Draft genome of the herbaceous bamboo Raddia distichophylla

Abstract

Bamboos are important nontimber forest plants widely distributed in the tropical and subtropical regions of Asia, Africa, America, and Pacific islands. They comprise the Bambusoideae in the grass family (Poaceae), including approximately 1700 described species in 127 genera. In spite of the widespread uses of bamboo for food, construction, and bioenergy, the gene repertoire of bamboo still remains largely unexplored. Raddia distichophylla (Schrad. ex Nees) Chase, belonging to the tribe Olyreae (Bambusoideae, Poaceae), is a diploid herbaceous bamboo with only slightly lignified stems. In this study, we report a draft genome assembly of the ∼589 Mb whole-genome sequence of R. distichophylla with a contig N50 length of 86.36 Kb. Repeat sequences account for ∼49.08% of the genome assembly, of which LTR retrotransposons occupy ∼35.99% of the whole genome. A total of 30,763 protein-coding genes were annotated in the R. distichophylla genome with an average transcript size of 2887 bp. Access to this herbaceous bamboo genome sequence will provide novel insights into biochemistry, molecular marker-assisted breeding programs, and germplasm conservation for bamboo species worldwide.

Introduction

Bamboos are important nontimber forest plants with a wide native geographic distribution in tropical, subtropical and temperate regions except Europe and Antarctica (Bamboo Phylogeny Group, 2012). Bamboos are of notable economic and cultural significance worldwide, and can be broadly used as food, bioenergy, and building materials. Bamboos comprise the Bambusoideae in the grass family (Poaceae), including approximately 1700 described species in 127 genera (Vorontsova et al. 2016; Soreng et al. 2017; Clark and Oliveira 2018). Molecular phylogenetic analysis suggested that Bambusoideae falls into the Bambusoideae–Oryzoideae–Pooideae (BOP) clade, which is phylogenetically sister to Pooideae (Saarela et al. 2018).

Bambusoideae may be divided into two morphologically distinct growth forms: woody bamboos and herbaceous bamboos (tribe Olyreae); woody bamboos can be further divided into two lineages: temperate woody (Arundinarieae) and tropical woody (Bambuseae) (Sungkaew et al. 2009; Bamboo Phylogeny Group 2012; Kelchner and Bamboo Phylogeny Group 2013). The tribe Olyreae comprises 22 genera and 124 described species native to South America except the genus Buergersiochloa and Olyra latifolia (Bamboo Phylogeny Group 2012; Clark et al. 2015). Herbaceous bamboos are characterized by usually weakly developed rhizomes and less lignification in the culms. Culm leaves and foliage leaves with the outer ligule are absent in herbaceous bamboos. In contrast to woody bamboos, herbaceous bamboos have at least functionally unisexual flowers and they flower annually or seasonally for extended periods (Gaut et al. 1997; Kelchner and Bamboo Phylogeny Group 2013; Oliveira et al. 2014; Clark et al. 2015; Wysocki et al. 2015). The tribe is fundamentally diploid, but chromosome counts indicating tetraploidy or hexaploidy or possibly even octoploidy (in Eremitis genus) are available (Soderstrom 1981; Judziewicz et al. 1999). The genus Raddia belongs to the Olyrinae, in which R. distichophylla (Schrad. ex Nees) Chase is a representative species and almost completely restricted to the forests of eastern Brazil (Oliveira et al. 2014). The species is a perennial plant growing in dense tufts. R. distichophylla is a monoecious plant with male and female spikelets in different inflorescences. It is delicate and much smaller than woody bamboos in height (usually 12–35 cm long) (Plants of the World online, http://powo.science.kew.org/) (Figure 1). In the last 25 years, R. distichophylla has seriously been threatened by a rapid deforestation and the conversion of cacao plantations (Giulietti et al. 2005). Besides, R. distichophylla is narrow endemic that has further led to a reduced effective population size.

