Cecal microbiota–host phenotype associations in young White Leghorn and Silky Fowl chickens

Abstract

Genetic background and intestinal region both shape the gut microbiota, yet their relative contributions before the onset of laying remain poorly defined. This study compared intestinal traits and luminal microbiota of the duodenum, jejunum, ileum, and cecum between White Leghorn (WL) and Silky Fowl (SF) chickens at 42 days of age (20 birds per breed; 160 samples). Microbiota were characterized by 16S rRNA gene sequencing, followed by diversity, differential abundance, co-occurrence network, functional prediction, and microbiota–phenotype association analyses. Body weight did not differ between breeds, whereas WL showed higher liver, spleen, and bursa of Fabricius indices and greater length and relative length in all four segments (P < 0.05); SF showed a higher eviscerating rate (P < 0.05). Intestinal segment explained more community variation than breed, with Lactobacillus dominating the small intestine and richness and phylogenetic diversity highest in the cecum. Community composition differed between breeds in the duodenum, jejunum, and cecum (q < 0.05) but not in the ileum. Cecal richness was higher in WL, whereas Shannon diversity did not differ between breeds. In the cecum, Bacteroides was enriched in SF and Megamonas and Lactobacillus in WL (P < 0.05). Co-occurrence networks and predicted functions were breed- and segment-specific, with pathways showing higher predicted abundance in WL concentrated in the duodenum and jejunum and predicted glycan- and carbohydrate-degradation pathways showing higher abundance in SF in the ileum and cecum. Of the four segments, only the cecal predicted functional profile was associated with the measured host phenotypes (RDA P = 0.022), with Bacteroides associated with body weight and eviscerating rate and Megamonas associated with intestinal length traits. Breed-associated microbial differences were therefore detectable at 42 days of age, well before the onset of laying, and were distributed unevenly along the intestinal tract, with the cecum the only compartment in which microbial composition, richness, and predicted functional profiles showed associations with the measured host phenotypes.

Introduction

The chicken gastrointestinal tract undergoes rapid morphological and functional maturation during the early growth period. In the first weeks after hatch, the duodenum, jejunum, ileum, and cecum develop distinct structural and physiological properties while microbial communities are progressively established and reorganized in response to diet, intestinal physiology, and host-derived factors (Rychlik, 2020; Xueyan et al., 2020). These early colonizers participate in nutrient utilization, gut barrier development, and immune maturation, and may influence subsequent growth performance and health status (Clavijo and Flórez, 2018; Yadav and Jha, 2019). Characterizing gut microbiota during this developmental window is therefore important for understanding host–microbe interactions before sexual maturity and the onset of production-related traits.

Microbial community structure along the intestinal tract reflects pronounced segment-to-segment variation in nutrient availability, transit time, oxygen tension, pH, bile acid exposure, and host secretions (Stanley et al., 2014; Glendinning et al., 2019). The small intestine is primarily involved in digestion and nutrient absorption, whereas the cecum provides an anaerobic, retention-favorable environment for microbial fermentation (Sergeant et al., 2014). Accordingly, cecal communities usually exhibit higher microbial richness and complexity than those in upper intestinal segments and are closely linked with fermentation-related functions and microbial metabolite production (Polansky et al., 2016; Zhang et al., 2023). Despite this recognized compartmentalization, many poultry gut microbiota studies still focus on fecal samples or a single intestinal compartment, which can overlook important segment-specific microbial patterns (Bajagai et al., 2024).

Host genetic background is another major determinant of microbial colonization and community assembly in chickens (Wen et al., 2019). Breeds differing in growth rate, digestive physiology, immunity, and metabolic status may provide distinct intestinal environments that shape the composition and function of gut microbiota (Wen et al., 2021; Jiang et al., 2023). White Leghorn (WL) and Silky Fowl (SF) represent two markedly different genetic backgrounds: WL is a highly selected commercial layer breed, whereas SF is an indigenous Chinese breed characterized by fibromelanosis and distinctive physiological and economic traits (Pourhamidi et al., 2024). Previous studies have linked chicken gut microbiota with growth performance, feed efficiency, intestinal health, immune traits, and production-related phenotypes (Yang et al., 2022; Yang et al., 2023; Liu et al., 2025). Our earlier work identified breed-associated cecal microbiota and metabolic differences between WL and SF hens during the laying period (Yang et al., 2026), but whether these breed-dependent microbial differences are established during early growth and how they are distributed across intestinal segments remain unclear (Shi et al., 2024; Zhang et al., 2026). WL and SF therefore provide a useful model for investigating how host genetic background is associated with segment-specific gut microbiota and functional potential during early life.

Beyond taxonomic composition, microbial interaction patterns and predicted functional potential may provide further insight into host–microbiota relationships. Co-occurrence network analysis can reveal differences in microbial community organization, including potential hub taxa and segment-specific interaction structures (Wu et al., 2024). Functional prediction based on 16S rRNA gene sequencing, although inferential, can help identify microbial pathways that may differ between groups and guide interpretation of taxonomic shifts in a functional context (Douglas et al., 2020). Integrating microbial composition, interaction networks, predicted functions, and host phenotypic traits can therefore provide a more comprehensive view of breed-associated microbiota differences during early intestinal development.

Accordingly, the present study characterized segment-specific gut microbial communities in 42-day-old WL and SF chickens by combining 16S rRNA gene sequencing, microbial co-occurrence network analysis, PICRUSt2-based functional prediction, and microbiota–phenotype association analysis. The study aimed to determine whether and how breed-associated microbial differences are distributed across the duodenum, jejunum, ileum, and cecum during early growth, and to explore potential associations among microbial taxa, predicted functions, and host phenotypic traits including intestinal development and organ indices. The findings are intended to provide a segment-resolved reference for breed-related gut microbial ecology during early chicken development, prior to the onset of laying.

Materials and methods

Animals, housing, and experimental design

Age-matched White Leghorn (WL) and Silky Fowl (SF) chicks were sourced from the conservation flocks of the Experimental Unit for Poultry Genetic Resource and Breeding, China Agricultural University. A total of 40 birds were included in the study, comprising 20 WL and 20 SF birds. All experimental birds were female. From hatch to 21 d of age, the chicks were group-housed at three birds per cage. At 21 d of age, the WL and SF birds were allocated to physically separated sections of the same poultry house and transferred to individual cages, where they were maintained until 42 d of age. The two breeds were maintained under the same feeding, management, lighting, and environmental conditions throughout the experimental period. Because the conservation populations were relatively large and managed by routine rotational mating to minimize close inbreeding, birds were selected to avoid known full-sib and half-sib relationships. No known full-sib or half-sib relationships were present among the selected individuals, although complete individual pedigree information was not available. Sampled birds were distributed across multiple cage rows and tiers to avoid confounding by cage position. Housing density during the group-housing period was three birds per cage. The experiment followed a completely randomized design, with the individual bird serving as the biological experimental unit for host phenotypic measurements and microbiota sampling. The study examined four intestinal segments, namely, the duodenum, jejunum, ileum, and cecum, with one sample collected from each segment of each bird. Thus, each breed contributed 20 biological replicates per intestinal segment, resulting in 160 intestinal content samples in total. The same sample numbers were used for alpha-diversity, beta-diversity, and genus-level differential-abundance analyses unless otherwise stated. Unlike our previous work on the same two breeds at peak lay (Yang et al., 2026), the present study targeted the early growth phase (0–42 d), during which intestinal development and microbial colonization are still being established. Day 42 was selected as a juvenile pre-laying time point, allowing breed-associated intestinal and microbial differences to be evaluated before the physiological changes associated with sexual maturation and egg production.

All procedures were approved by the Animal Experimental Ethics Committee of China Agricultural University (Approval No. AW32802202-1-2) and complied with the Laboratory Animal Welfare guidelines of the same institution (Permit No. SKLAB-2012-04-07). Birds received a mash corn–wheat–soybean meal diet formulated to meet the nutrient requirements for chickens during the 0–42 d period according to GB/T 5916-2020 and ZBB43005-86; ingredient composition and analyzed nutrient levels are given in Fig. S1. Feed and water were available ad libitum throughout. Artificial lighting was provided using warm-white LED lamps. The daily photoperiod was 24 h on d 1, gradually reduced from 23 to 18 h during d 2 to 7, maintained at 16 h during d 8 to 21 and at 18 h during d 22 to 35, and gradually increased from 18 to 23 h during d 36 to 42. Light intensity was maintained at 10–20 lux during the first 3 d and at approximately 5 lux thereafter.

