ABSTRACT
In the present study, we investigated whether reducing the particle size of wheat bran affects the colonizing microbial community using batch fermentations with cecal inocula from seven different chickens. We also investigated the effect of in-feed administration of regular wheat bran (WB; 1,690 μm) and wheat bran with reduced particle size (WB280; 280 μm) on the cecal microbial community composition of broilers. During batch fermentation, WB280 was colonized by a lactic acid-producing community (Bifidobacteriaceae and Lactobacillaceae) and by Lachnospiraceae that contain lactic acid-consuming butyric acid-producing species. The relative abundances of the Enterobacteriaceae decreased in the particle-associated communities for both WB and WB280 compared to that of the control. In addition, the community attached to wheat bran was enriched in xylan-degrading bacteria. When administered as a feed additive to broilers, WB280 significantly increased the richness of the cecal microbiota and the abundance of bacteria containing the butyryl-coenzyme A (CoA):acetate CoA-transferase gene, a key gene involved in bacterial butyrate production, while decreasing the abundances of Enterobacteriaceae family members in the ceca. Particle size reduction of wheat bran thus resulted in the colonization of the bran particles by a very specific lactic acid- and butyric acid-producing community and can be used to steer toward beneficial microbial shifts. This can potentially increase the resilience against pathogens and increase animal performance when the reduced-particle-size wheat bran is administered as a feed additive to broilers.
IMPORTANCE Prebiotic dietary fibers are known to improve the gastrointestinal health of both humans and animals in many different ways. They can increase the bulking capacity, improve transit times, and, depending on the fiber, even stimulate the growth and activity of resident beneficial bacteria. Wheat bran is a readily available by-product of flour processing and is a highly concentrated source of (in)soluble dietary fiber. The intake of fiber-rich diets has been associated with increased Firmicutes and decreased Proteobacteria numbers. Here, we show that applying only 1% of a relatively simple substrate which was technically modified using relatively simple techniques reduces the concentration of Enterobacteriaceae. This could imply that in future intervention studies, one should take the particle size of dietary fibers into account.
INTRODUCTION
The greatest determinant of the gut microbiota composition of production animals, in addition to age and breed, is diet. The dietary levels and quality of fat, protein, and carbohydrates and the use of exogenous feed enzymes all impact the gut microbiota (1–3). Dietary fibers (DF) are carbohydrate polymers, including lignin and other plant-associated substances, that escape digestion in the small intestine and pass into the hindgut where they are (partially) fermented by the microbiota (4, 5). They can be classified as being either soluble or insoluble (6). Most water-soluble DF (such as pectins) are fermentable, while insoluble DF (such as cellulose, lignin, and hemicellulose) are thought to be less fermentable (7–9). Certain DF are known to improve the gastrointestinal health of both humans and animals by stimulating the growth of resident beneficial bacteria, which in turn can, for example, increase the resilience of the host against pathogens (10–13).
Wheat bran, a by-product of the dry milling of wheat grains into flour, is a highly concentrated source of DF. It consists of the outer layers (cuticle, pericarp, and seed coat) of the wheat kernel retaining only small amounts of aleurone and starchy endosperm. Wheat bran as such consists of both fermentable (starch, protein, and arabinoxylans with low substitution degrees) and nonfermentable (cellulose, lignin, and arabinoxylans with high substitution degrees) components (14–17). The polysaccharide fraction consists mainly of insoluble DF, while only a small proportion is soluble (15, 16). Within the intestinal environment, the insoluble fraction shows a great resemblance to (in)organic macroaggregates in aquatic ecosystems (18). It has been shown that these aggregates are heavily colonized by bacteria and other heterotrophic microorganisms (18). The results from several studies have suggested that bacterial community structures are dissimilar between particle-associated and free-living bacteria, the former exhibiting higher enzymatic activity than the latter (18). It is evident that these aggregates and their surroundings are hot spots of microbial processes, with the solutes leaking out of the aggregates (18). In accordance, several studies have shown that particulate matter in the intestine harbors a unique and distinct bacterial community which differs from the free-living community (14, 19, 20). The particle-associated community composition appears to depend strongly on the substrate (14, 20). While no such data are available for poultry, in ruminant animals, it has been shown that specific ruminal bacteria develop a dynamic biofilm upon digesta particles and possess higher fibrolytic activity than the luminal population (21–23). With regard to wheat bran, it has been shown in a very recent in vitro study that it can be colonized by a specific population of bacteria from the human gut microbiota (19). Attached to wheat bran, the close proximity between the different bacteria might foster an enhanced cross-feeding and more efficient use of substrate and thus more elaborate energy harvest. In a previous study, we demonstrated that particle size reduction affects fermentability and that smaller wheat bran fractions are more efficiently fermented into short-chain fatty acids (24). It is unknown whether the size of the particles matters with regard to colonization patterns. In this study, we examined whether wheat bran with a reduced particle size is colonized by a specific bacterial community and whether these bran particles can induce shifts in the composition of the cecal microbiota when supplemented to the feed of broilers.
