Introduction

In recent years, the health impact of intestinal and cecal microbiome composition has become a prominent research focus in poultry science [29], 42]. Understanding and modulating the intrinsic and extrinsic interplay between differential microbial populations and their host environment has led to improved animal health and greater profitability in agricultural endeavours [23]. At present, broiler chickens (Gallus gallus subsp. domesticus; “broilers”) constitute the most consumed meat worldwide, with an approximate 100 million tons of poultry meat produced annually [67]. Due to their economic importance, high nutritive value, and accessibility of their meat, broilers have been extensively subjected to, and immensely benefitted from, intestinal microbiome composition and modulation analyses [11,12,13, 74]. The combined efforts of such research endeavours have reduced chick mortality, increased growth rates, and reduced the microbial load of major poultry and human pathogens [17, 52, 93]. Efforts of particular importance (and success) involve modulating microbiome composition using feed supplements [11].

The holobiont theory suggests that the health, metabolic prowess, and overall success and survivability of an organism is largely influenced by the composition, diversity, and complexity of their associated microbiomes [80]. Most previous studies of the chicken gut microbiome have focused on the ceca due to their dense bacterial populations which aid in digestion of otherwise indigestible residues remaining in chyme, bioconverting them to digestible metabolites for host absorption (eg. digestion of cellulose to glucose; [50, 69, 78, 84]). Many studies have found that differing microbiome compositions are strongly correlated with disease states across Metazoan lineages [22, 48, 78], and their modulation (via nutrient supplementation or transplantation) has resulted in profound improvements in human and animal health [11, 29, 52].

Due to increases in antimicrobial and metal or biocide resistance arising from their systematic use and misuse as livestock growth promoters and over-prescription in human medicine, alternative growth promotion techniques and supplements are being explored without using clinically relevant compounds [96]. One of the most promising poultry feed supplements are prebiotics containing mannan, such as mannan rich fraction (MRF) derived from Saccharomyces cerevisiae cell wall residues [8, 41]. These compounds display particular efficacy in binding to type-1 fimbriae in Gram-negative bacterial pathogens, specifically Enterobacteriaceae [28]. Reduction of such populations allows mutually symbiotic and commensal microbiota such as Lactobacillus to flourish [10, 29, 74, 93].

While 16S microbiome studies are highly informative, there may be some bias and lost significance due to the disparate number of 16S rRNA genes between species [95]. As these differences can be quite pronounced, we constructed a large 16S rRNA dataset from publicly available bacterial genomes and devised a simple weighting system, where each read from each taxon was divided by the number of 16S rRNA genes available in each taxon. The purpose of this procedure was to reduce bias caused by the widely uneven 16S rRNA gene counts commonly observed across Domain Bacteria.

Avian gut microbiome reports display considerable animal-to-animal variation which has the potential to incorrectly bias post hoc statistical comparisons [30, 99]. To counter this problem, we employed isolation forests (a common machine learning technique) and median imputation to each sample to remove and replace any outliers to decipher any previously unseen underlying trends [56]). The aim of this study was to investigate the impact of MRF addition on the microbial communities of the three main GI nutrient absorption sites (duodenum, jejunum, and ileum) of the intestinum tenue (“small intestine”) and the cecum in broilers. By removing intrinsic and extrinsic biases from 16S rRNA gene counts and cumulative community structures, we aim to highlight otherwise overlooked microbial taxa that may be of importance in food safety microbiology.

Methods

Sample collection and preservation

This broiler trial was performed at a commercial production site within the European Union. On the day of hatch, chicks were taken from a commercial hatchery and transported to an associated commercial farm. Approximately 35,000 birds were placed from the hatchery into each of two sheds where they received a control standard commercial wheat-soya diet or a standard diet plus MRF (Alltech Biotechnology) at the following inclusion rates; 1300:1000:600 gt−1 starter, grower, and finisher rations respectively. Birds were raised and fed as per typical commercial production conditions receiving feed and water ad libitum. All other conditions were kept uniform for both sheds. At day 35 (post-hatch) the intact gastrointestinal tracts of 12 randomly caught birds per shed were excised immediately after humane euthanisation. Intestinal contents from the duodenum, jejunum, ileum, and cecum were massaged into individual sterile tubes, immediately frozen on dry ice, transported within 8 h and stored at −80 °C for downstream processing.

