How HMOs are built, which bacteria use them and what is established for 2′-FL, intestinal barriers, infections and infant feeding. Primary studies are considered alongside adult, synbiotic, allergy and safety evidence, with their practical limitations.
Human milk oligosaccharides (HMOs) are a family of carbohydrates with different structures and biological activities. Most resist digestion in the small intestine and reach microbes capable of using them. Some also interact with epithelial cells, immune receptors and microbial carbohydrate-binding proteins. Binding in a laboratory assay, changing a microbial community and preventing illness in a person are distinct outcomes.
2′-Fucosyllactose (2′-FL) is one of the best studied members. Adding it or several HMOs to formula reproduces particular components of human milk. Human milk also contains many other glycans, proteins, lipids and antibodies; the properties of this whole system cannot be assigned to one ingredient.
HMOs are built from glucose, galactose, N-acetylglucosamine, L-fucose and N-acetylneuraminic acid, usually called Neu5Ac. Their reducing end contains a lactose core, Galβ1–4Glc. Extension, branching, fucosylation and sialylation generate different structures. Lacto-N-tetraose (LNT) ends in a type-1 Galβ1–3GlcNAc unit, lacto-N-biose. Lacto-N-neotetraose (LNnT) has a type-2 Galβ1–4GlcNAc terminal unit, N-acetyllactosamine. Larger structures include branched lacto-N-hexaoses.
Neutral non-fucosylated, fucosylated and acidic sialylated groups help describe a sample, but fucose and sialic acid can coexist in one molecule. There is no fixed percentage distribution across mothers or stages of lactation. Total concentrations are generally higher in colostrum and fall with lactation; individual structures such as 3-FL may follow a different trajectory (Thurl et al., 2010).
FUT2 transfers fucose from GDP-fucose to form α1–2 linkages, including those in 2′-FL and lacto-N-fucopentaose I. The 2′-FL structure is Fucα1–2Galβ1–4Glc. FUT2 activity defines secretor status. The rs601338 variant explains a non-secretor phenotype in some populations, but it is not the sole worldwide cause. FUT3 contributes to Lewis-related glycosylation, including α1–4-fucosylation. Secretor and Lewis combinations produce different HMO profiles; they do not make non-secretor milk deficient or imply that every fucosylated HMO, including 3-FL, is absent (Thurl et al., 2010; Lewis et al., 2015).
3′-Sialyllactose (3′-SL) contains Neu5Acα2–3Galβ1–4Glc; 6′-SL contains Neu5Acα2–6Galβ1–4Glc. CMP-Neu5Ac provides the sialic acid donor. Disialyllacto-N-tetraose, DSLNT, carries two sialyl groups on a tetraose backbone. Negative charge affects molecular interactions but does not provide universal electrostatic virus repulsion. A linkage alone does not establish which human mammary sialyltransferase isoform produced it.
HMO resistance to digestion reflects their structures and the limited repertoire of matching digestive enzymes. It does not mean infants lack all β1–4-cleaving activity: lactase digests lactose. Some HMOs are found in blood and urine, so colonic fermentation is not their only fate (Goehring et al., 2014; Underwood et al., 2015).
The genome of Bifidobacterium longum subsp. infantis ATCC 15697 contains an approximately 43-kb H1 region encoding HMO-related transport and breakdown functions, with additional enzymes elsewhere. This helps explain the strain’s substrate range. It does not establish identical abilities in every member of the subspecies or every commercial probiotic (Sela et al., 2008).
One strategy is to capture intact glycans with extracellular substrate-binding proteins and import them through ATP-dependent ABC transporters. In Gram-positive bifidobacteria, these proteins should not be described as periplasmic in the Gram-negative sense. Distinct fucosyllactose uptake systems have overlapping but different specificities (Sakanaka et al., 2019).
