Recent advances in the science of human milk oligosaccharides

Highlights

• •The lists of human milk oligosaccharides (HMOs) are updated to incorporate the identified structures since 2018.
• •Laboratory-scale preparations of HMOs by isolation from human milk, enzymatic synthesis, and chemoenzymatic synthesis are reviewed.
• •Several biological functions of HMOs, including anti-infection, prevention of necrotizing enterocolitis, and enhancement of cognitive ability, are reviewed.
• •Industrial production of fucosyllactose, lacto-N-tetraose/lacto-N-neotetraose, and sialyllactose by fermentation process are reviewed.
• •Intervention trials to use several HMOs are reviewed.

Abstract

Human colostrum and mature milk contain oligosaccharides (Os), designated as human milk oligosaccharides (HMOs). Approximately 200 varieties of HMOs have been characterized.

Although HMOs are not utilized as an energy source by infants, they have important protective functions, including pathogenic bacteria and viral infection inhibitors and immune modulators, among other functions, and HMOs stimulate brain-nerve development. The Os concentration is average 11 g/L in human milk but >100 mg/L in mature bovine milk, which is used to manufacture infant formula, suggesting that human-identical milk oligosaccharides (HiMOs) should be incorporated into milk substitutes. Some infant formulas incorporating 2′-fucosyllactose and lacto-N-neotetraose are now commercially available, and intervention trials have been concluded.

We review basic HMO information, including their chemical structures and concentrations, attempts to synthesize HMOs at small and plant scale, studies that clarified HMO biological functions, and interventions with milk substitutes incorporating HiMOs in formula-fed infants.

1. Basic information about human milk oligosaccharides (HMOs): chemical structures and concentrations

Human milk contains approximately 7 % carbohydrate, comprising 80 % lactose (Galβ1−4Glc) and 20 % other oligosaccharides (Os). In this paper, milk oligosaccharides are defined for oligosaccharides in milk excepting for lactose. The Os in human milk, called HMOs, are complex carbohydrate mixtures, most of which have a lactose unit in the reducing end and contain monosaccharides, such as glucose (Glc), galactose (Gal), N-acetylglucosamine (GlcNAc), N-acetylneuraminic acid (Neu5Ac) or fucose (Fuc) [1,2]. The estimated total HMO concentrations are 17.7 g/L in the colostrum (0–5 days postpartum), 13.3 g/L in transition milk (6–14 days postpartum), and 11.3 g/L in mature milk (15–90 days postpartum), as averaged from the data from 12 papers [3]. HMOs are the third highest percentage of solid components after lactose and lipids in human milk [1,2]. HMOs are unique among milk oligosaccharides (MOs) in terms of the prevalence of neutral oligosaccharides, particular fucosyl HMOs, in comparison to acidic HMOs. In contrast, acidic MOs are more abundant than neutral MOs in the milks or colostrums of cows, sheep, goats, horses and pigs. Furthermore, the type 1 MOs, which contain Galβ1–3GlcNAc (lacto-N-biose I), are more prevalent than the type 2 MOs, which contain Galβ1–4GlcNAc (N-acetyllactosamine, LacNAc). This is in contrast to the milks of other mammals, including primates, where the type 2 MOs are more abundant than the type 1 [4].

In infants, the lactose from breast milk is hydrolyzed by lactase at the micro villi of the brush border along the small intestine, and the resulting Gal and Glc are absorbed and used for energy. By contrast, most HMOs remain unabsorbed and reach the colon, where they serve as prebiotics to stimulate the growth of beneficial colonic bacteria, anti-infection agents against pathogenic microorganisms, colonic immune-modulating factors, colonic-barrier-strengthening factors, and preventive factors for necrotizing enterocolitis (NEC). After small portion of HMOs is absorbed and enters the circulation, they are used in brain-nerve development or act as immune-modulation factors during the circulation process [1,2].

1.1. Chemical structures of HMOs

Approximately 200 HMOs have been characterized to date. Urashima et al. listed the HMO structures in reviews published in Trends in Glycoscience and Glycotechnology (2018) [1] and Comprehensive Glycoscience, second edition (2021) [2]. These structures have been classified into 20 series according to their core structures [2], and approximately 170 varieties of HMOs have been constructed with Fuc and/or Neu5Ac attached to the specific positions of Gal, GlcNAc or Glc in the cores. After these reviews were published, other HMOs, which are present at trace levels, have been identified by mass spectrometry (MS) by Remoroza et al. [5], Samuel et al. [6], Weng et al. [7], Shi et al. [8], and Zhang et al. [9]. In the present review, these additional structures are presented in the list of HMOs in Table S1, and their structural diversity is discussed. The nomenclatures of each oligosaccharide are shown in the supplementary file, but their abbreviations are shown in Table S1. Shi et al. presented tentative HMO structures because the anomeric configurations of the glycosidic bonds and linked positions of the residues could not be definitively characterized.

Although Os, which contain Galβ1–4GlcNAc at the reducing end, have been found in milk/colostrum of cows, sheep, and goats [10], they had never been identified in human milk. The identification of such type neutral and sialyl (N-acetylneuraminyl) Os in human milk by Remozora et al. [5] is noteworthy, although the concentrations were at trace levels. It can be hypothesized that these oligosaccharides are formed from the glycoproteins or glycolipids in milk by peeling of partial sugar chains.

The HMOs having Galβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4Glc (lacto-N-novopentaose I, novo LNP-I) as a core are minor saccharides [5]. Previously, only one fucosyl and sialyl oligosaccharide of this type was identified [11]. Recently, underivatized novo LNP-I and Galβ1–3Galβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4Glc (Galnovo LNP-I) were found in human milk by Remoroza et al. [5] and Shi et al. [8]. These Os have been found in marsupials, such as the tammar wallaby [12], brushtail possum [13], koala [14], and in several domestic animals, including cows [10,15], goats [10], sheep [10], horses [10,16], and camels [10,17]. Identification of this type oligosaccharide in human milk further expands our knowledge of the structural diversity of HMOs, even though this type is very minor in human milk.

Shi et al. found Galα1–3(Fucα1–2)Galβ1−4Glc (B tetrasaccharide) and Galα1–3(Fucα1–2)Galβ1–3/4GlcNAcβ1–3Galβ1–4Glc (B hexasaccharide), which contain B antigen (Galα1–3[Fucα1–2]Gal), in human milk initially [8]. It is assumed that these OSs should exist only in the milk of Type B blood donors, and the concentrations should be similar or lower levels to those of A tetrasaccharide (GalNAcα1–3[Fucα1–2]Galβ1–4Glc) and A hexasaccharide (GalNAcα1–3[Fucα1–2]Galβ1–3GlcNAcβ1–3Galβ1–4Glc) in the milk of Type A blood donors. A-tetra was consistently found in cohorts with mean concentrations at 30–60 mg/L depending on time of lactation with maximum levels in the range of >200 mg/L observed in some individuals [18,19]. It is also noteworthy that Neu5Acα2–8Neu5Acα2–3Galβ1–4Glc (DSL) was found in human milk at first. DSL is an abundant milk oligosaccharide in bovine colostrum [10], but it is only at a trace concentration in human milk.

Weng et al. characterized novel Os as Galβ1–4(Galβ1–2)Glc, Galβ1–4(Glcα1–2)Glc, GalNAcβ1–4GlcNAcβ1–6Galβ1–4Glc, and Galβ1–4Glcβ1–4Glc [7] but could not speculate on their biosynthesis by typical mechanisms. Possibly, they are biosynthesized by unidentified novel glycosyltransferases or by a monosaccharide-transfer reaction by mammary-cell glycosidases [20]. Notably, an unprecedented trisaccharide with Glcβ1–4Glc in the reducing end was identified.

