A Chromosome-Level Genome Assembly of Sitotroga cerealella (Olivier, 1789) (Lepidoptera: Gelechiidae), a Global Pest of Stored Grains

Abstract

Sitotroga cerealella, commonly known as the Angoumois grain moth, is a highly destructive, cosmopolitan pest affecting stored cereal grains worldwide. Despite its economic importance, high-quality genomic resources for this species have been lacking, hindering advanced molecular studies aimed at developing effective control strategies. Here, we present a highly contiguous, chromosome-level reference genome assembly for male S. cerealella using a multi-platform sequencing approach combining Oxford Nanopore Technologies (ONT) ultra-long reads, Illumina short reads, and Hi-C chromatin interaction data. The final assembly, which lacks the female-specific W chromosome, achieved a total size of 325.99 Mb and was anchored into 30 pseudochromosomes, representing an anchoring rate of 99.89%. The assembly continuity is exceptionally high, with a Scaffold N50 of 12.046 Mb, approaching telomere-to-telomere (T2T) completion for many chromosomes. Comprehensive structural and functional annotation, supported by multi-stage RNA-seq data, identified 13,840 protein-coding genes and revealed that 40.81% of the genome consists of repetitive elements. Benchmarking Universal Single-Copy Orthologs (BUSCO) analysis confirmed the high quality and completeness, achieving a score of 99.7% (Insecta). This high-quality genomic resource provides a critical foundation for studying the molecular mechanisms of insecticide resistance, detoxification metabolism, and sustainable pest management of this destructive stored-grain pest.

Background & Summary

The Angoumois grain moth, Sitotroga cerealella (Lepidoptera: Gelechiidae), is one of the most destructive primary pests of stored cereal grains, including wheat, rice, corn, and barley, worldwide¹. Its internal feeding behavior throughout the larval stage leads to significant weight loss, nutritional degradation, and economic damage to stored commodities, particularly in warm, humid climates. The effective management of this pest is crucial for minimizing post-harvest losses and ensuring food security². However, the development of sustainable control strategies, such as targeted gene regulation or new pesticide development, is significantly impeded by the limited availability of high-quality genomic information.

Genome sequencing has proven to be an indispensable tool for understanding insect biology, identifying pest-specific vulnerabilities, and informing molecular pest management strategies³. While several Lepidopteran genomes have been sequenced, the Gelechiidae family, to which S. cerealella belongs, remains underrepresented, particularly with chromosome-level assemblies. The genomic architecture of S. cerealella is of particular interest due to its obligate association with stored products and its documented ability to develop resistance to various insecticides^1,4. A high-quality reference genome is essential for comparative genomics studies, identifying genes involved in detoxification and chemoreception, and supporting functional genomic research.

In this study, we leveraged a multi-platform approach, integrating Oxford Nanopore Technologies (ONT) long reads, Illumina short reads, and Hi-C proximity ligation data to assemble a highly contiguous, chromosome-level reference genome for S. cerealella. Subsequently, we performed comprehensive annotation of repeats, non-coding RNAs, and protein-coding genes, thereby providing a critical foundation for subsequent molecular and genomic research on the Angoumois grain moth.

Methods

Sample collection, preparation, and sequencing

Samples of S. cerealella were collected from a grain warehouse in Mengcheng, Anhui, China in 2025 and subsequently reared in the laboratory on whole barley grains. The rearing conditions were controlled at a relative humidity of 65 ± 5% relative humidity and a photoperiod of 14:10 h (light: dark).

Genomic DNA (gDNA) and RNA were extracted to construct distinct sequencing libraries. For the genome survey, gDNA was extracted from a single male pupa using MagPure Tissue&Blood DNA LQ Kit. A short-insert library (200–400 bp) was constructed utilizing the Agencourt AMPure XP-Medium kit and sequenced on the BGI/MGI DNBSEQ-T7 platform. For long-read sequencing, high-molecular-weight gDNA was extracted from a single male pupa using the QIAGEN Genomic-tip kit to maximize yield and integrity. The library was prepared using the Ligation Sequencing Kit V14 (SQK-LSK114, Oxford Nanopore Technologies plc., UK) with a targeted insert size of 40 kb and sequenced on the ONT PromethION platform in Super High-Accuracy (SUP) model. For Hi-C sequencing, gDNA was extracted from six male pupae. The Hi-C library was prepared using established protocols involving MboI restriction enzyme digestion, followed by deep sequencing on the BGI/MGI DNBSEQ-T7 platform. For RNA-seq, total RNA was extracted from eggs, larvae, pupae, and distinct male and female adults using TRIzol Reagent. Equal amounts of RNA from each stage were pooled, and the library was constructed using the VAHTS mRNA-seq v2 Library Prep Kit and sequenced on the BGI/MGI DNBSEQ-T7 platform.

