Effects of phages on the gut microbiota and their impact on probiotic ecacy in broiler chickens

Background Probiotics are the most recognised alternative to antibiotic growth promoters (AGPs) for poultry production. They adhere to the intestinal epithelial cells to exert their benecial effects on the host. However, this is often restricted by indigenous gut microbes that compete for adhesion sites and nutrients. Therefore, phage application has been proposed to reduce the competition of probiotics with gut microbiota. The main objective of this study was to investigate the effects of phages on the gut microbiota and their impact on probiotic ecacy in chickens. Results Four best Escherichia coli lytic phages were selected and characterised. The in vivo trial was conducted to investigate the effects of dietary treatments of phage cocktail at 0.1% (1 (cid:0) ) and 0.2% (2 (cid:0) ), 0.1% PrimaLac ® probiotic (P) and their combinations (1 (cid:0) P, 2 (cid:0) P) on growth performance and gut microbiota of chickens. The results from the in vivo trial showed that chickens in the 1 (cid:0) P group had signicantly (P < 0.05) better body weight (BW), body weight gain (BWG), and feed conversion ratio (FCR), lower serum triglycerides and higher ileal villus height than the control. Chickens in the 1 (cid:0) P group also had higher (P < 0.05) ileal villus height, and lower crypt depth than those in the probiotic group. The gut microbiota proles revealed that 1 (cid:0) P supplementation signicantly (P < 0.05) reduced E. coli population, increased (P < 0.001) short chain fatty acids (SCFAs) producers, and improved (P < 0.001) gene expressions that are related to carbohydrate and amino acid metabolisms, and nutrient uptake. Conclusions The results from this study showed that 1 (cid:0) P treatment could modulate the gut microbiota and enhance growth performance in chickens. This nding could be considered as a potential alternative to AGPs for poultry. This study investigated the effects of dietary treatments of isolated phage cocktail at different dosages and their combinations with probiotic on growth performance, health and gut microbial communities of chickens. The assessment on phages as dietary supplementation at various dosages and their combinations with probiotic in broiler chickens has never been reported. However, similar experiments have been conducted on pigs in which commercial phages and probiotic were used, and phages were found to be more effective [14, 16]. This study showed that chickens fed with 0.1% phage cocktail and probiotic combination (1 (cid:0) P) can positively modulate the gut microbiota based on the signicant reduction of E. coli.

Phage is a bacteria-infecting virus that can multiply by utilising bacteria replication systems, and eventually kill them at the end of the phage lytic cycle [11]. They have a very narrow killing range (e.g., strain, species or rarely genera) [12]. This feature can then be exploited to design a targeted gut microbiota modulation that precludes the need of AGPs, and thus reduce the formation of bacterial resistance [12]. The 'auto-dosage' system is also one of the intriguing characteristics, where more phages will be produced from the initial infection, increasing their effectiveness on eliminating target bacteria [12]. Recently, phage has been studied for modulation of gut microbiota that has shown improvement on performance and health of chickens and swines [8,14,16].
The gut microbiota can be one of the main factors affecting probiotic e cacy in the gut where they can in uence the availability of adhesion sites and nutrients. This is because probiotic need to adhere to the intestinal epithelial cells before they can exert their bene cial effects to the host [9,10,23]. Little is known about the effects of phages on probiotic e cacy in chickens. Therefore, the aim of this study was to investigate the effects of dietary supplementation of isolated phage cocktail at low 0.1% (1 ) and high 0.2% (2 ) dosages, 0.1% PrimaLac® probiotic (P) and combinations of 0.1% PrimaLac® probiotic with 1 (1 P) and 2 (2 P) on chicken growth performance and gut microbial dynamics.

Isolation of bacterial hosts and phages
Both Escherichia coli (E. coli) and phages were originally isolated from ileal and caecal contents of broiler chickens. Four phages (C1, C2, C3 and C4) were selected where they speci cally infect unique E. coli strains.

Phage morphological study by transmission electron microscopy (TEM)
The concentrated and puri ed phage lysates were dropped onto a carbon-coated copper grid and incubated at room temperature for 6 min. The lysates were then negatively stained with 2% (w/v) uranyl acetate and left for another 10 min before being observed under Philips HMG 400 transmission electron microscope (TEM) (Philips, The Netherlands) at magni cation of 300,000 x. The morphology of each phage was analysed, compared and classi ed based on International Committee of Taxonomy of Viruses (ICTV) that has been described in details by Ackermann [24].