Sampling and phenotypic measurements

At 42 d of age, 20 clinically healthy female birds per breed (n = 20) were randomly selected from the larger conservation populations. Birds with similar body weights were selected in both breeds to reduce body-weight-related variation at sampling. Although the sample size was limited to 20 birds per breed, the birds were randomly selected from relatively large conservation populations. Selection was performed to avoid known full-sib and half-sib relationships, thereby reducing the potential influence of family structure on the breed comparison. Host phenotypic measurements and intestinal samples were obtained from the same individual birds. Individual body weight was recorded, after which birds were anesthetized and euthanized by jugular exsanguination; all tissue and content sampling was completed within 15 min post mortem. Hatch body weight and mortality from hatch to 42 d of age were not recorded.

The abdominal cavity was opened aseptically and the heart, liver, spleen, and bursa of Fabricius were dissected free of adhering tissue and weighed individually. The eviscerated carcass was weighed for calculation of eviscerating rate. The gastrointestinal tract was then removed and the duodenum, jejunum, ileum, and cecum were delimited with hemostatic clamps at the pylorus–duodenum junction, the duodenal flexure, the yolk sac diverticulum, and the ileocecal junction, respectively. Segment length was measured with a flexible ruler along the mesenteric border without stretching, after the lumen had been gently emptied. Luminal contents from the mid-duodenum, mid-jejunum, mid-ileum, and cecum were collected aseptically into sterile cryotubes, snap-frozen in liquid nitrogen, and stored at −80°C. This yielded one content sample per segment per bird (n = 20 per breed per segment, 160 samples in total); the four segment-specific observations from the same bird were retained as separate intestinal-segment observations, while their within-bird dependence was accounted for in the mixed-effects models used for alpha-diversity analysis. For beta-diversity and genus-level differential-abundance analyses, comparisons were performed within intestinal segments, with 20 birds per breed contributing one sample each.

Phenotypic indices were derived as: organ index (%) = organ weight (g) / body weight (g) × 100; bursa of Fabricius index (%) = bursa weight (g) / body weight (g) × 100; intestinal length index (%) = segment length (cm) / body weight (g) × 100; eviscerating rate (%) = eviscerated carcass weight (g) / live body weight (g) × 100. The measured host traits represented early-growth, organ-development, carcass, and intestinal morphological phenotypes. Egg production and other laying-performance traits were not measured because the birds were sampled at 42 d of age as a pre-laying, early-growth time point, before the onset of egg production.

DNA extraction, amplicon library preparation, and sequencing

Total microbial DNA was isolated from each content sample with the OMEGA Soil DNA Kit (D5625-01, Omega Bio-tek, Norcross, GA, USA) following the manufacturer's protocol, then held at −20°C. Yield and purity were checked on a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and integrity by agarose gel electrophoresis. No samples were excluded during DNA extraction, library preparation, sequencing, or downstream quality control; all 160 intestinal content samples were retained for subsequent analyses. The sequencing data were newly generated in the present study from the intestinal content samples collected from this 42-day-old WL and SF cohort.

Library preparation followed the procedure described previously (Yang et al., 2026). Briefly, the 16S rRNA gene V4–V5 region was amplified with primers 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 907R (5′-CCGTCAATTCMTTTRAGTTT-3′) in 25 μL Q5 High-Fidelity reactions (New England Biolabs) under the following program: 98°C for 2 min; 25 cycles of 98°C for 15 s, 55°C for 30 s, 72°C for 45 s; 72°C for 5 min. Amplicons were bead-purified (VAHTS DNA Clean Beads, Vazyme, Nanjing, China), quantified with the Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen, Carlsbad, CA, USA), assigned sample-specific index sequences during library preparation, and then pooled in equimolar amounts before sequencing as 2 × 250 bp paired-end reads on an Illumina MiSeq (MiSeq Reagent Kit v3; Shanghai Personal Biotechnology, Shanghai, China). All raw sequencing data were deposited in the NCBI Sequence Read Archive under BioProject accession PRJNA1514282.

Sequence processing and community analysis

Amplicon data were processed in QIIME 2 (v2019.4; Bolyen et al., 2019). The QIIME 2 release used in this study was retained because the sequencing data were generated and analyzed approximately using a previously established workflow; all samples were processed with the same software version, plugins, and parameters to ensure analytical consistency. Reads were demultiplexed with the Demux plugin and primer sequences trimmed with Cutadapt (Martin, 2011). DADA2 (Callahan et al., 2016) was then applied for quality truncation, denoising, paired-read merging, and chimera screening, yielding amplicon sequence variants (ASVs). ASVs represented by more than one read were aligned in MAFFT (Katoh et al., 2002), and the alignment was used for phylogenetic reconstruction in FastTree2 (Price et al., 2010). Taxonomy was assigned from phylum to genus with the feature-classifier naïve Bayes classifier (Bokulich et al., 2018) trained against SILVA Release 132 (Koljalg et al., 2013).

ASV membership shared between breeds within each intestinal compartment was compared on presence–absence data and summarized as Venn diagrams (Zaura et al., 2009). Before alpha-diversity calculation, the ASV table was rarefied by random subsampling without replacement to a common sequencing depth selected for the dataset. This normalization retained all 160 samples. Within-sample diversity was quantified from the rarefied ASV table using Chao1, observed species, Shannon, Simpson, Faith's PD, Pielou's evenness, and Good's coverage, and displayed as violin plots. Good's coverage was also examined as a sequencing-depth diagnostic. Each intestinal segment was represented by one sample from each bird, resulting in 20 biological replicates per breed per segment. Between-sample dissimilarity was calculated using Bray–Curtis distances and visualized by principal coordinates analysis (PCoA). Differences in community composition among the eight breed-by-segment groups were assessed using an omnibus permutational multivariate analysis of variance (PERMANOVA) with 999 permutations. Pairwise PERMANOVA comparisons were subsequently used to assess differences among intestinal segments and between breeds within matched intestinal segments. Where applicable, P values from multiple pairwise comparisons were adjusted using the Benjamini–Hochberg procedure and are reported as q values.

Taxa distinguishing the two breeds were identified by three complementary approaches. Genus-level relative abundances were compared between breeds within each segment using the Wilcoxon rank-sum test, with nominal P < 0.05. Taxa were additionally ranked by variable importance in a Random Forest classifier (Breiman, 2001), and screened by linear discriminant analysis effect size (LEfSe; Segata et al., 2011), in which a Kruskal–Wallis test (α = 0.05) preceded linear discriminant analysis (LDA); an LDA score >2 was required for a taxon to be regarded as discriminative.

Microbial co-occurrence network analysis

Genus-level relative abundance data were used to reconstruct co-occurrence networks separately for each breed within each intestinal segment, using the R packages NetCoMi (v1.1.0) and igraph (v1.3.5). Genera undetected across all samples were discarded beforehand. Associations were inferred with SPIEC-EASI under the Meinshausen–Bühlmann neighborhood-selection model (nlambda = 15, lambda.min.ratio = 1e−2); the penalty parameter was selected via pulsar.select (rep.num = 15), and the association matrix was sparsified at a threshold of 0.3 so that only edges with absolute association values >0.3 were retained.

Network topology was described at both global and node level. Global descriptors comprised the numbers of nodes and edges, the numbers of positive and negative correlations, average degree, network density, transitivity, and modularity. Node-level descriptors comprised degree, betweenness, closeness, and eigenvector centrality. Networks were rendered with the Fruchterman–Reingold layout algorithm, node size scaled to degree centrality and node color assigned by module membership. Network construction, metric computation, and visualization were implemented with custom scripts in R (v4.5.1).

Microbial functional prediction and contribution analysis

Functional potential was inferred from the ASV table with Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt2; Douglas et al., 2020), which predicts the abundance of KEGG Orthologs (KOs) and metabolic pathways from ASV profiles and a sequenced-genome reference database (https://www.kegg.jp/). Pathway-level enrichment between breeds was assessed by generalized reporter score-based analysis (GRSA; Peng et al., 2024), which aggregates the coordinated shift of constituent KOs into a Reporter Score; a pathway was regarded as significantly enriched when |Reporter Score| > 2.

Genus-level contributions to pathway abundance were then evaluated by jointly analyzing the genus abundance, pathway abundance, and KO abundance tables. Genera and pathways detected in fewer than 10% of samples were filtered out, and all tables were converted to relative abundance to make samples comparable. Two criteria defined the target pathway set: the 20 pathways with the highest mean relative abundance, and pathways differing between breeds by Wilcoxon rank-sum test with Benjamini–Hochberg correction (FDR < 0.05 together with FC > 2 or FC < 0.5).

Two independent lines of evidence were generated for every target pathway. Spearman's rank correlation was used to quantify the association between each genus and pathway abundance, retaining pairs with FDR-adjusted P < 0.05 and |ρ| > 0.5. In parallel, a Random Forest regression model (ntree = 1000) was fitted with genus abundances as predictors and the percentage increase in mean squared error (%IncMSE) as the importance metric, and the 10 most important genera per pathway were kept as candidate microbial features. The two were combined into a single index: Comprehensive Score = (|Correlation Coefficient| × 0.6 + Standardized Importance Score × 0.4) × 100.

