When bacteria become resistant to antibiotics, it becomes difficult or impossible to treat infections in both people and animals. Antibiotic resistance is a growing problem, and many fear a “post-antibiotic era” in which common infections become life threatening.

In order to slow the spread of antibiotic resistance, it is important to understand how and where this resistance develops. In general, using antibiotics increases the likelihood that bacteria will develop resistance. Therefore, locations where antibiotics are commonly used – such as hospitals, long-term care facilities, livestock facilities (such as feedlots) and crop production areas (such as orchards) – may help antibiotic resistance to develop and spread. However, it is largely unknown how much each location promotes the emergence of antibiotic-resistant bacteria.

Until recently, we have only been able to investigate how resistance develops in bacteria grown in a laboratory; or to look for a handful of specific resistance genes in a sample of bacteria collected from people, animals or the environment. Fortunately, a technology called next-generation sequencing now allows us to look at all the resistance genes within all the bacteria in a sample. This may help us to improve our understanding of how and where resistance develops and spreads.

Noyes et al. have now used next-generation sequencing to describe the antibiotic resistance potential (known as the “resistome”) found in various types of samples collected from feedlots and slaughterhouses involved in producing beef. This showed that the number of different resistance genes in the samples decreased while cattle were in the feedlot and during the slaughter process. Several groups of resistance genes that were detected when the cattle first arrived in the feedlot were not detected at all at the end of the feedlot period. However, some resistance genes were detected throughout the feedlot period, and these tended to be resistance genes that allow the bacteria to evade the same antibiotics that were used in the cattle. In addition, no resistance genes of any type were detected in the samples collected after the cattle had been slaughtered.

As well as providing insights into the resistome of beef production, Noyes et al.’s study also highlights the fact that we need to develop a deeper understanding of the data that come from next-generation sequencing. This may involve developing new laboratory techniques and creating new methods to analyze such data.

Food production and food products are important potential sources of antimicrobial resistant (AMR) infections in humans. Beef is a widely consumed protein commodity, and production and consumption is expected to increase in the United States and globally (Daniel et al., 2011; OECD/Food and Agriculture Organization of the United Nations, 2015). In North American beef production, several critically important antimicrobial drugs (AMDs) such as fluoroquinolones, macrolides and third-generation cephalosporins are used, while others are not, e.g., carbapenems (World Health Organization, 2011; Food and Drug Administration, 2003). Use of these AMDs is thought to increase the risk of AMR being transmitted to humans through environmental exposures (i.e., air, water and soil), occupational exposures (Levy et al., 1976; Moon et al., 2015), as well as through consumption of beef products (Antibiotic Resistance from the Farm to the Table [Internet], 2014). While surveillance for foodborne AMR pathogens has been part of North American food safety systems for decades (National Antimicrobial Resistance Monitoring System – Enteric Bacteria (NARMS), 2011; Canadian Integrated Program for Antimicrobial Resistance Surveillance (CIPARS), 2013), we have yet to fully understand and quantify the public health risk posed by transmission of non-pathogenic bacteria that carry antimicrobial resistance determinants (ARDs). These ARDs could pose a risk to human health if the bacteria carrying them become established within the microbiome of the human host, subsequently enabling horizontal gene transfer to pathogens (Forsberg et al., 2012; Rolain, 2013); or if these ARDs are present in opportunistic pathogens that become established within an immunocompromised individual. Establishment within an individual’s microbiome could occur either through the ingestion of contaminated food products or through exposure to environmental effluents disseminated from beef feedlots (Antibiotic Resistance from the Farm to the Table [Internet], 2014), i.e., a facility where cattle are aggregated, reared in pens (i.e., outdoor enclosures) and fed a high-energy ration before being slaughtered. The rate at which ARDs from beef products or production facilities become established within humans is unknown, largely due to an historical reliance on culture and isolation of pathogens and an inability to access the microbial community and its complete repertoire of ARDs (i.e., the resistome).

Several steps in the beef production system could play crucial roles in the transmission of AMR from beef production to humans via environmental interfaces and beef products. In North America, use of AMDs is much greater in feedlots than any other phase of beef production (Gow et al., 2008; Rao et al., 2010), a fact that has raised concerns that these operations could represent the principal phase of beef production in which AMR is acquired or maintained. Furthermore, feedlots are intricately linked to environmental exposure pathways such as air, manure, soil, and water, enabling indirect human exposure to feedlot effluents (McEachran et al., 2015). In North America, abattoirs (i.e., slaughterhouses) are a potential control point for the transmission of AMR , as they employ sequential antibacterial interventions to reduce pathogen contamination in beef products; testing of these interventions has demonstrated that they are effective in reducing not only pathogens, but also total bacterial contamination of beef products (Bacon et al., 2000). We hypothesized that the antimicrobial interventions and other procedures used in feedlots and abattoirs would exert a measurable effect on the presence, abundance and composition of ARDs in the bacterial populations of cattle, the feedlot environment and market-ready beef products. Furthermore, we hypothesized that the use of a metagenomics approach would enable us to quantify these changes at an ecological level and therefore better understand the risk to public health, compared to the use of a culture-based approach.

In order to understand how feedlots and abattoirs affect the transmission of ARDs , it is imperative to track pens of cattle through the beef production system (Figure 1), documenting AMD use and antimicrobial interventions and describing resistome changes over time. However, research in this area has been constrained by the challenges of tracking beef products and environmental effluents from individuals or pens of cattle, and collecting detailed records of AMD exposures for the cattle being studied. Specific challenges include lack of unique animal identification, use of non-computerized or hard-to-access AMD treatment records, effluents that are difficult to trace (e.g., air and runoff water), disassembly of carcasses into hundreds of non-linked parts, and the sheer difficulty of obtaining relevant, representative samples from feedlot steers, which weigh over 450 kg. Because of these complexities, studies in this area have been constrained to descriptions of AMR prevalence in isolated sectors of the beef production process without access to relevant AMD exposure data (Sheikh et al., 2012); or they have relied on the use of AMD and AMR data at very abstract levels such as the nation-state (Chantziaras et al., 2013). To our knowledge, no studies have specifically tracked antimicrobial use in cattle while investigating antimicrobial resistance in market-ready products or consumers. This dearth of evidence greatly complicates efforts to develop effective policies related to antimicrobial use in livestock with the goal of protecting public health (Landers et al., 2012). The objective of this study was to perform a prospective longitudinal analysis of antimicrobial use and resistance in beef production and to exploit shotgun metagenomics to characterize resistome dynamics in the environment and the products of cohort cattle from feedlot through to market-ready product.

