Bacteriocin gene-mediated ecological adaptation of Bifidobacterium breve in the adult human gut

Summary

The ecological persistence of Bifidobacterium breve across life stages reflects adaptive strategies beyond the classical infant- versus adult-type dichotomy, historically attributed to differential nutrient utilization. Here, comparative genomics revealed no major differences in shared carbohydrate-related genes or accessory genome content between infant- and adult-derived strains. Instead, a distinct type III lanthipeptide bacteriocin cluster, lanKC, was specifically detected in adult-derived isolates. Functional assays combining gene knockout, in vitro co-cultivation, and human intervention demonstrated that lanKC enhances strain-level competitive fitness and promotes community stability. Phylogenetic and metagenomic analyses of 5,475 lanKC homologs and 6,122 infant gut metagenomes further suggested a possible early-life acquisition via intra-genus horizontal gene transfer. These findings uncover a previously unrecognized genetic basis underlying B. breve adaptation to the gut environment and support a multi-factorial model in which metabolic flexibility and interference competition jointly sustain bifidobacterial persistence and host-microbe symbiosis throughout life.

Graphical abstract

Highlights

• •B. breve colonization in adults is not solely determined by nutritional factors
• •The accessory genome of specific adult-derived B. breve harbors the lanKC gene
• •LanKC drives targeted inhibition and reinforces community structural stability
• •LanKC was likely acquired via intra-genus horizontal transfer early in life

A type III lanthipeptide bacteriocin gene cluster, lanKC, identified in adult-derived Bifidobacterium breve, enhances competitive fitness and stabilizes gut microbial communities, likely contributing to its persistent colonization since early life. This discovery reveals a complementary mechanism to the nutrient utilization-driven model of bifidobacterial ecological adaptation.

Introduction

Bifidobacterium represents one of the most widely used probiotic genera, and naturally inhabits both the infant and adult gastrointestinal tracts.¹ More than ten Bifidobacterium species and subspecies coexist in the human gut, exhibiting distinct prevalence patterns across life stages.² Traditionally, bifidobacterial taxonomy and ecological classification have been defined according to their host origin, leading to the well-established distinction between “infant-type” and “adult-type” species.³ The former group includes Bifidobacterium longum (subsp. longum and subsp. infantis), Bifidobacterium bifidum, and Bifidobacterium breve, whereas the latter primarily comprises Bifidobacterium adolescentis, Bifidobacterium pseudocatenulatum, and Bifidobacterium catenulatum. This framework has been largely shaped by the presumed link between host age and carbohydrate metabolism: infant-type species are generally characterized by their ability to metabolize human milk oligosaccharides (HMOs), whereas adult-type species preferentially utilize plant- or diet-derived polysaccharides.^4,5 Previous studies have also proposed other determinants of bifidobacterial persistence—such as cross-feeding, strain-specific adhesion, and host immune interactions^6,7,8,9—but these hypotheses remain largely inferential and lack direct experimental validation. Consequently, carbohydrate utilization capacity is still considered the primary driver of bifidobacterial niche differentiation.

However, Bifidobacterium breve does not fully conform to this metabolic-ecological paradigm. Despite its traditional designation as an infant-type species, B. breve consistently occurs in adults and even in elderly populations. For instance, a study of healthy Asian adults identified B. breve as the second most abundant Bifidobacterium species, constituting 26.3% of the intra-genus composition—surpassing B. catenulatum (17.0%) and B. bifidum (11.3%).¹⁰ Similar trends were observed in other cohorts, with B. breve accounting for 22.1% of Bifidobacterium in young adults (20–28 years) and up to 67.6% in centenarians (>90 years).¹¹ From a metabolic perspective, B. breve paradoxically exhibits limited HMO utilization, characterized by selective isomer preference, poor degradation of fucosylated HMOs, and dependence on cross-feeding from fucosidase-producing commensals such as Ruminococcus gnavus.^12,13 Moreover, substantial intraspecies heterogeneity in carbohydrate metabolism has been observed even among canonical HMO utilizers (e.g., B. infantis and B. longum), while B. breve appears to harbor distinct gene clusters adapted to specific host lifestyles.¹⁴ These observations suggest that B. breve has developed additional adaptive strategies beyond nutrient-based competition, potentially contributing to its sustained colonization across host life stages.

Our previous genomic analyses based on the Early-Life Gut Genome (ELGG) catalog revealed a decline in Bifidobacterium relative abundance from birth to three years of age, accompanied by increased strain-level heterogeneity within individual hosts.¹⁵ Moreover, the early-life Bifidobacterium genomes displayed greater conservation, whereas later developmental stages were characterized by the acquisition of more strain-specific genes.¹⁵ These findings raise an important question: do strain-specific genetic determinants, beyond broad metabolic traits, underpin the distinctive ecological persistence of B. breve throughout the human lifespan? Yet, the genetic mechanisms governing this adaptability remain largely unexplored in Bifidobacterium.

In this study, we conducted a comprehensive genomic and functional investigation of B. breve strains isolated from Chinese cohorts and publicly available datasets across different host ages. We identified a bacteriocin biosynthesis gene, lanKC, exclusively detected in adult-derived isolates. Targeted gene knockout, combined with in vitro and in vivo (human) microbiota assays, demonstrated that lanKC contributes to the competitive fitness of its host strain within the gut microbial ecosystem. Phylogenetic and genetic tracing analyses further indicated that lanKC was likely acquired via intra-genus horizontal gene transfer during the evolutionary diversification of Bifidobacterium. Collectively, these findings reveal a previously underappreciated interference-competition mechanism that complements nutrient-based adaptation and contributes to the ecological diversification of B. breve. Rather than redefining bifidobacterial ecology, this work refines the existing framework by proposing a multi-factorial model in which both metabolic adaptability and interference competition jointly support persistence across different life stages of the host.