DNA extraction and 16S rRNA gene sequencing

DNA was extracted from intestinal contents using the QIAamp DNA Stool Mini Kit according to the manufacturer’s instructions using 0.05 g of intestinal content (QIAamp DNA Stool Mini Kit, Qiagen). Genomic DNA concentration was determined at a wavelength of 260 nm using a NanoDrop (NanoDrop). Isolated DNA was then used as a template in PCR amplification for construction of 16S rDNA libraries which were prepared and sequenced by BaseClear genomics. Sequencing libraries were prepared by amplification and barcoding of the 16S rRNA gene V3–V4 region and the resulting amplicons were sequenced on an Illumina MiSeq platform generating 10–50 k PE300 reads per sample. The mean library size used was 580 bp (inclusive of barcodes and adapters) and the insert size was approximately 460 bp (580–120 = 460 bp). A total 3,988,410 reads were achieved. In the control dataset, average reads for each of the duodenum, jejunum, ileum and cecum were observed to be 41,749.42 ± 6442.53, 45,074.75 ± 6468.97, 42,135.83 ± 7449.29, and 48,489.92 ± 4364.9 respectively. Comparatively, in the MRF-treated dataset average reads of 35,883.92 ± 4765.3, 35,644.5 ± 9590.25, 43,873.5 ± 6593.51, 39,495.67 ± 8224.7 for the duodenum, jejunum, ileum, and cecum were observed.

Dataset construction

Each sample was adapter and quality trimmed using TrimGalore! v.0.6.6 [54]under default settings and powered with cutadapt v.3.0 [64] and FastQC v.0.11.9 [7]. Between 14,044 and 54,118 reads were observed pre-quality-control and between 13,868 and 53,830 after, with an observed percentage read discard range between 0.325% and 9.66%. Chimeras were identified using UCHIME v.4.2.40 [36] and removed. Quality controlled reads were merged using the “–fastq_merge” function in VSEARCH v.2.14.2 [37, 79 to give a single entry for each read pair in FASTA format.

16S rRNA database construction

A database of 16S rRNA genes was constructed by downloading all bacterial genome assemblies (n = 274,268) from NCBI assembly [51] and extracting all 16S rRNA genes using Barrnap v.0.9 (as used for rRNA detection by Prokka v.1.1.14 [83]) with default settings. Taxonomic lineages were assigned to each genome (and their associated genes) using the “lineage” function in TaxonKit v.0.6.0 [87] and standardised to the seven ranks (Domain, Phylum, Class, Order, Family, Genus, and Species) using the TaxonKit “reformat” function. Sequences with length less than 1200 nucleotides (nt) were discarded to mirror the strict filtering methods employed during the construction of the SILVA database [75]. Remaining sequences were searched against all other remaining sequences using the “–usearch-global” function in VSEARCH v.2.14.2 with a minimal percentage identity stringency score of 0.97 (97%), self-hits were excluded, and, with the exception of Escherichia, Shigella, and Salmonella (ESS), top-hit pairs where sequences were observed to be from different genera were discarded. The ESS species were excluded from further filtration during this step due to the close evolutionary relatedness of these clinically relevant genera [38, 43, 91]. Finally, exact duplicates of 16S rRNA genes were removed resulting in a database of 68,724 16S rRNA genes from 21,928 species from 37 definite phyla and 70 candidate phyla/divisions (107 in total). This dataset is available for download at (https://github.com/RobLeighBioinformatics/Broiler_GI_microbiome).

Database weighting

Bacterial genomes are highly dynamic due to rapid gene duplication, loss, and horizontal transfer events which may result in varying numbers of 16S rRNA genes [95]. Alien and spurious 16S rRNA genes were removed during database construction, so it is anticipated that all genes in the database were chromosomal in origin. Species were weighted by the number of 16S rRNA genes remaining in each genome after the strict filtration steps during database construction. The median number of 16S rRNA genes was taken where multiple genomes from the same species were retained. Genera weighting was calculated by excluding all genomes not definitively identified to species level (eg. genomes labelled “Salmonella sp.” (as opposed to, for example, Salmonella enterica or “undefined Lactobacillaceae”) and assumed to be the median for all species in a given genus. For higher taxonomic ranks, the median of rank medians was taken (eg. for families, the median of all genera medians in each family was taken). This method was employed to prevent biasing from well sampled species in a genus compared to less common species (eg. Escherichia coli vs. Escherichia marmotae). This weighting table is available at https://github.com/RobLeighBioinformatics/Broiler_GI_microbiome.