Figure 1.

The sequenced R. distichophylla plant.

Since the first comparative DNA sequence analysis of bamboos by Kelchner and Clark (1997), a number of studies have been further carried out through combining various biotechnology platforms (Das et al. 2005; Sharma et al. 2008; Sungkaew et al. 2009; Gui et al. 2010; Zhang et al. 2011; Peng et al. 2013; Oliveira et al. 2014; Wysocki et al. 2016). Peng et al. (2013) reported the first draft genome of tetraploid moso bamboo (Phyllostachys edulis, 2n = 4x = 48). Recently, Guo et al. (2019) have released four draft bamboo genomes, including Olyra latifolia (2n = 2x = 22), Raddia. guianensis (2n = 2x = 22), Guadua angustifolia (2n = 4x = 46), and Bonia amplexicaulis (2n = 6x = 72). Whole-genome sequencing of herbaceous bamboo is limited to R. guianensis with a relatively fragmented assembly (contig N50 = 11.45 Kb and scaffold N50 = 12.09 Kb) and a relatively low BUSCO completeness rate (∼72.0%). Thus, the lack of a high-quality genome sequence for the diploid bamboo has been an impediment to our understanding of the bamboo diversification and evolution. It is recognized that genomics allows novel insights into the evolutionary history of species and offers basic information for taking efficient conservation strategies (Silva-Junior et al. 2018). In this study, we generated a draft genome assembly of R. distichophylla through sequencing on an Illumina HiSeq 2000 platform. The availability of the fully sequenced and annotated genome assembly will provide functional, ecological, and evolutionary insights into the bamboo species and greatly enhance conservation genetics of this endangered species.

Materials and methods

Sample collection, total DNA, and RNA extraction and sequencing

The source plant was an individual of R. distichophylla grown in cultivation at the R.W. Pohl Conservatory, Iowa State University. Fresh and healthy leaves were harvested and immediately frozen in liquid nitrogen, followed by storage at −80°C in the laboratory prior to DNA extraction. A modified CTAB method (Porebski et al. 1997) was used to extract high-quality genomic DNA. The quantity and quality of the extracted DNA were examined using a NanoDrop D-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and electrophoresis on a 0.8% agarose gel, respectively. A total of three paired-end and six mate-pair sequencing libraries, spanning 180, 300, 500, 2000, 5000, 10,000, and 20,000 bp, were prepared using Illumina’s paired-end and mate-pair kits, respectively (Illumina, San Diego, CA). Over 5 µg genomic DNA was fragmented by nebulization with compressed nitrogen gas for paired-end libraries. About 10–30 µg of high-quality genomic DNA was required for the long-insert mate-pair libraries. The DNA fragments were circularized by self-ligation, while after the digestion of linear DNA, circularized DNA was again fragmented. The fragmented DNA was purified using streptavidin-coated magnetic beads before adapter ligation. After quality control and concentration estimation of DNA samples with an Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA, USA), libraries were sequenced on Illumina HiSeq 2000 platform.

Total RNA was extracted from four tissues (root, stem, young leaf, and female inflorescence), using a Water Saturated Phenol method. RNA libraries were built using the Illumina RNA-Seq kit (mRNA-Seq Sample Prep Kit P/N 1004814). The extracted RNA was quantified using NanoDrop-1000 UV-VIS spectrophotometer (NanoDrop), and RNA integrity was checked using Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). For each tissue, only total RNAs with a total amount ≥15 μg with a concentration ≥400 ng/μl, RNA integrity number (RIN) ≥7, and rRNA ratio ≥1.4 were used for constructing cDNA library according to the manufacturer’s instruction (Illumina, USA). The libraries were then sequenced (100 nt, paired-end) with Illumina HiSeq 2000 platform.

