orthologues (BUSCO, version 5.4.3)33 with the plant dataset (embryophyte) was used to assess the genome completeness and continuity. BUSCO analysis showed that 95.11% (0.50% fragmented and 4.40% missing BUSCOs) of the BUSCO genes in the C. iria genome were successfully identified as complete BUSCO, which indicates the high completeness of the genome assembly (Table S3). Then, the assembly continuity was determined by analyzing the LTR Assembly Index (LAI), the LAI score was 9.77. The heatmap of Hi-C interaction revealed a higher intensity of interactions in diagonals is higher than that in nondiagonal positions in each group, indicating the chromosome-level genome assembly was complete and robust (Fig. 2a). The BUSCO analysis also used to evaluate the gene set completeness, revealing that 96.28% (1,554 genes) of the BUSCO genes were present in the C. iria gene set. This finding confirms the high quality of gene prediction.