The time-course of broiler intestinal microbiota development after administration of cecal contents to incubating eggs

Animal trials

Two identical trials were conducted eight weeks apart. For each of the trials, Cobb 500 fertile eggs (total 150) were obtained from the Baiada Hatchery (Willaston, South Australia) from the same breeder stock, and incubated for 18 days in a single incubator (IM Incubators model IM288; 38 °C, 55% relative humidity, turning hourly). Fertile eggs were allocated by weight into three inoculation treatment groups and transferred to three separate incubators (Intensive Farming Supplies, Cavan, SA 5094, model MPS24 A; 36.7 °C and 66% relative humidity). On the day before hatching, 12–15 eggs at a time were removed from the incubators, wiped with 70% ethanol then swabbed with (1) control PBS solution, (2) cecal contents inoculum 1, diluted in PBS, and (3) cecal contents inoculum 2, diluted in PBS.

The inocula for both cecal content transfer trials were selected from our previous trial (Crisol-Martinez et al., 2017). The donors were selected from four birds with relatively high FCR for inoculum 1, and four birds with low FCR for inoculum 2. The cecal samples selected for use in the inocula were also selected based on the previously determined microbiota composition with the cecal samples used in inoculum 2 having particularly high levels of Bacteroides. Thus the two inocula used were derived from birds with different FCR performance and distinctly different micorbiotas. Each inoculum was prepared by mixing equal weights cecal content from each of four birds and diluting in sterile PBS. The inocula (0.3 ml) were painted onto the upper half of the shells (blunt end up) of each egg using a cotton bud. Eggs were returned to the incubator after less than 10 min.

Newly hatched chickens were weighed individually then transferred in inoculation treatment groups to three separate floor pens in a temperature controlled room (day 0). There was no difference in mortality or hatchability between the three groups. There was physical separation between treatment groups at all times and staff wore disposable gloves when handling chickens, feeders and drinkers. All chickens received a proprietary diet formulated for this breed comprised of 44.4% wheat, 17% soybean meal, 15% barley, 10% canola meal, 5% peas, 3.2% meat meal, 3% tallow, 1% limestone, 0.5% vitamin mix, and traces of salt, lysine HCl, DL-methionine and threonine. The feed was antibiotic and ionophore free. There was no specific coccidiosis control method used. The same batch of commercially prepared starter crumbles (Ridley Agriproducts, Murray Bridge, South Australia) was used in both trials and was stored under controlled cool and dry conditions for the duration of the trials. Birds were fed ad libitum and had access to water from drinking nipples at all times. Birds had 23 h of light for the first three days and then 12 h of light for the remainder of the experimental period.

At 15 days of age, the birds were transferred in pairs to metabolism cages in a temperature controlled room. The initial pairing was done to minimise stress and allow the birds to adjust to cages. At day 17, birds were moved into individual metabolism cages. Individual caging allowed the precise assessment of individual feed intake, energy in feed, and unused energy remaining in feces. The experimental design eliminated competition for feed and reduced behavioural issues affecting feed intake. Single bird caging and individual measurements and sampling were implemented in order to allow direct correlation of microbiota structure and productivity measurements on a bird by bird basis.

The apparent metabolisable energy (AME) values of the commercial broiler diet were determined in a classical 7-day AME study involving measurements of total feed intake and total excreta output and subsequent measurement of gross energy (GE) values of feed and excreta by isoperiobol bomb calorimetry. AME in MJ/kg dry matter, was calculated as (AMEdiet  = [(GEdiet × feed eaten) × (GEexcreta × dry excreta)]/feed eaten/dry diet content) (Stanley et al., 2013a).

Feed intake (FI) was measured during the adaptation and collection phases of the study. All excreta were collected daily during the 4-day collection phase and dried overnight at 90 °C. Birds were weighed at the start and end of the 7-day period then retained in the cages until 27 days of age. Feed consumption was measured by weighing the feed at the start (day 17) and end (day 24) of the AME period. On day 24, feed was weighed into hoppers and the residue weighed on day 27 in order to provide feed intake and feed conversion ratio data up until the time the experiment was terminated. All of the above measurements were taken from day 17 to day 24, during the time when single birds were housed in metabolic cages. The performance analysis was completed on both trials separately, however sequencing was performed using the samples from the trial 1. Fecal samples (via cloacal swab) were collected at multiple time points from day 0 to day 23 and frozen immediately for DNA extraction. Although all birds were sampled at each time-point, due to lower fecal content some swabs yielded too little DNA to be sequenced. A total of 160 samples were successfully sequenced in this study. All 96 birds were killed by cervical dislocation. The sex of each chicken was determined by visual observation of gonads when dissected.

