Colibacillosis, caused by pathogenic Escherichia coli and characterized by diarrhea or systematic infection, is a common bacteriosis and generates huge economic losses in the animal husbandry industry. In the last two decades, researchers have investigated alternative treatment methods for colibacillosis because of the antibiotics residues and the increasing concern over drug-resistant bacterial strains. Among the promising new agents are bacteriophages, which can specifically lyse host bacterial cells and reproduce themselves approximately one 100-fold, thereby preventing and controlling the disease. Several studies demonstrated that a single orally administrated bacteriophage or bacteriophage cocktail reduced the levels of pathogenic bacteria in the gastrointestinal tract of sheep or mice experimentally infected with E. coli (Tanji et al., 2005; Raya et al., 2011) and decreased the rate of diarrhea and the mortality of poultry in farms (Xie et al., 2005).

Rabbits have suffered from colibacillosis caused by enteropathogenic E. coli (EPEC) since the beginning of the 1980s (Milon et al., 1999). We recently isolated an atypical EPEC strain (ZR1) and one of its specific bacteriophages (ZRP1) from rabbits. ZRP1 is a member of myoviridae family and its genome is a 68201-bp dsDNA molecule with a GC content of 46.16%. The result of blast showed that the DNA shared 96% of identity with the Escherichia phage ECML-117 (Query coverage = 93%; E-value = 0.0), which is a lytic bacteriophage with a broad target range (Abuladze et al., 2008) and was proved free of bacterial toxin genes, antibiotic resistance encoding genes, and bacterial 16S rRNA genes (Carter et al., 2012). An in vivo trial with ZRP1 as an intravenous therapeutic phage showed that it can significantly prolong the survival time of rabbits simultaneously received lethal dose of ZR1 (Zhao et al., 2017). Furthermore, we conducted a trial in rabbits to compare the efficacy of orally administered bacteriophage with antimicrobials and to examine the effect of this phage therapy on the intestinal microbiota profiles.

Atypical EPEC causes intestinal lesions characterized by attaching and effacing (A/E), and the intimin encoded by the attaching and effacing gene (eae) is thought to contribute to the attachment and destruction of microvilli (Hernandes et al., 2009). In our study, histopathological features in the ileum were similar to those previously described (Al-Mamun et al., 2013). Normal intestinal barriers are breached by pathogenic E. coli colonization; then bacteria and their products, such as LPS, infiltrate into blood and recruit polymorphonuclear (PMN) leukocytes. Moreover, as reported previously (Gallois et al., 2007; Al-Mamun et al., 2013), growth performances of rabbits are retarded by the infection. And, in our case, the retardation was caused by the decrease in the feed conversion ratio as the feed supply was constant. LPS induces an inflammatory response from animal immune systems by binding to toll-like receptor 4. Intraperitoneal injection of E. coli LPS has been used to stimulate an inflammatory state and trigger the increase in WBCs 24 or 48 h after the injection (Brecchia et al., 2010, 2014). In our study, the elevation of blood endotoxin levels and the increase in WBC number were also observed in infected rabbits. In additional, the rise in W-LCR indicated that neutrophils responded quickly to the infection during the acute phase of inflammation.

Phage therapy predates the discovery of penicillin and has been investigated as an alternative to antibiotics recently. Bacteriophages are harmless to animals, and their efficacy is superior to that of antibiotics in certain respects (Smith and Huggins, 1982; Bull et al., 2002; Zhang et al., 2015). In our experiment, application of ZRP1 successfully reduced the E. coli ZR1 load in the intestine. Either the eaeA gene abundance or the Enterobacteriaceae tag numbers were substantially lower in treated as compared to in untreated rabbits. qRT-PCR is more sensitive and specific than conventional culture method for diagnosis and quantification (Bischoff et al., 2005). Therefore, the eae gene was selected to identify certain EPEC strain in feces using this method (Al-Mamun et al., 2013). Theoretically, the eaeA gene should not have been detected in the control group. One possible reason we detected in our control group is because the rabbits we used in the experiment were non-SPF animals and might have carried very few bacteria with the eae gene prior to the start of the experiments. With the elimination of ZR1 in the gastrointestinal tract, serum LPS concentrations and the W-LCR values decreased accordingly, hence the body weight gain recovered after the phage treatment. Ciprofloxacin lactate was more powerful than we expected in treating ZR1 in vivo. This treatment nearly wiped out the pathogenic E. coli; and a consequent compensatory growth was observed in the rabbits. We assumed that the inferior phage therapy efficacy could be the results of using an insufficient phage dose. The influence of bacteriophage-to-bacterium ratio was reviewed (Hraiech et al., 2015); and a previous study concluded that a larger phage dose was superior in passive treatments (Tsonos et al., 2014). The sensitivity of ZR1 to ciprofloxacin lactate may play a key role as well. The studies we cited above (Smith and Huggins, 1982; Bull et al., 2002; Zhang et al., 2015) did not mention the relationships between the pathogens and the antibiotics, but we chose ciprofloxacin lactate by antimicrobial susceptibility testing. Nevertheless, bacteriophages demonstrate a virtue not apparent in antibiotics: phages self-replicate rather than being metabolized in vivo. Thus, dosing bacteriophages repeatedly is unnecessary (Mandal et al., 2014). In our research, numerous active bacteriophages (>10⁵ PFU/g feces, data not shown) were found 3 days after the treatment. Taken the coprophagy of rabbits into account, these bacteriophages had a great chance to re-enter the digestive tract. Moreover, these bacteriophages might play a role in limiting host E. coli numbers in the environmental reservoir. Thus, the potential benefits of phage therapy may be seen long term.