It is also interesting that Fucα1–2Galβ1–3GlcNAcβ1–3(Galβ1–6Galβ1-?Galβ1–4GlcNAcβ1–6)Galβ1–6Galβ1–4Glc, which contains 6′-GL (Galβ1–6Galβ1–4Glc) as a core unit reported by Hanisch et al. [21] even though the structure was not definitively characterized.

Wu et al. found an undeca- and a dodeca-saccharides, which each had two potential structures [22,23]. Undecasaccharide; Galβ1–3GlcNAcβ1–3Galβ1–4GlcNAcβ1–3[Galβ1–4GlcNAcβ1–3Galβ1–4(Fucα1–3)GlcNAcβ1–6]Galβ1–4Glc (F-novo-lacto-N-neodecaose) or Galβ1–3(Fucα1–4)GlcNAcβ1–3[Galβ1–3GlcNAcβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4GlcNAcβ1–6]Galβ1–4Glc (F-LND-II), dodecasaccharide; Galβ1–3GlcNAcβ1–3Galβ1–4(Fucα1–3)GlcNAcβ1–3[Galβ1–4GlcNAcβ1–3Galβ1–4(Fucα1–3)GlcNAcβ1–6]Galβ1–4Glc (DF-novo-lacto-N-neodecaose) or Galβ1–3(Fucα1–4)GlcNAcβ1–3[Galβ1–3GlcNAcβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4(Fucα1–3)GlcNAcβ1–6]Galβ1–4Glc (DF-LND-VII). Novo lacto-N-neodecaose is a potential HMO core structure in addition to the previously proposed 20 cores. Twenty-one core structures of HMOs are shown in Table S2, which include novo lacto-N-neodecaose.

Among HMOs, a few Os, including DF-para-LNH sulfate I, and II, TF-para-LNH sulfate (structure in Table S1) [24] and 3′-SL-6′-O-sulfate [25], have sulfate groups, although at trace concentrations. Only OH-6 of GlcNAc or Gal could be substituted with a sulfate group among HMOs. Although some Os sulfated at the OH-3 of a non-reducing Gal residue, such as lactose-3′-O-sulfate have been identified in the milk of several mammals, including house dogs [26], red kangaroos [27], and brushtail possum [13], this type of sulfated oligosaccharide has never been identified in HMOs.

Although CMP-Neu5Ac hydrolase, which catalyzes the conversion of CMP-Neu5Ac to CMP-Neu5Gc, has been knocked out in humans, Quin et al. found Neu5Gc- containing Os in human milk; it is hypothesized that the origin of Neu5Gc might be the glycoproteins/glycolipids in the diet, which had been fed by the mothers [28]. However, the structures have not been characterized.

Porfirio et al. characterized HMO compositions by permethylation followed by liquid chromatography with high-resolution tandem mass spectrometry (LC-MS/MS) analysis in human milk samples from Bangladeshi mothers. They found >100 different glycoforms, with a highest detected permethylated >5000 mass units [29].

1.2. Potential biosynthetic pathway of HMOs

The potential biosynthetic pathways of the core structures of HMOs, excepting those containing a reducing LacNAc unit, are shown in Fig. 1. These include the pathway of lacto-N-novopentaose I, which was not included in our previous reviews [1,2], and novo lacto-N-neodecaose (structure in Table S2), which is a potential structure that has not been definitively elucidated. Two varieties of β6-N-acetylyglucosaminyltransferase (IGnT) are assumed to be related to the biosynthesis: one has substrate specificity to GlcNAcβ1–3Galβ1–4Glc (lacto-N-triose-II, LNTri-II) to produce GlcNAcβ1–3(GlcNAcβ1–6)Galβ1–4Glc, designated as dIGnT [30], and the other can transfer GlcNAc to Gal at the third position from the nonreducing end of Galβ1–4GlcNAcβ1–3Galβ1–4Glc (lacto-N-neotetraose, LNnT), designated as cIGnT [30]. The potential pathways of dIGnT and cIGnT are shown by solid and dotted lines, respectively (Fig. 1). In addition, the third IGnT, which transfers GlcNAc to Gal at the second position of Galβ1–3Galβ1–4Glc, is hypothesized. The activity of this IGnT type has been found in the lactating mammary glands of the tammar wallaby, a marsupial [31,32]. McDonald et al. [30] hypothesized that the following glycosyltransferases are related to the biosynthesis of HMOs and constructed 206 oligosaccharide structures in silico; β-N-acetylglucosaminyl-glycopeptide β-1,4-galactosyltransferase (β4GalT, EC 2.4.1.38), 3-galactosyl-N-acetylglucosaminide 4-α-l-fucosyltransferase (α4FucT, EC 2.4.1.65), galactoside 2-α-l-fucosyltransferase (Type 1 & Type 2) (α2FucT, EC 2.4.1.69, EC 2.4.1.344), N-acetyllactosaminide β-1,3-N-acetylglucosaminyltransferase [β3GnT (iGnT), EC 2.4.1.149], 4-galactosyl-N-acetylglucosaminide 3-α-l-fucosyltransferase (α3FucT, EC 2.4.1.152), N-acetyl-β-d-glucosaminide β-1,3-galactosyltransferase (β3GalT, EC 2.4.1.86), β-galactoside α-2,6-sialyltransferase (ST6Gal, EC 2.4.3.1), N-acetyllactosaminide α-2,3-sialyltransferase (ST3Gal, EC 2.4.3.6), N-acetylglucosaminide α-2,6-sialyltransferase (ST6GlcNAc, EC 2.4.3.10), N-acetyllactosaminide β-1,6-N-acetylglucosaminyltransferase (cIGnT, EC 2.4.1.150), GlcNAcβ1,3Galβ-1,6-N-acetylglucosaminyltransferase (distally acting) (dIGnT, EC 2.4.1.386). However, the existence of type 2 α2FucT might not be possible, because almost all fucosyl HMOs, which have a non-educing α1–2 linked Fuc, contain Fucα1–2Galβ1–3GlcNAc unit, excepting for Fucα1–2Galβ1–4Glc (2′-FL) and Fucα1–2Galβ1–4(Fuca1–3)Glc (DFL).

Fig. 1.

The potential biosynthetic pathway of the 21 core structures of HMOs (A) and a comprehensive scheme to show characteristic features of HMOs with focus on branching and elongation events (B). Fig. 1(A) and (B) are adapted from Fig. 1, Urashima, T. et al., Comprehensive Glycosci. (second edition), vol.5 (2021) pp389–439, ref. [2].

Kellman et al. used a system for biology framework that integrates glycan and RNA expression data to construct an HMO biosynthetic network and predict the glycosyltransferases involved. They constructed models describing the most likely pathways for the synthesis of the Os accounting for >95 % of the HMO content in human milk. They also proposed candidate genes for elongation, branching, fucosylation, and sialylation of HMOs [33].

Ambalavanan et al. addressed the complex and dynamic nature of the mother-milk-infant triad by investigating maternal genomic factors regulating HMOs. Nineteen HMOs are quantified from 980 mothers of the CHILD Cohort study. Genome-wide association studies identified HMO-associated loci on the chromosomes, spanning several fucosyltransferases (FUT) genes, and also novel associations on another chromosome for 6′-sialyltransferase (6′-SL) in the sialyltransferase (ST6GAL1) gene [34].