RESULTS
Wheat bran particles of different particle sizes are colonized by distinctive microbial communities.Wheat bran (WB) and wheat bran with a reduced particle size (WB280) were in vitro fermented using the cecal inocula of seven broilers. After 24 h, all non- and loosely attached bacteria were removed by washing, and DNA of the attached bacteria was isolated and used for 16S rRNA sequencing. In total, 374,832 merged 16S rRNA sequences were used for analysis, with a mean number of 18,742 sequences per sample. Using a principal-coordinate analysis, it was shown that the bacterial diversity of the attached communities differed greatly from that of the control community. Moreover, the attached microbial community was influenced by the particle size of the wheat bran particles (Fig. 1A). The bacterial richness (number of observed operational taxonomic units [OTUs]) on both bran fractions was smaller than that of the control community (Fig. 1B). Significant differences were observed in the relative abundances of specific 16S rRNA sequences at different taxonomic levels (see Table S1 in the supplemental material). Interindividual differences were relatively limited, as illustrated at the family level (see Fig. S1). Sample WB280 4 is characterized by a larger relative abundance of Actinobacteria and lower abundance of Lachnospiraceae members than the other WB280 samples (Fig. S1C). For control samples 5, 6, and 7, an increased relative abundance of Defluviitaleaceae can be noticed in comparison to the other control samples (Fig. S1A). Also, WB sample 3 shows a deviating composition compared to the other WB samples (Fig. S1B).
Diversity and richness measures for the control bacterial community or the communities attached to WB and WB280. (A) PCoA biplot of the abundance-based Bray-Curtis dissimilarity index revealed distinct microbial communities attached to WB (▲; n = 6), WB280 (●; n = 7), and the control (■; n = 7) (P = 0.001). (B) Richness of the different bacterial communities, expressed as numbers of observed OTUs. *, P < 0.05.
Members of the phylum Actinobacteria were significantly enriched on WB (34%) and WB280 (20%) compared to that on the control (4%, P < 0.001 and P = 0.039, respectively). Members belonging to Bacteroidetes preferred to attach to WB (8%) compared to WB280 (3%, P = 0.006). No significant differences were observed compared to the control (3%; P > 0.05). For WB280, 70% of the attached community belonged to Firmicutes, as opposed to only 46% for WB (P = 0.004) and 61% for the control (P > 0.05). Proteobacteria appeared to be reluctant to attach to bran material, since their relative abundance was lowered with 20% on WB and 29% (P < 0.001) on WB280 compared to that of the control community (Fig. 2).
Relative abundances of the most abundant bacterial phyla present in the control community or attached to WB or WB280. Bran fractions were fermented in a nutrient-poor medium with a chicken cecal inoculum for 24 h at 37°C. Bran was omitted in the control fermentations. Non- and loosely attached bacteria were removed by washing prior to DNA extraction.
The following significant differences in relative abundances at the family and genus levels were observed: members of the Enterobacteriaceae family were relatively less abundant on the bran. Compared to the control (27%), only 8% (P > 0.05) of the bacteria on WB were Enterobacteriaceae, and the difference was statistically significant for WB280 (3.5%, P < 0.001). This difference was completely explained by the high relative abundance of the genus Escherichia/Shigella in the control community (23%) compared to those on WB280 (3%; P < 0.001) and WB (8%; P > 0.05). Members of the Bacteroidaceae were enriched on WB (4%) when comparing the relative abundances in the control community (1%; P = 0.008) and the community attached to WB280 (2%; P = 0.038). This fully coincided with the different relative numbers of members of the genus Bacteroides when comparing control (2%; P = 0.07) and WB280 (2%; 0.043) with WB (4%). Also, WB (4%) was colonized to a higher extent than WB280 (2%) by members of the Porphyromonadaceae family (P = 0.048) and those belonging genus Parabacteroides (WB, 4%; WB280, 2%; P = 0.030). Relatively, Bifidobacteriaceae members showed a significantly higher abundance on WB (34%; P < 0.001) and WB280 (20%; P = 0.039) than the control (4%). This was partly explained by an enrichment of the genus Bifidobacterium on WB280 (20%) and WB (34%; P = 0.001) compared to that of the control (6%). Two important families of butyrate-producing bacteria showed significant differences in their relative abundances. First, members of the Lachnospiraceae were enriched on WB280 (44%) compared to that on WB (15%; P = 0.001) but not compared to the control community (32%; P > 0.05). This was mainly explained by a higher abundance of a group of uncultured Lachnospiraceae and Lachnospiraceae that could not be assigned to a specific genus within the family (WB, 12%; WB280, 41%; P < 0.001). Members from the genus Roseburia showed a 10-fold higher abundance on WB280 (0.11%) than the control community (0.01%; P = 0.048). No significant differences were observed for WB (0.03%; P > 0.05). Second, members of the Ruminococcaceae showed significantly lower abundances on WB280 (14%) than on WB (22%; P = 0.034) and the control (23%; P = 0.005). The most important genera contributing to this difference were the genus Anaerotruncus, a group of uncultured Ruminococcaceae, and a group of OTUs that could not be assigned to a specific genus within the Ruminococcaceae family. Members of the genus Subdoligranulum showed a deviating trend and were enriched on both WB (1.70%; P = 0.007) and WB280 (1.20%; P = 0.041) compared to the control community (0.49%). Within the Enterococcaceae family, Enterococcus species showed a tendency to be enriched on WB280 (3%; P = 0.05) compared to the control (0.4%). A significantly higher relative abundance Lactobacillaceae was observed on bran. Compared to a relative abundance of 0.37% of Lactobacillaceae in the control community, abundances of 2% and 3% were observed in the communities associated with WB (P = 0.009) and WB280 (P = 0.003), respectively. This was solely caused by an increase in the relative numbers of the genus Lactobacillus (Fig. 3).