The integrity and purity of all extracted nucleic acids were assessed using standardized procedures, including gel electrophoresis, a NanoDrop One spectrophotometer, and a Qubit fluorometer, ensuring quality standards were met for downstream library preparation. Sequencing services were provided by Annoroad Gene Technology (Beijing, China), with the exception of ONT sequencing, which was performed by GrandOmics Biosciences (Wuhan, China).

The sequencing effort yielded a total of 62.51 Gb of clean Illumina short-read data, providing approximately 191.76 × coverage for genome survey; 36.99 Gb of ONT long-read data, providing 113.46 × coverage for de novo assembly; and 73.06 Gb of ultra-deep Hi-C data, equating to 224.11 × coverage for scaffolding (Table 1).

Table 1.

Summary of genomic libraries, sequencing data output, and coverage statistics for the Sitotroga cerealella genome assembly.

Library type	Total reads	Total yield (bp)	Average length (bp)	N50 (bp)	Coverage (X)
WGS	416,745,370	62,511,805,500	150	—	191.76
ONT-sup	1,911,157	36,986,461,361	19,352.92	27,808	113.46
Hi-C	487,060,258	73,059,038,700	150	—	224.11
RNA-seq	150,575,932	22,586,389,800	150	—	—

WGS: Whole Genome Sequencing (Illumina short reads); ONT-sup: Oxford Nanopore Technologies Super High-Accuracy model; Hi-C: Chromosome conformation capture data. Coverage (X) is calculated based on the estimated genome size. “-” indicates data not applicable for coverage calculation or length statistics in this context.

Genome survey and K-mer analysis

A genome survey was conducted to estimate genome size, heterozygosity, and repeat content, which facilitated the optimization of parameters for downstream assembly. First, raw short-read data were quality-controlled and filtered using fastp v0.23.4⁵ with specific parameters “-q 20 -D -g -x -u 10 -5 -r -c”. This process involved retaining bases with a Phred quality score ≥20, removing PCR duplicates “-D”, trimming poly-G/X tails “-g -x”, and discarding reads containing >10% unqualified bases “-u 10”. Additionally, base correction in overlapping regions of paired-end reads was enabled “-c”. The quality-filtered reads were subjected to k-mer frequency distribution analysis to model intrinsic genomic complexity. A 21-mer (k = 21) distribution was generated using Jellyfish v2.2.3⁶, and the resulting histogram was analyzed using GenomeScope v2.0⁷. The analysis estimated the haploid genome size to be approximately 288.37 Mb with a high heterozygosity rate of 1.47% (Fig. 1). We also systematically performed k-mer frequency analysis with four alternative k-mer lengths (k = 17, 19, 23, 25) using the same workflow, yielding genome size estimates ranging from 280.95 Mb to 293.79 Mb, with corresponding heterozygosity estimates between 1.37% and 1.52%.

Fig. 1.

K-mer frequency distribution and genome size estimation using GenomeScope. The plot illustrates the 21-mer (k = 21) spectral analysis generated from Illumina short-read data.

Genome assembly and scaffolding

The initial de novo assembly was performed using Hifiasm v0.25.0⁸ with parameters “-l 3 -dual-scaf -ont -rl-cut 10000 -sc-cut 15”. The assembly was run using the highly accurate ONT mode “-l 3” and the dual scaffolding option “-dual-scaf”. To enhance contig quality, low-quality reads were filtered out based on two criteria: length < 10,000 bp and mean base quality score < 15. Assembly continuity was further refined by filtering contigs based on sequencing depth. Hifiasm was instructed to retain only contigs with a sequencing depth greater than 4 × while removing contigs assembled from low-depth reads (approximately 1/10 of the expected depth). Although the k-mer analysis indicated a high heterozygosity rate, subsequent BUSCO analysis (see below) did not reveal significant duplication or unphased haplotypes in the preliminary assembly. Therefore, the standard purging of redundant heterozygous contigs was not required.