The 0.6/0.4 weighting slightly favors the interpretability of the direct genus–pathway association while retaining the non-linear predictive signal captured by the Random Forest model. Robustness to this choice was tested with alternative weightings of 0.5/0.5 and 0.7/0.3, and the top-ranked microbial features were highly consistent across schemes. The score is intended to prioritize putative functional associations and does not establish causality. Correlation heatmaps were drawn with pheatmap, Random Forest importance plots with ggplot2, and genus–pathway association networks with igraph.

Integrative analysis of microbiota, predicted functions, and host phenotypes

Relationships among microbial composition, predicted function, and host phenotype were examined within each intestinal segment on the GenesCloud platform (https://www.genescloud.cn). Genus-level relative abundances were correlated with host phenotypic parameters, and with PICRUSt2-predicted KO abundances, using Spearman's rank correlation; correlations with |R| ≥ 0.5 were retained.

The multivariate relationship between predicted function and host phenotype was assessed by redundancy analysis (RDA), with predicted functional profiles as response variables and host phenotypic parameters as explanatory variables, namely body weight, eviscerating rate, organ indices, bursa of Fabricius index, intestinal length, and intestinal length index. RDA was run separately for each intestinal segment, and the significance of each model was evaluated by permutation testing.

Statistical analysis of data

Phenotypic data were compiled in Microsoft Excel 2016 and are expressed as mean ± standard error of the mean (M ± SEM). Normality was checked by the Shapiro–Wilk test. Breed comparisons used Student's t-test for normally distributed variables and the Wilcoxon rank-sum test for variables departing from normality, with significance defined at P < 0.05. For host phenotypic traits, each individual bird was considered one biological replicate. For intestinal length and intestinal index, breed comparisons were performed separately for each intestinal segment, with one measurement obtained from each bird. These analyses and the corresponding plots were produced in GraphPad Prism 8.3.0 (GraphPad Software, San Diego, CA, USA).

Alpha-diversity indices were analyzed by mixed-effects models fitted by restricted maximum likelihood (REML), specifying breed, intestinal segment, and their interaction as fixed effects and individual bird as a random effect to accommodate within-bird dependence across the four segments. Main effects and the interaction term were evaluated at P < 0.05, followed by Tukey's post hoc test for multiple comparisons. Thus, the four intestinal samples collected from the same bird were treated as repeated observations across intestinal segments rather than as samples from four independent birds.

Between-sample dissimilarity was calculated using Bray–Curtis distances and visualized by PCoA. Differences in microbial community composition were assessed using PERMANOVA with 999 permutations. The overall analysis included the eight breed-by-segment groups, and pairwise comparisons were used to evaluate differences among intestinal segments and between breeds within matched intestinal segments. Genus-level relative abundances were compared between breeds separately within each intestinal segment using the Wilcoxon rank-sum test.

Results

Breed differences in organ and intestinal traits at 42 days of age

At 42 d of age, body weight did not differ between breeds (WL 0.744 ± 0.133 kg vs. SF 0.698 ± 0.080 kg, P > 0.05; Fig. 1A), and heart index was likewise comparable (0.477 ± 0.049 vs. 0.472 ± 0.045, P > 0.05). Liver index (2.69 ± 0.35% vs. 2.24 ± 0.33%), spleen index (0.223 ± 0.032% vs. 0.170 ± 0.037%), and bursa of Fabricius index (0.421 ± 0.062 vs. 0.220 ± 0.045) were all higher in WL (all P < 0.001), the bursa showing the largest relative difference between breeds. Eviscerating rate was higher in SF (51.37 ± 1.55% vs. 47.82 ± 1.81%, P < 0.001). The two breeds therefore differed in organ and carcass composition at a comparable body weight.

Fig. 1.

Physiological traits and intestinal development of White Leghorn and Silky Fowl chickens at 42 days of age. (A) Body weight, carcass trait, and organ indices of White Leghorn (WL) and Silky Fowl (SF) chickens at 42 days of age, including body weight, eviscerating rate, heart index, liver index, spleen index, and bursa of Fabricius index. Each individual bird represented one biological replicate. (B) Intestinal length and intestinal index of the duodenum, jejunum, ileum, and cecum in WL and SF chickens. For each intestinal segment, one measurement was obtained from each bird. Data are presented as mean ± SEM (n = 20 birds per breed). Statistical significance was assessed using Student’s t-test for normally distributed variables or the Wilcoxon rank-sum test for variables departing from normality. Nominal P values of < 0.05, < 0.01, and < 0.001 are indicated by one, two, and three asterisks, respectively. Abbreviations: WL, White Leghorn; SF, Silky Fowl; D, duodenum; J, jejunum; I, ileum; C, cecum.

Intestinal length was greater in WL in all four segments (all P < 0.001; Fig. 1B): duodenum 23.91 ± 2.60 vs. 20.03 ± 1.18 cm, jejunum 51.54 ± 5.90 vs. 41.98 ± 2.80 cm, ileum 50.87 ± 3.64 vs. 41.16 ± 2.32 cm, and cecum 12.21 ± 1.02 vs. 10.75 ± 0.62 cm. The same ranking held after normalization to body weight (all P < 0.001): duodenum 3.795 ± 0.413 vs. 3.180 ± 0.188, jejunum 8.180 ± 0.937 vs. 6.663 ± 0.444, ileum 8.075 ± 0.578 vs. 6.533 ± 0.368, and cecum 1.939 ± 0.161 vs. 1.706 ± 0.098, indicating that the greater intestinal size of WL was not attributable to the small, non-significant difference in body weight.

Breed separation was compartment-dependent

ASV sharing differed between the small intestine and the cecum (Fig. 2A, B). Only a limited number of ASVs were shared across all six small-intestinal breed–segment groups, whereas each group retained a substantial set of private ASVs. More segment-specific ASVs were observed in SF than in the corresponding WL segments. In contrast, the cecum showed greater ASV sharing between breeds, together with breed-specific ASV sets in both WL and SF.

Fig. 2.

Microbial diversity across intestinal segments in White Leghorn and Silky Fowl chickens at 42 days of age. (A, B) Shared and unique amplicon sequence variants (ASVs) in the small intestine and cecum, respectively. The small intestine includes the duodenum, jejunum, and ileum. (C–F) Alpha-diversity indices, including Good’s coverage, Chao1 richness, Shannon diversity, and Faith’s phylogenetic diversity (Faith_pd), across intestinal segments. (G, H) Principal coordinates analysis (PCoA) based on Bray–Curtis dissimilarity for small intestinal and cecal microbiota, respectively. For alpha-diversity analysis, breed, intestinal segment, and their interaction were included as fixed effects, and bird identity was included as a random effect to account for the dependence among the four intestinal segments collected from the same bird. Each intestinal segment was represented by one sample per bird, resulting in 20 biological replicates per breed per segment. For beta-diversity analysis, differences in community composition were assessed using PERMANOVA based on Bray–Curtis distances with 999 permutations. Different letters indicate significant pairwise differences among intestinal segments within the same breed based on adjusted q values, whereas asterisks indicate nominal breed-level P values: P < 0.05, P < 0.01, P < 0.001. P values from pairwise PERMANOVA comparisons were adjusted using the Benjamini–Hochberg procedure and are reported as q values. Abbreviations: WL, White Leghorn; SF, Silky Fowl; D, duodenum; J, jejunum; I, ileum; C, cecum; ASV, amplicon sequence variant; PCoA, principal coordinates analysis.

All four alpha diversity indices differed among intestinal segments (P < 0.001, REML mixed-effects model; Fig. 2C–F). Chao1, Shannon, and Faith's phylogenetic diversity were highest in the cecum in both breeds, whereas Good's coverage was lowest there, consistent with the cecum harbouring the richest and more incompletely sampled community. Breed main effects were significant for Good's coverage (P < 0.001) and Chao1 (P = 0.01) but not for Shannon (P = 0.29) or Faith's PD (P = 0.46), and the segment × breed interaction was significant for Good's coverage (P < 0.001), Chao1 (P < 0.001), and Faith's PD (P = 0.005) but not for Shannon (P = 0.16). Pairwise breed differences were restricted to the cecum, where Chao1 was higher in WL and Good's coverage was higher in SF (both P < 0.001); no breed difference was detected in any small intestinal segment.