Figure 1

Download asset Open asset

Figure 1—source data 1

Sample collection details, by location, sample matrix and pen.

https://doi.org/10.7554/eLife.13195.004

Download elife-13195-fig1-data1-v1.xlsx

A convenience sample of beef feedlots in Texas (n=2) and Colorado (n=2) was selected for study enrollment based on their willingness to participate. Capacity in these feedlots was 69,000, 73,000, 74,000, and 98,000 cattle, which is typical for large U.S. feedlots. While small feedlots (defined as feeding <100 cattle) comprise >60% of the ~26,500 feedlot operations in the U.S., large feedlots (defined as feeding >1000 cattle) contain >75% of all cattle that are reared in feedlots in the U.S. (Census of Agriculture, 2012). Animal handling and management procedures in study feedlots were typical for large feedlot operations in North America, and included daily observations of all pens by feedlot personnel trained to identify clinically ill cattle. Sick cattle were moved from their home pen to the cattle handling chute and/or a hospital pen to receive treatment. Rations in these 4 feedlots were corn-based and conformed to the National Research Council diet requirements for beef cattle (National Research Council, 2000). In each feedlot, 2 pens of cattle were enrolled in the study. Because cattle are typically raised in feedlots for at least 5 months, pens of newly arrived cattle are not always available for study participation. Therefore, study pens were a convenience selection from all pens that had been most recently filled to capacity at or near the start of the study. Feedlot owners and managers were aware of the identity of enrolled pens, but personnel involved in managing the cattle on a daily basis were not and therefore it is unlikely that standard operating procedures were altered during the study period. Pen capacities ranged from 150 to 281 cattle, and a total of 1741 cattle were housed in these 8 pens. Antimicrobial exposures were recorded throughout the feeding period, i.e., the total time that animals were housed in the feedlot, which ranged from 117 to 227 days (see Figure 1—source data 1). This duration is typical for North American beef cattle that enter feedlots at this age/weight. Analysis of these records indicates that all cattle received in-feed macrolides (tylosin), but use of parenteral AMDs was relatively infrequent at the individual animal level (Table 1). All pens contained at least one animal that was exposed to individual doses of parenteral tetracyclines and macrolides.

In order to assess the resistome throughout the feedlot and slaughter processes, we collected pooled, ecological-level samples from pens of cattle and their environment as they moved through the beef production system (Figure 1). Pooled samples provide a representative resistome for the unit of animal management in feedlots and abattoirs, i.e., the pen groupings; they do not provide insight into individual animal resistome variability, and therefore inferences in this study were made at the group level. Pooled fecal, soil and drinking water samples from each pen were collected within 6 weeks after the arrival of cattle at the feedlots ('arrival samples', n=24, one pooled sample per pen per sample type; see Figure 1—source data 1), as well as within 1 day before the same pens of cattle were shipped for slaughter ('exit samples', n=24, one pooled sample per pen per sample type). Cattle were then transported (<8 hr transport time) by truck to 2 abattoirs (one in Texas and one in Colorado) and the walls, floors and ceilings of the trucks were swabbed ('truck samples', n=8, one per pen) immediately after the cattle were unloaded; each abattoir slaughtered ~5500 cattle per day. Outside of the abattoirs, cattle were placed in holding pens (i.e., outdoor enclosures that can hold hundreds of cattle for a short amount of time), where pooled fecal and drinking water samples were collected after cattle had been moved into the abattoir ('holding samples', n=16, one per pen per sample type). Cattle were then euthanized and the carcasses disassembled into beef products. Both abattoirs employed multiple-hurdle interventions to reduce pathogen and other bacterial contamination on carcasses; this included hot water pasteurization, lactic and peroxyacetic acid spray, as well as knife trimming and spot steam vacuuming of the carcass (Figure 1). At the end of this process, pooled swab samples were taken from the conveyor belt used to transport disassembled market-ready carcass parts (n=8, one per pen). In addition, beef trimmings (i.e., the parts of the carcass used to make ground beef) were collected and rinsed to obtain a pooled sample of the parts of the carcass with the highest food-safety risk (n=8, one per pen). The metagenomic sequences obtained from table and trimming samples represented the microbiome and resistome after antibacterial interventions had been applied to the carcass, and just before the beef products were packaged for retail distribution ('market-ready samples'). All sampling locations represented points in the beef production process when cattle or their end-products were actively managed and/or antimicrobial interventions were applied, both of which could influence the abundance and/or transmission of ARDs.

Total DNA was extracted from 88 samples and sequenced on an Illumina HiSeq, resulting in 407.7 Gb of sequence data (average 46.3 M reads per sample, range 12.0–93.4 M, Supplementary file 1). One drinking water sample did not contain enough DNA (i.e., <1 ng) to be sufficiently sequenced. Reads were trimmed and filtered for quality, and reads classified as host genome (Bos taurus) were removed from further analysis (Supplementary file 1). Non-host reads were then aligned to a custom non-redundant database of ARD sequences compiled from publicly available sources. ARDs with a gene fraction of >80% across all alignments were considered to be positively identified in a sample. We identified 319 unique ARDs across all 87 samples (Supplementary file 2), representing 1.2 M individual read alignments (Figure 2—source data 1). Most ARDs were present in low numbers (Figure 2A), and the proportion of trimmed, non-host reads aligning to ARDs in the database was correspondingly low across all samples (range 0.00% to 0.12%, Supplementary file 1). The median number of unique ARDs identified per sample was 33 (range: 0 to 136; Figure 2B). While this may be an underestimate due to sequencing coverage, recent studies using functional metagenomic screening of dairy cattle feces reported a maximum of 26 unique ARDs per animal (Wichmann et al., 2014). Functional metagenomics offers increased sensitivity compared to shotgun metagenomics (Forsberg et al., 2014), and therefore our observation of a relatively large number of unique ARDs per sample suggests that the pooling of samples undertaken in this study may have increased sensitivity. For comparison, a shotgun metagenomic study of 252 individual human fecal samples identified an average of 21 unique ARDs per sample (Forslund et al., 2013). The 319 ARDs identified in this study represented 42 AMR mechanisms within 17 drug classes (Supplementary file 3). Reads aligning to genes that encode resistance to tetracyclines and the macrolide-lincosamide-streptogramin classes of antimicrobials were most abundant, with the TetQ, TetW, TetO, and mefA gene families comprising the majority of alignments within these two classes (see Source Data for Figure 2A).