DNA preparation, sequencing, and data analysis

DNA was prepared, sequenced and analysed as detailed by Stanley et al. (2013a). The primers used to PCR amplify the V1–V3 region of the 16S rRNA gene were (forward primer, 5′ AGAGTTTGATCCTGG 3′; reverse primer, 5′ TTACCGCGGCTGCT 3′). Sequencing was performed using a Roche/454 FLX+ Genome Sequencer and Titanium chemistry. Sequences were analysed using QIIME (Caporaso et al., 2010), denoised and error corrected with Acacia (Bragg et al., 2012) and chimera checked using Pintail (Ashelford et al., 2005). The sequences were then quality trimmed with length from 300 to 600 bases, no ambiguous nucleotides and a maximum of 6 nucleotides in homopolymer runs. OTU picking was done at 3% divergence level using the UCLUST (Edgar, 2010) algorithm. OTUs represented with less than 10 sequences and present in less than 5 samples were filtered out of the analysis. QIIME generated abundance table based data were further analysed and visualised using Primer-E (http://www.primer-e.com/) and Calypso (Zakrzewski et al., 2017). The sequencing data is publicly available on MG-RAST database under library number mgl564762 and a project ID mgp21821.

Statistical analysis

SAS for Windows version 9.4 software package (SAS Institute Inc., Cary, NC, USA) was used to determine whether data were normally distributed (Univariate procedure; Shapiro–Wilk test), then two-way analysis of variance with the General Linear Model (GLM) procedure was used to examine the factors inoculation treatment and sex of chicken, and the interaction between treatment and sex. Significant differences between treatments and sex were determined by Duncan’s Multiple Range Test. Microbiota statistical analysis was done in Calypso (Zakrzewski et al., 2017) for Richness, Evenness, ANOSIM, ANOVA, network, correlations and LEfSe. UniFrac matrixes were calculated in QIIME (Caporaso et al., 2010) and PERMANOVA was done in Primer-E (http://www.primer-e.com/).

Animal ethics statement

Animal ethics approvals were obtained from the University of Adelaide (S-2013-149) and Primary Industries and Regions South Australia (14/13).

Live weight, feed intake, feed conversion and metabolisable energy

The effects of inoculation treatments and sex of chicken on live weight, feed intake, feed conversion ratio and apparent metabolisable energy of the diet are summarised in Tables 1 and 2. Inoculation treatment 1 significantly reduced live weight at 15 and 22 days of age compared with the control, but had no effect on feed intake, feed conversion, or apparent metabolisable energy value of the diet in the first experiment (Table 1). In contrast, in the second experiment, inoculation treatments had no effect on live weight, feed intake, and feed conversion, but appeared to have a small but significant effect on apparent metabolisable energy with inoculant 2 depressing AME value by 1.6% for male chickens only (Table 2). In both experiments, male chickens were heavier, ate more feed, and converted more efficiently than female chickens, but there were no significant differences in AME values due to gender alone.

Cecal transfer

The genus level cecal microbiota profiles of control and the two cecal inocula treated bird groups, compared to their respective donor inocula are shown in Fig. 1A. The high performing bird derived inoculum 2 was dominated with Bacteroides that were absent in the birds of poor performance represented in inoculum 1. However, Bacteroides, an obligate anaerobic genus, was not transferred to the recipient birds. The presence of this dominant genus and its absence in other groups resulted in clear distinction of the microbiota in inoculum 2 group of donor birds, away from all other groups, including inoculum 1 (Fig. 1B). Based on microbiota profiles and the lack of significant difference in FCR between the control and either of inoculum treated groups (Fig. 1C), under the present conditions and with the current microbiota profiles used as inoculum, FCR-based performance was not affected in the recipient birds.

Microbiota composition of cloacal swabs

The birds were sampled at multiple time points from day 0 to day 23. However, only very low levels of DNA could be isolated from the cloacal swabs on day 0 and most of the samples could not be PCR amplified and hence no sequence data was obtained. Therefore, only results from day 1 to day 23 are presented. 16S amplicons from 160 cloacal samples were sequenced and presented in the analysis. Although nine phyla were detected across the three treatment groups, there was a clear domination of Firmicutes in all three groups (Figs. S1, S2A). At a genus level Lactobacillus was the dominant genus (mean abundance of 40.39%), followed by Enterococcus (23.32%), Clostridium (3.7%), Faecalibacterium (2.27%), Ruminococcus (2.07%), Staphylococcus (1.77%), Coprobacillus (1.62%), Coprococcus (1.52%) and Sphingomonas (1.29%) with all other genera being of much lower abundance (<1%) (Fig. S1). There was a noticeable bird-to-bird variation in a few of the birds in each group (Fig. S1), for example, minimum abundance for the genus Lactobacillus was 0% (in a bird dominated by Enterococcus at 98.86%) while maximum of Lactobacillus abundance was 99.54%.

