Pandemic preparedness is a strategic priority for the European Union driven by the substantial morbidity and mortality associated with recent viral emergence events such as SARS-CoV-2 in humans and Avian influenza virus H5N1 in birds and cattle [1]. Zoonotic spillover between host species has underpinned the major pandemics of the last 50 years including those caused by SARS-CoV-2, HIV and Influenza H1N1 [2].

In response, the EU and World Health Organisation have adopted One Health strategies to pandemic preparedness and response frameworks, [1, 3]. Surveillance initiatives, such as One Health – ALL Ireland for European Surveillance (OH-ALLIES), are being established to build capacity for detection of high-risk viral families circulating in animal populations while also enhancing the ability to discover previously unknown viruses in Ireland [4]. Within this context, “Pathogen X” and “Pathogen Y” refer to hypothetical, unknown human and veterinary agents, respectively, with pandemic potential and potential to cause significant health and socio-economic disruption [4–6]. Viruses are considered the most plausible Pathogen X or Pathogen Y agents due to their rapid evolution, host switching capacity and ability to alter their virulence before recognition [7–10].

Viral metagenomic next-generation sequencing (mNGS) enables unbiased detection of both known and novel viruses directly from clinical samples addressing gaps in targeted molecular diagnostics, which rely on prior knowledge of pathogen sequences and are not optimum when viral genomes are highly divergent or co-infections and unexpected pathogens are present [11, 12].

mNGS has already made important contributions to the detection and characterisation of emerging viral pathogens, exemplified by the identification of Schmallenberg virus as the aetiological agent of large bovine and ovine abortion storms in Europe in 2011 [13], and the rapid characterisation of SARS-CoV-2 from the first COVID-19 cases in China [7–10], where early sequences facilitated PCR assays[14, 15] and vaccine design [16–18], phylogenetic analyses to track viral spread [19, 20], and inference of phenotype from genotype [9]. Co-infections can also be identified using mNGS [21]. These strengths distinguish mNGS from targeted methods, yet there are several challenges to mNGS-based pathogen discovery and surveillance. Practical concerns, especially in resource-limited settings, include its cost and requirements for specialist equipment, computing and personnel [22–24]. The issue of reduced analytical sensitivity relative to PCR due to low viral fraction in many samples and the need for careful interpretation to ensure specificity. Recently, technical advances in mNGS have made it more accessible[22–24] and various approaches, including methods of host depletion[25–27] and viral enrichment [28, 29], have been developed to improve the sensitivity of viral mNGS.

There is a particular need for rigorously validated, practical mNGS workflows that can operate reliably on the heterogeneous specimens encountered at the animal–human interface spanning both simple swab matrices and complex tissues. This study develops and validates two optimised mNGS workflows, one for swab samples and the other for complex tissue matrices (Fig. 1), each integrating comprehensive quality control (QC) measures, from nucleic acid extraction through library preparation and sequencing to bioinformatic analysis, to support robust virus discovery and characterisation. The methods detailed here cover the quality metrics, performance for detection of known and unknown viruses, and application of these workflows for consensus genome assembly, providing a framework for deployment in One Health surveillance and outbreak investigation.

Reference Biospecimen Repository

A biospecimen repository of clinical samples positive for known viruses was assembled from five independent sources: the Department of Agriculture, Food and the Marine (DAFM), the Agriculture and Food Development Authority (Teagasc), UCD School of Veterinary Medicine, St Vincent’s University Hospital (SVUH) and Mater Misericordiae University Hospital (MMUH). Eight swab and six tissue samples representing nine anatomical sites, 12 viral species and nine host species were included for method development and evaluation (Table 1). All samples from non-human animals were obtained from naturally deceased animals during routine postmortem examinations without the use of anaesthesia or euthanasia.

Swab specimens were collected using flocked swabs and a range of transport and storage conditions. DAFM collected swabs into universal transport medium (COPAN, 330C), Teagasc collected swabs (COPAN, 552C) into PrimeStore molecular transport media containing guanidine thiocyanate lysis buffer (Thermo Fisher Scientific, R13905), while SVUH and MMUH collected dry swabs (COPAN, 552C) without transport medium. SVUH and MMUH samples were extracted within 2 days of collection to avoid freeze-thaw cycles. DAFM and Teagasc samples were stored at -80°C and − 20°C, respectively, until extraction (Table 1).

Tissue biopsies were collected during routine postmortem examinations. Grey seal (Halichoerus grypus) carcasses processed by UCD School of Veterinary Medicine were stored at -20°C prior to necropsies after which tissue biopsies were stored at -80°C before processing. Sika deer (Cervus nippon) and Asian elephant (Elephas maximus) tissues collected from fresh animals by UCD Veterinary Medicine were similarly stored at -80°C before processing. DAFM tissue samples were stored at -80°C and had undergone at least one freeze-thaw cycle before extraction (Table 1).