Taxonomic assignment and weighting

Each read entry was searched against our 16S rRNA database using the “–usearch-global” function in VSEARCH and top hits with an alignment stringency cut-off of 0.97 (97%) were extracted (Additional file 1: Tables S1–S6). To mitigate taxonomic misassignment, the stringency cut off was increased to 0.99 (99%) for species level assignment. Read counts were then weighted using the 16S rRNA gene counts calculated above (Additional file 1: Tables S7–S12). The proportion of each weighted taxon in each sample was computed and normalised (closed) by dividing by a “closure constant” (CC) for each sample and dividing each weighted read count per taxa by the closure constant (Additional file 1: Tables S13–S18). This standardisation ensures all samples have the same number of reads for downstream comparative analysis. The standardisation constant was constructed using the formula:

$${\text{CC}} = \frac{{\Sigma_{x} }}{{\max \left( {\Sigma_{{x_{1} }} , \Sigma_{{x_{2} }} , ..., \Sigma_{{x_{n} }} } \right)}}; \;{\text{CC}} \le 1$$

where x: Series of reads in a sample/replicate.

Outlier processing

Due to the extensive intersample variation observed in microbiome studies [99], as discussed previously, we endeavoured to remove extreme outliers to examine potential underlying trends that may be otherwise obfuscated. Outliers were removed and imputed with the median of the remaining inliers using uniForest v.1 with default parameters [56].

Fold changes

For all comparisons made below, median fold changes (ηFC) were calculated using the formula:

$$\eta_{{{\text{FC}}}} = \frac{{\eta_{(b)} - \eta_{(a)} }}{{\eta_{(a)} }}$$

where η(x): Median observation for group x.

Fold changes have a lower limit of −1 (complete depletion) and no change is represented by 0. A FC is incalculable if η(a) = 0 as this represents a complete introduction.

Statistical analysis

Kolmogorov-Smironov tests [53, 92] using a Lilliefors’ distribution [59] were used to determine sample series distribution normality (H0:X ~ N(μ,σ2;HA:X≁N(μ,σ2); P > 0.05: X ~ N(μ,σ2)) and as all distributions were determined to follow a non-normal distribution, Brunner–Munzel tests [20] were used to compare taxa between the control and MRF treated datasets H0:B = 0.5;HA:B ≠ 0.5). A Brunner–Munzel test was used instead of a Mann–Whitney U test [62] as the data was assumed to have unequal variance due to the high level of variability usually observed in microbiome analyses [99]. A Bonferroni–Dunn (BD; PBD) correction [16, 35] was applied to each test (PBD = P × ncomparisons) and instances where PBD ≤ 0.05 were considered to be statistically significant (Additional file 1: Table S19) and the FC (as described above) was used to indicate the trend changes. Different ncomparisons were used to calculate PBD (by taxonomic rank) to strengthen confidence in results at lower taxonomic ranks, however, to restrict an overly stringent correction, statistical comparisons were only performed when ηControl or ηMRF > 20 (or ηsite(a) or ηsite(b) > 20).

Ecological statistics

A bias-corrected Chao1 richness estimator [24], Simpson’s D index [90], Simpson’s E index [90], and Shannon’s H index [85] was calculated for each anatomical site in each dataset at each taxonomic rank using the sklearn-bio (skbio) v.0.2.0 Python library (http://scikit-bio.org/). A Brunner–Munzel test (H0:B = 0.5;HA:B ≠ 0.5) was performed between diversity indices at each rank. A Bonferroni–Dunn correction was performed for each subset (ncomparisons = 4) and instances where PBD ≤ 0.05 were considered statistically significant (Additional file  1: Table S20). Statistical trend changes were determined using the FC calculation described above.

A principal component analysis [47, 73] (PCA) was performed between all data subsets at each site using the “PCA” module in the “sklearn.decomposition” Python machine learning library. A permutational analysis of variance [4] (PERMANOVA) was used to compare control vs MRF treated samples. A PERMANOVA is used to compare the centroid and dispersion of two groups based on the 2 dimensional (2D) or 3D coordinates of their points using 999 iterations (in = 999). A Bonferroni–Dunn correction was applied (ncomparisons = 4) and a PBD ≤ 0.05 was considered statistically significant (Additional file 11: Table S21).

