Rapid Oxford Nanopore Technologies MinION Sequencing Workflow for Campylobacter jejuni Identification in Broilers on Site—A Proof-of-Concept Study

Simple Summary Current culture methods for Campylobacter enumeration in poultry meat require at least five days. To improve this outcome, we have assembled a portable sequencing kit consisting of the Bento Lab and the MinION and developed a workflow for on-farm use that can detect and report the presence of Campylobacter in less than five hours from sampling to data. In addition, this workflow allows for the determination of microbiota profiles within those five hours. Overall, this workflow and approach can be helpful in the control of zoonotic agents, with a particular focus on poultry meat, where it takes less than 24 h from the time the animals arrive at the slaughterhouse to the time the carcasses are available on the supermarket shelves. Abstract Campylobacter is recognised as one of the most important foodborne bacteria, with a worldwide health and socioeconomic impact. This bacterium is one of the most important zoonotic players in poultry, where efficient and fast detection methods are required. Current official culture methods for Campylobacter enumeration in poultry usually include >44 h of culture and >72 h for identification, thus requiring at least five working shifts (ISO/TS 10272-2:2017). Here, we have assembled a portable sequencing kit composed of the Bento Lab and the MinION and developed a workflow for on-site farm use that is able to detect and report the presence of Campylobacter from caecal samples in less than five hours from sampling time, as well as the relationship of Campylobacter with other caecal microbes. Beyond that, our workflow may offer a cost-effective and practical method of microbiologically monitoring poultry at the farm. These results would demonstrate the possibility of carrying out rapid on-site screening to monitor the health status of the poultry farm/flock during the production chain.


Introduction
Since 2005, Campylobacter has been the most commonly diagnosed zoonotic pathogen in the European Union (EU), and one of the major causes of foodborne illness worldwide [1,2]. Poultry is considered the main reservoir of the bacteria, with the consumption of contaminated and/or inadequately prepared poultry products as the main source of outbreaks in humans [1, 3,4]. Thus, in January 2018, the EU established Regulation (EC) No 1495/2017, to set a process hygiene criterion for Campylobacter control at slaughterhouses, and began to monitor the bacterial burden, with the limit of 1000 CFU/g of neck skin. Despite this control, approximately 220,000 cases of Campylobacter from meat consumption continue to occur every year [1]. The use of prevalence target C. jejuni models

Campylobacter Jejuni Infection
At day 11 days old, 20% of birds/pen were orally infected with C. jejuni. The strain was selected from a database of Campylobacter strains isolated in a longitudinal study of the whole poultry production cycle (breeders and their progeny) in the Valencia region (Eastern Spain) [38]. To generate the inoculum, an overnight inoculum was diluted to obtain a wavelength of 0.2 (OD 600, 10 8 CFU/mL). Then, tenfold dilutions were made until a concentration of 10 6 CFU/mL was achieved, and animals were infected with 100 µL (10 5 CFU/bird).

Sampling Collection
The Campylobacter status of the chicken houses was tested before the arrival of the animals taken from environmental samples with sterile swabs from the floor and walls from six different points of the house. All collected samples were analyzed within 24 h according to 10272-2:2017.
To assess Campylobacter status of the animals upon arrival at the farm, 10 birds were randomly selected and euthanised. After necropsy, the pair of ceca were removed and placed in an individual sterile jar. Caeca samples were pooled into a composite sample for the detection of Campylobacter. On days 14 and 42, three birds per pen were randomly selected and euthanised. Individual caecum samples were obtained and placed in sterile pots. Briefly, the surface of each intact caecum was spray treated with 70% ethanol, as directed by Hansson et al. [39], before caeca contents were extracted. Then, the caeca content was removed and homogenized. To this end, the caeca content was placed in a sterile Petri dish and homogenized vigorously with a sterile swab until a homogenized product was obtained [40,41]. Then, pools of the three animals from the same pen were prepared and divided into two aliquots. One aliquot was analysed by the culture method and the other aliquot was analysed by 16S rDNA amplicon sequencing. Workflows for the ISO method and MinION nanopore sequencing performed are shown in Figure 1. the whole poultry production cycle (breeders and their progeny) in the Valencia region (Eastern Spain) [38]. To generate the inoculum, an overnight inoculum was diluted to obtain a wavelength of 0.2 (OD 600, 10 8 CFU/mL). Then, tenfold dilutions were made until a concentration of 10 6 CFU/mL was achieved, and animals were infected with 100 μL (10 5 CFU/bird).