Inside ATCC 15697, GH95 α-fucosidase Blon_2335 mainly cleaves α1–2 linkages, whereas GH29 Blon_2336 acts on α1–3/4 linkages. NanH1 is Blon_0646; NanH2 is Blon_2348 and is associated with the HMO region. β-Galactosidases and N-acetylhexosaminidases dismantle other parts of the glycan. AfcA/AfcB names used for B. bifidum enzymes should not be transferred casually to B. infantis proteins. Lacto-N-biose phosphorylase and lacto-N-biosidase are also different enzymes (Sela et al., 2011, 2012).
Cleaving 2′-FL releases fucose and lactose; lactose can subsequently yield glucose and galactose. Released fucose is not automatically routed through the same pathway as those hexoses. B. bifidum uses important extracellular glycan-processing activities, releasing products that other microbes may consume. Such cross-feeding is an ecological interaction, not evidence that the species is harmful. Bacteroides have their own glycan capture and degradation systems, including mucus-related pathways (Marcobal et al., 2011). B. breve and B. animalis subsp. lactis, including BB-12, likewise require strain-specific assessment rather than a universal label of HMO competence or clinical inferiority.
Fructose-6-phosphate/xylulose-5-phosphate phosphoketolase participates in central bifidobacterial metabolism. An idealized hexose carbon balance is:
This gives three acetate and two lactate molecules per two hexoses. It is a carbon-allocation model, not a complete ATP balance or the fixed product yield of every HMO. Actual ratios vary with substrate and growth conditions.
Acetate and lactate can lower intestinal pH and alter conditions for competing organisms. They do not sterilize the gut or guarantee pathogen elimination. Other bacteria may consume these products, including in pathways producing butyrate; the relevant communities change with age and diet. Fecal pH is a limited measurement of a complex intestinal environment, not a stand-alone definition of a healthy microbiome (Frese et al., 2017).
Some HMOs resemble carbohydrate motifs displayed on cells or in mucus. A soluble glycan can compete with a host ligand if the pathogen’s adhesin recognizes it. The effect depends on the HMO structure, microbial protein, concentration, receptor accessibility and interaction valency. Reduced attachment does not necessarily mean the organism has been killed.
For scale, the nominal molar mass of 2′-FL is about 488.44 g/mol. At 1 g/L:
This conversion is not a therapeutic dose or proof that the same concentration reaches a receptor after ingestion.
Weichert et al. (2016) examined binding of 2′-FL and 3-FL to the VP1 capsid P-domain of norovirus GII.10 and competition by virus-like particles for glycan ligands. These structural and surrogate-assay findings do not establish infection prevention in humans or protection across all norovirus genotypes, including GII.4.
Rotavirus uses different proteins, including the VP8* fragment of VP4. Ramani et al. (2018) found that some HMOs, including 2′-FL, LNT and 6′-SL, enhanced cell infection by neonatal G10P[11] rotavirus; milk composition was also associated with symptomatic infection in the observed population. This is a crucial exception to blanket antiviral claims. Different pathogens and experimental systems can respond differently to the same family of molecules.
Campylobacter jejuni has been linked to recognition of the type-2 H(O) motif Fucα1–2Galβ1–4GlcNAc (Ruiz-Palacios et al., 2003). Observational work associates some fucosylated HMOs with less diarrheal illness, but feeding, milk composition and other factors complicate causal interpretation (Morrow et al., 2004). It cannot establish equal protection against every Campylobacter strain, enteropathogenic or enterotoxigenic E. coli, Salmonella or Helicobacter pylori. Attachment, virulence expression, bacterial growth and inflammatory responses are separate endpoints.
Cholera toxin and E. coli heat-labile toxin recognize the complex ganglioside GM1. Free sialyllactose is not the complete GM1 molecule. Shared sugar fragments do not establish reliable blockade of these toxins or Clostridioides difficile toxins A/B. Direct binding and functional evidence, followed by clinical studies, are needed.
Milk glycan binding to DC-SIGN has also been considered in HIV transmission. Bode et al. (2012) reported an association between HMO concentration and postnatal HIV transmission during breastfeeding. This observational comparison was not a trial of HMO prophylaxis and cannot replace antiretroviral prevention. For cytomegalovirus, a particular HMO likewise does not establish a quantified neutralization effect or safe feeding under all conditions. HIV and CMV decisions require clinical context and applicable guidance.
