Human milk microbiome as a modulator of the early-life gut–brain axis: mechanisms and translational opportunities for neurodevelopment Authors: Fatima Alnuaimi, Lidya K. Yassin, Salma Alketbi, Jurga Skrabulyte-Barbulescu, Saif Almazrouei, Dana Alremeithi, Noud Alahbabi, Sultan Almarzooqi, Shamma H. Alkuwaiti, Mohammad I. K. Hamad Published: 2026 ABSTRACT The first 1000 days of life represent a critical developmental window during which early microbial colonization contributes to immune, metabolic, and neurodevelopmental programming. Human milk is increasingly recognized as a biologically active fluid that shapes infant gut microbiome assembly through microorganisms, human milk oligosaccharides (HMOs), immune factors, and microbial metabolites. Within this framework, the human milk microbiome has emerged as a potential contributor to the milk–gut–brain axis (MGBA), a bidirectional communication network linking the gut microbiota with neural, immune, endocrine, and metabolic pathways involved in brain development. This review summarizes current evidence regarding the origins, determinants, and functional relevance of the human milk microbiome and its potential role in early-life neurodevelopment. Proposed microbial sources include maternal skin, the infant oral cavity, environmental exposure, and the entero-mammary pathway, while key determinants include lactational stage, delivery mode, antibiotic exposure, maternal diet, obesity, and prematurity. Mechanistic pathways linking milk-associated microbes with neurodevelopment are discussed, including microbial colonization, immune and barrier maturation, vagal and neuroendocrine signalling, and production of short-chain fatty acids (SCFAs) and tryptophan-derived metabolites. We further evaluate evidence relating breastfeeding and milk-associated microbial exposures to cognitive, behavioural, and neurodevelopmental outcomes, particularly in preterm and medically vulnerable infants. Although experimental and observational evidence supports biologically plausible links between the human milk microbiome and the developing microbiota–gut–brain axis, major uncertainties remain regarding microbial viability, sustained colonization, causality, and long-term functional significance. Methodological limitations, including low microbial biomass, contamination susceptibility, and heterogeneity in analytical approaches, continue to complicate interpretation across studies. Future progress will require longitudinal, mechanistically informed studies integrating microbiome profiling with metabolomics, immune phenotyping, neuroimaging, and validated neurodevelopmental outcomes. Introduction The first 1000 days of life, extending from conception to the child’s second birthday, represent a critical developmental window during which the microbiome, immune system, metabolism, and brain undergo rapid and coordinated maturation [1, 2]. During this period, early microbial colonization is shaped by maternal health, delivery mode, antibiotic exposure, gestational age, feeding practices, and environmental factors, all of which may influence long-term developmental trajectories [3, 4]. Increasing evidence suggests that these early-life microbial exposures interact with the microbiota–gut–brain axis, a bidirectional communication network linking the gut microbiota with the central nervous system (CNS) through neural, immune, endocrine, and metabolic pathways [5]. Recent reviews further highlight the microbiota–gut–brain axis as a central framework linking microbial dysbiosis with neurodevelopmental and neurological disorders across the lifespan [6–9]. Experimental studies indicate that microbial signals can influence stress regulation, immune maturation, blood–brain barrier (BBB) integrity, neurotransmitter-related pathways, and microglial development during sensitive developmental windows [10–12]. These effects are mediated through pathways involving microbial metabolites, vagal signalling, cytokine regulation, and neuroendocrine interactions [13]. Although much of the evidence remains preclinical, these findings support the concept that early microbial ecology may contribute to neurodevelopmental processes. Within this framework, human milk is increasingly recognized as more than nutrition alone. Human milk contains microorganisms, HMOs, immunoglobulins, antimicrobial proteins, lipids, cytokines, and other bioactive compounds that shape infant gut colonization and immune development [14, 15]. HMOs selectively support milk-adapted taxa, including Bifidobacterium, and thereby influence infant microbial succession and metabolite production [16–19]. The human milk microbiome itself has emerged as an important area of investigation. Multiple studies have identified bacterial taxa in human milk, including Staphylococcus, Streptococcus, Lactobacillus, and Bifidobacterium, although considerable variability exists across studies [20, 21]. Proposed sources include maternal skin, the infant oral cavity, environmental exposure, and a possible entero-mammary pathway linking maternal gut microbes with the mammary gland [22, 23]. However, important methodological challenges complicate interpretation across studies [24, 25]. Consequently, important questions persist regarding microbial viability, sustained colonization, and functional relevance. Breastfeeding is consistently associated with distinctive infant gut microbiome development. Compared with formula feeding, breastfeeding supports a milk-adapted infant gut microbiome, whereas formula-fed infants often exhibit earlier microbial diversification [26, 27]. Breastfeeding may also partially buffer microbial disruptions associated with caesarean delivery and antibiotic exposure [28]. These observations are particularly relevant because early microbial colonization coincides with critical periods of immune and neurodevelopment. Emerging evidence further suggests that human milk components, including HMOs and microbial metabolites, may be associated with cognitive, behavioural, and neurodevelopmental outcomes, particularly in preterm and medically vulnerable infants [29, 30]. However, most available evidence remains observational, and direct causal relationships between the human milk microbiome and neurodevelopment have not been established. This review examines the human milk microbiome as a potential modulator of the early-life MGBA. The strength of evidence differs across this pathway. Links between the infant gut microbiome and neurodevelopment are supported by experimental, mechanistic, and human observational evidence [10–12, 19, 31–52]. Associations between breastfeeding and neurodevelopmental outcomes are also supported by human observational studies, although residual confounding remains important [29, 30, 34–38, 41–45, 48–60]. In contrast, direct evidence that milk-derived microbes themselves causally influence neurodevelopment is limited. Therefore, the human milk microbiome is considered a biologically plausible contributor to early-life gut–brain signalling rather than an established causal determinant of neurodevelopment. We discuss the origins and determinants of the human milk microbiome, mechanisms linking microbial colonization with immune and neurodevelopmental processes, the role of microbial metabolites and signalling pathways, clinical evidence relating breastfeeding and neurodevelopment, and emerging translational opportunities and challenges. Throughout, emphasis is placed on distinguishing mechanistic plausibility from demonstrated clinical benefit and on identifying the key knowledge gaps that must be addressed before microbiome-targeted interventions can be translated into clinical practice. Literature identification strategy This article is a comprehensive review rather than a systematic or scoping review. Therefore, a formal PRISMA screening, risk-of-bias assessment, and meta-analysis were not performed. Relevant literature was identified through searches of PubMed, Scopus, and Google Scholar. The search strategy was designed to identify publications issued between 2000 and 2026. Search terms included combinations of “human milk microbiome”, “breast milk microbiota,” “milk virome,” “milk mycobiome,” “human milk oligosaccharides,” “infant gut microbiota,” “microbiota–gut–brain axis,” “early-life microbiome,” “short-chain fatty acids,” “tryptophan metabolism,” “immune development,” “preterm infants,” “breastfeeding,” “neurodevelopment,” “cognition,” “autism spectrum disorder,” and “attention-deficit/hyperactivity disorder.” Eligible publications included original human studies, relevant animal and mechanistic studies, clinical trials, meta-analyses, and recent systematic reviews addressing human milk composition, microbial exposure and colonization in infants, microbial metabolites, immune and barrier maturation, gut–brain axis pathways, and neurodevelopmental outcomes. Priority was given to studies providing greater mechanistic insight, including longitudinal investigations, mother–infant dyad analyses, studies involving preterm infants, and studies assessing clinically neurodevelopmental endpoints. We excluded publications if they were not available in English, had poor fit with human milk and early-life microbial development, centered on unrelated adult microbiome questions, did not give enough methodological context, or enough biological detail to interpret them. Since this review brings together mechanistic, observational, and translational evidence, the evidence was synthesized narratively. The interpretation was guided by biological plausibility and the recognized limitations of causal inference in the included study designs. To clarify the types of evidence integrated in this comprehensive review, Table 1 summarizes the main evidence categories, their relevance to the human milk microbiome–MGBA framework, and whether they directly address neurodevelopmental outcomes.Table 1Categories of evidence informing the relationship among human milk, the early-life microbiome, the MGBA, and neurodevelopmentEvidence CategoryEvidence Considered in the ReviewContribution to the ReviewRelevance to NeurodevelopmentAnimal and mechanistic studiesGF, colonization, and experimental models examining microbial effects on HPA-axis regulation, BBB integrity, microglial maturation, vagal signalling, SCFA-related pathways, and tryptophan metabolism [10–13, 33, 39, 47, 61–73]These studies provide biological plausibility and some mechanistic support for MGBA signalling during early development.These studies address neurobiological processes directly, but their preclinical nature limits direct extrapolation to human infant neurodevelopment.Human observational studies of human milk and infant microbiotaCohort and mother–infant dyad studies examining associations among breastfeeding, milk-associated microbes, HMOs, delivery mode, antibiotic exposure, prematurity, and infant gut microbial composition [20–30, 74–98]These studies characterize associations between human milk exposures and infant microbial colonization, microbial succession, and immune-related pathways.Relevance to neurodevelopment is generally indirect because most studies do not include validated neurodevelopmental endpoints.Human observational studies with neurodevelopmental outcomesStudies and meta-analyses evaluating associations between breastfeeding, HMOs, early microbial exposures, or stool-derived microbial features and cognitive, behavioral, language, socioemotional, or preterm neurodevelopmental outcomes [29–31, 34–38, 41–45, 48–60, 99–102]These studies provide the strongest human evidence linking milk-related exposures or microbiome-associated markers with developmental outcomes.These studies directly assess neurodevelopmental outcomes, but the evidence remains primarily associative and vulnerable to residual confounding.Human intervention and translational studiesStudies evaluating probiotics, prebiotics, synbiotics, donor milk, maternal interventions, and NICU-based microbiome-targeted strategies [103–119]These studies examine whether early microbial composition or function can be modified in clinically relevant populations, particularly preterm infants.Direct evidence for neurodevelopmental benefit is limited, as most interventions assess microbiome, inflammatory, feeding, or short-term clinical endpoints rather than long-term validated neurodevelopment.Abbreviations: BBB, blood-brain barrier; GF, germ-free; HMO, human milk oligosaccharide; MGBA, milk-gut-brain axis; NICU, neonatal intensive care unit; SCFA, short-chain fatty acid Table 1Categories of evidence informing the relationship among human milk, the early-life microbiome, the MGBA, and neurodevelopmentEvidence CategoryEvidence Considered in the ReviewContribution to the ReviewRelevance to NeurodevelopmentAnimal and mechanistic studiesGF, colonization, and experimental models examining microbial effects on HPA-axis regulation, BBB integrity, microglial maturation, vagal signalling, SCFA-related pathways, and tryptophan metabolism [10–13, 33, 39, 47, 61–73]These studies provide biological plausibility and some mechanistic support for MGBA signalling during early development.These studies address neurobiological processes directly, but their preclinical nature limits direct extrapolation to human infant neurodevelopment.Human observational studies of human milk and infant microbiotaCohort and mother–infant dyad studies examining associations among breastfeeding, milk-associated microbes, HMOs, delivery mode, antibiotic exposure, prematurity, and infant gut microbial composition [20–30, 74–98]These studies characterize associations between human milk exposures and infant microbial colonization, microbial succession, and immune-related pathways.Relevance to neurodevelopment is generally indirect because most studies do not include validated neurodevelopmental endpoints.Human observational studies with neurodevelopmental outcomesStudies and meta-analyses evaluating associations between breastfeeding, HMOs, early microbial exposures, or stool-derived microbial features and cognitive, behavioral, language, socioemotional, or preterm neurodevelopmental outcomes [29–31, 34–38, 41–45, 48–60, 99–102]These studies provide the strongest human evidence linking milk-related exposures or microbiome-associated markers with developmental outcomes.These studies directly assess neurodevelopmental outcomes, but the evidence remains primarily associative and vulnerable to residual confounding.Human intervention and translational studiesStudies evaluating probiotics, prebiotics, synbiotics, donor milk, maternal interventions, and NICU-based microbiome-targeted strategies [103–119]These studies examine whether early microbial composition or function can be modified in clinically relevant populations, particularly preterm infants.Direct evidence for neurodevelopmental benefit is limited, as most interventions assess microbiome, inflammatory, feeding, or short-term clinical endpoints rather than long-term validated neurodevelopment.Abbreviations: BBB, blood-brain barrier; GF, germ-free; HMO, human milk oligosaccharide; MGBA, milk-gut-brain axis; NICU, neonatal intensive care unit; SCFA, short-chain fatty acid Categories of evidence informing the relationship among human milk, the early-life microbiome, the MGBA, and neurodevelopment Abbreviations: BBB, blood-brain barrier; GF, germ-free; HMO, human milk oligosaccharide; MGBA, milk-gut-brain axis; NICU, neonatal intensive care unit; SCFA, short-chain fatty acid Human milk microbiome: what it is, where it comes from, and what shapes it Defining the human milk microbiome The human milk microbiome refers to the microorganisms detected in human milk and their potential biological roles within the maternal–infant interface [15]. This concept reflects an important shift in the understanding of human milk, which is increasingly recognized not only as a source of nutrition but also as a biologically active fluid containing microbial, immunological, metabolic, and bioactive components that may contribute to infant gut and immune maturation [15]. Sequencing-based studies have shown that bacterial communities can be detected in human milk and may display relative stability within individuals over time [20]. Complementary culture-dependent and culture-independent approaches have further demonstrated that bacteria can be recovered directly from human milk samples [21]. However, microbial detection must be interpreted cautiously. Detection of bacterial DNA does not necessarily confirm viability, metabolic activity, or biological function, particularly in low-biomass samples such as human milk [24, 25, 120]. Therefore, the key question is not simply whether microorganisms can be detected in