Background & Summary

With approximately 15,000 described species, Trichoptera (caddisfly) represents the second most diverse monophyletic group of aquatic insects1. The most species diversity for Trichoptera is found in the Indomalayan and Neotropical regions, with 47–77% of widespread genera recorded. Five families comprise 55% of the global Trichoptera species, while 19 families each contain fewer than 30 species2. Trichoptera larvae (or caddisfly larvae) build their protective cases using secreted silk combined with selected foreign materials, such as sand grains, mollusk shells, or plant fragments3,4,5. This case-building behavior has enabled their ecological diversification, allowing them to inhabit environments that are otherwise inaccessible to many other species3. Among caddisflies, Stenopsyche angustata stands out due to its large size and preference for fast-flowing water environments. Its larvae are typically dark brown, with long, narrow heads and short antennae6. S. angustata produces adhesive silk, which is used to construct intricate underwater composite structures7.

Advancements in sequencing technology have significantly enhanced our understanding of the genome, leading to the decoding of Trichoptera species, such as Himalopsyche anomala and Eubasilissa splendida8, as well as Cheumatopsyche charites9. However, the genome of S. angustata has not yet been sequenced. High-quality reference genomes are crucial for advancing genetic and evolutionary research on this species. In this study, we employed PacBio long-read sequencing and Hi-C techniques to achieve chromosome-level genome assembly for S. angustata. The final assembly totaled 510.47 Mb, with a scaffold N50 of 39.81 Mb. Hi-C scaffolding anchored 99.63% of the initial sequences to 13 pseudochromosomes. Repetitive elements accounted for 44.29% (226.08 Mb) of the genome, and a total of 10,699 protein-coding genes were identified. This high-quality genome facilitates our understanding of the adaptive evolution in Trichoptera.

Methods

Sample collection and sequencing

S. angustata larvae were collected form the Beilun River (21.80 N, 107.89E), Guangxi Zhuang Autonomous Region, China on November 15, 2022. Nine live larval individuals were immediately frozen in liquid nitrogen and stored at −80 °C until DNA extraction. Due to their small size, DNA was extracted from the whole bodies of four of the collected individuals, and prepared for both second-generation and third-generation sequencing using the classic phenol–chloroform method. The quality and quantity of the extracted DNA were assessed using an Agilent 2100 bioanalyzer (Agilent Technologies, Santa Clara, CA, USA), and integrity was evaluated using agarose gel electrophoresis with ethidium bromide staining. Second-generation sequencing was performed on Illumina platform conducted by Novogene Bioinformatics Technology (Novogene, Beijing, China) (Table 1).

Table 1 Genomic sequencing data.
Full size table

For PacBio sequencing, high-quality DNA samples were randomly fragmented into smaller pieces using a Covaris ultrasonic disruptor (Covaris, Woburn, MA, USA). Large DNA fragments were enriched and purified with magnetic beads, after which they underwent damage repair and end repair. Adapters were then ligated to both ends of the DNA fragments, forming stem-loop structures. Unligated fragments were removed via exonuclease treatment. The constructed libraries were then sequenced using the PacBio Sequel system (PacBio, Menlo Park, CA, USA). Consensus sequences were generated by aligning subreads obtained from a single Zero-Mode Waveguide (ZMW), omitting the need for a reference genome. Circular Consensus Sequence (CCS) reads were obtained using the CCS algorithm, requiring at least two full-pass subreads from the insert. The raw sequencing data yielded 97.12 Gb (approximately 180.9 × coverage) with a scaffold N50 of 33.78 kb (Table 1).

