This study was approved by the Shanxi Agricultural University Animal Experiment Ethics Committee (the license number: SXAU-EAW-2017-002Chi.001). In total, 108 1-day-old IBH were chosen. Chickens were randomly divided into nine replicates, with 12 chickens per replicate. Chickens were fed three different diets (Table 1) during brooding periods (0–8 weeks), growing periods (9–20 weeks), and laying periods (21–50 weeks).

Chickens were given free access to water and diet. The management of the temperature, light, and humidity was conducted according to the breeding manual. No conventional immunization schedule was performed to avoid impacts on the gut microbiota.

A chicken from each replicate was chosen at the end of 8, 20, and 50 weeks. They were slaughtered humanely using oral bloodletting slaughtering method. The contents of the left cecum per bird were collected into multiple cryogenic tubes, and they were put into a liquid nitrogen tank and then preserved at −80°C until the 16S rRNA gene sequence of the gut microbiota and the determination of the concentration of SCFAs.

In terms of gut microbiota, the comparisons of the relative abundance phyla and genera in groups were performed by Metastats (41) (version 20090414). Metastats showed significantly different relative abundances of bacteria using P < 0.01 or 0.05. Multivariate statistical techniques including principal component analysis (PCA), principal coordinates analysis (PCoA), and non-metric multidimensional scaling (NMDS) of weighted uniFrac distances were calculated and plotted in R project. Statistical analysis of Welch's t-test and Anosim test was calculated using R project. The β-diversity analyses between groups were calculated by the Kruskal–Wallis using Vegan package in R project. Heatmap analysis was performed using the R package. Analysis of function difference between groups was calculated by Welch's t-test in R package (version 2.5.3). Statistical analyses of SCFAs were performed using a one-way ANOVA with SPSS 22.0 software. The results are expressed as the means and standard error of the mean (SEM).

For simplicity, “ISA Brown Hens-8 weeks” was named IBHE for short, “ISA Brown Hens-20 weeks” was named IBHT, and “ISA Brown Hens-50 weeks” was named IBHF.

The total and unique numbers of OTUs at 50 weeks (1,629 and 823) were more than those at 8 (963 and 193) and 20 weeks (958 and 144) (Figure 1). Moreover, the α-diversity indices including ACE and Chao1 significantly increased with age (P < 0.01) (Table 3). It is suggested that the microbial community richness significantly increased with age. The gut microbial diversity of IBHF was the highest among three periods (Table 3); this was also consistent with the number of OTUs.

For the β-diversity indices, the PCA (Figure 2A) and PCoA (Figure 2B) showed that the samples were separated in the first principal component. NMDS showed that the stress = 0.091 < 0.1, indicating that the accuracy of the model was good (Figure 2C). In addition, Anosim test showed that R = 1 > 0, indicating that the difference of microorganisms between groups was greater than that within groups; however, P = 0.1 > 0.05 represented that there was no significant difference in the pairwise comparison of three periods (Figure 2D). On the whole, the samples were clustered by different growth periods, but there was no significant difference between different periods (P > 0.05).

At the phylum level (Figure 3A), the dominant phyla of all samples were both Firmicutes and Bacteroidetes. The abundance of “generalist” Bacteroidetes (44.40, 62.55, and 73.65%) (Figure 3B) significantly increased at 8, 20, and 50 weeks, respectively (P < 0.01), but Firmicutes (40.47%, 23.41, and 16.76%) (Figure 3C) and Proteobacteria (7.27, 3.37, and 2.70%) (Figure 3D) significantly decreased with age (P < 0.01). Special cellulose-degradation phylum Fibrobacteres was only found at 50 weeks (P < 0.01) (Figure 3E).

At the genus level (Figure 4A), the dominant genus of all samples was Bacteroides (27.97, 24.84, and 26.29%) (P > 0.05). The dominant bacteria at 8 weeks included the bile-tolerant bacterium Alistipes (7.38%) (Figure 4B) and low-abundance butyrate-producing genus Anaerostipes (0.4%) (Figure 5A) compared with 20 weeks (1.23%) (0.033) and 50 weeks (0.76%) (0.00021) (P < 0.01), respectively. In addition, Bacteroides thetaiotaomicron (0.006%), which is one of the best fiber-degrading species, was only detected at 8 weeks (Figure 5B). The dominant bacteria genera at 20 and 50 weeks included Rikenellaceae_RC9_gut_group (14.56 and 15.23%) (P < 0.01) (Figure 4C), propionate-producing genus Phascolarctobacterium (1.92 and 2.76%) (P < 0.01) (Figure 4E), fiber-degradation bacteria Prevotellae_UCG_001 (1.81 and 2.13%) (P < 0.01) (Figure 4D), and Alloprevotella (0.56 and 0.92%) (P < 0.05) (Figure 4F) compared with those at 8 weeks (1.09, 0.80, 0.26, and 0.21%), respectively.