milk, but which taxa are reproducibly observed, whether they are viable, and whether they contribute meaningfully to infant microbial colonization and downstream biological processes. Milk virome and mycobiome Although most human milk microbiome studies have focused on bacteria, human milk also contains viral and fungal components that may contribute to early microbial ecology. The milk virome appears to be dominated by bacteriophages, which can influence bacterial community structure through host-specific predation, lysogeny, and horizontal gene transfer. Metagenomic analysis of human milk has shown that bacteriophages represent a major fraction of detected viral reads, with composition varying according to lactational stage, delivery mode, prematurity, and birth weight [121]. Human milk and infant stool viromes also show shared and distinct viral features, supporting the possibility that breastfeeding contributes to early-life viral exposure and infant gut virome development [122]. However, the biological significance of the milk virome remains insufficiently defined, and its direct relevance to gut–brain axis signalling or neurodevelopmental outcomes has not yet been established. The human milk mycobiome is also emerging as a relevant but understudied component of milk microbial ecology. Fungal DNA and viable fungi have been detected in milk from healthy lactating women, with reported genera including Malassezia, Candida, Saccharomyces, Rhodotorula, and others [123]. Larger comparative studies suggest that milk fungal profiles may vary by geographic location, delivery mode, and bacteria–fungi interactions, indicating that the mycobiome may be shaped by maternal, environmental, and perinatal factors [124]. Because fungi can interact with bacteria and host immune pathways, the milk mycobiome may contribute to infant gut microbial succession and immune education. Nevertheless, evidence remains limited, and future studies should integrate bacteriome, virome, and mycobiome profiling to clarify inter-kingdom interactions, microbial viability, transmission to the infant gut, and relevance to MGBA-related outcomes [125]. Core bacterial genera reported in human milk Across studies, Staphylococcus and Streptococcus are among the most consistently reported genera in human milk and were identified as dominant taxa in early milk microbial profiling studies [20]. Corynebacterium has also been repeatedly detected, supporting the possibility that certain milk-associated taxa recur across different cohorts and analytical approaches [20]. Earlier studies also reported Propionibacterium, a group now largely reclassified within Cutibacterium [20]. Other genera, including Lactobacillus and Bifidobacterium, have been described in human milk, although their detection varies across studies and may depend on maternal, infant, geographic, and methodological factors [15]. The repeated identification of these genera suggests that human milk microbial profiles may not simply reflect random contamination; rather, they may contain recurring microbial signatures with potential biological relevance [23]. Nevertheless, recurrence alone does not establish function. The significance of these taxa depends on whether they are viable, transferred during breastfeeding, and capable of interacting with the infant gut ecosystem. This issue is particularly relevant for Bifidobacterium, as related phylotypes have been identified in both maternal milk and infant fecal samples within mother–infant dyads [126]. Understanding why these microorganisms are detected in milk requires consideration of their proposed origins. Proposed origins of the human milk microbiome The origin of microorganisms detected in human milk is likely multifactorial. One proposed route is the entero-mammary pathway, in which selected maternal gut bacteria may reach the mammary gland through immune-cell-mediated trafficking during late pregnancy or lactation [22]. This model is biologically plausible because it provides a potential mechanism through which maternal gut-derived microorganisms could be delivered to the infant through breastfeeding. However, direct evidence in humans remains limited, and the pathway should be interpreted as a proposed mechanism rather than an established route of microbial transfer [22]. Importantly, human evidence supporting direct trafficking of maternal gut bacteria to the mammary gland remains limited and debated. Much of the entero-mammary pathway hypothesis is based on biological plausibility, animal data, immune-cell trafficking models, and the detection of shared or related taxa across maternal gut, human milk, and infant stool samples [22, 23, 74, 75]. A distinct Bifidobacterium breve strain has been detected in maternal rectal samples, breast milk, and infant stool, providing strain-level evidence consistent with possible entero-mammary transfer [127]. However, such findings remain limited and do not by themselves prove that bacteria first translocated from the maternal intestine to the mammary gland. Shared taxa may instead reflect maternal skin, infant oral retrograde transfer during breastfeeding, environmental exposure, or postnatal mother–infant contact [128, 129]. Therefore, the entero-mammary pathway should be interpreted as a plausible but incompletely demonstrated route in humans. Confirming this pathway will require longitudinal maternal–infant studies that sample the maternal gut, mammary/areolar skin, milk, infant oral cavity, and infant stool at multiple time points, combined with strain-resolved metagenomics, viability assessment, source-tracking analysis, and rigorous contamination control. Maternal skin microbiota represents another plausible source, as many taxa commonly detected in milk are also associated with skin microbial communities [129]. The infant oral cavity may also contribute through retrograde transfer during breastfeeding, whereby oral bacteria are introduced into the mammary environment during suckling [129]. In addition, environmental, perinatal, and maternal factors may shape milk microbial patterns [128]. Current evidence therefore supports a multifactorial model in which the human milk microbiome likely reflects combined maternal, infant, and environmental influences rather than a single exclusive source [23, 130]. Determinants and dynamics of the human milk microbiome Changes across lactation If human milk microbial composition reflects multiple maternal, infant, and environmental inputs, it would be expected to change across lactation rather than remain static. Current evidence supports this view, suggesting that the milk microbiome is dynamic and shifts alongside physiological changes in lactation, maternal biology, infant feeding behaviour, and maternal–infant interaction. Human milk may therefore be more appropriately understood as a dynamic ecological interface rather than a fixed microbial reservoir, with its composition changing from colostrum to mature milk as shown in (Fig. 1).Fig. 1Human milk bioactives across lactation and their potential roles in infant gut and immune development. Colostrum is enriched in immune-active components, transitional milk exhibits microbial and compositional shifts, and mature milk may show greater microbial stability. Human milk-derived microbes, immune factors, HMOs, extracellular vesicles, and other bioactive components may be associated with gut microbiome establishment, host immune programming, and neurodevelopmental processes. These pathways represent hypothesized mechanisms based on current evidence and should not be interpreted as established causal relationships. Created with BioRender Human milk bioactives across lactation and their potential roles in infant gut and immune development. Colostrum is enriched in immune-active components, transitional milk exhibits microbial and compositional shifts, and mature milk may show greater microbial stability. Human milk-derived microbes, immune factors, HMOs, extracellular vesicles, and other bioactive components may be associated with gut microbiome establishment, host immune programming, and neurodevelopmental processes. These pathways represent hypothesized mechanisms based on current evidence and should not be interpreted as established causal relationships. Created with BioRender Colostrum in the immediate postpartum period often contains microbial profiles that differ from those observed in transitional and mature milk [76, 77]. During transitional lactation, microbial composition changes alongside increasing milk volume, altered milk biochemistry, and evolving infant feeding patterns. This period may be particularly important because it coincides with rapid infant gut microbiome assembly, when microbial ecosystems remain highly responsive to early-life exposures. Once mature milk is established, microbial profiles may stabilize to some extent, although they continue to reflect ongoing maternal, infant, and environmental influences. Breastfeeding may further influence these dynamics through retrograde microbial exchange between the infant oral cavity and mammary gland [18]. Together, these findings indicate that lactational stage is a major determinant of milk microbial ecology and may influence the microbial exposures delivered to the infant over time. Maternal and perinatal determinants Beyond lactational stage, milk microbial composition is influenced by multiple maternal and perinatal factors acting through interconnected biological and environmental pathways. Maternal diet has been associated with differences in milk microbial diversity and composition, potentially through interactions with maternal metabolism and gut microbial ecology [78]. Maternal metabolic status, including obesity, has also been linked to altered microbial diversity and shifts in relative taxonomic composition [76]. Perinatal factors such as antibiotic exposure, mode of delivery, and gestational age further contribute to microbial variability [76, 79]. Caesarean delivery may alter maternal microbial exposure and perinatal physiology, while antibiotic exposure may disrupt maternal microbial ecosystems during a critical transmission window. Pregnancy complications, psychosocial stress, and geographic variation may additionally shape microbial patterns in milk [77, 80]. Collectively, these findings suggest that the human milk microbiome represents a biologically responsive interface shaped by maternal physiology, clinical exposures, and environmental context. However, before these determinants can be confidently interpreted as biological effects, it remains essential to distinguish true biological variation from methodological artifacts. Methodological challenges in human milk microbiome Research Low-biomass and contamination A defining limitation of human milk microbiome research is that milk is a low-biomass biological sample, meaning microbial DNA is often present in very small quantities relative to host-derived nucleic acids [25]. This creates substantial technical vulnerability because even small amounts of contaminant DNA introduced through laboratory reagents, extraction kits, collection materials, or sample handling can distort microbial profiles [81]. Rigorous negative controls, extraction blanks, contamination-aware analytical pipelines, and transparent methodological reporting are therefore essential. DNA detection versus viability Most sequencing approaches detect microbial DNA rather than viable organisms. Consequently, identification of bacterial DNA in milk does not confirm that microorganisms are alive, metabolically active, or capable of colonizing the infant gut [82]. This distinction fundamentally limits interpretation because the biological relevance of the milk microbiome depends not only on microbial detection, but also on viability, functional activity, and transmission potential. Sequencing and analytical limitations Sequencing strategy further influences interpretation. Many studies rely on 16S rRNA gene sequencing, which provides taxonomic profiling but limited strain-level and functional resolution. Shotgun metagenomics may offer greater mechanistic insight, including strain-level sharing between maternal milk and infant gut microbiota, but it is often constrained by low microbial biomass and high host DNA content [25, 75]. Resolving these challenges will require standardized protocols, contamination-resilient workflows, and longitudinal study designs specifically optimized for low-biomass samples. Functional relevance of the human milk microbiome Direct microbial transfer Despite methodological limitations, current evidence supports the possibility that human milk contributes directly to infant microbial exposure. Breastfeeding repeatedly introduces milk-associated microorganisms into the infant gastrointestinal tract, and multiple mother–infant cohort studies have reported overlap between taxa detected in maternal milk and infant stool [74]. Higher-resolution analyses have also identified shared bacterial strains between maternal milk and infant gut microbiota [75], supporting the hypothesis that human milk may contribute to early microbial exposure and potentially to aspects of infant microbial seeding. Strain-level transmission and source-attribution analyses are particularly important because taxonomic overlap alone cannot establish milk to-infant microbial transfer. Source-tracking approaches have estimated that, during early infancy, breast milk and areolar skin can contribute measurable proportions of bacteria detected in the infant gut, although these estimates vary according to breastfeeding intensity, sampling time, and analytical method [74]. In addition, strain-resolved studies have reported shared strains between maternal milk and infant stool, including Bifidobacterium and Streptococcus-related taxa, supporting possible mother–milk–infant microbial sharing [75, 127]. However, strain sharing does not by itself prove directionality, sustained colonization, or functional activity. Shared strains may reflect transfer from milk to the infant gut, retrograde transfer from the infant oral cavity to milk, maternal skin, or areolar sources, or shared environmental exposure. Therefore, future studies should combine longitudinal sampling of maternal stool, areolar skin, milk, infant oral cavity, and infant stool with strain-resolved metagenomics to clarify the contribution of human milk as a microbial source. Indirect effects through human milk oligosaccharides and bioactive components The biological significance of human milk extends beyond direct microbial transfer. Human milk contains HMOs, immune factors, antimicrobial proteins, and other bioactive compounds that shape microbial colonization indirectly by selectively supporting beneficial microbial taxa and influencing ecological succession in the developing infant gut [18]. Through these mechanisms, milk may function not only as a microbial source, but also as an ecological regulator of early microbial development. Implications for infant gut colonization and the MGBA Taken together, human milk should be understood as a dynamic maternal–infant biological interface with potential relevance to infant gut colonization through both direct microbial transfer and indirect ecological modulation. Early microbial colonization contributes to immune maturation, inflammatory regulation, and gut barrier development, all of which may influence broader developmental systems. Through microbial metabolites, immune signalling pathways, and gut–brain communication networks, these early-life processes may also have relevance for neurodevelopment. However, major uncertainties remain regarding microbial viability, sustained colonization, and functional significance. Clarifying these questions is essential for determining how the human milk microbiome contributes to the broader MGBA and whether these microbial exposures exert meaningful long-term developmental effects. Breastfeeding, milk exposure, and infant gut colonization Establishment of the infant gut microbiome begins immediately after birth and represents one of the earliest developmental processes shaping immune, metabolic, and epithelial maturation. Although the traditional “sterile womb” paradigm has been challenged by reports of microbial DNA in placental and fetal compartments, it remains uncertain whether these findings reflect true prenatal colonization, transient microbial exposure, or methodological contamination [131]. Regardless of this debate, the most substantial and sustained microbial colonization occurs after birth, when maternal, dietary, clinical, and environmental exposures shape the developing gut ecosystem. At birth, the neonatal gut contains minimal microbial biomass, but