Contigs were anchored and oriented into chromosome-level pseudomolecules using YAHS v1.2⁹ with Hi-C data. First, the Hi-C raw data were processed using Chromap v0.2.6¹⁰ for quality control, which included read alignment, duplicate removal, and Hi-C contact extraction, resulting in an alignment file. Two rounds of scaffolding were performed using YAHS with default parameters. Following the first round of scaffolding, the resulting assembly was manually reviewed and corrected using Juicebox v1.11.08¹¹ to resolve potential misjoins, translocations, and inversions before the final second round of scaffolding. The sequencing depth for each pseudochromosome was assessed using SAMtools v1.10¹² after aligning the WGS reads (either HiFi reads using “minimap2 -ax map-hifi” or Illumina short reads using “minimap2 -ax sr”) to the assembly. Since the sequenced individual was male (ZZ sex chromosomes), sex/autosomal assignments were not determined based on sequencing depth alone. The high quality of the resulting chromosome anchoring was validated by the Hi-C contact heatmap (Fig. 2), which demonstrated clear inter- and intra-chromosomal interactions and yielded 30 chromosome-level assemblies.

Fig. 2.

Hi-C contact map showing genome-wide chromatin interactions. The x and y-axes represent the 30 assembled pseudochromosomes. The heatmap visualizes the contact intensity of chromatin interactions across the genome.

The assembly achieved a total size of 325.99 Mb and a GC content of 39.07% (Table 2, Fig. 3). It exhibited extremely high continuity, with only 32 scaffolds and 33 contigs. The longest scaffold and contig both reached 21.042 Mb, and the Scaffold N50 and Contig N50 lengths were both 12.046 Mb. This high continuity resulted in the majority of the chromosomes achieving a telomere-to-telomere (T2T) level of completeness. The assembly was highly anchored, with 30 pseudochromosomes totaling 325.64 Mb, representing an anchoring rate of 99.89%. Among the 30 assembled pseudochromosomes, 22 chromosomes contain (TTAGG)n telomere repeats detected at both ends, and 7 chromosomes have telomere repeats detected at one end. In total, 51 telomere ends were identified across the 30 chromosomes, with only 9 ends missing detectable telomere repeats. Detailed information on the length and sequencing depth of each chromosome is provided in Table 3.

Table 2.

Statistical summary of the final S. cerealella genome assembly and quality assessment.

Content	Values
Assembly size (bp)	325,990,210
Number of pseudo-chromosomes (sizes)	30 (325,642,844)
Number of scaffolds/contigs	32/33
Longest scaffold/contig (Mb)	21.042/21.042
N50 scaffold/contig length (Mb)	12.046/12.046
GC content (%)	39.07
BUSCO completeness (%)	99.7
S	99.0
D	0.7
F	0.1
M	0.2
Mapping ratio of reads (%)
WGS	96.27
ONT-sup	99.75
RNA-seq	95.43

N50: The shortest sequence length at 50% of the genome. BUSCO: Benchmarking Universal Single-Copy Orthologs (Insecta_odb10 dataset, $n = 1,367$); S: Complete and single-copy; D: Complete and duplicated; F: Fragmented; M: Missing2. Mapping ratio indicates the percentage of reads from each library type successfully aligned to the final assembly3.

Fig. 3.

Circos plot characterizing genomic features of the Sitotroga cerealella assembly. The distribution of genomic features is visualized in concentric circles with a non-overlapping window size of 100 kb. From the outermost to the innermost circles are: chromosome length, GC content, gene density, DNA transposon density, SINE density, LINE density, and LTR retrotransposon density.

Table 3.

Length distribution and sequencing depth coverage across the 30 pseudochromosomes.

