Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

The Poultry-Associated Microbiome: Network Analysis and Farm-to-Fork Characterizations

  • Brian B. Oakley ,

    brian.oakley@ars.usda.gov

    Affiliation Poultry Microbiological Safety, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • Cesar A. Morales,

    Address Current address: Kimberly-Clark Corporate Research & Engineering, Roswell, Georgia, United States of America

    Affiliation Poultry Microbiological Safety, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • J. Line,

    Affiliation Poultry Microbiological Safety, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • Mark E. Berrang,

    Affiliation Bacterial Epidemiology and Antibiotic Resistance Research Unit, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • Richard J. Meinersmann,

    Affiliation Bacterial Epidemiology and Antibiotic Resistance Research Unit, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • Glenn E. Tillman,

    Address Current address: USDA FSIS OPHS Eastern Laboratory, Athens, Georgia, United States of America

    Affiliation Poultry Microbiological Safety, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • Mark G. Wise,

    Address Current address: BioMerieux, Centre Christophe Merieux, Grenoble, France

    Affiliation Poultry Microbiological Safety, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • Gregory R. Siragusa,

    Address Current address: Dupont, Waukesha, Wisconsin, United States of America

    Affiliation Poultry Microbiological Safety, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • Kelli L. Hiett,

    Affiliation Poultry Microbiological Safety, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

  • Bruce S. Seal

    Affiliation Poultry Microbiological Safety, USDA Agricultural Research Service, Richard B. Russell Agricultural Research Center, Athens, Georgia, United States of America

The Poultry-Associated Microbiome: Network Analysis and Farm-to-Fork Characterizations

  • Brian B. Oakley, 
  • Cesar A. Morales, 
  • J. Line, 
  • Mark E. Berrang, 
  • Richard J. Meinersmann, 
  • Glenn E. Tillman, 
  • Mark G. Wise, 
  • Gregory R. Siragusa, 
  • Kelli L. Hiett, 
  • Bruce S. Seal
PLOS
x

Abstract

Microbial communities associated with agricultural animals are important for animal health, food safety, and public health. Here we combine high-throughput sequencing (HTS), quantitative-PCR assays, and network analysis to profile the poultry-associated microbiome and important pathogens at various stages of commercial poultry production from the farm to the consumer. Analysis of longitudinal data following two flocks from the farm through processing showed a core microbiome containing multiple sequence types most closely related to genera known to be pathogenic for animals and/or humans, including Campylobacter, Clostridium, and Shigella. After the final stage of commercial poultry processing, taxonomic richness was ca. 2–4 times lower than the richness of fecal samples from the same flocks and Campylobacter abundance was significantly reduced. Interestingly, however, carcasses sampled at 48 hr after processing harboured the greatest proportion of unique taxa (those not encountered in other samples), significantly more than expected by chance. Among these were anaerobes such as Prevotella, Veillonella, Leptrotrichia, and multiple Campylobacter sequence types. Retail products were dominated by Pseudomonas, but also contained 27 other genera, most of which were potentially metabolically active and encountered in on-farm samples. Network analysis was focused on the foodborne pathogen Campylobacter and revealed a majority of sequence types with no significant interactions with other taxa, perhaps explaining the limited efficacy of previous attempts at competitive exclusion of Campylobacter. These data represent the first use of HTS to characterize the poultry microbiome across a series of farm-to-fork samples and demonstrate the utility of HTS in monitoring the food supply chain and identifying sources of potential zoonoses and interactions among taxa in complex communities.

Introduction

As the source of a majority of emerging infectious diseases, animal-associated microbiomes represent a nexus of food safety, animal health, and public health [1], [2], [3]. The poultry microbiome is of particular interest as a major source of foodborne infections both worldwide and in the U.S., where foodborne pathogens are estimated to have accounted for 47 million episodes of illness at an economic cost of $77 billion in 2011 [4], [5]. In the course of commercial poultry production, chickens typically progress in a single age cohort from a microbiologically depauperate gastrointestinal (GI) tract in newly-hatched chicks to market age adults at 6–8 weeks harbouring up to 1011 bacterial cells g-1 of intestinal contents [6], representing hundreds [7], [8] to thousands [9] of distinct taxa. The poultry GI microbiome differs ecologically from mammals in that colonization occurs primarily from the surrounding environment and individuals of the same age reared in close proximity rather than through direct contact with adults [10], [11], [12], [13], [14], [15]. Colonization of poultry by microbes from environmental sources associated with commercial poultry production may have important biosecurity and management implications if human pathogens are transferred from environmental reservoirs through the poultry supply chain to consumers. This potential is recognized in the recently implemented Food Safety Modernization Act which emphasizes prevention of foodborne illness via monitoring of the entire food supply chain [16], and thus serves as a compelling mandate for a microbial census along what has been termed the ‘farm-to-fork’ continuum.

Detection and enumeration of poultry-associated bacteria at various stages of commercial poultry production have been the subject of several decades of research, generally focused on known pathogens such as Campylobacter. For example, transmission of Campylobacter between sequential flocks has been demonstrated [17], [18], , and specific Campylobacter genotypes present on farms have been linked to human illness [21], [22]. Although such research has answered important questions, pathogen surveys have typically relied either on cultivation of a specific organism of interest or molecular assays designed from strains already in cultivation - effectively limiting such surveys to the small proportion of microbial diversity recoverable by cultivation [23]. High-throughput sequencing (HTS) offers a solution to this previously intractable problem by rapidly taking a census of a microbial community independently of the ability to culture resident organisms. Although the exact magnitude of taxonomic richness that is actually measured by current sequencing technologies is still a matter of some debate [24], [25], [26], the fact that only a small proportion (<1%) of microbial taxa are cultivable as first noted by Staley and Konopka [27] remains true today. For human pathogens, fewer than half of clinically important pathogens are thought to have been positively identified by cultivation [2], [28], and unknown agents are estimated to cause four times as many episodes of foodborne illness as 31 major known pathogens [29]. HTS is an important new tool to identify emerging pathogens and explore previously uncharacterized microbial habitats.

As sequencing throughput has continued to increase and costs decrease in recent years, an opportunity has been created to greatly expand on previous characterizations of the poultry gut microbiome [30], [31], [32], compare communities, and identify specific taxa throughout the farm-to-fork continuum of poultry rearing and consumption. HTS-based characterizations of microbial communities along the poultry supply chain will help to provide a baseline census, document the dynamics of known pathogens, identify emerging pathogens, and help target specific interventions to improve animal and human health. HTS data sets are also highly amenable to network analysis which can offer novel insights into community composition, taxonomic interactions, and ecological niche associations in complex microbial communities [33], [34], [35].

In the current study, we combined 454 and Illumina sequencing of 16S rRNA gene amplicons with quantitative-PCR focused on putative pathogenic organisms and virulence genes to compare bacterial community structure and abundance of specific pathogens from poultry fecal samples, litter samples, carcass rinses, and carcass weeps, including the metabolically active bacterial community from fresh poultry products purchased at retail. For a subset of the samples, two flocks were followed longitudinally from rearing through processing. Our main objectives were to answer the following questions: 1) Which taxa are shared between bacterial communities in the litter of typical commercial poultry houses and the birds themselves? 2) How similar are these communities to those associated with chicken carcasses following commercial processing? 3) How does the carcass-associated community change during shelf-storage conditions? 4) what bacterial taxa are associated with poultry products as typically sold to the consumer?, and 5) what taxa, if any, are significantly associated with the foodborne pathogen Campylobacter?