The dominant bacteria at 20 weeks also included low-abundance fiber-degradation and acetate-producing genus Bifidobacterium (0.028%) (Figure 5C) and potential fiber-degradation bacteria Prevotellae_Ga6A1_group (0.62%) (Figure 5D) compared with those at 8 and 50 weeks. Bifidobacterium was also a famous probiotic genus. The dominant bacteria at 20 weeks also included some lactate-producing species such as Lactobacillus aviaries (0.014%), Lactobacillus agilis (0.029%), and Lactobacillus alvi (0.043%) (Figure 5E). Moreover, many unique fiber-degradation bacteria including Fibrobacter (0.00072%), Prevotellae_UCG_003 (0.13%), Prevotella_1 (0.11%), and Prevotella_9 (0.0026%) were only found at 50 weeks (Figures 5F,G).

The KO of OTUs was further predicted (Figure 6A). The heatmap showed that ATP-binding cassette (ABC) transports including K06147 (0.0066, 0.0063, and 0.0056%), K02003 (0.0042, 0.0039, and 0.0036%), and K01990 (0.0031, 0.0030, and 0.0027%) significantly decreased at 8, 20, and 50 weeks (P < 0.05) (Figure 6B), respectively. In contrast, β-galactosidase K01190 (0.0039, 0.0050, and 0.0056%), β-glucosidase K05349 (0.0034, 0.0042, and 0.0051%), and two-component system (TCS) including K00936 (0.0060, 0.0071, and 0.0075%) and K07636 (0.0029, 0.0035, and 0.0039%) significantly increased with age (P < 0.01) (Figure 6C).

Then, the enriched metabolic pathways of OTUs were predicted (Figure 7A). Amino acid metabolism pathways ko00330 (0.021, 0.019, and 0.018%), the energy metabolism pathway ko00190 (0.017, 0.016, and 0.015%), and the membrane transport of ABC transporters ko02010 (0.075, 0.066, and 0.029%) significantly decreased at 8, 20, and 50 weeks (P < 0.01) (Figure 7B), respectively. In contract, carbohydrate metabolism pathways ko00500 (0.026, 0.027, and 0.029%) and ko00051 (0.017, 0.021, and 0.021%) (P < 0.01) and the TCS pathway ko02020 (0.068, 0.070, and 0.072%) (P < 0.05) significantly increased with age. In addition, the carbohydrate metabolism pathway ko00051 (0.021 and 0.021%) was enriched at 20 and 50 weeks compared with 8 weeks (0.017%) (P < 0.01) (Figure 7C).

Next, the concentration of gut microbial metabolites SCFAs was measured (Table 4). The concentration of SCFAs at 20 weeks was higher than that at 8 weeks (P < 0.01). Interestingly, although the microbial diversity at 50 weeks was the highest, the concentration of SCFAs at 20 weeks was also almost twice as high as that at 50 weeks (P < 0.01).

The gastrointestinal tract of newly hatched chickens immediately has microbial colonization (42). With the growth of age, the cecal microorganisms form a complex community (43). In this experiment, the numbers of OTUs and gut microbial α-diversity including ACE and Chao1 of chickens significantly also increased with age. The larger the value of ACE and Chao1, the higher the community richness. This is consistent with previous reports that increasing taxonomic richness and diversity were observed in chickens through time (9). The α-diversity of the gut microbiota of pigs also increased with age (21). In this experiment, the β-diversity of the samples was clustered by different growth periods, but NMDS analysis showed the difference between different periods was not great. It was consistent with a report that NMDS showed that 0- to 42-day-old chicken gut microbiota could be clustered according to different ages, but microbial clusters became quite similar after 28 days (11).

At the level of phyla, the predominant phyla were Firmicutes and Bacteroidetes throughout the period in this experiment, which was consistent with previous studies (6). In addition, the relative abundance of Firmicutes was higher than that of Bacteroidetes at 8 weeks in this study. It was also reported that there were around 37% Firmicutes and 10% Bacteroidetes in 8-week-old chickens (1). Firmicutes species are regarded as “specialists” for storage plant polysaccharides (starch and fructose) and oligosaccharides. In contrast, Bacteroidetes is a kind of “generalist” that degrades dietary fiber. It can utilize more a wide range of plant polysaccharides than does Firmicutes (19). In this experiment, Bacteroidetes increased from 8 to 50 weeks. This was supported by studies that reported the gradual increase of Bacteroidetes at the expense of Firmicutes during chicken rearing and egg production (18). Members of Bacteroidetes were present mainly in adult hens (44).