Chromosome	Length (bp)	N (bp)	Telomere motif (TTAGG)n	ONT-sup (X)	WGS (X)
Chr01	21042213	0	Both ends	105.908	189.64
Chr02	13885405	0	Both ends	103.046	177.681
Chr03	13825472	0	Both ends	103.588	171.963
Chr04	13597665	0	Both ends	101.276	173.077
Chr05	13504154	0	Both ends	102.986	173.076
Chr06	13022873	0	Both ends	101.521	172.641
Chr07	12909381	0	Both ends	103.257	177.741
Chr08	12855467	0	Both ends	102.069	175.084
Chr09	12497446	0	One end	102.893	170.333
Chr10	12483135	0	One end	102.337	173.437
Chr11	12273011	0	Both ends	102.523	174.38
Chr12	12046246	0	No	104.237	171.492
Chr13	11929006	0	Both ends	100.621	168.801
Chr14	11799521	0	Both ends	100.992	170.533
Chr15	11306274	0	One end	100.164	177.793
Chr16	11088485	0	One end	102.559	168.374
Chr17	11003518	100	One end	102.116	167.966
Chr18	10915838	0	Both ends	102.557	170.348
Chr19	10902425	0	Both ends	100.974	171.023
Chr20	9787900	0	Both ends	99.1623	165.071
Chr21	9286679	0	Both ends	101.702	163.747
Chr22	8359135	0	One end	100.184	166.77
Chr23	8351648	0	Both ends	99.1914	168.337
Chr24	8042400	0	Both ends	98.9423	164.33
Chr25	7881224	0	Both ends	96.6183	168.424
Chr26	7783196	0	Both ends	101.75	170.577
Chr27	6453854	0	Both ends	101.52	167.096
Chr28	5918331	0	Both ends	98.0363	158.731
Chr29	5600947	0	Both ends	96.4336	166.444
Chr30	5289995	0	One end	99.226	165.228

Chromosomes are numbered Chr01 through Chr30 based on length. ONT-sup (X): Average sequencing depth derived from Nanopore long reads. WGS (X): Average sequencing depth derived from Illumina short reads.

Genome annotation

To construct the species-specific repeat library, RepeatModeler v2.0.5¹³ was utilized with the “-LTRStruct” option enabled to incorporate an additional LTR search pipeline based on de novo prediction and characteristic repeat structures. The resulting de novo repeat library was then merged with known repetitive element databases, Dfam v3.8¹⁴ and RepBase-20181026¹⁵, to form the final comprehensive repeat reference database. Repetitive sequences within the genome were subsequently predicted and classified by aligning the assembly against the final merged repeat database using RepeatMasker v4.1.5¹⁶. In total, 133.05 Mb of sequences were identified as repetitive elements, accounting for 40.81% of the entire genome (Table 4). The major classified families included 8.56% Long Terminal Repeats (LTRs), 7.54% Long Interspersed Nuclear Elements (LINEs), 7.03% Short Interspersed Nuclear Elements (SINEs), 2.74% Rolling-circles, 2.69% DNA transposons, and 11.18% Unknown (Fig. 3).

Table 4.

Summary statistics of structural and functional annotations for protein-coding genes.

Annotation	Number
Structure annotation
Number of protein-coding genes	13,840
Number of predicted protein sequences	20,106
Mean protein length (aa)	610.6
Mean gene length (bp)	12,867.7
Gene ratio	54.63%
Number of exons per gene	7.9
Mean exon length (bp)	302.7
Exon ratio	10.24%
Number of CDSs per gene	7.6
Mean CDS length (bp)	220.6
CDS ratio	7.15%
Number of introns per gene	6.9
Mean intron length (bp)	1,664.5
Intron ratio	44.39%
Function annotation
Number of genes matching Uniprot records	13,413
Number of genes labelled as “Uncharacterized protein”	2,363
Number of genes labelled as “unknown function”	490
Number of genes with InterProScan annotations	10,893
Number of genes with GO items from InterProScan annotations	6,404
Number of genes with eggNOG annotations	13,017
Number of genes with GO items from eggNOG annotations	8,273
Number of genes with Enzyme Codes (EC) from eggNOG annotations	2,773
Number of genes with KEGG ko terms from eggNOG annotations	7,626
Number of genes with KEGG pathway terms from eggNOG annotations	4,729
Number of genes with COG Functional Categories from eggNOG annotations	11,350
Number of genes with GO items (combining InterProScan and eggNOG results)	9,799
Number of genes with KEGG pathways items (combining InterProScan and eggNOG results)	4,729

CDS: Coding DNA Sequence; GO: Gene Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes; EC: Enzyme Commission number; COG: Clusters of Orthologous Groups.