Beta diversity, assessed using Bray–Curtis distances and visualized by PCoA, differed among the eight breed × segment groups in the omnibus PERMANOVA (pseudo-F = 9.526, P = 0.001; all pairwise comparisons in Table S1). Segment was the dominant source of variation: within both breeds, the small-intestinal communities differed from the cecal community, whereas the duodenum and jejunum were not distinguishable within either breed. Breed effects were smaller and segment-dependent. In matched-segment comparisons, WL and SF differed in the duodenum (q = 0.0037), jejunum (q = 0.0082), and cecum (q = 0.0325), but not in the ileum (q = 0.220). In the small intestinal ordination the two breeds overlapped broadly despite the significant duodenal and jejunal tests (Fig. 2G), whereas cecal communities separated along PCo2 (15.0%) rather than PCo1 (26.3%) (Fig. 2H). Breed-associated compositional differences were therefore detectable in three of the four segments, but were accompanied by differences in community richness only in the cecum. The dispersion and partial overlap of individual samples in the PCoA plots indicated appreciable inter-individual variability within both breeds, although breed separation was more evident in the cecum than in the small intestine.

Lactobacillus dominated the small intestine, Bacteroides and Megamonas the cecum

Composition differed markedly between the small intestine and the cecum in both breeds (Fig. 3A–D). At the phylum level, Firmicutes was the most abundant phylum in all six small intestinal groups, but its share differed between breeds: approximately 80% in WL_D and WL_J and 71.5% in WL_I, compared with 53.2% in SF_D and 57.1% in SF_J (Fig. 3A). Proteobacteria were correspondingly more abundant in the SF duodenum and jejunum (36.2%, 35.4%) than in the corresponding WL segments (18.1%, 20.0%), with Bacteroidetes present as a minor component in SF_D (9.2%). The SF ileum was the exception, showing the highest Firmicutes proportion of any group (88.1%) and low Proteobacteria (9.1%). In the cecum, Bacteroidetes and Firmicutes together accounted for more than 95% of sequences in both breeds, with Bacteroidetes predominating (63.4% in WL_C, 67.3% in SF_C) over Firmicutes (34.1% and 30.1%, respectively); the remaining phyla, including Proteobacteria, Tenericutes, Synergistetes, Deferribacteres, and Elusimicrobia, each represented less than 2% (Fig. 3B).

Fig. 3.

Relative abundance and differential analysis of gut microbiota in White Leghorn and Silky Fowl chickens at 42 days of age. (A, B) Phylum-level relative abundance of gut microbiota in the small intestine and cecum, respectively. (C, D) Genus-level relative abundance of gut microbiota in the small intestine and cecum, respectively; “Others” represents low-abundance genera not individually displayed. (E–H) Differentially abundant genera identified by Wilcoxon rank-sum test in the duodenum, jejunum, ileum, and cecum, respectively. Statistical significance was defined as nominal P < 0.05. Abbreviations: WL, White Leghorn; SF, Silky Fowl; D, duodenum; J, jejunum; I, ileum; C, cecum.

At the genus level, Lactobacillus was the most abundant identified genus in all six small intestinal groups, but its abundance was strongly segment-dependent (Fig. 3C): 65.1% in WL_D and 76.1% in WL_J, falling to 20.8% in WL_I, and 45.6% in SF_D and 41.8% in SF_J, falling to 15.7% in SF_I. The ileum of both breeds therefore had a far more even composition than the proximal segments, with Planococcaceae_Bacillus, Solibacillus, Psychrobacter, Paenibacillus, Rubrivivax, Ralstonia, Candidatus_Arthromitus, and Helicobacter each contributing and the pooled low-abundance fraction exceeding 45%. In the cecum, Bacteroides was the most abundant identified genus and was higher in SF_C (40.6%) than in WL_C (24.5%), followed at lower abundance by Faecalibacterium, Parabacteroides, Oscillospira, Phascolarctobacterium, Subdoligranulum, Blautia, Lactobacillus, Ruminococcus, and Mucispirillum (Fig. 3D). The pooled low-abundance fraction remained large in both breeds (50.4% in WL_C, 42.0% in SF_C), indicating that a substantial part of the cecal Bacteroidetes was not resolved to genus level.

Genus-level differences between breeds were segment-specific (Fig. 3E–H). In the duodenum, Pseudomonas (P = 0.008), Flavobacterium (P = 0.005), Rubrivivax (P = 0.003), and Ralstonia (P = 0.034) were more abundant in SF, as was the pooled low-abundance fraction (P = 0.036), whereas Lactobacillus did not differ significantly (P = 0.096; Fig. 3E). In the jejunum, Flavobacterium (P = 0.002), Rubrivivax (P = 0.032), and the pooled low-abundance fraction (P = 0.041) were higher in SF, whereas Lactobacillus was higher in WL (P = 0.005; Fig. 3F). No genus differed significantly in the ileum, where the smallest P value was 0.136 (Candidatus_Arthromitus; Fig. 3G). In the cecum, Bacteroides was enriched in SF (P = 0.009), whereas Megamonas (P = 0.038) and Lactobacillus (P = 0.027) were enriched in WL; Faecalibacterium (P = 0.199), Parabacteroides (P = 0.968), and Phascolarctobacterium (P = 0.989) did not differ (Fig. 3H). Thus the only compartment in which the differing genera were abundant, strictly anaerobic residents was the cecum; in the duodenum and jejunum the differing genera were aerobic or facultatively aerobic taxa present at low relative abundance.

Megamonas and Bacteroides were the principal cecal discriminants

Random Forest was applied across the eight breed–segment groups, so importance scores rank taxa that separate compartments and breeds jointly rather than breeds within a segment (Fig. 4A, B). At the phylum level Firmicutes, Bacteroidetes, and Proteobacteria ranked highest (importance 0.138, 0.115, and 0.105), followed by Synergistetes, Actinobacteria, and Tenericutes (Fig. 4A). At the genus level the highest-ranked taxa were Bacteroides (0.049), Faecalibacterium, Lactobacillus, Rubrivivax, Ralstonia, Caulobacter, Oscillospira, Acidovorax, Flavobacterium, Phascolarctobacterium, Helicobacter, Alistipes, Sporosarcina, Streptococcus, and Subdoligranulum (Fig. 4B). This ranking combined cecum-associated anaerobes (Bacteroides, Faecalibacterium, Oscillospira, Phascolarctobacterium, Alistipes, Subdoligranulum) with the aerobic and facultatively aerobic taxa that distinguished the proximal segments (Rubrivivax, Ralstonia, Caulobacter, Acidovorax, Flavobacterium), consistent with the compartmentalized pattern above. Bacteroides ranked first among all genera, whereas Megamonas did not appear among the 15 highest-ranked taxa.

Fig. 4.

Identification of discriminative gut microbial taxa by Random Forest and LEfSe analyses. (A, B) Random Forest analysis showing the relative importance of microbial taxa at the phylum and genus levels, respectively. The color gradient indicates increasing importance scores. (C–F) LEfSe cladograms highlighting taxa significantly enriched in either breed in the duodenum, jejunum, ileum, and cecum, respectively. Discriminative taxa were identified using segment-specific LDA score thresholds of 3.5 for the duodenum, 3.0 for the jejunum, and 2.0 for the ileum and cecum. Abbreviations: WL, White Leghorn; SF, Silky Fowl; D, duodenum; J, jejunum; I, ileum; C, cecum; LDA, linear discriminant analysis; LEfSe, linear discriminant analysis effect size.

LEfSe was performed separately for each segment, using LDA thresholds of 3.5 for the duodenum, 3.0 for the jejunum, and 2.0 for the ileum and cecum (Fig. 4C–F). Complementing the Wilcoxon analysis, LEfSe ranked Lactobacillus as the leading WL-associated duodenal taxon and identified additional discriminative taxa in the duodenum and jejunum (Fig. 4C, D). The ileum yielded the fewest discriminative taxa and the lowest LDA scores: Bacillus and Lactobacillus were associated with WL, whereas Pseudomonas was associated with SF, together with family- or higher-level taxa, all with LDA scores below 3.9 (Fig. 4E). None of these ileal genera differed significantly in the Wilcoxon analysis (Fig. 3G), indicating that the ileal LEfSe findings were supported only at the lower LDA threshold used for this segment.

In the cecum, Bacteroides had the highest LDA score, whereas Megamonas had the highest score among WL-associated taxa (Fig. 4F). Both LEfSe and the Wilcoxon analysis associated Bacteroides with SF and Megamonas with WL, making them the most consistently supported breed-associated cecal taxa; Bacteroides was also the highest-ranked genus in the Random Forest model. Notably, Megamonas was associated with SF in the jejunum but with WL in the cecum, further demonstrating that its breed association was segment-dependent.

SF networks were larger and less modular than WL in all segments

Co-occurrence networks were constructed separately for WL and SF in each intestinal segment; topological properties are summarized in Table 1, and full visualizations are provided in Fig. S2. SF networks consistently contained more nodes and edges than WL across all four segments, while WL networks had higher modularity in every segment (Table 1). Hub taxa were entirely non-overlapping in the duodenum, jejunum, and ileum, but two of three cecal hubs were shared between breeds.

