A high-quality chromosome-level genome assembly of Antiaris toxicaria

Abstract

Objectives

Antiaris toxicaria is a tall tree belonging to the Moraceae family, known for its medicinal value. Its latex contains various cardiac glycosides, which hold significant research and potential application value. However, the lack of genomic resources for A. toxicaria currently hinders molecular genetic studies on its medicinal components. For its effective conservation and elucidation of the distinctive genetic traits of and medical components, we present its chromosome-level genome assembly.

Data description

Here, we assembled two haplotypes of A. toxicaria, including a 671.73-Mb HapA subgenome containing 27,213 genes and a 666.41-Mb HapB subgenome containing 28,840 genes. Their contig N50 sizes were 90.18 and 90.29 Mb, respectively. The transposable elements represented 61.15% and 64.13% of the total assembled genome in HapA and HapB subgenome, respectively. A total of 27,213 and 28,840 genes were predicted in the two haplotypes. Hopefully, this chromosome-level genome of A. toxicaria will provide a valuable resource to enhance understanding of the biosynthesis of medicinal compounds.

Objective

Antiaris is a genus in the Moraceae, all species are large trees with tall and straight trunk and plank-like roots. There are approximately seven species and three varieties worldwide, with only one species, Antiaris toxicaria, found in China, distributed in Guangdong, Hainan, Guangxi, and southern Yunnan. A. toxicaria is well known for its ornamental value, ecological function and cultural importance. Previous studies on the secondary metabolites of this species have indicated that its latex [1, 2] and seeds [3] are rich in cardiac glycosides. The sap of the leaves and branches of A. toxicaria contains highly toxic substances [4], with the main toxic components being cardiac glycosides such as α-antiarin [5], antioside [6, 7], and convallatoxin [8, 9], which have effects including enhancing heart function, inducing vomiting and diarrhea, and possessing anesthetic properties [10]. Additionally, HPLC screening of A. toxicaria extracts revealed the presence of gallic acid, catechins, chlorogenic acid, caffeic acid, ellagic acid, epigallocatechin, rutin, isoquercitrin, quercitrin, quercetin and kaempferol [11]. Therefore, A. toxicaria holds significant research and commercial value. However, our understanding of the biosynthesis and regulatory mechanisms of secondary metabolites in A. toxicaria is limited, and further research is needed on the candidate genes and transcription factors involved in cardiac glycoside biosynthesis pathways.

In this study, we successfully assembled the A. toxicaria chromosome-level genome using high-fidelity (HiFi) reads and high-throughput chromosome conformation capture (Hi-C) sequencing technologies. This study reports the high-quality genome of A. toxicaria. We believe that this research will provide important resources for studying the biosynthetic mechanisms of this species.

Data description

A. toxicaria samples were obtained from the South China Botanical Garden (23.18°N, 113.36°E), Guangzhou, China. Fresh leaves of A. toxicaria were collected for PacBio HiFi, ONT ultralong, and Hi-C sequencing. A PCR-free SMRTBell library was constructed using high-quality purified long reading DNA for PacBio HiFi sequencing. The ONT PromethION sequencer was used to generate ONT ultralong reads. Hi-C libraries were constructed and sequenced using BGI platform. Stems, leaves, and seeds of A. toxicaria were frozen in liquid nitrogen and stored at − 80 °C for transcriptome analyses. All Illumina sequencing data were filtered to obtain clean data using the fastp v0.23.1 [12] for subsequent analysis. A total of 128.34 Gb (~ 191.06 × coverage) paired-end Illumina reads (Table 1; Data set 1), 32.7 Gb (~ 48.68 × coverage) PacBio HiFi long reads (Table 1; Data set 2), 16.71 Gb ONT Ultra-long reads (~ 24.87 × coverage) (Table 1; Data set 3), and Hi-C reads (~ 126.14×coverage) (Table 1; Data set 4) were generated for the genome survey, and assembly (Table 1; Data file 1).

Table 1 Overview of data files/data sets

Before genome assembly, we used the GCE (Genomic Charactor Estimator) v 1.0.2 [30] to assess the genome size based on Illumina short reads. The genome size of A. toxicaria was estimated to be approximately 729.84 Mb based on the assessment results when using kmer length of 17 bp, showing a high degree of repeat content (70.62%) and heterozygosity (0.57%) (Table 1; Data file 2).