Relative abundances of the most important bacterial families significantly differing between the control and/or bran-associated communities. Bran fractions were fermented in a nutrient-poor medium with a chicken cecal inoculum for 24 h at 37°C. Bran was omitted in the control fermentations. Non- and loosely attached bacteria were removed by washing prior to DNA extraction. Mean relative abundances are shown on the y axes while the x axes indicate the bran fractions or controls for Bifidobacteriaceae (A), Bacteroidaceae (B), Porphyromonadaceae (C), Lactobacillaceae (D), Streptococcaceae (E), Defluviitaleaceae (F), Lachnospiraceae (G), Ruminococcaceae (H), and Enterobacteriaceae (I), as determined by 16S rRNA V3 to V4 amplicon sequencing. “+” represents the mean values, and the horizontal lines at the bottoms, middles, and tops of the boxes represent the first quartile, median, and third quartile, respectively. The whiskers indicate the min/max values. *, P < 0.05; **, P ≤ 0.01; ***, P < 0.001.
Wheat bran-attached bacterial community potentially exerts different metabolic functions than the control community.Exploring the potential functions of the attached and control communities reveals potentially distinct metabolic profiles. The communities attached to WB and WB280 were enriched in OTUs assigned to xylan degraders (Fig. S2A). Since one of the major constituents of wheat bran is arabinoxylan, the NCBI protein database was searched for putative glycoside hydrolases and carbohydrate esterases which enable (partial) xylan degradation in taxa that were significantly enriched on bran material. Endoxylanases are responsible for the breakdown of the xylan backbone into smaller fragments. Bacteroides thetaiotaomicron encodes many enzymes, distributed over the different families: putative endoxylanases, xylosidases, arabinofuranosidases, and glucuronidases can be traced back in the chosen reference genome. Also, Bifidobacterium pseudolongum features mostly arabinofuranosidases, which are accessory enzymes, assisting the xylosidases and endoxylanases (Fig. S2B). In addition, it harbors a β-xylosidase. OTU5, showing 97% 16S rRNA sequence similarity with Ruminococcus torques, putatively expresses an endoxylanase and glucuronidase (see Table S2).
Inclusion of wheat bran with different particle sizes in the feed of broilers induces particle-size-dependent shifts in the cecal microbiota.One-day-old chickens received a standard diet for 10 consecutive days supplemented with WB or WB280. The control group received nonsupplemented feed. Chickens were euthanized on day 10, and cecal contents were collected, from which DNA was extracted. DNA samples were sent for Illumina sequencing. In total, 662,579 merged 16S rRNA sequences were used for analysis, with a mean number of 26,503 sequences per sample. A principal-coordinate analysis (PCoA) plot based on the Bray-Curtis dissimilarity index shows a specific clustering pattern of the cecal samples (Fig. 4A). The cecal bacterial diversity of chickens showed clustering according to the sample origin. Also, the supplementation with either of two bran fractions had different effects on the bacterial community compositions of the ceca (P < 0.001). The richness, expressed as the number of observed OTUs, was significantly increased, with 15% (P = 0.027) more OTUs for chickens receiving WB280 than for control chickens, and was 33% lower (P < 0.001) for chickens receiving WB than for chickens receiving WB280 (Fig. 4B).
Diversity and richness measures for the cecal bacterial communities of chickens in the control group and the groups receiving 1% WB and 1% WB280. One-day-old chickens received a nonsupplemented standard diet (control ■; n = 9) or a standard diet supplemented with 1% WB (▲; n = 6) or 1% WB280 (●; n = 10) for 10 days. (A) PCoA biplot of the abundance-based Bray-Curtis dissimilarity index revealed distinct cecal microbial communities for chickens receiving a nonsupplemented diet, 1% WB, and 1% WB280 (P < 0.001). (B) Richness of the different bacterial communities, expressed as numbers of observed OTUs. *, P < 0.05; ***, P < 0.001.
The relative abundances of members of the Firmicutes phylum were significantly higher in cecal contents from chickens receiving WB280 (94%) than from the control group (82%, P = 0.008). A significantly lower abundance in Actinobacteria was observed for chickens fed WB (0.45%, P = 0.002) and WB280 (1.95%, P = 0.006) than for control chickens (12.5%) (Fig. 5; see also Table S3).
Relative abundances of the most important bacterial phyla present in cecal contents of broilers receiving bran with different particle sizes. A standard broiler feed was supplemented with either 1% WB or WB280 for 10 days. DNA was extracted from cecal contents and sent for 16S rRNA V3 to V4 amplicon sequencing. Cecal community compositions were compared between chickens from the control group receiving a standard feed and groups of chickens receiving WB or WB280.