Materials and Methods

Sample Collection

Longitudinal samples.

Initial collections of litter and fecal samples were made in October 2010 and February 2011 from two commercial broiler chicken production houses in the southeastern U.S., each containing a single flock of ca. 16,000 birds (Ross x Hubbard) six weeks of age. The flocks were selected to represent typical industry practices which include the administration of sub-therapeutic amounts of antibiotic growth promoters (AGPs). Litter (pine shavings) and fresh fecal droppings were collected to be representative of the entire flock by collecting from 25 locations within a semi-regular grid covering each poultry house. Samples were pooled into five sets of five, diluted 1∶3 (w:v) with 1× PBS, homogenized for 1 min, and 400 µl of the filtrate from sterile blending bags with full-width membrane filters (Model R36840-56, Cole Parmer, Vernon Hills, IL) used for DNA extractions as described below. Litter was collected as dry samples with ambient moisture content and wet samples saturated with water from directly underneath drip lines. Samples were stored at 4°C and transported to the laboratory and processed within 6 hours of sampling. DNA extractions were performed on each of the five sub-samples according to the protocol for pathogen detection from human stool with the QIAamp DNA Stool Kit (Qiagen, Valencia, CA, USA). For each sample type (fecal, wet litter, dry litter), the five sub-samples were normalized and pooled after DNA quantification to form a single sample for each poultry house.

Two weeks after initial sample collection, at ca. 8 weeks of age, these same flocks were commercially processed per typical industry practices including the immersion of carcasses in chlorinated water as the final step in processing. Following processing, 25 carcasses were collected from the chlorinated chill tank, aseptically bagged, transported to the lab, and shaken in an automated shaker [36] for 2 min after addition of 100 mL of sdH2O. The resulting fluid (hereafter referred to as ‘carcass rinse’) was collected aseptically. After remaining in the same bags for 48 hr at 4°C, an additional 50–75 mL (hereafter ‘carcass weep’) was collected from each carcass. For both carcass rinse and weep samples, samples from all 25 birds were pooled in sets of five, concentrated to ca. 2 mL by centrifugation and DNA extractions performed as above with the QIAamp DNA Stool Kit (Qiagen, Valencia, CA, USA) using 200 µL of the cell suspension. For each sample type (carcass rinse, carcass weep), the five sub-samples were normalized and pooled as above to form a single sample.

Retail samples.

To characterize the microbial community with potential metabolic activity (defined by the presence of intact rRNA molecules) in poultry products as typically purchased by consumers, poultry packages were purchased at retail in April 2012 and total nucleic acids extracted using the Griffiths method [37]. Aqueous exudate was collected directly from six packages of uncooked chicken parts (drumsticks, thighs and breasts); each package included multiple parts from chicken flocks processed in different slaughter plants on different dates from three different producers. From each package, 3 mL of exudate (hereafter ‘retail weep’) was collected, filtered through a coarse membrane filter as above, and centrifuged at 10,000 g for 10 min. Nucleic acid extractions were re-suspended in 35 µL of DEPC-treated TE, and incubated with DNase (Promega M6101) at 37°C for 30 min followed by a 10 min denaturation at 65°C per manufacturer’s recommendations. Subsequently, 2.75 µL of this reaction was incubated with 5 pmol of the primer 530R (5′- CCGCNGCNGCTGGCAC - 3′) at 70°C for 5 min and reverse-transcribed by incubation with random hexamers and reverse transcriptase (Invitrogen Superscript II) at 42°C for 50 min per manufacturer’s recommendations. PCR was performed with the primers 104F (5′-CCGCNGCNGCTGGCAC-3′) and 530R targeting the V1–V3 region of the 16S rRNA gene as previously described [8]; PCR products representing retail weep fluid from six packages were pooled for sequencing. For all amplifications of cDNA, corresponding negative results were confirmed for DNase treated samples without reverse transcription.

Samples for network analysis.

Samples collected for the longitudinal study described in detail above were combined with samples previously collected by our research group to build an adequate data set for network analysis. The previously collected samples were comprised of cecal contents (pooled from five individuals each) from 1, 2, and 3 week-old birds as described [38]; cecal samples from 3 week-old birds [39]; and carcass rinses collected during commercial processing as previously described [40]. Each sample was screened to ensure it was Campylobacter positive.

No specific permits were required for the described studies. Verbal permission to collect samples for this study was obtained from the commercial owners of private property in Georgia. Sample collection did not involve endangered or protected species. Additional details for all samples are provided in Table S1.

Sequencing and Data Analysis

PCR and 454 pyrosequencing were performed by Research and Testing Laboratory (Lubbock, TX) and USDA Agricultural Research Service (Athens, GA) using the same primers as above and tagged amplicon methods as previously described [8], [41]. Sequences were processed in mothur [42] using quality files with a 50 bp moving window at an average quality score cutoff of 35 and a maximum number of homopolymers below eight. Further processing per recent recommendations [24], [25], [43] and standard protocols (http://www.mothur.org/wiki/Schloss_SOP) was completed using Perl and Bioperl scripts to trim pyrosequencing tag sequences, screen for presence of the forward PCR primer sequence, and remove sequences with any ambiguous base calls. Based on expected amplicon sizes and frequency distributions of sequence lengths in v104 of the Silva reference database [44], sequences were further limited to a range of 325–425 bp. Sequences which passed these screens were then aligned to the Silva reference dataset using PyNAST [45] implemented in QIIME [46] and putative chimeric sequences were identified with ChimeraSlayer in mothur [42].

Illumina HiSeq2000 sequencing was performed using two flow cell lanes to generate paired-end reads from PCR products of the 16S rRNA V3 region obtained with the primers 341F (5′- CCTACGGGNGGCWGCAG-3′) and 519R (5′- ATTACCGCGGCTGCTGG-3′). from the 14 cecal and carcass rinse samples collected previously [38], [39],[40] as detailed above and in Table S1. Paired-end reads were merged with flash [47] and processed with the fastx toolkit [48] in a command line implementation with default settings. After random sub-sampling of 20,000 sequences per sample, the merged and quality-trimmed reads were trimmed to between 160 and 190 bp in length. The effects of the shorter Illumina reads on taxonomic classification and clustering was determined by trimming 454 reads to the same region (160–190 bp upstream of the 341F primer) and comparing to untrimmed reads of the same set of sequences. The misclassification rate determined by this approach was consistent with previous observations [49], and well within the acceptable range for the analysis presented here, as no Campylobacter sequences were misclassified. Rarefaction curves for all samples at each of three similarity cutoffs are shown in Figure S1.

Taxonomic classifications were based on the EMBL taxonomy from the Silva project (v104) curated seed database [44] using usearch [50] with the global alignment option. To assess phylotype richness and diversity independent of taxonomic classifications, sequences which passed all the screens described above were grouped into similarity clusters (operational taxonomic units; OTUs), using similarity cutoffs of 90%, 95%, and 97%. As a methodological comparison, both CD-HIT [51] and usearch [50] were used, both run with default parameters. Using control data sets derived from pyrosequencing of single colonies, we have previously shown CD-HIT to be a much more conservative clustering algorithm than approaches based on distance-matrices derived from multiple sequence alignments [41], which are known to falsely inflate richness and diversity estimates [25]. The output from CD-HIT and usearch provided the inputs for a data analysis pipeline we constructed to parse the clustering results, provide input for mothur [42], and produce graphical and statistical summaries of the data for the desired sampling units using R [52]. More information and open-source code can be found at http://go.warwick.ac.uk/thermophyl/pipeline. Sequence data have been deposited in GenBank with BioSample accession numbers: SAMN01853131–SAMN01853156 and MG-RAST as ID 4511219.3–4511244.3.