The PacBio HiFi, ONT Ultra-long, and Hi-C data were assembled using Hifiasm [31] with the default parameters. Then, the Hi-C data was aligned to the HapA and HapB subgenomes, respectively, and classified as valid or invalid interaction pairs using the Juicer pipeline [32] and YaHS v1.2 [33]. Meanwhile, misassembled contigs were detected, corrected manually and oriented to chromosomes through Juicebox v1.11.08 [32]. The corrected ONT and PacBio HiFi reads were used to replace the gap region using TGS-GapCloser v1.2.1 [34], and then obtained the haplotype-resolved gap-free genome of A. toxicaria. Finally, the A. toxicaria genome was ultimately phased into two haplotypes, comprising a total of 26 pseudochromosomes, with HapA spanning approximately 671.73 Mb and featuring a contig N50 of 90.18 Mb (Table 1; Data files 3–4; Data set 5). Similarly, HapB spans around 666.41 Mb with a contig N50 of 90.29 Mb (Table 1; Data files 3–4; Data set 6). Moreover, the GC content of HapA was 35.65%, while that of HapB was 35.61% (Table 1; Data file 4).

The genome completeness was assessed by searching the gene content of the embryophyta_odb10 database (1,614 expected genes from the embryophyta) with BUSCO v4.1.2 [35], showed that, the proportions of complete BUSCOs (including single-copy and multi-copy) of these two haplotypes were 98.5% and 98.6%, respectively (Table 1; Data file 4). The quality of repetitive genomic regions was assessed using the LAI v3.2 program [36], which exhibited LAI values of 16.4 (HapA) and 14.72 (HapB) (Table 1; Data file 4). Then the per-base consensus accuracy (QV) was estimated with Merqury v1.365 [37] using PacBio HiFi long reads, resulting in QV values of 47.13 and 47.1 (Table 1; Data file 4). Short-reads and long-reads were mapped to the genome with BWA v0.7.13-r1126 [38] and Minimap2 v2.21 [39], and we found that the genome coverage of sequencing data exceeded 99% (Table 1; Data file 4).

Protein-coding genes was predicted using homology-based, transcriptome-based, and ab initio prediction methods. First, we used homologies as protein-based evidence for predicting gene sets using GeneWise v2.4.1 [40]. Transcriptome data were mapped using HISAT2 v2.1.0 [41] (Table 1; Data set 7). ab initio prediction using packages AUGUSTUS v3.4.0 [42], trained by the transcriptome data. To generate a comprehensive protein-coding gene set, we used the GETA v2.6.1 (Genome-wide Electronic Tool for Annotation) pipeline (https://github.com/chenlianfu/geta) to integrate annotations from all homology-based, transcriptome-based, and ab initio predictions. Then Functional annotation of the protein-coding genes was carried out by blast searches against databases, including the NCBI nr [43], Swiss-Port [44], KOG [45], eggNOG [46], Pfam [47], GO [48], and KEGG [49]. In total, we obtained 27,213 and 28,840 protein-coding genes of the HapA and HapB subgenomes, respectively (Table 1; Data file 4; Data sets 8–9). Moreover, 26,906 (98.87%) genes of the HapA subgenome and 26,360 (98.8%) genes of the HapB subgenome were supported by multiple functional databases (Table 1; Data files 5–7).

To identify Transposable elements (TEs), we used the pipeline of Extensive de-novo TE Annotator (EDTA) v2.1.0 [50], which combines both structural-based and homology-based predictions. For noncoding RNA prediction, the tRNA genes were predicted using tRNAscan-SE v2.0.6 [51]. Others, including miRNA, rRNA and snRNA genes, were detected by comparison with the Rfam database [52] using CMsearch v1.1.3 [53] with the default parameters. A total of 427.39 Mb of TEs were identified, accounting for 64.13% of the HapB subgenome, which was higher than the HapA subgenome (Table 1; Data file 8; Data sets 10–11). In addition, the long terminal repeat retrotransposons (LTRs) were the predominant repeats covering 55.63% (370.77 Mb) of the HapB subgenome, and the Copia and Gypsy-type LTRs were the largest LTR subfamilies, accounting for 15.89% (105.89 Mb) and 39.10% (260.58 Mb), respectively (Table 1; Data file 8; Data sets 10–11). Moreover, 456 tRNAs and 111 miRNAs were identified in the A. toxicaria subgenome (Table 1; Data file 9; Data sets 12–13). 1,637 and 1,182 rRNAs were identified in the HapA and HapB subgenomes, respectively (Table 1; Data file 9; Data sets 12–13).

Limitations

Genome and transcriptome data are available in this study, but there is a lack of proteome and metabolome data from different tissues, as well as multi-omics correlation analysis.