At the genus level, the composition of the gut microbiota in different growth periods was also different in this experiment. The dominant bacteria genus at 8 weeks did not include fiber-degradation bacteria but included the bile-tolerant bacterium Alistipes compared with that at 20 and 50 weeks. This was maybe attributed to a high-fat and low-fiber diet (45). Alistipes specially increased in the persons who consumed animal-based diet instead of plant-based diet (46). In contrast, fiber-degradation bacteria genus Prevotellae_UCG_001 and Alloprevotella obviously increased at 20 and 50 weeks compared with 8 weeks. For 20 weeks with a high-fiber diet, this because the fiber content of it was higher than that at 8 weeks, resulting in increased fiber-degradation bacteria and Prevotellae_Ga6A1_group and Bifidobacterium. Research on human showed that the abundance of Prevotella is enriched in a fiber diet (47) and has a strong ability to utilize fiber (48). Prevotella gradually became the most diverse and predominant genus with the increase of dietary fiber and age in pigs (21). The research showed that increasing dietary fiber could increase the abundance of Bifidobacterium (49). For 50 weeks with a low-fiber diet, the feed intake was greater than that at 8 weeks, resulting in more dietary fiber being ingested. This may also help to explain why fiber-degradation bacteria Prevotella_1 and Prevotella_9 and excellent special cellulose-degradation genus Fibrobacter were only detected at 50 weeks in this experiment. Notably, as far as we know, it seems that Fibrobacter has not been reported in chickens. Fibrobacter was once believed to only exist in mammalian intestines (50), and it was first reported in the cecum of birds (ostrich) in 2010 (51).

Software packages such as Tax4Fun can be used to predict the functional profiles of OTUs using 16S rRNA gene sequences based on the SILVA rRNA database (52). In this experiment, functional annotation of the gut microbiota showed that ABC transporters decreased at 8, 20, and 50 weeks. This was related to the dominant phylum, which was Firmicutes, decreasing during this period. Firmicutes has gram-positive polysaccharide utilization loci (PULs) (gpPULs), which encode ABC transporters and other transporters to introduce small sugar into the periplasm for processing (15, 53). ABC transporter is a type of transport ATPase on the bacterial plasma membrane, and it transfers glucose to the other side of the membrane through the change of conformation. Consistently, the energy metabolism pathway and amino acid metabolism pathway were rich at 8 weeks in this study. Research also reported that the microbiota in the hindgut of 42-day-old chickens were enriched in amino acids metabolism and energy metabolism according to KEGG functional analysis (11).

In contrast, functional annotation of OTUs included a TCS; and β-glycosidase, carbohydrate metabolism pathway, and TCS pathway increased at 8, 20, and 50 weeks. This was related to the fact that the abundance of Bacteroidetes increased with age. Contrary to the gpPULs of Firmicutes, Bacteroidetes can utilize a series of plant-derived dietary polysaccharides via unique PUL (54, 55). PUL has been identified in all members of Bacteroidetes such as B. thetaiotaomicron and Bacteroides ovatus (19). PUL encodes a hybrid TCS, extracellular glycoside hydrolase such as β-glycosidase (56), and other enzymes to degrade dietary fiber by cleaving glycosidic bonds (19). Consistently, carbohydrate metabolism pathway was increased with age in this experiment.

SCFAs are the major fermented metabolites of dietary fiber for SCFA-producing bacteria; primarily acetate, propionate, and butyrate account for 90–95% (57). SCFAs are very important for the growth and health of host.

In this study, the concentration of SCFAs at 20 weeks was higher than that at 8 weeks. This may be attributed to the level of fiber increase, and the numbers of some dominant fiber-degradation bacteria and SCFA-producing bacteria were more in 20 weeks. The dominant fiber-degradation bacteria at 20 weeks such as Prevotellae_UCG_001 and Alloprevotella broke down dietary fiber into more monosaccharides, and then dominant SCFA-producing bacteria Phascolarctobacterium, Bifidobacterium, and Lactobacillus fermented monosaccharides into more SCFAs than those at 8 weeks. Phascolarctobacterium can ferment monosaccharide into propionate (58), and Bifidobacterium produces acetate using “bifid-shunt” (59). Members of Lactobacillus use dietary fiber via ABC transporters to produce acetate (60), which can be further fermented into butyrate by some butyrate-producing bacteria (61).

In addition, the concentration of SCFAs at 20 weeks was also greater than that at 50 weeks in this experiment. Given that the relative abundance of the dominant SCFA-producing bacteria showed no great difference between them, we speculated that it was more likely that although the microbial α-diversity including ACE and Chao1 at 50 weeks was higher than that at 20 weeks, most of the bacteria were not fiber-degradation bacteria or SCFA-producing bacteria; this means that more bacteria compete for limited glucose as the carbon for growth and that less glucose was fermented into SCFAs by few SCFA-producing bacteria. This may also help to explain why obese people have less microbial diversity but have more SCFAs than lean people (62).