Table 1.

Topological properties of microbial co-occurrence networks in the intestinal segments of White Leghorn and Silky Fowl.

Intestinal Segment	Breed	Nodes	Edges	Density	Transitivity	Modularity	Key hub taxa
Duodenum	WL	58	55	0.0684	0.2118	0.8915	Bacteroidaceae, Bacteroides, Ruminococcaceae
	SF	74	86	0.0592	0.1989	0.8756	Burkholderia, Hyphomicrobiaceae, Magnetospirillum, Veillonellaceae
Jejunum	WL	37	29	0.0991	0.1154	0.8928	Caulobacteraceae, Comamonadaceae
	SF	70	101	0.0708	0.2167	0.8435	Acidovorax, Brevundimonas, Burkholderia, Xanthobacteraceae
Ileum	WL	46	43	0.0860	0.1552	0.8837	Erysipelotrichaceae, Lachnospiraceae, Streptococcaceae
	SF	55	60	0.0774	0.2655	0.8679	Christensenellaceae, Clostridium, Synergistetes
Cecum	WL	55	59	0.0768	0.2213	0.8636	[Mogibacteriaceae]; cc_115; Dorea
	SF	54	67	0.0846	0.2209	0.8421	[Mogibacteriaceae]; cc_115; Dehalobacteriaceae

Microbial co-occurrence networks were constructed separately for each intestinal segment and breed using genus-level relative abundance data. Each breed-by-segment network was based on one sample per bird from 20 birds. Network inference was performed using the SPIEC-EASI algorithm with the Meinshausen–Bühlmann method implemented in the NetCoMi package. The association matrix was sparsified by thresholding, retaining only associations with absolute values greater than 0.3. Nodes represent microbial taxa retained in the co-occurrence network, and edges represent retained microbial associations. Density, transitivity, and modularity describe overall network connectivity, local clustering, and community structure, respectively. Key hub taxa were identified based on empirical quantiles of centrality measures. WL, White Leghorn; SF, Silky Fowl.

In the jejunum, the SF network was substantially larger than the WL network (Fig. 5A, B). The WL network was fragmented into isolated node pairs with a single module centred on Caulobacteraceae and Comamonadaceae, whereas SF formed one large interconnected component. Hub taxa differed completely: Caulobacteraceae and Comamonadaceae in WL; Acidovorax, Brevundimonas, Burkholderia, and Xanthobacteraceae in SF.

Fig. 5.

Breed-specific microbial co-occurrence networks in the jejunum and cecum of White Leghorn and Silky Fowl. (A, B) Microbial co-occurrence networks in the jejunum of WL and SF, respectively. (C, D) Microbial co-occurrence networks in the cecum of WL and SF, respectively. Each network was constructed from one sample per bird, with 20 birds contributing to each breed-by-segment group. Nodes represent microbial taxa, and edges represent significant co-occurrence associations. Node size represents network centrality, and node color indicates network module. Representative hub taxa and taxa identified in previous differential abundance, Random Forest, or LEfSe analyses are labeled. Abbreviations: WL, White Leghorn; SF, Silky Fowl.

In the cecum, the two networks were similar in node number, but the SF network contained more connections than the WL network (Fig. 5C, D). Unlike the small intestinal segments, the two cecal networks shared most hub taxa: [Mogibacteriaceae] and cc_115 were hubs in both breeds, and only the third differed (Dorea in WL, Dehalobacteriaceae in SF). The cecum therefore showed the most conserved hub structure despite being the compartment with the clearest breed separation in community composition (Fig. 2H, Fig. 3H).

Functional enrichment shifted from WL proximally to SF distally

PICRUSt2-predicted KEGG orthologs were analyzed using GRSA (|Reporter Score| > 2; Fig. 6A; Table S2). Differences in predicted pathway abundance reversed along the intestinal tract: in the duodenum, 23 of 37 pathways were predicted to be enriched in WL; in the jejunum, 17 of 20; in the ileum, only 6 of 17; and in the cecum, only 4 of 26.

Pathways with higher predicted abundance in WL in the duodenum and jejunum were dominated by genetic information processing, nucleotide metabolism, and central carbohydrate metabolism. Nine pathways were predicted to be enriched in WL in both segments, including gga03010 (Ribosome), gga00970 (Aminoacyl-tRNA biosynthesis), gga00230 (Purine metabolism), gga00240 (Pyrimidine metabolism), gga00010 (Glycolysis/Gluconeogenesis), gga00030 (Pentose phosphate pathway), and gga00052 (Galactose metabolism). gga03010 carried the highest Reporter Score of any pathway in the dataset, followed by gga00970. Pathways with higher predicted abundance in SF in the proximal segments were mainly related to amino acid and lipid metabolism, including gga00380 (Tryptophan metabolism), gga00330 (Arginine and proline metabolism), and gga00071 (Fatty acid degradation), with the highest SF score an order of magnitude below the leading WL score.

In the ileum and cecum, SF showed higher predicted abundance of the same core metabolic pathways that had been predicted to be enriched in WL proximally, including gga03010, gga00970, gga00010, gga00030, and gga00230. The cecum showed the strongest SF bias, with pathways predicted to be enriched in SF including gga00520 (Amino sugar and nucleotide sugar metabolism), gga00511 (Other glycan degradation), gga00531 (Glycosaminoglycan degradation), and gga00500 (Starch and sucrose metabolism). Only four cecal pathways were predicted to be enriched in WL: gga00280 (Valine, leucine and isoleucine degradation), gga00650 (Butanoate metabolism), gga00982 (Drug metabolism–cytochrome P450), and gga00380 (Tryptophan metabolism).

The reversal was reciprocal. The ribosome, aminoacyl-tRNA biosynthesis, and central carbohydrate and nucleotide metabolism module (gga03010, gga00970, gga00010, gga00030, gga00230, gga00250, gga00052) switched from WL in the duodenum and jejunum to SF in the ileum and cecum. Conversely, the branched-chain amino acid and xenobiotic degradation module (gga00280, gga00380, gga00982) switched from SF in the duodenum to WL in the cecum. The same pathways therefore distinguished the breeds in opposite directions depending on compartment.

UpSet analysis showed that the duodenum contained the largest number of differentially abundant predicted pathways (37), followed by the cecum (26), jejunum (20), and ileum (17; Fig. 6B). Pathways recurring across multiple segments were dominated by ribosome function, aminoacyl-tRNA biosynthesis, and central carbohydrate metabolism, but most reversed direction between the proximal and distal intestine. WL showed a greater number of predicted enriched pathways in the duodenum and jejunum, whereas SF displayed a greater number of predicted enriched pathways in the ileum and cecum, indicating that breed-associated differences in predicted microbial functional potential were strongly dependent on intestinal compartment.

Fig. 6.

GRSA-based functional pathway enrichment analysis of intestinal microbiota in White Leghorn and Silky Fowl chickens. (A) GRSA enrichment scores of KEGG pathways across intestinal segments. Pathways with |ReporterScore| > 2 were considered significantly enriched. Numbers displayed in white represent the ratio of significantly differential KEGG orthologs (KOs) to total detected KOs within each pathway. (B) UpSet plot showing unique and shared enriched pathways among intestinal segments. The bar plot indicates the number of pathways in each intersection, and the left horizontal bars indicate the total number of enriched pathways in each segment comparison. Abbreviations: WL, White Leghorn; SF, Silky Fowl; D, duodenum; J, jejunum; I, ileum; C, cecum; KO, KEGG ortholog; GRSA, Generalized Reporter Score-based Enrichment Analysis.

Microbial functional associations differed between the small intestine and the cecum

To identify genera potentially associated with the predicted functional differences detected by GRSA, we performed a functional contribution analysis; genus–pathway correlation heatmaps are shown in Fig. 7 and Fig. S3. In the duodenum, the genera separated into two clusters with opposite correlation profiles (Fig. 7A). Lactobacillus was positively correlated with most of the retained pathways, including ko03010 (Ribosome), ko00970 (Aminoacyl-tRNA biosynthesis), ko00030 (Pentose phosphate pathway), ko00550 (Peptidoglycan biosynthesis), ko00900 (Terpenoid backbone biosynthesis), ko00620 (Pyruvate metabolism), ko00670 (One carbon pool by folate), ko00250 (Alanine, aspartate and glutamate metabolism), ko00300 (Lysine biosynthesis), and ko00730 (Thiamine metabolism), but negatively correlated with ko00072 (Synthesis and degradation of ketone bodies) and ko00290 (Valine, leucine and isoleucine biosynthesis). Pseudomonas, Ralstonia, Flavobacterium, Caulobacter, and Rubrivivax formed a block with the inverse profile, showing uniformly strong negative correlations with the same pathways and positive correlations with ko00072 and ko00290. The remaining genera, including Faecalibacterium, Helicobacter, Bacteroides, Blautia, Oscillospira, Parabacteroides, Subdoligranulum, Psychrobacter, and Streptococcus, followed the same direction as this block but with weaker coefficients.