De novo assembly of the R. distichophylla genome

Two orthogonal methods were used to estimate the genome size of R. distichophylla, including k-mer frequency distribution and flow cytometric analysis. Firstly, we generated the 17-mer occurrence distribution of sequencing reads using GCE v1.0.0 (settings: -m 1 –D 8 –H 1) (Liu et al. 2013), and the genome size was then calculated with the equation $G=\frac{N\times (L-K+1)}{L\times D}$, where G represents the genome size; N is the total number of bases from sequencing data; L is the average length of reads; K is set to 17; and D indicates expected coverage depth for bases. Secondly, the genome size was further estimated and validated using flow cytometry analysis. We employed the rice cultivar Nipponbare as an inner standard with an estimated genome size of ∼389 Mb (IRGSP 2005).

Paired-end sequencing reads were processed to remove adaptor and low-quality sequences using Trimmomatic v0.33 (Bolger et al. 2014). Reads were retained only if both paired reads passed quality control filtering. We assembled the clean reads using Platanus v1.2.1 software (Kajitani et al. 2014), which is optimized for highly heterozygous diploid genomes. First, the high-quality paired-end Illumina reads from short-insert size libraries (≤500 bp) were assembled into contigs using Platanus. The initial K-mer size was set 37, K-mer coverage cutoff was 2, and the step size was 10. The assembled contigs were then scaffolded using SSPACE v3.0 (settings: –k 5 –x 0 –g 3 –a 0.7) (Boetzer et al. 2011) using Illumina mate pair data. GapCloser v1.12 (settings: -a scaffolds.fasta -b reads.lib -o gapcloser_scaffolds.fasta −l 149 −t 4) (Li et al. 2010) was finally used to fill gaps within the scaffolds with the paired-end sequencing data.

Three methods were employed to assess the quality and completeness of the R. distichophylla genome. First, high-quality reads from short-insert size libraries were mapped to the genome assembly using BWA v0.7.15 with default parameters for paired-end reads (Li and Durbin 2009). Second, the genome assembly was checked with Benchmarking Universal Single-Copy Orthologs (BUSCO) (Simao et al. 2015). Third, the RNA sequencing reads generated in this study were assembled using Trinity v r20131110 with default parameters (Grabherr et al. 2011). The assembled transcripts were then aligned back to the assembled genome using GMAP v2014-10-2 (Wu and Watanabe 2005) at 30% coverage and 80% identity thresholds.

Annotation of repetitive sequences and noncoding RNA genes

We used a combination of homology-based and de novo approaches to identify the repetitive sequences in the R. distichophylla genome. RepeatModeler v1.0.10 (Tarailo-Graovac and Chen 2009), which included two de novo repeat finding programs, RECON (Bao and Eddy 2002) and RepeatScout (Price et al. 2005), was used for the construction of the repeat library. This produced library, along with the Poaceae repeat library, was used as the reference database for RepeatMasker (Tarailo-Graovac and Chen 2009). Simple sequence repeats (SSRs) were identified in the genome sequence using the MISA perl script (Thiel et al. 2003) with the default settings: monomer (one nucleotide, n ≥ 12), dimer (two nucleotides, n ≥ 6), trimer (three nucleotides, n ≥ 4), tetramer (four nucleotides, n ≥ 3), pentamer (five nucleotides, n ≥ 3), and hexamer (six nucleotides, n ≥ 3).

Noncoding RNA genes play important roles in many cellular processes. The five different types of noncoding RNA genes, namely transfer RNA (tRNA) genes, ribosomal RNA (rRNA) genes, small nucleolar RNA (snoRNAs) genes, small nuclear RNA (snRNAs) genes, and microRNA (miRNAs) genes, were predicted using various de novo and homology search methods. We used tRNAscan-SE algorithms (version 1.23) (Lowe and Eddy 1997) with default parameters to identify the tRNA genes. The rRNA genes (8S, 18S, and 28S), which is the RNA component of the ribosome, were predicted by using RNAmmer algorithms (v1.2) (Lagesen et al. 2007) with default parameters. The snoRNA genes were annotated using snoScan v1.0 (Lowe and Eddy 1999) with the yeast rRNA methylation sites and yeast rRNA sequences provided by the snoScan distribution. The snRNA genes were identified by INFERNAL software (v1.1.2) (Nawrocki et al. 2009) against the Rfam database (release 9.1) with default parameters.