Network analysis was conducted as previously described [34] using normalized OTU tables at various levels of clustering and removing OTUs or taxa represented by fewer than five sequences or <0.5% total relative abundance across all samples. Spearman correlation coefficients of 0.7 and p-values of 0.001 were required to establish valid co-occurrence among OTUs. Network analysis was performed in R with the igraph package and visualized with the program Gephi.

Quantitative PCR

Quantitative PCR for 16S rRNA genes was performed with SYBR green chemistry (ABI, Foster City, CA, USA) using primer sets previously validated as specific for Clostridial Group I [53] and the genus Campylobacter [7] with thermal cycling protocols as previously described [53]. PCR products were obtained from Clostridium perfringens ATCC 13124 and Campylobacter jejuni NCTC 11168 and purified with Qiaquick spin columns (Qiagen, Valencia, CA, USA). The purified products were subsequently cloned into vector pCR4-TOPO using the TOPO TA Cloning Kit (Invitrogen, Carlsbad, CA, USA) following manufacturer’s recommendations and used as standards for absolute quantification for each respective assay.

To target C. jejuni specifically, a new probe and primer set targeting the beta-subunit of the cytolethal distending toxin gene (cdtB) was designed with the program ThermoPhyl [54] using publicly available sequences retrieved with BioPerl from Genbank, and assay sensitivity and specificity were confirmed with ARB [55]. ABI TaqMan chemistry was used with standards as above and universal cycling conditions per manufacturer’s recommendations. For all reactions, R2 values of standard curves were at least 0.98 and reaction efficiency was between 90–110%.

Results and Discussion

Longitudinal Farm-to-fork Data Set

Comparisons of core microbiome.

To address the potential for microbes from the poultry-rearing environment to be transferred through the poultry supply chain to consumers, we followed two flocks of ca. 16,000 birds each from the farm through poultry processing, collecting samples from the chickens and their immediate environments. These samples included fecal, litter, and carcass rinse and weep samples as detailed in the methods section and Table S1. As a simplified conceptual model of community interactions, we hypothesized that a large proportion of taxa would be shared between the litter and the fecal samples, and additionally, that the majority of sequences recovered from carcass rinses and subsequent weeps should represent a subset of those present in the litter and fecal samples (Figure 1A). These hypotheses were based on the following observations (reviewed in [12], [14], [56]: 1) newly-hatched chicks have a naïve intestinal community, 2) chickens are hunt-and-peck feeders, frequently contacting their litter, 3) chickens are coprophagous, 4) litter from previous flocks with a well-developed microbial community is commonly re-used, and 5) poultry houses are confined environments in which birds are exclusively reared after ca. 2 d of age until harvest 6–8 weeks later. Additionally, our fecal samples were collected as fresh droppings - because they did have some contact with the litter, overlap between the fecal and litter communities should be expected. Based on all of these factors, we expected to observe well-mixed microbial communities with many taxa shared across sample types.

thumbnail
Figure 1. Overview of sample types collected and similarity of samples across the farm-to-fork continuum.

A) Sample types collected with a simplified model of microbial community interactions. B) Non-metric dimensional scaling representation of overall community similarity of sample types based on 3% OTU abundance. Samples represent two farms (F2, F3) with litter samples split into dry and wet litter. C) Number of shared and unique OTUs between the fecal and litter microbiomes at a 3% cutoff. D) Shared and unique 3% OTUs among fecal and litter samples, carcass rinse, and carcass weep samples. For both (C) and (D), Venn diagrams are scaled according to OTU membership with singleton and doubleton OTUs excluded. Line drawing is from http://www.allaboutdrawings.com/bird-drawings.html.

http://dx.doi.org/10.1371/journal.pone.0057190.g001

Contrary to these expectations, neither hypothesis above was well supported. UniFrac analysis rejected a null hypothesis of no phylogenetic clustering by sample type (p<0.001), and fecal and litter samples generally clustered with each other as distinct from the rinse and weep samples in ordinations representing overall community similarity (Figure 1B). Moreover, each of the fecal, litter, carcass rinse, and carcass weep samples contained more unique OTUs than OTUs shared with any other sample type, even when uncommon sequence types were excluded from consideration (Figure 1C, 1D). Poultry-associated microbial communities along the farm-to-fork continuum clearly have major differences in community structure. We expected to find differences in relative abundance of various taxa among the sample types, but the extreme differences observed for taxon presence/absence were surprising. Ecological mechanisms such as dispersal limitation, differences in colonization ability, habitat filtering, niche exclusion, or competition may contribute in varying degrees to the observed differences in community structure and some (such as niche exclusion) may offer potential to suppress specific human pathogens at various stages of the poultry production process.

To determine in more detail the nature of the shared (core) versus unique (satellite) bacterial community by sample type, we next performed a taxonomic classification of sequences against a reference database using both blastn and usearch as described in the methods. The two methods gave very similar results with only minor discrepancies (data not shown). We concluded that either method is appropriate although usearch has a significant advantage of speed (in our tests, nearly 300× faster than blastn run locally). At a species level, a core set of 52 taxa were common to all sample types (Figure 2A). Within this core group, the most abundant sequence types in the fecal sample were most closely related to the anaerobic gram-negative Fusobacterium, gram-positive Actinomycete genera such as Brevibacterium, Bacteroides, and the Firmicute genera Clostridium, Faecalibacterium, and Lactobacillus (Figure 2B), consistent with previous observations [9], [31], [57], [58], [59]. For the wet litter samples, the relative abundance of these taxa was fairly similar to the fecal samples, with the exception of higher proportions of sequences most closely related to Staphylococcus and sequences most closely related to Shigella (Figure 2C). Shigella has not generally been considered to be associated with poultry [60], but our results are consistent with other observations [32], [61] and may expand the list of potential poultry-associated pathogens.

thumbnail
Figure 2. Core microbiome as classified to the species level with usearch against the Silva reference database as described in the methods.

A) Number of shared and unique taxa for each of the five sample types shown. The relative abundance of the 52 taxa belonging to a core microbiome common to all samples is shown for B) Feces, C) Wet litter, D) Dry litter, E) Carcass Rinses, and F) Carcass Weeps.

http://dx.doi.org/10.1371/journal.pone.0057190.g002

In contrast to the general similarity of the fecal and wet litter samples, one of the dry litter samples was unusual in its domination by sequences most closely-related to C. jejuni (Figure 2D). Campylobacter has been shown to quickly lose viability when inoculated onto fresh, dry litter [62], and although DNA-based PCR can amplify 16S rRNA genes from dead cells, it is difficult to argue against C. jejuni as abundant in the dry litter either at the time of sampling or in the relatively recent past given the predominance of these sequences in this sample type (Figure 2D). Core taxa detected in the carcass rinse and weep samples were generally similar in their relative abundance (Figure 2E, 2F). These two sample types were distinguished from the fecal and litter samples by several relatively abundant taxa, notably Gallibacterium anatis, strains of which have been shown to cause several diseases in poultry including peritonitis, salpingitis, and septicemia [63].