A Bray–Curtis distance matrix [19] was constructed between control and MRF-treated datasets for each anatomical site using the “beta_diversity” driver function from the “skbio.diversity” Python library and a principal coordinate analysis (PCoA) was performed on each distance matrix using the “pcoa” function from the “skbio.stats.ordination” package. A PERMANOVA was used to compare control vs MRF treated PCoA groups using 999 iterations (in = 999) as is common practice. A Bonferroni–Dunn correction was applied (ncomparisons = 4) and a PBD ≤ 0.05 was considered statistically significant (Additional file  1: Table S21).

Short chain fatty acid concentration analysis

The concentrations of three short chain fatty acids (SFCA; acetate, propionate, and butyrate) in cecal digesta was measured using gas chromatography after metaphosphoric acid derivation as previously described with minor modifications [77]. Briefly, 0.20 g of thawed sample was diluted with 2 mL double-distilled water in a sterile screw-capped tube, then homogenized, and centrifuged at 4000 × g for 10 min at 10 °C. A volume of 1 mL of supernatant was then transferred to another Eppendorf tube and mixed with 0.2 mL, 25% (wt/vol) ice-cold metaphosphoric acid solution. Subsequently, this solution was kept at − 20 °C for 4 h. Samples were then thawed, 0.1 mL 4 M sodium hydroxide solution added and centrifuged at 4000 × g for 10 min at 10 °C before analysis. The supernatant was then filtered with a 0.22 μm membrane, and an injection volume of 0.4 μL of sample solution was analyzed using a gas chromatography (Agilent 7890A system) coupled with a CP-Wax 58 FFAP CB column (Agilent) and flame ionization detector to determine SCFA concentrations in cecal content. The concentrations of acetate, propionate, and butyrate were calculated and expressed as μmol/g of wet cecal digesta.

Again, Kolmogorov-Smironov tests (using a Lilliefors’ distribution) were used to determine sample series distribution normality (H0:X ~ N(μ,σ2);HA:X≁N(μ,σ2); P > 0.05: X ~ N(μ,σ2)) for control and MRF-treated SFCA concentration series. Equivarience was assessed using a Levene’s test (H0:σ2a = σ2b;σ2a ≠ σ2b) [57]. As equivariance was not observed between any pair and as one distribution (MRF-treated acetic acid) was determined to follow a non-Gaussian distribution, Brunner-Munzel tests were used to compare each taxon between the control and MRF treated datasets H0:B = 0.5;HA:B ≠ 0.5) (Additional file 1: Table S22).

Results

Broiler growth characteristics

The growth indices of the MRF supplemented broilers were compared with the control (Table 1). Feed conversion ratios and average live weights did not differ significantly between the two groups however, the MRF supplemented birds were on average 5 g heavier and finished 1 day earlier than the control group. Birds supplemented with MRF tended to have a greater European production efficiency factor (EPEF).

Table 1 Comparison of growth indices of broiler commercial units with and without MRF dietary supplementation
Full size table

Effect of diet and GI tract section on α- and β- diversity

A total 3,988,410 sequence reads were recovered from the 96 samples analysed. In the control dataset, average reads for each of the duodenum, jejunum, ileum, and cecum were observed to be 41,749.42 ± 6442.53, 45,074.75 ± 6468.97, 42,135.83 ± 7449.29, and 48,489.92 ± 4364.9, respectively. Comparatively, in the MRF supplemented dataset average reads of 35,883.92 ± 4765.3, 35,644.5 ± 9590.25, 43,873.5 ± 6593.51, 39,495.67 ± 8224.7 for the duodenum, jejunum, ileum, and cecum, respectively.

Microbial diversity at the four anatomical sites was estimated using α-diversity indices (Chao1 index, Simpson’s E (evenness), and Shannon’s H’ index). Chao1 was used to estimate richness (Fig. 1a), Shannon's H’ index was used to indicate diversity (Fig. 1(b..)) and Simpson’s E was used to indicate evenness (Fig. 1(c.); Additional file 1: Table S20). Richness was observed to be significantly increased in the MRF-treated ceca (Chao1:ηFC = 0.1311) and significantly lower in MRF-treated duodena (Chao1:ηFC = -0.3072) and jejuna (Chao1:ηFC = −0.2241) respectively. Evenness was not observed to be significantly affected by MRF-addition and the ileum was not observed to be modulated post-treatment.