Chicken trial (i) Chicken management
A total of 288 one-day-old male Cobb 500 broiler chickens (initial body weight (BW) = 42.95 ± 2.26 g) were obtained from a local commercial hatchery. They were housed in stainless steel three-tiered battery cages with raised wire oors (Dimension: 116 cm width x 89 cm length x 46 cm height) in an open house facility at Animal Research Centre (ARC), Institute of Tropical Agriculture, Universiti Putra Malaysia, Malaysia. The cages were cleaned and disinfected through fumigation beforehand, and strict hygiene and biosecurity measures were practiced throughout the experiment. The feeders and drinkers were cleaned and lled with fresh feed and water daily. The temperature and relative humidity were recorded twice daily in the morning and the afternoon. For the chicken brooding period that lasts for the rst 14 days, lighting from 100 W bulb per cage was provided for each replicates cages. The raised wire oors were covered with newspaper and cleaned daily. Procedures pertaining to chicken management, experimental design, procedures  (ii) Experimental design One-day-old chicks were randomly allotted to 1 of 6 dietary treatments with six replicates cages containing eight chicks per cage. The dietary treatments consisted of: (i) basal diet (BD)(Control, C), (ii) BD + 0.1% phage cocktail (1 ), (iii) BD + 0.2% phage cocktail (2 ), (iv) BD + 0.1% probiotic (P), (v) BD + 0.1% phage cocktail + 0.1% probiotic (1 P), (vi) BD + 0.2% phage cocktail + 0.1% probiotic (2 P). The phage cocktail consisted of freeze-dried powder of C1, C2, C3 and C4 phages at a titre of 10 10 PFU/g each. The PrimaLac® probiotic (Starlabs, USA) used was consisted of Lactobacillus acidophilus, Lactobacillus casei, Bi dobacterium termophilum, Enterococcus faecium and Aspergillus oryzae at a concentration of 10 9 CFU/g each. Both phage cocktail and probiotic for respective treatments were supplemented at the expense of corn to achieve an equal nal percentage for each treatment group. The basal diet formulated for starter (1 to 21 days) and nisher (22 to 35 days) periods were antibiotic-free, in mash form, and ensured to meet or exceed the energy and nutrient requirements as recommended by Nutrient Requirements of Poultry (NRC 1994) for each growing phase. Both feeds and water were provided ad libitum. The phage cocktail and commercial probiotic were mixed with basal feed fresh daily before being fed to the chickens. The formulated starter feed, nisher feed, phage cocktail and probiotic were chemically analysed, based on feed proximate analyses (Malaysian Agricultural Research and Development Institute (MARDI)) on crude protein, crude fat, crude bre, calcium, phosphorus and sodium.