Comparisons of satellite microbiome.

Comparisons of the satellite microbiome – defined as taxa found only in a single sample type – revealed that the most dramatic difference occurred in the carcass weep which had many more unique taxa than expected by chance (p<0.0001; chi-square test). Of 41 taxa accounting for >0.1% of sequences and found only in a single sample type, 33 of these were unique to the carcass weep (Figure 3A–F). Dominant taxa in the weep samples included the anaerobes Prevotella, Veillonella, and Leptotrichia commonly present in the oral microbiome of humans [64], [65] as well as cats and dogs [66], and the aerobe Neisseria subflava, shown to dominate experimental biofilms [67]. Of particular relevance to food-safety concerns among taxa uniquely present in the weep were two species of Campylobacter – C. consisus, and C. showae (Figure 3F). Current dogma generally holds that poultry-associated Campylobacters are C. jejuni or C. coli, but our results show multiple sequence types present at low relative abundance more closely-related to C. consisus and C. showae than any known C. jejuni sequence. Deep sequencing of the poultry microbiome may challenge existing dogma which is largely based on cultivation with media selective for C. jejuni or C. coli [8].

thumbnail
Figure 3. Taxa unique to each sample type as classified to the species level with usearch against the Silva reference database as described in the methods for taxa representing greater than 0.1% of all sequences.

A) Number of shared and unique taxa for each of the five sample types shown. The relative abundance of the taxa unique to each sample type is shown for B) Fecal samples (blue), C) Wet litter (none), D) Dry litter (none), E) Carcass Rinses (black), and F) Carcass Weeps (red).

http://dx.doi.org/10.1371/journal.pone.0057190.g003

The discovery of taxa unique to the carcass weep samples is particularly interesting as, barring contamination from external sources, taxa present at this stage along the farm-to-fork continuum should logically represent a subset of those encountered in other sample types, particularly the carcass rinse samples. Although we can infer from richness estimates that some low-abundance OTUs have not been recovered (Table S1), incomplete sampling does not adequately explain the differences observed for the carcass weep (Figure 3F), as the taxa uniquely present are relatively abundant sequence types (0.1–8% of the total). Given roughly equivalent sampling effort, these taxa should have been encountered in the other sample types had they been present at similar population sizes. Although we did not sample all components of the poultry microbiome (e.g. skin, feathers, etc.), the carcass rinse and weep samples do represent directly paired samples taken from the same 25 carcasses. The relative abundance of taxa in the core microbiome was similar for the rinse and weep samples (Figure 2E–F), but nearly 2/3 of the taxa in the weep samples were not found in the rinse samples collected 48 hr earlier. These sequence types most likely represent taxa which persisted at low abundance through commercial anti-microbial treatment and were able to subsequently multiply during the 48 hr between the two sample collections.

Although it is common for commercial poultry processors in the U.S. to apply antimicrobial chemicals to carcasses during processing (for example, up to 50 ppm sodium hypochlorite in immersion chill tanks; www.fsis.usda.gov/OPPDE/rdad/FSISDirectives/7120.1.pdf), complete sterilization is not the goal and persistence of viable bacterial cells on carcasses following processing has been well documented. For example, cultivable Salmonella attached to chicken skin can be recovered after commercial processing [68], and incubations of whole carcasses in enrichment broth can recover bacteria not found by vigorous rinsing of the same carcasses [69]. Sequential carcass rinses have shown that viable bacteria can be continually recovered even after 40 rinses of the same carcass [70], and processed carcasses have been shown to carry the same subtypes of cultivable C. jejuni as present in a flock [71]. Our data are consistent with these observations and support a model in which a complex community of viable bacterial cells in and on a carcass can be transferred from the farm into the retail food chain despite current anti-microbial interventions utilized in poultry processing.

Detection and quantification of Clostridium and Campylobacter.

To follow specific foodborne pathogens along the food supply chain, we used qPCR to quantify the abundance of two of the most important pathogenic genera, Campylobacter and Clostridium, across our longitudinal samples using previously validated primer sets [7], [53]. Because of the importance of C. jejuni as a human pathogen, we also designed a novel TaqMan qPCR assay specific and sensitive for C. jejuni targeting the beta-subunit of the cytolethal distending toxin gene (cdtB). CDT is associated with C. jejuni pathogenesis, although knowledge of its exact role remains incomplete [72]; here we use it simply as a marker for cdt-positive C. jejuni. Of 18 C. jejuni genomes in the current version of the IMG database [73], 17 contain at least one cdt gene copy, and phylogenetic analysis of the assay designed here showed it to be highly specific and sensitive (Figure S2).

Clostridium was detected in all samples, but was significantly reduced (p<0.01, one-sided pairwise t test) in the weep sample relative to the other sample types (Figure 4A). For the genus-level Campylobacter assay, inter-farm variability was particularly high for the fecal and wet litter samples with 1.5–2 log differences between the two farms for these sample types. Importantly, significant reductions (p<0.004) were observed for the carcass rinse and carcass weep relative to the on-farm samples for both farms (Figure 4B). For the C. jejuni-specific cdtB assay, the results were quite similar, with absolute numbers reduced significantly (p<0.05) for the carcass rinse and weep samples compared to the on-farm sample types (Figure 4C). The qPCR results were consistent with the pyrosequencing data in which Campylobacter sequence types were only present at very low relative abundance in the rinse and weep samples, and with previous demonstrations of reductions in C. jejuni abundance and detection frequency after commercial anti-microbial treatments [74], [75].

thumbnail
Figure 4. Clostridium and Campylobacter abundance across sample types as determined by quantitative-PCR for Farm 2 (grey symbols) and Farm 3 (white symbols).

Number of 16S rRNA gene copies by sample type and farm for A) Clostridium group I assay, and B) Campylobacter genus-level-assay, and C) C. jejuni-specific cdtB assay. Horizontal bars represent means for sample types for all points within range. Letters denote statistical significance (p<0.05) of pairwise t tests performed in R with the Holm correction for multiple comparisons.

http://dx.doi.org/10.1371/journal.pone.0057190.g004

Bacterial Community Associated with Retail Products

In addition to reductions of Campylobacter, decreases in taxonomic richness in the rinse and weep samples demonstrated a community-level effect of residual anti-microbial treatments as applied in the chill tank. However, the uniqueness of the community in the retail weep samples as shown above (Figure 3), strongly suggested that a metabolically active community associated with poultry carcasses can persist through processing and pass into the retail food chain. If this is true, one would expect a complex community of metabolically active bacteria to be associated with poultry products packaged for retail sale. We explicitly tested this hypothesis by purchasing a variety of retail poultry products, extracting RNA from these samples, and performing PCR and 454 sequencing of amplicons generated from the resulting cDNA. Positive amplification from cDNA reverse-transcribed from rRNA with corresponding negative results for controls without reverse-transcriptase confirmed the presence of intact rRNA molecules, presumed to represent potentially metabolically active bacteria (Figure 5A). Following the same data processing steps described above, 10512 sequences were recovered from the DNA fraction, and 25165 sequences were obtained from the cDNA fraction. Taxonomic classification of these sequences showed the presence of 28 different genera in the retail weep fluid, of which 21 were judged potentially metabolically active as determined by sequence recovery from the cDNA fraction (Figure 5B). Interestingly, Campylobacter was not detected but the closely-related genus Arcobacter was present at low relative abundance in both fractions (Figure 5B). Campylobacter has been detected in surveys of retail poultry products at a prevalence of 20–100% of samples, depending on the method of recovery [76]. Of the 28 genera present in the retail weep samples, 20 were also present in the longitudinal data set (Figure 5B). Of the sequences in the retail weep, Pseudomonas was by far the most dominant genus, representing 98% of sequences from both the DNA and cDNA fractions (Figure 5B), with a high proportion of sequences most closely related to psychrophilic species such as P. psychrophila and P. antarctica (Figure 5C). These results are consistent with classification of Pseudomonas as a dominant poultry spoilage organism [77], and to our knowledge, represent the first HTS-based characterization of the microbiome associated with retail poultry products.

thumbnail
Figure 5. Characterization of bacterial community present in retail poultry products.