Figure 2

Download asset Open asset

Figure 2—source data 1

Raw and normalized count matrix of ARDs (rows) identified by sample (columns).

https://doi.org/10.7554/eLife.13195.007

Download elife-13195-fig2-data1-v1.xlsx

To assess systematic changes in resistome composition during the feeding period (i.e., from arrival to exit, truck and holding samples), non-metric multidimensional scaling (NMDS) ordination at the ARD level was performed using Hellinger transformation and Euclidean distances to avoid overweighting of rare ARDs, which were prevalent in these data (Figure 2A) (Legendre and Gallagher, 2001). Samples with only one ARD (n = 2) were removed from clustering analyses. Pre-slaughter samples clustered by sample matrix (i.e., feces, soil, water and swabs, the latter of which comprised all truck samples) based on ARD composition (Figure 3A), suggesting that the resistomes in these sample types differed significantly. Therefore, to avoid confounding, we performed ordination separately on fecal, soil and water samples, all of which exhibited a significant shift from arrival to exit or holding (Figure 3B–D); the truck resistome could not be compared owing to complete confounding between sampling location and matrix type. While this shift could result from AMD exposures, culture-based studies of phenotypic resistance have reported mixed results when investigating potential associations between feedlot AMD use and AMR (Rao et al., 2010; Morley et al., 2011). In addition to AMD exposures, cattle undergo numerous changes during the feeding period, including maturation and a gradual shift from forage-based to high-energy rations, all of which have been shown to affect the fecal microbiome in swine, although little is known about these factors in beef cattle (Umu et al., 2015; Mach et al., 2015; Kim et al., 2012). Therefore, changes in the resistome could also be driven by changes in bacterial community composition, a phenomenon recently reported for a set of functionally confirmed metagenomic soil samples (Forsberg et al., 2014). In order to measure correlation between the microbiome and resistome, and thus to investigate the extent to which microbiome changes may have been driving the resistome, we performed post-hoc procrustes analysis, which can be used to determine the degree of correlation between two ordinations (Peres-Neto and Jackson, 2001). This analysis confirmed a high correlation between the resistome (ARD level) and the microbiome (species level) in arrival and exit samples (Figure 4A–B). However, there was tighter correlation on arrival than on exit, leading to the hypothesis that additional factors such as AMD exposures may have influenced resistome changes independently of the microbiome.

Figure 3

Download asset Open asset

Figure 4

Download asset Open asset

Resistome richness at the ARD level decreased significantly during the feeding period (Wilcoxon paired signed-rank p=0.002). Interestingly, this decrease occurred primarily through the loss of resistance to AMDs that were not used on cattle in this study population or in the feeding systems in which they were raised (e.g., phenicols, aminocoumarins, elfamycins, rifampin, bacitracin and polymyxin B, Figure 5). In contrast, resistance to macrolides and tetracyclines remained prevalent in exit and holding pen samples, and these were the two classes of AMDs most commonly administered to cattle within study pens (Table 1). In addition, 5 of the 8 pens contained several cattle that received fluoroquinolones, and 3 of the 8 contained several cattle that received β-lactam AMDs. One hypothesis for the finding of reduced ARD richness is that the pattern of antimicrobial exposures during the feeding period created pressure on microbes that favored maintenance of ARDs that conferred a fitness advantage (i.e., ARDs protecting against macrolides and tetracyclines). Conversely, ARDs conferring resistance to antimicrobials that were not used in this study population (e.g., phenicols, aminocoumarins, elfamycins) or non-drug-specific ARDs (e.g., porin modification genes) did not confer a competitive advantage and may have exacted a fitness cost to bacteria carrying them (Vogwill and MacLean, 2015). Thus, such bacteria may have gradually diminished in the population, leading to disappearance of these ARDs within the pan-microbial population and thus an observed decrease in ARD richness. These pan-microbial population genetic changes may also explain the apparent decoupling of the resistome and the microbiome that was observed from feedlot arrival to exit (Figures 4A–B). However, more research is needed to understand how selective pressures acting on specific bacteria may manifest at the pan-microbial level.

Figure 5

Download asset Open asset

Interestingly, the aminoglycoside class was an exception to this pattern, as aminoglycoside resistance remained prevalent throughout the feeding period despite the absence of these drugs in this study population and in beef production in general. While this study was not designed to definitely test for AMD-ARD associations, these findings suggest that the relationship between AMD use and AMR is not straightforward. In addition, we identified several ARDs that confer resistance to critically important antimicrobial drugs in humans when expressed in disease-causing agents (Table 2); however, these AMDs were not used in this study population and most are not sold for use in cattle. An arrival soil sample contained the functionally-confirmed carbapenemase (bla)OXA-235 (Higgins et al., 2013) as well as vgaD and vatG ARDs, which together confer resistance to quinupristin-dalfopristin (Jung et al., 2010). Another arrival soil sample and a water sample from a holding pen contained (bla)OXA-347, which has only been shown to confer resistance to ampicillin but is classified as a carbapenemase based on 53% amino acid identity (Cheng et al., 2012); therefore, it is unknown whether this is a true carbapenemase ARD. Of the 7 sequenced water samples collected from holding pens, 3 contained reads aligning to the strict carbapenemase class bla(cphA). Additionally, the 4 truck samples collected in Texas all contained a CfrA 23S rRNA methyltransferase, which confers resistance to phenicols, lincosamides, oxazolidinones (linezolid), pleuromutilins, and streptogramin A (PhLOPS_A). Despite the presence of these ARDs, study cattle were not exposed to aminoglycosides, carbapenems, streptogramins, phenicols, lincosamides, linezolid, or pleuromutilins while in the feedlot (Table 1). In addition, carbapenems, pleuromutilins and linezolid are not approved for use in cattle production, and therefore antimicrobial use practices cannot directly explain the presence of these important ARDs. We also identified bla(TEM)-116, an extended-spectrum β-lactamase (ESBL) ARD, in 1 water holding pen sample. Third-generation cephalosporins (a class of β-lactams) are used in beef cattle production, although other drugs are often used more commonly as reflected in this study (Table 1). We did not identify ARDs from the bla(SHV), bla(CTX-M) or bla(CMY) classes of ESBLs, or the carbapenemases bla(SME), bla(IMI), bla(NDM), bla(GES) or bla(KPC) (Table 2).