Ethical and regulatory oversight was secured through institutional review processes. The UCD Animal Research Ethics Committee granted exemptions for the use of samples collected as part of routine diagnostics or postmortem from DAFM (AREC-E-24-39-Gautier), Teagasc (AREC-E-25-25-Gautier) and UCD School of Veterinary Medicine (AREC-E-22-04-Jahns), and the Human Research Ethics Committee provided approval to work with human clinical samples provided by hospitals (306-LS-CSD-25-Mallon).

Dry Mpox-positive swabs were processed by adding 850 µl Buffer AVL, incubating for 10 minutes, and using 700 µl lysate for TNA extraction with the QIAamp Viral RNA Mini Kit (Qiagen, 52906) following the manufacturer’s instructions with the sole modification of replacing 6 µl carrier RNA with 6 µl linear acrylamide (5 mg/ml, Thermo Fisher Scientific; AM9520). Linear acrylamide was used instead of carrier RNA to prevent sequencing of carrier RNA. All other swabs were processed by vortexing for 5 seconds then 300 µl transport media was used for TNA extraction with the Liferiver Viral RNA Isolation Kit (P20211009) on the Liferiver automated extractor platform, following the manufacturer’s instructions except that 6 µl linear acrylamide (5 mg/ml) was substituted for the carrier RNA.

RNA was extracted from tissue samples using the RNeasy Mini Kit (Qiagen, 74106). Tissue was cut on dry-ice into 10–20 mg pieces and transferred into 600 µl Buffer RLT supplemented with 10% β-mercaptoethanol. Samples were sonicated on the high setting of the Biorupter NextGen sonicator (diagenode) for three cycles of 30 second ON and 30 second OFF at 4°C followed by column-based homogenisation (BioTech, HCR003) at 14,000 x g for 120 seconds. RNA was purified using the Qiagen RNeasy Mini Kit and residual DNA was digested with the DNA-free DNA Removal Kit (Thermo Fisher Scientific, AM1906) according to manufacturer’s instructions.

Nucleic acid extract purity and quantity was assessed using the ND-1000 NanoDrop spectrometer (Labtech International). RNA concentrations were determined using the Qubit RNA High Sensitivity RNA Kit (Thermo Fisher Scientific, Q32852) and DNA concentrations using the Qubit DNA High Sensitivity Kits (Thermo Fisher Scientific, Q33230). RNA from tissue extracts was assessed using the Bioanalyzer RNA Nano chip (Agilent, 5067 − 1512) or the Tapestation RNA ScreenTape (Agilent, 5067–5579). Swab extracts with concentrations below the limit of detection of these platforms were not subjected to integrity assessment.

First-strand cDNA synthesis was performed using SuperScript IV (Thermo Fisher Scientific, 18090050) by combining 11 µl TNA extract with 1 µl 10 mM deoxynucleotides and 1 µl 50 ng/µl random hexamers (Thermo Fisher Scientific, 51709), then incubating at 65°C for 5 minutes. Then, 4 µl 5X SSIV Buffer, 1 µl 100 mM DTT, 1 µl RNase OUT (Thermo Fisher Scientific, 100000840) and 1 µl SuperScript IV enzyme were added to reaction mixes, which were incubated at 23°C for 10 minutes, 50°C for 30 minutes and 80°C for 10 minutes.

Second-strand cDNA synthesis was carried out using the Klenow Fragment (Thermo Fisher Scientific, EP0051) by addition of 0.5 µl 10 U per 1 µl Klenow Fragment, 1 µl 10 mM deoxynucleotides, 5 µl Klenow Fragment Buffer and 33.5 µl nuclease-free water to first-strand cDNA synthesis reaction mixes. The reaction was incubated at 37°C for 30 minutes, then heat inactivated at 80°C for 5 minutes. The final product contained genomic DNA and ds-cDNA.

DNA libraries were generated from 50 pg to 50 ng of DNA (genomic DNA plus ds-cDNA) using the ThruPLEX DNA-Seq Kit (Takara Biosciences, R400674) according to the manufacturer’s instructions. Sequencing adapters and unique dual indexes were incorporated using PCR with tailed primers. Cycle numbers were adjusted based on DNA input: 8 cycles for 30–50 ng, 9 cycles for 10–30 ng, 10 cycles for 3–10 ng, 11 cycles for 1.5-3 ng, 12 cycles for 0.5–1.5 ng, 14 cycles for 0.2–0.5 ng and 16 cycles for 0.05–0.2 ng. Final DNA libraries were purified using bead:sample ratio of 1.0.