At the family and genus levels, several significant differences were observed. The number of Enterobacteriaceae was significantly lower in cecal contents from chickens receiving WB280 (0.67%) than in those from the control group (3.5%, P = 0.021) and the group fed WB (8.4%, P < 0.001) (Fig. 6F). This was fully explained by a lower abundance of the group Escherichia/Shigella. In accordance with the lower abundance of Actinobacteria in chickens fed wheat bran of either size, a difference in Bifidobacteriaceae and members of the genus Bifidobacterium was observed (Fig. 6A). Overall numbers of Lachnospiraceae did not change when wheat bran was supplemented, but specific genera within this family showed a significant difference. An enrichment of the genus Pseudobutyrivibrio was observed in the cecal contents of chickens receiving WB (2%) compared to that of control chickens (0.3%; P = 0.004) and chickens receiving WB280 (0.08%; P = 0.002). Members of the genus Blautia were enriched in the ceca of chickens receiving unsupplemented feed (14%; P = 0.003) or feed with 1% WB280 (11%; P = 0.03) compared to that of the group receiving WB (1%). A number of OTUs that could not be assigned to a specific genus within the Lachnospiraceae or uncultured Lachnospiraceae were higher in abundance in the cecal contents from chickens receiving bran than in the cecal contents from control chickens. This was significant for the group receiving WB (P = 0.015; P = 0.001). The intake of WB280 (20%) resulted in a higher number of Ruminococcaceae than with the intake of regular wheat bran (8%; P = 0.018) and the control feed (12%; P = 0.032, respectively) (Fig. 6E). Within the Ruminococcaceae, the genus Anaerotruncus was represented less in the ceca of chickens receiving WB (0.06%) than in the ceca of control chickens (1%; P = 0.005) and in the group of chickens receiving WB280 (1.7%; P < 0.001). Members of the genus Subdoligranulum were specifically enriched in the cecal contents of chickens receiving WB280 (7%) compared to those from the groups receiving nonsupplemented feed (1%; P = 0.015) and WB (1%; P = 0.012). When summing the relative abundances of Lachnospiraceae and Ruminococcaceae, a specific difference was observed: chickens fed WB280 showed higher numbers in their cecal contents than control chickens (73% versus 54%, P = 0.024) (Fig. 6G). In accordance, the number of gene copies encoding the butyryl-CoA:acetate CoA-transferase, as determined with quantitative PCR (qPCR), was significantly higher in the cecal contents of chickens receiving WB280 than in those of control chickens (P = 0.02), while the total numbers of bacteria, determined by quantifying the total bacterial 16S rRNA gene copies, remained unchanged (Fig. 7).
Relative abundances of the most important bacterial families significantly differing between groups. Chickens were fed a standard commercial feed for 10 days. Group 1 received an unsupplemented feed (control), while groups 2 and 3 were fed the standard feed supplemented with 1% WB and 1% WB280, respectively. On day 10, chickens were euthanized and DNA was extracted from the cecal contents. Mean relative abundances are shown on the y axes while the x axes indicate the treatments for Bifidobacteriaceae (A), Enterococcaceae (B), Defluviitaleaceae (C), Peptostreptococcaceae (D), Ruminococcaceae (E), Enterobacteriaceae (F), and the sum of Lachnospiraceae and Ruminococcaceae (G), as determined by Illumina sequencing. “+” represents the mean values, and the horizontal lines at the bottoms, middles, and tops of the boxes represent the first quartile, median, and third quartile, respectively. The whiskers indicate the min/max values. *, P < 0.05; **, P ≤ 0.01; ***, P < 0.001.
Numbers of total bacteria (A) and the relative numbers of butyryl-CoA:acetate CoA-transferase (CoAt) gene copies (B). The number of CoAt gene copies was expressed relatively to the number of total bacterial 16S rRNA gene copies of 10-day-old broilers fed a standard feed (control) or a standard feed supplemented with either 1% WB or 1% WB280. “+” represents the mean values, and the horizontal lines at the bottoms, middles, and tops of the boxes represent the first quartile, median, and third quartile, respectively. The whiskers indicate the min/max values. *, P < 0.05.
DISCUSSION
Previous human microbiota studies indicated that wheat bran is colonized by a specific bacterial community (14, 19, 20). In the present study, we investigated whether reducing the particle size of wheat bran altered and/or optimized the colonizing community by using a cecal inoculum from broilers. We also investigated the effect of the administration of regular wheat bran (WB) and wheat bran with reduced particle size (WB280) on the cecal community compositions of broilers in vivo to evaluate their potential to induce a microbial community shift deemed to be beneficial.