Results

B. breve strains from infants and adults exhibit no distinct genetic variations in carbohydrate utilization genes

We isolated 100 B. breve strains, categorizing them by the host age, i.e., adult-derived (n = 9, subjects >18 years) and infant-derived (n = 91, infants aged 0–3 years). Additionally, 57 infant-derived and 8 adult-derived B. breve genomes with high-quality assemblies were obtained from NCBI according to host metadata (Figure 1A; Table S1). Despite the distinct gut environments implied by the host age and the potential for divergent ecological adaptation strategies, these two groups exhibited a high genomic similarity, ranging from 97.0% to 100% of average nucleotide identity (ANI) (Figure S1A). Moreover, a phylogenetic tree based on 930 core genes revealed that adult-derived strains were interspersed across multiple evolutionary branches, without forming a discrete phylogenetic cluster (Figure 1B).

Figure 1.

Lack of distinct genomic signatures between adult- and infant-derived B. breve strains

(A) Schematic workflow of comparative genomic analysis between adult- (n = 17) and infant-derived (n = 148) B. breve strains.

(B) Maximum-likelihood phylogenetic tree based on 930 core genes. Branch lengths represent evolutionary distances; red circles in the outer ring indicate adult-derived strains.

(C) Distribution of carbohydrate-active enzyme (CAZy) gene families. The density plot shows the overall distribution across the two groups, with dashed lines marking group medians. The heatmap displays the normalized abundance of 49 CAZy families that were present in ≥50% of strains in each group. The abundance of each family was normalized across all strains using the PercentScaler method. Gene family names are indicated on the outer ring.

(D) Phylogenetic and evolutionary distance analysis of representative carbohydrate utilization genes. The left panel shows a phylogenetic tree of five selected genes across strains, with red marking adult-derived strains. The right panel presents boxplots comparing average evolutionary distances: the adult and control groups are based on 1,500 random samplings of 17 and 165 CAZy genes. Statistical significance was assessed by the permutation test; p < 0.05 was considered significant.

Given the critical role of carbohydrate utilization in shaping the gut microbial niches, we analyzed the distribution and evolutionary patterns of carbohydrate-active enzyme (CAZyme) genes within B. breve strains. Overall, the CAZyme gene profiles were highly similar between adult- and infant-derived strains (Figure 1C). Notably, genes implicated in the metabolism of HMOs, raffinose, and starch, such as GH112 lnbP,¹⁶ GH20 nahA,¹⁶ GH2 lacZ,¹⁷ GH42 bgaB,¹⁸ and GH36 rafA,¹⁹ have been proposed to influence B. breve colonization in the adult gut. However, all 165 strains in our collection harbored these genes, showing no differential distribution linked to their origin (Table S2).

Then, we employed a bootstrap resampling strategy to construct the intra-group evolutionary distances of these genes and evaluate whether they exhibited phylogenetic clustering by comparing all strains with the subset of adult-derived strains. We found that the evolutionary distances of these genes in the adult group were significantly higher than those in the reference set of all strains (p ≤ 0.001) (Figure 1D), suggesting that these carbohydrate utilization genes in adult-derived B. breve have undergone greater evolutionary divergence but do not exhibit a clear clustering tendency.

Identification of an accessory bacteriocin synthesis gene in adult-derived B. breve strains

To explore potential ecological adaptation mechanisms in adult-derived B. breve strains, we constructed an accessory genome of 165 B. breve comprising 9,445 genes. Within this dataset, 46 accessory genes were consistently shared across all 17 adult-derived strains (Table S3). However, these genes were present in over 75% of infant-derived strains, and their detection frequencies did not differ significantly between adult- and infant-derived groups (adj-p > 0.05, Fisher’s exact test) (Figure 2A; Table S4), suggesting that adaptation to the adult gut environment may not depend on universally conserved accessory genes but rather on strain-specific genetic determinants.

Figure 2.

B. breve strain CCFM1025 harbors a conserved accessory gene cluster without impacting basic growth or metabolism

(A) Presence-absence heatmap of accessory genes shared among adult-derived strains and their distribution across infant-derived strains. Each column represents a gene, each row represents a strain. Differential gene distributions were analyzed using Fisher’s exact test, with p values corrected for the false discovery rate using the Benjamini-Hochberg method.

(B) Number of unique genes identified in each adult-derived strain. Bar labels indicate strain-specific gene counts.

(C) Distribution of unique genes in CCFM1025, highlighting the type III lanthipeptide biosynthetic gene cluster. The upper panel shows the genomic organization of unique gene regions; the lower panel depicts the gene cluster structure, with arrows indicating gene orientation.

(D) Workflow for the construction and validation of the lanKC insertion mutant. Gel image: marker 1, 100-bp ladder; Marker 2, 1-kb ladder. Bands A–D correspond to PCR products of 1373, 4932, 1513, and 754 bp, respectively.

(E) Growth curves of CCFM1025 wild-type and lanKC mutant strains in vitro. OD₆₀₀ was monitored hourly from 0 to 18 h. Data were analyzed by two-way ANOVA with Bonferroni’s post-hoc test; ns indicates p > 0.05.

Considering the strain heterogeneity and the need for subsequent experimental validation, we conducted a pan-genome analysis based on the 100 B. breve isolates obtained in this study. Subsequent comparative analysis identified 110 genes that were unique to adult-derived strains, absent in any infant-derived counterparts. Notably, strain CCFM1025 harbored the greatest number of these unique genes (n = 34), representing 30.9% of the total (Figure 2B). Among these, 11 co-linearly arranged genes formed a complete biosynthetic gene cluster encoding a type III lanthipeptide bacteriocin. This cluster included a critical modifying enzyme, a class III lanthionine synthetase designated lanKC-1025 (hereafter lanKC) (Figure 2C). In contrast, while the other eight adult-derived strains possessed several unique genes, none contained a fully assembled and functionally annotated modular cluster (Table S5).

