The Infectious Diseases Society of America is among the organizations that have recommended that the U.S. government limit the use of antibiotics in agriculture (1
). The European Union has banned the use of all agricultural antibiotics that are used for growth promotion (3
). However, various factors, notably the cost-effectiveness of antibiotic use for performance benefits in modern, conventional agricultural practices (5
), have forestalled similar measures in the United States. Partly driving the ongoing debate on the prudence of the widespread use of antibiotics to improve feed efficiency in agricultural animals (6
) is a growing body of research on the collateral effects of these antibiotics, ranging from an increased abundance of antibiotic resistance genes in pigs fed antibiotics (T. P. Looft, T. Johnson, H. K. Allen, D. O. Bayles, D. P. Alt, R. D. Stedtfeld, W.-J. Sul, T. M. Stedtfeld, B. Chai, S. A. Hashsham, J. M. Tiedje, and T. B. Stanton, submitted for publication) to the modulation of bacterial gene expression by subinhibitory concentrations of antibiotics (8
). Another collateral effect could be the induction of gene transfer among bacteria.
Horizontal gene transfer is a mechanism by which bacteria exchange genetic material and is known to occur among gut bacteria (9
). One type of horizontal gene transfer is mediated by phages. Some antibiotics are known to affect prophage-mediated gene transfer in certain bacteria in vitro
. The virus-like gene transfer agent VSH-1 is induced by carbadox in the swine pathogen Brachyspira hyodysenteriae
and transfers antibiotic resistance genes (10
). Also, beta-lactam antibiotics and fluoroquinolones induce prophages in Staphylococcus aureus
, some of which package pathogenicity islands and therefore transfer virulence traits (12
). However, the in vivo
effects of antibiotics on phages in the gut are unknown.
Phage diversity and function have been studied in water, soil, and animal-associated environments, but only a small fraction of phages have been characterized (14
). Metagenomic analyses enable the study of phages without isolating them (15
). The phage metagenome, or virome, is the sequenced assemblage of the total phages of a microbial community. Recently, phage metagenomic analyses have launched studies comparing phage ecology between environments. In such studies, phage diversity and functions can be elucidated despite a limited understanding of specific phages in a given community, such as the demonstrated increase in phage diversity in the lungs of cystic fibrosis patients (16
Our goal was to examine the fecal viromes over time in swine that were fed the common antibiotics carbadox and ASP250. The viromes were compared to those of nonmedicated swine and to the corresponding bacterial communities. The data show that in-feed antibiotics induce prophages in the swine intestine and cause significant shifts in both phage and bacterial community structures. Additionally, analysis of the relative abundance of Firmicutes bacteria and phages, specifically the Streptococcus spp., unexpectedly revealed that the predator-prey population dynamics model called “kill-the-winner” might apply to the swine microbiome.
This is the first report of the effect of antibiotics on total phage diversity in a microbial community. The results show that a collateral effect of antibiotic treatment is increased abundance of phage integrase-encoding genes, reflecting the induction of prophages from gut bacteria. Integrases are an appropriate marker for prophage induction because they are required for temperate phages transitioning from lysis to lysogeny (22
). Integrases are also associated with pathogenicity islands, which are often mobilized by prophages (23
). A greater abundance of integrases with antibiotic treatment, therefore, indicates that antibiotics are inducing phage-mediated bacterial lysis in the gut. Integrase abundance increased regardless of the type of in-feed antibiotic; further research is required to determine the specificity of perturbations that result in increased integrase abundance.
Additional consequences of in-feed antibiotic-mediated phage induction could be increased abundance of bacterial fitness or virulence genes, such as those encoding antibiotic resistance. Various homologues of antibiotic resistance genes were detected in the swine viromes at a frequency corresponding to approximately 1/50 of the frequency of antibiotic resistance genes in an Escherichia coli
genome. Particularly in the context of no detection of 16S rRNA genes by PCR in the viromes, the apparent low number of resistance-encoding virome reads seems surprisingly frequent. Resistance genes were identified slightly more frequently in human fecal viromes (0.1% [24
]). With selective pressure, any phage-transferred resistance genes could accelerate the evolution of resistance in the gut microbiome. Despite the potential relevance of the transduction of resistance genes in an antibiotic-containing environment, the swine viromes provided no evidence of a treatment effect. Increased sequencing depth may be required to detect differences among the viromes, such as the effect of generalized transduction on a virome. Additionally, transcriptomic analyses would demonstrate which phage genes have altered expression as a result of antibiotic treatment, revealing those genes important for fitness in an antibiotic environment.