Genome annotation

The gene prediction pipeline combined the de novo method, the homology-based method and the transcriptome-based method. Augustus v2.5.5 (Stanke et al. 2004) and Fgenesh (Salamov and Solovyev, 2000) were used to perform the de novo prediction. To improve the quality of gene prediction, we performed self-training with Augustus. RNA-seq reads were de novo assembled using Trinity and refined with PASA (Haas et al. 2008) to produce additional genome-guided transcriptome assemblies. Manual curation was performed with the training set. Genes were retained if: (1) they have the complete gene structure without inner stop codons; (2) they have multiple exons and the CDS length exceed 800 bp. CD-Hit (Li and Godzik 2006) was used to remove the training set with over 70% sequence similarity. The protein sequences of moso bamboo (http://server.ncgr.ac.cn/bamboo/) (Peng et al. 2013), stiff brome (GenBank, assembly accession GCA_000005505.4) (The International Brachypodium Initiative, 2010), barley (https://webblast.ipk-gatersleben.de/barley_ibsc/) (Mascher et al. 2017), maize (http://ensembl.gramene.org/Zea_mays/Info/Index) (Jiao et al. 2017), Oropetium thomaeum (GenBank, assembly accession LFJQ00000000) (VanBuren et al. 2015), foxtail millet (GenBank, assembly accession AGTC01000000) (Zhang et al. 2012), rice (http://rice.plantbiology.msu.edu/index.shtml) (IRGSP 2005), and sorghum (GenBank, assembly accession QWKM00000000) (Deschamps et al. 2018) were mapped to the genome using Exonerate (settings: genome2protein option) (Slater and Birney 2005). To further aid the gene annotation, Illumina RNA-seq reads were assembled using the Trinity software (v20131110) with default parameters (Grabherr et al. 2011). The resulting transcripts were then aligned to the soft-masked genome assembly using GMAP v2014-10-2 (Wu and Watanabe 2005) and BLAT v35 (Kent 2002). The potential gene structures were derived using PASA v20130907 (Program to Assemble Spliced Alignments) (Haas et al. 2003). All gene models produced by the de novo, homology-based, and transcriptome-based methods were integrated using GLEAN (Elsik et al. 2007).

The predicted genes were searched against Swiss-Prot database (Boeckmann et al. 2003) using BLASTP (e-value cutoff of 10⁻⁵). The motifs and domains within gene models were identified by InterProScan (Jones et al. 2014). Gene Ontology terms and KEGG pathway for each gene were retrieved from the corresponding InterPro entry. Gene functions were also assigned with TrEMBL database (Boeckmann et al. 2003) using BLASP with an e-value threshold of 10⁻⁵.