Non-coding RNAs (ncRNAs) were annotated using a dual strategy. Homology-based annotation of rRNAs, snRNAs, miRNAs, and other ncRNAs was performed by aligning the genome against the Rfam database using Infernal v1.1.5. Transfer RNA (tRNA) prediction was conducted de novo using tRNAscan-SE v2.0.12¹⁷, with low-confidence tRNAs filtered by the built-in EukHighConfidenceFilter script. The final annotation identified 815 ncRNA genes, comprising 271 tRNAs, 150 rRNAs, 86 miRNAs, 70 snRNAs, 69 ribozymes, and 2 lncRNAs. The snRNA set included 48 major spliceosomal RNAs (U1, U2, U4, U5, U6, U11), 4 minor spliceosomal RNAs (U4atac, U6atac), 9 C/D box snoRNAs, and 4 HACA-box snoRNAs.

Protein-coding gene structures were predicted using the MAKER pipeline (v3.01.04)¹⁸, which integrates three lines of evidence. First, ab initio gene prediction was performed using BRAKER v3.0.6¹⁹ and GeMoMa v1.9²⁰. The RNA-seq short reads were first aligned to the genome using minimap2 “-x splice:sr” to generate BAM alignment files. BRAKER automatically trained the ab initio predictors Augustus v3.4.0²¹ and GeneMark-ETP²², leveraging arthropod protein sequences from the OrthoDB12 v1 database²³ and transcriptome data to enhance accuracy. GeMoMa utilized protein homology and intron positions from high-quality annotated genomes of related species, including Drosophila melanogaster (Diptera) and four Lepidoptera (Amyelois transitella, Bombyx mori, Spodoptera frugiperda, Pectinophora gossypiella). GeMoMa was performed with parameters “GeMoMa.c = 0.4 GeMoMa.m = 65000”. The results from BRAKER and GeMoMa were merged to provide the ab initio input for MAKER, broadening the candidate gene pool. Second, transcriptome-based evidence was generated by performing reference-guided transcript assembly of the RNA-seq BAM files using StringTie v3.0.1²⁴. Third, protein homology evidence was supplied by aligning the same homologous protein sequences used in the GeMoMa analysis to the genome. The MAKER pipeline predicted a total of 13,840 protein-coding genes (Table 4, Fig. 3). Key structural statistics revealed an average gene length of 12,867.7 bp, containing an average of 7.9 exons (302.7 bp average length) and 6.9 introns (1,664.5 bp average length), resulting in 7.6 CDS per gene (220.6 bp average length).

Gene functions were assigned using two strategies. First, homology-based annotation was performed by searching the UniProtKB database (v202503, encompassing SwissProt and TrEMBL) using Diamond v2.1.7.161²⁵ in high-sensitivity mode “–very-sensitive -e 1e-5”. Second, integrated functional analysis was conducted to predict conserved protein sequences, domains, Gene Ontology (GO) terms, and pathways (KEGG, Reactome). This involved searching the Pfam database²⁶ using InterPro v5.70-102.0²⁷ and utilizing eggNOG-mapper v2.1.12²⁸ against the eggNOG database v5.0.2²⁹. Overall, 13,413 genes (96.91%) matched records in the UniProtKB database (Table 4). InterPro successfully identified protein domains for 10,893 protein-coding genes. Furthermore, InterPro and eggNOG-mapper collectively assigned GO terms to 9,799 genes and KEGG pathway entries to 4,729 genes.