Library fragment size distribution was assessed using either the Bioanalyzer DNA High Sensitivity Chip (Agilent, 5067 − 4626) or the DNA D1000 TapeStation (Agilent, 5067–5584). Libraries were quantified with the Qubit DNA Broad Range Kit (Thermo Fisher Scientific, Q33260) or Qubit DNA High Sensitivity Kits (Thermo Fisher Scientific, Q33230).

These mNGS workflows are intended for deployment in scenarios where prior knowledge of circulating viruses and reference sequences may be incomplete or absent. Detection of 12 secondary infections demonstrates the capacity to identify unknown knowns, while recovery of Gammaherpesvirus genus sequences corresponding to PHV7 illustrate that the mNGS workflows can also detect and partially characterise known unknowns.

High-quality genomes suitable for downstream analyses were those with at least 75% genome coverage at 1x and mean read depth of at least 10 (Table 5) [39]. For nine non-segmented viruses meeting these criteria, coverage plots showed uniformly high depth across the genome length, except Enterovirus G from Sample 3, which had several gaps and areas of low coverage (Fig. 5). All other consensus genomes, which had at least 60% coverage at ≥ 10X read depth (the reduced coverage from 75% to 60% was caused by increased read depth requirements), were annotated with VAPiD [40] and uploaded to GenBank (Supplementary Table 2).

For segmented viruses, coverage plots revealed uneven depth and coverage among segments (Fig. 6). Consensus genomes were annotated with VAPiD [40] and submitted to GenBank for segments with at least 60% coverage at ≥ 10X read depth (Table 6).

Assembly of consensus genomes from mNGS data enabled verification of viral hits from the metagenomics pipeline based on the mapping of reads to multiple loci along reference genomes. High-quality consensus genomes were also assembled and deposited to GenBank.

In preparation for future viral outbreaks of unknown aetiology, mNGS workflows were designed and evaluated for swab and tissue samples positive for a range of known DNA and RNA viruses, collected from multiple anatomical sites and host species and subjected to diverse handling and storage timelines. QC analyses confirmed that these clinical specimens provided nucleic acid extracts spanning a wide range of concentrations, purity and integrity, yet the workflow remained sufficiently robust to detect viruses from poor-quality extracts. Applying the threshold of at least 2 unique matching reads for nucleotide-level searches (NT) and at least 10 unique matching reads for amino acid-level searches (NR), 89.5% of the 19 known viruses were detected, demonstrating that the selected extraction, library preparation and sequencing protocols are suitable for viral mNGS.

Detection of previously unrecognised or poorly characterised viruses is a key requirement of any mNGS-based pathogen surveillance system. The workflows established here generated data of sufficient quality to identify additional viruses in samples that had not previously tested positive for these agents, illustrating the capacity to detect previously unrecognised co-infections (“unknown knowns”). In addition, members of the Gammaherpesvirus genus were identified in Sample 14 using an amino acid search of the NR database, consistent with the presence of PHV7 – a virus lacking complete reference genome in the NT database. The logical next step is to apply these mNGS workflows to cases of suspected infectious disease of unknown aetiology, to assess performance in identifying truly novel or unanticipated viral infection, including unknown unknowns [42].

Assembly of viral consensus genomes was used to verify metagenomic hits by examining how reads distributed along reference genomes, enabling the identification of potential false positives. Identification of potential false positives in this study highlights the importance of complementing automated metagenomic classification with an independent verification step or other orthogonal methods.

High-quality consensus genomes were generated for several of the primary and secondary viral infections identified. An important advantage of mNGS over targeted virus identification methods, such as PCR and ELISA, is the breadth of genetic information obtained, which can be leveraged for downstream applications. Previous studies have shown how consensus genomes assembled from mNGS data can support further analyses by both the originating investigators and the wider research community once deposited in public open access repositories [12]. The workflows presented here can generate data of sufficient quality for the assembly of genomes with at least 65% coverage at 10X depth or greater, providing a foundation for applications such as phylogenetic analysis, molecular epidemiology, and exploration of genotype-phenotype relationships [43].

Previous studies have described mNGS workflows for viral pathogen identification across multiple sample matrices and body sites [26]. The present study extends this by demonstrating for the first time, an mNGS workflow tested on clinical samples from multiple host species, while explicitly documenting its robustness across nucleic acid extracts of varying concentrations, purity and integrity. Collectively, the tissue and swab mNGS workflows developed here have proven effective at detecting known viruses, previously unrecognised infections (unknown knowns), and partially characterised agents (known unknowns) from a spectrum of clinically relevant samples. These workflows are now ready for integration into passive and syndromic surveillance networks to evaluate their performance in real-world investigations of suspected infectious disease cases with unknown aetiology.