It is important to note that these ARDs were present in extremely low relative abundance, with fewer than a dozen of the 1.2 M ARD-assigned reads aligning to each (Figure 2—source data 1). While high sequence homology and the resistance-conferring functional residues were intact (Table 2), we cannot infer phenotypic expression from these data. Furthermore, while this is the first published report of these ARDs in the feedlot setting, it is also the first study to utilize a shotgun metagenomics approach on feedlot samples; therefore, we cannot contextualize these findings with respect to previous research and we cannot determine whether presence of these ARDs in feedlot samples is a novel or long-standing phenomenon. Identifying these ARDs in metagenomic data may provide important insight above and beyond a culture- or PCR-based approach, but additional work is needed to understand the biological, ecological, and public health consequences of these findings (Martínez et al., 2015). For instance, the presence of these ARDs in the feedlot soil could be explained by transfer into the feedlot environment through either other cattle or fomites (e.g., feedlot workers, feedlot working dogs and horses), or through air or water. Use of other AMD classes could also co-select for these ARDs within the cattle population. These findings highlight the complexity of the AMD–AMR relationship, as well as the fact that food production is intrinsically linked to other ecosystems via diffuse environmental contacts. Given these complexities, we believe an ecological and metagenomic approach is necessary to thoroughly and comprehensively research this important public and human health issue.

ARD composition did not differ between pens of cattle (n=8) or feedlots (n=4) when ordinated using NMDS (Figure 6). This was notable given the geographic separation of the 4 feedlots, the fact that the pens did not have contact with one another, and the variability in parenteral AMD exposures across the 8 study pens (Table 1). However, common management strategies used in all 4 feedlots and the ubiquitous use of in-feed macrolides within study pens could explain this lack of difference. In addition, water runoff, windborne dust and fomites within feedlots could contribute to a mixing of pen resistomes, despite differential AMD exposures across pens. Resistome composition did show statistical difference by geography at the ARD level (i.e., Colorado versus Texas feedlot, truck, and holding pen samples), but not at the level of resistance mechanism or class. Despite statistical significance, both the R² value from adonis results and the R-statistic from anosim results suggest that the biological relevance of the differences between states was not high (see Figure 6—source data 1). These findings support the idea of environmental connectivity within feedlots, and perhaps within and between geographic regions. However, further study is needed to determine what effect common management practices may have had on these results.

Figure 6

Download asset Open asset

Figure 6—source data 1

NMDS ordination, adonis, and anosim results at the ARD, mechanism and class levels, by pen, feedlot, and state variables.

https://doi.org/10.7554/eLife.13195.013

Download elife-13195-fig6-data1-v1.xlsx

When examining the 16 post-slaughter samples obtained from the belts, tables, and meat trimmings, no ARDs were identified in any of these market-ready samples (n = 8 pooled belt/table samples and n=8 trimming rinses). These samples yielded large amounts of DNA, but >99% of the reads aligned to the bovine genome (Supplementary file 1); therefore, the lack of detection of ARDs could likely be attributable to low-sequencing coverage of bacterial DNA. However, there are also plausible biological explanations for the lack of bacterial DNA (and thus ARDs) in these samples. The bacterial contamination of beef during slaughter occurs primarily during the removal of the hide and gastrointestinal tract (GIT), at which point the surfaces of carcasses can routinely be contaminated with aerobic bacterial counts of 6.1 – 9.1 log₁₀ CFU/100 cm² (Bacon et al., 2000). To decrease this contamination, it is standard in North America for carcasses to undergo several highly effective antibacterial interventions after hide and GIT removal, including steam vacuuming, carcass washing, application of organic acid rinses and thermal pasteurization (Bacon et al., 2000). All carcasses in this study underwent each of these interventions sequentially, a process that has been shown to reduce bacterial loads by >5 log₁₀ CFU/100 cm² total plate count (Greig et al., 2012). Nevertheless, nationwide food safety surveys in the U.S. suggest that a relatively low level of bacterial presence is common on post-slaughter beef trimmings, with aerobic plate counts routinely containing 10²–10³ CFU/g of trimmings (USDA, FSIS, 2011). It is likely that some unknown proportion of these organisms harbor ARDs, suggesting that the shotgun metagenomics assay as used in these high-host background samples may have had low sensitivity to detect these very low-abundance ARDs.

To differentiate between the effects of incomplete sequencing and true reduction of the microbial population, we conducted a microbiome analysis of samples collected pre- and post-slaughter, the former comprising all arrival, exit, truck, and holding pen samples. We found that microbiome (i.e., total microbial community) diversity was reduced in the post-slaughter samples compared to the pre-slaughter samples, suggesting a dramatic alteration in the composition of the microbial community, which could reflect the impact of antibacterial interventions that are applied during slaughter (Figure 7A). However, low diversity could also be the result of low-coverage sequencing of microbial DNA in the post-slaughter samples (Smith et al., 2014). Therefore, we conducted a closer analysis of differential microbial abundance between pre- and post-slaughter samples using zero-inflated Gaussian mixture models to account for distinct processes for zero-count data (i.e., true absence vs. incomplete sequencing depth) (Paulson et al., 2013). Pairwise comparisons between pre- and post-slaughter samples were performed using limma’s makeContrasts function (Smyth, 2004), and pen ID was added to all models as a covariable to account for repeated measurements. We identified 416 (of 763) differentially abundant genera and 840 (of 1821) differentially abundant species, the vast majority of which were more abundant in the pre-slaughter samples (Figure 7B). Of the 19 genera and 68 species that were more abundant in post-slaughter samples, many are known to be heat-tolerant and/or environmentally hardy bacteria, archaea, viruses, and phages (seeFigure 7—source data 1); e.g., Staphylothermus, Pyrococcus, Pseudomonas and Pleurocapsa, suggesting that they were able to withstand the high heat and other harsh conditions utilized as part of the slaughter antimicrobial interventions. However, sample matrices in the pre- and post-slaughter samples were confounded with location. Therefore, we performed the same analysis solely on swab samples (pre-slaughter swabs from the transport trucks and post-slaughter swabs from the abattoir conveyor belts). Interestingly, the same overall pattern of change was observed; the majority of bacterial genera (105/757) exhibited decreased abundance post-slaughter (see Figure 7—figure supplement 1). However, unlike the comparison across all matrices, the genera with increased relative abundance in post-slaughter swab samples were not characterized by a large proportion of environmentally hardy bacteria ( Figure 7—source data 1). Given that the truck environment is also relatively harsh (i.e., extreme temperatures, lack of nutrients, dry conditions), this result was not unexpected. Taken together, these results suggest that the multiple-hurdle interventions decrease overall bacterial load, and thus greatly diminish the abundance of ARDs in post-slaughter beef. However, the incomplete sequencing depth in the post-slaughter samples cannot be fully discounted and future metagenomic investigations of samples with high-host background should consider methods to overcome this limitation. We were unable to compare these results to previous studies in the abattoir setting, as most of these bacteria are difficult or impossible to culture and therefore have not been investigated. Interestingly, culture-based resistance surveillance in beef trimmings reported >30% prevalence of resistance amongst Salmonella isolates (USDA, FSIS, 2011). However, Salmonella was recovered from <1% of these trimming samples (22 of 1,791 samples collected in 2011), suggesting that the recovery of resistant Salmonella is a very rare event; in this respect, the results of these culture-based efforts concur with our findings. Furthermore, samples taken as part of the culture-based surveillance program were enriched prior to isolation, likely resulting in increased sensitivity for Salmonella detection compared to the shotgun metagenomics approach. Further work is needed to compare the sensitivity and specificity of shotgun metagenomics with more traditional assays.

