Background & Summary

The genus Bougainvillea (Caryophyllales: Nyctaginaceae) is indigenous to the South American tropics and subtropics. It is an immensely showy, colourful and floriferous plant, virtually pest/disease-free and exhibits strong adaptability to multiple stresses of drought, salinity, UV radiation and industrially contaminated soil.

Due to the nature of its ornamental value, as well as the minimum requirement of water and care necessary for maintenance, bougainvillea is one of the most prominent and popular perennial ornamental crops in floriculture for multitudinous use. The genus Bougainvillea has approximately 11 to18 species based on historical specimens and botanical publications, of which three (B. glabra Choisy, B. spectabilis Willdenow, and B. peruviana Humboldt & Bonpland) are of horticultural importance with very showy and colourful bracts. Almost all existing hundreds of varieties were derived from these species and their hybrid groups viz. B. × buttiana (B. × glabra-peruviana), B. × specto-peruviana, and B. × specto-glabra (2n = 2× = 34)1,2,3). Among which, B. glabra is the most common and prominent species, due to the diversity in bract colour and shape, continuous flowering, strong tolerance to biotic and abiotic stresses, and high cross-compatibility in breeding (Fig. 1a).

Fig. 1
figure 1

The overview features of the B. glabra Choisy. (a) Morphological characters of floral organs in B. glabra plants. The blooming shoots are densely arranged with inflorescences arise as terminal or axillary cymose. The inflorescences show asynchronous flowering, with some differences in timing among individuals. The flower is salver-shaped, with tubular perianths constricted in the middle, its pedicel is attached to the central midrib of bract. Axillary thorns can become hard through lignification during vegetative growth stage, or differentiate into inflorescences during reproductive growth stage. (b) The genome survey results based on the kmer size of 51. (c) and (d) exhibited the Hi-C heatmap of Hap1 and Hap2 respectively. The darker colours indicate that contigs in the region have more interactions with each other. (e and f) exhibited the kmer spectrum analysis performed by KAT, the black region represents the proportion of kmers present in the HiFi reads but missing in the assemblies. Other colours indicate how many times the specific kmer occurred in the assembled genome.

Full size image

The most remarkable decorative feature of bougainvillea is its bract colour, which is primarily produced by betalains and are virtually only found in species of the core Caryophyllales4,5,6. Most recently, researchers published the high-quality genome and annotation results for B. × buttiana ‘Ms Butt’7, and also studied the correlation between the bract colour and gene expression levels in the betalains biosynthesis pathway. But the lower sequence coverage (~13×) still caused a large number of gaps, and we need more genomes to analyse the great genomic diversity in the genus Bougainvillea. Moreover, as an interspecific hybrid, ‘Ms Butt’ has a doubled genome (2n = 68, >5 Gbp), as well as hybrid sterility, all of which limit its application in molecular evolution and genetics research. Therefore, constructing a high-quality genome for Bougainvillea species is essential for offering key insights into the habitat adaptability and horticultural traits of this significant lineage.

In this study, we provided the first haplotype-aware chromosomal-level assemblies for the B. glabra by using the state-of-art sequencing technology from Revio platform combined with Hi-C sequencing data. The genome size was 2,889,511,499 and 2,855,692,602 bp for Hap1 and Hap2 respectively, with N50 of 167,905,837 and 165,910,972 bp. Through the BUSCO evaluation results, the assembly and annotation results exhibited the completeness value of 96% and 90.8% for Hap1, while 96% and 90.7% for Hap2. These high-quality assemblies will provide valuable resource for future functional analysis, evolution and genetic improvement in bougainvillea.