1.3. Concentration of HMOs

HMOs profiles are not homogeneous among all donors’ milk, and the heterogeneity has been found in either secretors or non–secretors, and also in either Lewis-positive or -negative types. Secretor or nonsecretor donors are identified by the detection/nondetection of ABO blood group antigens in the donor's bodily fluids, including milk; this is caused by the presence/absence of α2FucT (FUT2) activity, which transfers Fuc to nonreducing Gal, on lactating mammary glands. Nonsecretor donor's milk does not contain or contains only trace levels of 2′-FL, lacto-N-fucopentaose-I (LNFP-I), or lacto-N-difucohexaose-I (LNDFH-I), which have a nonreducing Fucα1–2 residue, although these are the predominant HMOs in secretor's milk. Lewis positivity or negativity is caused by the presence/absence of α3FucT (FUT3) activity, which transfers Fuc to GlcNAc of Galβ1–3/1–4GlcNAc via α1–3- or α1–4-linkage. Lewis-negative donor milk does not contain or contains only trace levels of LNFP-II and LNDFH-I, II, which have a Fuc residue linked to GlcNAc via α1–4-linkage [35]. Although the ratios of secretor/nonsecretor and of Lewis-positive/Lewis-negative donors differ according to the donor's ethnicity and geographic location [3,36], it is estimated to total approximately 80 % in secretors, 15 % in nonsecretors, and 5 % in Lewis-negative individuals on average. Table 1 presents the mean concentrations (g/L) of the abundant HMOs in the mature milk of secretor, nonsecretor, and Lewis-negative donors, as determined by Thurl et al. [37].

Table 1.

Concentrations of HMOs (g/L) in mature milk of three donor groups, as reported by Thurl et al. (2010) [37] and those in colostrum, transitional, mature, and late milks, as reported by Soyyilmaz et al. (2021) [3].

	2′FL	LNDFH-I	LNFP-I	LNFP-II	LNT	3FL	6′SL	DSLNT	LNnT	DFL	FDS-LNH-I	LNFP-III	3′SL	LST c	TF-LNH-I
Colostrum (0–5 days)	3.18	1.03	0.83	0.78	0.73	0.72	0.4	0.38	0.37	0.29	0.28	0.26	0.19	0.17	0.25
Transitional (6–14 days)	2.07	1.06	1.11	0.33	1.07	0.59	0.71	0.67	0.47	0.56		0.37	0.13	0.55	0.17
Mature (15–90 days)	2.28	1.1	0.83	0.78	0.74	0.72	0.403	0.38	0.372	0.293	0.29	0.26	0.19	0.17	0.15
Late (90 days ∼)	1.65	0.87	0.41	0.27	0.64	0.92	0.3	0.22	0.19	0.27	0.27	0.23	0.13	0.08	0.2
Secretors	3.13	1.28	1.58	0.22	0.87	0.42	1.22	0.39	0.23	0.41		0.38	0.27	0.29
Non-secretors	ND	ND	ND	1.25	1.18	1.79	1.14	0.42	0.08	ND		0.38	0.24	0.21
Lewis negatives	4.57	ND	3.18	ND	0.84	0.15	1.31	0.41	0.23	0.17		0.31	0.31	0.31

Soyyilmaz et al. estimated the mean concentrations (g/L) of the abundant HMOs in the colostrum (0–5 days), transitional milk (6–10 days), mature milk (15–90 days), and late milk (>90 days), independent of secretor status, geographic location, mother's ethnicity, and analytical method [3]. The concentrations were averaged from the values collected from 51 articles and 30 countries (Table 1). In mature pooled milk, the top 15 HMOs in decreasing order of magnitude are 2′-FL, LNDFH-I, LNFP-I. LNFP-II, LNT, 3-FL, 6′-SL, DSLNT, LNnT, DFL, FDS-LNH-I, LNFP-III, 3′-SL, LST c, and TF-LNH-I. The most abundant HMO in all lactation periods is 2′-FL, the level of which decreases from 3.18 g/L in colostrum to 1.64 g/L in late milk, whereas the 3-FL level increases from 0.37 g/L in colostrum to 0.92 g/L in late milk. The order of abundance of the concentrations (mmol/L) is 2′-FL (32 %), 3-FL (10 %), LNDFH-I (8 %), LNT (7 %), LNFP-I (7 %), LNFP-II (6 %), 6′-SL (4 %), LNnT (4 %), and DFL (3 %). The data on the concentrations of the top 15 HMOs demonstrate that fucosyl HMOs are highly abundant among HMOs and that type 1 HMOs are more prevalent than type 2, as previously described. The data presented by Thurl et al. demonstrated that the ratios of LNT vs LNnT, as well as the ratios of saccharides containing the LNT core vs those containing the LNnT core, are higher in the former than the latter in the breast milk of secretor donors. Furthermore, these ratios were higher still in the milk of non-secretor donors [37]. It is hypothesised that the prevalence of type 1 HMOs in human milk is associated with the formation of the colonic bifidus flora in colon of breastfed infants, given that the human infant type bifidobacteria possess the specific metabolic pathway required to metabolise type 1 HMOs [4]. Furthermore, the prevalence of fucosyl HMOs would confer an advantage to the bifidus flora formation, as these type bifidobacterial strains possess the requisite transporters for the internalization of fucosyl HMOs [38]

Conze et al. estimated the mean concentrations of 2′-FL, 3-FL, LNT, 3′-SL, 6′-SL as 2.58, 0.57, 0.94, 0.28, and 0.39 g/L, respectively, by weighted analysis [39]. Please refer to the excellent reviews of the HMOs concentrations [3,36].

2. Laboratory-scale preparation of HMOs for in vitro functional studies

HMOs have over 200 structures [5,32] and are known for their diverse functions [1,2,40]. With the growing body of knowledge on HMOs, many efforts have been made to incorporate them into infant formulae [[41], [42], [43]] including preterm infants in NICU [44] or clinical supplements for adults [45]. Currently, several infant formula manufacturers have begun to introduce HMOs into their products [46]. The details of industrial HMO production will be discussed in the Section 4. In this section, we summarize the representative methods for the preparation of HMOs on a laboratory scale for studies on their functions and characteristics.

2.1. Isolation of HMOs from human milk

Human milk contains approximately 11 g/L of HMOs together with large amounts of lactose (67–78 g/L), lipids (32–36 g/L), and proteins (9–12 g/L) [47]. Therefore, it is necessary to remove large amounts of lipids, proteins, and lactose before isolating HMOs from the donated milk. Briefly, lipids are first removed by centrifugation and filtration through, for example, a glass wool column. Next, precipitated proteins, generated by the addition of two volumes of ethanol to the retentate, are removed by centrifugation. Finally, a large amount of lactose is removed by solid phase extraction [48].

Since the late 1960s, Kobata and Ginsburg et al. have reported the isolation and purification of various HMOs by performing a combination of gel and paper chromatography [[49], [50], [51]], which has accelerated the study of HMOs. In gel chromatography, the crude HMOs mentioned above are dissolved in water, and the solution is applied to a Sephadex G-25 or Bio-Gel P-4 column. A broad peak of acidic HMOs elutes initially, followed by a broad peak of neutral HMOs, regardless of the molecular sizes of the oligosaccharides [52]. The similar elution pattern was reported also by Ranjan et al. in the isolation of sheep MOs [53]. The obtained acidic, neutral, or the entire HMO group can be used in research on the functions or characteristics of HMOs.

Another method for isolating HMOs is to use carbon columns, such as an activated carbon column [54] or a graphitized carbon column [48]. The defatted and deproteinized HMOs containing a large amount of lactose are dissolved in 4 %(v/v) aqueous ethanol, and the solution is applied to the carbon column pre-equilibrated with 4 % (v/v) aqueous ethanol. Under these conditions, lactose elutes without adsorbing to the column while HMOs are retained. Then, the HMOs in the column are isolated by gradient elution, first with a mobile phase containing an aqueous ethanol concentration ranging from 4 % to 20 % (v/v) to obtain neutral Os. Then with a similar gradient using a similar solution that contains 0.1 % trifluoroacetic acid to obtain acidic Os [53]. Acetonitrile may be used instead of ethanol in these separations [54].