Sampling Collection
The Campylobacter status of the chicken houses was tested before the arrival of the animals taken from environmental samples with sterile swabs from the floor and walls from six different points of the house. All collected samples were analyzed within 24 h according to 10272-2:2017.
To assess Campylobacter status of the animals upon arrival at the farm, 10 birds were randomly selected and euthanised. After necropsy, the pair of ceca were removed and placed in an individual sterile jar. Caeca samples were pooled into a composite sample for the detection of Campylobacter. On days 14 and 42, three birds per pen were randomly selected and euthanised. Individual caecum samples were obtained and placed in sterile pots. Briefly, the surface of each intact caecum was spray treated with 70% ethanol, as directed by Hansson et al. [39], before caeca contents were extracted. Then, the caeca content was removed and homogenized. To this end, the caeca content was placed in a sterile Petri dish and homogenized vigorously with a sterile swab until a homogenized product was obtained [40,41]. Then, pools of the three animals from the same pen were prepared and divided into two aliquots. One aliquot was analysed by the culture method and the other aliquot was analysed by 16S rDNA amplicon sequencing. Workflows for the ISO method and MinION nanopore sequencing performed are shown in Figure 1.

Culture Method
Enumeration of Campylobacter in samples was performed according to ISO/TS 10272-2:2017. As required by the ISO method, tenfold dilutions of samples on BPW were performed. Then, 0.1 mL from each inoculum was plated onto mCCDA (Modified Charcoal Cefoperazone Deoxycholate Agar, Oxoid, Dardilly, France) and incubated at 41.5 ± 1 • C in a micro-aerobic atmosphere (84% N 2 , 10% CO 2 , 6% O 2 ) for 44 h ± 4 h. For further confirmation analysis, five Campylobacter-like colonies were purified on blood agar (AES La-boratories ® , Bruz Cedex, France) at 41.5 ± 1 • C in a microaerobic atmosphere (84% N 2 , 10% CO 2 , 6% O 2 ) over 24 h. Then, colony morphology and motility under dark field microscopy were evaluated. Confirmation of presumptive Campylobacter colonies was assessed by oxidase and catalase tests and plating at different temperatures and atmospheres (25 • C under aerobic conditions and 41.5 • C under microaerophilic conditions for 24 h) in Columbia blood agar (AES Laboratories).

MinION Sequencing: Genomic DNA Extraction, Sequencing, and Analysis
On-farm DNA analysis was performed using a mobile device called Bento Lab, which combines a portable PCR machine, a microcentrifuge, gel electrophoresis, and a transillumina-tor [37]. Genomic DNA was extracted from 1 g of caecal content with the Wizard Genomic DNA Purification Kit (Promega, Madison, WI, USA). After extraction, DNA was purified with one volume of Agencourt AMPure XP beads (Beckman Coulter, Inc., L'Hospitalet de Llobregat Barcelona, Spain). The quality and average size of genomic DNA were visualised by gel electrophoresis with a 1% agarose gel. For microbial community analysis, 10 ng of genomic DNA was used as an input for the Nanopore 16S Barcoding Kit (SQK-RAB204), which amplified the 16S rDNA gene by PCR using specific 16S primers 27F and 1492R containing barcodes and 5 tags, following the manufacturer's instructions. The sequencing was performed in a portable Min-ION sequencing device, with FLO-MIN106 flow cell (ONT Research, Oxford, UK). A total of 12 samples were sequenced in a run. Raw reads were base called into fast5 files with the Albacore program version 2.1.7 (ONT Research, Oxford, UK). The fast5 files were converted to fastq format using poretools version 0.6.0 [43]. For taxonomy analyses, taxonomy was assigned to each individual sequence read using the RDP classifier (v2.13, Trainset No: 18, 07/2020) [44]. The statistical analyses were performed in R software version 4.1.1 (R Core Team 2018) using the gdata [45], vegan [46], tydyr (https://CRAN.R-project.org/package=tidyr, accessed on 1 April 2022) packages. Reads' time points were obtained from Nanopore fastq headers using the "start_time" field expressed as "YYYY-MM-DD HH:MM:SS". This field value was extracted from every read and converted on a second-scale count. A taxonomy contingency table was built every 10 min using in-house R scripts.
ClustVis software was used to perform the PCA [47]. InteractiVenn software was used for Venn diagram construction [48]. Differences in relative abundances of taxonomic units between rearing days (14 vs. 42 days) were estimated using Bayesian interference. The sampling time was included as a treatment with two levels (14 vs. 42 days). The dendrogram obtained by Bayesian interference was created by 60,000 interactions of Markov chain Monte Carlo, with a burn-in period of 10,000 and saving only 1 of every 10 samples for inference. The parameters obtained from the marginal posterior distributions of the relative abundance between groups were the mean of the difference (D [14][15][16][17][18][19][20][21][22][23][24] ; computed as , the probability of the difference being greater than 0 when D 14-24 > 0 or lower than 0 when D 14-24 < 0 (P 0 ), and the highest posterior density region at 95% of probability (HPD95%). D 14-24 estimated the mean of the differences between 14 and 42 traits, P 0 estimated the probability of D 14-24 = 0, and HPD95% estimated the accuracy. Statistical differences were considered if P 0 > 0.8 (80%). Statistical analyses were computed with the Rabbit program developed by the Institute for Animal Science and Technology (Valencia, Spain). To determine the shortest MinION sequencing time that could generate sufficient data for Campylobacter taxa prediction, raw reads generated after 360 min of sequencing were divided into 10 min batches using read timestamps included in the read's names during the data production per se. The generated and analyzed datasets are available at NCBI's BioProject PRJNA814618.