Fig. 1
figure 1

a–c Four α-diversity metrics displayed for the four anatomical sites explored in this study. Statistically significant (PBD ≤ 0.05) results are highlighted with an asterisk

Full size image

Differences in β-diversity within the intestinal microbial population between groups and between intestinal sections within groups were assessed using PCoA (Figs. 2 and 3). The PCoA plots shown in Fig. 2a–d show that the bacterial community composition at the species level differed significantly (PBD ≤ 0.05) as a result of diet in each intestinal section with PC1 accounting for 60.1%, 69.28%, 49.13% and 91.32% of the total variation; PC2 accounting for 18.61%, 8.36%, 17.78% and 3.17%; and PC3 accounting for 7.38%, 5.63%, 13.48%, and 1.74% in the duodenum, jejunum, ileum, and cecum respectively. The bacterial community composition between intestinal sections was also analysed for differences and showed that each intestinal section harboured a distinct bacterial community structure regardless of diet (Fig. 3a, b, PBD ≤ 0.05).

Fig. 2
figure 2

a–d Species-level Bray–Curtis distance matrices (β-diversity) expressed as PCoA between control (red) and MRF-supplemented broilers (blue) at each anatomical site

Full size image
Fig. 3
figure 3

a, b Species-level Bray–Curtis distance matrices (β-diversity) expressed as PCoA between duodenal (red), jejunal (orange), and ileal (grey) anatomical sites across the control (left) and MRF-supplemented (right) datasets

Full size image

Effect of diet and GI tract section on bacterial community composition

To determine which bacterial taxa contributed to separating bacterial communities based on diet and intestinal section, the phylum level relative abundances of each GI tract were considered (Table 2). At the phylum level, four main bacterial phyla were identified within each gastrointestinal section, Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria (newly renamed as Actinomycetota Bacteroidota, Bacillota, and Pseudomonadota, respectively [72]). Phylum Firmicutes was the predominantly abundant phylum within each GI section. Following MRF supplementation, Firmicutes were significantly lower in the duodenum, and significantly greater in the cecum. Actinobacteria was identified as the second most abundant phylum in all control group anatomical sites but was significantly lower in the duodenum and cecum as a result of MRF supplementation. Proteobacteria were significantly greater in the duodenum and significantly lower in the ileum following MRF addition to the diet. Finally, Bacteroidetes was predominantly detected in the cecum compared to any other site.

Table 2 Relative abundances of bacterial phyla obserbed in each anatomical site in both control and MRF supplemented broilers
Full size table

The top 10 most abundant bacterial genera and species for each GI tract section in control and MRF supplemented groups are shown in Tables 3 and 4 respectively. At the genus level the most abundant genera within the intestinum tenue in both control and MRF supplemented groups were Lactobacillus followed by Bifidobacterium (> 90% abundance combined). In the MRF supplemented birds the duodenum samples were dominated by Proteobacterial genera Pseudomonas, Halomonas, and Shewanella. For the control dataset the most abundant species within the intestinum tenue were Bifidobacterium animalis, Lactobacillus crispatus, and Lactobacillus salivarius (accounting for a η% > 65%). Comparatively, in the MRF-treated sample dataset, each intestinum tenue site had a distinct set of predominant species (Bifidobacterium animalis, Lactobacillus aviarus, Lactobacillus crispatus, and Lactobacillus kitasatonis; (however these were observed in highly divergent η% between sites)) and alongside other species (listed below) accounted for η% < 60% in all sites. For the duodenum, Pseudomonas veronii, and Pseudomonas sp. TKP were highly observed, and for the ileum, Lactobacillus vaginalis was also highly observed. In the cecum the most abundant genus was Faecalibacterium in both control and MRF supplemented groups (> 50%) followed by Bifidobacterium and Blautia in the control group and Blautia and Lactobacillus in the MRF supplemented group. For the control cecal dataset, the most abundant observed species were Faecalibacterium sp. An122, Bifidobacterium gallinarum, and Bifidobacterium pullorum, (accounting for a η% > 65%). In the cecal MRF supplemented dataset, the most prominent species (accounting for η% > 69%) were Faecalibacterium sp. An122, Blautia sp. An81, and Eubacterium sp. An11.