To elucidate the physiological role of lanKC in the host strain, we generated a lanKC insertion mutant (ΔlanKC) (Figure 2D). With standard culture conditions, ΔlanKC exhibited no significant differences compared to the wild-type strain in the growth kinetics, acid production, or maximum viable cell counts (Figures 2E, S2A, and S2B), indicating that lanKC is dispensable for basic growth. Given its localization within the accessory genome, lanKC likely represents an acquired gene that confers enhanced ecological fitness rather than essential metabolic functions.

LanKC enhances host strain’s ecological fitness and microbiota structural stability

To assess the impact of lanKC on the ecological fitness of its host strain, we compared the ecological competitiveness of the ΔlanKC mutant and the wild-type strain, as well as their effects on the structural stability of the microbial community, using an in vitro fecal fermentation model simulating the human gut environment (Figure 3A). The relative abundance of the wild-type strain increased significantly more than that of the ΔlanKC mutant at multiple time points (Figure 3B), indicating a pivotal role for lanKC in ecological competitiveness. Co-abundance network analysis revealed that the weighted interaction strength between B. breve and other gut microbes was markedly higher in the wild-type intervention group compared to the ΔlanKC group, with differences most pronounced at 12 h and subsequent time points (Figure 3C). This suggests that lanKC enhances the interaction of its host strain with the commensal microbiota, thereby boosting its ecological fitness.

Figure 3.

The lanKC gene enhances B. breve ecological adaptation and stabilizes microbial community structure

(A) Schematic of the in vitro fecal fermentation setup, supplemented with viable wild-type (WT) or ΔlanKC strains.

(B) Changes in B. breve relative abundance over time. y axis indicates a change from 0 h. Statistical significance assessed by two-tailed Student’s t test; ∗, p < 0.01; ∗∗, p < 0.001; data are mean ± SD.

(C) Total interaction strength between B. breve and other species in co-abundance networks. Bar plots show summed weighted interaction strengths (p < 0.05, two-tailed Student’s t test); the accompanying heatmap presents Z score-normalized interaction strengths.

(D) Microbial co-abundance networks at 6, 12, and 18 h in WT and ΔlanKC groups. n, number of nodes; L, number of edges. Node size reflects connectivity; edge thickness denotes correlation strength. Gold and brown edges represent positive and negative correlations, respectively. Only edges with a correlation coefficient > 0.4 and p > 0.05 were included.

(E) Pearson correlations between microbial cohesion and indices of community complexity and stability. Gray indicates non-significance (p > 0.05); ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001.

(F) Positive and negative cohesion values of microbial communities in WT and ΔlanKC groups. Two-tailed Student’s t test was used to assess group differences; ∗∗∗, p < 0.001.

(G) Robustness indices of microbial co-abundance networks (50% node removal). Group differences determined via two-tailed Student’s t test; ∗∗∗, p < 0.001.

Regarding microbial diversity, both the wild-type and ΔlanKC groups increased the alpha diversity relative to the solvent control; however, no significant differences in alpha or beta diversity were observed between these two groups (Figures S3A and S3B). Despite limited effects on the overall diversity metrics, network structure analysis uncovered distinct ecological dynamics, where the wild-type group consistently exhibited a greater number of network nodes and edges at multiple time points (except for nodes at 18 h, which were lower in the wild-type group), indicating a more complex and interconnected microbial network (Figure 3D).

Further evaluation of the microbiota network topology, including average degree, centrality, average path length, modularity, global efficiency, number of edges and nodes, and robustness, demonstrated superior performance of the lanKC-containing microbiota network across multiple complexity indices compared to the ΔlanKC group (Table S6). Notably, community cohesion analysis showed that both positive cohesion (reflecting cooperative interactions²⁰) and negative cohesion (indicative of competitive interactions²⁰) were elevated in the lanKC group. Moreover, negative cohesion positively correlated with multiple network complexity parameters (Figures 3E and 3F), indicating that lanKC promotes stronger and more balanced ecological interactions within the microbiota. Finally, a composite microbiota robustness index, derived from topological parameters, revealed that the lanKC-containing microbiota network exhibited greater structural stability under simulated perturbations (Figure 3G).

LanKC-encoded product targets and inhibits competitor species

To elucidate the mechanisms by which lanKC enhances strain ecological fitness and modulates the microbial community stability, we systematically analyzed changes in the co-abundance relationships between B. breve and other microbial species from wild-type and ΔlanKC groups (Table S7). This revealed 129 microbial species exhibiting differential co-abundance patterns with B. breve between the two groups, with 29 species being unique to the wild-type treatment, and 100 exclusively to the ΔlanKC group (Figure 4A). These taxa likely constitute the primary targets of lanKC-mediated ecological interactions, with their abundance presumably modulated directly or indirectly by the lanKC-encoded product.

Figure 4.

The lanKC gene modulates interactions between B. breve CCFM1025 and gut symbiotic species

(A) Co-abundant species correlated with B. breve in WT and ΔlanKC viable strain intervention groups. Pearson correlation coefficients were calculated based on species-level relative abundances at 6, 12, and 18 h. Significant differences were determined as adj-p < 0.05. Differentially co-abundant species were identified as potential responders to lanKC modulation.