Direct epithelial effects are studied separately from fermentation. In Caco-2Bbe and HT-29 models, particular HMOs reduced proliferation and altered differentiation markers; they did not simply stimulate cell division through EGFR (Holscher, Davis and Tappenden, 2014). MAPK/ERK and PI3K/Akt participate in many cellular processes, but a single obligatory “HMO → EGFR → intestinal repair” pathway has not been established.
In studied enterocyte models, 2′-FL modified CD14 expression and attenuated responses to LPS (He et al., 2016). This does not demonstrate high-affinity direct DSLNT binding to TLR4–MD-2. Receptor binding, expression changes and downstream inflammatory signals must be distinguished.
Transepithelial electrical resistance, TEER, measures an aspect of ion passage; tracer flux, such as dextran passage, measures another aspect of permeability. Higher TEER alone does not prove that bacteria cannot translocate in a person. Natividad et al. (2020) found protection in cell models challenged under specific conditions, but those experimental concentrations and exposures cannot automatically be mapped onto human feeding.
ZO-1 organizes tight-junction components with intracellular structures. Occludin and different claudins have distinct functions, including ion-selective passage. MLCK regulation can affect barrier function, but universal suppression by all HMOs has not been shown. The distinction between markers, tight junctions and clinical outcomes is explored in the article on intestinal permeability and claudins.
MUC2 is the major secreted goblet-cell mucin; MUC1 is membrane-associated and contributes to the surface glycocalyx. A change in one mucin marker does not establish a completely sterile mucus layer. Non-secretor maternal milk does not necessarily produce a thin or defective mucus barrier.
Necrotizing enterocolitis (NEC) is a serious disease of particular importance in premature infants. Jantscher-Krenn et al. (2012) showed protection by DSLNT in a neonatal rat NEC model. This supports investigation of mechanisms, but it does not determine supplement efficacy in a premature infant.
Autran et al. (2018) followed 200 mother–very-low-birth-weight infant pairs; eight infants developed NEC. Lower milk DSLNT was associated with NEC. This event count and observational design do not establish a universal diagnostic concentration or guarantee absence of disease above a threshold. The result supports a potential biomarker, not a ready milk-selection or dosing algorithm.
DSLNT is not the only HMO investigated in this context. Experimental 2′-FL effects on mesenteric perfusion have also been reported (Good et al., 2016). Vascular perfusion and endothelial nitric oxide synthase signaling are a different compartment from epithelial tight junctions. Their background is covered in the article on eNOS and uncoupling.
The endothelial glycocalyx contains proteoglycans and glycosaminoglycans. Finding HMOs in blood does not demonstrate incorporation into this structure or replacement of heparan sulfate. Syndecan-1 release can reflect glycocalyx injury; it does not establish specific HMO repair or thrombosis prevention.
DC-SIGN is a calcium-dependent membrane lectin. Noll et al. (2016) identified binding of particular milk glycans; free 2′-FL had an inhibitory concentration around 1 mM in the assay used. An IC50 depends on assay design and is not a dissociation constant, Kd. This cannot be described as nanomolar affinity or effective binding in every tissue.
Selectins and Siglecs recognize specific motifs. Spatial presentation and combined fucosylation/sialylation, as in some sialyl-Lewis structures, matter for selectins; simple 2′-FL and 6′-SL are not automatically universal antagonists. Bode et al. (2004) observed changes in platelet–neutrophil complexes in an experimental system. That is mechanistic evidence, not established sepsis therapy.
ITIM signaling by some Siglecs can recruit SHP-1/SHP-2 phosphatases and limit activation. This general biology does not prove that dietary free 6′-SL suppresses Siglec-9-related processes, NADPH oxidase or neutrophil extracellular traps. The relevant ligand, functional effect and attainable exposure require separate evidence.