Table 3 The (ten) most prevalent bacterial genera observed at each anatomical site in both control and MRF-treated datasets
Full size table
Table 4 The (ten) most prevalent bacterial species observed at each anatomical site in both control and MRF-treated datasets
Full size table

The relative abundances of several bacterial genera and species were significantly different with MRF supplementation (Tables 5 and 6, respectively). Notably, the bacterial genus Escherichia was significantly lower in the duodenum and ileum (numerically lower in jejunum and cecum, Additional file 1: Table S19). Genus Shigella was significantly lowered in the ileum, while the genus Bifidobacterium was significantly lowered in the duodenum and cecum. Whilst the genus Lactobacillus was noted to be significantly lower in the duodenum it was significantly greater in the cecum in MRF supplemented birds. Similarly, the genera Anerostipes, Kineothrix, and Blautia were noted to be significantly greater whilst Alistipes was significantly lower in the cecum of MRF supplemented birds when compared to the control. Genus Clostridioides was noted to be significantly lowered while other genera including Shewanella, Pseudomonas, and Halomonas were greater in the duodenum. Genera Streptococcus and Agarivorans were also significantly lower in the ileum of broilers supplemented with MRF. At the species level, the relative abundances of several bacteria were significantly different with MRF supplementation (Table 6). Of note, Escherichia coli and Clostridoides difficile were significantly lower across all three intestinum tenue sites following MRF supplementation. In the duodenum and jejunum, Bifidobacterium gallinarum was significantly lower, whereas Bifidobacterium gallinarum and Bifidobacterium pullorum were significantly lower in the cecum. Modulations in Lactobacillus species were observed throughout the GI tract following MRF supplementation. Of interest, L. reuteri, was observed to be significantly lower in the duodenum but significantly greater in the ileum and cecum and L. salivarius, was observed to be lower across the entire GI tract. The species Barnesiella intestihominis was noted to be significantly lower in the caeca of MRF-treated birds (compared to control birds), whereas Blauta sp. An81, which is strongly associated with weight gain, was observed to be significantly greater in both the cecum and jejunum. As mentioned above, Escherichia coli and Clostridoides difficile were observed to be significantly lower in the duodenum whereas Pseudomonas veronii, Halomonas axialensis, and Shewanella algae were significantly greater. After MRF-treatment, Shigella flexneri was observed to be significantly lower in the ileum.

Table 5 Significantly altered (increased or decreased) genera observed at each anatomical site
Full size table
Table 6 Significantly altered (increased or decreased) species observed at each anatomical site
Full size table

To investigate the gut microbial community in different GI tract sections analysis of the common and unique OTUs was conducted, shown in the Venn diagrams (Fig. 4). A total of just 22 OTUs were shared by all 4 chicken gut sections in both the control and MRF supplemented groups. The number of OTUs observed in only one chicken gut section varied from 1 to 84, with the jejunum having the least amount of unique OTUs in both control (2) and MRF (1) supplemented groups and the cecum having the greatest amount of unique OTUs in both control (66) and MRF (84) supplemented groups. Neighbouring GI tract sections shared very few common OTUs with duodenum-jejunum sharing 8 and 4 OTUs, jejunum-ileum sharing 4 and 9 OTUs and ileum-cecum sharing 2 and 4 OTUs in control and MRF supplemented groups, respectively.

Fig. 4
figure 4

a, b Venn diagram showing common and shared species-level OTUs within each GI tract section for both control and MRF supplemented broilers (99% sequence identity)

Full size image

Effect of diet on cecal short chain fatty acids

Cecal propionate was significantly greater (ηFC = 0.176) and cecal butyrate was numerically greater (ηFC = 0.009; PBD = 1) in MRF supplemented birds when compared to the control (Fig. 5). No significant statistical differences in the concentrations of cecal acetate or total SCFA concentrations were observed between the control and MRF supplemented birds (PBD > 0.05).