(B) Experimental design of in vitro fecal fermentation using cell-free fermentation supernatants from WT (WT-FS) and ΔlanKC (ΔlanKC-FS) strains.

(C) Differential species showing significant differences in relative abundance between WT-FS and ΔlanKC-FS groups (n = 9). Relative abundances across 6, 12, and 18 h were compared using a two-tailed Student’s t test. Fold change (FC) was calculated as WT-FS/ΔlanKC-FS. Species with p < 0.05 and |log₂FC| > 1 were considered lanKC-responsive.

(D) Temporal dynamics of three lanKC-responsive species at 0, 6, 12, and 18 h in WT-FS and ΔlanKC-FS groups. Statistical comparisons at each time point were performed using two-tailed Student’s t test; ns, p > 0.05; ∗, p < 0.05; ∗∗, p < 0.01; and ∗∗∗, p < 0.001. Data are shown as mean ± SD.

To validate this hypothesis, we assessed the effects of fermentation supernatants from the wild-type and ΔlanKC strains, containing or lacking the lanKC product, respectively, on in vitro cultured microbiota (Figure 4B). Consistent with the live bacterial interventions, the wild-type supernatant significantly enhanced the microbiota network’s positive and negative cohesion as well as overall robustness relative to the ΔlanKC supernatant (Figures S4A and S4B), confirming that the lanKC-encoded product acts as a direct modulator of microbiota interaction structure.

Among the 129 co-abundance species, nine showing significant abundance shifts under the fermentation supernatant treatment were selected for detailed analysis (Figure 4C; Table S8). Temporal profiling revealed that Amedibacterium intestinale, Eubacterium hominis, and Ruminococcus lactaris consistently exhibited negative responses throughout the intervention (Figure 4D), indicating a stable and potent inhibitory effect exerted by the lanKC product. In contrast, other responding species, such as Ruminococcus callidus, an unclassified Clostridiaceae (SGB15090), and Anaerostipes rhamnosivorans, showed inconsistent trends, suggesting their dynamics may be governed by more complex, multi-factorial regulatory mechanisms within the microbiota (Figure S5A).

To characterize the phylogenetic context of these inhibited species, we performed linear discriminant analysis effect size (LEfSe) analysis on their higher-order taxonomic groups, which identified the family Erysipelotrichaceae as the most significantly responsive lineage to the lanKC product (Figure S5B). Notably, multiple members of this family have been implicated as key players in the “gut microbiota health index” (GMHI) and “two competing guilds” (TCG) models, often associated with pathogenicity or dysbiosis.^21,22 A. intestinale, confirmed here as a lanKC target, belongs to this family, though functional characterization remains limited to preliminary 16S and metagenomic annotations.²³ Similarly, E. hominis, another species strongly inhibited by lanKC and classified within the Eubacteriaceae family, showed high taxonomic significance in LEfSe analysis (linear discriminant analysis, LDA > 3) (Figure S5B), highlighting its potential role as an ecological target of lanKC-mediated modulation within the host gut.

Persistent colonization of lanKC-harboring B. breve contributes to the gut microbiota homeostasis in adults

To evaluate the in vivo ecological impact of lanKC-harboring strains, a clinical intervention was conducted wherein healthy adults ingested B. breve CCFM1025 daily for 14 days (Figure 5A). Analysis of the gut microbiota structure revealed no significant changes in alpha diversity between the baseline (day 0) and post-intervention (day 14) (Figure S6A). However, relative abundances of several genus-level taxa within Erysipelotrichaceae and Terrisporobacter (a representative genus of Peptostreptococcaceae) were significantly reduced following intervention (Figures 5B and S6B), closely recapitulating the in vitro findings.

Figure 5.

Differences in colonization and alterations in microbial network structure and stability following B. breve CCFM1025 intervention in humans

(A) Experimental design of the human intervention study with B. breve CCFM1025. Participants were categorized as colonizers or non-colonizers based on the detection of the strain on day 28.

(B) LEfSe analysis of gut microbiota profiles before (day 0) and after (day 14) intervention. Taxa significantly enriched following CCFM1025 consumption (LDA > 2) are shown.

(C) Microbial co-abundance networks of colonizers and non-colonizers on days 14 and 28. n and L indicate the number of nodes and edges, respectively. Edges represent significant correlations (coefficient > 0.15, p < 0.05).

(D) Weighted negative interaction strengths in co-abundance networks on days 14 and 28. Group comparisons were conducted using two-tailed Student’s t test; ns, p > 0.05; and ∗∗, p < 0.01.

(E and F) Microbial network robustness in colonizers and non-colonizers on day 14 (E) and day 28 (F), measured by node retention during random removal. Line plots represent means from 100 replicates, with shaded areas indicating 95% confidence intervals. Statistical analysis was performed by two-way ANOVA; ns, p > 0.05; ∗∗∗, p < 0.001.

On day 28, a strain-specific quantitative PCR assay targeting the lanKC gene was established to determine the absolute abundance of B. breve CCFM1025 in fecal samples. Based on the detection results, participants were classified as colonizers or non-colonizers (details provided in prior work^24,25). Longitudinal analysis during the intervention (days 0 to 14) showed that only colonizers exhibited a significant decrease in Erysipelotrichaceae abundance, a trend absent in non-colonizers (Figures S6C and S6D).

Comparison of the microbiota network structures on days 14 and 28 revealed divergent trajectories. In non-colonizers, network complexity markedly declined by day 28 (Figure 5C), accompanied by a pronounced weakening in the weighted strength of negative correlation edges (Figure 5D), indicative of diminished antagonistic microbial interactions. Conversely, colonizers maintained stable microbiota network complexity and interaction strength, exhibiting no significant structural shifts. While robustness indices did not differ significantly between groups on day 14 (Figure 5E), by day 28, the robustness of non-colonizers’ microbiota networks substantially decreased, falling below that of colonizers (Figure 5F). These findings demonstrate that a persistent colonization by lanKC-harboring B. breve supports the maintenance of gut microbiota structural integrity and resilience against perturbations.

