Optimisation of rapid untargeted nanopore DNA virus metagenomics using cell cultures and calves experimentally infected with bovine herpes virus-1

Bovine respiratory disease (BRD) is the leading cause of morbidity and mortality in cattle in Ireland, and internationally. This disease is caused by many well-known, and an ever-increasing number of newly associated viruses and bacteria. Consequently, diagnosis of BRD pathogens by targeted real-time polymerase chain reaction (qPCR) diagnostics is too expensive and slow to enable a same-day response that is targeted at the causative pathogen(s). To address this, we developed a same-day, sample to result, untargeted metagenomic MinION sequencing protocol for the identification of DNA viruses associated with BRD from nasal swabs. The procedure comprises non-viral nucleic acid depletion, nucleic acid extraction, rapid transposase-based tagmentation with barcoded adapters, non-biased PCR amplification of tagmented nucleic acid, sequencing on a MinION device, then rapid analysis of resulting sequences on cloud-based software EPI2ME WIMP. The protocol was developed using BoHV1-infected foetal lung cell cultures where we achieved 96% enrichment of the BoHV-1 sequence. Subsequently, the protocol was successfully applied to untargeted detection of BoHV-1 in nasal swabs from calves experimentally challenged with BoHV-1.


Introduction
Bovine respiratory disease (BRD) is the leading cause of morbidity, mortality and economic loss in cattle of all ages both in Ireland, and internationally [1][2][3][4][5] . The extensive use of vaccines against BRD-associated viral and bacterial pathogens has not reduced the incidence or severity of BRD in cattle. Consequently, large quantities of antimicrobials are still used for therapeutic treatment of BRD in Europe and the US 6,7 . Easy and economical methodologies that enable the rapid and reliable on-farm detection of viral and bacterial pathogens has the potential to increase effective use of vaccines by informing production and appropriate use of relevant up to date vaccines.
New pathogens e.g. influenza D virus (IDV) 11 , and Sneathia amnii 12 are continually added to the list of BRD aetiology. Viruses are known to initiate the disease by weakening the animal's defences which commonly leads to secondary bacterial infection 13 .
A diagnosis of BRD is generally based on clinical signs which are assessed by visual inspection of the animal, measurement of the rectal temperature and pulmonary auscultation 2,3,14,15 . If identification of the causative pathogen/s is attempted (which is often not the case), a nasal swab from the affected animal is sent to a centralised laboratory where targeted qPCR diagnostics of only four or five of the most likely previously mentioned bacterial or viral pathogens is conducted 16,17 . As there are at least 40 possible bacterial and viral pathogens associated with BRD, it is too expensive and time consuming to test for all of these by targeted real-time PCR (qPCR) diagnostics. Consequently, there are considerable delays in receiving the results of aetiological diagnosis of BRD cases which are often inconclusive 18 . For pathogen identification to be of direct practical use in preventing a BRD outbreak, results would need to be available to a vet within 24 h. Therein, current methods used for BRD-associated pathogen identification fall far short of what is required.
The Oxford Nanopore Technologies (ONT) MinION DNA/RNA sequencer is a portable sequencing device that allows real-time data analyses 19 and has the potential to enable untargeted same-day (sample to result) viral diagnostics. Sequencing all of the nucleic acid in a sample potentially allows detection of all organisms, including pathogens that are present in a single assay. The MinION sequencer uses a flowcell comprising a membrane containing approximately 1200-1600 active biological nanopores. Less expensive flowcells are also available that contain approximately 80-160 active nanopores. The membrane is flanked on either side by opposing electrical charges which drive negatively charged individual single chain strands of DNA or RNA through the nanopores towards the positive charge. A sensor registers the unique change of current produced by each individual base