Using a modified standard protocol as described previously10, we constructed Hi-C libraries using the whole bodies of the remaining five S. angustata larvae individuals. Larval individuals were ground in liquid nitrogen and cross-linked with a 4% formaldehyde solution at room temperature under vacuum for 30 min. The crosslinking reaction was quenched by adding 2.5 M glycine and incubating for 5 min, followed by placing the sample on ice for 15 min. The samples were then centrifuged at 2500 rpm at 4 °C for 10 min, and the pellet was washed with 500 μl PBS and centrifuged again at 2500 rpm for 5 min. The pellet was resuspended in 20 μl of lysis buffer (1 M Tris-HCl, pH 8, 1 M NaCl, 10% CA-630, and 13 units of protease inhibitor) and centrifuged at 5000 rpm at room temperature for 10 min. The pellet was washed twice with 100 μl ice-cold 1x NEB buffer and centrifuged at 5000 rpm for 5 min. The nuclei were resuspended in 100 μl of NEB buffer, solubilized with dilute sodium dodecyl sulfate (SDS), and incubated at 65 °C for 10 min. SDS was neutralized with Triton X-100, and the solution was incubated overnight with the 4-cutter restriction enzyme MboI (400 units) at 37 °C on a rocking platform to digest the DNA into smaller fragments. The DNA ends were then labeled with biotin-14-dCTP, followed by blunt-end ligation of the cross-linked fragments. Proximal chromatin DNA was re-ligated using a ligation enzyme, and the nuclear complexes were reverse cross-linked by incubation with proteinase K at 65 °C. DNA was then purified through phenol-chloroform extraction, and biotin was removed from non-ligated fragment ends with T4 DNA polymerase. The ends of sonicated fragments (200–600 bp) were repaired with a mixture of T4 DNA polymerase, T4 polynucleotide kinase, and Klenow DNA polymerase. Biotin-labeled Hi-C samples were enriched with streptavidin C1 magnetic beads. After adding A-tails to the fragment ends and ligating Illumina paired-end (PE) sequencing adapters, Hi-C sequencing libraries were amplified by polymerase chain reaction (PCR) (12–14 cycles) and sequenced on an Illumina PE150 platform by Novogene Bioinformatics Technology (Novogene, Beijing, China) (Table 1).

Genome size estimation and assembly

The genome size, heterozygosity rate, and repeat content of S. angustata were estimated through k-mer analysis using Jellyfish (v2.3.0)11. A total of 17, 977, 255, 521 17-mers with a depth peak of 33 were analyzed. Using the formula: genome size = K-num/K-depth, the genome size of S. angustata was estimated to be 544.77 Mb. After removing contaminated and erroneous sequences, the revised genome size was determined to be 536.86 Mb. Meanwhile, the estimated heterozygous ratio and repeat content were approximately 0.97% and 43.07%, respectively (Table 2).

Table 2 K-mer based genome survey of S. angustata.
Full size table

PacBio subreads were used for de novo genome assembly using the wtdbg2 software12. Initially, DNA sequences were randomly sheared into 1,024 bp fragments for clone sequencing. Reads were then used to construct a vertex sequence based on their similar relationships. Sequencing reads were then analyzed to identify overlaps, and vertex sequences were constructed based on their similarity relationships. These vertex sequences were subsequently concatenated based on their position on the reads. Contigs were obtained by segmenting sequences at repeat region boundaries to improve assembly accuracy. Scaffold sequences were constructed using the paired-end relationships of large-fragment sequencing data. The chromosome number and ploidy were determined through our previous karyotype analysis. Chromosome-level genome assembly was achieved using the ALLHiC pipeline13, which consists of five steps: pruning, partition, rescue, optimization, and building. In the pruning step, crosslinking signals between homologous chromosomes were trimmed to separate alleles and homologous sequences into their respective haplotypes, enabling independent assembly and minimizing errors. During partition, Hi-C interaction signals between contigs were analyzed to cluster them into groups, effectively distinguishing homologous chromosomes. The rescue step addressed assembly inconsistencies by identifying collapsed regions, which are highly similar sequences that were not properly separated, and by detecting the strongest interaction signals between non-collapsed regions. Additionally, contigs that remained unassigned during pruning were reassessed and correctly placed. In the optimization step, genetic algorithms were used to iteratively refine the sorting and orientation of contigs within each chromosome group. Finally, in the building step, a chromosome-level assembly was generated and visualized using a contact map to assess Hi-C interaction patterns, confirming the accuracy of the final genome assembly.