colonization proceeds rapidly following exposure to maternal and environmental microorganisms. Within days, microbial density increases substantially, and colonization follows a structured ecological trajectory rather than a random process [1, 132]. Early communities are initially enriched in facultative anaerobes, which consume oxygen and help create conditions that favor the later expansion of obligate anaerobes as intestinal oxygen tension declines [133]. This transition reflects coordinated ecological succession influenced by host physiology, delivery mode, feeding pattern, medication exposure, and environmental context. Maternal transmission is a major contributor to early microbial exposure. Increasing evidence also supports broader parental microbiome contributions to early neurodevelopment through multi-niche microbial transfer pathways involving maternal gut, vaginal, oral, skin, and milk-associated microbiota [9, 130]. Vaginal delivery exposes infants to maternal vaginal and intestinal microbiota, facilitating early colonization by taxa such as Lactobacillus, Bacteroides, and Prevotella [3, 134]. In contrast, caesarean delivery is associated with reduced exposure to maternal-associated microbes and increased colonization by skin- and hospital-associated taxa, including Staphylococcus and Enterococcus [135]. These differences have been associated with delayed colonization by Bacteroides and altered microbial maturation that may persist beyond the neonatal period [136]. Following initial colonization, infant feeding becomes a dominant determinant of microbial succession. Breastfeeding provides not only macronutrients, but also a complex biological matrix containing live microbes, HMOs, secretory immunoglobulin A (sIgA), antimicrobial proteins, immune mediators, and growth factors that collectively shape microbial composition and function [14, 83]. HMOs support the growth of specialized milk-adapted taxa, particularly bifidobacteria, which contribute to microbial succession and metabolite production [16, 84, 85]. Through fermentation of HMOs, bifidobacteria produce acetate and lactate, contributing to acidification of the intestinal environment, suppression of potential pathogens, and support of epithelial barrier integrity [85]. HMOs may also act as soluble decoy receptors that inhibit pathogen adhesion and may directly modulate host immune responses [86, 87]. sIgA further refines microbial colonization by promoting host–microbe homeostasis and limiting pathogen expansion [88]. Infant formula generally lacks the full structural diversity and biological complexity of human milk, particularly with respect to HMOs and immunological factors. Consequently, formula-fed infants tend to exhibit earlier microbial diversification and a less distinctly milk-adapted microbial profile [27]. These differences are associated with variation in intestinal pH, microbial metabolic output, and immune signalling, indicating that early microbial development is shaped by functional organization rather than diversity alone. Antibiotic exposure represents another major determinant of early colonization. Maternal intrapartum antibiotics and early-life antibiotic use can disrupt microbial succession, reduce colonization by key taxa such as Bifidobacterium and Bacteroides, and alter microbial metabolic activity [4, 89]. These effects may extend beyond microbial composition to influence immune regulation and metabolic signalling. Early-life antibiotic exposure has been associated with increased risk of allergic and metabolic disorders, although causal relationships remain difficult to establish [90]. Breastfeeding may partially support microbial recovery by providing selective substrates and bioactive components, but restoration is often incomplete and context-dependent [28]. Prematurity further amplifies vulnerability during microbial seeding. Preterm infants are exposed to multiple factors that disrupt colonization, including immature intestinal barrier function, reduced gastric acidity, frequent antibiotic exposure, delayed enteral feeding, and prolonged hospitalization. Their microbiota is often dominated by facultative anaerobes and opportunistic taxa, particularly members of Enterobacteriaceae, with reduced abundance of beneficial microbes [91, 92]. This dysbiotic profile is associated with increased inflammatory signalling, impaired barrier integrity, and heightened susceptibility to necrotizing enterocolitis. In this context, human milk provides not only microbial and prebiotic inputs, but also immunological factors, including sIgA and anti-inflammatory cytokines, that support barrier integrity and immune regulation [93]. Studies in preterm infants suggest that exposure to maternal milk is associated with increased abundance of beneficial taxa and reduced prevalence of inflammation-associated organisms, although findings vary according to clinical context [94]. The neonatal intensive care unit (NICU) environment further shapes microbial exposure, often introducing hospital-associated organisms that interact with medical interventions and feeding practices [95]. Beyond local intestinal effects, early microbial colonization may influence systemic physiology through bioactive metabolites, including SCFAs and amino acid derivatives [17, 96]. Their roles in MGBA signalling are discussed in Section “Microbial metabolites and neurodevelopmental effects”. These processes are particularly relevant during infancy, when immune and neurodevelopmental systems remain highly plastic. Breastfeeding may also help buffer disruptions in early colonization. Infants born by caesarean delivery or exposed to antibiotics often exhibit altered microbial profiles, but exclusive breastfeeding can partially restore microbial maturation over time [97]. Similarly, breast milk may support recovery following antibiotic-induced disruption by promoting microbial resilience and reducing the burden of antibiotic resistance genes [98]. These buffering effects highlight the role of human milk not only in shaping initial colonization, but also in maintaining microbial stability under adverse conditions. Taken together, early-life gut colonization reflects the interaction of maternal transmission, feeding practices, medical exposures, and environmental conditions. Within this system, breastfeeding acts as a central modulator by providing selective substrates, microbial inputs, and bioactive signals that shape microbial composition and function. Early microbial development is determined not only by the presence of specific taxa but also by their metabolic capacity and interactions with the host. Collectively, these processes support the concept that early microbial colonization contributes to intestinal and immune maturation and may participate in broader signalling networks involved in gut–brain communication and neurodevelopment. These factors highlight that early colonization is influenced by a combination of biological inputs and environmental conditions, as shown in (Fig. 2).Fig. 2Determinants of early infant gut microbial colonization and potential developmental implications. Maternal, perinatal, dietary, genetic, and environmental factors may influence early gut microbiome establishment. Factors including delivery mode, breastfeeding, maternal health, gestational age, antibiotic exposure, and maternal microbial transmission may be linked to beneficial microbial colonization or dysbiotic patterns, which have been associated with impaired immune maturation and altered microbial diversity. The mechanisms shown are hypothesized mechanisms based on current evidence. Created with BioRender Determinants of early infant gut microbial colonization and potential developmental implications. Maternal, perinatal, dietary, genetic, and environmental factors may influence early gut microbiome establishment. Factors including delivery mode, breastfeeding, maternal health, gestational age, antibiotic exposure, and maternal microbial transmission may be linked to beneficial microbial colonization or dysbiotic patterns, which have been associated with impaired immune maturation and altered microbial diversity. The mechanisms shown are hypothesized mechanisms based on current evidence. Created with BioRender The MGBA: core mechanisms Microbial colonization and immune–barrier programming Early microbial colonization represents a central interface through which human milk may influence infant development within the MGBA. Beyond its nutritional role, human milk contributes to establishment of the neonatal gut microbiome, thereby shaping early microbial ecology, immune programming, and barrier development [137]. Milk-associated microbes and HMOs act together with perinatal and environmental factors to shape early microbial ecology [138]. These early colonization processes establish microbial and immune conditions that may contribute to later gut–brain communication. The establishment of the gut microbiome is also essential for maturation of the intestinal barrier. At birth, the neonatal intestinal barrier is immature and relatively permeable compared with later life. Over time, exposure to commensal microbes supports epithelial maturation, mucus production, tight junction regulation, and immune homeostasis within the intestinal mucosa [139]. Commensal bacteria contribute to epithelial integrity and balanced immune responses, reducing vulnerability to inflammation. In contrast, disrupted microbial colonization may impair barrier development, potentially contributing to prolonged intestinal permeability and inflammatory risk. Microbial influences may extend beyond the gut to the BBB, which is critical for protecting the CNS. Preclinical studies show that absence of gut microbiota is associated with increased BBB permeability and reduced expression of tight junction proteins, including occludin and claudin-5. Importantly, microbial colonization can partially restore these protective functions in experimental models [11]. Microbial metabolites may contribute to intestinal and BBB maturation and to neonatal immune regulation [69, 140]. The specific roles of SCFAs in these processes are discussed in Section “Microbial metabolites and neurodevelopmental effects”. These immune effects are closely connected to brain development because microbial metabolites, cytokines, and other neuroactive compounds can influence neural signalling through microbiota–gut–brain communication pathways [40]. Collectively, these findings support the concept that human milk contributes to early microbial, immune, and barrier development. However, the extent to which these processes directly mediate downstream neurodevelopmental outcomes remains incompletely defined. Neural and neuroendocrine communication pathways A central question in the MGBA is how microbial signals generated in the gut are transmitted to the developing brain. Although the gut–brain axis is often described as bidirectional in broader physiology, the neonatal evidence reviewed here primarily supports gut-to-brain signalling rather than clearly demonstrated brain-to-gut or brain-to-mammary signalling. Therefore, in this review, MGBA mechanisms are interpreted mainly as milk- and gut-derived influences on infant microbial colonization, immune and barrier maturation, microbial metabolite production, vagal signalling, neuroendocrine regulation, and early neurodevelopment. Direct neonatal evidence for reciprocal brain-to-gut or brain-to-mammary regulation remains limited. In early life, the intestine functions as a neuroendocrine interface in which microbial metabolites and molecular signals are detected and relayed through neural and endocrine pathways. These signals are transmitted to the CNS through two closely linked routes: rapid neural signalling, primarily via vagal afferent pathways, and slower neuroendocrine signalling involving gut peptide release and regulation of the hypothalamic–pituitary–adrenal (HPA) axis. Across the evidence considered here, three mechanistic observations support gut-to-brain signalling: loss of microbial effects after vagotomy, exaggerated stress responses in germ-free animals that normalize following microbial reconstitution, and direct microbial metabolite regulation of enteroendocrine hormone production [5, 13]. Vagus nerve signalling The vagus nerve represents the principal neural pathway linking the gut and brain. Approximately 80% of vagal fibers are afferent, transmitting sensory information from visceral organs to central autonomic centers [5]. Vagal afferent terminals do not directly sample the intestinal lumen. Instead, microbial metabolites and luminal signals act on epithelial and enteroendocrine cells, which release mediators such as serotonin and gut peptides that activate receptors on vagal afferent endings [5]. Experimental studies support the functional importance of this pathway. In mice, ingestion of Lactobacillus rhamnosus JB-1 altered central gamma-aminobutyric acid (GABA) receptor expression, reduced anxiety- and depression-related behaviours, and lowered stress-induced corticosterone levels. These neurochemical and behavioural effects were absent after vagotomy, indicating that intact vagal signalling was required for microbial signals to influence the brain [13]. In addition to transmitting sensory information, vagal efferent pathways contribute to intestinal homeostasis by reinforcing epithelial barrier integrity and dampening inflammation, processes that may indirectly support microbial stability within the gut ecosystem [5]. Enteric Nervous System–microbiome interactions Before signals reach the vagus nerve, microbial cues are likely processed locally within the enteric nervous system (ENS), a neural network embedded within the gastrointestinal wall that regulates intestinal motility, secretion, and sensory activity. Increasing evidence suggests that microbial colonization contributes to shaping ENS function. One proposed mechanism involves serotonergic signalling. Microbiota-dependent stimulation of serotonin release from enterochromaffin cells has been shown to promote ENS maturation through activation of 5-hydroxytryptamine 4 (5-HT4) receptors. Additional evidence suggests that microbial modulation of ENS activity may influence brain function through vagal pathways. For example, administration of Bifidobacterium longum NCC3001 produced anxiolytic effects in mice that depended on intact vagal signalling and was associated with altered excitability of enteric neurons [61]. Together, these findings suggest that microbial signals may first be integrated within ENS circuits before being relayed centrally via vagal afferents. HPA axis programming and stress reactivity Microbial colonization also appears to influence neuroendocrine regulation of the stress response. Germ-free mice exhibit exaggerated HPA axis activation following stress exposure, characterized by increased corticosterone secretion [10]. This phenotype can be partially normalized by microbial reconstitution, particularly when colonization occurs early in life, indicating a developmental window during which microbial exposure contributes to stress-axis calibration [10]. Reconstitution with Bifidobacterium infantis, a bacterium commonly associated with the infant gut microbiome, has been shown to attenuate exaggerated stress responses in germ-free animals [10]. Human evidence remains limited but broadly consistent with this concept. In healthy one-month-old infants who were vaginally delivered and exclusively breastfed, gut microbiome diversity was associated with cortisol reactivity during a heel-stick stress test [62]. Although causality cannot be inferred, these findings suggest that early microbial variation within a relatively uniform feeding context may relate to infant stress physiology. Endocrine signalling Microbial metabolites can also influence host physiology through endocrine pathways originating in the intestinal epithelium. Enteroendocrine cells detect luminal microbial signals and translate them into hormonal outputs that act locally, systemically, and through neural circuits. SCFAs produced through microbial fermentation are particularly important in this context. Experimental studies show that propionate and butyrate increase transcription and secretion of peptide YY (PYY) in human enteroendocrine cell models [63]. Mechanistically, this effect appears to be driven largely by histone deacetylase (HDAC) inhibition, although SCFA-responsive receptors such as free fatty acid receptor 2 (FFA2) and free fatty acid receptor 3 (FFA3) also contribute to signal detection [63]. Gut peptides, including PYY and glucagon-like peptide-1 (GLP-1), may subsequently influence brain physiology through circulation