Sequencing Costs
We also calculated the sequencing costs associated with MinION to examine how they compared with the conventional culture. Our calculations did not consider the entry cost of sequencing hardware (i.e., the price of a MinION sequencer, EUR 900). For MinION sequencing, we assumed that reagents and flow cells were purchased separately because this provides a more meaningful gauge, as the MinION starter pack is usually a one-time purchase for most users. In addition, the personal labour cost was excluded.

Results
During this study, a total of 12 caecal pools (6 per sampling time) were collected and processed through microbiological culture and sequencing. No clinical signs were ob-served during rearing, and the productive parameters obtained were in accordance with the breed standards. Environmental samples and day-old chick samples were confirmed as Campylobacter negative. During rearing, all caeca samples were positive for Campylo-bacter with both methods (Table 1). For the ISO 1027-2:2017 method, starting at the enumeration to identification, a total of four days was necessary to obtain the result. All samples were positive, regardless of the sampling day, with the mean number of Campylobacter being 7.4 log CFU/g on day 14 and 6.7 log CFU/g on day 42 (Table 1). We found significant differences between 14 and 42 days of rearing in the number of Campylobacter ( Table 2). The complete MinION sequencing process, starting from the DNA extraction to the extraction to a minimum limit of 239,988 reads, lasted 10.5 h (average of 20,000 reads by sample, Figure S1). Read length ranged from 1542.00 ± 241.50 nt to 1563.02 ± 243.71 nt, with an average of 1553.55 ± 237.55 nt (mean ± SD). As in the ISO 1027-2:2017 method, the mean relative abundance for Campylobacter was conditional by sampling day (Table 2), with the relative abundance for Campylobacter being less on day 42. . P 0 = probability of the difference (D  ) being greater than 0 when D 14-42 > 0 or lower than 0 when D 14-42 < 0. HPD95% = the highest posterior density region at 95% of probability. Statistical differences were assumed if |D 14-42 | surpass the R value and its P 0 > 0.80.

Taxonomic Diversity
Alpha diversity was assessed through species evenness and species richness measures. Rarefaction analysis indicated that the number of amplicon sequence variants (ASVs) per sample reached a maximum, indicating that the sequencing effort captured the diversity of the majority of the samples. The observed sequences and Chao1 alpha diversity indexes revealed a significant difference between the caecal microbiota depending on the age of the animals (155.2 ± 5.6 vs. 213.83 ± 5.6 for observed sequences, p ≤ 0.01 and 193.2 ± 13.3 vs. 284.5 ± 13.3 for Chao1, p ≤ 0.01, for 14-and 42-day-old chicks, respectively). Finally, to assess differences in microbiota between rearing sample times, beta diversity was analysed. To this end, principal coordinate analysis (PCoA) was conducted based on Bray-Curtis dissimilarity, and the results are shown in Figure 2. In both 14-and 42-day-old chicks, the six replicates from each rearing day clustered together. These results showed differences in the microbial community composition among the two broiler-rearing times.