Fig. 7.

Microbial functional association analysis reveals segment-specific associations between key genera and predicted metabolic pathways. (A–D) Heatmaps showing integrated association scores between the top microbial genera and selected KEGG pathways in the duodenum, jejunum, ileum, and cecum, respectively. The integrated association score combines Spearman correlation strength and feature importance derived from Random Forest regression. Rows represent microbial genera, and columns represent KEGG pathways. Genera and pathways were clustered based on their association patterns. Red indicates positive associations, whereas blue indicates negative associations. The integrated association score is intended to prioritize putative functional associations rather than demonstrate direct causality. Abbreviations: D, duodenum; J, jejunum; I, ileum; C, cecum; KEGG, Kyoto Encyclopedia of Genes and Genomes.

The jejunum showed the same two-cluster structure in its most pronounced form (Fig. 7B). Lactobacillus carried strong positive correlations across the main pathway cluster (ko03430, ko04112, ko00030, ko00970, ko03010, ko00061, ko00300, ko00471, ko00473, ko00550, ko00900, ko00670, ko00620, ko00250, ko00730, ko00121), and Pseudomonas the mirror-image negative correlations, with Azospirillum, Streptococcus, Afipia, Ralstonia, Acinetobacter, Sphingomonas, Flavobacterium, Caulobacter, and Rubrivivax negative across the same cluster at intermediate strength. The direction reversed for ko00072, ko00290, and ko01051 (Biosynthesis of ansamycins), where Lactobacillus was negative and the environment-associated genera positive. Gallibacterium was near zero throughout.

Correlations in the ileum were weaker overall and no single genus dominated the positive side (Fig. 7C). Lactobacillus was positively correlated with ko00030, ko00970, ko03430 (Mismatch repair), and ko00471 (D-glutamine and D-glutamate metabolism), and negatively correlated with ko00290 and ko00660 (C5-branched dibasic acid metabolism). The strongest positive correlations in this panel belonged to a Bacillales block comprising Bacillaceae_Bacillus, Paenisporosarcina, Sporosarcina, Paenibacillus, Planococcaceae_Bacillus, and Solibacillus, and were restricted to ko02030 (Bacterial chemotaxis), ko00770 (Pantothenate and CoA biosynthesis), ko00785 (Lipoic acid metabolism), ko00061 (Fatty acid biosynthesis), and ko00072.

The cecum again showed two blocks of opposite sign, but composed of different taxa (Fig. 7D). Bacteroides, Mucispirillum, and Prevotella, and more weakly Parabacteroides, were positively correlated with ko00730 (Thiamine metabolism), ko00511 (Other glycan degradation), and ko00780 (Biotin metabolism), and negatively correlated with the larger central cluster containing ko03010, ko00550, ko00471, ko00473, ko00290, ko00670, ko00300, and ko03430. Faecalibacterium, Ruminococcus, Oscillospira, Odoribacter, Subdoligranulum, Butyricicoccus, Coprobacillus, Barnesiella, Alistipes, and Lactobacillus showed the inverse pattern, and Barnesiella was additionally positively correlated with ko00121 (Secondary bile acid biosynthesis), ko00710, and ko00770. Dorea, Coprococcus, Clostridium, Phascolarctobacterium, Sutterella, [Ruminococcus], and Blautia were near zero across all pathways. Lactobacillus was the main taxon positively correlated with the pathway set in the duodenum and jejunum, whereas positive correlations were distributed among Bacillales genera in the ileum and among two mutually opposed groups of anaerobes in the cecum.

Microbiota–function–phenotype associations were strongest in the cecum

Genus-level taxa, PICRUSt2-predicted KOs, and host phenotypic traits were analyzed by Spearman correlation and redundancy analysis across the four intestinal segments. The jejunum and cecum are shown in Fig. 8; duodenal and ileal results are in Fig. S4.

In the jejunum, Lactobacillus was positively correlated with liver index, spleen index, eviscerating rate, and all four intestinal length and index traits except jejunal index, whereas Rubrivivax, Acinetobacter, Ralstonia, and Flavobacterium were negatively correlated with intestinal length and index traits (Fig. 8A). Genus–KO correlations were extensive: 10 genera and 29 KOs were retained in the analysis (Fig. 8B), with Lactobacillus negatively correlated with most KOs and the environment-associated genera (Pseudomonas, Rubrivivax, Acinetobacter, Ralstonia, Flavobacterium, Subdoligranulum, Blautia, Sphingomonas) positively correlated.

Fig. 8.

Integrative association analysis of jejunal and cecal microbiota, predicted KOs, and host phenotypic traits. (A) Spearman correlation heatmap showing associations between jejunal genus-level microbial abundance and host physiological or intestinal traits. (B) Spearman correlation heatmap showing associations between jejunal genus-level microbial abundance and PICRUSt2-predicted KEGG orthologs (KOs). (C) Spearman correlation heatmap showing associations between cecal genus-level microbial abundance and host physiological or intestinal traits. (D) Spearman correlation heatmap showing associations between cecal genus-level microbial abundance and PICRUSt2-predicted KOs. (E) Redundancy analysis (RDA) of cecal KO profiles constrained by host phenotypic traits. In the correlation heatmaps, red and blue indicate positive and negative Spearman correlations, respectively, and asterisks indicate statistically significant correlations. In the RDA plot, points represent individual samples, arrows indicate host phenotypic variables or KOs, and the percentages on the axes indicate the proportion of constrained variation explained by each RDA axis. Complete association results for the duodenum and ileum are provided in the supplementary figures. Abbreviations: WL, White Leghorn; SF, Silky Fowl; KO, KEGG ortholog; RDA, redundancy analysis.

In the duodenum and ileum (Fig. S4), Lactobacillus and Gallibacterium were positively correlated with intestinal development traits. Genus–KO analysis identified fewer significant associations in these segments than in the jejunum or cecum.

The cecum displayed the most extensive association structure of any segment. Bacteroides, Oscillospira, Phascolarctobacterium, and Blautia were positively correlated with body weight, and Megamonas, Lactobacillus, and Clostridium with intestinal length and index traits (Fig. 8C). Alistipes and Dorea were positively correlated with bursa of Fabricius index. Cecal genus–KO analysis retained 10 genera and 27 KOs, the largest set of any segment (Fig. 8D). Bacteroides, Phascolarctobacterium, and Oscillospira were positively correlated with multiple KOs, including K00012, K00020, K00128, K00252, K00253, K00261, K00452, K00764, K00799, K00850, K00901, K01187, K01206, K01432, K01443, K01907, K01965, K01968, K01969, K03111, K03392, K03426, and K05349; Megamonas and Lactobacillus were positively correlated with K01965, K01969, K03392, K03426, and K05349; and Alistipes was positively correlated with K00012, K00850, K01187, K01206, K01432, K01443, K10747, and K12373.

RDA showed a significant association between the predicted KO profile and host traits only in the cecum (permutation test, P = 0.022; Fig. 8E, Table S3). The RDA plots for all four intestinal segments are provided in Fig. S5. RDA1 explained 88.19% of the constrained variation and separated the breeds, with WL samples distributed toward positive RDA1 values and SF samples toward negative values. Body weight, heart index, liver index, and organ indices loaded toward the negative (SF) side of RDA1, whereas duodenum length and ileum length loaded toward the positive (WL) side; jejunum length loaded strongly negative along RDA2. KO loadings were distributed in both directions along RDA1, with K09128, K00252, K01965, K01969, K01968, K00253, K01965, K01187, K01206, K03392, K03426, K05349, K01907, K00261, and K01432 toward the positive side and K05349, K00764, and K431 toward the negative side. The duodenal and ileal models were not significant. Pairwise genus–phenotype and genus–KO associations were therefore detected in all four segments, but the cecum was the only compartment in which the KO profile as a whole was significantly associated with host phenotype.

Discussion

The present study shows that at 42 days of age, well before sexual maturity, White Leghorn and Silky Fowl chickens already exhibit distinct physiological, intestinal, and gut microbial phenotypes (Oakley et al., 2014; Mohd Shaufi et al., 2015). Although body weight did not differ significantly, WL showed higher liver, spleen, and bursa of Fabricius indices, as well as consistently greater intestinal length and intestinal indices across all four segments. These findings indicate that breed-specific developmental trajectories are established early in life (Han et al., 2015). Notably, microbial community structure was strongly compartmentalized: the cecum, not the small intestine, displayed the highest alpha diversity, the clearest breed-level separation in beta diversity, and the most robust associations among taxa, predicted functions, and host phenotypes (Oakley et al., 2014; Glendinning et al., 2019). The cecum therefore emerges as the compartment in which breed background is most closely associated with gut microbial ecology during juvenile development (Yang et al., 2022).