(iii) Data collection and analyses
The chicken body weight (BW) and body weight gain (BWG) were individually recorded on a weekly basis (1 d, 7 d, 14 d, 21 d, 28 d and 35 d), while feed intake (FI) and replicate cages were recorded daily.
Mortality rate was checked and recorded daily. The BWG, feed conversion ratio (FCR and mortality rate were calculated as described by Naidoo and McGaw [25] and Wang and Xu [26]. For every sampling period on day 21 and 35, twelve chickens per treatment (2 chickens per replicate cage) were randomly selected, weighed and euthanised by severing the jugular veins. Blood samples were immediately collected from the jugular veins into 10 ml non-coagulated clot activator tubes (BD, USA), and coagulated K2EDTA tubes (BD, USA) for biochemical and serum immunoglobulins analyses and hematological analysis, respectively. Gastrointestinal tract such as ilea and caeca were cut open, and gently scraped by using sterile microscope slide. The samples were used for gut microbiota study based on real-time quantitative PCR (qPCR) and high-throughput next-generation sequencing (HT-NGS) of 16S rRNA gene amplicons. The ilea tissues were also collected for measurement of the villus height and crypt depth. All samples were kept on ice every time before the respective samples were processed, and frozen in -80°C freezer until analysis.
(iv) Serum lipid The blood samples from non-coagulated clot activator tubes (BD, USA) were processed immediately by centrifugation using JA-20.1 rotor of Beckman Avanti J-25I Centrifuge (Beckman, USA) at 3,000 x g for 10 min. The serum samples were collected and transferred into sterile 2 ml microcentrifuge tubes and used for biochemistry analyses of triglycerides, cholesterol, high-density lipoprotein (HDL) and low-density lipoprotein (LDL) using COBAS reagents (Roche Diagnostics, USA) and analysed using Hitachi 902 Automatic Biochemistry Analyser (Hitachi, Japan). The same serum samples were also stored at -20°C for serum immunoglobulins analysis.
(v) Serum immunoglobulins Serum samples of three chickens from separate cages were selected for IgG, IgA and IgM analyses of every treatment for day 21 and 35 sampling. Serums were diluted at appropriate dilutions and then quanti ed in technical duplicates for each sample. Serum immunoglobulins were quanti ed by using double-antibody sandwich enzyme-linked immunosorbent assay (ELISA) kit of chicken IgG, IgA and IgM (Bethyl Laboratories Inc., USA) by following manufacturer instructions. Absorbance was measured by using BioTek EL800 Absorbance Reader (BioTek, USA) at a wavelength of 450 nm. The standard curve was plotted from the graph of average absorbance measured versus known concentration of chicken IgG, IgA and IgM, to determine and calculate the concentration of unknown samples. The data was then multiplied by the dilution factor used earlier to determine the amount of antibody in the original undiluted sample.
(vi) Ileal villus height and crypt depth The 1 cm midpoint ileum was excised for measurements of the villi height and crypt depths from 12 chickens per treatment (2 chickens per replicate cage) for day 21 and 35 sampling. The tissue samples were rinsed in phosphate-buffered saline (PBS) buffer (pH 7.4) and xed in 10% buffered formalin overnight. There were dried in an automatic tissue processor for 16 h using Leica ASP 3000 fullyenclosed tissue processor (Leica, Japan) and embedded in para n wax by using Leica EG1160 para n embedding station (Leica, Japan). The tissue samples with the thickness of 4 µm were cut using Leica RM2155 microtome (Leica, Japan), placed on glass slides and then stained with haematoxylin and eosin. The sections were examined under Olympus light microscope QG2-32 (Olympus, Japan) with digital camera attached. The villus height was measured by using Fiji ImageJ [27] from the top end of the villus to the bottom of the villus-crypt junction, while crypt depth from the basement membrane to the villuscrypt junction. An average of 3 villus heights and crypt depth were measured for each sample.
Chicken gut microbial dynamics (i) DNA extraction Genomic DNA was extracted by using QIAamp Fast DNA Stool Mini Kit (QIAGEN, Germany) according to the manufacturer instructions with some modi cations. The mucosal contents weighing between 180-220 mg were treated with lysozyme lysis buffer [25 mg/ml lysozyme (Vivantis, Malaysia); 20 mM Tris-Cl, pH 8.0; 2 mM EDTA, pH 8.0; 1% Triton X-100) and incubated for 30 min at 37°C to facilitate the lysis of Gram-positive bacteria. The 1 ml InhibitEX Buffer was added to each sample where it was vortexed continuously for 1 min. The mixtures were heated for 5 min at 95°C and vortexed for 10 s to lyse Grampositive bacteria. Subsequently, the solution was centrifuged at 16,100 x g for 1 min to pellet intestinal debris. The supernatants were treated with 4 µl of 5 µg/ml RNase A (Epicentre, USA) and incubated for 30 min at 37°C to remove RNA. The eluted genomic DNA was then stored in -20°C freezer.