Data availability

All sequencing reads have been deposited in the NCBI Sequence Read Archive SRR8759078 to SRR8759084 (2019) (under accession number PRJNA528150) and BIG Genome Sequence Archive CRR049770 to CRR049776 (2019) (under accession number PRJCA001330). The assembled genome sequence is available at the NCBI and BIG Genome Warehouse under accession number SPJY00000000 and GWHAAKD00000000, respectively. Gene prediction and peptide fasta of R. distichophylla may also be accessed through the BIG Genome Warehouse under accession number GWHAAKD00000000. Supplementary Figure S1 presents evolutionary history of TE super-families in the R. distichophylla and moso bamboo genomes. Supplementary Figure S2 shows comparisons of gene features among R. distichophylla and three other plant species. Supplementary Table S1 shows the whole genome sequencing (WGS) reads used to assemble the R. distichophylla genome. Supplementary Table S2 shows the summary of RNA sequencing (RNA-Seq) of R. distichophylla. Supplementary Table S3 shows the summary of genome assembly. Supplementary Table S4 shows the validation of the R. distichophylla genome assembly using reads mapping and transcript alignments. Supplementary Table S5 shows the assessment of the R. distichophylla genome assembly using BUSCO. Supplementary Table S6 shows the statistics of repeat sequences in the R. distichophylla and moso bamboo genomes. Supplementary Table S7 shows the statistics of typical transposable elements (TEs) between R. distichophylla and R. guianensis. Supplementary Table S8 shows the summary of types and number of SSRs in the R. distichophylla and moso bamboo genomes. Supplementary Table S9 shows the noncoding RNA genes in the R. distichophylla genome. Supplementary Table S10 shows the statistics of predicted protein-coding genes in the R. distichophylla genome. Supplementary Table S11 shows the functional annotation of the R. distichophylla protein-coding genes.

Supplementary material is available at figshare DOI: https://doi.org/10.25387/g3.13186907.

Results

We performed whole-genome sequencing with the Illumina sequencing platform. A total of 253.94 Gb short sequencing reads were generated (∼272.89-fold coverage) (Supplementary Table S1). A total of 21.99 Gb RNA-seq data were obtained from root, stem, young leaf, and female inflorescence (Supplementary Table S2). Based on the K-mer analysis, we estimated the genome size of R. distichophylla to be ∼608 Mb (Figure 2). Flow cytometry analysis estimated the genome size of R. distichophylla to be ∼589 Mb, which is close to the obtained result from the k-mer analysis. The final assembly amounted to ∼580.85 Mb, representing for 95.56% of the estimated genome size. The N50 lengths of the assembled contigs and scaffolds were ∼86.36 kb and ∼1.81 Mb, respectively (Table 1; Supplementary Table S3). The contig N50 and scaffold N50 sizes represent ∼7.14-fold and ∼150.83-fold improvement compared with the previously reported R. guianensis genome assembly (Guo et al. 2019), respectively.

Figure 2.

The 17-mer distribution of sequencing reads from R. distichophylla. The occurrence of 17-mer was calculated using GCE based on the sequencing data from short insert size libraries (insert size ≤500 bp) of R. distichophylla. The sharp peak on the left with low depths represents the essentially random sequencing errors. The middle and right peaks indicate the heterozygous and homozygous peaks, the depths of which are 46 and 94, respectively.

Table 1.

Summary of the genome assemblies and annotations of R. distichophylla and R. guianensis

	R. distichophylla	R. guianensis
Assembly
Estimated genome size (Mb)	608	685
Assembled sequence length (Mb)	581	626
Scaffold Number	38,269	12,824
Scaffold N50 (Mb)	1.81	0.012
Contig Number	45,206	13,300
Contig N50 (Kb)	86.36	12.09
Annotation
Number of predicted protein-coding genes	30,763	24,275
Average gene length (bp)	2,886.93	2,635.83
tRNAs	727	923
rRNAs	90	743
snoRNAs	242	NA
snRNAs	127	358
miRNAs	256	387
Transposable elements (Mb)	285.10	339.26
Transposable elements (%)	49.08	54.15

To assess the genome assembly quality, we first mapped ∼211 Mb of high-quality reads to the genome sequences. Our results revealed that nearly 89.25% Illumina reads were mapped to the genome assembly (Supplementary Table S4); second, BUSCO was used to assess the completeness of the genome assembly. The percentage of completeness for our assembly was 92.08% in the Embryophyta lineage (Supplementary Table S5); and finally, we mapped the assembled transcripts to the genome sequences. Approximately 78.04% of the transcripts could be mapped to the genome (Supplementary Table S4).