and through activation of receptors expressed on vagal afferent neurons [5]. Together, these findings support a model in which microbial signals generated within the gut are detected by epithelial, endocrine, and neural interfaces and relayed to the brain through coordinated vagal, enteric, and neuroendocrine pathways. Within the current evidence base, the most direct brain-facing outcomes include alterations in neurotransmitter-related gene expression and stress-hormone reactivity rather than definitive downstream cellular neurodevelopmental effects. These interactions may be particularly important during early life, when both the microbiome and host regulatory systems remain developmentally plastic. Microbial metabolites and neurodevelopmental effects Microbial metabolites are among the primary mediators through which early-life microbiome activity may influence host development. SCFAs and tryptophan-derived metabolites have been most extensively studied, although their specific roles in neurodevelopment remain incompletely defined. These metabolites regulate host physiology through immune, epithelial, metabolic, and neuroactive pathways, including modulation of regulatory T-cell activity and immune homeostasis [47]. To improve readability and clarify the main metabolite-linked pathways discussed in this section, Table 2 summarizes key microbial metabolites, their principal biological targets, and the primary evidence base supporting their relevance to early-life MGBA signalling. As described in Section “Breastfeeding, milk exposure, and infant gut colonization”, HMO-utilizing bifidobacteria contribute substantially to acetate production during early infancy [141–143]. SCFA profiles also change over time. In predominantly breastfed infants, acetate remains dominant during breastfeeding and increases toward six months alongside Bifidobacteriales-dominant microbiota, whereas propionate and butyrate increase after cessation of breastfeeding with emergence of Clostridiales [19]. This shift may reflect reduced lactoferrin after weaning, increasing luminal iron availability and promoting expansion of butyrate-producing taxa [19].Table 2Key microbial metabolites relevant to early-life MGBA signallingMetabolite GroupRepresentative MetabolitesMain Biological TargetsPrimary Evidence TypeSCFAsAcetate, propionate, butyrateEpithelial barrier integrity, T cell differentiation, BBB function, microglial maturationPreclinical and human observational evidenceIndole derivativesIndole, indole-lactic acid, indole-propionic acidAryl hydrocarbon receptor signalling, epithelial barrier regulation, immune modulationMainly mechanistic and preclinical evidenceKynurenine pathway metabolitesKynurenine, kynurenic acid, quinolinic acidNeuroimmune signalling, glutamatergic pathways, inflammatory regulation, excitotoxicity-related mechanismsMechanistic, preclinical, and limited human observational evidenceSerotonin-related metabolitesSerotonin, tryptophan derived serotonin precursorsEnteroendocrine signalling, vagal pathways, synapse maturation, stress-related signallingPreclinical and limited translational evidenceAbbreviations: BBB: Blood-brain barrier; SCFA: Short-chain fatty acid Table 2Key microbial metabolites relevant to early-life MGBA signallingMetabolite GroupRepresentative MetabolitesMain Biological TargetsPrimary Evidence TypeSCFAsAcetate, propionate, butyrateEpithelial barrier integrity, T cell differentiation, BBB function, microglial maturationPreclinical and human observational evidenceIndole derivativesIndole, indole-lactic acid, indole-propionic acidAryl hydrocarbon receptor signalling, epithelial barrier regulation, immune modulationMainly mechanistic and preclinical evidenceKynurenine pathway metabolitesKynurenine, kynurenic acid, quinolinic acidNeuroimmune signalling, glutamatergic pathways, inflammatory regulation, excitotoxicity-related mechanismsMechanistic, preclinical, and limited human observational evidenceSerotonin-related metabolitesSerotonin, tryptophan derived serotonin precursorsEnteroendocrine signalling, vagal pathways, synapse maturation, stress-related signallingPreclinical and limited translational evidenceAbbreviations: BBB: Blood-brain barrier; SCFA: Short-chain fatty acid Key microbial metabolites relevant to early-life MGBA signalling Abbreviations: BBB: Blood-brain barrier; SCFA: Short-chain fatty acid Acetate modulates macrophage and regulatory T-cell activity, supporting immune homeostasis and potentially contributing to neurodevelopmental regulation [144]. Regulation of microglial maturation represents one plausible mechanism through which SCFAs may influence neurodevelopment. In germ-free mice, SCFA supplementation restored microglial density, morphology, and maturation, including normalization of colony-stimulating factor 1 receptor expression [12]. However, these findings were obtained in adult animals, limiting direct transferability to early-life development. Evidence from neonatal models provides additional support. In neonatal mice, Bifidobacterium colonization increased amoeboid microglia and phagocytic markers, including CD68, CD36, Marco, and Msr1, and normalized microglial maturation by postnatal day 20 [64]. Because these taxa produce acetate through the bifid shunt [145], the findings support a potential link between human milk-shaped microbial metabolism and microglial development. Similarly, neonatal Bifidobacterium colonization suppressed elevated expression of synaptic plasticity-associated genes, including Bdnf and Ngf, prevented excessive synapse formation, and restored synaptic density and neuronal activity by postnatal day 20 in mouse models [64]. However, the specific contribution of SCFAs remains unclear. SCFA signalling alone is unlikely to explain all human milk-mediated neurodevelopmental effects. In a neonatal rat model of preterm brain injury, HMO supplementation restored cortical myelination by increasing myelin basic protein-positive area without increasing oligodendrocyte lineage cell numbers [65]. A prebiotic intervention that induced similar microbial and SCFA changes failed to restore myelination [65], suggesting that additional human milk-specific mechanisms, including sialylated glycans, may be involved [30]. Collectively, these findings indicate that human milk influences early microbial metabolism and SCFA production, which may contribute to immune and neurodevelopmental processes. Experimental studies further suggest that microbiome disruption induced by antibiotics may directly alter neuronal morphology and inhibitory circuit organization, supporting the possibility that dysbiosis may influence structural neurodevelopmental processes [66–68]. However, the extent to which SCFAs directly mediate these effects remains unresolved. Human evidence supports associations among breastfeeding, human milk components, infant gut microbial composition, and stool metabolite profiles [19, 26, 27, 32, 33, 47, 60, 84, 85, 140–143]. In contrast, mechanisms linking SCFAs to BBB maturation, microglial development, synaptic remodelling, and stress-axis programming are based mainly on rodent, germ-free, antibiotic-treated, or neonatal animal models [10–12, 64–69, 144, 145]. Tryptophan–serotonin signalling is also supported largely by neonatal porcine and mechanistic data, with limited direct evidence in human infants [70–73, 146]. These pathways should therefore be interpreted as biologically plausible rather than established causal mechanisms in humans, with key translational gaps related to species differences, dose, timing, and developmental stage. Tryptophan-derived metabolites Tryptophan metabolism represents another pathway through which human milk and the microbiome may influence neurodevelopment [71]. Tryptophan availability in human milk exhibits circadian variation and declines across lactation [147, 148], and reduced availability has been associated with altered neurodevelopmental processes [71]. Tryptophan metabolism occurs through three principal pathways that generate serotonin, kynurenine metabolites, and microbiota-derived indole derivatives. These pathways are regulated by enzymes and signalling systems including indoleamine 2,3-dioxygenase 1 (IDO1), tryptophan hydroxylase 1 (TPH1), and aryl hydrocarbon receptor (AhR) signalling [149]. Approximately 90% of tryptophan is metabolized through the kynurenine pathway [146]. Regulation of tryptophan metabolism links human milk to neurotransmitter systems through the microbiota–gut–brain axis. In a neonatal porcine model, sow feeding maintained TPH1-mediated serotonin synthesis, whereas formula feeding reduced enterochromaffin cell number and serotonin levels [70]. These findings suggest that human milk-shaped, bifidobacteria-enriched microbial ecology may support intestinal tryptophan–serotonin balance. Serotonin regulates dendritic spine formation and synapse maturation through 5-HT2A and 5-HT7 receptor signalling [73]. More broadly, extracellular molecular signalling pathways are increasingly recognized as critical regulators of dendritic architecture and neural circuit maturation during brain development [72, 150, 151]. However, whether human milk-driven alterations in tryptophan metabolism directly regulate synaptogenesis in human infants remains to be established. Overall, available evidence suggests that human milk may influence tryptophan metabolism and downstream neurotransmitter systems through microbiome-dependent pathways, but direct evidence linking these processes to neurodevelopmental outcomes in humans remains limited. To distinguish the major human milk-mediated mechanisms discussed above, Table 3 summarizes microbiome-mediated, HMO-mediated, immune factor-mediated, and nutritional pathways and their proposed relevance to early-life MGBA development.Table 3Conceptual separation of human milk-mediated pathways relevant to early-life MGBA developmentPathway CategoryPrincipal Milk-Related Components or ExposuresProposed Biological MechanismsRelevance to NeurodevelopmentMicrobiome-mediated effectsMilk-related microbes, infant gut microbial colonization, microbial metabolites with SCFAs, and tryptophan derivative metabolites.Direct microbial transfer and microbial succession, plus immune maturation and barrier development, metabolite generation, vagal signalling, HPA axis modulation, and neuroimmune regulation [19, 32, 33, 39, 46, 47, 64–68, 71, 74, 75, 141–145]Biologically plausible and backed by preclinical work. However, direct human evidence tying milk-associated microbes to neurodevelopment is limited.HMO-mediated effectsHuman milk oligosaccharides, including fucosylated and sialylated HMOs.Selective enrichment of Bifidobacterium, microbial metabolic modulation, epithelial and immune regulation, and potential effects on myelination and cognitive outcomes [16, 18, 29, 30, 55, 56, 58, 84–87, 143, 152–155]Backed by human observational data, where certain HMOs are tied to cognitive, language, socioemotional, and even neuroimaging outcomes, although causality remains uncertain.Immune factor-mediated effectssIgA, lactoferrin, antimicrobial proteins, cytokines, and immune cells.Host–microbe homeostasis, pathogen restriction, epithelial barrier support, inflammatory regulation, and gut–brain immune signalling [14, 40, 69, 83, 88, 93, 137–140]Relevant to neurodevelopment through immune maturation and inflammatory regulation, but direct milk immune factor – neurodevelopment links remain unclear.Nutritional effectsLC-PUFAs, amino acids, hormones, peptides, and growth factors.Support of brain growth, neurotransmitter synthesis, myelination, synaptogenesis, and gastrointestinal maturation [70–73, 146–148, 156–161]Supported by broader breastfeeding and preterm-infant literature, but effects are difficult to separate from microbiome, immune, and caregiving-related influences.Abbreviations: HMO, human milk oligosaccharide; HPA, hypothalamic–pituitary–adrenal; LC-PUFA, long-chain polyunsaturated fatty acid; SCFA, short-chain fatty acid; sIgA, secretory immunoglobulin A Table 3Conceptual separation of human milk-mediated pathways relevant to early-life MGBA developmentPathway CategoryPrincipal Milk-Related Components or ExposuresProposed Biological MechanismsRelevance to NeurodevelopmentMicrobiome-mediated effectsMilk-related microbes, infant gut microbial colonization, microbial metabolites with SCFAs, and tryptophan derivative metabolites.Direct microbial transfer and microbial succession, plus immune maturation and barrier development, metabolite generation, vagal signalling, HPA axis modulation, and neuroimmune regulation [19, 32, 33, 39, 46, 47, 64–68, 71, 74, 75, 141–145]Biologically plausible and backed by preclinical work. However, direct human evidence tying milk-associated microbes to neurodevelopment is limited.HMO-mediated effectsHuman milk oligosaccharides, including fucosylated and sialylated HMOs.Selective enrichment of Bifidobacterium, microbial metabolic modulation, epithelial and immune regulation, and potential effects on myelination and cognitive outcomes [16, 18, 29, 30, 55, 56, 58, 84–87, 143, 152–155]Backed by human observational data, where certain HMOs are tied to cognitive, language, socioemotional, and even neuroimaging outcomes, although causality remains uncertain.Immune factor-mediated effectssIgA, lactoferrin, antimicrobial proteins, cytokines, and immune cells.Host–microbe homeostasis, pathogen restriction, epithelial barrier support, inflammatory regulation, and gut–brain immune signalling [14, 40, 69, 83, 88, 93, 137–140]Relevant to neurodevelopment through immune maturation and inflammatory regulation, but direct milk immune factor – neurodevelopment links remain unclear.Nutritional effectsLC-PUFAs, amino acids, hormones, peptides, and growth factors.Support of brain growth, neurotransmitter synthesis, myelination, synaptogenesis, and gastrointestinal maturation [70–73, 146–148, 156–161]Supported by broader breastfeeding and preterm-infant literature, but effects are difficult to separate from microbiome, immune, and caregiving-related influences.Abbreviations: HMO, human milk oligosaccharide; HPA, hypothalamic–pituitary–adrenal; LC-PUFA, long-chain polyunsaturated fatty acid; SCFA, short-chain fatty acid; sIgA, secretory immunoglobulin A Conceptual separation of human milk-mediated pathways relevant to early-life MGBA development Abbreviations: HMO, human milk oligosaccharide; HPA, hypothalamic–pituitary–adrenal; LC-PUFA, long-chain polyunsaturated fatty acid; SCFA, short-chain fatty acid; sIgA, secretory immunoglobulin A Neurodevelopmental and clinical outcomes This section examines human evidence linking breastfeeding, human milk components, and microbiome-related exposures to cognitive, behavioural, and neurodevelopmental outcomes. While important from a clinical perspective, these associations do not support a causal role for the human milk microbiome. To further clarify the strength and type of evidence supporting each proposed mechanism, Table 4 summarizes whether the main pathways are supported by in vitro evidence, animal evidence, human observational evidence, or human interventional evidence.Table 4Evidence level supporting proposed human milk-related mechanismsProposed MechanismIn Vitro EvidenceAnimal EvidenceHuman Observational EvidenceHuman Interventional EvidenceMicrobial colonization and infant gut microbial successionLimitedModerate [10, 13, 33, 39, 47, 61–64, 69]Moderate [20–30, 74, 76, 83–88, 90, 95, 96, 98]Limited [103–105, 113, 117]HMO-mediated bifidobacterial enrichment and microbial metabolismModerate [84–87]Moderate [65, 143]Moderate to strong [16, 18, 29, 30, 55, 56, 58, 153–155]LimitedSCFA-mediated immune, barrier, and neuroactive signalingModerate [63]Moderate [12, 64, 69, 140, 144, 145]Limited to moderate [19, 32, 33, 46, 47, 60, 96, 141–143]LimitedTryptophan-derived metabolite and serotonin-related signallingLimitedModerate [70–73]Limited [71, 146–148, 161, 162]LimitedImmune factor mediated host microbe regulationModerateModerate [40, 69, 137–140]Moderate [14, 83, 88, 93]LimitedNutritional and bioactive milk component effectsLimitedModerate [65, 70–73, 146–148, 161]Moderate [25, 36, 37, 41–43, 48–52, 54, 101, 156]Limited to moderate [103, 104, 106–111, 113–119, 149] Table 4Evidence level supporting proposed human milk-related mechanismsProposed MechanismIn Vitro EvidenceAnimal EvidenceHuman Observational EvidenceHuman Interventional EvidenceMicrobial colonization and infant gut microbial successionLimitedModerate [10, 13, 33, 39, 47, 61–64, 69]Moderate [20–30, 74, 76, 83–88, 90, 95, 96, 98]Limited [103–105, 113, 117]HMO-mediated bifidobacterial enrichment and microbial metabolismModerate [84–87]Moderate [65, 143]Moderate to strong [16, 