Retraction Note: Antibiotic growth promoters virginiamycin and bacitracin methylene disalicylate alter the chicken intestinal metabolome

Editor's Note: this Article has been retracted; the Retraction Note is available at https://doi.org/10.1038/s41598-021-82285-2.

In a mouse model of antibiotic growth promotion, administration of dietary antibiotics altered the composition and metabolic capability of the gut microbiota by selecting for bacterial species capable of metabolizing complex carbohydrates to short-chain fatty acids, thus extracting a higher proportion of available calories for energy expenditure 31 . Subsequently, Cox et al. 32 reported that exposure of mice to antibiotics early in life induced long-term metabolic effects by accelerating the development of a normal, age-related microbiota. However, definitive linkage of particular gut bacterial populations to intestinal metabolic changes remains to be established 33 . The current study was undertaken to characterize the combined host-and microbiome-derived metabolic alterations in the chicken gut following dietary antibiotic supplementation to identify potential chemical metabolites that might be used in lieu of dietary antibiotics to improve poultry growth performance.

Results
Effect of dietary antibiotics on broiler growth performance. Dietary supplementation with 20 g/ton of the broad spectrum antibiotic, virginiamycin, increased chicken body weight gain by 10.1% between days 0 and 21 of age compared with chickens fed an unsupplemented diet (p < 0.05). Similarly, chickens fed a diet containing 50 g/ton of the narrow spectrum antibiotic, bacitracin methylene disalicylate, had 7.9% greater body weight gain compared with birds given an unsupplemented diet (p < 0.05).

Effect of dietary antibiotics on intestinal global metabolite levels. A total of 706 biochemicals
were identified in the intestinal contents of chickens fed an unsupplemented, control diet, or a diet supplemented with virginiamycin or bacitracin methylene disalicylate. In the virginiamycin vs. control groups, the levels of 156 chemicals were increased and 62 were decreased; in the bacitracin vs. control groups, 96 chemicals were increased and 23 were decreased; in the bacitracin vs. virginiamycin groups, 43 chemicals were increased and 36 were decreased; and in the control vs. both antibiotics groups, 132 chemicals were increased and 46 were decreased. Table 1 lists the Random Forest Analysis (RFA) data for metabolite signatures and biochemical importance of the 30 most statistically significantly altered metabolites for distinguishing the virginiamycin vs. control, bacitracin vs. control, and virginiamycin vs. bacitracin groups. RFA of the virginiamycin vs. control groups gave a predictive accuracy of 85.7%, while that of bacitracin vs. control groups was 78.5%. Among 7 samples tested from each dietary group, 6 samples from both the virginiamycin and bacitracin groups were predicted to belong to their respective group, while the remaining sample was predicted to belong to the control group. Of 7 control group samples, one was predicted to belong to the virginiamycin group and two were predicted to belong to the bacitracin group. By contrast, RFA of the virginiamycin vs. bacitracin groups gave a predictive accuracy of 65.0%, suggesting that when compared with each other, dietary supplementation with either antibiotic produced a less characteristic biochemical signature compared with the antibiotic vs. control comparisons. Among the biochemicals classified as the most biochemically important for distinguishing between the 3 dietary groups, metabolites of amino acids (33.0%), fatty acids   Fatty acids and their metabolites also contributed to the biochemical signatures separating chickens given the antibiotic-supplemented diets, particularly the bacitracin-supplemented group, from unsupplemented controls (Fig. 2B). Many long chain saturated and polyunsaturated fatty acids, as well as several lysophospholipids, were increased in the bacitracin vs. control groups. Most notable in this comparison were oleate/vaccinate (18:1) (2.96-fold increase), eicosapentaenoate (2.55-fold increase), 16-hydroxypalmitate and stearate (both 2.42-fold increases), arachidate (2.39-fold increase), 10-nonadecenoate (2.30-fold increase), palmitate (2.24-fold increase), and 3-hydroxylaurate (1.51-fold increase).

