The upper respiratory tracts of the cattle sampled in this study were dominated by Moraxella bovoculi, Mycoplasma dispar, and Pasteurella multocida, a finding consistent with other studies of the bovine respiratory microbiome in healthy [24, 25] and unhealthy [26, 27] feedlot cattle in North America. Although M. bovoculi is frequently found in the bovine respiratory tract, it is more commonly associated with bovine keratoconjunctivitis [28], though the species from these two niches are genetically distinct [29]. M. bovoculi and the broader Moraxella genus do not have a recognized role in bovine respiratory disease. In contrast, both Mycoplasma dispar and P. multocida are not only frequently detected in the nasopharynx as commensal microbiota, but are known opportunistic pathogens that actively contribute to respiratory disease [30–32]. The frequent co-occurrence of these species is remarkable in that several studies suggest synergism between Mycoplasma sp. and P. multocida [31, 33, 34] as well as a possible initiative role of M. dispar in the development of respiratory tract disease in dairy cattle [35–38]. While the pathogenicity of M. dispar is well described in dairy calves, particularly in countries free of M. bovis, it has received little attention in BRD feedlot studies. However, this could be a consequence of the challenges associated with the fastidious nature of this organism rather than its potential importance in BRD [38]. In this regard, culture-independent approaches like metagenomic sequencing offer the opportunity to expand pathogen detection beyond the more typical predetermined BRD pathogens of interest (M. haemolytica, P. multocida, H. somni) for which culture-based techniques are well-established.
In this study, P. multocida was the most abundant species in one third of all samples, representing a larger fraction of reads than what has been previously reported in other metagenomic studies on Canadian feedlots. This abundance can be partially explained by the population of chronically diseased animals sampled in this study. Unlike the primary insulting pathogens identified from acutely ill cases of BRD during the early stages of infection, P. multocida is an opportunist, and is more often implicated in cattle with subacute or chronic pneumonia [31, 39, 40]. Chronically diseased cattle are less likely to be shedding high numbers of pathogens as in acute cases. Therefore, the relative abundance of bacteria in this study’s chronically ill cattle was expected to be different compared to other studies which describe apparently healthy animals or acute cases of BRD.
The multifactorial nature of treatment failure (due to host, pathogen, drug, drug administration, and environmental factors) [23], as well as the slightly mixed population of cattle considered unresponsive to treatment, may further account for the pathogens observed. While most cattle in the chronic pens sampled in this study were considered chronic pneumonias, cattle with nonresponsive lameness were also present. In additional to naturally harbouring a different respiratory microbiome, chronically lame cattle likely received different antimicrobial protocols to their respiratory counterparts.
The differences seen between the dominant organisms in this study and those of previously published studies could potentially also be due to the relatively limited sequencing coverage of bacterial genomes in these samples. However, it is worth noting that most previous studies exploring the nasal microbiome have relied on sequencing one or two hypervariable regions of the 16S rRNA gene, and oftentimes this approach cannot be reliably used to classify sequences at the species or subspecies level [41].
The ARGs detected in these samples were associated with resistance to drugs commonly used in the cattle industry [1], including macrolides (erm35, ermC, mphE), phenicols (florR, cmx), and tetracyclines (tet34, tetB, tetH, tetQ, tetW, tetX, tetY). While these genes were detectable in the sequence data, relatively short fragment sizes and limited coverage in this preliminary proof of concept trial often impeded the classification of ARG-containing reads beyond the level of phylum. The ARGs detected in this study are similar to those of previous DNA-based surveys of the bovine respiratory tract [8, 9, 21] and even those of other bovine microbiome sites [42]. In particular, tet(H), the ARG present in the greatest number of samples, has been detected in integrative and conjugative element (ICE)-containing strains of M. haemolytica, P. multocida, and H. somni isolated from confirmed BRD cases [43].
The other tetracycline resistance genes (tet34, tetB, tetQ, tetW, tetX, tetY) detected in this study were found in fewer samples and with the exception of tet(B), are not known to occur in pathogens of the bovine respiratory system in feedlot cattle. Tet(B), which encodes a tetracycline efflux pump, has been found in E.coli isolated from bovine feces [44], feedlot fecal composite samples [45], feedlot wastewater lagoons [46], and P. multocida isolates derived from cattle [47]. Tet(W), tet(Q), and tet(X) have been detected in bronchoalveolar lavage and deep nasopharyngeal swab samples collected from feedlot cattle, but it was not evident that these genes were present in BRD pathogens rather than environmental or commensal populations [8, 21, 48, 49].
The macrolide phosphotransferase gene, mphE, was also detected in several samples. This gene has been found in M. haemolytica isolates derived from pneumonic bovine lung tissue [22, 50], and bronchoalveolar lavage samples collected from BRD cattle confirmed to have died of BRD [21]. This gene is frequently detected with msrE, as the two occur together in an operon on the ICE, ICEPmu1, which has been found in several members of the Pasteurellaceae family [50, 51]; however, msrE was not detected in any sample in this study. The aadA31 gene was also detected in several samples and encodes a spectinomycin/streptomycin adenylyltransferase. This gene has previously been detected in P. multocida and H. somni recovered from confirmed BRD mortalities where it was located inside a variant of the ICE, ICEMh1 [52]. A notable unifying theme among the most abundant ARGs in these samples is that many have been detected on ICE found in BRD-associated pathogens. The presence of these ARGs on ICE could explain why they are present in the higher abundance, but additional sequencing coverage would be necessary to confirm this.
Most M. bovis isolates recovered in this study had high MICs for gamithromycin, tildipirosin, tilmicosin and tylosin. While WGS demonstrated that some of these isolates had point mutations in the 23S rRNA gene that could be responsible for the resistant phenotype, this genetic pattern was not present in all resistant isolates and differed from the SNP signature of previously reported resistant M. bovis isolates in other studies [53, 54]. Antimicrobial resistance in Mycoplasma spp. differs from that of the other BRD pathogens, as members of this genus typically achieve resistance through point mutations rather than through the acquisition of ARGs [54, 55]. In this study, M. bovis reads were detected in 13 samples, ranging in abundance from 1 to 344 reads; the detection of genetic determinants of resistance in the metagenome was therefore not feasible for M. bovis, due to the low sequencing coverage. While the accurate detection of SNPs in metagenomes has been previously documented [56], it requires high sequence coverage of target genomes (typically in excess of 100x).
For M. haemolytica, P. multocida and H. somni, sequencing detected the presence of the pathogens more frequently than did culture. Indeed, DNA-based approaches are often better equipped to identify the presence of pathogens, as they can detect bacteria that are growth-inhibited or dead following antimicrobial therapy [57]. A comparison of these methods is more complex for M. bovis, insofar that the pathogen was recovered from culture but not the sequence data on four occasions. Given that the genome of M. bovis is approximately one-third of the length of the other BRD pathogens, it would have a slightly higher limit of detection on a per-cell basis and might therefore have been missed by metagenomic sequencing. Further and owing to its lack of cell wall, M. bovis may have been absent from some samples due to inherent incompatibilities with sample processing. This is probably unlikely given the comparatively large representation from other Mycoplasma spp. detected in sequence, such as M. dispar.