A) PCR results for amplification from total nucleic acids (lane 1), DNase treated reverse-transcriptase negative control (lane 2), and reverse-transcribed cDNA (lane 3). PCR reactions were performed with the broad-range primers 104F/530R as described in the text. Figure shows pooled products obtained from six retail packages of uncooked chicken parts; MW marker is NEB 100 bp ladder. B) Relative abundance of sequence types recovered from total nucleic acids (upper graph) as compared to sequence types present in the presumed potentially metabolically active fraction of the community (lower graph) classified to the genus level. Genera marked in bold (21 of 28) were designated potentially metabolically active as determined by sequence detection in the cDNA fraction. Genera marked with vertical arrows were also present in the carcass rinse or weep samples of the longitudinal data set. Genera marked with black circles were present in the fecal or litter samples of the longitudinal data set. Note break in Y axis for Pseudomonas which comprised 98% of both fractions. C) Comparison of species-level classification of sequences for Pseudomonas in DNA (upper graph) and cDNA (lower graph) fractions.

http://dx.doi.org/10.1371/journal.pone.0057190.g005

Network Analysis

We focused our network analysis on the relationship of Campylobacter to co-occurring taxa with several interesting results. First, when classified to the species level by comparison to the Silva reference database, the only significant linkage for any Campylobacter taxon was between C. jejuni and C. coli, neither of which had any significant associations with other taxa (Figure S3). This result was consistent even for relaxed cutoffs for statistical significance and Pearson correlation coefficients. This result may have important implications for Campylobacter competitive exclusion (CE) strategies which were first proposed in 1982 [78] but have had only limited success in reducing Campylobacter colonization in poultry [79]. If confirmed by additional sequencing of Campylobacter-positive samples, the limited efficacy of CE for Campylobacter may be explained by a lack of co-occurring taxa, which are presumed to occupy an ecological niche space overlapping with that of Campylobacter. To confirm that the inclusion of samples sequenced with the Illumina platform did not influence these results, these samples were excluded from the analysis, with no change in results.

To investigate this question with more discriminatory power, we next performed the same analysis using OTUs mapped to taxonomic classifications. Consistent with the previous analysis, most of the OTUs classified as Campylobacter had no significant associations with any other taxonomic group (Figure 6, orange clusters). Interestingly, however, two OTUs, including the 2nd most abundant Campylobacter OTU (Clstr97) were significantly co-associated with a suite of other taxa (Figure 6, yellow clusters). Included among these were OTUs most closely related to taxa such as the anaerobe Faecalibacterium prausnitzii, known to be abundant in the chicken cecum [59] and the human colon where it produces the short chain fatty acid butyrate and has been shown to employ a flavin-dependent electron transfer scheme to exploit oxic-anoxic boundaries [80]. Another sequence type significantly co-associated with these Campylobacter OTUs was Megamonas hypermegale which has been previously shown to be associated with Campylobacter exclusion in poultry using an elegant antibiotic selection scheme [81]. Interestingly, in this work, Scupham et al. [81] classified M. hypermegale sequences as either type I or II and found a significant correlation with C. jejuni suppression only from type I sequences. The two Megamonas OTUs (Clstr630 and Clstr2939; Figure 6) we found with significant association to Campylobacter were clearly more closely related to type II sequences from Scupham et al. (Figure S4). If Megamonas type I sequences actually do have a suppressive effect on C. jejuni, the predominance of type II sequences associated with C. jejuni sequence types in these samples may not be surprising.

thumbnail
Figure 6. Network analysis of Campylobacter OTUs within poultry-associated microbiomes.

A) Significant network associations for OTUs classified as C. jejuni and C. coli shown in detail, and B) in context of entire network. C) C. jejuni OTUs with connections only to other Campylobacter as shown in orange accounted for a majority of Campylobacter sequences. D) Two Campylobacter OTUs, Clstr3618 and Clstr97, shown in yellow and denoted by arrows, were connected to a much larger suite of taxa including Megamonas hypermegale and Faecalibacterium prausnitzii as discussed in the text.

http://dx.doi.org/10.1371/journal.pone.0057190.g006

Our results are consistent with previous demonstrations of intra-specific diversity of sequence types most closely related to Campylobacter [8] and provide further evidence that these sequence types have significantly different ecological niches. Taken together, our network analyses suggest that the limited efficacy historically demonstrated for CE of Campylobacter may be due to the variety of ecological niches inhabited by Campylobacter. These Campylobacter sequence types may represent novel diversity currently undescribed by cultivation; their associations to human illness and food safety remain an interesting and important topic for future research.

Conclusions

By exploiting the capacities of next-generation sequencing, the approach taken here provides a rapid method to characterize and compare poultry-associated microbial communities independent of biases associated with cultivation. High-throughput sequencing of 16S rRNA gene amplicons has become a common method to investigate microbial communities in complex samples [26], [82], [83], but to our knowledge, this study demonstrates the first use of HTS to characterize the poultry microbiome across a series of farm-to-fork samples.

Analysis of samples along the so-called ‘farm-to-fork’ continuum showed several genera containing recognized pathogens (Clostridium, Campylobacter, and Shigella) belonging to a core microbiome common to all sample types. Quantitative-PCR assays demonstrated that Clostridium abundance was significantly reduced in weep samples, while Campylobacter generically and C. jejuni specifically were significantly reduced in carcass rinse and weep samples relative to fecal and litter samples. The weep samples were the most unique, largely due to the presence of anaerobes such as Prevotella, Veillonella, Leptotrichia, C. concisus, and C. showae, suggesting that taxa present at low absolute and relative abundance are able to persist through poultry processing in a viable state. Network analysis revealed that most Campylobacter sequence types do not have significant associations with other taxa which may explain the historically poor efficacy of attempts to competitively exclude Campylobacter. Two C. jejuni sequence types did have significant linkages to a suite of other taxa, including Megamonas hypermagale, consistent with previous observations [81]. High-throughput sequencing provides a powerful tool to identify potential reservoirs of foodborne pathogens and analyse interactions within complex microbial communities.

Supporting Information

Figure S1.

Rarefaction curves by sample for OTU designations at 97%, 95%, and 90% similarity cutoffs defined with uclust as described in the text.

doi:10.1371/journal.pone.0057190.s001

(EPS)

Figure S2.

Taxonomic specificity and sensitivity of cdtB TaqMan qPCR assay. Maximum-likelihood tree contains all non-redundant publicly available sequences from the IMG database retrieved by annotations of ‘cdt’, ‘cytolethal distending toxin’, and IMG genome blast with default parameters. Sequences with perfect matches to the newly-designed assay are shown in red; all other sequences have >5 mismatches to each primer and probe.

doi:10.1371/journal.pone.0057190.s002

(EPS)

Figure S3.