It was previously demonstrated that the community composition of insoluble substrates strongly depends on the nature of the substrate (14). In a fermentation experiment, mucus, wheat bran, and starch were colonized by very distinct bacterial communities (14, 20). In the present study, we showed that also the size of the particles has an influence on the community composition. By reducing the particle size, one does not alter the chemical composition of the bran material, but the availability of some of its constituents (e.g., damaged starch content and water-extractable arabinoxylans) is increased (25). Distinct colonization patterns for WB and the technically modified WB280 fraction were observed: both fractions were enriched with lactate-producing genera such as Lactobacillus and Bifidobacterium, but compared to that with WB, there was a significant enrichment of Lachnospiraceae on WB280. In addition to a large number of OTUs which could not be appointed to an established genus within the Lachnospiraceae, the genus Roseburia was responsible for this shift. Several members of this Lachnospiraceae family have the ability to consume lactate and produce butyrate (3, 26). Cross-feeding, either of lactate or partial breakdown products, and thus butyrate production, could potentially be enhanced between these metabolic groups through their cooccurrence on WB280. It may be questioned whether the transit rate in the chicken intestine is sufficiently low to enable the formation of particle-associated communities. The longer retention time in the cecal pouches poses a probable answer to this question (27, 28).
Concordant with the increased relative abundances of Lachnospiraceae members, one could expect increased levels of butyric acid, as some members of this family are well-known butyrate producers. Indeed, in a previous study, we showed that a 24-h fermentation of WB280 in a carbohydrate-poor medium with cecal inocula from broilers yielded significantly higher levels of acetic (36.58 ± 2.29 mM; P < 0.05), propionic (10.72 ± 0.88 mM; P < 0.05), and butyric (6.40 ± 1.38 mM; P < 0.01) acids than control fermentations which were depleted of wheat bran (27.5 ± 2.80 mM, 7.95 ± 0.44 mM, and 1.60 ± 0.25 mM, respectively) (24). These results may be considered an affirmation of the presently observed increased relative abundances of Lachnospiraceae members in vitro. The effects of WB supplementation also led to higher concentrations of acetic and butyric acids but to a lesser extent than WB280 (24).
The specific reluctance of members of the Enterobacteriaceae family to attach to small-particle-size bran may be explained by the presence of this butyrogenic consortium. It is known that short-chain fatty acids such as butyrate and propionate have inhibitory effects on Salmonella, a zoonotic agent belonging to the Enterobacteriaceae (24, 29–32). These effects appear to radiate from the bran material to the entire cecal content, since less Enterobacteriaceae were present in the cecal contents of WB280-fed broilers compared to that of control-fed chickens. This might be a consequence of the cooccurring increases in the relative numbers of butyrate-producing bacteria (sum of Lachnospiraceae and Ruminococcaceae). We previously showed that the addition of 1% WB280 to the feed of broilers reduced the cecal colonization of Salmonella, whereas this effect was not achieved when administering WB (24). It was hypothesized that through particle size reduction, the bran constituents become more available for fermentation by intestinal bacteria due to the breaking of cell wall barriers and/or an increased accessible surface (24, 33). In addition, the presence of a specifically attached butyrogenic community might be responsible for the more efficient fermentation of WB280 and increased production of short-chain fatty acids (SCFA). However, since we did not measure SCFA concentrations in the cecal contents of WB280- and control-fed chickens, caution is advised when drawing conclusions. At this point, we base our conclusions solely on indirect evidence of the increased relative abundances of butyryl-CoA:acetate CoA-transferase gene copies in the cecal contents from WB280-fed chickens. Nonetheless, considering both in vivo and in vitro results, we believe that all of these observations indirectly but strongly hint at an actual effect in the cecum induced by WB280 supplementation. Conclusive evidence would rely on actual measurements of changes in SCFA concentrations in the ceca. Another shortcoming of the in vivo study is the inclusion of only one pen per treatment; thus, we cannot exclude cage effects.
When comparing in vitro and in vivo data, aside from the similarities, several differences are noted. For example, Actinobacteria were enriched in vitro on WB or WB280 compared to the control community, while in vivo, a decrease of Actinobacteria was evident in chickens that were fed WB- or WB280-supplemented feed compared to that in control chickens (Fig. 1 and 4). The main message and consistent observation in this paper, however, is that WB280 promotes Lachnospiraceae and repels bacteria belonging to the Enterobacteriaceae family. The observed differences between in vivo and in vitro studies may have been caused by the distinct ages of the chickens used in the experiments (4 weeks versus 10 days). The colonization of the gastrointestinal tract is thought to start immediately after hatching, but a typical and stable adult gut microbiota is established only after several weeks (34–36). The rapid development of the microbiome in the early days and weeks of life may have caused a bias, preventing a correct comparison between the presented in vitro and in vivo experiments. As an example, representatives of the phylum Bacteroidetes are important degraders of poly- and oligosaccharides (19, 37–40), and their presence is underrepresented in the guts of young chickens (41).