as it passes through the nanopore. These changes in current are translated into nucleotide sequence information by neural network basecallers 20 and approximately 4000 FASTQ files are generated every minute as the sequence run progresses 21 . As soon as these FASTQ files are generated they can be uploaded to a cloud-based software platform called Epi2ME which contains a number of intuitive point and click sequence analysis applications called 'work flows'. In the present work, the WIMP workflow was used as it identifies viral, bacterial, fungal and yeast sequence using 'Centrifuge' software. This means that a pathogen can, depending on its abundance in the samples, be identified within approximately 15 to 30 min of loading a sequencing library on a flowcell.
The size of most viral genomes is several orders of magnitude lower than those of bacteria and eukaryotes. In nasal swabs taken from cattle infected with a BRD-associated virus the vast majority of nucleic acid will therefore be prokaryotic or eukaryotic and just a fraction will be viral. As such, the current depth of nucleotide sequence achievable using next generation sequencing approaches is still insufficient to economically run multiple samples at the same time and reliably detect virus without prior enrichment of viral genomes.
Non-viral nucleic acid removal can be achieved by separation of eukaryote and prokaryote cells from the much smaller viral capsids by ultracentrifugation 22 . More recently it has been found that some giant viruses such as mimiviruses, which are associated with pneumonia in humans, are larger in size than some bacteria and pellet at low centrifugation speeds 23 . As intact viral capsids are nuclease resistant, RNaseA and DNase1 can be used to selectively digest non-viral nucleic acids. DNase1 and RNaseA are applied following cell disruption so that the eukaryotic and prokaryotic nucleic acids are exposed to the nucleases 24 .
However, in a cell infected with virus (i.e. virocell), much of the virus nucleic acid is not protected by a capsid and this unprotected viral sequence can also be lost if cell disruption and nuclease treatment are applied.
Following depletion of non-viral nucleic acid there is often insufficient total nucleic acid to generate enough sequencing library for NGS and TGS platforms, so following doublestranded cDNA synthesis, whole genome amplification (WGA) approaches are usually applied to amplify all of the remaining total nucleic acid in a depleted nucleic acid preparation. These approaches include Sequence-Independent, Single-Primer Amplification (SISPA) and Linker Amplified Shotgun Library (LASL) 25,26 , which both employ PCR, and isothermal multiple displacement amplification (MDA) using podovirus φ29 polymerase 26 .
Not surprisingly, each WGA method has been shown to preferentially amplify different families of viruses and MDA is prone to generation of chimeric sequence 26 .
The protocol we developed employs the LASL WGA approach. Compared to MDA and SISPA, LASL sequencing requires fewer reagents, thus lower cost, and fewer steps, so less time from sample to loading the flowcell. LASL simply comprises tagmentation of nucleic acid with a sequencing adapter, followed by a 70 minute, 30 cycle PCR amplification with barcoded primers.
BoHV-1 is an enveloped DNA virus and was selected for this study as it is of economic importance both to Ireland and internationally [27][28][29] . In addition, this virus has a well characterised and predictable pattern of infection. Firstly, non-viral nucleic acid depletion and subsequent non-biased LASL WGA of viral nucleic acids were optimised using bovine foetal lung cells (bFLC) infected with BoHV-1. These optimised procedures were then applied to nasal swabs collected from Holstein-Friesian calves that were experimentally challenged with BoHV-1.
The objective of the present study was to develop and validate, using Holstein-Friesian calves experimentally infected with a known virus, BoHV-1, a protocol for sameday, sample to result, untargeted identification of DNA viruses in nasal swabs from cattle using the portable MinION sequencer and Epi2ME cloud based software. Nasal swabs are the most common sample collected for the identification of pathogens associated with BRD outbreaks.

