African swine fever virus (ASFV) is a highly pathogenic virus that produces a deadly hemorrhagic disease in both domestic and feral pigs, with extremely high mortality rates. Since 2007, ASFV has spread across several countries in Europe and Asia, wiping out hundreds of millions of pigs (Dixon et al., 2019). This transboundary virus causes severe economic losses and poses a significant threat to the global pig industry and food security. Despite extensive research, no effective vaccine or treatment is currently approved for ASFV. Moreover, the complete molecular mechanisms underlying ASFV pathogenicity remain to be fully elucidated.

ASFV is a large, complex DNA virus belonging to the Asfarviridae family. Its genome structure has been described as linear, double-stranded DNA of approximately 170–190 kb that encodes 151–167 open reading frames (ORFs) (Dixon et al., 2013). The ASFV genome can be divided into three major regions: the central conserved region (CCR, ∼125 kb), which is flanked by the left variable region (LVR, ∼40 kb) and the right variable region (RVR, ∼20 kb). Genes in the CCR play significant roles in viral replication, assembly, and immune evasion. Most of the genes in the LVR and RVR are members of five multigene families (MGF), including MGF 100, 110, 300, 360 and 505, which are involved in virus replication, cell and host tropism, immune evasion, and virus virulence (Chapman et al., 2008; Dixon et al., 2013).

Both ends of the linear double-stranded genome contain inverted terminal repeats (ITRs) and hairpin loops, which play important roles in genome stability, replication, and packaging (Gonzalez et al., 1986; Dixon et al., 2013). These regions of the genome are difficult to assemble when using only short-read sequencing data (Forth et al., 2019). The BA71V genome assembly, the only genome with complete ITR and hairpin loop sequences, revealed four open reading frames (ORFs) within the ITRs (Yanez et al., 1995). This emphasizes the importance of a full-length ASFV genome assembly with complete ITRs and hairpin loop sequences. The limitations of short-read sequencing highlight the need for long-read sequencing such as Oxford Nanopore technology (Oxford Nanopore Technologies, Oxford, United Kingdom). Long reads (> 10 kb) aid in creating complete, high-quality ASFV genome assemblies (Scarano et al., 2024), and resolving the ambiguity in assembling the ITR and hairpin loop regions. In this study, we aimed to generate a full-length ASFV genome assembly of Chonburi_2024_209-MA that includes intact terminal hairpin structures using Oxford Nanopore long-read sequencing, and to identify potential methodological challenges specific to sequencing genomes with terminal hairpin loops.

Aiming to get a full-length ASFV genome, we employed Oxford Nanopore Technologies (ONT) to sequence the genome of a cell culture-adapted ASFV originally isolated from infected Large White pigs in Phanat Nikhom District, Chonburi, Thailand in August 2023. The isolated ASFV was cultured in MA-104 cells (ATCC CRL-2378.1) in DMEM (Cytiva; Cat# SH30243.02) with 10% fetal bovine serum (FBS; Sigma; Cat# F7524) and was serially passaged until passage 50. Details of ASFV isolation and passaging can be found in a previous study (Thaweerattanasinp et al., 2024). At passage 50, the cell suspension was centrifuged at 5,000 × g for 10 min at 4°C. Cell supernatant was then harvested, filtered through 0.45-μm syringe filters, and concentrated to a volume of 20 mL using Amicon Ultra centrifugal filters (Merck; Cat# UFC9100). The resulting supernatant was ultracentrifuged at 150,000 × g for 3 h at 4°C. The supernatant was then carefully decanted into a waste container, while the pellet was resuspended in 1 mL of PBS. ASFV DNA was extracted from the resuspended pellet using the High Pure Viral Nucleic Acid Kit (Roche; Cat# 11858874001) according to the manufacturer’s instructions. This cell culture-adapted ASFV isolate was named Chonburi_ 2021_209-MA.

The extracted DNA was sequenced with both long-read and short-read technologies. Oxford Nanopore Technologies was applied for long-read sequencing. The PCR-free library was prepared from 250 ng of high molecular weight genomic DNA using a 1D Native Barcoding Kit 24 V14 (SQK-NBD114.24; Oxford Nanopore Technologies, Oxford, United Kingdom). The library was loaded onto a R10.4.1 flow cell and ran for 12 h on a MinION Mk1C sequencer. Regarding short-read sequencing, genomic DNA was used for library preparation with MGIEasy FS DNA Library Prep Set (MGI Tech, Shenzhen, China) according to the manufacturer’s instructions. The library was sequenced using DNASEQ-G400 (150-bp paired-end reads).