Species colonizing the surfaces of dietary particles are likely to harbor a specific enzymatic armory that enables both attachment and degradation of complex insoluble polysaccharides (42). In comparison to the control community, the wheat bran-attached community may be metabolically adapted to degrade xylan. The community attached to wheat bran was enriched in OTUs which were taxonomically assigned to xylan-degrading species such as Bacteroides thetaiotaomicron, Bifidobacterium pseudolongum, and Ruminococcus torques, while the control community was depleted in OTUs linked to taxa known for exerting this metabolic function. Samples WB280 4 and WB 2 appear to be the major drivers of this observation. Yet, most wheat bran samples were characterized by the presence of this xylan-degrading community as opposed to the control community (Fig. S2 in the supplemental material). Bacteroides thetaiotaomicron is known to possess extensive carbohydrate degradation activities, harboring glycoside hydrolases (GHs) belonging to as many as 56 different families (37). Many members of the Bacteroidetes phylum are equipped with an extended armory of carbohydrate active enzymes (CAZymes) (37–40), and so it is likely that the Bacteroidetes strain present in the wheat bran-attached community also has a high enzymatic potential. The carbohydrate catabolism of Firmicutes should not be underestimated either. Roseburia intestinalis, for example, is reported to have extensive xylan utilization (43). In addition to Bacteroidetes and Firmicutes species, El Kaoutari et al. have shown that Bifidobacterium species, such as B. longum, are also involved in carbohydrate metabolism by harboring several putative GHs (40). Our results show an enrichment of an OTU assigned to B. pseudolongum on both WB and WB280, which harbor, according to the NCBI database, xylosidases and arabinofuranosidases. These enzymes enable both the breakdown of the xylan backbone and the release of the most important substituents (44).
Conclusions.We conclude that the technical modification of wheat bran, such as particle size reduction as described here, induces the colonization of a very specific butyrogenic community compared to that of nonmodified wheat bran. Bacteria with the appropriate enzyme arsenal to breakdown xylans appear to be attracted to the bran fractions. When WB280 was administered to broilers, shifts in the total cecal community were induced. The intake of fiber-rich diets has been associated with increased Firmicutes and decreased Proteobacteria numbers (45). Here, we show that applying only 1% of particle-size-reduced wheat bran increased the ratio of Firmicutes to Proteobacteria and specifically reduced cecal Enterobacteriaceae levels. This implies that in future intervention studies, one should take the particle size of DF into account.
MATERIALS AND METHODS
Modified wheat bran fractions.The particle size of wheat bran was reduced as described previously (33). In short, the particle size of commercial wheat bran (Dossche Mills, Deinze, Belgium) was reduced with a Cyclotec 1093 sample mill (Foss, Höganäs, Sweden). By changing the grinding ring and/or the mesh size of the final sieve of the mill, bran fractions with different particle sizes were obtained. The particle size distribution of the resulting fractions was determined by sieving 20.0 g of bran on a set of sieves with mesh sizes of 4,500, 2,000, 1,000, 710, 500, 400, 250, 200, 160, 125, 112, 90, 50, and 38 μm. The set of sieves was shaken for 30 min at a frequency of 1.5 s−1 with a Retsch vibratory sieve shaker (Aartselaar, Belgium), after which, each sieve was gently brushed to avoid the clogging of the sieve pores. The mass of the bran that remained on each sieve was determined and used to calculate a mass-based average particle size of dav = ∑di × mi, where di = (upper size limit of bran fraction i + lower limit of bran fraction i)/2 and mi is the mass fraction, i.e., mass on sieve i/sum of masses of all fractions.
Regular wheat bran (1,690 μm) and wheat bran with an average reduced particle size of 280 μm, here referred to as WB and WB280, respectively, were used in the experiments described below.
In vitro digestion and fermentation.The different wheat bran fractions (WB and WB280) were predigested in vitro using a protocol previously described by Wu et al. (46) that was slightly adapted. First, 1 g was incubated with 1.5 ml 0.03 M HCl (40°C for 30 min) to mimic the initial stages of digestion in the crop. Second, the digestion in the proventriculus and gizzard was simulated by incubating the substrate with 3,000 U of pepsin (Sigma-Aldrich, St. Louis, MO) in 1 ml 1.5 M HCl (40°C for 45 min). To simulate digestion in the duodenum, 1 ml 1 M NaHCO3 and 3.7 mg pancreatin (Sigma-Aldrich) were added and incubated for 2 h at 40°C. Subsequently, the predigested bran fractions were centrifuged (5 min at 5,000 × g) and washed twice with 20 ml of Aqua Dest. The resulting pellet was retained and lyophilized. The in vitro fermentations were performed using a nutrient-poor medium described by Moura et al. (47) with minor modifications, previously described by De Maesschalck et al. (3). The pH of the medium was adjusted to 6.5. Four-week-old Ross 308 broilers were euthanized by intravenous injections of an overdose of sodium pentobarbital 20% (Kela, Hoogstraten, Belgium). The ceca were isolated and put immediately in an anaerobic cabinet (Ruskinn Technology, Bridgend, United Kingdom) with 84% N2, 8% H2, and 8% CO2 at 37°C. The cecal content of an individual chicken, diluted 1,000-fold, was used as the inoculum. The lyophilized predigested fractions were supplemented to the nutrient-poor medium at a concentration of 1% (wt/vol). Nonsupplemented medium was used as a control (here referred to as control). After a 24-h anaerobic incubation at 37°C, the bran fraction was collected by centrifugation (5 min at 700 × g). To remove all non- and loosely attached bacteria, the bran fractions were washed using a protocol previously described by Leitch et al. (14) and slightly adapted by De Paepe et al. (19). The experiment was repeated seven times using cecal contents obtained from seven different chickens.