LanKC in B. breve likely originates from intra-genus horizontal gene transfer during early life

We comprehensively screened 796 B. breve genomes (664 publicly available from NCBI and 132 laboratory isolates) for the presence of the lanKC gene, irrespective of the host origin or age of the strains (Table S9). The gene was detected in only nine strains, indicating a rare distribution within the species (Table S10). To elucidate its evolutionary origin, we integrated all available lanKC-containing sequences from both previously isolated strains and NCBI datasets, totaling 5,475 sequences (Tables S10 and S11), including ten from B. breve, and performed phylogenetic analyses.

The resulting phylogeny revealed that lanKC sequences in B. breve segregate into two distinct evolutionary clades. One clade, containing lanKC-1025, was highly conserved and predominantly comprised of Bifidobacterium species, showing over 90% sequence identity with lanKC homologs from other congeneric species (Figure 6A). The closest non-Bifidobacterium homologs originated from the families Brevibacteriaceae and Actinomycetaceae but shared less than 50% sequence similarity, indicating distant evolutionary relatedness (Figure 6A). The second clade similarly displayed genus-level conservation, further supporting the notion of intra-genus transmission (Figure S7A). Notably, strains harboring lanKC-1025-related genes were isolated from diverse mammalian hosts, including humans, monkeys, and bats, suggesting a potential ecological advantage facilitating gene retention across host species (Figure 6A).

Figure 6.

The lanKC gene in B. breve is evolutionarily derived from the Bifidobacterium genus and is predominantly found in B. infantis

(A) Maximum-likelihood phylogenetic tree of lanKC amino acid sequences from the NCBI database and our laboratory isolate collections (n = 5,475). The outer circle of the left tree indicates source phyla. Asterisks and red labels mark the positions of B. breve-derived lanKC sequences. The right panel shows 65 sequences with potential homology to lanKC-1025, annotated at the family level. The far-right panel displays host-associated Bifidobacterium species harboring lanKC, with dashed boxes indicating host origin and a red dashed line highlighting the position of lanKC-1025. Red branches denote evolutionary clusters exclusively derived from Bifidobacterium.

(B) Workflow of pediatric metagenomic cohort analysis and taxonomic annotation, outlining the steps used to identify lanKC within child gut microbiomes.

(C) Prevalence of the lanKC gene across eight Bifidobacterium species in the ELGG genome catalog (n = 4,863). Species are ranked by detection rate; corresponding genome counts are indicated to the right of each bar.

(D) Pairwise amino acid sequence similarity matrix of Bifidobacterium-derived lanKC genes (ELGG catalog), with rows and columns color-coded by species. Clustering was performed using complete linkage.

(E) Maximum-likelihood phylogenetic tree of lanKC sequences from Bifidobacterium species (ELGG catalog), with species origin color-coded below the tree.

To substantiate the hypothesis of intra-genus horizontal gene transfer (HGT), we leveraged the ELGG catalog, which comprises 4,863 metagenome-assembled genomes (MAGs) from eight Bifidobacterium species (Figure 6B). Annotation of lanKC genes within these MAGs revealed widespread presence in B. infantis and B. pseudocatenulatum (detection rates > 50%), occasional occurrence in B. breve, B. bifidum, B. longum, and B. kashiwanohense, and absence in B. adolescentis and B. dentium (Figure 6C). Further phylogenetic analysis of lanKC amino acid sequences delineated two major clades within the genus, i.e., one predominantly comprising B. infantis sequences and the other mainly B. pseudocatenulatum. These clades shared less than 30% sequence identity, indicating distinct evolutionary origins (Figure 6D). All lanKC genes identified in B. breve clustered within the B. infantis-dominated clade, which also included sequences from B. bifidum, B. longum, and B. kashiwanohense. Conversely, the B. pseudocatenulatum clade lacked B. breve sequences, with only sporadic representation of B. kashiwanohense (Figure 6E). These findings suggest that the lanKC gene in B. breve was most likely acquired from B. infantis via intra-genus HGT during early life. However, we were unable to directly detect the coexistence of lanKC-positive B. breve with their putative donor species within a single host sample, limiting direct verification of the HGT events. In addition, the precise ecological context and temporal dynamics of lanKC acquisition within the Bifidobacterium genus remain to be fully elucidated.

Discussion

Conventional theories posit that the ecological adaptation of Bifidobacterium within the human gut is largely driven by the efficiency of carbohydrate metabolism, particularly the utilization of host-derived oligosaccharides. However, this nutrition-centric model alone appears insufficient to explain the ecological behavior of B. breve. Although broad-spectrum carbohydrate metabolism, mediated by glycoside hydrolase families such as GH36,¹⁹ has been presumed to underlie its adaptability, our comparative genomic analysis of infant- and adult-derived B. breve strains revealed a high level of conservation in core genome architecture and carbohydrate utilization genes across host age groups.