18, 29, 30, 55, 56, 58, 153–155]LimitedSCFA-mediated immune, barrier, and neuroactive signalingModerate [63]Moderate [12, 64, 69, 140, 144, 145]Limited to moderate [19, 32, 33, 46, 47, 60, 96, 141–143]LimitedTryptophan-derived metabolite and serotonin-related signallingLimitedModerate [70–73]Limited [71, 146–148, 161, 162]LimitedImmune factor mediated host microbe regulationModerateModerate [40, 69, 137–140]Moderate [14, 83, 88, 93]LimitedNutritional and bioactive milk component effectsLimitedModerate [65, 70–73, 146–148, 161]Moderate [25, 36, 37, 41–43, 48–52, 54, 101, 156]Limited to moderate [103, 104, 106–111, 113–119, 149] Evidence level supporting proposed human milk-related mechanisms Breastfeeding, neurodevelopment, and cognitive outcomes Human evidence is largely observational, linking breastfeeding and infant feeding patterns to later cognitive or behavioural outcomes [48]. Although breastfeeding is associated with modest cognitive benefits, the strength and independence of this relationship remain uncertain because of residual confounding [36]. Prospective cohort studies following children from infancy into school age have shown that longer breastfeeding duration is associated with better performance on standardized cognitive assessments, including measures of memory, language, and problem-solving ability. A systematic review with network meta-analysis found that breastfed children had intelligence-test scores approximately 4% higher than those of children who had never been breastfed (ratio of means, 1.04; 95% confidence interval [CI], 1.02–1.06) [34]. Longer breastfeeding durations were generally ranked more favourably, although the observational nature of the included studies limits causal interpretation. However, across cohorts and meta-analyses, the reported cognitive advantage is typically small and is partly explained by maternal, familial, and socioeconomic factors [37]. Breastfeeding duration and exclusivity may also influence neurodevelopmental outcomes. Systematic reviews suggest that exclusive breastfeeding, particularly during the first six months of life, is associated with more favourable neurodevelopmental outcomes compared with partial or no breastfeeding [50]. A dose–response relationship has been observed in several studies, with longer breastfeeding duration correlating with higher cognitive scores [41]. A 2024 systematic review and meta-analysis of four observational studies involving 1,783 infants found that any breastfeeding, compared with no breastfeeding, was associated with higher cognitive scores in preterm infants (standardized mean difference [SMD], −0.19; 95% CI, −0.31 to −0.07; p = 0.002; heterogeneity [I2] = 44%) [163]. The negative direction of the SMD reflected the comparison coding used by the authors and favoured the breastfed group. However, the certainty of evidence was rated very low because the included studies were susceptible to confounding. In the same meta-analysis, two observational studies involving 1,133 infants found that never-breastfed infants had a higher risk of neurodevelopmental impairment than infants who received any breastfeeding (relative risk [RR], 1.17; 95% CI, 1.07–1.29; p = 0.0008; I2 = 0%). In contrast, when predominant formula feeding was compared with predominant breastfeeding, the pooled difference in cognitive scores across 10 observational studies involving 1,903 participants was not statistically significant (SMD, 0.17; 95% CI, −0.02 to 0.36; p = 0.08; I2 = 68%), indicating substantial heterogeneity. Heterogeneity may reflect differences in breastfeeding definitions, exposure levels, assessment timing, gestational age, and outcome measures. Associations may be stronger in high-risk populations, but interpretation remains limited by confounding from maternal, socioeconomic, and home-environment factors [34]. Studies that carefully adjust for these variables often report attenuation of the association, and in some cases the effect becomes non-significant [43]. Sibling-comparison designs, which partially control for shared genetic and environmental factors, also tend to show smaller effect sizes. These findings suggest that part of the observed cognitive advantage may reflect underlying family characteristics rather than breastfeeding alone. However, several studies still report a residual association after adjustment, indicating that breastfeeding may contribute independently, although modestly, to cognitive development. Beyond global measures such as IQ, longitudinal studies have examined specific cognitive domains. Breastfeeding has been linked to enhanced performance in language-related abilities, including verbal fluency and reading comprehension, as well as executive function domains such as attentional control and working memory [38]. Several cohort studies report that longer breastfeeding duration corresponds to higher scores in language-based assessments and academic achievement extending into adolescence [35]. Neuroimaging studies provide additional support by showing associations between breastfeeding exposure and structural brain differences, including increased cortical thickness and more organized white matter tracts in frontal and temporal regions involved in language processing, executive regulation, and higher-order cognition [45]. Overall, breastfeeding and early microbial exposures are associated with modest differences in cognitive, behavioral, and preterm neurodevelopmental outcomes. However, the extent to which these associations are directly mediated by the infant microbiome remains uncertain. Addressing this question will require longitudinal human studies integrating microbial, metabolic, immune, neuroimaging, and neurodevelopmental phenotyping [42, 51, 52]. Breastfeeding and neurodevelopmental conditions: ASD and ADHD Beyond cognition, breastfeeding has been examined in relation to ASD-related traits and ADHD-related behaviors. Evidence in these areas is less consistent than for general cognitive outcomes and is particularly vulnerable to confounding, reverse causation, and heterogeneity in outcome definitions. A meta-analysis of seven studies, including 1,463 children with ASD and 1,180 children without ASD, found that ASD was associated with lower odds of having been breastfed (pooled OR, 0.61; 95% CI, 0.45–0.83; p = 0.002) [164]. However, the pooled association was derived from observational studies and may be influenced by residual confounding, variation in breastfeeding definitions, and reverse causation. A multicenter Spanish birth-cohort study reported that longer breastfeeding duration was associated with lower autistic trait scores after adjustment for multiple confounders, although effect sizes were small [53]. However, these findings do not establish causality. Breastfeeding duration is strongly patterned by maternal IQ, smoking during pregnancy, parental education, social class, maternal mental health, and the home environment, all of which may also influence developmental outcomes [165–169]. Furthermore, reverse causation is possible because early infant behavior, feeding difficulties, or emerging neurodevelopmental differences may affect breastfeeding initiation and duration [170, 171]. Accordingly, breastfeeding should not be interpreted as preventing ASD, although it may be associated with developmental conditions or traits through biological and socio-environmental pathways. Evidence relating breastfeeding to ADHD is similarly mixed. Some prospective studies have suggested reduced ADHD-related behaviors, hyperactivity, or broader externalizing problems with longer breastfeeding duration, whereas others have not replicated these findings [172–177]. Differences in breastfeeding definitions, exposure duration, outcome measures, and confounder adjustment likely contribute to inconsistent findings. Current evidence supports, at most, a modest association between breastfeeding and some ADHD-related behaviours rather than a clear protective effect against ADHD as a clinical disorder. Several biological pathways may help explain these associations. Breast milk contains LC-PUFAs, including docosahexaenoic acid and arachidonic acid, which are important for brain development and have been linked to cognitive and behavioural outcomes [156–158]. Breastfeeding also involves close mother–infant physical and social interaction, which may influence oxytocin signalling, stress regulation, and attachment-related developmental pathways [159, 160]. Within the framework of this review, breastfeeding is also relevant because it shapes the early gut ecosystem, which may influence neurodevelopment through immune, metabolic, and gut–brain signalling pathways. However, current evidence does not demonstrate that breastfeeding prevents ASD or ADHD, nor does it establish that observed associations are microbiome mediated. The key clinical studies supporting these breastfeeding-neurodevelopment associations are summarized in Table 5.Table 5Selected clinical studies linking breastfeeding, human milk components, microbiome-related markers, and neurodevelopmental outcomesStudy DesignSample SizeMain Effect EstimateAdjustmentInterpretationReferenceProspective cohort study from the French PELAGIE birth cohort286 mother-child pairs assessed at age 6Breastfeeding for >4 months was associated with higher verbal comprehension scores at age 6 (β = 4.95; 95% CI, 0.54 9.37)The analysis adjusted for maternal IQ, education, home environment, socioeconomic deprivation, birth factors, parity, delivery mode, and maternal employmentLonger breastfeeding was associated with improved verbal and selected neurodevelopmen tal outcomes, although causality cannot be inferred[48]Longitudinal cohort study using Growing Up in Australia dataN = 8643 for language skills, N = 8560 for non verbal IQ, and N = 6213 for executive function analysesEach additional month of breastfeeding was associated with higher language scores (β = 0.05; 95% CI, 0.03–0.08) and higher non verbal intelligence scores (β = 0.02; 95% CI, 0.01–0.04), but not executive function at age 15Models adjusted for child sex, birth factors, age at testing, maternal pregnancy exposures, postnatal depression, parental education, and parental age at deliveryLonger breastfeeding was associated with better language and non-verbal cognitive outcomes, although causality cannot be established[37]Retrospective cohort study82 mother–infant pairs, including 59 typically developing children and 23 children at risk for neurodevelopmen tal delay, using human milk samples collected before 9 months of ageMetabolomic and exposomic profiling identified 453 metabolites and 61 environmental chemicals, and predictive models showed diagnostic accuracy of 0.81 for females and 0.79 for malesThe study matched controls to at-risk infants by storage time, sex, and age at milk collection, then used machine-learning and pathway analyses to identify predictive biomarkersHuman milk metabolic and exposure profiles may predict later neurodevelopmen tal delay before symptoms emerge, but larger validation studies are needed[57]Prospective population-based birth cohort study500 children from two Spanish cohorts, assessed at age 4 yearsLonger breastfeeding was associated with better executive function and fewer ADHD symptoms, including RR = 0.56; 95% CI, 0.37 0.85 for attention deficit hyperactivity symptoms after > 12 weeks of breastfeedingModels adjusted for parental socioeconomic and educational factors, maternal characteristics, child sex, evaluator, and key birth-related factors for cognitive outcomesLonger breastfeeding was associated with fewer attention hyperactivity symptoms and better social and executive functioning at age 4, although causality remains uncertain[172]Large cluster randomized trial follow-up from the PROBIT study17,046 breastfeeding mother–infant pairs originally enrolled and 13,889 children followed at age 6.5 yearsThe breastfeeding promotion intervention increased exclusive breastfeeding at 3 months (43.3% vs 6.4%) but showed no significant effects on parent- or teacher-rated child behavior scoresAnalyses used intention-to-treat multilevel models accounting for clustering, with findings robust to adjustment for demographic and parental factorsBreastfeeding promotion had no clear effect on child behavior or maternal adjustment at 6.5 years, but increased breastfeeding of the next-born child[173]Prospective pregnancy cohort study from the Western Australian Raine Study2900 pregnant women and following children’s mental health outcomes to age 14Breastfeeding for<6 months versus≥6 monthspredicted moremental healthproblems, withhigher total CBCLscores (EE = 1.45;95% CI, 0.59–2.30)and higher odds ofCBCL morbidity(OR = 1.33; 95% CI,1.09–1.62)Models adjusted for maternal demographics, prenatal exposures, socioeconomic factors, birth weight adequacy, child sex, and postnatal depressionShorter breastfeeding was associated with poorer mental health through age 14, although residual confounding remains possible[174]Prospective UK Millennium Cohort Study analysis10,037 mother child pairs, including 9,525 term and 512 preterm children assessed at age 5In term children, breastfeeding for ≥4 months was associated with lower odds of abnormal total SDQ behaviour scores compared with never breastfeeding (fully adjusted OR = 0.67; 95% CI, 0.54–0.83)Models adjusted for socioeconomic, maternal, neonatal, attachment, birth order, childcare, and relevant prenatal exposure factorsLonger breastfeeding was associated with fewer parent-rated behavioural problems at age 5 among term children, but evidence for preterm children was less precise[175]Abbreviation: ADHD = Attention-deficit/hyperactivity disorder; β = beta coefficient / regression coefficient; CBCL = Child Behaviour Checklist; CI = confidence interval; EE = estimate of effect; HOME score = Home Observation for Measurement of the Environment score; IQ = intelligence quotient; N = sample size / number of participants; OR = odds ratio; PROBIT = Promotion of Breastfeeding Intervention Trial; RR = relative risk; SDQ = Strengths and Difficulties Questionnaire Table 5Selected clinical studies linking breastfeeding, human milk components, microbiome-related markers, and neurodevelopmental outcomesStudy DesignSample SizeMain Effect EstimateAdjustmentInterpretationReferenceProspective cohort study from the French PELAGIE birth cohort286 mother-child pairs assessed at age 6Breastfeeding for >4 months was associated with higher verbal comprehension scores at age 6 (β = 4.95; 95% CI, 0.54 9.37)The analysis adjusted for maternal IQ, education, home environment, socioeconomic deprivation, birth factors, parity, delivery mode, and maternal employmentLonger breastfeeding was associated with improved verbal and selected neurodevelopmen tal outcomes, although causality cannot be inferred[48]Longitudinal cohort study using Growing Up in Australia dataN = 8643 for language skills, N = 8560 for non verbal IQ, and N = 6213 for executive function analysesEach additional month of breastfeeding was associated with higher language scores (β = 0.05; 95% CI, 0.03–0.08) and higher non verbal intelligence scores (β = 0.02; 95% CI, 0.01–0.04), but not executive function at age 15Models adjusted for child sex, birth factors, age at testing, maternal pregnancy exposures, postnatal depression, parental education, and parental age at deliveryLonger breastfeeding was associated with better language and non-verbal cognitive outcomes, although causality cannot be established[37]Retrospective cohort study82 mother–infant pairs, including 59 typically developing children and 23 children at risk for neurodevelopmen tal delay, using human milk samples collected before 9 months of ageMetabolomic and exposomic profiling identified 453 metabolites and 61 environmental chemicals, and predictive models showed diagnostic accuracy of 0.81 for females and 0.79 for malesThe study matched controls to at-risk infants by storage time, sex, and age at milk collection, then used machine-learning and pathway analyses to identify predictive biomarkersHuman milk metabolic and exposure profiles may predict later neurodevelopmen tal delay before symptoms emerge, but larger validation studies are needed[57]Prospective population-based birth cohort study500 children from two Spanish cohorts, assessed at age 4 yearsLonger breastfeeding was associated with better executive function and fewer ADHD symptoms, including RR = 0.56; 95% CI, 0.37 0.85 for attention deficit hyperactivity symptoms after > 12 weeks of breastfeedingModels adjusted for parental socioeconomic and educational factors, maternal characteristics, child sex, evaluator, and key birth-related factors for cognitive outcomesLonger breastfeeding was associated