Species level network analysis showing significant associations for Campylobacter only between C. coli and C. jejuni as shown in boxed area.

doi:10.1371/journal.pone.0057190.s003

(EPS)

Figure S4.

Representative sequences classified as Megamonas recovered by our sequencing compared to Megamonas sequences in Scupham et al. (2010) shown in bold. All sequences found here were classified as cluster II, not cluster I which were found to be significantly associated with C. jejuni suppression by Scupham et al. (2010).

doi:10.1371/journal.pone.0057190.s004

(EPS)

Table S1.

Sample sources, and taxonomic richness and diversity observed at a 95% clustering level for samples sequenced in this study.

doi:10.1371/journal.pone.0057190.s005

(DOCX)

Acknowledgments

We acknowledge the generous support of a commercial poultry processing facility, and the excellent technical assistance of Manju Amin, Susan Brooks, Johnna Garrish, Linda Genzlinger, Susan Mize, and Latoya Wiggins. Emily Hollister generously provided advice and analytical tools for network analysis.

Author Contributions

Conceived and designed the experiments: BBO CAM MEB RJM GET MGW GRS KLH JEL BSS. Performed the experiments: BBO CAM MEB GET MGW GRS. Analyzed the data: BBO. Contributed reagents/materials/analysis tools: BBO CAM RJM GRS KLH BSS. Wrote the paper: BBO.