(ii) Real-time quantitative PCR (qPCR) The population of speci c gut microbes which are Escherichia coli, Lactobacillus spp., Bi dobacterium spp. and Clostridium perfringens in ileal and caecal contents at 21 and 35 d were analysed based on absolute quanti cation in real-time qPCR. The E. coli population was quanti ed as it was the phagetarget bacteria of the dietary treatments (1 , 2 , 1 P and 2 P). The Lactobacillus spp. and Bi dobacterium spp. were part of supplemented Primalac® probiotic dietary treatments (P, 1 P and 2 P). The Clostridium perfringens is the common indicator of pathogens.
The genomic DNA from samples of each treatment for ilea and caeca at 21 d and 35 d (three replicate cages (1 chicken per cage) were used as the templates for real-time qPCR study. The procedure was conducted by using SensiFAST SYBR No-ROX Kit (Bioline, UK) in CFX96 Real-Time PCR Detection System (Bio-Rad, USA). The 20 µl reaction mixtures contained 10 µl SensiFAST SYBR No-ROX mix, normalised genomic DNA template at 80 ng/µl, 1 µl of forward and reverse primers (10 pmol/µl) (refer Table 4) and sterile Mili-Q water. Both standard curve samples and no template control (NTC) were prepared in every run. The technical duplicates were prepared for each sample. The PCR cycling reaction included an initial denaturation step at 95°C for 3 min followed by 40 cycles of denaturation at 95°C for 5 s, annealing and extension at 72°C for 20 s. The speci city of amplicons were then veri ed based on the melting curve analysis where they were incubated by increasing the temperature from 65°C to 95°C (0.5°C increment) and hold for 5 s. The nal results were then analysed by using Bio-Rad CFX Manager version 3.1 (Bio-Rad, USA).
(iii) Illumina sequencing of the V3-V4 region of the 16S rRNA gene V3-V4 hypervariable region of 16S rRNA gene was ampli ed using forward primer (5'-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGNNNNCCTACGGGNG GCWGCAG-3') and reverse primer (5'-GTCTCGTGGGCTATAAGAGACA GGACTACHVGGGTATCTAATCC-3') (Integrated DNA Technologies (IDT), Singapore) as per described by Klindworth, Pruesse et al. (2013) with some modi cations. Four degenerate bases (N) were added to maximise the diversity for unique clusters identi cation. The library preparation condition and procedure is given in Table S2. The puri ed amplicons were quanti ed by using Qubit Fluorometer (ThermoFisher Scienti c, USA) and normalised to 2 nM and subjected to Illumina Miseq desktop sequencer by using paired 300 bp reads Miseq Reagent Kit v3 (600-cycle) (Illumina, USA) at Monash University Malaysia Genomic Facility.
(iv) 16S sequence analyses Sequences in FASTQ format were were assembled and quality ltered using Mothur software package (v. 1.38.1) [28]. The processed sequences were subsampled to 9998 prior to alpha diversity analyses. The raw OTUs Table was normalised using 'cumNorm' command. In addition, OTUs with signi cant difference in abundance were selected based on zero-in ated log-normal model using ' tFeatureModel' command. Both 'cumNorm' and ' tFeatureModel' were implemented in metagenomSeq package [29]. The normalised OTUs Table was  Microbial predicted functional metagenomes based on PICRUSt and STAMP. The prediction of metagenome function from 16S rRNA marker gene was performed by using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) version 1.1.0 [37]. Brie y, the Greengenes-based biom le was generated using Mothur software and uploaded into online galaxy PICRUSt (http://huttenhower.sph.harvard.edu/galaxy/). The OTUs table provided was rst normalised for multiple 16S copy number where the genome was then predicted based on Kyoto Encyclopedia of Genes and Genomes (KEGG) ortholog abundances. The output les were then analysed by using Statistical Analysis of Metagenomic Pro les (STAMP) version 2.1.3 bioinformatics software package [30]. The Storey's FDR of multiple test correction methods was used to calculate the statistical signi cance and the features were ltered by effect size at 0.05.
(v) Nucleotide sequence accession numbers The V3-V4 region of 16S rRNA gene sequences from this study have been deposited in the NCBI sequence read archive (https://www.ncbi.nlm.nih.gov/sra) under BioSample Accession numbers of SAMN06027949-SAMN06028092.