Taxonomic Diversity
Alpha diversity was assessed through species evenness and species richness measures. Rarefaction analysis indicated that the number of amplicon sequence variants (ASVs) per sample reached a maximum, indicating that the sequencing effort captured the diversity of the majority of the samples. The observed sequences and Chao1 alpha diversity indexes revealed a significant difference between the caecal microbiota depending on the age of the animals (155.2 ± 5.6 vs. 213.83 ± 5.6 for observed sequences, p ≤ 0.01 and 193.2 ± 13.3 vs. 284.5 ± 13.3 for Chao1, p ≤ 0.01, for 14-and 42-day-old chicks, respectively). Finally, to assess differences in microbiota between rearing sample times, beta diversity was analysed. To this end, principal coordinate analysis (PCoA) was conducted based on Bray-Curtis dissimilarity, and the results are shown in Figure 2. In both 14-and 42-day-old chicks, the six replicates from each rearing day clustered together. These results showed differences in the microbial community composition among the two broiler-rearing times.

Microbial Community Dynamics
The complete list of all bacterial taxa (average relative abundance of more than 2%) identified for each age group in the caecal content samples is provided in Table 3. Microbial taxa consistently present over time (core microbiome) represented by 22 genera (mainly belonging to the order Clostridiales, phylum Firmicutes) were identified in both sampling times with variations in the relative abundance ( Table 3). The highest count of Campylobacter was found in 14-day-old broilers (1.36%, p = 1.00, HPD 95% [0.79, 1.94]).

Microbial Community Dynamics
The complete list of all bacterial taxa (average relative abundance of more than 2%) identified for each age group in the caecal content samples is provided in Table 3. Microbial taxa consistently present over time (core microbiome) represented by 22 genera (mainly belonging to the order Clostridiales, phylum Firmicutes) were identified in both sampling times with variations in the relative abundance ( Table 3). The highest count of Campylobacter was found in 14-day-old broilers (1.36%, p = 1.00, HPD 95% [0.79, 1.94]).

Effect of Sequencing Time on Campylobacter spp. Identification
To investigate the potential robustness to detect Campylobacter spp. throughout the sequence time, we examined changes in the relative abundance of Campylobacter spp. every 10 min from 0 to 360 (Figure 3). There was a rapid increase in the relative abundance of Campylobacter spp. detection from 0 to 20 sequencing minutes, after which it was constant until 280 min, regardless of the rearing moment, with a relative abundance average of 1.78 ± 0.105% and 0.13 ± 0.037% for 14-and 42-day-old chick samples (mean ± standard deviation), respectively (Figure 3). After 320 min of sequencing, 14-day-old chick samples slightly increased the relative abundance of Campylobacter, while that of the 42-day-old chick samples slightly decreased.

Sequencing Costs
Overall, the per sample costs for MinION sequencing were found to be approximately eighteen-fold more expensive than the conventional culture method (Table 4), costing an estimated EUR 75 per sample considering 12 samples per run. This cost can be further reduced down to about EUR 41 per sample pooling up to 24 barcodes.

Sequencing Costs
Overall, the per sample costs for MinION sequencing were found to be approximately eighteen-fold more expensive than the conventional culture method (Table  4), costing an estimated EUR 75 per sample considering 12 samples per run. This cost can be further reduced down to about EUR 41 per sample pooling up to 24 barcodes.

Discussion
When broilers were artificially infected with Campylobacter jejuni, Firmicutes and Campilobacterota were the most dominant phyla in the caecal microbiota. Notably, the current study reports fast Campylobacter (and core microbiota) screening under farm conditions. The decision-action gap is crucial in intensive farming systems such as poultry production, where the high throughput of animal husbandry presents a high risk of developing and transmitting zoonotic agents [49][50][51]. Campylobacter infection is one of the