The greater intestinal length and indices in WL suggest a breed-specific bias toward enhanced absorptive capacity, possibly reflecting divergent resource allocation strategies. Since body weight was only marginally higher in WL and not statistically significant, the proportional increase in intestinal dimensions implies that WL prioritizes gut development relative to overall body mass. Longer intestines may extend nutrient exposure time and increase surface area, thereby influencing microbial colonization dynamics (Choi et al., 2015). The elevated liver index in WL may reflect higher metabolic activity, while the larger spleen and bursa of Fabricius indices point to breed-related differences in immune organ development. Because the bursa approaches its maximum relative size in the weeks preceding sexual maturity and regresses thereafter (Ciriaco et al., 2003), the 42-day sampling point captures this organ near its developmental peak, making the observed breed difference in bursal index unlikely to reflect differences in the timing of involution (Pourhamidi et al., 2024). In contrast, SF exhibited a significantly higher eviscerating rate, which may reflect a greater relative allocation of body mass to the carcass and a lower allocation to visceral organs, rather than more efficient carcass deposition.

Intestinal segment remained the dominant factor structuring microbial composition, with the small intestine and cecum forming two distinct ecological niches, consistent with previous studies showing pronounced spatial differences in chicken intestinal microbiota (Mohd Shaufi et al., 2015; Glendinning et al., 2019). Despite the segment-dependent breed-associated differences, the partial overlap among individual samples indicated substantial within-breed variability, suggesting that breed was not the sole determinant of microbial community composition. The small intestinal microbiota was uniformly dominated by Lactobacillus across both breeds, consistent with its adaptation to carbohydrate-rich, fast-transit environments and with previous evidence highlighting its prominent role in chicken gut microbial ecology and feed efficiency (Stanley et al., 2014; Yan et al., 2017). In contrast, the cecum harbored a phylogenetically richer community, with Firmicutes and Bacteroidetes as dominant phyla and a diverse assemblage of anaerobic genera, including Bacteroides, Megamonas, Prevotella, Parabacteroides, Faecalibacterium, and Alistipes, reflecting its role as the primary site of microbial fermentation (Oakley and Kogut, 2016; Fan et al., 2023). High Bacteroides abundance in young chicks has been linked with increased short-chain fatty acid production and reduced markers of gut inflammation (Zhang et al., 2023), consistent with the enrichment of Bacteroides-related taxa observed here. Chao1, Shannon, and Faith's PD were all highest in the cecum, confirming its status as a permissive niche for microbial diversification (Polansky et al., 2016). Good's coverage was correspondingly lowest in the cecum, indicating that even at the sequencing depth applied here the cecal community was the least completely sampled of the four segments; cecal richness estimates should therefore be read as lower bounds.

Breed effects were present along most of the tract, but the depth to which they extended differed between compartments. The segment × breed interaction was significant for Chao1 (P < 0.001), Good's coverage (P < 0.001), and Faith's PD (P = 0.005), and pairwise breed contrasts in richness were confined to the cecum, where Chao1 was higher and Good's coverage lower in WL (both P < 0.001); no small intestinal segment differed between breeds in any richness metric. Shannon diversity, by contrast, showed neither a breed main effect (P = 0.29) nor a segment × breed interaction (P = 0.16). This dissociation suggests that the cecum-specific breed difference in richness may primarily reflect variation in low-abundance taxa rather than major differences among dominant community members. The greater complexity of the cecal community may make subtle breed-associated variation more detectable, although this interpretation remains tentative and requires validation. The larger private ASV pool of the WL cecum (15,463 vs. 10,156 in SF) is consistent with this interpretation, although unique-ASV counts are sensitive to prevalence filtering and sequencing depth and should be interpreted alongside the formal richness estimates calculated after sequencing-depth normalization. Community composition was the more widely responsive readout: matched-segment comparisons separated the breeds in the duodenum (q = 0.0037), jejunum (q = 0.0082), and cecum (q = 0.0325), but not in the ileum (q = 0.220). Breed-associated differences were therefore not confined to the cecum at the compositional level; what set the cecum apart was that they extended beyond composition to community richness and to the only significant association with host phenotype, whereas the proximal small intestinal differences were compositional only. Because pseudo-F scales inversely with within-group dispersion, its magnitude is not directly comparable across segments that differ in community variability, and the higher duodenal and jejunal values therefore do not imply that breed explains more variation there than in the cecum. That the cecal separation was expressed along PCo2 (15% of variance) rather than PCo1 (26.3%) further indicates that breed accounts for a real but secondary share of cecal community variation at this age. The pattern in the proximal small intestine, compositional differences without richness differences, is consistent with the strong ecological filtering of the small intestine, where rapid digesta transit, bile acids, oxygen tension, and digestive enzymes constrain microbial assembly (Tellez et al., 2006; Wen et al., 2021), since such filtering restricts assembly to a narrow, Lactobacillus-dominated set of taxa whose relative abundances can shift between breeds without a change in the number of taxa supported. The ileum was the only segment in which no breed effect was detected by any measure; whether this reflects stronger convergent selection in this segment or limited power at the present sample size cannot be resolved from a single time point. That breed effects penetrated furthest in the cecum is consistent with the principle that communities in stable, anaerobic environments are more responsive to host-derived signals (Jiang et al., 2023; Yan et al., 2023).

At the taxonomic level, breed-associated differences were segment-specific. In the duodenum, Pseudomonas, Flavobacterium, Ralstonia, and Rubrivivax were significantly enriched in SF; in the jejunum, Flavobacterium and Rubrivivax were also SF-enriched, whereas Lactobacillus was WL-enriched. Most notably, in the cecum, Bacteroides was significantly enriched in SF, while Megamonas and Lactobacillus were enriched in WL (all P < 0.05) (Yan et al., 2017; Fan et al., 2023). This pattern, SF favoring Bacteroides, WL favoring Megamonas and Lactobacillus, was supported across differential abundance testing, LEfSe, and functional contribution analyses, and underscores that breed-associated microbial signatures are not random but may reflect coherent ecological strategies (Zhang et al., 2023).

Co-occurrence network analysis further indicated that breed-associated divergence extends beyond abundance to community organization. Network properties, including node number, edge number, density, modularity, and hub taxa, are commonly used to infer potential microbial associations and community structure (Faust et al., 2012; Gao et al., 2022). In the jejunum, the SF network was larger (70 nodes, 101 edges) than WL's (37 nodes, 29 edges), while WL exhibited higher modularity (0.893 vs. 0.844), suggesting a more compartmentalized topology. In the cecum, although node numbers were similar (55 vs. 54), SF had more edges (67 vs. 59) and higher density (0.085 vs. 0.077), while WL showed marginally higher modularity (0.864 vs. 0.842). Hub taxa also differed: Bacteroides was a high-centrality node in the SF cecal network, whereas Megamonas and Alistipes showed elevated centrality in the WL network, mirroring the differential abundance results and suggesting that discriminative taxa also occupy central positions within their respective networks (Deng et al., 2012). Because a single network was reconstructed per breed and segment, these topological comparisons are descriptive and were not subjected to statistical testing; the differences in modularity in particular were small and should be regarded as tentative.

Functional profiling integrated taxonomic patterns into a predicted functional framework. PICRUSt2 prediction combined with GRSA revealed a striking spatial inversion: WL-associated predicted pathways showed higher enrichment in the duodenum and jejunum, primarily involving ribosome, aminoacyl-tRNA biosynthesis, nucleotide metabolism, and glycolysis—pathways related to processes such as microbial growth and central carbon metabolism, although these functions were not directly measured (Douglas et al., 2020). Conversely, SF-associated predicted pathways showed stronger enrichment in the ileum and especially the cecum, with enrichment in glycan degradation, starch/sucrose metabolism, amino sugar metabolism, and lysosome-related pathways (Flint et al., 2008; Ravcheev et al., 2013). This pattern is consistent with differences in the predicted functional potential of the two breed-associated microbiota, with WL showing greater enrichment of these pathways in the upper gut and SF showing greater enrichment of pathways related to complex-substrate processing in the hindgut(White et al., 2023). These results should not be interpreted as evidence that either microbiota directly performs enhanced energy harvesting or substrate processing, and direct metagenomic, metatranscriptomic, metabolomic, or culture-based validation is needed.