Phage morphology
A nal four phages were used for the phage cocktail study. They were observed under transmission electron microscopy (TEM) and their morphologies were analysed. All phages exhibit a hexagon nucleocapsid head structure with a short tail for C1 and a long tail for C2, C3 and C4 phages (Fig. 1).

Chicken trial
Growth performance The effects of phage cocktail at different dosages, probiotic and their combinations on chicken growth performance are shown in Table 1. Chickens supplemented with 1 P had signi cantly (P < 0.05) better BW (35 d) and BWG (22-35 d, 1-35 d) than those of the control, but not chickens of the other supplemented groups (1 , 2 , P and 2 P). There were no signi cant differences in the FI among all groups throughout the experimental period. Chickens of all supplemented groups at 22-35 d and 1-35 d had signi cantly (P < 0.05) lower FCR than the control group. However, at 1-21 d, the FCR of the P group (1.40) was not signi cantly different from that of the control (1.43). The highest mortality rate was recorded in both control and 2 groups at 4.17%, while no mortality was observed in the P group.

Serum Lipid Analysis
Chickens in all supplemented groups, except the 1 group, showed signi cantly (P < 0.05) lower triglycerides than control chickens at 35 d, but at 21 d, only chickens in the1 and P groups had signi cantly (P < 0.05) lower triglycerides than the control. There were no signi cant (P > 0.05) differences in total cholesterol and HDL among all groups of chickens at 21 and 35 d. For LDL, chickens in the 1 and 2 groups showed signi cantly (P < 0.05) lower levels than the control group at 21 d, but there were no signi cant differences among all groups at 35 d. The concentrations of IgA and IgM in chickens were not affected (P > 0.05) by dietary treatments at 21 and 35 d ( Table 2). This was also similar for IgG at 35 d.

Ileal Villus Height And Crypt Depth
There were no signi cant differences between the supplemented treatment groups and the control in villus heights at 21 d, but at 35 d, the 1 P and 2 P groups had signi cantly (P < 0.05) higher villus heights than the 1 , P and control groups. Chickens fed probiotic (P group) had signi cantly (P < 0.05) greater crypt depth than those of the control and other supplemented groups at 21 d (Table 4). At 35 d, there were no signi cant differences between all supplemented groups and control group. The villus to crypt ratio was not signi cantly (P > 0.05) different among all groups of chickens at 21 d. At 35 d, only chickens of the 2 P group had signi cantly (P < 0.05) higher villus to crypt ratio than control chickens. Each value is mean ± SE of 6 replicate cages with 2 chickens each. a,b,c Means within the same row that have different superscripts differ signi cantly (P < 0.05). 1 C = control (basal diet); 1 = BD + 0.1% phage cocktail; 2 = BD + 0.2% phage cocktail; P = BD + 0.1% probiotic; 1 P = BD + 0.1% phage cocktail + 0.1% probiotic; 2 P = BD + 0.2% phage cocktail + 0.1% probiotic.
Signi cant OTUs present in the P compared to the non-P groups The OTUs that were signi cantly expressed in phage cocktail and probiotic combinations treatment groups ( P) were identi ed based on ' tFeatureModel' in the metagenomeSeq package ( Table 8). The value of logFC is directly related to their abundance in the P groups. Out of the top 50 OTUs selected, the most common bacterial genera or species that were signi cantly elevated in P groups, compared to non-P groups were Bacteroides, Odoribacter, Alistipes, Anaerotruncus, Ruminococcaceae, Lachnospiraceae, Ruminococcus, Desulfovibrio, Anaerostipes, Clostridium, Coprobacillus, Butyricimonas, Faecalibacterium prausnitzii and Oscillopira. Note: LogFC is directly related to their abundance in P group, where the highest number showed the highest abundance. For phage cocktail and probiotic combinations groups ( P; 1 P and 2 P) and other groups (Non-P; C, 1 , 2 and P).

Microbial predicted functional metagenomes
The principal component analysis (PCA) of predicted functional metagenomes study corroborated the earlier ndings that the P groups had distinctive gut microbial communities compared to those of non-P