Data Records

The complete genome assembly and raw sequencing data for S. cerealella have been deposited at the National Center for Biotechnology Information (NCBI). The project is registered under BioProject accession number PRJNA1308772. The final genome assembly is available through GenBank under accession JBQQXG000000000³⁰. All raw sequencing data used for the assembly, including Illumina short reads, ONT long reads, Hi-C reads, and RNA-seq reads, are accessible via the NCBI Sequence Read Archive (SRA) under accession numbers SRR35946759³¹ (Illumina WGS), SRR35832401³² (Hi-C data), SRR35832037³³ (RNA-seq data), SRR35831679³⁴ and SRR35831680³⁵ (ONT long reads). The comprehensive annotation results, including the gene structure file (GFF3), coding sequences and protein sequences have been deposited in the FigShare database³⁶.

Technical Validation

To assess data utilization and assembly integrity, raw sequencing reads from Illumina survey, RNA-seq, and ONT-sup platforms were mapped back to the genome assembly using Minimap2, with statistics calculated via SAMtools. All datasets exhibited high mapping rates exceeding 95% (Illumina survey: 96.27%; RNA-seq: 95.43%; ONT-sup: 99.75%), confirming strong consistency between the raw data and the final consensus. The resulting assembly comprised 30 pseudochromosomes totaling 325.64 Mb, achieving a chromosomal anchoring rate of 99.89%, with Hi-C interaction heatmaps (Fig. 2) demonstrating high-quality chromosome-level scaffolding. Genome completeness was evaluated using BUSCO v5.7.19 against the insecta_odb10 lineage (n = 1,367), yielding a completeness score of 99.7% with a minimal duplication rate of 0.7%. Notably, although the final assembly size (325.64 Mb) exceeds the GenomeScope k-mer estimate (288.37 Mb), the negligible BUSCO duplication rate confirms that this discrepancy is not driven by the retention of redundant haplotypes despite high heterozygosity (1.47%). Rather, it likely reflects the superior resolution of complex repetitive elements by long-read sequencing that were underestimated by short-read k-mer analysis. The short-read k-mer analysis underestimates the genome size mainly due to the high content of complex repetitive elements in the S. cerealella genome. Our assembly shows that repetitive elements account for 40.81% of the genome, most of which are long, highly similar repetitive sequences. Short reads (150 bp) cannot span these long repetitive regions, leading to the collapse of repetitive sequences in k-mer analysis, thus significantly underestimating the actual genome size. The ONT ultra-long reads (average length 19.35 kb, N50 27.8 kb) used in our de novo assembly can fully span these long repetitive regions, accurately resolving and assembling these sequences that were missed by short-read k-mer analysis. Consistent with these genomic metrics, the predicted protein set achieved 99.1% completeness (C:99.1% [S:80.2%, D:18.9%], F:0.1%, M:0.8%). These results collectively demonstrated the high quality and reliability of the S. cerealella genome assembly and annotation.

Author contributions

Z.Y. and F.Q. contributed to the conceptualization and research design. Q.L. and Y.L. collected the samples and performed the preliminary biological work. Z.Y. and Y.L. conducted the primary data analysis, assembly, and annotation procedures. Y.L. and Z.Y. wrote and revised the draft manuscript. All co-authors reviewed the manuscript and provided their final approval.

Funding

This study was supported by National Key R&D Program of China (2023YFD1400604), Natural Science Foundation of Hainan Province (Grant no. 323QN262), and the Fundamental Research Funds for the Central Universities (Grant no. KJYQ2026038).

Data availability

The complete genome assembly and raw sequencing data for S. cerealella have been deposited at the National Center for Biotechnology Information (NCBI). The project is registered under BioProject accession number PRJNA1308772. The final genome assembly is available through GenBank under accession JBQQXG000000000. All raw sequencing data used for the assembly, including Illumina short reads, ONT long reads, Hi-C reads, and RNA-seq reads, are accessible via the NCBI Sequence Read Archive (SRA) under accession numbers SRR35946759 (Illumina WGS), SRR35832401 (Hi-C data), SRR35832037 (RNA-seq data), SRR35831679 and SRR35831680 (ONT long reads). The comprehensive annotation results, including the gene structure file (GFF3), coding sequences and protein sequences have been deposited in the FigShare database.

Code availability

No specific custom scripts were utilized in this study. All bioinformatics analyses were performed using standard, publicly available software tools and pipelines, with versions and parameters fully detailed in the Methods section.

Competing interests

The authors declare no competing interests.