In vivo experiment.Thirty 1-day-old Ross 308 broilers were obtained from a local hatchery and housed in three containers on wood shavings. A commercial standard wheat/rye-based feed and drinking water were provided ad libitum. Group one received no supplement. The feeds of groups two and three were supplemented with 1% unmodified wheat bran (1,690 μm) and 1% wheat bran with an average reduced particle size of 280 μm, respectively. The chickens were euthanized by intravenous injections of an overdose of sodium pentobarbital 20% on day 10 of the experiment, and cecal contents were collected for microbiota analysis. There was no need for approval of the ethical committee, since no experimental procedures were performed on live animals.
DNA extraction.DNA was extracted from cecal contents and wheat bran-associated bacteria using the cetyltrimethylammonium bromide (CTAB) method as described previously by Griffiths et al. (48) and Kowalchuk et al. (49). To 100 mg of cecal content or 100 mg of washed bran, 0.5 ml CTAB buffer (hexadecyltrimethylammonium bromide 5% [wt/vol], 0.35 M NaCl, 120 nM K2HPO4) and 0.5 ml phenol-chloroform-isoamyl alcohol (25:24:1) (Sigma-Aldrich) was added, followed by homogenization in Eppendorf tubes. The samples were shaken 6 times for 30 s using a beadbeater (Magna Lyser, Roche, Basel, Switzerland) at 6,000 × g with 30 s between shakings. After centrifugation (10 min at 8,000 × g), 300 μl of the supernatant was transferred to a new tube. Then, samples were reextracted with 250 μl CTAB buffer. After homogenization, the samples were centrifuged (10 min at 8,000 × g), and 300 μl of supernatant was added to the first 300 μl. Phenol was removed by adding an equal volume of chloroform-isoamyl alcohol (24:1) (Sigma-Aldrich). The aqueous phase was transferred to a new Eppendorf tube. Nucleic acids were precipitated with 2 volumes of PEG 6000 solution (polyethylene glycol 30% [wt/vol], 1.6 M NaCl) for 2 h at room temperature. After centrifugation (20 min at 13,000 × g), the pellet was rinsed with 1 ml of ice-cold 70% (vol/vol) ethanol. The pellet was dried and resuspended in 100 μl RNase-free water (VWR, Leuven, Belgium).
16S rRNA amplicon sequencing.The 16S rRNA sequencing using MiSeq v2 technology (2 × 250 bp) technology from Illumina was performed at the GenoToul Genomics and Transcriptomics facility (Auzeville, France). The primers used were 343F (5′-CTTTCCCTACACGACGCTCTTCCGATCTACGGRAGGCAGCAG-3′) and 784R (5′-GGAGTTCAGACGTGTGCTCTTCCGATCTTACCAGGGTATCTAATCCT-3′) targeting the hypervariable 16S rRNA V3 to V4 region (50). The amplification mix contained 5 U of FastStart high-fidelity polymerase (Roche Diagnostics, Vilvoorde, Belgium), 8 μl of deoxynucleoside triphosphate (dNTP) mix (250 μM; Eurogentec, Liège, Belgium), 2 μl of each primer (20 μM), and 100 ng of genomic DNA in a volume of 100 μl. The thermocycling conditions consisted of a denaturation at 94°C for 2 min followed by 30 cycles at 94°C for 60 s, 65°C for 40 s, and 72°C for 30 s and a final elongation step of 10 min at 72°C. These amplifications were performed on a Mastercycler Pro (Eppendorf, Hamburg, Germany). Subsequently, DNA was purified using HighPrep PCR (MagBio Genomics Inc., Gaithersburg, MD) according to the protocol of the manufacturer. Single multiplexing was performed using a 6-bp index, which was added during a second PCR with 12 cycles. PCR products were purified, and the quality and the fragment length were checked using Agilent DNA 7500 DNA chip (Agilent Technologies, Santa Clara, CA) according to the manufacturer's protocol. The resulting products were purified and loaded onto an Illumina MiSeq cartridge according to the manufacturer's instructions (Illumina Inc., San Diego, CA). The quality of the run was internally checked using control libraries generated from the PhiX virus (Illumina PhiX control; Illumina Inc.) as recommended by Illumina.
Sequence processing.Demultiplexing of the amplicon data set and deletion of the barcodes were performed by the sequencing provider. Primers were removed using the FastX toolkit (51). Forward and reverse reads were merged using PEAR (52). Raw Illumina forward and reverse reads were also trimmed and merged by the sequencing provider. In the following steps, different programs of the Usearch software v7.0.1090 were used (53). Merged sequences were quality filtered with a maximum expected error of 3 with the “fastq_filter” option. Next, the sequences of all samples that needed to be compared to each other were merged, dereplicated, and sorted by size. The reads were clustered into operational taxonomic units (OTUs) using Uparse, with a similarity level of 97% (53, 54). Chimeras were removed using “uchime_ref” with the RDP Gold database as a reference (55). Finally, the sequences of individual samples were mapped back to the representative OTUs using the “usearch_global” algorithm at 97% identity and converted to an OTU table (56). OTU tables of the 16S rRNA amplicon sequencing were analyzed using the QIIME software package (v1.9.0) (57). Taxonomy was assigned with the script “assign_taxonomy.py” using the uclust method considering a maximum of 3 database hits, with the Silva v119 97% rep set (as provided by QIIME) as a reference. Representative bacterial OTU sequences were aligned to the SILVA 97% rep set using the PyNast algorithm with QIIME default parameters (58, 59). Rarefaction analysis was conducted using the “alpha_rarefaction.py” script and indicated that a sequencing depth of 10,000 reads was sufficient to analyze both the attached and planktonic communities in vitro and the bacterial community in the cecal samples from broilers.