References

  1. 1. Taylor LH, Latham SM, Woolhouse ME (2001) Risk factors for human disease emergence. Philos Trans R Soc Lond B Biol Sci 356: 983–989. doi: 10.1098/rstb.2001.0888
  2. 2. Woolhouse MEJ (2002) Population biology of emerging and re-emerging pathogens. Trends in Microbiology 10: S3–S7. doi: 10.1016/s0966-842x(02)02428-9
  3. 3. DuPont HL (2007) The growing threat of foodborne bacterial enteropathogens of animal origin. Clinical Infectious Diseases 45: 1353–1361. doi: 10.1086/522662
  4. 4. Scallan E, Hoekstra RM, Angulo FJ, Tauxe RV, Widdowson MA, et al. (2011) Foodborne illness acquired in the United States-major pathogens. Emerg Infect Dis 17: 7–15. doi: 10.3201/eid1701.09-1101p1
  5. 5. Scharff RL (2012) Economic Burden from Health Losses Due to Foodborne Illness in the United States. Journal of Food Protection 75: 123–131. doi: 10.4315/0362-028x.jfp-11-058
  6. 6. Gabriel I, Lessire M, Mallet S, FGuillot JF (2006) Microflora of the digestive tract: critical factors and consequences for poultry. World’s Poultry Science Journal 62: 499–511. doi: 10.1017/s0043933906001115
  7. 7. Oakley BB, Line JE, Berrang ME, Johnson JM, Buhr RJ, et al. (2012) Pyrosequencing-based validation of a simple cell-suspension polymerase chain reaction assay for Campylobacter with application of high-processivity polymerase and novel internal amplification controls for rapid and specific detection. Diagnostic Microbiology and Infectious Disease 72: 131–138. doi: 10.1016/j.diagmicrobio.2011.11.001
  8. 8. Oakley BB, Morales CA, Line JE, Seal BS, Hiett KL (2012) Application of high-throughput sequencing to measure the performance of commonly used selective cultivation methods for the foodborne pathogen Campylobacter. FEMS Microbiology Ecology 79: 327–336. doi: 10.1111/j.1574-6941.2011.01219.x
  9. 9. Danzeisen JL, Kim HB, Isaacson RE, Tu ZJ, Johnson TJ (2011) Modulations of the Chicken Cecal Microbiome and Metagenome in Response to Anticoccidial and Growth Promoter Treatment. PLoS One 6.
  10. 10. Hald B, Skovgard H, Bang DD, Pedersen K, Dybdahl J, et al. (2004) Flies and Campylobacter infection of broiler flocks. Emerging infectious diseases 10: 1490–1492. doi: 10.3201/eid1008.040129
  11. 11. Kwan PS, Barrigas M, Bolton FJ, French NP, Gowland P, et al. (2008) Molecular epidemiology of Campylobacter jejuni populations in dairy cattle, wildlife, and the environment in a farmland area. Applied and Environmental Microbiology 74: 5130–5138. doi: 10.1128/aem.02198-07
  12. 12. Newell DG, Fearnley C (2003) Sources of Campylobacter colonization in broiler chickens. Applied and Environmental Microbiology 69: 4343–4351. doi: 10.1128/aem.69.8.4343-4351.2003
  13. 13. Jacobs-Reitsma WF, van de Giessen AW, Bolder NM, Mulder RW (1995) Epidemiology of Campylobacter spp. at two Dutch broiler farms. Epidemiology and infection 114: 413–421. doi: 10.1017/s0950268800052122
  14. 14. Lee MD, Newell DG (2006) Campylobacter in poultry: filling an ecological niche. Avian diseases 50: 1–9. doi: 10.1637/7474-111605r.1
  15. 15. Skov MN, Spencer AG, Hald B, Petersen L, Nauerby B, et al. (2004) The role of litter beetles as potential reservoir for Salmonella enterica and thermophilic Campylobacter spp. between broiler flocks. Avian Diseases 48: 9–18. doi: 10.1637/5698
  16. 16. Rittenberry R (2011) Farm-to-fork safety. The new federal food safety law aims to prevent contamination rather than simply respond to it. Occupational health & safety 80: 66–68.
  17. 17. Alter T, Weber RM, Hamedy A, Glunder G (2011) Carry-over of thermophilic Campylobacter spp. between sequential and adjacent poultry flocks. Veterinary Microbiology 147: 90–95. doi: 10.1016/j.vetmic.2010.06.005
  18. 18. Hiett KL, Stern NJ, Fedorka-Cray P, Cox NA, Musgrove MT, et al. (2002) Molecular subtype analyses of Campylobacter spp. from Arkansas and California poultry operations. Applied and Environmental Microbiology 68: 6220–6236. doi: 10.1128/aem.68.12.6220-6236.2002
  19. 19. Shreeve JE, Toszeghy M, Pattison M, Newell DG (2000) Sequential spread of Campylobacter infection in a multipen broiler house. Avian Diseases 44: 983–988. doi: 10.2307/1593076
  20. 20. Zweifel C, Scheu KD, Keel M, Renggli F, Stephan R (2008) Occurrence and genotypes of Campylobacter in broiler flocks, other farm animals, and the environment during several rearing periods on selected poultry farms. International Journal of Food Microbiology 125: 182–187. doi: 10.1016/j.ijfoodmicro.2008.03.038
  21. 21. Callicott KA, Hargardottir H, Georgsson F, Reiersen J, Frigriksdottir V, et al. (2008) Broiler Campylobacter contamination and human campylobacteriosis in Iceland. Applied and Environmental Microbiology 74: 6483–6494. doi: 10.1128/aem.01129-08
  22. 22. Damjanova I, Jakab M, Farkas T, Meszaros J, Galantai Z, et al. (2011) From farm to fork follow-up of thermotolerant campylobacters throughout the broiler production chain and in human cases in a Hungarian county during a ten-months period. International Journal of Food Microbiology 150: 95–102. doi: 10.1016/j.ijfoodmicro.2011.07.011
  23. 23. Rappe MS, Giovannoni SJ (2003) The uncultured microbial majority. Annual Review of Microbiology 57: 369–394. doi: 10.1146/annurev.micro.57.030502.090759
  24. 24. Huse SM, Welch DM, Morrison HG, Sogin ML (2010) Ironing out the wrinkles in the rare biosphere through improved OTU clustering. Environ Microbiol 12: 1889–1898. doi: 10.1111/j.1462-2920.2010.02193.x
  25. 25. Kunin V, Engelbrektson A, Ochman H, Hugenholtz P (2010) Wrinkles in the rare biosphere: pyrosequencing errors can lead to artificial inflation of diversity estimates. Environ Microbiol 12: 118–123. doi: 10.1111/j.1462-2920.2009.02051.x
  26. 26. Sogin ML, Morrison HG, Huber JA, Mark Welch D, Huse SM, et al. (2006) Microbial diversity in the deep sea and the underexplored “rare biosphere”. Proc Natl Acad Sci U S A 103: 12115–12120. doi: 10.1073/pnas.0605127103
  27. 27. Staley JT, Konopka A (1985) Measurement of in situ activities of nonphotosynthetic microorganisms in aquatic and terrestrial habitats. Annu Rev Microbiol 39: 321–346. doi: 10.1146/annurev.mi.39.100185.001541
  28. 28. Jones KE, Patel NG, Levy MA, Storeygard A, Balk D, et al. (2008) Global trends in emerging infectious diseases. Nature 451: 990–U994. doi: 10.1038/nature06536
  29. 29. Scallan E, Griffin PM, Angulo FJ, Tauxe RV, Hoekstra RM (2011) Foodborne Illness Acquired in the United States-Unspecified Agents. Emerging Infectious Diseases 17: 16–22. doi: 10.3201/eid1701.p21101
  30. 30. Lu J, Idris U, Harmon B, Hofacre C, Maurer JJ, et al. (2003) Diversity and succession of the intestinal bacterial community of the maturing broiler chicken. Appl Environ Microbiol 69: 6816–6824. doi: 10.1128/aem.69.11.6816-6824.2003
  31. 31. Qu A, Brulc JM, Wilson MK, Law BF, Theoret JR, et al. (2008) Comparative metagenomics reveals host specific metavirulomes and horizontal gene transfer elements in the chicken cecum microbiome. PLoS One 3: e2945. doi: 10.1371/journal.pone.0002945
  32. 32. Sekelja M, Rud I, Knutsen SH, Denstadli V, Westereng B, et al. (2012) Abrupt temporal fluctuations in the chicken fecal microbiota are explained by its gastrointestinal origin. Appl Environ Microbiol 78: 2941–2948. doi: 10.1128/aem.05391-11
  33. 33. Foster JA, Krone SM, Forney LJ (2008) Application of ecological network theory to the human microbiome. Interdisciplinary perspectives on infectious diseases 2008: Article ID 839501.
  34. 34. Barberan A, Bates ST, Casamayor EO, Fierer N (2012) Using network analysis to explore co-occurrence patterns in soil microbial communities. ISME J 6: 343–351. doi: 10.1038/ismej.2011.119
  35. 35. Faust K, Raes J (2012) Microbial interactions: from networks to models. Nature Reviews Microbiology 10: 538–550. doi: 10.1038/nrmicro2832
  36. 36. Dickens JA, Cox NA, Bailey JS, Thomson JE (1985) Automated microbiological sampling of broiler carcasses. Poultry Science 64: 1116–1120. doi: 10.3382/ps.0641116
  37. 37. Griffiths RI, Whiteley AS, O’Donnell AG, Bailey MJ (2000) Rapid method for coextraction of DNA and RNA from natural environments for analysis of ribosomal DNA- and rRNA-based microbial community composition. Applied and Environmental Microbiology 66: 5488–5491. doi: 10.1128/aem.66.12.5488-5491.2000
  38. 38. Wise MG, Siragusa GR (2007) Quantitative analysis of the intestinal bacterial community in one- to three-week-old commercially reared broiler chickens fed conventional or antibiotic-free vegetable-based diets. J Appl Microbiol 102: 1138–1149. doi: 10.1111/j.1365-2672.2006.03153.x
  39. 39. Tillman GE, Haas GJ, Wise MG, Oakley B, Smith MA, et al.. (2011) Chicken intestine microbiota following the administration of lupulone, a hop-based antimicrobial. FEMS Microbiol Ecol.