Phages have been shown to play an important role in ecosystem dynamics (25
), and one dynamic is the relationship between phages and bacteria. A widely investigated model for this relationship is called kill-the-winner (26
). This model predicts that an increase in a given bacterial host population (winning prey) results in an increase in its phages (predators) and subsequent predation of the winner. Kill-the-winner population dynamics have been supported in marine ecosystems (27
), but it is unclear if the model holds true for gut ecosystems. The extended sampling (seven viromes over time) of nonmedicated animals’ viromes presented a ripe data set for investigating phage-bacterial population dynamics. Overall, the nonmedicated swine viromes showed taxonomic and functional stability over time, as seen in aquatic microbiomes (27
). Despite this apparent stability, examination of the relative abundances of Streptococcus
phages compared to Streptococcus
bacterial abundances over time revealed a dynamic process resembling kill-the-winner. Indeed, the swine viromes suggest that the kill-the-winner process might be detectable at the phylum level. This is consistent with other work that has shown kill-the-winner dynamics at the strain level in aquatic microbiomes (27
) and horse feces (28
). However, a fecal phage metagenomic study from pairs of twins and their mothers revealed little intravirome change across three sampling dates, and the authors refuted the model (29
). Two major differences between the present study and the twin study are that we isolated phages from fresh (not frozen) feces and that we did not employ a DNA amplification step prior to sequencing. These protocol improvements were designed to reduce bias in analyses of phage diversity (30
), enabling us to view population dynamics even in complex ecosystems such as the swine microbiome. The results tentatively support kill-the-winner dynamics in swine microbiomes, but more research is required to resolve the applicability of the kill-the-winner model across mammalian gut ecosystems.
Analysis of the in-feed ASP250 viromes suggests that there is an antibiotic effect on the relative abundance of fecal phages. The only component of ASP250 known to have an effect on phage lysis is penicillin. Subinhibitory concentrations of penicillin were shown to weaken Streptococcus
spp. such that even so-called phage-resistant strains in mixed cultures were susceptible to phage lysis by exogenous phages (31
). This could account for the significant decrease in Streptococcus
spp. with ASP250, although it provides no evidence for the concomitant increase in Streptococcus
phages. A related phenomenon is called phage-antibiotic synergy (PAS) and has been demonstrated with diverse phages of E. coli
in the presence of subinhibitory concentrations of various cephalosporin-type beta-lactam antibiotics (32
). The result of PAS is phage induction, and it is independent of an SOS response and dependent on a filamentation phenotype resulting from certain antibiotic treatments (32
). Taken together, these data suggest that penicillin is the active component of ASP250 that is affecting the phage population, perhaps by numerous and complex mechanisms in the bacterial milieu.
Analysis of the structure of the bacterial communities shows that a small, important fraction of the data is driving the shift in diversity with ASP250 treatment. The decrease in the lactic acid bacterium (LAB) Streptococcus
is particularly intriguing and agrees with the reported decrease in the abundance of LABs with certain oral antibiotic treatments (33–35
). This decrease is often accompanied by an increase in Proteobacteria
, specifically in Escherichia
populations as shown here and elsewhere (33; Looft et al., submitted). In addition to the immediate effects on the microbiota, oral antibiotic treatment was shown to decrease the immune response in mice, even in distant locations such as the lungs (34
). The interaction of LABs with the gut mucosa is thought to be immunomodulatory (36
), so perhaps there is a connection between the abundance of LABs and immune function. Interestingly, a recent study evaluating in-feed fumaric and formic acids showed a trend towards increased abundance of coliforms and decreased lactobacilli in plate counts (37
), mirroring the effect of ASP250 on the microbiota. Fumaric acid has been demonstrated to improve weight gain despite no changes in available energy in the gut (38
). Furthermore, in a study of irritable bowel syndrome, subjects with a higher body mass index than that of normal subjects had fewer lactobacilli (a type of LAB) (39
). Considering that one mechanism of antibiotic-mediated growth promotion could be suppressed immune response due to decreased bacterial load (40
), a decrease in immunomodulatory LABs might also decrease the energy spent on immunity and allow for increased feed efficiency.
The fecal bacterial diversity in the current study supports what has been shown previously: the swine gut is dominated by Firmicutes
(~50%), and Proteobacteria
(~10%). However, the proportion of assignable phage sequences does not mirror this distribution, with nearly 80% of reads called from phages of bacteria of the Firmicutes
phylum. Their inflation in our data set could result from the overrepresentation of phages of Firmicutes
in public databases compared to those of Bacteroidetes
, perhaps because of increased research interest due to their potential biotechnological applications (41
). Additionally, phage sequences are simply lacking in the databases compared to bacterial sequences, limiting the pool of potential homologues for the swine viromes.
A relatively large proportion of assignable virome sequences were of bacterial origin despite no detectable bacterial contamination of the viromes. Phages harbor more bacterial genes than previously appreciated (42
), making it reasonable that the assignable sequences of the swine viromes have 46% bacterial genes even in the absence of bacterial contamination. Also, generalized transducing phages package host bacterial DNA, contributing an unknown proportion to the counted bacterial genes.
No statistically significant effect of carbadox was detected on swine fecal phages or bacteria. Different sampling intervals and minor protocol improvements between the carbadox and ASP250 experiments (see Materials and Methods) may have affected the results. Our results suggest that 1 week following the commencement of in-feed antibiotics is an appropriate time to detect changes in the fecal microbiome.
This study provides evidence that a collateral effect of some in-feed antibiotics, such as ASP250, is the induction of prophages. Additionally, antibiotic resistance genes were detected in the phage metagenomes. Further work is required to determine the implications of prophage induction on the transfer of antibiotic resistance genes. Surprisingly, the data also support the kill-the-winner model for phage-bacterium population dynamics. Taken together, the data underscore the importance of phages in complex microbial communities.