Th1, Th2, Th17 and regulatory T-cell maturation depend on many signals. HMOs and microbial metabolites can influence this environment, but a universal immune “switch” is not established. Acetate and butyrate should not be treated as equivalent histone-deacetylase inhibitors. IL-22 can support epithelial defense, so suppressing it is not a general goal.
Goehring et al. (2016) measured cytokines in an infant feeding-study subgroup, with a separate breastfed reference group. Changes in plasma or stimulated-cell cytokines are immunological endpoints. They do not establish prevention of every infection or clinical equivalence between formula and human milk.
Some HMOs are detectable in infant blood and urine (Goehring et al., 2014; Underwood et al., 2015). This establishes possible systemic access, not a fixed 1–2% bioavailability for all structures. Extraintestinal effects require evidence about exposure, receptors and outcomes; RSV or urinary-tract infection prevention does not follow simply from detection.
Some HMOs are produced by fermentation with engineered strains, including derivatives of E. coli K-12. Engineering supplies GDP-fucose and transfers fucose to lactose with a suitable fucosyltransferase. manB/manC, gmd and wcaG describe precursor-pathway functions; human FUT2 is not a generic name for every bacterial fucosyltransferase. Specific changes, such as lacZ inactivation, depend on the production platform.
Cells are removed after fermentation and the product is purified and concentrated. Relevant specifications include identity, related sugars, residual protein, microbiological quality and, where appropriate, endotoxin and production-organism material. Identical yields, 98% purity or absolute freedom from every impurity cannot be promised across manufacturers. FDA’s GRN 852 response describes specifications for a particular 2′-FL process, illustrating product-specific control rather than a universal HMO standard (FDA, 2019).
| Study | Question | Interpretation |
|---|---|---|
| Marriage et al., 2015 | Growth and 2′-FL uptake in term infants receiving different formulas | Evidence for particular formulas and healthy participants; breastfeeding was a separate reference group |
| Goehring et al., 2016 | Immunological measurements in a feeding-study subgroup | Cytokine changes are not equivalent to clinical infection protection |
| Puccio et al., 2017 | Formula with 1 g/L 2′-FL and 0.5 g/L LNnT; 175 randomized infants | Growth was primary; illness and medication reports were secondary |
| Berger et al., 2020 | Microbiota community types and later antibiotic use | A community analysis does not establish causality for one bacterial species |
In Puccio’s study, the HMO-containing formula was given through six months. Weight gain at four months met the prespecified non-inferiority criterion. Follow-up beyond six months did not mean continuous HMO exposure through the first year. Secondary differences in bronchitis, lower respiratory infections and antibiotic use were interesting for follow-up research. Outcomes included parent reports and were analyzed without adjustment for multiple comparisons. They do not support universal percentages of protection or established otitis prevention. Manufacturer funding is also part of the study context (Puccio et al., 2017).
Microbiota changes may be intermediate mechanisms, but cannot guarantee a clinical outcome. A 16S rRNA community analysis does not always resolve closely related species or strains. More Bifidobacterium at genus level does not establish colonization specifically by B. infantis EVC001 (Berger et al., 2020).
Elison et al. (2016) studied 100 healthy adults for two weeks using 2′-FL, LNnT or combinations, including doses up to 20 g/day. The study addressed tolerability and microbiota. A short healthy-volunteer trial does not establish long-term safety in every patient or efficacy for irritable bowel syndrome (IBS).
Palsson et al. (2020) recruited 317 people with IBS in an open-label multicenter study; 245 completed it. Participants used 5 g/day of a 4:1 2′-FL:LNnT mixture for 12 weeks. Improvement from baseline cannot separate ingredient effects from placebo, changed habits or natural symptom variation because there was no control group.
Iribarren et al. (2020) conducted a separate randomized placebo-controlled microbiota-focused trial; a later publication examined metabolites and mucosal responses (Iribarren et al., 2021). Each paper should not be counted as an independent replication of clinical efficacy. Prespecified endpoints, between-group effects and trial design remain essential.