Discussion
This study investigated the effects of dietary treatments of isolated phage cocktail at different dosages and their combinations with probiotic on growth performance, health and gut microbial communities of chickens. The assessment on phages as dietary supplementation at various dosages and their combinations with probiotic in broiler chickens has never been reported. However, similar experiments have been conducted on pigs in which commercial phages and probiotic were used, and phages were found to be more effective [14,16]. This study showed that chickens fed with 0.1% phage cocktail and probiotic combination (1 P) can positively modulate the gut microbiota based on the signi cant reduction of E. coli.
The selection of phage-target bacteria was based on the previous study where both genera of Escherichia and Shigella were observed at a moderate concentration in ilea (9-32%) and caeca (1-5%) at different ages [34]. E. coli was selected as the target bacteria due to their consistent representations in various ages in ilea and caeca, the ease of handling (e.g., aerobic, fast growing bacteria), and also due to their non-pathogenic nature for some of the strains.
Chicken has a range of body temperature between 40.5 to 42°C [59]. The survivability of phage within this temperature is required to ensure the e cacy of phage application. In the current study, all phages had an optimal lytic activity at 37°C. Although phage titre for all phages remained high at a high temperature, phage lytic activity (host reduction) was reduced (at 42°C compared to at 37°C). This observation suggested that phage activity will still persist at 42 °C, but phage may take a longer time to clear targeted hosts. The use of an additive such as skim milk might be useful to preserve the phage activity at a high temperature, where it has been demonstrated in Lactococcus phage [53]. At a lower temperature, phage lytic activity was slightly reduced but was stable at 25°C, especially for C2, C3 and C4 phages when compared to C1 phage. This showed that they were stable at room temperature. The E. coli host was nulli ed at 60°C and hence replications of phages were not possible.
The results of the present study showed that chickens fed with a combination of 0.1% phage cocktail and probiotic (1 P) had signi cantly better BW (35 d), BWG (22-35 d, 1-35d), and lower FCR compared to the groups (Fig. 5). The supplementation of phage cocktail and probiotic combination resulted in changes of gut microbial dynamics. The predicted functional metagenomes by STAMP analysis were performed based on the previous results from multivariate analysis of gut microbiota diversity which has shown signi cant differences between gut microbiota dynamics between P and non-P groups (Fig. 5). The predicted functional metagenomes of ltered Storey's FDR multiple test correction analyses showed signi cant differences between P and non-P groups in metabolic pathways. The pairwise comparison of the microbial predicted functional metagenomes showed that 14 out of 21 KEGG features were signi cantly higher in P, while another 6 were signi cantly higher in the non-P groups (Fig. 6). The metabolic pathways of P groups that were related to carbohydrate (e.g., fructose and mannose metabolism), amino acids (e.g., amino sugar and nucleotide sugar metabolism) and tyrosine metabolisms were signi cantly elevated compared to those of non-P groups.
control. However, the BW (35 d) and BWG (22-35 d, 1-35 d) of chickens supplemented with phage cocktail singly (1 , 2 ) or probiotic (P) singly, or a combination of 0.2% phage cocktail and probiotic (2 P) were not signi cantly different from those of control chickens. This showed that the use of 0.1% phage cocktail with a combination of probiotic (1 P) has the potential to improve the growth performance of broiler chickens. Although supplementation of phage cocktail alone did not signi cantly improve growth performance in broiler chickens, studies on laying hens reported that supplementation of phage cocktail signi cantly increased their growth performance and egg production [70]. Studies in pigs had also shown that supplementation of phage cocktail singly or in combinations with probiotic signi cantly improved the growth performance [14,15].
The results of the present study revealed that 0.2% phage cocktail (2 ), probiotic (P) and a combination of 0.1% or 0.2% phage cocktail and probiotic (1 P, 2 P) signi cantly reduced serum triglycerides in chickens, but the reduction was not signi cantly different between the treatments. This suggests that the addition of a phage cocktail to the probiotic did not further reduce the serum triglycerides level. The reduction of serum triglycerides by probiotic has been widely reported in chickens [5,65,31,66,38], but there is very little information on the effects of phages or their combinations with probiotic on serum triglycerides.
The exact mechanism(s) on the reduction of triglycerides by probiotic is still not well understood. Taranto [67] suggested that reduction of triglycerides in mice supplemented with Lactobacillus reuteri could be due to hypolipemic effect, which involved the reduction of lipid assimilation or increase of lipid catabolism. Santoso [65], on the other hand, suggested that reduction of triglycerides by Bacillus subtilis in broiler chickens was due to their ability to decrease the acetyl coenzyme A carboxylase activity, which has a crucial function on restricting the synthesis of fatty acids. As the fatty acids decrease, it will in turn reduce the triglycerides esteri cation process [38].
None of the dietary treatments in the present study had any signi cant effects on serum total cholesterol, HDL and LDL of chickens at 35 d, though LDL was signi cantly reduced by 0.1% and 0.2% phage cocktails at 21 d and the serum total cholesterol was numerically the lowest in the 1 P group. This is rather surprising as many studies have shown the ability of probiotic to reduce serum cholesterol in broiler chickens [5,38,44,35].
Immunoglobulins are antibodies that are released by plasma cells in response to foreign materials with antigens such as bacteria or viruses. Chickens immunoglobulins can be classi ed into three main classes: IgA, IgY (IgG) and IgM [47]. IgA is found mainly in bile and intestines of chickens [47]. IgG is the most dominant immunoglobulins found in chickens and it carries out an important function of protecting the host from infection [47]. IgM, on the other hand, is present in the rst week post-infection and also appears during immunisation [64].