To study the functions and characteristics of HMOs, two approaches are used. One approach is to use neutral Os, acidic Os, and/or entire Os individually [54]. The other approach is to use specific HMO samples with clearly defined structures to investigate structure-function relationships [55]. In the case requiring further isolation of individual HMOs, various preparative high-performance LC methods are used depending on the nature of the research [56]. Examples include hydrophilic-interaction HPLC using a silica-based amide column (TSKgel Amide-80), a polymer-based amino-column (Asahipak NH2P-50) [57], a porous graphitized carbon stationary-phase LC [58], or anion-exchange LC with a pulsed amperometric detection system [59]. The processes explained above can be used to isolate and purify Os synthesized by chemical or enzymatic methods as well.

Some representative HMOs isolated and purified by the above procedures have been commercialized by various reagent companies and have been used in basic research, such as functional or mechanistic studies of glycosyltransferases or galectins. However, in recent years, it has become increasingly challenging for reagent companies to sell materials derived from human origin, including HMOs, due to ethical considerations.

2.2. Transglycosylation reactions using natural glycosidases

Since the mid-1980s, enzymatic oligosaccharide syntheses have been developed mainly by transglycosylation reactions using glycosidases derived from natural sources [60]. For example, Fucα1–2Galβ-OMe [61], Fucα1–3GlcNAc [62], Galβ1–4GlcNAc [63], Galβ1–6GalNAc [64], 3′-SL [65], and 6′-SL [65] have been synthesized by transferring, for example, p-nitrophenyl glycosides as the donor sugar to the corresponding acceptor sugars. However, most of the reaction yields are <40 %, which are insufficient for preparing Os on a large scale.

Transglycosylation reaction using glycosidases from natural sources have several problems: (i) the yield is generally low due to the hydrolysis of glycosyl donor and products during the reaction, (ii) the amounts of the enzyme derived from natural sources are not always sufficient, and (iii) if the microorganism is pathogenic, the produced Os cannot be used as food additives. To solve these problems, genetic engineering of glycosidases has been challenged since the late 1990s. The results of these endeavors are detailed in 2–3–2–6.

2.3. Transglycosylation reactions using recombinant glycosidases

Bacillus circulans mainly produces a β−1,4-galactosidase, which catalyzes the transglycosylation reaction to yield Galβ1–4GlcNAc using Gal-β-pNP as the donor and GlcNAc as the acceptor [63]. However, production of β−1,3-galactosidase by this microorganism is currently unknown. When Ito and Sasaki [66] tried to clone the gene for β−1,4-galactosidase from Bacillus circulans, they also obtained the gene for β−1,3-galactosidase from the same microorganism, and each gene was expressed in Escherichia coli separately. The resulting recombinant β−1,3-galactosidase, bgaC, exhibited high transglycosylation activity, producing Galβ1–3GlcNAc (LNB) regioselectively with a 36 % yield from Gal-β-pNP and GlcNAc [67].

Guo et al. cloned six genes encoding sialidases, three from Bifidobacterium fragilis NCTC9343 and three from Clostridium perfringens ATCC 13124 [68]. These six genes were subjected to heterogeneous expression in E. coli. The recombinant enzymes were screened for transglycosylation activity. Among these six mutant enzymes, sialidase BfGH33C possessed the best transglycosylation activity. The obtained recombinant enzyme produced 6′-SL by a reaction using a sialic acid dimer as the donor and lactose as the acceptor, with a yield of 26.2 %.

As shown in these examples, the transglycosylation reaction yield using the recombinant glycosidase was not improved significantly but remained within the range of 20–40 %.

2.4. Reactions using glycosynthase: glycosidase eliminated hydrolytic activity

The low yield of the transglycosylation reaction using glycosidases from natural sources or from recombinant enzymes is caused by the hydrolysis of the donor sugar and products during the reaction by the glycosidases used. To address this issue, Withers and co-workers introduced the concept of “glycosynthase” in 1998 [69]. The amino acid residue l-Glu 358 was responsible for the hydrolytic activity of β-glucosidase from Agrobacterium sp. Withers et al. substituted this nucleophilic l-Glu at the active site with non-nucleophilic amino acids such as l-Ala or l-Ser, creating the mutants E358A and E358S, respectively. These enzymes are incapable of hydrolyzing glucosidic bonds; therefore, α-glucosyl fluoride for example, which mimics the high-energy covalent intermediate, is used as the donor substrate for the β-glucosyl synthase. In the reaction using the E358S mutant with Glc-α-F as the donor and Glc-β-pNP as the acceptor, Glcβ1–4Glc-β-pNP was obtained at >90 % yield [70].

Following their report, many glycosynthases have been created [71,72] including β-mannosyl synthase using Man-α-F as the donor [73] and α-fucosyl synthase using Fuc-β-N₃ as the donor [74]. These engineered enzymes have enabled the synthesis of complex carbohydrates without hydrolysis of the donor and products; consequently, the glycosylation reaction yield was improved significantly.

2.5. Reactions using protein-engineered glycosidases and various strategies

After the reports by Withers et al., the use of protein engineering technology for glycosidases was accelerated with a strategy different from that of Withers et al. One potential strategy was the “directed evolution” proposed by Feng et al. [75]. Briefly, the genes encoding amino acids near the active site were randomly mutagenized and screened for variants displaying improved transglycosylation properties. After several cycles of in vitro recombination or even just one cycle, the mutant enzyme showed a much higher transglycosylation activity/hydrolysis activity ratio. Using this strategy, Osanjyo et al. converted α-l-fucosidase from Thermotoga maritima into α-l-transfucosidase variants, T264A, Y267F, L322P, which were all located within the second amino acid shell of the active site. In the transglycosylation reaction using Fuc-α-pNP as the donor and Gal-β-pNP as the acceptor, Fucα1–2Gal-β-pNP was obtained with a yield of >60 % compared to 12 % in the reaction using the wild-type enzyme [76]. Vuillemin et al. [77] initially expressed 13 variants of lacto-N-biosidase from Bifidobacterium bifidum JCM 1254. Among them, five LnbB variants–W394F, W394H, Y419N, W465F, and W465H– were selected on the basis of LNT yield in the transglycosylation reaction using Galβ1–3GlcNAc-β-oxazoline as the donor and lactose as the acceptor. These variants were further screened by monitoring transglycosylation reactions using Galβ1–3GlcNAc-β-pNP as the donor and lactose as the acceptor. Ultimately, the mutant W394F was selected as it afforded the best yield of 91 %.

Another strategy, “loop engineering”, was proposed by Zeuner et al. [78,79]. The α1–3/4-fucosidases from Bifidobacterium bifidum (BbAfcB) and Clostridium perfringens (CpAfc2) have transglycosylation activity that can produce Galβ1–4(Fucα1–3)GlcNAcβ1–3Galβ1–4Glc (LNFP-II) by a reaction using Galβ1–4(Fucα1–3)Glc (3-FL) as the donor and LNT as the acceptor. Although the original CpAfc2 exhibits higher activity than that of BbAfcB, C. perfringens is highly pathogenic. Zeuner et al. replaced the α-helical loops consisting of 17 amino acids of CpAfc2 with the corresponding 23 amino acids close to the active site of BbAfcB. In the reaction using the engineered recombinant enzyme, the yield of LNFP-II increased from 14 % to 39 % with the recombinant enzyme compared to the original BbAfcB. In addition, Yang et al. recently reported that the transglycosylation reaction yield for synthesizing LNFP-II using wild-type SpGH29^C α1–3/4-fucosidases increased to 91 % from the previously reported value of 51 % by controlling the solvent pH and reaction time [80].