To explore these potential complementary mechanisms, our investigation turned to the accessory genome, which increasingly is recognized as a reservoir of genes associated with niche-specific fitness traits, including interspecies competition, environmental sensing, and stress resilience.^26,27 In an adult-derived strain CCFM1025, we identified a complete type III lanthipeptide bacteriocin biosynthetic gene cluster, featuring a key modification enzyme, LanKC. Such bacteriocin biosynthetic modules, typically encoded within the accessory genome, are well-documented mediators of interference-like competition. By exerting narrow-spectrum antimicrobial activity, they enable producer strains to selectively inhibit competing taxa and thereby enhance ecological stability.²⁸

While bacteriocin-mediated adaptation has been postulated previously, direct experimental evidence has remained scarce due to two major challenges: (1) technical limitations in expressing and purifying complete bacteriocin biosynthetic gene clusters for functional validation²⁹ and (2) difficulties in evaluating their in situ ecological impacts within the gut environment.³⁰ To circumvent these constraints, we employed a genetic knockout approach to generate a lanKC-deficient mutant and assessed its ecological impact using fecal fermentation systems and a human intervention study.

Our results provide two lines of direct evidence suggesting a potential ecological role of lanKC. First, deletion of lanKC significantly reduced the colonization efficiency of its host strain within the human gut microbiota. Second, the absence of lanKC-harboring strains weakened microbiota network robustness without markedly altering alpha diversity. Specifically, the presence of lanKC was associated with decreased abundance of bacterial families such as Erysipelotrichaceae and Eubacteriaceae, implying a selective inhibitory effect rather than broad community disruption. These effects were recapitulated in human participants who sustained colonization of the lanKC-positive strain post-intervention, reinforcing the reproducibility of this interference-based advantage.

Gene phylogenetic analyses further revealed that lanKC homologs are evolutionarily conserved within the Bifidobacterium genus. Metagenomic data from over 6,000 infant fecal samples supported the hypothesis that lanKC was horizontally transferred to B. breve from infant-type species such as B. infantis or B. bifidum (whereas B. pseudocatenulatum shows a distinct distribution pattern and is therefore not grouped here). Although lanKC acquisition events appear to be rare, their evolutionary relevance may lie in providing strain-level advantages that enhance persistence in specific host environments.

From an ecological perspective, the acquisition of lanKC may thus represent an auxiliary adaptation—a supplement rather than a replacement to nutrient-based colonization strategies. Unlike B. infantis, whose persistence in infants is largely determined by nutrient availability, B. breve in adults may rely on multi-factorial determinants encompassing both metabolic flexibility and competitive interactions. This supports the notion that bacteriocin biosynthesis acts as an accessory ecological trait, contributing to community stabilization and competitive resilience rather than serving as a universal determinant of colonization. Moreover, such an acquisition may reflect a beneficial co-adaptive process between the microbe and its host. Given that bifidobacterial abundance and diversity decline with age—coinciding with elevated risks of metabolic, inflammatory, and neurodegenerative disorders^31,32,33—the persistence of lanKC-harboring B. breve strains could provide compensatory ecological or immunomodulatory functions reminiscent of infant-type bifidobacteria. Although direct evidence linking lanKC to host benefits remains to be established, its maintenance within certain adult-associated lineages may indicate an adaptive advantage that supports gut homeostasis during host aging.

In conclusion, this study identifies a previously unrecognized, strain-specific auxiliary mechanism that complements the classical nutrient-driven model of B. breve colonization. By integrating comparative genomics, targeted gene knockout, microbiota network analysis, and human validation, we provide experimental evidence that interference-like interactions, mediated by lanKC, can contribute to the ecological competitiveness of specific B. breve strains. Importantly, lanKC should not be viewed as a universal determinant of bifidobacterial ecology but as a supplementary adaptive factor that refines the broader framework of nutritional and metabolic colonization theories.

Limitations of the study

Nonetheless, several limitations merit emphasis. First, due to the limited number of B. breve isolates available in public databases and incomplete host metadata, this study primarily focused on Chinese-derived strains and employed B. breve CCFM1025 as a model. Whether similar accessory genome-mediated strategies operate across broader geographic or ecological contexts remains to be determined. Second, due to ethical and technical considerations, we were unable to directly compare the long-term colonization dynamics between wild-type and lanKC-deficient strains in human hosts, and this limits a more direct in vivo validation of the ecological observations made in vitro. Third, as lanKC was detected in only one adult-derived strain in our dataset, the ecological advantage observed may reflect strain-specific rather than species-wide characteristics. The low prevalence of lanKC-positive isolates and the overall rarity of adult-type B. breve suggest that colonization in adults likely arises from heterogeneous, multi-factorial mechanisms, among which bacteriocin production represents only one adaptive route. Future studies should explore how accessory functions such as bacteriocin biosynthesis integrate with core metabolic traits to shape bifidobacterial persistence under different host and environmental contexts.

Resource availability

Lead contact

Correspondence and requests for materials should be addressed to Peijun Tian (pjtian@jiangnan.edu.cn).

Materials availability

No unique reagents were developed as part of this study.

Data and code availability

The 32,277 genome assemblies and 2,172 representatives of ELGG reported in this paper were downloaded from the Zenodo repository under https://doi.org/10.5281/zenodo.6969520. The sequence data reported in this paper have been deposited in the NCBI database (genome resequencing and metagenomic sequencing data: PRJNA1338785 and PRJNA1338803). The other data supporting the findings of this study are available within the paper and additional files. All the mentioned tools used for data analysis in this study are publicly available, and the version and parameters used have been indicated. This paper does not report original code or custom pipelines.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (32394051, 32201988, and 82241036), Special Fund for Science and Technology Program of Jiangsu Province (BM2022019), and the Program of Collaborative Innovation Center of Food Safety and Quality Control in Jiangsu Province.