In the current study, the IgA and IgM concentrations of chickens were not signi cantly different among all the dietary treatment groups. However, IgG was signi cantly lower in the 1 , 1 P and 2 P groups when compared with the control at 21 d. The current study also showed that there were no signi cant differences in IgG between the probiotic (P) group and the rest of the supplemented groups. This showed that supplementation of probiotic singly or in combination with a phage cocktail did not improve antibody titres. These ndings are in contrast to those of other studies, in which probiotic either enhance [63,39,40] or have no effects [36] on the chicken humoral immune response. Bene cial effects of probiotic in modulating immune response at the cellular and humoral levels have also been reported in chickens [56]. Immunomodulatory activities of probiotics can be in uenced by various factors such as type of probiotic strains, host genetics and types of antigens present [39,62]. There were also no signi cant effects on immunoglobulin levels when phage cocktail was fed to pigs [15]. Probiotics are known to enhance immunomodulatory activities to the host by improving antibody response, but this was not observed in this study. It is however important to note that Mountzouris [36] had pointed out that studies which highlighted on enhancement of antibody response with probiotic supplementation [63,39,40] were performed against speci c model antigens. The immunoglobulin levels in this study showed a general overview of systemic antibody response with supplementation of different diets, but not against speci c antigens.
Phage supplementation has been known to initiate an immune response in chickens. This had been demonstrated by Huff [13] who reported that chickens supplemented with phages had signi cantly higher IgG due to activation of immune responses. Repetitive supplementations of phages would lead to an increase in mortality of birds challenged with pathogenic E. coli [13]. This shows that repetitive doses of phages would limit phage action due to the immune response that was initiated. In the current study, signi cantly lower IgG was observed in birds supplemented with phage cocktails or their combinations with probiotic (1 , 1 P and 2 P), as compared to the control at 21 d. This indicated that IgG was not increased even with ad libitum supplementation of phages, suggesting that supplementation of very high titre of phage cocktail at 10 10 PFU/g might have overwhelm the host antibody response on the phages.
However, this can only be con rmed by measuring the antibodies at short intervals after supplementation of phages.
A study of intestinal villus height and crypt depth would provide valuable information on the changes of chicken intestinal epithelial cells in response to supplementation of phage cocktail, probiotic and their combinations. An increase in villus height is associated with greater nutrient absorption due to the larger surface area for e cient nutrient uptake [55]. Villus crypt is recognised as an important site for generation of intestinal villus [69]. A shorter villus is correlated with accumulation of toxins and a high rate of tissue turnover of villus production due to pathogens, while deeper crypt is associated with toxins that cause in ammation or sloughing [69,52,33].
In the current study, supplementation of phage cocktail and probiotic combination to chickens signi cantly improved in villus height (35 d) [46] have reported that villus height was signi cantly increased in the jejunum and ileum, but not duodenum in probiotic-supplemented chicken. Aliakbarpour [51] also found that Lactobacillus supplementation increased the intestinal villus height, but this was not demonstrated in chickens that were supplemented with Bacillus subtilis. To date, there has been no previous study on the effect of phage supplementation on intestinal villus height and crypt depth study. These ndings show that the effects of probiotic on intestinal villi and crypts are in uenced by the types of probiotic used and the part of the intestinal studied.
E. coli population was signi cantly lower in chickens fed with phage cocktail and probiotic combination (1 P) than the control in ilea of 21 d chickens and also lower in phage cocktail singly groups (1 and 2 ), although this was not signi cant. This showed that E. coli phage cocktail employed was potentially effective in reducing E. coli, especially in combination with probiotic. Based on the current study, it was di cult to deduce the phage cocktail alone which was responsible for the gut modulation as the concentration of phage titer was not quanti ed. The caveat is due to the challenge in quantifying C2, C3 and C4 phages as they can cross-infect between each target bacterial hosts. Nevertheless, the qPCR results showed that the E. coli population in the phage singly group was signi cantly reduced, compared to the control. This result potentially suggests that phage cocktail and probiotic worked synergistically in ablating E. coli in the gut. The milder reduction of E. coli population was also detected in the 1 P group at 21 d caeca and 35 d caeca than the control, although this was not signi cant. The lack of difference may be explained by the fact that gut microbiota became more complexed as chicken aged, thus was more resistant to modulation. Indeed, several studies have shown that gut modulation was effectively performed in chickens of young age between 3 to 25-day-old, where these chickens still harboured transient gut microbiota than the adult chickens that had complex gut microbial communities [45,50,68].
Furthermore, E. coli population was abundant in ilea compared to caeca, where it was expected that greater reduction of the bacterium was mainly detected in ilea [34].
The PrimaLac® probiotic used in this study was comprised of Lactobacillus acidophilus, Lactobacillus casei, Bi dobacterium termophilum, Enterococcus faecium and Aspergillus oryzae. In the current study, gut population study based on real-time qPCR showed that there was no signi cant different for Lactobacillus spp. population between dietary treatments for all sampling (age and part of intestine). The populations of Lactobacillus spp. and Bi dobacterium spp. were expected to be higher in probiotic supplemented groups (P, 1 P and 2 P), but Bi dobacterium population was found to be higher in chickens fed with phage cocktail singly instead (1 and 2 ). This suggests that the indigenous gut commensals replaced the niche created by the clearance of E. coli in phage cocktail singly groups. However, this was not observed in probiotic supplemented groups. It could be due to the ine cacy of PrimaLac® probiotic to colonise in the intestines. This result was in contrast with the previous studies, where probiotic supplementation was found to increase the level of probiotic bacteria in chicken intestines [36,49,54]. It also need to be highlighted that in the current study, PrimaLac® probiotic microbes such as Enterococcus faecium and Aspergillus oryzae were not quanti ed, where their populations might be elevated in probiotic supplemented groups.
In the current study, chickens from probiotic supplemented groups (P, 1 P and 2 P) had signi cantly lower population of C. perfringens compared to the control at 21 d caeca. This result was consistent with previous studies, in which C. perfringens population was signi cantly lower in chickens supplemented with probiotic [49]. This was due to the protective roles of probiotic in reducing pathogens based on the reported mode of actions of competitive exclusion [48,42] and antagonistic activity [44,43,32,60].