Methods

Library construction and genome sequencing

All plant materials used to sequence in this study were collected at the bougainvillea nursery of the Chinese Academy of Tropical Agricultural Sciences in Danzhou, China (109°42’E/19°35’N). Total genomic DNA was extracted by using the CTAB8 method, then the Revio library was constructed based on the high molecular weight DNA from the standard preparation protocol, further sequenced by the Pacbio Revio system and got ~92 G HiFi data. The Hi-C library was constructed based on the 2 g plant leaves which were cut into 1- to 2-mm strips by following the previous research9, and sequenced by the Illumina HiSeq X Ten platform (San Diego, CA, United States) with 150PE mode. By filtering the low-quality sequencing data, a total data size of ~248 G Hi-C data was used for genome assembly and scaffolding. Transcriptome sequencing of mixed tissues from B. glabra ‘Formosa’ including fresh flower, fresh leaves, mature leaves, root and stem was conducted to improve gene prediction and annotation, through the quality control performed by the fastp10 with default parameters, we got a total RNA data size of 10.82 G for genome annotation.

Genome survey

All HiFi reads generated by the Revio were calculated the kmer frequency by the Jellyfish v2.3.0 (count -m 51; and histo with the default parameters). Then the findGSE (https://github.com/schneebergerlab/findGSE) was used for estimating genome size. For its results, the haplotype’s genome size is nearly 2896.646 Mb, and the heterozygous is 0.01493859 (Fig. 1b).

Genome assembly

These high-quality HiFi and Hi-C reads were supplied to the hifiasm v0.19.8 (-s 0.1–f-perturb 0.99–n-perturb 100000) to assemble the two haplotypes at the same time. After the contig of each haplotype was generated, HiC reads were then mapped to the contig assembly results generated by the hifiasm by bwa (version 0.7.17-r1188) with mem -5SP, further removed the PCR duplicates by the samblaster v0.1.2611 with default parameters, samtools v1.6 was used to filter out secondary and supplementary alignments. Finally, the HapHiC (version 1.0.2)12 was used for scaffolding the contig into the chromosomes, and manually corrected by the JUICER13, the gaps were filled with length of 500 bp of Ns by defaults.

According to the assembly pipeline, we get the genome size of 2,889,271,299 and 2,855,518,402 bp, with N50 size of 3,347,701 and 3,461,675 bp for Hap1 and Hap2 respectively. After scaffolding these contigs into 34 chromosomes (for each haplotype is 17 chromosomes) (Figs. 1c,d, 2a), the genome size turned into the 2,889,511,499 and 2,855,692,602 bp, and the N50 increased into 167,905,837 and 165,910,972 bp for Hap1 and Hap2, respectively.

Fig. 2
figure 2

The assembly and assessment results of two different haplotypes in B. glabra Choisy. (a) The circos plot for Hap1 and Hap2 assembled results, the same chromosome numbers indicated that the homologous chromosome. From the outside to the inside, each circos represents the chromosome, GC ratio, the gene density and the repeat content density. (b) The BUSCO evaluation results of the assembly and annotation results. (c) The ratio of different variation type in the total genome size between the Hap1 and Hap2.

Full size image

To further evaluate our assembly quality, the BUSCO (Benchmarking Universal Single-Copy Orthologs) v 5.4.714 were used based on the eudicots_odb10. Both haplotypes exhibited a high score of completeness with 96% (Fig. 2a). Also we performed the kmer spectrum analysis by KAT15, exhibited the haplotype-resolved assemblies results (Fig. 1e and f). The LTR Assembly Index (LAI)16 was also used to evaluate the assembly quality, the value of 12.8 and 12.9 for Hap1 and Hap2 respectively based on the EDTA17 annotation results.

Genome annotation

For protein-coding gene prediction, we combined three different evidence from the homology, de novo and transcriptome prediction to get the non-redundant high-confidence prediction results. Homologue proteins from the protein belongs to plant of Swiss-Prot18, B. × buttiana ‘Mrs. Butt’7, Beta vulgaris19, Spinacia oleracea20 and TAIR1021. These proteins were aligned to the repeat-masked genome by miniprot22 with default parameters.

For de novo gene prediction, first, we masked the repeat region by the RepeatMasker v 4.1.2-p1 and RepeatModeler v2.0.3. Then, the BRAKER v 2.1.623 pipeline which combined the GeneMark-ET v 4.68_lic24 and AUGUSTUS v 3.4.025 was used to perform the de novo annotation by default parameters. We used repeat-masked genomic data and quality-controlled RNA-Seq data to automatically generate full gene structure in de novo approach. The RNA-Seq reads were mapped to the repeat-masked genome using Hisat226 by default parameter.