Figure 7 with 1 supplement see all

Download asset Open asset

Figure 7—source data 1

Shannon’s diversity, taxa table and model output for Figure 7.

Sheet 1 ('Figure 7A') includes sample metadata and Shannon's Diversity Index calculated at the species level. Sheet 2 ('Figure 7B taxa table') includes the taxa count matrix, by sample, that was used to model log₂-fold changes in abundance for Figure 7B and Figure 7—figure supplement 1. Sheet 3 ('Figure 7B model output') shows model output for log₂-fold change in abundance at the genus level across all pre- and post-slaughter samples. Sheet 4 ('Figure 7 supp. Model output') shows model output for log₂-fold change in abundance at the genus level between pre- and post-slaughter swab samples.

https://doi.org/10.7554/eLife.13195.015

Download elife-13195-fig7-data1-v1.xlsx

1. 1. Microbial Ecology Group, Colorado State University

   (2015) Beef Production Metagenome Raw sequence reads

   Publicly available at the NCBI BioProject database (Accession no: PRJNA292471).

   http://www.ncbi.nlm.nih.gov/bioproject/292471

1. 1. Anderson MJ

   (2001) A new method for non-parametric multivariate analysis of variance

   Austral Ecology 26:32–46.

   https://doi.org/10.1111/j.1442-9993.2001.01070.pp.x

   • Google Scholar
2. Website

   1. Antibiotic Resistance from the Farm to the Table [Internet]

   (2014) The Centers for Disease Control and Prevention

   http://www.cdc.gov/foodsafety/from-farm-to-table.html
3. 1. Bacon RT
   2. Belk KE
   3. Sofos JN
   4. Clayton RP
   5. Reagan JO
   6. Smith GC

   (2000)

   Microbial populations on animal hides and beef carcasses at different stages of slaughter in plants employing multiple-sequential interventions for decontamination

   Journal of Food Protection 63:1080–1086.

   • PubMed
   • Google Scholar
4. 1. Benedict KM
   2. Gow SP
   3. Checkley S
   4. Booker CW
   5. McAllister TA
   6. Morley PS

   (2013) Methodological comparisons for antimicrobial resistance surveillance in feedlot cattle

   BMC Veterinary Research 9:216.

   https://doi.org/10.1186/1746-6148-9-216

   • PubMed
   • Google Scholar
5. 1. Bolger AM
   2. Lohse M
   3. Usadel B

   (2014) Trimmomatic: a flexible trimmer for Illumina sequence data

   Bioinformatics 30:2114–2120.

   https://doi.org/10.1093/bioinformatics/btu170

   • PubMed
   • Google Scholar
6. Website

   1. Canadian Integrated Program for Antimicrobial Resistance Surveillance (CIPARS)

   (2013) http://www.phac-aspc.gc.ca/cipars-picra/2013/annu-report-rapport-eng.php

   http://www.phac-aspc.gc.ca/cipars-picra/2013/annu-report-rapport-eng.php
7. Website

   1. Census of Agriculture US

   (2012) Internet]. USDA National Agricultural Statistics Service; 2014

   http://www.agcensus.usda.gov/Publications/2012/#full_report
8. 1. Chantziaras I
   2. Boyen F
   3. Callens B
   4. Dewulf J

   (2014) Correlation between veterinary antimicrobial use and antimicrobial resistance in food-producing animals: a report on seven countries

   The Journal of Antimicrobial Chemotherapy 69:.

   https://doi.org/10.1093/jac/dkt443

   • PubMed
   • Google Scholar
9. 1. Cheng G
   2. Hu Y
   3. Yin Y
   4. Yang X
   5. Xiang C
   6. Wang B
   7. Chen Y
   8. Yang F
   9. Lei F
   10. Wu N
   11. Lu N
   12. Li J
   13. Chen Q
   14. Li L
   15. Zhu B

   (2012) Functional screening of antibiotic resistance genes from human gut microbiota reveals a novel gene fusion

   FEMS Microbiology Letters 336:11–16.

   https://doi.org/10.1111/j.1574-6968.2012.02647.x

   • Google Scholar
10. 1. Daniel CR
    2. Cross AJ
    3. Koebnick C
    4. Sinha R

    (2011) Trends in meat consumption in the USA

    Public Health Nutrition 14:575–583.

    https://doi.org/10.1017/S1368980010002077

    • PubMed
    • Google Scholar
11. 1. Durso LM
    2. Harhay GP
    3. Bono JL
    4. Smith TPL

    (2011) Virulence-associated and antibiotic resistance genes of microbial populations in cattle feces analyzed using a metagenomic approach

    Journal of Microbiological Methods 84:278–282.

    https://doi.org/10.1016/j.mimet.2010.12.008

    • Google Scholar
12. Website

    1. Food and Drug Administration

    (2003) Guidance for Industry #152: Evaluating the Safety of Antimicrobial New Animal Drugs with Regard to Their Microbiological Effects on Bacteria of Human Health Concern [Internet]. 2003

    http://www.fda.gov/downloads/AnimalVeterinary/GuidanceComplianceEnforcement/GuidanceforIndustry/ucm052519.pdf
13. 1. Forsberg KJ
    2. Reyes A
    3. Wang B
    4. Selleck EM
    5. Sommer MOA
    6. Dantas G