The annotation of repeat sequences showed that approximately 49.08% of the R. distichophylla genome consists of TEs, lower than the amount (63.15%) annotated in the moso bamboo genome (Peng et al. 2013) with the same methods (Supplementary Table S6; Supplementary Figure S1). The R. distichophylla showed a similar repeat content (∼285.10 Mb; ∼49.08%) compared with R. guianensis in the same genus (∼339.26 Mb; ∼54.15%) (Table 1) (Guo et al. 2019). LTR retrotransposons were the most abundant type of TEs, occupying roughly 35.99% of the R. distichophylla genome. Specifically, the R. guianensis genome showed significantly expansion of Ty3/Gypsy retrotransposons compared with the R. distichophylla genome (Supplementary Table S7). In total, 220,737 and 496,819 SSRs were found in the R. distichophylla and moso bamboo genomes, respectively, with trimer and tetramer as the most abundant SSR types (Supplementary Table S8). Among the trimer motifs, (CCG/GGC)n were the predominant repeat in R. distichophylla, whereas (CCG/GGC)n, (AGG/CCT)n, and (AAG/CCT)n showed a similar proportions in the moso bamboo genome. The identified SSRs will provide valuable molecular resources for germplasm characterization and genomics-based breeding programs. In total, we identified 727 tRNA genes, 90 rRNA genes, 242 snoRNA genes, 127 snRNA genes, and 256 miRNA genes, respectively (Table 1; Supplementary Table S9).

In combination with ab initio prediction, protein and transcript alignments, we obtained a gene set consisted of 30,763 protein-coding genes (Table 1; Supplementary Table S10), with an average length of 2,887 bp and an average coding sequence length of 1,099 bp (Supplementary Table S10; Supplementary Figure S2). Among these genes, 88.85% had significant similarities to sequences in the public databases (Supplementary Table S11).

Discussion

In this study, we present a draft genome assembly of the herbaceous bamboo R. distichophylla to supplement the currently existing bamboo genomic resources. The assembled genome was ∼580.85 Mb in size, with a contig N50 length of ∼86.36 Kb, ∼7.14 times longer than the previously reported genome assembly of R. guianensis (Guo et al. 2019). The genome assembly comprised 38,269 scaffolds with a N50 length of ∼1.80 Mb, which is far more contiguous than that of R. guianensis (Guo et al. 2019). Validation of genome assembly using reads mapping, transcripts alignments, and BUSCO assessment together showed that the assembled R. distichophylla draft genome is accurate and complete. Polyploidy has been proved to be an important evolutionary force for the speciation as well as trait specialization in flowering plants (Wolfe 2001; Blanc and Wolfe 2004; Jiao et al. 2011; Jiang et al. 2013), which is commonly present in bamboos. Thus, the availability of a much more contiguous diploid genome will greatly facilitate the reconstruction of the evolutionary history of polyploidy in different bamboo clades.

SSRs were commonly utilized to develop molecular markers (Kumari et al. 2013; Pandey et al. 2013; ,Sugita et al., 2013 ), which have been extensively applied to exploring the molecular phylogeny and taxonomy of bamboo species (e.g., Zhao et al. 2015). In this study, we identified the whole genome-based SSR loci. Our results suggested that about ∼59.2% of SSRs were the tri- and tetra-nucleotide repeats. Of these, trinucleotide repeats were the predominant class of repeat type. It is well known that the taxonomy of bamboos has long puzzled the researcher community because of bamboos’ reproductive characteristics. The SSRs identified in this study would particularly benefit population genetics and phylogenetics studies on bamboos toward efficient conservation of the bamboo germplasms.

Funding

This work was supported by Yunnan Innovation Team Project and the start-up grant from South China Agricultural University (to L.-Z.G.), Guangdong Special Support Program (to Q.-J.Z.), and the Presidential Foundation of Guangdong Academy of Agricultural Sciences (201611) (to Q.-J.Z.).

Conflicts of interest: None declared.