Discussion
Virginiamycin and bacitracin methylene disalicylate are common growth enhancers used in the poultry industry. Virginiamycin is a streptogramin antibiotic produced by Streptomyces virginiae as a mixture of two macrocyclic lactone peptolides, virginiamycin M and virginiamycin S, both of which bind to the bacterial 50 S ribosomal subunit to synergistically inhibit protein synthesis 4 . Virginiamycin M is a polyunsaturated cyclic peptolide while virginiamycin S is a cyclic hexadepsipeptide 34 . Dietary supplementation of chickens with virginiamycin decreased intestinal colonization by Clostridium perfringens 35 , and decreased the severity and mortality due to necrotic enteritis caused by C. perfringens 36 , both compared with unsupplemented controls. Bacitracin is a mixture of more than 10 related cyclic peptides produced by Bacillus subtilis and B. licheniformis that disrupt bacterial cell wall synthesis by inhibiting dephosphorylation of lipid pyrophosphate 4 . Dietary supplementation of chickens with bacitracin reduced gut colonization by C. perfringens and Enterococcus faecalis 19,37 , but increased the number of Salmonella enterica, compared with unsupplemented controls 38 . Compared with chickens fed an unsupplemented diet, intestinal microbiome analyses of chickens fed virginiamycin-and/or bacitracin-supplemented diets have generally revealed a decreased in microbial diversity, with an increase in Enterococcus and Lactobacillus spp., although a decreased frequency of L. salivarius has been noted 19,20,24,25,27,28 . Other investigators have reported an altered bacterial composition, but no change in gut microbiome richness or diversity, associated with virginiamycin-or bacitracin-supplemented diets, compared with antibiotic-free diets 22,39 .
The levels of amino acid metabolites, particularly those of lysine and tryptophan, were substantially altered by dietary supplementation with virginiamycin or bacitracin methylene disalicylate. Tryptophan is metabolized by two major pathways, either through kynurenine leading to niacin and associated cofactors, including nicotinamide adenine dinucleotide (NAD), or through a series of indole-related compounds leading to serotonin and melatonin. Dietary supplementation with either virginiamycin or bacitracin methylene disalicylate increased the levels of kynurenine, as well as its metabolites, kynurenate and quinolinate, in the chicken gut. Kynurenine and kynurenate play important roles in the regulation of inflammation and the adaptive immune response, as well as multiple neurological pathways 40,41 . Increased activity of the kynurenine pathway is internally consistent with decreased levels of indolelactate and serotonin following antibiotic supplementation. Serotonin (5-hydroxytryptamine) receptors are found throughout the intestinal ileum and associated smooth muscle 42 . In the small intestine, serotonin enhances the rate at which intestinal contents move through the digestive system. Increased body weight gain in antibiotic-supplemented diets might be related, in part, through decreased serotonin levels leading to increased residence time and absorption of intestinal nutrients. Consistent with the increased levels of quinolinate following antibiotic supplementation, nicotinamide metabolism and the NAD biosynthetic pathway were also shown to be increased in chickens given the virginiamycin-or bacitracin-containing diets. Interestingly, NMN, which is produced from nicotinamide, was decreased in virginiamycin-supplemented birds, with a trend for decreased levels following bacitracin supplementation, suggesting that nicotinamide might be shuttled to nicotinate biosynthesis. Indeed, animals in the virginiamycin-supplemented group had significantly increased nicotinate levels compared with unsupplemented controls.
One of the most striking features of the current dataset is the increase in levels of many long chain fatty acids, particularly polyunsaturated fatty acids (PUFAs), in the intestine of bacitracin-supplemented, but not virginiamycin-supplemented, chickens. PUFAs are not commonly found in bacteria, and while chickens can synthesize PUFAs from dietary linolenate and linoleate, much of the PUFA content in chicken tissues is thought to originate from ingested sources 43 . Increased levels of PUFAs in the ileum of bacitracin-supplemented birds, therefore, might be the result of decreased intestinal absorption. PUFAs are important as substrates for inflammatory and anti-inflammatory fatty acids, such as the prostaglandins, leukotrienes, and thromboxanes 44 . Omega-3 fatty acids with a C=C double bond at the third carbon atom from the end of the carbon chain, such as eicosapentaenoate, are thought to have more anti-inflammatory properties, while omega-6 fatty acids, such as arachidonate, contribute to inflammatory reactions 44 . Increased levels of eicosapentaenoate following antibiotic supplementation in the current study lends support to the non-antibiotic, anti-inflammatory theory of antibiotic growth promotion 14 . Those growth-related metabolites are shown in Kyoto encyclopedia of genes and genomes pathway (KEGG) and human metabolome database (HMDB), further studies are required to summarizing important/ abundant metabolites pathway in chickens. In summary, this study compared the metabolome profiles of the intestinal contents of chickens fed an unsupplemented diet with animals given a diet containing the antibiotic growth promoters virginiamycin or bacitracin methylene disalicylate. The results demonstrated that antibiotic supplementation had profound effects on the levels of a wide variety of chemical metabolites, particularly amino acids, fatty acids, nucleosides, and nicotinamide-related compounds. Further, these altered metabolite levels provided a biochemical signature unique to each antibiotic supplementation group when compared with unsupplemented controls. Future investigations of the chemical compounds identified in this study might provide new approaches to enhance food animal growth without the use of antibiotics.