Jers et al. [81] cloned the genes encoding a seven amino acid motif (197–203) at the border of the substrate-binding cleft of pathogenic trans-sialidase from Trypanosoma cruzi. They expressed this motif in the α-sialidase from Trypanosoma rangeli to create the non-pathogenic mutant Tr13. This mutant, whose amino acids at 197–203 of T. rangeli were substituted from IADMGGR to VTNKKKQ, was used in a transglycosylation reaction with casein glyco-macropeptide (CGMP), separated from cheese whey, as the donor and lactose as the acceptor, giving Neu5Acα2–3Galβ1–4Glc in 31 % yield.

From the perspective of using HMOs for food additives, the source of sugar donor in the transglycosylation reaction is important. It would be preferable to use, for example, CGMP, fetuin, or sialic acid dimer for the sialylation reaction; partial hydrolysates of fucoidan (fucosyl polysaccharide isolated from brown seaweed) for the fucosylation reaction; and oxazoline derivatives for attaching GlcNAc derivatives.

2.6. Reactions using glycosyltransferases

Glycosyltransferase is an enzyme that governs the synthesis of sugar chains in living organisms by using sugar nucleotides as donors. Therefore, the reaction yield is very high in vivo, though it is not always 100 %. For the synthesis of d-Gal-, l-Fuc-, d-GlcNAc-, and l-Neu5Ac-linked Os by transglycosylation reaction, UDP-Gal, GDP-Fuc, UDP-GlcNAc, and CMP-Neu5Ac are used, respectively, as the sugar donors. These sugar nucleotides and most of glycosyltransferases are commercially available, though they can be costly. Utilizing high reaction yield and high regioselectivity of the glycosyltransferase assisted reaction, many HMOs have been synthesized at laboratory scale [82]. Two examples of the syntheses of long-chain and complex HMOs are described below.

Miyazaki et al. synthesized Fucα1–2Galβ1–3(Fucα1–4)GlcNAcβ1–3Galβ1–4Glc (LNDFH-I) by a reaction using three glycosyltransferases and one glycosidase as shown in Fig. 2 [83]. In step 1; β−1,3-N-acetylglucosaminyltransferase (β−1,3-GnT), in step 3; α−1,2-fucosyltransferases (FUT1), and in step 4; α-(1,3/1,4)-fucosyltransferase (FUT3) were used. In step 2, recombinant β−1,3-galactosidase from B. circulans, bgaC [[66], [67]] was used instead of β1,3-galactosyltransferase because the appropriate enzyme was not easily obtainable. The isolation yields for the four steps were 44 %, 22 %, 71 %, and 85 %, respectively.

Fig. 2.

Enzymatic synthesis of lacto-N-difucohexaose-I (LNDFH-I) [83]. The enzymes used are; Step 1: β−1,3-N-acetyl glucosaminyl transferase, Step 2: recombinant β-galactosidase from Bacillus circulans [66,67], Step 3: FUT 1 (GeneBank No M35531), Step 4: FUT 3 (GeneBank No 81,485).

Prudden et al. [84] synthesized an HMO consisting of 15 sugars by the reaction using glycosyltransferases. The structure of the synthesized oligosaccharide and the order of the reaction steps are shown in Fig. 3 together with the used enzymes and sugar nucleotides. It is noteworthy that a high molecular-weight HMO composed of 15 sugars was synthesized through a 12-step reaction using only eight glycosyltransferases. As shown in Fig. 3, the order of the reaction was carefully considered. If the order of any reaction step is changed, the structure of the product would be completely different from that of the target compound.

Fig. 3.

Syntheses of triantennary HMOs composed of 15 saccharides [84]. (1) – (12) denote the order of the reactions. (1) UDP-GlcNAc, β1,3-N-acetylglucosaminyltransferase 2 (B3GNT2), (2) UDP-Gal, β1,4-galactosyltransferase from bovine milk (GalT 1), (3) UDP-GlcNAc, N-acetyllactosaminide β1,6-N-acetylglucosaminyltransferase GCNT2, (4) CMP-Neu5Ac, α−2,6-sialyltransferase 1 (ST6Gal 1), (5) UDP-Gal, GalT1, (6) UDP-GlcNAc, (B3GNT2), (7) UDP-Gal, GalT1, (8) UDP-GlcNAc, GCNT2, (9) UDP-Gal, GalT1, (10) UDP-GlcNAc, (B3GNT2), (11) UDP-Gal, GalT1, (12) GDP-Fuc, galactoside α1,2-fucosyltransferase 1 (FUT1).

As mentioned above, the reaction involving glycosyltransferase is effective for the syntheses of complex or long chain HMOs at the laboratory scale. Although the synthesis of Os using glycosyltransferases generally yields high results, the high cost of enzyme and sugar nucleotides limits their use in the practical production of HMOs. The challenges in solving these issues through bacterial production system started around 2000 [85], and are discussed in Section 4.

2.7. Reactions using phosphorylase with multiple-enzyme combinations

Nishimoto and Kitaoka demonstrated a new strategy for the synthesis of HMOs using phosphorylase instead of glycosidases or glycosyltransferases [86]. The principle of the process is based on the fact that lacto-N-biose (LNB, Galβ1–3GlcNAc) is phosphorolyzed by Galacto-N-biose/Lacto-N-biose-1-phosphorylase (GLNBP) to yield Galα−1-phosphate (Galα−1-P) and GlcNAc. Since this reaction is reversible, LNB can be synthesized from Galα−1-P and GlcNAc. To generate Galα−1-P, recombinant UDP-galactose–hexose-1-phosphate uridyltransferase (Gal-T), UDP-glucose-4-epimerase (Gal-E), and sucrose phosphorylase (SP) are involved. Finally, in a one-pot reaction utilizing these four recombinant enzymes, LNB was obtained with an 83 % yield from GlcNAc, sucrose, and a catalytic amount of UDP-Glc and inorganic phosphate (Pi) (Fig. 4) [87]. Furthermore, 1.5 Kg of LNB was obtained from a 10-L scale reaction [88]. Additionally, aiming to use the produced LNB as a food additive, they reported a new enzyme system from natural strains of B. breve MCC1320 and B. longum subsp. longum MCC135 to synthesize LNB with a yield of 77 % [89].

Fig. 4.

Multi-enzymatic production of LNB from sucrose and GlcNAc [86]. Abbreviations: SP; sucrose phosphorylase, GLNBP; Galacto-N-biose/Lacto-N-biose-1-phosphorylase, Gal-T; UDP-glucose–hexose-1-phosphate uridyltransferase, Gal-E; UDP-glucose-4-epimerase.

2.8. Chemical and chemoenzymatic reactions

Since the 1960s, numerous studies on the chemical syntheses of various HMOs have been reported, including kilogram-scale synthesis of 2′-FL by Agoston et al. [90]. Recently, Bandara et al. synthesized a GlcNAcβ1–3Galβ1–4Glc (LNTri- II) derivative by combining protected GlcNAc with a protected lactose derivative, followed by combining protected d-galactose derivative with the aforementioned trisaccharide derivative to obtain LNT [91]. In a successive synthesis of Galβ1–4GlcNAcβ1–3Galβ1–4Glc (LNnT) [92], Galβ1–3GlcNAcβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4Glc (LNH) [92], Galβ1–4GlcNAcβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4Glc (LNnH) [93], Bandara et al. used the concept of “block synthesis” to systematically combine rationally protected mono-, di-, or trisaccharide blocks. This strategy is effective for the construction of an HMO library.