    (2012) The Shared Antibiotic Resistome of Soil Bacteria and Human Pathogens

    Science 337:1107–1111.

    https://doi.org/10.1126/science.1220761

    • PubMed
    • Google Scholar
14. 1. Forsberg KJ
    2. Patel S
    3. Gibson MK
    4. Lauber CL
    5. Knight R
    6. Fierer N
    7. Dantas G

    (2014) Bacterial phylogeny structures soil resistomes across habitats

    Nature 509:612–616.

    https://doi.org/10.1038/nature13377

    • Google Scholar
15. 1. Forslund K
    2. Sunagawa S
    3. Kultima JR
    4. Mende DR
    5. Arumugam M
    6. Typas A
    7. Bork P

    (2013) Country-specific antibiotic use practices impact the human gut resistome

    Genome Research 23:1163–1169.

    https://doi.org/10.1101/gr.155465.113

    • Google Scholar
16. 1. Fu L
    2. Niu B
    3. Zhu Z
    4. Wu S
    5. Li W

    (2012) CD-HIT: accelerated for clustering the next-generation sequencing data

    Bioinformatics 28:3150–3152.

    https://doi.org/10.1093/bioinformatics/bts565

    • Google Scholar
17. 1. Gibson MK
    2. Forsberg KJ
    3. Dantas G

    (2015) Improved annotation of antibiotic resistance determinants reveals microbial resistomes cluster by ecology

    The ISME Journal 9:207–216.

    https://doi.org/10.1038/ismej.2014.106

    • Google Scholar
18. 1. Gow SP
    2. Waldner CL
    3. Rajić A
    4. McFall ME
    5. Reid-Smith R

    (2008)

    Prevalence of antimicrobial resistance in fecal generic Escherichia coil isolated in western Canadian cow-calf herds. Part I--beef calves

    Canadian Journal of Veterinary Research = Revue Canadienne De Recherche VéTéRinaire 72:82–90.

    • PubMed
    • Google Scholar
19. 1. Greig JD
    2. Waddell L
    3. Wilhelm B
    4. Wilkins W
    5. Bucher O
    6. Parker S
    7. Rajić A

    (2012) The efficacy of interventions applied during primary processing on contamination of beef carcasses with Escherichia coli: A systematic review-meta-analysis of the published research

    Food Control 27:385–397.

    https://doi.org/10.1016/j.foodcont.2012.03.019

    • Google Scholar
20. 1. Gupta SK
    2. Padmanabhan BR
    3. Diene SM
    4. Lopez-Rojas R
    5. Kempf M
    6. Landraud L
    7. Rolain J-M

    (2014) ARG-ANNOT, a New Bioinformatic Tool to Discover Antibiotic Resistance Genes in Bacterial Genomes

    Antimicrobial Agents and Chemotherapy 58:212–220.

    https://doi.org/10.1128/AAC.01310-13

    • Google Scholar
21. 1. Hall RM
    2. Schwarz S

    (2016) Resistance gene naming and numbering: is it a new gene or not?

    Journal of Antimicrobial Chemotherapy 71:569–571.

    https://doi.org/10.1093/jac/dkv351

    • PubMed
    • Google Scholar
22. 1. Higgins PG
    2. Perez-Llarena FJ
    3. Zander E
    4. Fernandez A
    5. Bou G
    6. Seifert H

    (2013) OXA-235, a Novel Class D -Lactamase Involved in Resistance to Carbapenems in Acinetobacter baumannii

    Antimicrobial Agents and Chemotherapy 57:2121–2126.

    https://doi.org/10.1128/AAC.02413-12

    • Google Scholar
23. 1. Jung Y-H
    2. Shin ES
    3. Kim O
    4. Yoo JS
    5. Lee KM
    6. Yoo JI
    7. Chung GT
    8. Lee YS

    (2010) Characterization of Two Newly Identified Genes, vgaD and vatG, Conferring Resistance to Streptogramin a in Enterococcus faecium

    Antimicrobial Agents and Chemotherapy 54:4744–4749.

    https://doi.org/10.1128/AAC.00798-09

    • Google Scholar
24. 1. Kim HB
    2. Borewicz K
    3. White BA
    4. Singer RS
    5. Sreevatsan S
    6. Tu ZJ
    7. Isaacson RE

    (2012) Microbial shifts in the swine distal gut in response to the treatment with antimicrobial growth promoter, tylosin

    Proceedings of the National Academy of Sciences 109:15485–15490.

    https://doi.org/10.1073/pnas.1205147109

    • Google Scholar
25. 1. Landers TF
    2. Cohen B
    3. Wittum TE
    4. Larson EL

    (2012)

    A review of antibiotic use in food animals: perspective, policy, and potential

    Public Health Reports 127:4–22.

    • PubMed
    • Google Scholar
26. 1. Legendre P
    2. Gallagher E

    (2001) Ecologically meaningful transformations for ordination of species data

    Oecologia 129:271–280.

    https://doi.org/10.1007/s004420100716

    • Google Scholar
27. 1. Levy SB
    2. FitzGerald GB
    3. Macone AB

    (1976) Changes in Intestinal Flora of Farm Personnel after Introduction of a Tetracycline-Supplemented Feed on a Farm

    New England Journal of Medicine 295:583–588.

    https://doi.org/10.1056/NEJM197609092951103

    • PubMed
    • Google Scholar
28. 1. Li H
    2. Durbin R

    (2009) Fast and accurate short read alignment with Burrows-Wheeler transform

    Bioinformatics 25:1754–1760.

    https://doi.org/10.1093/bioinformatics/btp324

    • PubMed
    • Google Scholar
29. 1. Mach N
    2. Berri M
    3. Estellé J
    4. Levenez F
    5. Lemonnier G
    6. Denis C
    7. Leplat J-J
    8. Chevaleyre C
    9. Billon Y
    10. Doré J
    11. Rogel-Gaillard C
    12. Lepage P

    (2015) Early-life establishment of the swine gut microbiome and impact on host phenotypes

    Environmental Microbiology Reports 7:554–569.

    https://doi.org/10.1111/1758-2229.12285

    • Google Scholar
30. 1. Martínez JL
    2. Coque TM
    3. Baquero F

    (2015) What is a resistance gene? Ranking risk in resistomes

    Nature Reviews Microbiology 13:116–123.