with fewer attention hyperactivity symptoms and better social and executive functioning at age 4, although causality remains uncertain[172]Large cluster randomized trial follow-up from the PROBIT study17,046 breastfeeding mother–infant pairs originally enrolled and 13,889 children followed at age 6.5 yearsThe breastfeeding promotion intervention increased exclusive breastfeeding at 3 months (43.3% vs 6.4%) but showed no significant effects on parent- or teacher-rated child behavior scoresAnalyses used intention-to-treat multilevel models accounting for clustering, with findings robust to adjustment for demographic and parental factorsBreastfeeding promotion had no clear effect on child behavior or maternal adjustment at 6.5 years, but increased breastfeeding of the next-born child[173]Prospective pregnancy cohort study from the Western Australian Raine Study2900 pregnant women and following children’s mental health outcomes to age 14Breastfeeding for<6 months versus≥6 monthspredicted moremental healthproblems, withhigher total CBCLscores (EE = 1.45;95% CI, 0.59–2.30)and higher odds ofCBCL morbidity(OR = 1.33; 95% CI,1.09–1.62)Models adjusted for maternal demographics, prenatal exposures, socioeconomic factors, birth weight adequacy, child sex, and postnatal depressionShorter breastfeeding was associated with poorer mental health through age 14, although residual confounding remains possible[174]Prospective UK Millennium Cohort Study analysis10,037 mother child pairs, including 9,525 term and 512 preterm children assessed at age 5In term children, breastfeeding for ≥4 months was associated with lower odds of abnormal total SDQ behaviour scores compared with never breastfeeding (fully adjusted OR = 0.67; 95% CI, 0.54–0.83)Models adjusted for socioeconomic, maternal, neonatal, attachment, birth order, childcare, and relevant prenatal exposure factorsLonger breastfeeding was associated with fewer parent-rated behavioural problems at age 5 among term children, but evidence for preterm children was less precise[175]Abbreviation: ADHD = Attention-deficit/hyperactivity disorder; β = beta coefficient / regression coefficient; CBCL = Child Behaviour Checklist; CI = confidence interval; EE = estimate of effect; HOME score = Home Observation for Measurement of the Environment score; IQ = intelligence quotient; N = sample size / number of participants; OR = odds ratio; PROBIT = Promotion of Breastfeeding Intervention Trial; RR = relative risk; SDQ = Strengths and Difficulties Questionnaire Selected clinical studies linking breastfeeding, human milk components, microbiome-related markers, and neurodevelopmental outcomes Breastfeeding for <6 months versus ≥6 months predicted more mental health problems, with higher total CBCL scores (EE = 1.45; 95% CI, 0.59–2.30) and higher odds of CBCL morbidity (OR = 1.33; 95% CI, 1.09–1.62) Abbreviation: ADHD = Attention-deficit/hyperactivity disorder; β = beta coefficient / regression coefficient; CBCL = Child Behaviour Checklist; CI = confidence interval; EE = estimate of effect; HOME score = Home Observation for Measurement of the Environment score; IQ = intelligence quotient; N = sample size / number of participants; OR = odds ratio; PROBIT = Promotion of Breastfeeding Intervention Trial; RR = relative risk; SDQ = Strengths and Difficulties Questionnaire Breastfeeding, the gut–brain axis, and preterm neurodevelopment Preterm infants, particularly those with very low birth weight (VLBW), are at high risk of disrupted early microbial colonization because of intestinal immaturity, intensive medical exposures, and prolonged hospitalization. Their early microbial development is shaped by delayed enteral feeding, antibiotic exposure, NICU care, and interruption of direct maternal milk feeding [178–182]. The preterm population is therefore especially important for studying how human milk may influence neurodevelopment through MGBA mechanisms, as shown in (Fig. 3). Human milk provides microbial, prebiotic, and immune components that may support gut colonization, barrier development, and gut–brain communication in preterm infants [26, 74]. Maternal milk may also serve as a direct microbial source for VLBW infants, with studies reporting microbial sharing between milk and infant stool and greater likelihood of maternal microbial transfer among breastfed infants [183–185].Fig. 3Human milk exposure and gut–brain axis signaling in VLBW preterm infants. In very-low-birth-weight (VLBW) preterm infants, human milk may support beneficial microbial colonization, short-chain fatty acid production, immune regulation, and blood–brain barrier integrity, potentially supporting gut–brain axis signalling and neurodevelopment. In contrast, limited human milk exposure, formula feeding, and NICU-related factors may be associated with dysbiosis, inflammation, barrier disruption, and neurodevelopmental vulnerability. These pathways represent hypothesized mechanisms based on current evidence and do not establish direct causality. Created with BioRender Human milk exposure and gut–brain axis signaling in VLBW preterm infants. In very-low-birth-weight (VLBW) preterm infants, human milk may support beneficial microbial colonization, short-chain fatty acid production, immune regulation, and blood–brain barrier integrity, potentially supporting gut–brain axis signalling and neurodevelopment. In contrast, limited human milk exposure, formula feeding, and NICU-related factors may be associated with dysbiosis, inflammation, barrier disruption, and neurodevelopmental vulnerability. These pathways represent hypothesized mechanisms based on current evidence and do not establish direct causality. Created with BioRender Formula-fed preterm infants often exhibit microbial profiles characterized by lower abundance of Bifidobacterium and Lactobacillus and higher abundance of Firmicutes and Proteobacteria, including Staphylococcus, Klebsiella, Enterobacter, Enterococcus, and other Enterobacteriaceae members [186, 187]. Despite greater taxonomic diversity, this pattern may be functionally less favorable because it is associated with altered microbial metabolism and a more inflammation-prone intestinal environment [188]. In preterm infants, intestinal dysbiosis may further increase intestinal permeability and alter cytokine production, providing a plausible pathway through which microbial imbalance could contribute to systemic inflammation and downstream neurodevelopmental vulnerability [187, 189]. In preterm infants, human milk should therefore be viewed not only as nutrition but also as a biologically active exposure that may shape microbial, immune, and neurodevelopmental trajectories. Human milk contains LC-PUFAs, including arachidonic acid and docosahexaenoic acid, which support cognitive development, physical growth, and visual function in preterm infants [190–192]. It also contains hormones, peptides, amino acids, and growth factors that support gastrointestinal maturation and host defence [193–197]. Together, the nutritional, immunological, microbial, and metabolic properties of human milk provide a biologically plausible framework linking breastfeeding to preterm neurodevelopment. Clinical evidence generally suggests that human milk exposure benefits neurodevelopment in preterm infants, although findings are not uniform. A nonrandomized study of former preterm infants found that those who received human milk during hospitalization had higher IQ scores at 7.5 to 8 years than infants who received other milk types [99]. A meta-analysis included 20 eligible studies, of which 11 reported both unadjusted and covariate-adjusted estimates [102]. Breastfed children showed an unadjusted cognitive advantage of 5.32 points compared with formula-fed children (95% CI, 4.51–6.14). After adjustment for at least five covariates, the pooled difference was attenuated but remained statistically significant at 3.16 points (95% CI, 2.35–3.98). The adjusted estimate was reported as statistically homogeneous, whereas the unadjusted estimate showed greater between-study heterogeneity. The cognitive advantage was larger among low-birth-weight infants than among normal-birth-weight infants; however, residual confounding remained an important limitation. Not all studies have demonstrated clear benefit; for example, breastfeeding showed no significant neurodevelopmental advantage for VLBW infants at 6 and 12 months after adjustment for sociodemographic and infant variables [100]. Overall, the evidence suggests potential benefit, particularly in vulnerable preterm populations, but results vary across cohorts, exposure definitions, and outcome measures. Breastfeeding may support preterm neurodevelopment by promoting a more favourable gut microbial balance, protecting against dysbiosis, and regulating inflammatory and metabolic signalling along the gut–brain axis [26, 74, 101, 186, 188]. However, specific microbiome changes have not yet been causally linked to defined neurodevelopmental outcomes. Current evidence supports human milk as a biologically beneficial exposure for preterm infants, but further studies are needed to establish direct relationships among milk composition, microbial transfer, metabolite production, and long-term neurodevelopment. Translational opportunities for prevention and intervention From mechanism to translation: defining the clinical question The growing literature on the human milk microbiome supports a biologically plausible role for early microbial exposures in shaping infant neurodevelopment through the gut–brain axis, including effects on microbial colonization, immune programming, barrier maturation, and metabolite production. However, translation from mechanistic insight to clinical application remains limited. The central translational question is not simply whether the human milk microbiome is biologically relevant, but which microbiome-associated processes are modifiable, safe to target, and clinically meaningful within defined developmental windows and patient populations. This issue is especially important in preterm and medically vulnerable populations, where disrupted early colonization, antibiotic exposure, and altered milk feeding patterns may reshape the developing gut ecosystem. Recent intervention studies show microbiome-level effects, but direct neurodevelopmental benefit remains unproven [103, 104]. Accordingly, translational interpretation requires clear distinction among mechanistic plausibility, microbiome-level intervention effects, and demonstrated clinical or neurodevelopmental benefit. Evidence for the latter remains limited [105]. Mechanistic and preclinical findings should therefore not be interpreted as equivalent to clinical efficacy. Maternal-targeted interventions during pregnancy or lactation Maternal-targeted strategies, including dietary modification, prebiotics, probiotics, and stress-reduction interventions, represent upstream approaches for modifying the intrauterine and early-life microbial environments, as shown in (Fig. 4). Although biologically plausible, evidence supporting direct neurodevelopmental benefit remains limited and is predominantly observational or preclinical. Maternal dietary patterns have been associated with offspring neurodevelopmental outcomes. Western dietary patterns have been linked to increased risk of neurodevelopmental disorders and reduced gut microbiota diversity, whereas Mediterranean-style diets are associated with enrichment of SCFA-producing taxa [106, 107]. However, these associations are vulnerable to confounding by socioeconomic, metabolic, and environmental factors, and causal inference remains uncertain. For example, a study using three large independent mother–child cohorts reported associations between maternal Western diet and offspring neurodevelopmental disorders, particularly ADHD and ASD, suggesting that maternal nutrition may represent a modifiable factor influencing offspring development [107]. Dietary intake of the mother could affect the microbial composition and metabolism, which would have downstream impacts on the maternal immune and metabolic pathways; nevertheless, associations with fetal neurodevelopment have not been elucidated fully. Preclinical evidence supports the notion that maternal intake of prebiotics could alter the microbial composition and reduce maternal immune activation. However, these findings are derived primarily from animal models, and clinical evidence demonstrating reduced neurodevelopmental disorder risk in humans is currently lacking.Fig. 4Maternal-targeted interventions and modulation of breast milk bioactives. Maternal dietary patterns, prebiotics, probiotics, and stress-related interventions during pregnancy or lactation may influence breast milk bioactive composition, including human milk oligosaccharides, microbial communities, metabolites, antibodies, and immune cells. These factors may be associated with infant gut microbiome development, immune maturation, and early-life gut–brain axis signalling. The proposed pathways are hypothesized mechanisms based on current evidence. Created with BioRender Maternal-targeted interventions and modulation of breast milk bioactives. Maternal dietary patterns, prebiotics, probiotics, and stress-related interventions during pregnancy or lactation may influence breast milk bioactive composition, including human milk oligosaccharides, microbial communities, metabolites, antibodies, and immune cells. These factors may be associated with infant gut microbiome development, immune maturation, and early-life gut–brain axis signalling. The proposed pathways are hypothesized mechanisms based on current evidence. Created with BioRender Maternal probiotic supplementation during pregnancy and lactation is generally considered safe and can influence neonatal microbial colonization, particularly following caesarean delivery or antibiotic exposure [108]. While microbiome-level effects, including restoration of Bifidobacterium, have been observed, evidence for downstream neurodevelopmental benefit remains indirect. Preclinical studies demonstrate modulation of neuroinflammatory pathways involving interleukin-1 beta and interleukin-6, but translation to human neurodevelopmental outcomes has not been established [109]. Maternal stress is another potential upstream target. Fetal programming refers to the influence of the prenatal environment on fetal development, particularly during sensitive developmental windows. Maternal stress may affect fetal programming through activation of the HPA axis and increased glucocorticoid exposure. Experimental evidence also indicates that early-life stress can induce persistent neuroanatomical alterations in cortical inhibitory circuitry, supporting biological links between stress exposure and altered neurodevelopmental trajectories [110, 111]. These changes have been associated with alterations in fetal brain development, including amygdala structure and cortical maturation [112]. While stress-reduction interventions improve maternal psychological outcomes, evidence linking these interventions to sustained neurodevelopmental benefit in offspring remains limited. Bidirectional interactions between the maternal microbiome and stress pathways may further complicate these relationships, as shown in (Fig. 5). Antibiotic exposure and stewardship are also translationally relevant. Observational studies suggest associations between prenatal antibiotic exposure and increased risk of ASD and ADHD, although effect sizes are modest and heterogeneity is substantial. A systematic review and meta-analysis of 15 studies found that prenatal antibiotic exposure was associated with a modest increase in the risk of ADHD. The ADHD analysis included four studies and yielded a pooled OR of 1.14 (95% CI, 1.13–1.15), with no detected heterogeneity (I2 = 0%) [119]. In contrast, the pooled analysis of six studies found no statistically significant overall association with ASD (OR, 1.09; 95% CI, 0.88–1.31), with substantial heterogeneity (I2 = 78.9%) [119]. The association with ASD was statistically significant only in the prospective-cohort subgroup (OR, 1.17; 95% CI, 1.03–1.31; I2 = 48.1%), suggesting that the findings were sensitive to study design. These associations should be interpreted cautiously because maternal infection, the clinical indication for antibiotic treatment, antibiotic class, and timing of exposure, may confound the observed relationships. Therefore, the findings do not establish that prenatal antibiotic exposure causes ADHD or ASD.Fig. 5Maternal stress, lactation change, and fetal programming. Proposed pathways through which maternal stress may influence lactation, breast milk production, and fetal neurodevelopment. Stress-related