Oxford Nanopore read base calling was done using Guppy v6.3.8 (Oxford Nanopore Technologies, Oxford, United Kingdom). Oxford Nanopore reads were mapped on ASFV Georgia 2007/1 (GCF_003047755.2) using minimap2 v2.26 (Li, 2018). Mapped reads were extracted using SAMtools v1.17 (Danecek et al., 2021). Since Oxford Nanopore reads are prone to high error rates, we performed read correction, which denoised reads by generating consensus sequences of overlapping reads, and trimming using CANU v2.2 (Koren et al., 2017) prior to assembly with the same software. The corrected and trimmed Oxford Nanopore reads and MGI short reads were mapped onto the draft Chonburi_2024_209-MA assembly using minimap2 and Bowtie2 v2.5.1 (Langmead and Salzberg, 2012), respectively. Read depth supporting every nucleotide of the draft assembly was called using SAMtools. MUMmer4 (Marcais et al., 2018) was used to generate an alignment dot plot between the draft assembly and Georgia2007/1 genome.

To get a complete full-length genome assembly, we first determined the hairpin loop sequences at both termini of the draft assembly. The terminal DRC sequences extending beyond the hairpin loop sequences have been trimmed off. The trimmed assembly was polished with short reads using Pilon v1.24 (Walker et al., 2014). To maximize the accuracy of the assembly, a polishing step was repeated until there were no additional nucleotide changes. The trimmed and polished assembly, so called complete full-length Chonburi_2024_209-MA, was subsequently annotated using Prokka v1.14.6 (Seemann, 2014).

To investigate genetic variations, comparative analyses were performed. Complete full-length Chonburi_2024_209-MA assembly, Georgia2007/1, the first reported genotype II ASFV in China: Pig/HLJ/2018 (GCA_004338215), the first hairpin-to-hairpin ASFV genome: BA71V (GCF_000858485), and ONT-based HK_NT_202103 (GCA_030515915) were included in the analyses. To illustrate the advantage of long-read assembly, we constructed short-read assembly of Chonburi_2024_209-MA, so called Chonburi_2024_209-MA short-read assembly and included it in the comparative analyses. In short-read assembly, short reads were mapped on the draft Chonburi_2024_209-MA assembly with Bowtie2. Mapped reads were extracted using SAMtools and assembled with SPAdes v3.15.4 (Prjibelski et al., 2020).

Six ASFV genomes in total were aligned with MAFFT v7.525. To investigate structural variation among genomes, the boundaries of alignment blocks in each genome were extracted and visualized. We also assessed nucleotide-level variation i.e., SNPs, and indels. However, Chonburi_2024_209-MA short-read assembly and BA71V were excluded from the analysis. The alignment revealed no difference between the complete full-length Chonburi_2024_209-MA assembly and Chonburi_2024_209-MA short-read assembly. Meanwhile, BA71V, a genotype I ASFV, exhibited an unnecessarily high number of variations, making it unsuitable for comparative analysis in this level. The genomes of complete full-length Chonburi_2024_209-MA, Pig/HLJ/2018, and HK_NT_202103 were mapped to Georgia2007/1. Variants were called against a reference genome Georgia2007/1 using BCFtools v1.22 (Li, 2011). The effects of the variants on coding genes were assessed with SnpEff v5.2 (Cingolani et al., 2012).

Genome assemblies based on only short-read data present difficulty in resolving repetitive regions (Treangen and Salzberg, 2011). The presence of ITRs at both ends of ASFV genome together with low read depth at terminal sequences make it difficult to obtain a full-length ASFV assembly, as also noted by Chapman et al. (2008). Complete ITRs and hairpin sequences have been resolved in only one genome, BA71V, which utilized sequencing data from a combination of Sanger sequencing technology (Yanez et al., 1995), previously reported ITR sequences (de la Vega et al., 1994), and hairpin loop sequences obtained with a Maxam–Gilbert technique (Gonzalez et al., 1986). Here, we report another full-length ASFV genome with complete ITR and hairpin sequences. Consistent with the BA71V and HK_NT_202103 genomes, our genome contains four ORFs encoding two genes within both ITRs. These findings demonstrate the significance of a complete ASFV genome assembly in providing valuable insights into ASFV terminal region structure.

The hairpin loop sequences consist of 37 AT-rich nucleotides, as reported in BA71V (Gonzalez et al., 1986; Yanez et al., 1995). However, the nucleotide sequences are not identical. The secondary structure of the hairpin loop with the minimal free energy shows 3 single-stranded loops and 2 stems within the 37-nucleotide sequence (Figure 3). This structure is different from the structure reported in Gonzalez et al. (1986), plausibly because we took adjacent sequences at both sides of the hairpin loop into account in the evaluation of hairpin loop formation. The negative free-energy value indicates spontaneous formation and higher stability of the loop presented here, compared to that reported by Gonzalez et al. (1986). These secondary structures facilitate DNA folding and circularization, providing essential regulatory elements for viral functions and persistence (Chapman et al., 2008). Understanding these hairpin loop topologies is critical for interpreting the virus’s reproduction kinetics and investigating possible antiviral treatments (Gonzalez et al., 1986; Dixon et al., 2013).