Conclusions
Chicken gut microbiota was effectively modulated with supplementation of both E. coli phage cocktail and probiotic. Speci cally, an increased in bene cial bacteria such as SCFAs producers was detected.
Such positive modulation might facilitate carbohydrate and amino acid metabolism, and nutrient uptake, which provides energy for the chicken growth. This is the rst study that demonstrated a signi cant gut modulation using a combination of E. coli phage cocktail and probiotic. A further study is warranted to understand the mechanism involved in the gut modulation based on phage cocktail and probiotic combination. UPM/IACUC/AUP-R101/2015).

Consent of publication
All authors critically revised the manuscript for important intellectual contents and approved the nal manuscript.

Availability of data and material
Not applicable

Competing interests
The authors declare that they have no competing interests.  coli and Lactobacillus between treatments. Bi dobacterium were the most abundant in probiotics (P) group. C. perfringens were signi cantly lower in C, 2 and 1 P group. (b) There were no signi cant differences observed for E. coli, Lactobacillus, Bi dobacterium and C. perfringens between treatments.
Each bar is mean of 3 replicate cages with 1 chicken each. Error bar represents standard error (SE). a,bBars that have different superscripts differ signi cantly (P < 0.05). C = control (basal diet); 1 = BD +  Principal component analysis (PCA) of the predicted functional metagenomes based on P versus non-P groups. For phage cocktail and probiotic combinations groups ( P; 1 P and 2 P) and other groups (Non-P; C, 1 , 2 and P).

Figure 6
Pairwise comparison of the predicted functional metagenomes using Storey's FDR multiple test correction methods based on P versus non-P groups. For phage cocktail and probiotic combinations groups ( P; 1 P and 2 P) and other groups (Non-P; C, 1 , 2 and P).