Downstream data processing and statistics.Multivariate analysis was performed using the R package vegan (version 2.0-10) (60). The OTU tables were normalized by removing those OTUs with an abundance lower than 0.01% in all samples. Dissimilarity matrices (based on the Bray-Curtis dissimilarity index) were calculated from the OTU tables. The beta-diversity of the bacterial communities was studied by doing a permutational multivariate analysis of variance (PERMANOVA) and a principal-coordinate analysis (PCoA) on these dissimilarity indices.
To determine significant differences in the compositions of adhered and control communities, a Kruskal-Wallis test, followed by Dunn's multiple-comparison test, was performed using SPSS Statistics 23 (IBM Corp., Armonk, NY).
Illumina data obtained from the in vivo trial were log transformed, and a one-way analysis of variance (ANOVA) was used to determine the statistical differences in the relative abundances of the different bacterial families, followed by a Tukey post hoc test.
Functional differences based on bacterial 16S rRNA community composition were assessed with METAGENassist (61). Input files were created in QIIME. OTUs were assigned, mapped and condensed into functional taxa, and filtered on the basis of interquantile ranges after Pareto scaling. These data were then analyzed for “metabolism phenotype,” and the Euclidian distance measure was used to visualize the results in a heatmap.
OTUs that were significantly enriched on bran material were identified at the species level by using NCBI BLAST. For every hit, a reference genome was chosen and the NCBI protein databases were searched for the presence of xylan degradation enzymes, including xylanases, feruloyl esterases, arabinofuranosidases, acetyl xylan esterases, glucuronidases, and xylosidases.
Quantitative PCR for total bacteria and butyryl-CoA:acetate CoA-transferase gene.To quantify the number of total bacteria, the primers Uni331F (5′-TCCTACGGGAGGCAGCAGT-3′) and Uni797R (5′-GGACTAACCAGGGTATCTAATCCTGTT-3′) were used (62). Amplification and detection were performed using the CFX384 Bio-Rad detection system (Bio-Rad, Nazareth-Eke, Belgium). Each reaction was run in triplicates in 12-μl total reaction mixtures using 2× SensiMix SYBR No-ROX mix (Bioline, Kampenhout, Belgium), 0.5 μM final primer concentration, and 2 μl of (50 ng/μl) DNA. The amplification program consisted of 1 cycle at 95°C for 10 min followed by 40 cycles of 1 min at 94°C, 1 min at 53°C, and 2 min at 60°C. A melting curve analysis was performed after amplification and was obtained by slow heating from 60°C to 95°C at a rate of 0.5°C/5 s to confirm the specificity of the reaction. To quantify the number of gene copies encoding the butyryl-CoA:acetate CoA-transferase (CoAt) gene, the primers BCoATscrF (5′-GCIGAICATTTCACITGGAAYWS-3′) and BCoATscrR (5′-CCTGCCTTTGCAATRTCIACRAANGC-3′) were used (63). The final primer concentration was 2.5 μM, and the amplification program consisted of 1 cycle at 95°C for 10 min followed by 40 cycles of 30 s at 95°C, 30 s at 53°C, and 30 s at 72°C. The relative number of CoAt gene copies in the samples was obtained by calculating the ratio of CoAt gene copies to the total number of bacteria for each sample. To determine significant differences in relative CoAt gene copies, a Kruskal-Wallis test was performed, followed by a Dunn's multiple-comparison test. Statistical analysis was done with SPSS Statistics 23 (IBM Corp.).
ACKNOWLEDGMENTS
This work was supported by funding from Vlaams Agentschap Innoveren en Ondernemen (Vlaio, former IWT or Agentschap voor Innovatie door Wetenschap en Technologie) of the SBO BRANDING project (grant no. 130028).
We thank Evy Goossens for helping out with the sequencing data.
K. Vermeulen, J. Verspreet, C. M. Courtin, F. Van Immerseel, and R. Ducatelle are listed as coinventors on a patent application for a wheat bran fraction for use to combat Salmonella infection (international application number PCT/EP2017/067028).
K.V. conceived the study, performed animal trial and laboratory assays, and wrote the manuscript. J.V. provided wheat bran samples and edited the manuscript. C.M.C. conceived the study, provided wheat bran samples, and edited the manuscript. R.D., F.V.I., and F.H. provided resources, conceived the study, and edited the manuscript. A.H. and S.B. analyzed sequencing data and edited the manuscript. All authors have read and approved the final version of the manuscript.
FOOTNOTES
- Received 2 June 2018.
- Accepted 24 July 2018.
- Accepted manuscript posted online 24 August 2018.
Supplemental material for this article may be found at https://doi.org/10.1128/AEM.01343-18.
REFERENCES
- Copyright © 2018 American Society for Microbiology.