Author contributions

Conceptualization, P.T. and S.W.; investigation, J.W., Q.L., X.Q., and Z.J.; methodology, J.W., P.T., X.Q., and N.L.; software, J.W. and X.Q.; resources, S.W., J.Z., and W.C.; writing – original draft, J.W. and P.T.; writing – review & editing, P.T. and S.W.; visualization, J.W. and X.Q.; supervision, J.Z. and W.C.; funding acquisition, S.W., W.C., and P.T.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Bifidobacterium breve: CCFM1025	Culture Collection of Food Microorganisms (CCFM), Jiangnan University	N/A
Bacterial strains; see Table S1	This study	N/A
Biological samples
Human fecal samples	Jingjiang Chinese Medicine Hospital	ChiCTR2200057145
Critical commercial assays
Fast DNA Spin Kit for Feces	MP Biomedical	Cat# 6570200
FastPure Cell/Tissue Total RNA Isolation Kit V2	Vanzyme	Cat# DC102-01
HiScript III All-in-one RT SuperMix Perfect for qPCR	Vanzyme	Cat# R333-01
DNA gel purification Miniprep kit	Biomiga	Cat# DC3511-01
QIAampFastDNAStoolMiniKit	QIAGEN	Cat# 51604
Plasmid Extraction Kit	QINGEN	Cat# DC221-01
Deposited data
Bifidobacterium breve genome data	NCBI	# PRJNA1338785
Metagenomic sequencing data	NCBI	# PRJNA1338803
ELGG	Zeng et al.15	https://doi.org/10.5281/zenodo.6969520
Software and algorithms
SOAPdenovo (v2.0.4)	Xie et al.34	http://soap.genomics.org.cn/soapdenovo.html#intro2
GapCloser	Xu et al.35	https://sourceforge.net/projects/soapdenovo2/files/GapCloser/bin/r6/GapCloser-bin-v1.12-r6.tgztar
Prokka (v.1.14.6)	Seemann36	https://github.com/tseemann/prokka
Roary (v3.13.0)	Page et al.37	https://sanger-pathogens.github.io/Roary
OrthoFinder (v2.5.2)	Emms et al.38	https://github.com/davidemms/OrthoFinder
MAFFT (v7.453)	Katoh et al.39	https://mafft.cbrc.jp/alignment/software
dbCAN (v4.1.4)	Huang et al.40	https://bcb.unl.edu/dbCAN2
antiSMASH (v7.1.0)	Blin et al.41	https://antismash.secondarymetabolites.org
Kneaddata (v.0.10.0)	The Huttenhower Lab	https://github.com/biobakery/kneaddata
MetaPhlAn4 (v4.0.6)	Segata et al.42	https://github.com/biobakery/MetaPhlAn
R (v4.5.0)	R Foundation	https://www.r-project.org
GraphPad Prism 9.5	GraphPad Software	https://www.graphpad.com/

Experimental model and study participant details

Human subjects

Fresh stool samples were obtained from ten healthy adults (five men and five women, aged 20–30 years, BMI 18.5–23.9) who had not used antibiotics, probiotics, or prebiotics within the previous six months. Each participant consumed a daily dose of freeze-dried Bifidobacterium breve CCFM1025 powder containing 1 × 10¹⁰ CFU for 14 consecutive days. Fecal samples were collected at three time points: prior to administration (day 0), after 14 days of intake (day 14), and two weeks following discontinuation (day 28). Ethical approval for the study was granted by the Ethics Committee of Jingjiang Chinese Medicine Hospital (approval No. EN2021078), and all volunteers provided written informed consent. The study protocol was registered in the Chinese Clinical Trial Registry (ChiCTR2200057145).

Method details

Isolation and genome sequencing of B. breve

A total of 100 B. breve strains were isolated from fecal samples collected from individuals of various ages (n = 358, aged from 0 to 108 years) and provinces (n = 15) across China. To ensure that the isolates represented genuine host-associated commensals rather than transiently ingested strains, donors with a recent history of probiotic consumption were excluded. The isolates were anaerobically cultured on modified de Man, Rogosa, and Sharpe (mMRS)²⁵ agar and broth media supplemented with 0.05% (w/v) L-cysteine, using a Whitley DG250 anaerobic workstation (Don Whitley Scientific Limited, Shipley, UK). Cultures were incubated at 37°C for 24 h. Genomic DNA was extracted from pure cultures using a commercial DNA extraction kit (Qiagen, cat.51604) according to the manufacturer’s instructions. In addition, 65 publicly available B. breve genomes were retrieved from the NCBI database (Taxonomy ID: 1685), and only isolates with clearly annotated host information were included in the analysis.

Whole-genome sequencing was conducted using the Illumina HiSeq platform, generating 350-bp paired-end reads. Raw reads were assembled into high-quality contigs using SOAPdenovo (v2.0.4),³⁴ followed by scaffold gap closure and base-level error correction with GapCloser.³⁵

Phylogenetic and CAZyme annotation analyses of B. breve genomes

Genome annotation was performed using Prokka (v1.14.6)³⁶ to predict coding DNA sequences (CDSs). Pan-genome analysis was carried out using Roary (v3.13.0)³⁷ to define core and accessory genomes. Functional annotation of CDSs was achieved by BLASTP searches against the NCBI non-redundant (nr) and Clusters of Orthologous Groups (COG) databases.

Orthologous proteins across the 100 genomes were identified using OrthoFinder (v2.5.2)³⁸ with default parameters. Shared orthologs were aligned using MAFFT (v7.453),³⁹ and a phylogenetic tree was constructed based on the maximum likelihood method implemented in FastTree.⁴³ To identify host-associated genomic differences, orthologous proteins present in fewer than 50% of both adult- and infant-derived strains were excluded. Differences in ortholog presence between the two groups were assessed using Fisher’s exact test, with p-values corrected for the false discovery rate (FDR) using the Benjamini-Hochberg method.