Fig. 5
figure 5

Short-chain fatty acid (SCFA) concentration in broiler ceca. Statistical significance is denoted using an asterisk

Full size image

Discussion

A large and diverse microbial community inhabits the broiler GI tract and contributes to overall health and growth efficiency by controlling pathogens, enhancing nutrient availability, and modulating immunological pathways (Borda-Molina, Seifert and Camarinha-Silva, 2018). Gastrointestinal microbiome composition and diversity is influenced by many external factors (eg. environment, age, breed, antibiotic use or dietary supplementation) which may yield beneficial or maleficial consequence [102]. In this study, the impact of MRF dietary supplementation on broiler GI tract microbiota (across the intestinum tenue and ceca) was explored. Supplemented birds were observed to finish one day earlier with higher average weight (5 g) and EPEF than their control counterparts (Table 1; indicating improved bird health and producer economic potential.

Bacterial species α-diversity indices of richness, diversity and evenness are scalable metrics of health status with higher diversity negatively correlated with dysbiosis [31, 52, 98]. Comparatively, β-diversity metrics are also measures of health, where low values are expected between samples and higher values are expected between treatment groups [26, 27]. Increased α-diversity and lower β-diversity in broilers can be achieved using pre- and probiotics, and such strategies positively correlate with improved FCR and feed efficiency [2, 46, 49, 94]. The results from this study agree with previous studies, whereby α- and β-diversity differ between anatomical site [25, 86, 101]. In particular, the cecum was observed to be most diverse, and the ileum to be least diverse of the four sites, and MRF impacted cecum α-diversity more than any intestinum tenue site (Fig. 1(a.-c.)). Despite the lack of intersectional paries, each section of the unidirectional intestinum tenue displays differential absorptive properties, yields dynamic environmental conditions (e.g. pH, water content, chemical profiles, and available O2 content [60]) and microbial compositional profiles [65]. As the intestinum tenue maintains a continual flow, perhaps it is not surprising that α-diversity is less impacted than the cecum which displays a cul-de-sac architecture.

Abiotic stressors or infection can reduce α-diversity, leading to dysbiosis [23, 45]; broiler cecal α-diversity reduction typically coincides with reductions in Lactobacillaceae and an increase in Enterobacteriaceae [21, 39]. While MRF supplementation effect on the intestinum tenue has not been explored prior to this study, the observed cecal results (highlighting the dysbiotic amelioration effect of MRF via community composition alteration and increases in α-diversity) are in agreement with previously published cecal studies [26, 27]. Additionally, diversity metric trends between control group anatomical sites are also in agreement with previously published results [42, 101].

The major bacterial phyla identified in each of the four GI tract sections included Firmicutes, Actinobacteria, Bacteroidetes, and Proteobacteria, with Firmicutes being most dominant throughout each section (Table 2). Bacteroidetes was lowly represented in the intestinum tenue and was found in most abundance in the cecum, mirroring observations in previous studies [18, 101]. The major bacterial genera across the intestinum tenue were Lactobacillus and Bifidobacterium, with Lactobacillus accounting for 48%-92% across these intestinal sections. Early studies [15, 34] also reported that the intestinum tenue microbiota was dominated by Lactobacillus and their conclusions have been independently confirmed using metagenomic analyses [18, 58]. Interestingly, the most abundant species within the intestinum tenue were distinct between control and MRF supplemented groups. Bifidobacterium animalis, Lactobacillus crispatus, and Lactobacillus salivarius dominated the control dataset throughout; comparatively, in the MRF-treated dataset, each intestinum tenue site had a distinct set of predominant species (Table 3). Through efficient carbohydrate fermentation, Lactobacillus are known to provide substantial aid to host metabolism, yielding improved feed conversion ratios and reduced mortality in broilers [76],Lactobacillus also deter pathogen adhesion to the lumen walls [61, 81]. Previous studies have shown that Lactobacillus can positively influence villus height (VH), crypt depth (CD) and VH:CD in broiler intestines [6, 58]. Increased VH and VH:CD are thought to provide a larger surface area and enhance ability of nutrient absorption [32].