    https://doi.org/10.1038/nrmicro3399

    • PubMed
    • Google Scholar
31. 1. McArthur AG
    2. Waglechner N
    3. Nizam F
    4. Yan A
    5. Azad MA
    6. Baylay AJ
    7. Bhullar K
    8. Canova MJ
    9. De Pascale G
    10. Ejim L
    11. Kalan L
    12. King AM
    13. Koteva K
    14. Morar M
    15. Mulvey MR
    16. O'Brien JS
    17. Pawlowski AC
    18. Piddock LJV
    19. Spanogiannopoulos P
    20. Sutherland AD
    21. Tang I
    22. Taylor PL
    23. Thaker M
    24. Wang W
    25. Yan M
    26. Yu T
    27. Wright GD

    (2013) The Comprehensive Antibiotic Resistance Database

    Antimicrobial Agents and Chemotherapy 57:3348–3357.

    https://doi.org/10.1128/AAC.00419-13

    • Google Scholar
32. 1. McEachran AD
    2. Blackwell BR
    3. Hanson JD
    4. Wooten KJ
    5. Mayer GD
    6. Cox SB
    7. Smith PN

    (2015) Antibiotics, Bacteria, and Antibiotic Resistance Genes: Aerial Transport from Cattle Feed Yards via Particulate Matter

    Environmental Health Perspectives 123:337–343.

    https://doi.org/10.1289/ehp.1408555

    • PubMed
    • Google Scholar
33. 1. McMurdie PJ
    2. Holmes S

    (2014) Waste Not, Want Not: Why Rarefying Microbiome Data Is Inadmissible

    PLoS Computational Biology 10:e1003531.

    https://doi.org/10.1371/journal.pcbi.1003531

    • Google Scholar
34. 1. Moon DC
    2. Tamang MD
    3. Nam HM
    4. Jeong JH
    5. Jang GC
    6. Jung SC
    7. Park YH
    8. Lim SK

    (2015) Identification of livestock-associated methicillin-resistant Staphylococcus aureus isolates in Korea and molecular comparison between isolates from animal carcasses and slaughterhouse workers

    Foodborne Pathogens and Disease 12:327–334.

    https://doi.org/10.1089/fpd.2014.1868

    • PubMed
    • Google Scholar
35. 1. Morley PS
    2. Dargatz DA
    3. Hyatt DR
    4. Dewell GA
    5. Patterson JG
    6. Burgess BA
    7. Wittum TE

    (2011) Effects of Restricted Antimicrobial Exposure on Antimicrobial Resistance in Fecal Escherichia coli from Feedlot Cattle

    Foodborne Pathogens and Disease 8:87–98.

    https://doi.org/10.1089/fpd.2010.0632

    • PubMed
    • Google Scholar
36. 1. National Antimicrobial Resistance Monitoring System – Enteric Bacteria (NARMS): 2011 Executive Repor

    (2013)

    U.S. Department of Health and Human Services, Food and Drug Administration

    U.S. Department of Health and Human Services, Food and Drug Administration.

    • Google Scholar
37. Book

    1. National Research Council

    (2000)

    Nutrient Requirements of Beef Cattle (8th revised edition)

    The National Academic Press.

    • Google Scholar
38. Book

    1. OECD/Food and Agriculture Organization of the United Nations

    (2015)

    OECD-FAO Agricultural Outlook 2015-2024 [Internet

     OECD Publishing.

    • Google Scholar
39. Website

    1. Oksanen J
    2. Blanchet FG
    3. Kindt R
    4. Legendre P
    5. Minchin PR
    6. O’Hara RB

    (2014) vegan: Community Ecology Package [Internet]

    http://CRAN.R-project.org/package=vegan
40. 1. Paulson JN
    2. Stine OC
    3. Bravo HC
    4. Pop M

    (2013) Differential abundance analysis for microbial marker-gene surveys

    Nature Methods 10:1200–1202.

    https://doi.org/10.1038/nmeth.2658

    • Google Scholar
41. 1. Peres-Neto P
    2. Jackson D

    (2001) How well do multivariate data sets match? the advantages of a Procrustean superimposition approach over the Mantel test

    Oecologia 129:169–178.

    https://doi.org/10.1007/s004420100720

    • Google Scholar
42. 1. Rao S
    2. Van Donkersgoed J
    3. Bohaychuk V
    4. Besser T
    5. Song X-M
    6. Wagner B
    7. Hancock D
    8. Renter D
    9. Dargatz D
    10. Morley PS

    (2010) Antimicrobial Drug Use and Antimicrobial Resistance in Enteric Bacteria Among Cattle from Alberta Feedlots

    Foodborne Pathogens and Disease 7:449–457.

    https://doi.org/10.1089/fpd.2009.0400

    • PubMed
    • Google Scholar
43. 1. Rolain J-M

    (2013) Food and human gut as reservoirs of transferable antibiotic resistance encoding genes

    Frontiers in Microbiology 4:173.

    https://doi.org/10.3389/fmicb.2013.00173

    • PubMed
    • Google Scholar
44. 1. Sheikh AA
    2. Checkley S
    3. Avery B
    4. Chalmers G
    5. Bohaychuk V
    6. Boerlin P
    7. Reid-Smith R
    8. Aslam M

    (2012) Antimicrobial resistance and resistance genes in Escherichia coli isolated from retail meat purchased in Alberta, Canada

    Foodborne Pathogens and Disease 9:625–631.

    https://doi.org/10.1089/fpd.2011.1078

    • PubMed
    • Google Scholar
45. 1. Smith DP
    2. Peay KG
    3. Depth S
    4. Replication N

    (2014) Sequence Depth, Not PCR Replication, Improves Ecological Inference from Next Generation DNA Sequencing

    PLoS ONE 9:e90234.

    https://doi.org/10.1371/journal.pone.0090234

    • PubMed
    • Google Scholar
46. 1. Smyth GK

    (2004) Linear Models and Empirical Bayes Methods for Assessing Differential Expression in Microarray Experiments

    Statistical Applications in Genetics and Molecular Biology 3:1–25.

    https://doi.org/10.2202/1544-6115.1027

    • Google Scholar
47. 1. Song SJ
    2. Lauber C
    3. Costello EK
    4. Lozupone CA
    5. Humphrey G
    6. Berg-Lyons D
    7. Caporaso JG
    8. Knights D
    9. Clemente JC
    10. Nakielny S
    11. Gordon JI
    12. Fierer N
    13. Knight R

    (2013) Cohabiting family members share microbiota with one another and with their dogs

    eLife 2:e00458.

    https://doi.org/10.7554/eLife.00458

    • Google Scholar
48. 1. Tsai YL
    2. Olson BH

    (1992)

    Rapid method for separation of bacterial DNA from humic substances in sediments for polymerase chain reaction

    Applied and Environmental Microbiology 58:2292–2295.