The Trinity v2.8.527 was used to assemble the quality-controlled RNA-seq reads, and further cleaned by seqclean28, then mapped to the repeat-masked genome by minimap229, and the gene structure was predicted by PASA v2.5.030 as EST evidence.

In the end, the homology evidence, de novo prediction, and EST evidence were combined by EvidenceModeler31 to generate a non-redundant gene set, which containing 33,613 and 33,067 genes for Hap1 and Hap2 respectively. The BUSCO evaluation based on the eudicots_odb10 exhibited that the completeness score of 90.8% and 90.6% for Hap1 and Hap2 respectively (Fig. 2b). Moreover, we find that the ratio of monoexonic gene numbers to multiexonic gene numbers in Hap1 is 0.27, and 0.26 in Hap2, indicating reasonable and ideal prediction results32,33. The average lengths of all predicted genes were 6657.45 bp and 6710.67 bp, each gene had an average of 4.52 and 4.55 exons for Hap1 and Hap2 respectively.

The predicted genes’ functions were assigned according to the best match by aligning the protein sequences to the Swiss-Prot and NCBI nonredundant (NR)34 database using Blastp (Evalue = 1e−5)35. The motif and domains were annotated by Interproscan (v 5.52–86.0)36 and PfamScan (v 3.3.2)37. The Gene Ontology (GO) IDs for each gene were assigned according to the corresponding InterPro entry. In total, there were 29,208 and 28,473 genes could be annotated by the different databases for Hap1 and Hap2 respectively (Table 1).

Table 1 Genome annotation results and the genes could be functionally annotated by different databses.
Full size table

The EDTA v 2.1.017 was used to fully annotate the transposable elements (TEs) after completing genome structure prediction by using the parameters of “--sensitive 1 --species others --anno 1 --evaluate 1”. Through the EDTA pipeline, approximately 80% of the genome was occupied by the repetitive elements (Table 2). Among these, we found that the LTRs account for the most abundant repeat content, at ~60% of the total genome.

Table 2 Repeat contents annotations results.
Full size table

We evaluated the divergence between the Hap1 and Hap2 by syri (with parameter “-f -F S”), set Hap1 as reference and compared with Hap2 by minimap2 with “-ax asm10 –eqx”. The total ratio of syntenic region was 47.96% and 49.85% for Hap1 and Hap2 respectively (Fig. 2c). We also identified many inversions with 23.3% and 23.28% account for total genome size of Hap1 and Hap2 respectively. Exhibited the great divergence between the haplotypes.

Data Records

The raw sequencing data (HiFi, Hi-C and transcriptome) generated in this study have been deposited in NCBI (https://www.ncbi.nlm.nih.gov) under the Bioproject accession number PRJNA1146587 (SRP525956)38, and the genome assembles could be found in and JBIMPK000000000(PRJNA1172890)39 and JBIMPL000000000 (PRJNA1172889)40. The two haplotypes’ assemblies were deposited in National Genomics Data Center (NGDC) (https://ngdc.cncb.ac.cn/) under the Bioproject accession number of PRJCA02920341, with accession number of GWHFCQT00000000.142 and GWHFCQU00000000.143. The genome assembles and annotation pipeline used in this study, the assembly results and annotations result of both Hap1 and Hap2 were also deposit in Figshare44.

Technical Validation

The quality of the B. glabra assembly was assessed in three different approaches. First, the BUSCO evaluation results exhibited that the high completeness score of 96% for both Hap1 and Hap2 for their assembly results. Second, we performed the kmer spectrum analysis exhibited the haplotype-resolved results. Third, through the Hi-C interaction heatmap, there were no obvious sequence or contig direction errors in both Hap1 and Hap2 assemblies, also had features suggestive of intrachromosomal interaction enrichment, distance-dependent interaction decay and local interaction smoothness.