Results and Discussion
Optimisation of untargeted MinION sequencing of DNA viruses using in-vitro bovine foetal lung cells infected with BoHV-1

Non-viral nucleic acid depletion
Initially, optimisation of the non-viral nucleic acid depletion step was conducted using aliquots of the same batch of BoHV-1-infected bFLC culture. These aliquots were subjected to five different combinations of bead beating and nuclease treatment (A-E) (Fig. 1). Nucleic acids were then extracted and qPCR of the bACTB gene and UL27 gene was used to calculate the copy number of both the bovine and the BoHV-1 genomes respectively (Fig. 1). The most effective method identified for the depletion of non-viral nucleic acid comprised of an initial bead beating treatment, followed by a single treatment with RNaseA, then two 30 min incubation with Turbo DNase. However, whilst treatment E (bead-beating combined with single RNaseA and double DNase treatment) depletion resulted in the greatest reduction of bovine DNA, it also resulted in the greatest loss of viral nucleic acid. This could be due the exposure of non-capsid viral nucleic acid from infected bovine cells (i.e. virocells) to nucleases following disruption of the cells by bead beating. The number of BoHV-1 genome copies detected were also reduced approximately 20 fold for treatment B (bead beating, no nuclease) compared to treatment A (no bead beating, no nuclease) ( Figure 1). This suggests that bead beating alone leads to a considerable reduction in the amount of BoHV-1 DNA that can be subsequently be detected by qPCR. Bead beating may fragment some of the exposed viral DNA to sizes too small to be recovered by the QIAamp UltraSens Virus Kit or amplified by the UL27 primers. In order to achieve high sensitivity of BoHV-1 detection on the MinION it was necessary to increase the ratio of viral to non-viral nucleic acid even if that meant losing BoHV1 DNA.

Effect of non-viral nucleic acid depletion on nanopore sequencing
The effect of non-viral nucleic acid depletion treatment E ( Fig. 1) (bead beating and nuclease treatment) on nucleotide sequencing was then assessed. For this, non-barcoded PCR-free libraries were generated using the field sequencing library preparation kit (LRK001) from nucleic acid extracted from non-depleted samples (no bead beating or nuclease treatment) and depleted samples (bead beating and nuclease treatment) prior to extraction. The samples were either BoHV-1-infected bFLC in vitro cultures or a nasal swab from a BoHV-1-infected animal. These libraries were sequenced on a MinION (one library per R9 flowcell) and generated nucleotide sequences were analysed using the Epi2ME WIMP workflow. As expected, the percentage of reads that were assigned to viruses was significantly increased in the depleted cell culture libraries (mean = 96.94%, SD ±0.12, n = 3) compared to nondepleted cell culture libraries (mean = 45.61%, SD ±0.82, n = 3) ( Table 1). The percentage of viral reads was also dramatically increased in the depleted nasal swab library (12.91%) compared to the non-depleted (0.42%) nasal swab library (Table 1). However, the nondepleted cell culture libraries had higher viral read counts (mean = 22,626, SD ±8,888, n = 3) than the depleted cell culture libraries (mean = 4,164, SD ±1108, n = 3). Whereas the nondepleted library prepared from the swab had lower viral read counts (657) than the depleted library made from the same swab (1,903  We also tested if it was possible to reduce the time of the PCR extension step for each PCR cycle with the Q5 polymerase. The 5 minute extension step recommended by ONT for LongAmp Taq resulted in a PCR amplification step that took 4 hours and 10 min. Extension times tested for Q5 were 5 min (as recommended by ONT), 3 min, 2 min, and 40 s.
Unexpectedly, the longest reads were obtained with 40 s of extension, with an average read length of 1,400 bp. The reduction of the PCR extension step to 40 s led to a reduction of the overall 30 cycle PCR amplification to 80 min.

PBSBoHV-1 and PBS challenged calves
From the day prior to challenge to the sixth day post-infection, rectal temperatures and clinical signs were recorded for each calf. Rectal temperatures increased from day 2 postchallenge in calves challenged with BoHV-1 whereas rectal temperatures did not increase in the PBS challenged control calves (Fig. 3).

Figure 3. Rectal temperatures of BoHV-1 experimentally infected calves and PBS challenged
(control) calves from day -1 to day 6 relative to the challenge on day 0.