    • PubMed
    • Google Scholar
49. 1. Umu ÖC
    2. Frank JA
    3. Fangel JU
    4. Oostindjer M
    5. da Silva CS
    6. Bolhuis EJ
    7. Bosch G
    8. Willats WG
    9. Pope PB
    10. Diep DB

    (2015) Resistant starch diet induces change in the swine microbiome and a predominance of beneficial bacterial populations

    Microbiome 3:16.

    https://doi.org/10.1186/s40168-015-0078-5

    • PubMed
    • Google Scholar
50. 1. USDA, FSIS

    (2011)

    National Prevalence Estimate of Pathogens in Domestic Beef Manufacturing Trimmings (Trim)

    National Prevalence Estimate of Pathogens in Domestic Beef Manufacturing Trimmings (Trim).

    • Google Scholar
51. 1. Vogwill T
    2. MacLean RC

    (2015) The genetic basis of the fitness costs of antimicrobial resistance: a meta-analysis approach

    Evolutionary Applications 8:284–295.

    https://doi.org/10.1111/eva.12202

    • Google Scholar
52. 1. Wichmann F
    2. Udikovic-Kolic N
    3. Andrew S
    4. Handelsman J

    (2014) Diverse Antibiotic Resistance Genes in Dairy Cow Manure

    mBio 5:e01017-13–13.

    https://doi.org/10.1128/mBio.01017-13

    • Google Scholar
53. 1. Wood DE
    2. Salzberg SL

    (2014) Kraken: ultrafast metagenomic sequence classification using exact alignments

    Genome Biology 15:R46.

    https://doi.org/10.1186/gb-2014-15-3-r46

    • Google Scholar
54. Book

    1. World Health Organization

    (2011)

    Critically Important Antimicrobials for Human Medicine, 3rd Revision [Internet

    World Health Organization.

    • Google Scholar
55. 1. Yatsunenko T
    2. Rey FE
    3. Manary MJ
    4. Trehan I
    5. Dominguez-Bello MG
    6. Contreras M
    7. Magris M
    8. Hidalgo G
    9. Baldassano RN
    10. Anokhin AP
    11. Heath AC
    12. Warner B
    13. Reeder J
    14. Kuczynski J
    15. Caporaso JG
    16. Lozupone CA
    17. Lauber C
    18. Clemente JC
    19. Knights D
    20. Knight R
    21. Gordon JI

    (2012) Human gut microbiome viewed across age and geography

    Nature 486:222–227.

    https://doi.org/10.1038/nature11053

    • Google Scholar
56. 1. Zankari E
    2. Hasman H
    3. Cosentino S
    4. Vestergaard M
    5. Rasmussen S
    6. Lund O
    7. Aarestrup FM
    8. Larsen MV

    (2012) Identification of acquired antimicrobial resistance genes

    Journal of Antimicrobial Chemotherapy 67:2640–2644.

    https://doi.org/10.1093/jac/dks261

    • Google Scholar

Author details

1. Noelle R Noyes

   Department of Clinical Sciences, Colorado State University, Fort Collins, United States

   Contribution

   NRN, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   noelle.noyes@colostate.edu

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0001-6149-1008

2. Xiang Yang

   Department of Animal Sciences, Colorado State University, Fort Collins, United States

   Contribution

   XY, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

3. Lyndsey M Linke

   Department of Clinical Sciences, Colorado State University, Fort Collins, United States

   Contribution

   LML, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

4. Roberta J Magnuson

   Department of Clinical Sciences, Colorado State University, Fort Collins, United States

   Contribution

   RJM, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

5. Adam Dettenwanger

   Department of Computer Sciences, Colorado State University, Fort Collins, United States

   Contribution

   AD, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

6. Shaun Cook

   Agriculture and Agri-Food Canada Research Centre, Lethbridge, Canada

   Contribution

   SC, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

7. Ifigenia Geornaras

   Department of Animal Sciences, Colorado State University, Fort Collins, United States

   Contribution

   IG, Conception and design, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

8. Dale E Woerner

   Department of Animal Sciences, Colorado State University, Fort Collins, United States

   Contribution

   DEW, Conception and design, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

9. Sheryl P Gow

   Centre for Food-borne, Environmental Zoonotic Infectious Diseases, Public Health Agency of Canada, University of Saskatoon, Saskatoon, Canada

   Contribution

   SPG, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

10. Tim A McAllister

    Agriculture and Agri-Food Canada Research Centre, Lethbridge, Canada

    Contribution

    TAM, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

    Competing interests

    The authors declare that no competing interests exist.

11. Hua Yang

    Department of Animal Sciences, Colorado State University, Fort Collins, United States

    Contribution

    HY, Conception and design, Drafting or revising the article

    Competing interests

    The authors declare that no competing interests exist.

12. Jaime Ruiz

    Department of Computer Sciences, Colorado State University, Fort Collins, United States

    Contribution

    JR, Acquisition of data, Drafting or revising the article

    Competing interests

    The authors declare that no competing interests exist.

13. Kenneth L Jones

    Department of Biochemistry and Molecular Genetics, University of Colorado Denver School of Medicine, Aurora, United States

    Contribution

    KLJ, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

    Competing interests

    The authors declare that no competing interests exist.

14. Christina A Boucher

    Department of Computer Sciences, Colorado State University, Fort Collins, United States

    Contribution

    CAB, Analysis and interpretation of data, Drafting or revising the article

    Competing interests

    The authors declare that no competing interests exist.

15. Paul S Morley

    Department of Clinical Sciences, Colorado State University, Fort Collins, United States

    Contribution

    PSM, Conception and design, Analysis and interpretation of data, Drafting or revising the article

    Contributed equally with

    Keith E Belk

    Competing interests

    The authors declare that no competing interests exist.

16. Keith E Belk

    Department of Animal Sciences, Colorado State University, Fort Collins, United States

    Contribution

    KEB, Conception and design, Acquisition of data, Drafting or revising the article

    Contributed equally with

    Paul S Morley

    Competing interests

    The authors declare that no competing interests exist.

• 5,979

  views

• 1,159

  downloads

• 98

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.