For inflammatory bowel disease, experimental barrier protection or altered metabolites are not equivalent to remission of ulcerative colitis or Crohn’s disease. There is no established universal HMO treatment protocol for IBS, small intestinal bacterial overgrowth or IBD. Trial doses describe studied conditions, not individualized prescriptions.
Combining an HMO substrate with a suitable live strain is a synbiotic strategy, whose outcome depends on diet and the starting microbiome. Frese et al. (2017) reported persistence of EVC001 in breastfed infants after supplementation. This does not guarantee identical colonization under all conditions. In adults, B. infantis with HMOs produced reversible engraftment rather than assured lifelong persistence without substrate (Button et al., 2022). ATCC 15697, EVC001 and other strains need separate evidence.
A purified carbohydrate and a living probiotic have different risks. In 2023, FDA warned about invasive infections associated with some probiotic products used in hospitalized premature infants. This is not a warning about the HMO molecule itself, but matters when evaluating synbiotics for this vulnerable population (FDA, 2023).
An extensively hydrolyzed formula with two HMOs was studied in infants with cow’s milk protein allergy. Growth, tolerability and secondary infection outcomes do not establish faster allergy resolution or prevention of the entire atopic march (Vandenplas et al., 2022). The whole product matters: an HMO does not make an unsuitable protein ingredient hypoallergenic.
Sialic acid participates in gangliosides and NCAM polysialylation, both relevant to nervous-system development. Endogenous synthesis involves GNE, and polysialylation involves ST8SIA2/ST8SIA4. This biology does not prove an obligatory deficiency without sialyllactose supplementation. Dietary sialyllactose isomers changed ganglioside sialic acid in particular brain regions of piglets; this animal biochemical endpoint does not establish improved intelligence in children (Jacobi et al., 2016).
Microbial acids can participate in FFAR2/FFAR3 and gut-hormone signaling. This motivates HMO–GLP-1 and metabolic hypotheses, but does not establish human treatment of steatotic liver disease, guaranteed LPS reduction or improved insulin sensitivity. The broader mechanism appears in the article on GLP-1 and nutrient secretagogues.
Safety assessment concerns a specified structure, manufacturing process, composition, population and conditions of use. A no-questions response to a particular FDA GRAS notice addresses the notifier’s conclusion under those conditions; it is not drug approval or validation of every health claim (FDA, 2019). EFSA assesses novel-food safety, while EU authorization is a separate regulatory process. An opinion about 2′-FL and LNnT does not automatically cover every HMO (EFSA, 2022).
Large amounts of fermentable carbohydrate may cause gas, bloating or altered stools. Tolerability depends on the mixture, amount and person. Animal toxicology levels cannot be directly converted into an infant treatment dose. Adding an HMO independently to infant feeding does not reproduce a tested formula and can complicate assessment of its composition.
Interpret each conclusion in relation to the molecule or mixture, strain, age and endpoint actually studied. HMOs are an important area of nutrition and microbial-ecology research. Evidence for growth and tolerability in particular formulas, observational associations and mechanistic experiments carries different implications; together, these studies do not define a universal therapeutic protocol.
Method: this version, dated 8 October 2026, uses a selected check of primary studies and official documents. Accessible full-text sections, abstracts and indexed primary-source descriptions were used; complete access was not available for every paper. This is a narrative analysis, not a systematic review, a GRADE assessment or an independent peer review.
No. HMO describes a family of oligosaccharides; 2′-FL is one structure. Findings about it cannot automatically be applied to all HMOs or whole human milk.
Secretor status changes the profile of certain HMOs. It does not make the milk deficient or determine an infant’s health from one marker.
No. Effects depend on microbial proteins and glycan structures. Laboratory findings and secondary trial outcomes do not establish equal protection against every infection.
Animal and observational findings support further investigation. They do not provide a ready protocol for independently supplementing premature infants.
A universal therapeutic dose is not established. Published doses describe particular studies with different designs, mixtures and participants.
It supplies selected carbohydrate components. Human milk contains a more complex combination, and clinical equivalence cannot be inferred from one ingredient.