Short-chain fatty acids (SCFAs) play an important role in gut physiology. Increased intestinal butyrate in broilers has been shown to have many positive effects including improved energy supply, intestinal villi development, microbiome modulation, anti-inflammatory properties, and enteric pathogen control [9]. In this study, the cecum was shown to be dominated by the bacterial families Ruminococcaceae, Lachnospiraceae, and Bifidobacteriaceae in the control group and Ruminococcaceae, Lachnospiraceae and Lactobacillaceae in the MRF supplemented group, with the genera Faecalibacterium, Bifidobacterium, Blautia, and Lactobacillus being most prominent. Cecal microbiota are generally dominated by strict anaerobes with many of these bacteria belonging to SCFA producing families Lachnospiraceae and Ruminococacceae [81]. The genus Faecalibacterium is a prominent butyrate producer and is correlated with enhanced epithelial health and reduced intestinal inflammation [69, 70, 100]. Prebiotic genera Bifidobacterium, Blautia and Lactobacillus also bioconvert complex carbohydrates to SFCA for host energy utilisation [14]. Increased SFCA concentration results in a lower gastrointestinal tract pH and de-conjugated bile acids, which aid in pathogen control [9, 63], 55]. While an insignificant butyrate increase (+ 0.95%) was observed post MRF-treatment, propionate (+ 21.41%) and SFCA producing Blautia were significantly increased in the cecum (+ 69%). These results corroborate previous suggestions that increased abundance of Blautia and Faecalibacterium abundances may be related to improved growth performance [103].

Potential foodborne pathogens Escherichia coli and Clostridioides difficile were significantly lower across the intestinum tenue and Shigella flexneri in the ileum. Mannan rich fraction binds type-1 fimbriae of Enterobacteraceae, and has been shown to lower the prevalence of these pathogens in the intestine of animals [1, 8, 41]. Reducing foodborne pathogens (from any source) promotes food chain integrity, with Escherichia and Clostridioides reported as being amongst the most concerning from a One Health perspective [82, 88]. Additionally, as these species are potentially toxicogenic, synthesised toxins may travel to distal sites of the host organism and remain in meat products postprocessing [5, 44, 68, 71]. As such, any reduction in their prevalence should be viewed as a positive outcome.

The probiotic Bifidobacterium spp. were also shown to be significantly lower in the jejunum, ileum, and cecum of MRF supplemented broilers and was noted previously in the broiler cecum [27]. An interesting result observed in this dataset was a significantly greater relative abundance of Lactobacillus reuteri in the ileum and cecum. When supplemented with L. reuteri, both mammalian and poultry models were observed to have considerably reduced Enterobacteriaceae, specifically Salmonella enterica, compared to non-supplemented controls [33, 97]. In addition to bacteriological protection, L. reuteri supplementation is observed to confer antiprotozoal activity against Eimeria spp. in turkeys [33] and against another Eimeriorinan (Apicomplexan) parasite, Cryptosporidium parvum, in immunodeficient mice [3]. In previous studies, L. reuteri was strongly associated with weight gain whereas L. salivarius was strongly associated with lean maintenance [33, 89, 97]. Interestingly, L. reuteri was increased and L. salivarius was decreased in MRF supplemented birds.

Dietary MRF supplementation was observed to yield significantly greater relative abundances of cecal bacterial genera from families Lachnospiraceae, Ruminococcaceae and Lactobacillaceae. Whilst these are typical of the main bacterial families found in the broiler cecum, modulating their abundances can have profound health impacts, such as reduced inflammation, reduced intestinal atrophy, and improved mucosal barrier function [66, 81]. The significantly higher relative abundances of probiotic genera Lactobacillus and Blautia in the cecum, alongside higher relative abundances of jejunal and ileal Lactobacillus indicate MRF prebiotic action [40]. In essence, the comprehensive impact of prebiotics have important host health benefits beyond that of simple microbiota modulation.

Conclusion

This manuscript aimed to address the bird-to-bird (intersample) variation associated with microbiome studies and is the first to apply such corrections to a comparative supplementation study across intestinal geographies. Each GI tract section presented a distinct bacterial community composition which were altered as a result of MRF supplementation. Results from the present study indicated that Lactobacillus was the most abundant genus in the intentinum tenue and that the cecum was most bacterially divergent. Birds supplemented with MRF had significantly higher species richness in the cecum and significantly different bacterial community composition in each GI tract section. MRF supplemented birds had lower levels of the zoonotic pathogens Escherichia, Clostridioides, and Shigella which are of particular importance for food chain integrity. Higher levels of probiotic related bacteria, such as Lactobacillus and Blautia, were observed following MRF supplementation. Higher relative abundances of known SCFA producing bacteria (and SCFA concentrations) were also attributed to MRF supplementation. These bacterial and metabolite alterations highlight a protective role for dietary MRF inclusion to support broiler GI health and may allow safer meat to be produced.