In chemical synthesis, it is necessary to protect all hydroxyl groups with appropriate protecting groups, except for one hydroxyl group involved in the subsequent coupling reaction. When proceeding to the next step of the reaction, another single protecting group on the specific hydroxyl group of the acceptor sugar must be removed selectively. Since these processes are cumbersome and time-consuming, the combination of chemical synthesis and enzymatic synthesis, namely chemoenzymatic synthesis, was developed.

Ooi et al. first synthesized a GlcNH₂β1–3(GlcNAcβ1–6)Galβ1–4Glc derivative using the organic chemical method [94]. After deprotection, d-Galactose was attached to GlcNAc residue enzymatically using β1,4-galactosyltransferase from Helicobacter pylori 26695 (HP0826) with UDP-Gal. Subsequently, N-acetylation of the glucosaminyl residue was performed by the addition of acetic anhydride in aqueous sodium bicarbonate to obtain GlcNAcβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4Glc. Then, an enzymatic reaction using β1,3-galactosyltransferase from Escherichia. coli O55 (WbgO) with UDP-Gal was performed to obtain LNH.

Thus, the chemoenzymatic process considerably decreases the number of operations needed for protection and deprotection.

2.9. Summary

In Section 2, we summarized the methods for preparing HMO libraries, including isolation from human milk and in vitro syntheses by enzymatic, chemical, and chemoenzymatic methods. The oligosaccharide libraries prepared by these methods are expected to provide resources for functional studies or mechanistic studies on lectins such as galectins, selectins, and siglecs. Moreover, HMOs synthesized by enzymatic methods using natural sources are expected to extend their use as food additives.

3. Biological functions of HMOs

The in vitro studies to clarify the biological functions of HMOs were previously performed only with the HMO mixtures, which had been separated from human milk. The library of each HMO, separated and purified as in Section 2, have been utilized to study the sugar epitope of the glycoproteins/glycolipids on the host cells to which pathogenic microorganisms attach and to characterize the glycan affinity for some lectins, which are related to immune reactions. Such an HMO library could be used only for the in vitro study for HMO functions. Some HMOs, including 2′-FL, 3-FL, DFL, 3′-SL, 6′-SL, LNT, and LNnT, have been manufactured at plant scale by the fermentation method with the recombinant bacteria at present, and then used in vivo or preclinical studies. In Section 3, the biological functions, which have been observed and described by in vitro studies with epithelial cells or by in vivo studies with experimental animal models, will be introduced.

The following functions of HMOs have been reported: anti-infection against pathogenic microorganisms, strengthening of colonic barrier function, prevention of necrotizing enterocolitis (NEC), anti-inflammatory by immune modulation, enhancement of infant cognitive ability, and improvement of infant malnutrition, among others. Two potential mechanisms can be hypothesized: functional, caused by the fermentation products of HMOs by colonic beneficial bacteria, and by the interaction of HMOs with colonic epithelial cells or blood cells during the circulation of HMOs after absorption at low levels. The metabolisms of HMOs by Bifidobacteria, which are abundant beneficial colonic bacteria, have been described in previous reviews [2,32,38].

3.1. Anti–infection against pathogenic bacteria or viruses

Given that the chemical structures of HMOs are similar to those of the carbohydrate moieties of the glycoconjugates on epithelial cells, it has been hypothesized that HMOs may inhibit infections as the decoy receptors to which pathogenic bacteria and viruses attach. Such inhibitory ability has been traditionally studied by in vitro by co-culturing the cells with viral or bacterial strains and HMOs. After the epithelial cells are cultured with the bacteria or viruses in addition to HMOs, the attached amounts of the microorganisms on the cells are measured to compare between the cultures in the presence or absence of HMOs. When the attached levels are smaller in the culture in the presence of HMOs, one can speculate that the HMOs functioned as a decoy receptor against the infection.

For example, the following study was performed in vitro with a pathogenic virus. Laucirica et al. cultured MA104 cells from the kidney of an African green monkey with human rotavirus G1P(4) or G2P(4) strains in the presence/absence of 2.5 or 5.0 mg/mL of 2′-FL, 3′-SL or 6′-SL, and then counted the attached levels of the virus on the cells after the incubation [95]. These levels were lower with HMOs than without HMOs. However, when another rotavirus strain G10P(11) was used in a similar experiment with MA104 cells and 0.15–10 mg/mL of LNT or LNnT, the attached level was higher in the culture with LNT or LNnT than that in the control culture [96]. Consequently, caution is needed to decide if the attachment in vivo was inhibited in the presence of HMOs.

A new coronavirus SARS-Cov-2 uses the angiotensin-converting enzyme 2 receptor on the host cells to infect by binding to the receptor-binding domain (RBD), S-protein. Sheng et al. preincubated 1 mg/mL each of 6′-SL, 3′-SL, LNnT or 2′-FL with fluorescence-labeled RBD and then incubated with HepG2 cells from human liver cancer to investigate the inhibition of RBD binding [97]. By quantifying immune signal intensity, they observed increasing inhibition in the presence of HMOs in the order of strength by 6′-SL, LNnT and 2′-FL.

Similar in vitro experiments to assess inhibition of the attachment to the cell also have been performed with pathogenic bacteria. Manthey et al. cultured the enteropathogenic E. coli strain with HeLa cells, Hep-2 cells, or T84 cells in the presence or absence of HMO mixture from human milk, and then counted the attached levels of the E. coli strain on the cells [98]. The attached levels were lower in the culture with HMOs than that without HMOs. Kong et al. preincubated Caco-2 cells from human colonic cancer with 2 mg/mL of 2′-FL or 3-FL and then incubated with several E. coli or Klebsiella pneumoniae strains in the logarithmic- or constant-rate growth phase to count the attached bacterial levels to the cells [99]. Exposure of Coco-2 cells to 2′-FL decreased the attachment of the E. coli 0119 strain in the logarithmic-growth phase relative to that without 2′-FL, whereas the exposure of the cells to 3-FL did not decrease the attachment.

It has been shown that HMOs inhibit bacterial toxin binding to colonic epithelial cells. Cholera toxin (CT), which is secreted by Vibrio cholerae, has two subunits: when subunit B (CTB) binds to colonic epithelial cells surfaces, it is imported into the cell via endocytosis, and then cAMP is produced in the cytosols. Storage of cAMP in cytosol stimulates secretion of ions in the colon that causes diarrhea by elevating osmotic pressure. The cell-surface receptor to which CTB attaches is the glycolipid GM1a (Galβ1–3GalNAcβ1–4[Neu5Acα2–3]Galβ1–4Glc-Cer). However, the GM1a expression level is low on the surface of human colonic epithelial cells, so Wands et al. hypothesized that CTB should use another potential receptor on the cells for the attachment [100]. After the epithelial cell lines, colo205 or T84, were incubated with CTB in the presence/absence of 100 mM of several HMOs or l-fucose, the attached levels of CTB to both cells were compared by flow cytometry. The binding level of CTB was significantly lower in after incubation with 2′-FL, 3-FL or l-fucose than that without them, whereas inhibition was not observed when 3′-SL or 6′-SL was added into the culture. The inhibition level was stronger by 2′-FL than by 3-FL or l-fucose. When the incubation was performed after addition of 1 mM of DFL or 2′-FL, the inhibition level for CTB binding to the cells was higher with DFL than with 2′-FL. These results suggest that DFL or 2′-FL in human milk can prevent the serious diarrhea in breast-fed infants caused by infection with V. cholerae.