Genus-level contribution analysis identified taxa whose abundances were correlated with the predicted functional differences. In the small intestine, Lactobacillus showed broad positive correlations with core metabolic pathways, suggesting that its abundance was associated with the corresponding predicted functional profiles, rather than demonstrating that it acted as a functional keystone of the upper gut (Lee et al., 2010; Zou et al., 2018). In the cecum, however, no single taxon dominated: Bacteroides, Prevotella, and Parabacteroides were positively correlated with glycan-related pathways; Faecalibacterium, Ruminococcus, Oscillospira, Subdoligranulum, Butyricicoccus, Barnesiella, Alistipes, and Lactobacillus showed correlations with broader nucleotide, amino acid, and energy metabolism pathways (Sonnenburg et al., 2010; Schmidt et al., 2018). This distributed correlation structure is consistent with the complexity of anaerobic fermentation consortia and may partly explain the greater number of cecal taxon–function associations (Reichardt et al., 2018).

Only the cecal predicted KO profile showed a significant multivariate association with host phenotypes (P = 0.022), whereas the corresponding associations were not significant in the duodenum, jejunum, or ileum. At the genus level, the ordination suggested positive associations of Bacteroides with body weight and eviscerating rate and of Megamonas with intestinal length and intestinal index traits. Parabacteroides, Phascolarctobacterium, and Prevotella also showed positive links to growth- and development-related traits. These associations are biologically plausible: Bacteroides and Prevotella are proficient degraders of host- and diet-derived glycans (Glowacki et al., 2020; Crouch et al., 2022); Megamonas and Phascolarctobacterium belong to the Negativicutes, in which propionate is generated predominantly through the succinate pathway (Reichardt et al., 2014; Louis and Flint, 2017), and propionate in turn serves as both an energy substrate for the intestinal epithelium and a signalling molecule affecting epithelial turnover; and SCFAs can stimulate intestinal epithelial proliferation and modulate immune function (Koh et al., 2016). However, metabolite concentrations and pathway activities were not measured in the present study, so these mechanisms remain hypothetical. Moreover, the genus-level associations identified within the significant cecal RDA should be interpreted as exploratory correlations rather than validated functional predictions or evidence of causal effects. Together, these results identify the cecum as the compartment showing the strongest associations among microbial taxa, predicted functional potential, and host growth and intestinal development.

Comparison with our earlier study of the same two breeds at peak lay provides a developmental perspective on these patterns (Yang et al., 2026). Four features are informative. First, the small intestinal taxonomic signature was stable across stages: Lactobacillus was enriched in the WL jejunum at both 42 days and peak lay, consistent with the strong ecological filtering that constrains upper-gut assembly. At the whole-community level, however, the proximal small intestinal breed effect appears to attenuate with age: duodenal and jejunal communities separated between breeds at 42 days (q < 0.01), whereas no small intestinal segment separated between breeds at peak lay. Second, the cecal richness asymmetry was preserved in both location and direction: at both ages, breed differences in Chao1 and Good's coverage were restricted to the cecum, with the WL cecum showing the higher richness and the lower coverage. Third, the dominant cecal taxonomic signature was reversed. At peak lay, Bacteroides was enriched in the WL cecum and Faecalibacterium in SF, whereas at 42 days Bacteroides was enriched in SF, and Megamonas and Lactobacillus in WL. Fourth, and in contrast to this taxonomic inversion, the organizational contrast between the two cecal networks was preserved in direction: at both ages the WL cecal network was the more modular and the SF network the more densely connected, even though the taxa occupying hub positions differed entirely. Taken together, these comparisons suggest that what carries across developmental stages is not the identity of the discriminative taxa, but the compartment in which breed effects are expressed most fully, the direction of the cecal richness asymmetry, and the broad organizational contrast between the two communities. The taxonomic reversal is compatible with the substantial changes occurring between these stages, including the transition from grower to layer diets, the onset of ovarian steroid secretion, and the associated shift in calcium and energy demand, all of which alter cecal substrate supply and could favour different primary degraders (Zhou et al., 2021; Kuroda et al., 2026). These cross-stage comparisons are drawn from two independent cohorts analysed in separate studies, were not tested jointly, and are therefore descriptive. Resolving when and how this reversal occurs will require sampling the same birds across the pullet-to-layer transition.

This study has several limitations. First, functional profiles were inferred from 16S rRNA gene data rather than directly measured, a recognized limitation of marker-gene-based approaches (Langille et al., 2013; Matchado et al., 2024). The reported pathway differences therefore represent predicted functional potential rather than demonstrated metabolic activity and require direct functional validation. Second, association analyses identify statistical relationships but do not establish causality, as gut microbial patterns can be confounded by unmeasured host, environmental, and technical factors (Vujkovic-Cvijin et al., 2020). In addition, the cecal RDA model included multiple explanatory variables relative to the available sample size, and overfitting cannot be excluded. Although RDA was evaluated across all four intestinal segments, only the cecal model reached statistical significance, and the corresponding associations in the other three segments were not significant. The RDA findings should therefore be considered exploratory and require validation in larger, independent cohorts. Third, the single-timepoint design limits inference about the temporal trajectory of microbial establishment. Although birds with similar body weights were selected at 42 d of age, hatch body weight and mortality between hatch and 42 d of age were not recorded. Therefore, early growth trajectories, survival-related differences, and the contribution of initial body-weight differences to later phenotypes could not be fully evaluated. Longitudinal sampling from hatch through sexual maturity would help clarify when breed-associated divergence first emerges and how stable it remains (Liu et al., 2023). Fourth, several genera enriched in the SF duodenum and jejunum, including Flavobacterium, Ralstonia, Rubrivivax, and Pseudomonas, are widely distributed in water and environmental substrates and are recurrently reported among reagent- and laboratory-derived contaminant taxa in low-biomass samples (Salter et al., 2014). Because small intestinal contents carry substantially less microbial DNA than cecal contents, these signals should be interpreted with caution, and their biological relevance requires confirmation with quantitative or culture-based approaches. Fifth, co-occurrence networks were reconstructed from a single network per breed and segment, so topological differences could not be statistically evaluated and are reported as descriptive comparisons only. Despite these limitations, the integration of segment-resolved microbiota profiling, network analysis, predicted functions, and phenotype association provides a useful framework for understanding breed-related differences in early gut development.

Conclusion

This study shows that breed-associated gut microbiota differences between White Leghorn and Silky Fowl chickens were detectable at 42 days of age and varied substantially among intestinal compartments. The cecum exhibited the highest microbial diversity, the clearest breed-level separation, and the only significant association between predicted functional profiles and host phenotypes, whereas the small intestine showed conserved Lactobacillus dominance across breeds. These findings identify the cecum as a key site of breed-specific microbial organization at this pre-laying age and provide a segment-resolved framework for future longitudinal studies linking gut microbiota with host growth and intestinal maturation.

Author contributions

Xue Yang conceived and designed the study, conducted the investigation, developed the methodology, performed the formal analysis and data curation, validated the results, developed the software, prepared the visualizations, supervised the study, coordinated the project, and drafted the manuscript. Yurong Tai contributed substantially to the experimental investigation, methodology development, formal analysis, and data curation, and critically revised the manuscript. Xin Wu contributed substantially to the experimental investigation and methodology development, and critically revised the manuscript. Deping Han contributed substantially to the experimental investigation, provided supervision and critical resources, and critically revised the manuscript. Ganxian Cai contributed to the acquisition of experimental resources and critically revised the manuscript. Zihan Xu contributed to data acquisition and curation and critically revised the manuscript. Jiaqi Hao contributed to data organization and curation and critically revised the manuscript. Junying Li contributed to the acquisition of experimental resources and critically revised the manuscript. Jiankui Wang contributed to the experimental investigation, project administration, and acquisition of resources and funding, and critically revised the manuscript. Xuemei Deng contributed to the conception and design of the study, formal analysis, project administration, acquisition of resources and funding, and critical revision of the manuscript. All authors made substantial contributions to the work, critically reviewed and edited the manuscript for important intellectual content, approved the exact final version of the manuscript submitted for publication, and agreed to be personally and publicly accountable for all aspects of the work, including ensuring that any questions regarding its accuracy or integrity are appropriately investigated and resolved.

Ethic approval

This animal study was reviewed and approved by the Laboratory Animal Welfare Experiment License from China Agricultural University was obtained (permit number: SKLAB- 2012-04-07). All experiments were approved by the Committee on the Animal Experimental Ethical Inspection of China Agricultural University (issue number: AW32802202-1-2).

Data availability

The 16S rRNA gene sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1514282. The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.

Declaration of generative AI and AI-assisted technologies in the writing process

The authors did not use any artificial intelligence assisted technologies in the writing process.

Disclosures

The authors declare no personal or financial conflicts of interest.

Funding

This research was supported by the National Key Research and Development Program of China (2022YFF1000203), National Natural Science Foundation of China (32472896), Hainan Seed Industry Laboratory (B23CJ0521, B21HJ0506) and Independent Research Project of State Key Laboratory of Animal Biotech Breeding (2023SKLAB1-7). The authors also acknowledge the support from the 2115 Talent Development Program of China Agricultural University.