MinION sequencing of nasal swabs calves challenged with BoHV-1
Nasal swabs were collected from the six BoHV-1 challenged calves and six control (PBS challenged) calves from day -1 to day 6 relative to the challenge. Each swab was diluted in 3 mL of PBS and a 1 mL aliquot of this was used for depleted nucleic acid extraction for sequencing and library preparation. Nucleic acid extraction and library preparation were performed in batches comprising the 8 nasal swabs that were collected from each animal (one swab per day) plus a clean swab as negative control. Each batch of libraries was run on a separate flowcell. The libraries were sequenced on the MinION device attached to a MinIT for 24 h using rapid base calling and FASTQ files were uploaded to Epi2ME WIMP for taxonomic assignment.
BoHV-1 reads were identified by Epi2ME WIMP in all sequence libraries that had been generated from nasal swabs taken from the BoHV-1 challenge group from day 1 to day 6 post infection. BoHV-1 was not detected in any of the nasal swabs from calves challenged with PBS (Table 3). BoHV-1 was not detected on day -1 and 0 in four of the six calves challenged BoHV-1. One or two reads were identified as BoHV-1 for day -1 for one calf from the BoHV-1 challenge group, and for day 0 for two calves from the BoHV-1 challenge group. One or two reads were also identified as BoHV-1 in the negative extraction control that was included in the batch of extractions from those same two BoHV-1 challenged animals (BoHV1_1 and BoHV1_2). This could have resulted from either leaky barcodes during sequencing or cross contamination during sample processing.
Calf no.

Detection of viruses other than BoHV-1 in nasal swabs from BoHV-1 calf challenge model
For each swab, many single reads were assigned to viruses and phages other than BoHV-1 including bacteriophages and eukaryotic viruses (Supplementary Table S1). Where one or two reads were assigned to each of these eukaryotic viral taxa they were possibly a result of incorrect assignment of bovine, fungal or yeast genomes which were present in large amounts in these nasal swabs or inaccurate submissions to the RefSeq data base employed by WIMP.
Incorrect assignment in WIMP due to inaccurate submissions to RefSeq have been reported previously 31 . However, this was not checked in the current work. Bacteriophage taxonomic assignments were not surprising given the large numbers of bacteria in these nasal swab samples. More than 100 reads were assigned to Proteus phage VB_PmiS-Isfahan, Acinetobacter phage YMC13/03/R2096, Bubaline alphaherpesvirus and Bovine Alphaherpes Virus5 in some swabs. Proteus phage VB_PmiS-Isfahan seemed to be common in the nasal passages of these animals. Reads were also assigned to alphaherpes virus taxa other than BoHV-1 such as BoHV-5. These only occurred in animals that were challenged with BoHV-1 indicating that they were missassigned BoHV-1 sequences. We used rapid FASTQ basecalling in MinKnow software as our aim was to go from sample to result in a single day.
Accurate FASTQ base calling may have reduced the number of non-BoHV-1 viral reads but it was too slow for the same day protocol so we did not test it.

Conclusion
A 'same-day-sample-to-result' untargeted sequencing protocol to identify BRDassociated DNA viruses in nasal swabs was developed using the MinION nanopore sequencer and Epi2ME WIMP software. This protocol, which could be carried out within a six hour time frame, allowed correct identification of BoHV-1 in nasal swabs collected from calves that were experimentally infected with BoHV-1. The portability of the MinION nanopore sequencer means this protocol has potential for point-of-care viral pathogen testing in cattle.

Bovine foetal lung cells infected with BoHV-1
Bovine foetal lung cells were isolated from a bovine foetus and used as the in-vitro preparation for viral infection. The source, origin, and characteristics of these bFLC are shown in Supplementary Table S2 mL aliquots of the supernatant were transferred to sterile 1.5 mL microfuge tubes (Eppendorf, Hamburg, Germany) and frozen at -80°C.

Experimental calves
All animal experiments were carried out in accordance with the UK Animals (Scientific Procedures) Act 1986 and with the approval of the Agri-Food and Biosciences Institute Northern Ireland Ethical Review Committee. The study is reported in accordance with ARRIVE guidelines (https://arriveguidelines.org).
As part of a larger study, 12 Holstein-Friesian bull male calves (mean age 21