Methods
Animals and ethics statement. Forty-five-day-old commercial broiler chickens (Ross/Ross, Longenecker's Hatchery, Elizabethtown, PA) were housed in electrically-heated battery starter cages (Petersime, Gettysburg, OH). Chickens were raised in starter cages until 14 days of age and transferred to finisher cages where they were kept until the end of the experimental period. Feed and water were provided ad libitum. Animal husbandry followed guidelines for the care and use of animals in agricultural research 45 . All experimental protocols were approved by the Small Animal Care Committee of the Beltsville Agricultural Research Center.
Experimental diets and intestinal metabolomics analysis. Chickens (n = 15/group) were fed from hatch with a corn-and soybean meal-based unsupplemented, basal diet (control) formulated to meet or exceed the National Research Council's nutrient requirements for broiler chickens 46 , or the basal diet supplemented with 20 g/ton (22 ppm) virginiamycin (Phibro Animal Health, Teaneck, NJ) or 50 g/ton (55 ppm) bacitracin methylene disalicylate (Zoetis, Durham, NC) ( Table 2). Body weights and feed conversion ratios were measured daily until day 21. At 3 weeks of age, 7 chickens/group were euthanized by cervical dislocation and the intestinal ileum harvested. Ileal content was collected by gently fingers-stripping the ileal segment. Intestine contents were collected aseptically, immediately placed on dry ice, and stored at −80 °C. Global metabolomic profiling of the intestinal contents was performed by mass spectrometry (MS) (Metabolon, Durham, NC) as described [47][48][49] . Raw data was extracted and processed using the DiscoveryHD4 ™ global metabolomics platform. Compounds were identified by comparison to library entries of purified standards or recurrent unknown entities based on retention index, accurate mass match to the library ± 10 ppm, and MS/MS forward and reverse scores between experimental data and authentic standards. MS/MS scores were based on comparison of the ions present in the experimental spectrum to the ions present in the library spectrum.

Statistical analysis.
A two-tailed Student's t-test was used to compare body weight gains and feed conversion ratios of chickens fed the unsupplemented and virginiamycin-and bacitracin methylene disalicylate-supplemented diets. ANOVA was used to identify the biochemicals whose levels were significantly altered among the three dietary groups (virginiamycin vs. control, bacitracin methylene disalicylate vs. control, virginiamycin vs. bacitracin methylene disalicylate) following median scaling, log transformation, and imputation of missing values, if any, with the minimum value observed for each compound. Standard statistical analyses of log-transformed data were performed using Array Studio software (OmicSoft, Cary, NC). For analyses that were not standard in Array Studio, the programs R (R Foundation for Statistical Computing, Vienna, Austria) or JMP (SAS Institute, Cary, NC) were used. Changes in biochemical levels with p ≤ 0.05 were considered statistically significant. An estimate of the false discovery rate (FDR) was obtained by calculating the q-value to account for the false positives that normally occur in metabolomics-based studies. Random Forest Analysis (RFA) was performed by computing the Mean Decrease Accuracy (MDA) as a measure of biochemical importance to a classification.