The assembly had a total contig length of 510,457,233 bp and a contig N50 length of 13,847,508 bp. The total scaffold length was 510,471,733 bp, and a scaffold N50 length of 39,811,520 bp (Table 3). The genome anchoring rate was 99.63% (Table 4). Based on the Hi-C contact map (Fig. 1), both the genome metrics and anchoring rate were excellent.

Table 3 Chromosome-level genome assembly statistics for S. angustata.
Full size table
Table 4 Anchoring rate information for chromosome-level genome assembly.
Full size table
Fig. 1
figure 1

Hi-C contact map of the chromosome-level assembly of Stenopsyche angustata. The x- and y-axes represent the assembled chromosomes, arranged in numerical order. Red color indicates higher contact frequencies, while yellow color reflects lower frequencies. The strong diagonal pattern demonstrates predominant intra-chromosomal interactions, confirming accurate chromosome anchoring and assembly. The color scale bar shows the contact intensities (from 1 to 3).

Full size image

Repeat annotation

A comprehensive repeat annotation pipeline was applied, integrating homology-based alignment and de novo search strategies to identify genome-wide repeats. Tandem repeats were detected using Tandem Repeats Finder (TRF)14 through ab initio prediction, identifying approximately 3.97 Mb of sequences. For homology-based repeat identification, the Repbase database15 was employed in conjunction with RepeatMasker16 and its in-house script RepeatProteinMask with default parameters. This approach identified 222.37 Mb and 27.76 Mb of repeat regions, respectively (Table 5). The ab initio prediction generated a de novo repetitive element database using LTR_FINDER17, RepeatScout18, and RepeatModeler19 with default parameters (Table 6). Repeat sequences longer than 100 bp and with less than 5% ambiguous nucleotides (‘N’) were retained to construct the raw transposable element (TE) library. A customized, non-redundant library was then constructed by merging the de novo TE library with existing Repbase data, removing duplicate sequences using the UCLUST algorithm20. The obtained library was subsequently used for DNA-level repeat identification with RepeatMasker.

Table 5 Statistics of repeat sequences.
Full size table
Table 6 Transposable element (TE) distribution in the genome.
Full size table

Denovo + Repbase transposable elements (TEs) were predicted using de novo tools (RepeatModeler, RepeatScout, Piler21, and LTR_FINDER) and combined with the RepBase nucleic acid data. The results were integrated using UCLUST following the 80-80-80 rule to ensure high-confidence matches, and were finally annotated using RepeatMasker. TE proteins were identified by annotating the genome with the RepBase protein data using the RepeatProteinMask software. Combined TEs represent the results obtained by integrating the two aforementioned methods and removing redundancy. The ‘Unknown’ category includes repeat sequences that could not be classified by RepeatMasker. The ‘Total’ category represents the non-redundant result obtained after removing overlaps between the different classifications. DNA: DNA transposons; LINE: long interspersed nuclear elements; SINE: short interspersed nuclear elements; LTR: long terminal repeat.

The genome of S. angustata exhibited notable variations across chromosomes. GC content varied distinctly across the chromosomes, with a relatively higher level on chromosomes 12, potentially associated with gene enrichment and transcriptional activity in this region. Moreover, the distribution of interspersed repeats also showed clear chromosome-specific patterns, particularly on chromosomes 8 and 12, where an increased density of these sequences may suggest frequent replication or insertion events. In contrast, long repeats were predominantly concentrated on chromosomes 12 and 13, indicating the potential importance of these regions in maintaining genome structure and regulating chromosomal conformation. Additionally, we observed higher gene density on chromosome 13 (Fig. 2), which may contain a large number of functional genes or active transcription units. These findings provide valuable insights into the structural and functional characteristics of the S. angustata genome, offering a foundation for future research on functional gene characterization.