activation of the hypothalamic–pituitary–adrenal (HPA) axis may affect insulin sensitivity, glucocorticoid signaling, and oxytocin release, potentially influencing milk synthesis and ejection. Maternal stress may also contribute to fetal programming through neuroendocrine and microbiome-related pathways with potential implications for fetal brain development, amygdala structure, and cortical maturation. These relationships represent hypothesized mechanisms based on current evidence. Created with BioRender Maternal stress, lactation change, and fetal programming. Proposed pathways through which maternal stress may influence lactation, breast milk production, and fetal neurodevelopment. Stress-related activation of the hypothalamic–pituitary–adrenal (HPA) axis may affect insulin sensitivity, glucocorticoid signaling, and oxytocin release, potentially influencing milk synthesis and ejection. Maternal stress may also contribute to fetal programming through neuroendocrine and microbiome-related pathways with potential implications for fetal brain development, amygdala structure, and cortical maturation. These relationships represent hypothesized mechanisms based on current evidence. Created with BioRender Infant-targeted interventions: probiotics, synbiotics, and nutritional modulation Infant-targeted interventions represent the most direct approach to modifying early microbiome development. Evidence is strongest in preterm infants, where dysbiosis is common because of antibiotic exposure, delayed feeding, immature barrier function, and environmental factors. Probiotics are the most developed of these approaches and have been shown to increase Bifidobacterium dominance, promote microbiome maturation, reduce intestinal inflammation, and suppress multidrug-resistant organisms in some studies [103, 104]. These microbiome-level effects may influence immune, barrier, and metabolic pathways within the MGBA, but their neurodevelopmental and broader clinical significance remains uncertain. In the Priming Immunity at the Beginning of Life trial, microbiome improvements did not reduce the primary clinical endpoint [113]. Importantly, microbiome modulation does not inherently translate into functional neurodevelopmental benefit. Demonstrating clinical value requires evidence of meaningful effects on validated neurodevelopmental outcomes, not only microbial compositional change [114, 115]. Donor milk and milk processing as translational interfaces Donor human milk is one of the most practical ways to apply human milk research in neonatal care, especially when a mother’s own milk is unavailable for preterm or high-risk infants. Its importance lies not only in nutritional provision, but also in its potential to preserve aspects of the milk-related biological environment that support early gut colonization and may influence gut–brain axis development more favorably than formula. However, donor milk differs from fresh maternal milk because of processing requirements, particularly pasteurization, which reduces microbial load and alters bioactive molecules such as immunoglobulins and enzymes [105]. This requirement creates a trade-off, because procedures that reduce infection risk may also diminish some of the biologically active properties of human milk. Milk bank data also show that donor human milk use depends heavily on safety screening, contamination control, and handling procedures, meaning that its clinical value is shaped by implementation as well as biology [116]. Pooling milk from multiple donors may reduce some variation, but important differences in nutrients and bioactive factors remain, potentially leading to less consistent effects across infants [117]. Thus, the translational value of donor human milk lies not in replicating maternal milk, but in optimizing safe retention of functional bioactivity relevant to early microbiome development. NICU relevance and vulnerable infant populations The NICU represents the setting of highest translational relevance. Preterm infants exhibit physiological vulnerabilities, including immature immune systems, increased intestinal permeability, and underdeveloped microbial ecosystems, which increase susceptibility to dysbiosis and inflammatory dysregulation. For this reason, preterm infants are not only a high-risk group, but also the population in which microbiome-based prevention may be most clinically relevant. As discussed in Section “Infant-targeted interventions: probiotics, Synbiotics, and nutritional modulation”, microbiome-targeted interventions can modify microbial composition and inflammatory profiles in very preterm infants [103, 104]. These observations support the hypothesis that stabilizing early microbial ecology may influence downstream developmental processes, although direct neurodevelopmental benefit has not been consistently demonstrated. This distinction is important: improving microbial composition may reduce gut-related inflammatory stress, but current evidence remains insufficient to conclude that NICU microbiome interventions improve neurodevelopmental outcomes. Overall, the NICU should be viewed as the setting where microbiome-targeted interventions are most biologically plausible and clinically needed, but not yet as a setting where neurodevelopmental benefit has been clearly proven. Personalizing interventions based on microbial profile Personalized microbiome-based interventions remain a longer-term translational objective. Substantial inter-individual variability in human milk microbiota, infant gut composition, and microbial metabolic output suggests that uniform interventions may not be optimal. Emerging cohort evidence indicates that microbial metabolites, including SCFAs, may mediate associations between early exposures and neurodevelopmental trajectories [60]. However, these findings are largely observational, microbial biomarkers lack standardization, and predictive validity remains uncertain. Accordingly, precision microbiome strategies should currently be viewed as investigational, with future translation dependent on reproducible biomarker identification and demonstration of clinical utility. Translational challenges: heterogeneity, timing, safety, and overinterpretation Translation of microbiome-based interventions is limited by several factors. Study heterogeneity reduces generalizability, while timing remains critical because the early-life microbiome changes rapidly during the first days and weeks after birth [198]. An intervention may have different effects depending on when it begins, how long it continues, and whether it occurs during a sensitive developmental window for immune or brain development [60, 113]. Safety considerations are particularly important in preterm infants, including risks related to product variability, contamination, and regulatory oversight [105]. Importantly, the field remains at risk of overinterpretation. Direct causal pathways linking intervention, microbiome modulation, gut–brain axis signalling, and neurodevelopment remain largely unproven. Current translational position Overall, the translational potential of the human milk microbiome currently lies in informing early-life biological support strategies rather than serving as a direct therapeutic target for neurodevelopmental outcomes. Modulation of microbial composition, immune signalling, and metabolic pathways is biologically plausible and increasingly measurable, but consistent improvement in neurodevelopmental outcomes has not been demonstrated. Future progress will depend on rigorously designed, mechanistically informed clinical trials that integrate microbiome function with validated neurodevelopmental endpoints. Recent translational reviews similarly emphasize that successful microbiome-based interventions for paediatric neurodevelopmental disorders will require integration of microbial, immune, metabolic, and neurodevelopmental biomarkers within longitudinal clinical frameworks [8, 9, 199]. Until such evidence is available, microbiome-targeted interventions should be interpreted as biologically plausible but clinically unproven strategies for modifying infant neurodevelopment. Biomarkers, clinical readouts, and research tools Human milk is a dynamic biofluid whose composition varies according to maternal, environmental, and lactational factors, making it a measurable source of early-life exposure for biomarker discovery [200]. A biomarker is an objectively measurable feature that reflects biological processes or responses and must demonstrate reproducibility, predictive value, and clinical utility before adoption [201, 202]. Many neurodevelopmental disorders are diagnosed only at 2–3 years of age, limiting opportunities for early intervention despite evidence that earlier detection improves outcomes [57]. Advances in omics technologies, including metabolomics, proteomics, lipidomics, and glycomics, now enable system-wide characterization of human milk composition and support identification of candidate biomarkers associated with neurodevelopmental outcomes [203, 204]. Human milk-derived candidate biomarkers may therefore support earlier neurodevelopmental risk stratification and inform intervention studies, as shown in (Fig. 6).Fig. 6Human milk metabolomics workflow for biomarker discovery and validation. Schematic overview of the human milk metabolomics pipeline, including sample collection and processing, metabolite extraction, data acquisition, statistical and bioinformatic analyses, identification of differential metabolites, and validation of candidate biomarkers with potential relevance to early neurodevelopmental risk assessment. These biomarker relationships are hypothesized based on current evidence. Created with BioRender Human milk metabolomics workflow for biomarker discovery and validation. Schematic overview of the human milk metabolomics pipeline, including sample collection and processing, metabolite extraction, data acquisition, statistical and bioinformatic analyses, identification of differential metabolites, and validation of candidate biomarkers with potential relevance to early neurodevelopmental risk assessment. These biomarker relationships are hypothesized based on current evidence. Created with BioRender Human milk metabolome The human milk metabolome comprises low-molecular-weight metabolites, generally below 1,500 Da, present in human milk [205]. HMOs are among the most abundant components of the milk metabolome after lactose and lipids and are increasingly recognized as biologically relevant to infant neurodevelopment [152, 205]. Inter-individual variation in HMO composition is consistently observed across populations and reflects genetic and environmental influences relevant to biomarker development. HMO composition is strongly influenced by maternal secretor and Lewis status, which determine fucosylated HMO profiles [58, 153]. Secretor-positive mothers produce alpha-1,2-fucosylated HMOs, including 2′-fucosyllactose (2′-FL), whereas secretor-negative mothers produce little or none of these structures. Secretor status is associated with differences in infant gut microbiota, including variation in Bifidobacterium abundance [154]. Some studies also report higher neurodevelopmental scores among infants of secretor mothers, while breast milk 2′-FL levels have been positively associated with neurodevelopmental scores and head growth [58]. These findings support HMO profiles as candidate biomarkers of early-life exposure, although their role in individual risk prediction remains to be validated [155]. Across cohorts, 2′-FL, 3′-sialyllactose (3′-SL), and 6′-sialyllactose (6′-SL) show some of the most consistent associations with neurodevelopmental outcomes, whereas findings for other HMOs remain variable [152]. Higher fucosylated and total sialylated HMO abundance has been associated with higher language scores, particularly among infants of secretor mothers [56]. Greater exposure to 3′-SL has also been associated with higher early learning scores [55]. In another longitudinal cohort, higher 6′-SL during early lactation was inversely associated with socioemotional scores, with myelin water fraction partially mediating this association. In contrast, fucosylated HMOs were associated with language scores but showed no evidence of myelination-mediated effects, suggesting HMO-specific mechanisms [30]. Timing may be critical for biomarker interpretation. Higher milk 2′-FL at one month, but not at six months, predicted better cognitive scores at 24 months, suggesting an early window during which HMO exposure may be most informative [29]. However, translation is limited by marked variability. HMO concentrations vary substantially between individuals, in some cases by 20-fold to more than 100-fold [206]. This variation complicates cross-cohort comparison, reference range development, and patient-level interpretation. Collectively, HMO profiles are measurable and biologically plausible candidate biomarkers, but their clinical utility remains constrained by temporal variability, population heterogeneity, and lack of validated thresholds. Infant stool biomarkers The infant fecal metabolome provides a functional readout of gut microbial activity, shaped by early-life feeding and reflecting metabolites that may enter systemic circulation and influence development [207, 208]. It also captures maternal influences, including secretor status, through metabolites derived from HMO metabolism. Maternal secretor status is associated with shifts in infant fecal metabolites, including 1,2-propanediol and succinate, indicating that stool profiles can reflect biologically defined subgroups [154]. Early-life stool microbiota and metabolites have been associated with neurodevelopmental outcomes. Microbial composition at 30 days, but not at seven days, has been linked to later developmental scores, consistent with the importance of microbiome maturation [60]. Higher Firmicutes abundance has been associated with improved communication scores, whereas Campilobacterota and fecal acetic acid have been associated with lower motor outcomes, suggesting that stool-derived profiles may reflect early neurodevelopmental variation [60]. However, reliance on parent-reported outcomes and exploratory analyses limits causal inference. These findings support the investigation of stool metabolites as candidate noninvasive biomarkers; however, the reported associations remain exploratory and have not demonstrated consistent statistical robustness. Maternal stool may represent an upstream determinant of early-life microbial exposure, contributing to both breast milk and infant gut microbiota and supporting a gut–mammary–infant transfer pathway [209]. However, the absence of linked neurodevelopmental outcomes and integrated longitudinal datasets limits its current biomarker utility. Collectively, infant stool profiles provide a non-invasive, functional readout of microbiome activity, but clinical translation remains constrained by variability, limited causal inference, and lack of longitudinal validation. Standardization and reproducibility challenges Standardization and reproducibility challenges currently limit the clinical translation of HMO and stool-derived biomarkers. Quantification methods, including high-performance liquid chromatography (HPLC), high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC–PAD), and liquid chromatography–tandem mass spectrometry (LC–MS/MS), differ in sensitivity and structural resolution. Non-harmonized workflows for extraction, derivatization, and detection produce non-equivalent results across laboratories, limiting cross-cohort comparability [210]. Pre-analytical factors further contribute to variability. HMO concentrations vary with milk fraction, such as foremilk versus hindmilk, collection method, sample handling, storage temperature, freeze–thaw cycles, and timing relative to feeding or circadian rhythms [152, 211]. Clinical translation is further limited by the absence of validated reference ranges and thresholds. No established cut-offs exist for individual HMOs, including 2′-FL and 3′-SL, preventing patient-level interpretation [212]. In contrast, clinical biomarkers require defined context-of-use thresholds before implementation [201]. Collectively, analytical, pre-analytical, and threshold-related variability represent major barriers to clinical application and require harmonized protocols and validated reference standards across populations. Multi-omics biomarker Panels Biomarker panels integrate multiple independent markers and multi-omics approaches to generate more robust and reproducible insight into a defined clinical question than single biomarkers alone [213]. In a retrospective cohort of 82 infants, integration