Carbohydrate-active enzyme (CAZyme) families were annotated using hidden Markov models (HMMs) via dbCAN (v4.1.4).⁴⁰ Phylogenetic trees of individual CAZyme families were constructed using multiple sequence alignment followed by maximum likelihood estimation, with strain source indicated on the trees. Using genetic distance matrices, strains were classified into adult-derived clusters and a control group comprising all isolates. For each CAZyme gene, 1,500 random resamplings were performed within each group, and the mean inter-group genetic distance was calculated. The statistical significance of the observed clustering patterns was evaluated using permutation tests, with p-values derived from the resampling distribution.

Prediction of strain-specific gene clusters and phylogenetic analysis of B. breve

Genomes harboring lanKC homologs were subsequently analyzed with antiSMASH (v7.1.0)⁴¹ to predict secondary metabolite biosynthetic gene clusters (BGCs), enabling assessment of cluster integrity and similarity. To investigate the distribution of the lanKC gene within the Bifidobacterium genus, a total of 5,475 lanKC protein sequences were collected from the NCBI database (https://www.ncbi.nlm.nih.gov/) and our laboratory isolate collections. This profile was applied to evaluate the intra-species distribution of lanKC homologs within B. breve. Phylogenetic relationships were inferred by constructing a maximum likelihood tree based on aligned amino acid sequences of lanKC homologs, including a reference lanKC, to define phylogenetic clusters.

Construction of insertion mutants and growth curve analysis

Insertional mutagenesis of the lanKC gene in B. breve strain CCFM1025 was performed using a single-crossover homologous recombination approach.⁴⁴ The mutagenesis plasmid was constructed by inserting a 500-bp internal fragment of the target gene into the multiple cloning site (MCS) of the tetracycline-resistant plasmid pTET28, which was derived from a modified pFREM28 backbone.⁴⁵The recombinant plasmid was introduced into CCFM1025 via electroporation (2,500 V, 25 μF, 200 Ω). Cells were recovered in mMRS broth at 37°C under anaerobic conditions for 3 h, then plated on tetracycline-containing mMRS agar and incubated anaerobically at 37°C for 72 h. Colonies were screened via PCR with gene-specific primers to confirm insertion. The mutant strain (CCFM1025-ΔlanKC) and wild-type control (CCFM1025-WT) were cultured in antibiotic-free mMRS broth at 2% (v/v) inoculum. OD₆₀₀ was recorded hourly until stationary phase, with additional measurements of pH and viable cell counts to compare growth performance.

In vitro fecal fermentation

Each fecal sample was homogenized in 0.1 M PBS (pH 7.0) containing 0.1% (w/v) L-cysteine at a 1:10 (w/v) ratio. After centrifugation at 600 × g for 10 min at 4°C, supernatants from all samples were pooled in equal volumes and inoculated into GMM medium at a final concentration of 2% (v/v).⁴⁶ Specific strains or fermentation supernatants were then introduced into experimental groups, followed by anaerobic incubation for 18 h. Samples were collected every 6 h, and bacterial pellets were harvested for downstream analyses.

Shotgun metagenomic sequencing and analysis

DNA was extracted from fermentation pellets using the QIAamp Fast DNA Stool Mini Kit. Library construction and sequencing were performed using the MGI DNBSEQ-T7 platform. Raw reads were filtered using KneadData (v0.10.0) to remove low-quality sequences and host-derived contamination (mapped to GRCh38). Taxonomic profiling was conducted using MetaPhlAn4 (v4.0.6).⁴²

16S rRNA gene sequencing and microbiota profiling

Total bacterial DNA from fecal samples was used for PCR amplification of the V3–V4 regions of the 16S rRNA gene. Amplicons were visualized on a 1.5% agarose gel, purified, quantified, and pooled to construct sequencing libraries. Sequencing was conducted using the Illumina MiSeq platform. Raw reads were processed with QIIME2.⁴⁷ OTUs were clustered at 97% identity, and taxonomic assignments were made using the SILVA reference database.⁴⁸ Alpha diversity was assessed via the Shannon index calculated from rarefied OTU tables.

Construction of microbial co-abundance networks

Microbial interaction networks were inferred using SparCC, a correlation estimation method tailored for sparse, compositional microbiome data. Species-level relative abundances were obtained from metagenomic sequencing.⁴⁹ Taxa with prevalence <50% or relative abundance <0.01% were excluded. SparCC was run with 100 bootstrap replicates. Microbial pairs with absolute correlation coefficients >0.4 and p < 0.05 were retained for network construction.

Network complexity and stability analysis

Key topological properties—average degree, centrality, average path length, modularity, global efficiency, number of edges, and number of nodes—were calculated using the igraph package in R (v4.5.0). Network stability was assessed via robustness simulations, which evaluated the mean interaction strength of remaining nodes after random node removal.⁵⁰ Additionally, the abundance-weighted cohesion metric, corrected by a null model, was computed to quantify microbial community complexity. A modified SparCC-weighted interaction strength was also calculated to further characterize network resilience.

Quantification and statistical analysis

Statistical analyses were performed using GraphPad Prism (version 9.5) and R (version 4.5.0). Data were tested for normality, and outliers were identified using Grubbs’ test. Two-tailed Student’s t-tests were used to assess group differences in relative species abundance, total interaction strength, cohesion, and robustness. Pearson correlation analysis was applied to evaluate the relationships between cohesion and network parameters, as well as between B. breve and the relative abundance of other species. two-way ANOVA was used to assess differences in growth curves and community robustness between wild-type and ΔlanKC strains. Linear discriminant analysis (LDA) of LEfSe results was used to evaluate differences in microbial composition at different time points in human cohort experiments. p-values were adjusted for multiple comparisons using the Benjamini-Hochberg-FDR procedure (adj-p).

Published: December 19, 2025