Fig. 2
figure 2

Circos plot illustrating the genomic features of S. angustata. From the innermost to the outermost ring, the plot shows GC content, interspersed repeats, long repeats and gene density, highlighting variations in these figures across chromosomes. Chromosomes are labeled around the outermost ring.

Full size image

Gene structure and functional annotation

Homologous protein sequences were obtained from Ensembl22 and NCBI23. These sequences were aligned to the genome using TblastN (v2.2.26) with E-value ≤ 1e−5. Subsequently, GeneWise (v2.4.1) was used to align the matching proteins to the corresponding genome sequences, ensuring accurate spliced alignments and gene structure prediction of the identified protein regions. For ab initio gene prediction, an automated pipeline was employed, incorporating Augustus (v3.2.3)24, Geneid (v1.4)25, Genescan (v1.0)26, GlimmerHMM (v3.04)27, and SNAP (2013-11-29)28. The genome annotation was further refined using transcriptome read assemblies generated by Trinity (v2.1.1)29.

Gene functions were assigned by aligning the predicted protein sequences to the Swiss-Prot database30 using Blastp (E-value ≤ 1e−5). Motifs and domains were annotated using InterProScan70 (v5.31) against publicly available databases, including Swiss-Prot, Nr31, Interpro32, Pfam33, etc. Gene Ontology (GO) terms were assigned based on the corresponding InterPro entries. Protein functions were predicted by transferring annotations from the closest BLAST hit (E-value < 1e−5) in the Swiss-Prot database and DIAMOND (v0.8.22) or BLAST hits (E-value < 1e−5) in the NR20 database. Additionally, the gene set was mapped to Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways to identify the best match for each gene. The protein sequences derived from gene structure prediction were aligned to known protein databases, allowing functional prediction for 10,699 encoding genes, which represent 97.7% of the total genes (Table 7).

Table 7 Gene Functional Annotation Statistics.
Full size table

Non-coding RNA annotation

Non-coding RNA (ncRNA) annotation was performed to identify tRNAs, rRNAs, miRNAs, and snRNAs. tRNA genes were predicted using tRNAscan-SE34. Due to the high conservation of rRNA sequences, rRNAs were identified by aligning reference sequences from related species to the genome using BLAST. Other ncRNAs, including miRNAs and snRNAs, were detected by searching against the Rfam database35 using Infernal36 (http://infernal.janelia.org/) with default parameters (Table 8).

Table 8 Non-coding RNA Statistics.
Full size table

Data Records

This Whole Genome Shotgun project has been deposited at GenBank under the accession JBPJGE00000000037. Besides, the genome and raw sequencing data are publicly accessible in China National Gene Bank (https://db.cngb.org/) with the accession number CNP000649038. The genome assembly data and annotations have also been deposited at Figshare39. The PacBio reads are available in the NCBI SRA database under accession number SRR3208962140, while the Hi-C reads can be accessed under SRR3208962041.

Technical Validation

The integrity and accuracy of S. angustata genome assembly were evaluated through multiple approaches. First, the Hi-C contact map revealed strong intra-chromosomal interaction signals along the diagonal (Fig. 1), confirming the integrity of the genome structure. Second, the distribution of GC content demonstrated that there was no significant contamination in the assembly sequence (Fig. 2). To further assess genomic integrity, a BUSCO42 analysis was performed, showing that 98.8% of the complete single copy genes were assembled from a set of 1,013 single-copy orthologous genes (C: 98.8% [S: 98.0%, D: 0.8%], F: 0.4%, M: 0.8%, n: 1,013). At the same time, CEGMA43 was used to evaluate the completeness of the S. angustata genome. The results showed that 230 of the 248 full-length genes in the core gene set were included, achieving a 92.74% coverage. For accuracy assessment, small fragment library reads were mapped to the assembled genome using the BWA software44. The mapping rate and genome coverage rate were found to be 99.02% and 99.53%, respectively. Additionally, 10,456 (97.7%) gene models were successfully annotated in databases such as NR, KEGG, GO, Pfam and Interpro. Taken together, these results provide strong evidence that the obtained de novo S. angustata genome is of high quality.