of the human milk metabolome and exposome identified sex-specific multi-analyte models incorporating deoxyceramides, acylcarnitines, phosphatidic acids, plasmalogens, nicotinamide adenine dinucleotide, and caffeine. These models predicted developmental delay at 14.2 months with moderate accuracy in the discovery cohort [57]. These findings support the feasibility of using multi-omics biomarker panels for early risk prediction, although external validation is required before clinical translation. To clarify the translational readiness of candidate biomarkers discussed in this review, Table 6 summarizes the biomarker type, evidence level, validation status, and current limitations.Table 6Candidate biomarkers relevant to human milk, infant microbiome activity, and early neurodevelopmental risk assessmentBiomarker CategoryCandidate Biomarker or ReadoutEvidence LevelValidation StatusCurrent LimitationsHuman milk HMOs2′-FL, 3′-SL, 6′-SL, total fucosylated and sialylated HMOs.Human observational evidence linking these biomarkers to cognitive, language, socioemotional, and neuroimaging outcomes [30, 55, 56, 58, 152–155]Not clinically validated.Variable by individual and time point, influenced by secretor/Lewis status, with no validated clinical thresholds.Human milk metabolomeMilk metabolomic, including small metabolites, lipids, acylcarnitines, and phospholipids.Emerging human cohort evidence linking multi-analyte profiles with developmental-delay [57, 205]Exploratory; requires external validation.Small sample sizes, limited replication, analytical variability, and uncertain predictive performance.Infant stool microbiotaEarly-life microbial composition, Bifidobacterium abundance, Firmicutes, and Campilobacterota.Human observational evidence linking stool microbial profiles with later developmental scores [60, 154]Not clinically validated.Exploratory associations with limited causal inference and variability by feeding, delivery mode, and sampling age.Infant stool metabolitesSCFAs, acetic acid, 1,2- propanediol, and succinate.Human observational and mechanistic evidence supporting microbiome functional relevance [60, 154, 207, 208]Not clinically validated.Limited validation and variability by diet, microbiome maturity, and outcome measures.Abbreviations: 2′-FL, 2′-fucosyllactose; 3′-SL, 3′-sialyllactose; 6′-SL, 6′-sialyllactose; HMO, human milk oligosaccharide; SCFA, short-chain fatty acid Table 6Candidate biomarkers relevant to human milk, infant microbiome activity, and early neurodevelopmental risk assessmentBiomarker CategoryCandidate Biomarker or ReadoutEvidence LevelValidation StatusCurrent LimitationsHuman milk HMOs2′-FL, 3′-SL, 6′-SL, total fucosylated and sialylated HMOs.Human observational evidence linking these biomarkers to cognitive, language, socioemotional, and neuroimaging outcomes [30, 55, 56, 58, 152–155]Not clinically validated.Variable by individual and time point, influenced by secretor/Lewis status, with no validated clinical thresholds.Human milk metabolomeMilk metabolomic, including small metabolites, lipids, acylcarnitines, and phospholipids.Emerging human cohort evidence linking multi-analyte profiles with developmental-delay [57, 205]Exploratory; requires external validation.Small sample sizes, limited replication, analytical variability, and uncertain predictive performance.Infant stool microbiotaEarly-life microbial composition, Bifidobacterium abundance, Firmicutes, and Campilobacterota.Human observational evidence linking stool microbial profiles with later developmental scores [60, 154]Not clinically validated.Exploratory associations with limited causal inference and variability by feeding, delivery mode, and sampling age.Infant stool metabolitesSCFAs, acetic acid, 1,2- propanediol, and succinate.Human observational and mechanistic evidence supporting microbiome functional relevance [60, 154, 207, 208]Not clinically validated.Limited validation and variability by diet, microbiome maturity, and outcome measures.Abbreviations: 2′-FL, 2′-fucosyllactose; 3′-SL, 3′-sialyllactose; 6′-SL, 6′-sialyllactose; HMO, human milk oligosaccharide; SCFA, short-chain fatty acid Candidate biomarkers relevant to human milk, infant microbiome activity, and early neurodevelopmental risk assessment Abbreviations: 2′-FL, 2′-fucosyllactose; 3′-SL, 3′-sialyllactose; 6′-SL, 6′-sialyllactose; HMO, human milk oligosaccharide; SCFA, short-chain fatty acid Clinical endpoints and neurodevelopmental readouts An endpoint is a clearly specified variable that captures an outcome of interest and is analysed to address a defined research question [201]. Across existing cohorts, neurodevelopmental endpoints are highly heterogeneous. Clinician-administered assessments, including the Bayley Scales of Infant and Toddler Development, Third Edition (Bayley-III), and the Mullen Scales of Early Learning, are used alongside parent-reported screening tools such as the Ages and Stages Questionnaire, which identify risk but do not provide equivalent quantitative measures [29, 30, 55, 57, 58, 214]. Assessment timing also varies widely, often between approximately 10 and 24 months, such that similar early-life exposures are evaluated at different developmental stages. Anthropometric proxies, including head circumference gain, are also analysed alongside behavioural measures [58]. This heterogeneity limits replication and prevents the definition of clinically meaningful thresholds for biomarker-based risk stratification and monitoring. For a biomarker to be clinically useful, it must be reproducibly measurable and must inform decision-making, including identification of infants at increased risk, stratification of meaningful subgroups, and monitoring of response to nutritional or therapeutic interventions. In the MGBA, candidates such as HMO profiles, stool-derived metabolites, and multi-omics signatures are measurable but currently lack consistent predictive performance, external validation, and thresholds for individual-level interpretation. Analytical variability and heterogeneous endpoints further limit cross-cohort comparability and constrain translation. Without standardized measurement, validated thresholds, and a clearly defined context of use, these biomarkers cannot yet support clinical decision-making [215]. Challenges, controversies, and future directions As discussed in Section “Methodological challenges in human milk Microbiome Research”, low microbial biomass and methodological heterogeneity limit the reproducibility and interpretation of human milk microbiome studies [81, 82, 216]. More fundamentally, microbial detection does not establish viability, transmission, sustained colonization, or biological function. This distinction is critical because the biological relevance of milk-associated microbes to the MGBA depends not only on their presence, but also on whether they survive ingestion, integrate into the infant gut ecosystem, and exert measurable downstream effects on immune, metabolic, or neurodevelopmental pathways. At present, many of these links remain inferential. Evidence for vertical microbial transfer exists, but the degree to which milk microbes directly colonize the infant gut, rather than reflecting shared maternal, infant oral, skin, or environmental microbial exposure, remains unresolved. This uncertainty weakens broad claims regarding milk microbiome-driven programming of infant development. Causal interpretation is further limited by the predominance of observational research. Correlations between milk microbial profiles and infant outcomes are frequently biologically suggestive, but association alone cannot establish mechanistic relevance [217]. Maternal diet, delivery mode, antibiotic exposure, breastfeeding practices, geography, lactational stage, and socioeconomic context all shape maternal and infant microbial ecosystems [218]. Without rigorous control of these confounders, it remains difficult to determine whether reported infant outcomes are attributable to milk microbes themselves or to broader maternal–environmental systems. In many cases, current evidence cannot distinguish a causal microbial effect from an association arising from the broader maternal and environmental context. A further limitation is that most current human milk microbiome studies are not designed to support formal causal inference. Establishing causality requires an explicit definition of the exposure, mediator, outcome, target population, and time window of interest within a directed acyclic graph or target-trial framework that identifies confounding, selection bias, and measurement bias before analysis [219]. This is particularly important in MGBA research because breastfeeding, milk microbiome composition, infant gut colonization, microbial metabolites, immune maturation, and neurodevelopmental outcomes are temporally ordered but highly interdependent processes. Mediation analysis can test whether infant gut microbiota, microbial metabolites, or immune markers explain the link between human milk exposures and neurodevelopmental outcomes. However, valid causal mediation requires clear temporal ordering of exposure, mediator, and outcome; repeated measurement of relevant biological variables; and control of confounding at each step, including exposure–outcome, exposure–mediator, and mediator–outcome confounding [220, 221]. Because unmeasured mediator–outcome confounding can bias indirect-effect estimates, mediation findings should be supported by sensitivity analyses [222]. To clarify the major unresolved issues and research priorities in this field, Table 7 summarizes the key questions, current evidence, major limitations, and future directions relevant to the human milk microbiome and early-life MGBA.Table 7Key unresolved questions, current evidence, limitations, and future directions in human milk microbiome–MGBA researchKey QuestionCurrent EvidenceMajor LimitationFuture DirectionIs the human milk microbiome viable and functionally active?Bacterial DNA and culturable bacteria have been detected in human milk, with recurring taxa reported across studies [20, 21]Low biomass and contamination risk limit interpretation; sequencing does not confirm viability or function [24, 25, 81, 82]Apply standardized low-biomass protocols, negative controls, viability assays, and shotgun metagenomics.Are milk microbes transferred to the infant gut?Mother–infant studies show overlap between milk and infant stool microbiota [74, 75]Shared taxa do not prove sustained colonization and may originate from skin, oral, environmental, or perinatal sources [22, 23, 128, 129]Use longitudinal dyad sampling, strain-resolved metagenomics, and source-tracking approaches.Do milk microbes influence neurodevelopment through the MGBA?Preclinical evidence links early microbiota with immune maturation, barrier function, and neurodevelopment [10, 13, 19, 32, 33, 39, 40, 47]Human evidence remains mainly observational, and direct causal links are not established [42, 49, 51, 52]Integrate milk and infant gut microbiome profiling with metabolomics, immune markers, and validated developmental outcomes.Do microbial metabolites mediate milk – neurodevelopment relationships?SCFAs and tryptophan derived metabolites are plausible mediators of microbial, immune, and neural signalling [19, 32, 39, 47, 71, 146]Mediation requires temporal ordering, repeated measures, and adequate control of confounding [220, 221]Apply causal mediation frameworks with longitudinal sampling and sensitivity analyses.Can microbiome-targeted interventions improve neurodevelopment?Probiotics, prebiotics, synbiotics, and donor milk can modify microbial composition or function [103, 104, 108, 113]Microbiome modification does not necessarily translate into neurodevelopmental benefit [113–115]Conduct randomized or quasi-experimental studies with validated neurodevelopmental endpoints and mechanistic biomarkers. Table 7Key unresolved questions, current evidence, limitations, and future directions in human milk microbiome–MGBA researchKey QuestionCurrent EvidenceMajor LimitationFuture DirectionIs the human milk microbiome viable and functionally active?Bacterial DNA and culturable bacteria have been detected in human milk, with recurring taxa reported across studies [20, 21]Low biomass and contamination risk limit interpretation; sequencing does not confirm viability or function [24, 25, 81, 82]Apply standardized low-biomass protocols, negative controls, viability assays, and shotgun metagenomics.Are milk microbes transferred to the infant gut?Mother–infant studies show overlap between milk and infant stool microbiota [74, 75]Shared taxa do not prove sustained colonization and may originate from skin, oral, environmental, or perinatal sources [22, 23, 128, 129]Use longitudinal dyad sampling, strain-resolved metagenomics, and source-tracking approaches.Do milk microbes influence neurodevelopment through the MGBA?Preclinical evidence links early microbiota with immune maturation, barrier function, and neurodevelopment [10, 13, 19, 32, 33, 39, 40, 47]Human evidence remains mainly observational, and direct causal links are not established [42, 49, 51, 52]Integrate milk and infant gut microbiome profiling with metabolomics, immune markers, and validated developmental outcomes.Do microbial metabolites mediate milk – neurodevelopment relationships?SCFAs and tryptophan derived metabolites are plausible mediators of microbial, immune, and neural signalling [19, 32, 39, 47, 71, 146]Mediation requires temporal ordering, repeated measures, and adequate control of confounding [220, 221]Apply causal mediation frameworks with longitudinal sampling and sensitivity analyses.Can microbiome-targeted interventions improve neurodevelopment?Probiotics, prebiotics, synbiotics, and donor milk can modify microbial composition or function [103, 104, 108, 113]Microbiome modification does not necessarily translate into neurodevelopmental benefit [113–115]Conduct randomized or quasi-experimental studies with validated neurodevelopmental endpoints and mechanistic biomarkers. Key unresolved questions, current evidence, limitations, and future directions in human milk microbiome–MGBA research Current study designs compound these limitations. Small cohorts, cross-sectional approaches, and single-timepoint sampling are insufficient for resolving a dynamic biological system that changes across lactation and infant development. Because milk composition and microbial exposure are temporally variable, static study designs are poorly suited to answering questions of persistence, colonization, or developmental consequence. Longitudinal, mother–infant paired, multi-omics studies are essential to determine whether milk-derived microbes play transient, structural, or negligible roles in shaping infant biological systems. Methodological harmonization remains urgently necessary, but standardization alone will not resolve the field’s most important unanswered questions. Even with fully reproducible sequencing data, key questions would remain regarding whether detected microbes are viable, functional, and clinically meaningful. Future progress therefore requires a transition from descriptive sequencing toward mechanistic validation. Integrating culture-based viability testing, strain-resolved genomics, metabolomics, and functional host-response assays will be essential to determine whether milk microbes are biologically active participants in the MGBA or primarily biomarkers of maternal microbial exposure [25, 218]. Reporting frameworks such as strengthening the organization and reporting of microbiome studies (STORMS) may improve reproducibility [223], but reproducibility without mechanistic clarity will remain insufficient. Translational enthusiasm, particularly around probiotics or other microbiome-modulating interventions, should also be approached cautiously. The field has not yet resolved fundamental questions regarding which microbes matter, how they function, or whether intervention can safely replicate naturally occurring processes. Premature clinical application may risk outpacing biological understanding, especially in vulnerable neonatal populations where regulatory inconsistencies and safety considerations remain important [224]. The central challenge is to determine whether milk-associated microorganisms are viable, transmitted to the infant, and functionally relevant. Future research should therefore prioritize contamination-resilient, longitudinal, and mechanistically informed designs rather than additional descriptive sequencing studies alone. Until these questions are resolved, claims regarding the human milk microbiome’s role in shaping infant gut and neurodevelopment must remain scientifically cautious.