Discussion
When broilers were artificially infected with Campylobacter jejuni, Firmicutes and Campilobacterota were the most dominant phyla in the caecal microbiota. Notably, the current study reports fast Campylobacter (and core microbiota) screening under farm conditions. The decision-action gap is crucial in intensive farming systems such as poultry production, where the high throughput of animal husbandry presents a high risk of developing and transmitting zoonotic agents [49][50][51]. Campylobacter infection is one of the the last century's most widespread foodborne infectious diseases, and poultry has been considered the origin of 80% of human cases [52,53]. Thus, our study yields two main results. First, we can detect Campylobacter spp. in less than five hours, from sample collection to result, which has some important implications. Second, the workflow developed here shows remarkable potential for simultaneous detection of foodborne pathogens and provides an overview of the core microbiota that could be related at a given time to the health and welfare status of the animals. As on-farm detection represents the first step necessary to limit the introduction of Campylobacter into the food chain [54,55], our workflow can provide a cost-effective and practical method for on-farm real-time monitoring of microbial contamination of poultry. Although the costs of MinION sequencing seem high, they are generally inversely related to sample size [56]. One of the advantages of nanopore sequencing is that it is possible to stop sequencing as soon as sufficient data have been obtained and washed with a nuclease kit (100 euros for six washes), and the flow cell has been resued with other samples.
Based on the European Regulation, official Campylobacter detection in poultry must be carried out taking samples from neck skin during processing at the slaughterhouse and analysing them according to ISO 10272:2018 [53,[55][56][57]. The process involves a tenfold dilution of collected samples incubated at 41.5 • C in a microaerobic atmosphere for 48 h. Further characterisation was performed by morphology and motility and biochemical confirmation at different temperatures and atmospheres [57,58]. This technique makes these assays laborious, and they require several days of work (taking up to five days) [53,58,59]. The main problem is that the official samples are taken at the slaughterhouse, and the positive results have to be obtained as soon as possible to be able to make decisions about the destination of the flock [10,55]. It is important to note that, within just 24 h after the slaughter of the animals, the chickens are already on the supermarket shelves, ready to be consumed. Undoubtedly, it seems essential to assess the status of the flocks before they reach the slaughterhouse [30,[60][61][62][63][64]. This is why molecular methods for same-day diagnosis of pathogenic bacteria are increasingly being introduced into daily laboratory routines [53,65,66]. Overall, 100% of samples were positive for both methods, regardless of animal age, requiring a total of five days from enumeration to confirmation for ISO 1027-2:2017, while the sequencing method required only five hours between extraction and sequencing. It is notable that, regardless of the detection method, the colonisation of Campylobacter depended on the age of the bird after the infection challenge, with a relevant reduction observed in 42-day samples, in line with the literature [67][68][69].
Nanopore MinION has been proposed as a fast, efficient, sensitive, and cost-effective sequencing method capable of enabling the identification of pathogens based on the core microbiome and differentially abundant taxa at the field level [35,70]. In this study, the real-time availability of the data generated by MinION showed that, after 20 min of sequencing time, the relative abundance of Campylobacter remained constant up to 320 min, with a similar pattern observed at both rearing ages. In addition, this workflow makes it possible to analyse several samples at the same time in order to obtain a general overview of the microbiota, which could be useful for routine surveillance and early detection of foodborne pathogens [36]. Beyond that, as the microbiota influences various physiological and behavioural processes, a healthy microbiota would be correlated with an animal's fitness, making it possible to identify health and welfare deviations in the animals [71]. Though the workflow proposed here was based on data analysed after complete data collection, taxonomy sequences were obtained a posteriori, after the sequencing process finished, using per-sequence timestamps added by the nanopore sequencing system itself to each read. At the time of writing, the whole workflow can be easily transferred to an automated offline bioinformatics pipeline, as well as using the EPI2ME WIMP workflow: quantitative, real-time species identification from metagenomic samples. Taking our results together, the combination of mobile DNA barcoding with Bento Lab and MinION could be suitable for on-farm routine analysis of Campylobacter and other zoonotic pathogens, as the results can be obtained on the same day at the same site at which the samples have been taken [35,37].

Conclusions
Here, we proposed a MinION-based workflow for fast Campylobacter and other zoonotic pathogens' screening using a Bento Lab-MinION combination for portable se-quencing at the farm level. We have demonstrated that MinION-based workflow at the farm level is viable and highly accurate by comparing it with ISO/TS 10272-2:2017. We conclude that this method could provide a basis for future studies to help farmers implement active zoonotic bacterial screening in real-time diagnostics to take measures that will reduce the prevalence, which is indispensable for active surveillance.
Supplementary Materials: The following supporting information can be downloaded at https: //www.mdpi.com/article/10.3390/ani12162065/s1, Figure S1: Rarefaction curves for each sample, relating the number of amplicon sequence variants detected depending on the sequencing effort.