Inhibition by immune enhancement of infection caused by some viruses has been studied in vivo in model animals fed HMOs. Comstock et al. fed a milk substitute containing HMO mixture to piglets and then orally administered the rotavirus strain to them [101]. A few days later, the peripheral blood, intestinal lymph nodes, and Peyer's patch were collected, and then the counts of the immune competent cells, including interferon γ (IFN-γ)-producing cells and natural killer cells, were compared with those in the control animals not fed HMOs. The counts of the IFN-γ producing cells in peripheral blood were twice as high as those in the controls, suggesting that cellular immunity was enhanced by adding HMOs to the feed. Xiao et al. administered a diet containing 0.25–5 % of 2′-FL to 6-week-old female mice, and 2 weeks after the mice were vaccinated in their right ears with inactivated influenza virus and in their left ears with a subcutaneous injection of water [102]. The animals were re-vaccinated 9 days later, and the immune reaction was evaluated by measuring the difference between the ears’ thickness before re-vaccination and 24 days after vaccination. The 2′-FL fed animals showed dose-dependent the immune enhancement relative to the control animals, and the vaccine-specific IgG1 and IgG2 concentrations were higher than those of the controls. These results showed that both the cellular and acquired immunities were increased by feeding 2′-FL after the vaccination of influenza virus, suggesting that addition of 2′-FL can enhance the vaccination effect against influenza infection.

3.2. Anti–infection against pathogenic bacteria or viruses

Recently, HMOs have been shown direct growth inhibition of specific pathogenic bacteria, after culturing with HMOs. Some group-B Streptococcus (GBS) strains, which cause meningitis, and sometime sepsis, were incubated in the presence/absence of HMOs and the growth rates between incubation with and without HMOs were compared. The growth rate was slower with neutral HMOs than without them. Growth inhibition occurred after addition of specific HMOs, such as LNT, LNFP-Ⅰ, or LNDFH-Ⅰ [103].

Craft et al. studied the growth inhibition of GBS caused by addition of several antibiotics with 1 % of HMO mixture in the culture broth [104]. The minimum inhibitory concentration (MIC) by gentamycin or erythromycin decreased by 1/32 level with addition of HMOs compared with the incubation in absence of HMOs. The MIC was decreased by 1/16 level for clindamysin by addition of 1 % HMOs.

Whether or not such growth inhibition for GBS occurs when HMOs are fed, one needs to investigate a future clinical study using human-identical milk oligosaccharides (HiMOs).

3.3. Strengthening of colonic barrier function

It is thought that the secretion level of epithelial cell-surface mucin is increased by addition of HMOs to the feed during intestinal maturation, which strengthens colonic barrier function.

Barnett et al. reported the expression levels of mucin on epithelial cells in small and large-intestinal models with different ratios of Caco-2 cells and HT-29-MTX cells cultured in the presence/absence of an Os mixture separated from caprine milk [105]. They also measured the transepithelial electrical resistance (TEER) between incubation in the presence or absence of milk oligosaccharides (MOs). When the culture was incubated with MOs, the mucin expression level and TEER were increased. Although some MOs, such as sialyllactose (SL), exist in the milks of goats and humans, their MO profiles differ [106].

Perdijk et al. studied the cell-cycle control, growth rate, and differentiation of Caco-2 cells after addition of 10 mg/mL of SL, or galacto oligosaccharides (GOS) into the culture broth [107]. They also studied the epithelial regeneration affected by addition of SL or GOS to Ca9–22 cells from squamous cell carcinoma in the upper gingiva, after cellular damage. The expression levels of the genes of polo-like kinase, which help induce regulation of mitosis, were affected by the addition of SL to the culture broth, and the activity of alkaline phosphatase was also increased. The counts of Ca9–22 cells were increased in the damaged part of the growth broth by addition of SL.

Natividad et al. studied the effect on strengthening of the colonic epithelial barrier function in vitro by addition of a blend of six manufactured HMOs - 2′-FL (55 %), 3′-SL (7 %), 6′-SL (9 %), LNnT (5 %), LNT (18 %), DFL (6 %) (HMO6) - the ratios of which are similar to those in human milk [108]. After the mixture of Caco-2 and HT29-MTX cells was cultured in the presence/absence of 30 mg/mL or 60 mg/mL of the blend, the effect on the barrier function by addition of HMO6 was assessed by measuring TEER or fluorescein-isothiocyanate (FITC)-labeled dextran 4 kDa (FD4) translocation. The TEER value was higher when the blend was added to the culture broth than when it was not. The barrier function was reduced, as shown by the TEER value and FD4 translocation, when the culture was challenged with the inflammatory cytokines TNF (tumor necrosis factor)-α and INF-γ, but was protected by adding HMO6 to the culture broth. Next, protection against inflammatory disorder was assessed by addition of some HMO5 (in six separate experiments in which one of the six HMOs was removed from HMO6 in each experiment) into the broth and compared with the protection conferred by the culture containing all six HMOs (HMO6). The protection level was lowest in the culture containing HMO5 in which 2′-FL had been removed, suggesting that 2′-FL was most effective to protect the epithelial barrier function against inflammatory disorder.

Natividad et al. studied the protection of barrier function against inflammatory disorders by addition of the culture supernatant after incubation of HMO6 with fecal microflora obtained from a breast-fed infant [42] in an in vitro experiment similar to the previously described one with Caco-2: HT29-MTX cells. The protection was observed in the culture in the presence of the supernatant. It was suggested that butyric acid, which was produced by the fermentation of HMO6 by fecal microflora of the infant, might be effective for protecting this function against the disorder caused by the inflammatory cytokines [42].

Boll et al. assessed which HMOs among 2′-FL, 3-FL, LNT, LNTri-Ⅱ, 3′-SL and 6′-SL could enhance colonic barrier integrity or modulate immune responses [109]. To assess the effect on colonic barrier integrity, the TEER value was measured for the culture of Caco-2 cells, which had been incubated in the presence/absence of 1, 5, or 20 mg/mL of each HMO in the transwell. The TEER value increased in the cultures with fucosyl or nonfucosyl neutral HMOs, and did not increase in the cultures with sialyl HMOs (SHMOs). In addition, the secreted levels of some cytokines or chemokines were assayed in the supernatant of the HT29 cell cultures, incubated in the presence/absence of 0.1, 1, or 10 mg/mL of HMOs. These were also assayed in LPS-activated dendritic cells (DCs) or M1 macrophages (M1Mϕs) cultures, incubated in the presence/absence of 1, 2.5, or 5 mg/mL HMOs. The secretion levels of CXCL (C-X-C motif chemokine ligand) 10, CCL (chemokine ligand) 20, and CXCL8 increased in the culture of HT29 cells with two SHMOs, whereas those of IL (interleukin)−10, IL-12p70, and IL-23 increased in the cultures of DCs and M1Mϕs with the SHMOs. These activation effects were either not observed or observed at low levels in the cultures with fucosyl and nonfucosyl neutral HMOs. These results suggested that the colonic barrier integration is stimulated by the effect of neutral HMOs, whereas immune modulation in the colonic epithelium or in the circulation is mediated by the effect of SHMOs [109].

3.4. Prevention of necrotizing enterocolitis (NEC)

NEC is a serious disease in preterm infants whose body weights are <1500 g at birth and sometimes causes death. Even after recovery from NEC, nerve-brain development is sometimes delayed in infants with NEC. Since the frequency of NEC is lower in breast-fed infants than in bottle-fed infants, it has been hypothesized that some components in human breast milk should prevent NEC development. Recently, it has been shown that HMOs potentially can prevent NEC.

Jantscher-Krenn et al. induced NEC development in neonatal rats after birth by administration of a pro-inflammatory agent to the pregnant mother rats [110]. Then, the neonatal rats were fed a milk substitute with or without an HMO mixture, and the inflammation scores between animals fed or not fed (control) HMOs after they were sacrificed and their colons collected were compared. The survival rate was higher in the HMOs-fed animals than in the control animals, and the inflammation score was improved in the HMOs-fed rats than in the controls. This anti–inflammatory effect was also observed when the rats were fed disialyllacto-N-tetreoase (DSLNT), a two Neu5Ac-containing HMO.