  40. 40. Line JE, Oakley BB, Stern NJ (2013) Comparison of cumulative drip sampling with whole carcass rinses for estimation of Campylobacter species and quality indicator organisms associated with processed broiler chickens. Poultry science 92: 218–224. doi: 10.3382/ps.2012-02217
  41. 41. Oakley BB, Carbonero F, Dowd SE, Hawkins RJ, Purdy KJ (2011) Contrasting patterns of niche partitioning between two anaerobic terminal oxidizers of organic matter. ISME J.
  42. 42. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, et al. (2009) Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Appl Environ Microbiol 75: 7537–7541. doi: 10.1128/aem.01541-09
  43. 43. Huse SM, Huber JA, Morrison HG, Sogin ML, Welch DM (2007) Accuracy and quality of massively parallel DNA pyrosequencing. Genome Biol 8: R143. doi: 10.1186/gb-2007-8-7-r143
  44. 44. Pruesse E, Quast C, Knittel K, Fuchs BM, Ludwig W, et al. (2007) SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB. Nucleic Acids Res 35: 7188–7196. doi: 10.1093/nar/gkm864
  45. 45. Caporaso JG, Bittinger K, Bushman FD, DeSantis TZ, Andersen GL, et al. (2010) PyNAST: a flexible tool for aligning sequences to a template alignment. Bioinformatics 26: 266–267. doi: 10.1093/bioinformatics/btp636
  46. 46. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, et al. (2010) QIIME allows analysis of high-throughput community sequencing data. Nature Methods 7: 335–336. doi: 10.1038/nmeth.f.303
  47. 47. Magoc T, Salzberg SL (2011) FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 27: 2957–2963. doi: 10.1093/bioinformatics/btr507
  48. 48. Blankenberg D, Gordon A, Von Kuster G, Coraor N, Taylor J, et al. (2010) Manipulation of FASTQ data with Galaxy. Bioinformatics 26: 1783–1785. doi: 10.1093/bioinformatics/btq281
  49. 49. Soergel DAW, Dey N, Knight R, Brenner SE (2012) Selection of primers for optimal taxonomic classification of environmental 16S rRNA gene sequences. ISME J 6: 1440–1444. doi: 10.1038/ismej.2011.208
  50. 50. Edgar RC (2010) Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26: 2460–2461. doi: 10.1093/bioinformatics/btq461
  51. 51. Li W, Godzik A (2006) Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences. Bioinformatics 22: 1658–1659. doi: 10.1093/bioinformatics/btl158
  52. 52. R DCT (2010) R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.
  53. 53. Rinttila T, Kassinen A, Malinen E, Krogius L, Palva A (2004) Development of an extensive set of 16S rDNA-targeted primers for quantification of pathogenic and indigenous bacteria in faecal samples by real-time PCR. J Appl Microbiol 97: 1166–1177. doi: 10.1111/j.1365-2672.2004.02409.x
  54. 54. Oakley BB, Dowd SE, Purdy KJ (2011) ThermoPhyl: a software tool for selecting phylogenetically optimized conventional and quantitative-PCR taxon-targeted assays for use with complex samples. FEMS Microbiol Ecol 77: 17–27. doi: 10.1111/j.1574-6941.2011.01079.x
  55. 55. Ludwig W, Strunk O, Westram R, Richter L, Meier H, et al. (2004) ARB: a software environment for sequence data. Nucleic Acids Res 32: 1363–1371. doi: 10.1093/nar/gkh293
  56. 56. Hermans D, Pasmans F, Messens W, Martel A, Van Immerseel F, et al. (2012) Poultry as a Host for the Zoonotic Pathogen Campylobacter jejuni.. Vector-Borne and Zoonotic Diseases 12: 89–98. doi: 10.1089/vbz.2011.0676
  57. 57. Lu J, Sanchez S, Hofacre C, Maurer JJ, Harmon BG, et al. (2003) Evaluation of broiler litter with reference to the microbial composition as assessed by using 16S rRNA and functional gene markers. Appl Environ Microbiol 69: 901–908. doi: 10.1128/aem.69.2.901-908.2003
  58. 58. Zhu XY, Zhong T, Pandya Y, Joerger RD (2002) 16S rRNA-based analysis of microbiota from the cecum of broiler chickens. Appl Environ Microbiol 68: 124–137. doi: 10.1128/aem.68.1.124-137.2002
  59. 59. Lund M, Bjerrum L, Pedersen K (2010) Quantification of Faecalibacterium prausnitzii- and Subdoligranulum variabile-like bacteria in the cecum of chickens by real-time PCR. Poultry science 89: 1217–1224. doi: 10.3382/ps.2010-00653
  60. 60. Jay JM (2000) Modern Food Microbiology. Gaithersburg, MD: Aspen Publishers, Inc.
  61. 61. Hume ME, Barbosa NA, Dowd SE, Sakomura NK, Nalian AG, et al. (2011) Use of pyrosequencing and denaturing gradient gel electrophoresis to examine the effects of probiotics and essential oil blends on digestive microflora in broilers under mixed Eimeria infection. Foodborne Pathog Dis 8: 1159–1167. doi: 10.1089/fpd.2011.0863
  62. 62. Cox NA, Berrang ME, Stern NJ, Musgrove MT (2001) Difficulty in recovering inoculated Campylobacter jejuni from dry poultry-associated samples. J Food Prot 64: 252–254.
  63. 63. Bojesen AM, Nielsen OL, Christensen JP, Bisgaard M (2004) In vivo studies of Gallibacterium anatis infection in chickens. Avian Pathol 33: 145–152. doi: 10.1080/03079450310001652059
  64. 64. Ahn J, Yang L, Paster BJ, Ganly I, Morris L, et al. (2011) Oral microbiome profiles: 16S rRNA pyrosequencing and microarray assay comparison. PLoS One 6: e22788. doi: 10.1371/journal.pone.0022788
  65. 65. Yang F, Zeng X, Ning K, Liu KL, Lo CC, et al. (2012) Saliva microbiomes distinguish caries-active from healthy human populations. ISME J 6: 1–10. doi: 10.1038/ismej.2011.71
  66. 66. Abrahamian FM, Goldstein EJ (2011) Microbiology of animal bite wound infections. Clinical Microbiology Reviews 24: 231–246. doi: 10.1128/cmr.00041-10
  67. 67. Bradshaw DJ, Marsh PD, Allison C, Schilling KM (1996) Effect of oxygen, inoculum composition and flow rate on development of mixed-culture oral biofilms. Microbiology-UK 142: 623–629. doi: 10.1099/13500872-142-3-623
  68. 68. Lillard HS (1993) Bactericidal effect of chlorine on attached Salmonellae with and without sonication. Journal of Food Protection 56: 716–717.
  69. 69. Simmons M, Fletcher DL, Berrang ME, Cason JA (2003) Comparison of sampling methods for the detection of Salmonella on whole broiler carcases purchased from retail outlets. Journal of Food Protection 66: 1768–1770.
  70. 70. Lillard HS (1989) Factors affecting the persistance of Salmonella during the processing of poultry. Journal of Food Protection 52: 829–832.
  71. 71. Simmons M, Hiett KL, Stern NJ, Frank JF (2008) Comparison of poultry exudate and carcass rinse sampling methods for the recovery of Campylobacter spp. subtypes demonstrates unique subtypes recovered from exudate. Journal of Microbiological Methods 74: 89–93. doi: 10.1016/j.mimet.2008.03.007
  72. 72. Young KT, Davis LM, DiRita VJ (2007) Campylobacter jejuni: molecular biology and pathogenesis. Nat Rev Micro 5: 665–679. doi: 10.1038/nrmicro1718
  73. 73. Mavromatis K, Chu K, Ivanova N, Hooper SD, Markowitz VM, et al. (2009) Gene context analysis in the Integrated Microbial Genomes (IMG) data management system. PLoS One 4: e7979. doi: 10.1371/journal.pone.0007979
  74. 74. Stern NJ, Fedorka-Cray P, Bailey JS, Cox NA, Craven SE, et al. (2001) Distribution of Campylobacter spp. in selected US poultry production and processing operations. Journal of Food Protection 64: 1705–1710.
  75. 75. Berrang ME, Dickens JA (2000) Presence and level of Campylobacter spp. on broiler carcasses throughout the processing plant. Journal of Applied Poultry Research 9: 43–47. doi: 10.1093/japr/9.1.43
  76. 76. Jacobs-Reitsma WF (2000) Campylobacter in the food supply. In: Nachamkin I, Blaser MJ, editors. Campylobacter. Washington, D.C.: ASM Press.
  77. 77. Mead GC (2000) Fresh and further-processed poultry. In: Lund BM, Baird-Parker TC, Gould GW, editors. The microbiological safety and quality of food. Gaithersburg, MD: Aspen Publishers, Inc.
  78. 78. Soerjadi AS, Snoeyenbos GH, Weinack OM (1982) Intestinal colonization and competitive exclusion of Campylobacter fetus subsp. jejuni in young chicks. Avian Diseases 26: 520–524. doi: 10.2307/1589897
  79. 79. Stern NJ (1994) Mucosal competitive exclusion to diminish colonization of chickens by Campylobacter jejuni. Poultry Science 73: 402–407. doi: 10.3382/ps.0730402
  80. 80. Khan MT, Duncan SH, Stams AJ, van Dijl JM, Flint HJ, et al. (2012) The gut anaerobe Faecalibacterium prausnitzii uses an extracellular electron shuttle to grow at oxic-anoxic interphases. ISME J 6: 1578–1585. doi: 10.1038/ismej.2012.5
  81. 81. Scupham AJ, Jones JA, Rettedal EA, Weber TE (2010) Antibiotic manipulation of intestinal microbiota to identify microbes associated with Campylobacter jejuni exclusion in poultry. Appl Environ Microbiol 76: 8026–8032. doi: 10.1128/aem.00678-10
  82. 82. Hamady M, Walker JJ, Harris JK, Gold NJ, Knight R (2008) Error-correcting barcoded primers for pyrosequencing hundreds of samples in multiplex. Nat Methods 5: 235–237. doi: 10.1038/nmeth.1184
  83. 83. Fierer N, Hamady M, Lauber CL, Knight R (2008) The influence of sex, handedness, and washing on the diversity of hand surface bacteria. Proc Natl Acad Sci U S A 105: 17994–17999. doi: 10.1073/pnas.0807920105