Bird-to-bird microbiota variation

The high level of individual bird-to-bird variation in all three treatments implied that the issue of high intra flock variability was not remedied with cecal content treatments CCT1 and CCT2. To further investigate whether uniformity was increased using cecal content treatment ANOSIM was used to compare inter and intra group distances using Bray Curtis (Figs. S2B, S2D) and weighted and unweighted UniFrac at an OTU level (Figs. S2E, S2F). In all cases, the birds were most similar (or least dissimilar) in the CCT2 group; however there were no differences between the control and CCT1 treatment. All sample to sample distances within each group (bird-to-bird distances in the birds from the same group) were compared by weighted and unweighted UniFrac using ANOVA and Tukey Honestly Significant Differences test (using R). Sample to sample differences by weighted UniFrac were (ANOVA P = 1.165E⁻⁴) different, significantly between CCT1-CCT2 and CTRL-CCT2 (Fig. S2E). Similarly, using unweighted UniFrac the three groups had very different intra group distances (ANOVA, P = 2.2E⁻¹⁶), also significantly lower in CCT2 then in the other two groups and insignificant between CTRL and CCT1 (Fig. S2F). The above data indicates that CCT2 did achieve better uniformity and microbiota reproducibility; however this was not reproduced in CCT1 treatment and thus could depend on inoculum composition. In the development timeline, ANOSIM analysis showed that samples taken on day 1 were most dissimilar and samples taken on day 23 most similar (Fig. S2D) indicating less microbiota fluctuations in more mature birds.

Microbial community development over time

The microbiota community analysis was extended to investigate the influence of treatments on microbiota development over time. Two-way PERMANOVA (Primer-E) was used on weighted and unweighted UniFrac distance matrices. On both weighted and unweighted UniFrac the influence of treatment was significant (P = 8.00E⁻⁴ and P < 1E⁻⁵, respectively) and differences between the days of community development were also significant by both weighted and unweighted UniFrac (both with P < 1E⁻⁵). The interaction between day vs. treatment was not significant by weighted (P = 0.225) and was significant by unweighted UniFrac (P = 0.011).

Although there were significant differences in microbial communities between the treatments, the community development timeline was comparable between the 3 groups. Overall species Richness and Evenness were altered significantly (ANOVA) during development of the microbial community (P = 1.58E⁻⁴ and P = 2.02E⁻⁴, respectively, Fig. 2) and the trend of increasing Richness and Evenness with time was reproducible across the three trials (Fig. S3). There was no significant difference in either Richness or Evenness between the three flocks.

The microbiota community development timelines were then compared in the three groups using LDA Effect Size (LEfSe) timeline analysis and classic Pearson regression to find taxa strongly associated with time. LEfSe analysis identified OTUs that characterize and drive the differences between two or more biological conditions (Segata et al., 2011). Major microbiota perturbations at the genus level (Fig. 3A) occurred at days 1 and 2 and again from days 10–23. Enterococcus and Azospirillum influenced the development differences at days 1 and 2 and Blautia and Roseburia from day 10 to 13 while an increase in abundance of Faecalibacterium was evident in more mature microbiota towards the end of the four week trial at day 23 (Fig. 3B).

Pearson based regression analysis was used to identify genera (Table 3) and OTUs (Table 4) significantly (P < 0.001) correlated with time (days). At the genus level Faecalibacterium steadily increased in abundance and was the most significantly positively correlated with time (P = 1.80E⁻¹³, r = 0.54; Table 3, Fig. 4A, Fig. S4), and was identified by LEfSe analysis as the genus that most distinguishes the later stages of microbiota development (Day 23). Enterobacter was significantly (P = 9.8E⁻³) negatively (r =  − 0.20) correlated with time (Table 3, Fig. 4B). Enterococcus, identified by LEftSe as driving differences in timeline during the first 2 days was not significantly correlated with time. Instead, it showed a reproducible pattern of colonisation by being very high during the first two days, diminished at day 3 and steadily increasing afterwards (Fig. S5). Lactobacillus did not change significantly during the time-course (Fig. S6). Interestingly, there were a number of genera that showed very little oscillation in abundance during the time-course and those include the least affected Weissella (P = 0.99). A network diagram visualising the genera interactions with one another during time of development is shown in Fig. 5. OTUs highly correlated with the timeline reflect the above mentioned time-responsive genera (Table 4).