The deletion pattern in RVR of our complete full-length Chonburi_2024_209-MA assembly, when compared to Georgia 2007/1, was also similarly observed in the genome of ASFV strain Ratchaburi_2023_001-MA (GCA_032918185.1) (Thaweerattanasinp et al., 2024), and other ASFV genomes adapted to monkey kidney cells (Chapman et al., 2008; Krug et al., 2015; Mazloum et al., 2021), suggesting that these changes may have provided a selective advantage during the virus adaptation process, which is essentially a selection pressure for less lethal viral strains that can infect monkey kidney cells.

The genetic variation of four genotype II ASFV isolates provides insights into the mutation profile, revealing both shared and unique variants. The shared variants may represent conserved mutations within this lineage, whereas the unique mutations could contribute to strain-specific phenotypic differences or host adaptation mechanisms. The high frequency of C-to-T substitutions in MGF505-2R and MGF505-3R genes may be the result of APOBEC mutagenesis. The cytidine deamination activity of APOBEC/AID protein family, especially APOBEC3, can cause C-to-U alteration in ssDNA during replication, subsequently resulting in C-to-T substitution (Hoopes et al., 2016; Xu et al., 2020). APOBEC3 enzymes typically target cytosines within specific sequence motifs, most commonly 5′-TC-3′ or 5′-CC-3′, which is consistent with the pattern observed in this study. This mechanism is an innate immune response by hosts against viruses via deaminase-dependent hypermutation, which has been reported in several viruses e.g., HIV-1, human herpesviruses, human papillomavirus (Xu et al., 2020; Jonathan and Ikeda, 2023). However, this phenomenon has never been reported in ASFV. This study is the first to suggest that APOBEC mutagenesis may occur in the ASFV genome. Although other ASFV genomes adapted to MA-104 cell culture did not exhibit a similar pattern of substitutions (Thaweerattanasinp et al., 2024), the influence of the cell lines used for ASFV propagation cannot be overlooked. LaRue et al. (2008) reported that humans and non-human primates possess 7 APOBEC3 genes, while pigs have only 2 genes. This might be the reason that underlies the finding of this study. Nonetheless, the cause of this C-to-T hypermutation needs further investigation.

The MGF505-2R and MGF505-3R proteins can counteract the host immune system, especially through the cGAS-STING pathway and ferroptosis (Sunwoo et al., 2024; Niu et al., 2025). The deletion of MGF360/505 gene clusters including MG505-2R and MGF505-3R has been reported to attenuate ASFV virulence (Rathakrishnan et al., 2022; Sunwoo et al., 2024; Vu and McVey, 2024). Krug et al. (2015) also reported that the deletions of MGFs potentially influenced ASFV adaptation to Vero cells. Although there was no deletion of the LVR in our genome, similar attenuation and adaptation to MA-104 cells likely occurred in Chonburi_2024_209-MA as the truncated polypeptides of not only MGF505-2R and MGF505-3R, but also MGF300-2R, MGF110-1L, and MGF360-11L potentially resulted in loss of function. The attenuated, cell-adapted variant may have gained a fitness advantage, enabling it to outcompete other strains and become dominant in the population.

This study represents a significant advance in ASFV genomics as it provides one of the first complete hairpin-to-hairpin genome assemblies. Previous attempts to assemble the ASFV genome using short-read technologies were limited in their ability to resolve the terminal regions (Chapman et al., 2008). Even some long-read based assembly approaches struggled with the complex hairpin structures (Rodriguez et al., 2015; Licheri et al., 2024). Our ability to extract a sufficient amount of native ASFV DNA for Oxford Nanopore sequencing and our awareness of the existence of hairpin structures enabled us to accurately reconstruct the terminal hairpin loops without the artificial duplications that occurred in previous assembly attempts (Thaweerattanasinp et al., 2024). In addition, our assembly strategy reveals the presence of the multigene families near the terminal regions, which are often incompletely assembled or incorrectly annotated in previously published reference genomes. The complete genome structure presented here, including intact hairpin loops, provides a more accurate template for future comparative genomics and functional studies of ASFV. This study also underscores the importance of utilizing long-read sequencing platforms that can sequence native DNA to perform genome assembly, not only for the ASFV, but also for other double-stranded DNA viruses with a terminal cross-bridge sequence, such as poxviruses (Dixon et al., 2013).