After birth, rabbits are colonized by a complex and dynamic consortium of microorganisms in the digestive tract; and the bacterial community alteration could be caused by inflammation. In our study, the Venn diagram shows that Groups P, N, and A had eight orders in common. This may be attributed to the altered intestinal environment caused by the experimental infection of E. coli ZR1, creating niches for those bacteria to grow. However, the emergence of 13 particular orders in cecal contents of phage-therapy rabbits is still a mystery.

Using a highly effective sequencing technique, the representation of some cecal bacteria was established. Except Cyanobacteria, four phyla (Firmicutes, Bacteroidetes, Tenericutes, and Verrucomicrobia) in the top five and their relative abundances in the cecum of healthy rabbits were very similar to those previously reported (Bauerl et al., 2014). Likely, in rabbits infected but untreated, the increase of Enterobacteriales, Bacteroidales, and Verrucomicrobiales accompanied by a decrease of Clostridiales were observed, too (Bauerl et al., 2014). The results indicated a negative correlation between Clostridiales and the pathogens when the intestines of healthy rabbit were infected. Nevertheless, Gut bacteria plays an important role in food fermentation and therefore in health status. The rabbit’s cecal microbial composition was dominated by Ruminococcaceae (in the order Clostridiales) which succeeded Bacteroidaceae (in the order Bacteroidales) after weaning to adapt solid feed and such alteration reduced the mortality of weaned kits (Combes et al., 2014).

Studies using PCA for analyzing gut microbiota of rabbits receiving different treatments are very few. PCA is a method to covert a lot of observations into a set of linearly uncorrelated variables. Basically, the closer the dots on the PCA plot are, the more homologous they are. In our study, samples from Groups C and P had PCA plot points that appeared in a cluster. We assumed that the narrow antibacterial spectrum of bacteriophage ZRP1 preserved the existing microbiome. On the contrary, ciprofloxacin lactate, a broad-spectrum third-generation fluoroquinolone, might be harmful to YS2 (in the phylum Cyanobacteria) because ciprofloxacin is toxic to Cyanobacteria (Ebert et al., 2011). Antibiotics are widely used in animal disease treatment and prevention or feed efficiency improvement, but they destroy pathogens and commensal bacteria simultaneously. A short-term antibiotic treatment can shift the population structure of the microbiota and alter bacterial physiology, such as reduction of the amount and diversity of microbes, losses in the function of metabolism and the modulation of the immune system (Ferrer et al., 2017). Therefore, antimicrobial agents with minimal side-effects on commensal bacteria are badly needed (Toutain et al., 2016).

Normally, each individual phage has a very narrow host range that will benefit microbial community by affecting only a relatively small subset of bacteria (Dalmasso et al., 2014). The advantage of phage therapy is to specifically target pathogens while not to perturb the normal microbiota. If the normal gut microbiota is not disrupted in a designed phage therapy, they will provide protection to the gut from disease by stimulating the recovery of the immune response to pathogens (Belizario and Napolitano, 2015; McCarville et al., 2016).

Up to now, little was revealed about the role of phage therapy in shaping the gut microbiota and its possible influence on health. The potential impact of phages on bacterial population also relies on the opportunity of encounters between lytic phages and their host in vivo (Dalmasso et al., 2014). The result of our work suggested that the postulate that phage therapy has little impact on the gut bacteria ecology is well-founded.

Rabbits orally infected with pathogenic E. coli can be cured by a single specific phage therapy. The phage removed nearly all host cells in vivo and had few effects on other bacteria. However, comparing to those received antibiotic treatment; phage-therapy rabbits had a little higher residual ZR1 load in cecum contents at the end of the 3-day trial. It deserves further study using a longer period to see whether phage therapy is as effective as antibiotic treatment.

Conceived and designed the experiments: JZ, YL, HY, and GB. Performed the experiments: JZ, CX, and SH. Analyzed the data: JZ and YL. Contributed reagents/materials: CX. Contributed to the writing of the manuscript: JZ and YL. Revised the manuscript: GB.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.