Another study also showed that 2′-FL prevented NEC development for neonatal mice. Good et al. induced NEC for neonatal mice by administration of the milk substitute supplemented with the colonic microflora that had been collected from human infants with NEC, with hypoxic exposure [111]. The mice were fed 5 mg/mL of 2′-FL, and then the damage in the intestinal mucosa was scored after sacrifice and colon collection. The inflammation score was improved in the 2′-FL fed mice relative to that in the control animals. In addition, the fluorescent substance was injected intracardially into the mice, and then the collected colons were observed for blood perfusion under a microscope after staining with an endothelial cell marker. The results showed that the blood perfusion recovered in the 2′-FL fed mice since the expression of endothelial nitric oxide synthase was maintained. It is noteworthy that NEC was derived for the mice by administration of colonic microflora, collected from human infants with NEC. Given that NEC development is related to the dysbiosis of colonic microflora, prevention of NEC by administering 2′-FL might be due to the recovery from dysbiosis of colonic microflora.

The potential mechanisms prevention of NEC by feeding HMOs have been explored in a few studies. One potential mechanism might be the result of a strengthened colonic barrier function. Wu et al. induced NEC by hypoxic exposure along with feeding of lipopolysaccharide (LPS) to the mice, and they were then fed with the milk substitute in the presence or absence of 20 mg/mL HMO mixture [112]. After scarifying and collecting the colon, inflammation was scored, which showed that it was improved in the HMOs-fed mice relative to that in the control. In addition, after addition of blood FITC into the colon, the colon was ligated and soaked in water, and then the concentration of FITC released from the colon was determined. The concentration was lower in the HMOs-fed animals than in the controls, which showed that the barrier function was strengthened by feeding HMOs. The counts of epithelial cells producing Muc2+ were higher in the HMOs-fed animals than those in the controls. These results suggest that the prevention of NEC by feeding HMOs might be related to the strengthening of the colonic barrier function.

Another potential mechanism is related to control of the TLR (toll-like receptor) 4/KN (nuclear factor)-κB signaling inflammation pathway. Zhang et al. induced NEC by hypoxic exposure at low temperature in neonatal rats, which were subsequently fed a milk substitute with or without 1500 mg/L sialyl HMOs (SHMOs), and scored the inflammation of the colon after scarification [113]. The inflammation was improved in the SHMOs-fed rats relative to that in the controls. In addition, although the expression levels of the inflammatory cytokines, TLR4 and NLRP (NOD-like receptor) 3, and the activity of caspase-1 were increased by induction of NEC, they were decreased after feeding of SHMOs.

Another study determined whether or not the cranial nerve damage caused by development of NEC could recover after oral administration of 2′-FL or 6′-SL to neonatal mice. Sodhi et al. induced NEC for neonatal mice by gavage of the milk substitute, supplemented by colonic microflora from humans with NEC with hypoxic exposure [114]. Then, after oral administration of 2′-FL or 6′-SL to the mice, the brains were collected to compare the expression levels of the inflammatory cytokines TNF, IL-6, 3′-NT (nitrotyrosine), or DHE (dihydroethidium) as well as some brain-damaging inflammatory makers between 2′-FL- or 6′-SL-fed animals and nonfed controls. The expression levels of these inflammation cytokines were lower in the 2′-FL- or 6′-SL-fed animals than in the nonfed controls. In addition, the swimming learning ability was tested by the Morris test to determine if the reduced ability caused by developing NEC would recover in the animals fed 2′-FL or 6′-SL. The results showed that this ability was recovered by oral administration of 2′-FL or 6′-SL. In addition, the brain organoids, which had been generated from mice embryo, were cultured in addition of 2′-FL or 6′-SL with the supernatant of the growth broth of the colonic microflora from human patients who developed NEC, and then the expression levels of inflammatory cytokines and brain-damaging inflammatory markers were determined to compare the expression levels between the organoid cultured with or without 2′-FL or 6′-SL. The expression levels of TNF, IL-6, 3′-NT, or DHE, which had been increased by NEC induction, were decreased, and the damage to myelin was recovered in the corpus callosum and cerebellum by addition of 2′-FL or 6′-SL. These results suggest that the brain damage by NEC development could be repaired by suppressing brain inflammation after administration of 2′-FL or 6′-SL.

3.5. Immune-modulation effect

It has been hypothesized that HMOs have immune-modulation effects on intestinal epithelial cells or on blood cells during the circulation process after small parts are absorbed and enter the circulation.

Some studies have explored the anti-inflammation effect by HMOs. Castillo-Courtade et al. orally administered 2′-FL or 6′-SL to ovalbumin (OVA)-sensitized mice, who had a food allergy, with OVA and then compared the frequencies of diarrhea, malnutrition scores, secretion levels of inflammatory cytokines from spleen cells, and counts of mast cells with those of the control animals who were not feed 2′-FL or 6′-SL [115]. The results showed that the diarrhea frequency and malnutrition were improved, and mast cell counts and the production of inflammatory cytokines as TNF were decreased in the 2′-FL or 6′-SL-fed mice.

The mechanisms by which the anti–inflammatory effect occurred after feeding HMOs were studied in vitro with blood cells or endothelial cells. Newburg et al. cultured platelet cells with and without 2′-FL, 3-FL, 6′-SL, LNnT or DFL, and then after activation by thrombin, the production levels of chemokine RANTES (regulated on activation, normal T cell expressed and secreted) and soluble tumor necrosis factor-like CD 40 ligand (sCD40L) were determined [116]. The production levels of RANTES were much lower in the culture broth with DFL than in the control culture without HMOs.

Some studies have explored the anti-inflammatory effect on intestinal epithelium by addition of HMOs in vitro to epithelial cells. To clarify the mechanism by which HMOs alleviate a food allergy, Zehra et al. studied whether or not the inflammatory response after exposure to an antigen–antibody complex (Ag-IgE) on T84 or HT29 cells would be changed by addition of 2′-FL or 6′-SL in the culture broth [117]. The expression level of pro-inflammatory cytokine IL-8 in the presence of 10 mg/mL of 2′-FL was lower in both cell cultures than in the control broth without 2′-FL. In the culture broth of T84 cells, the expression levels of CCL20 and IL-8 were reduced in the presence of 10 mg/mL of 6′-SL. Cheng et al. determined the expression level of IL-8 in the supernatant of the culture broth of FHs 74Int cells, an immature colonic epithelial cell, and T89 cells, an adult-type colonic epithelial cell, which had been incubated with or without TNF-α, in the presence or absence of 2′-FL, 3-FL, LNT, LNnT, DFL, or LNTri-Ⅱ [118]. The results showed that the expression level of IL-8, elevated by the effect of TNF-α, was reduced in the culture broth of FHs 74Int cells in the presence of 3-FL, LNnT or DFL.

The immune-modulation effect of HMOs feeding on intestinal epithelium has been studied in model animal experiments. Rosa et al. administered 15 mg/day of an HMO mixture from human milk for 7 or 14 days to sterile mice from 21 days after birth and then performed morphological analysis of intestinal tissues, transcriptome analysis, and flow cytometry analysis of kidney and mesenteric lymph nodes (MLNs) at 28, 35, and 50 days after the start of HMOs feeding [119]. The small intestinal crypt was reduced at 28 and 35 days after HMOs feeding relative to that of the HMOs nonfed control animals, and the small intestinal villus height and depth of the colonic gland were decreased at 35 days after the feeding. The gene expression levels, which were related to exocellular matrix formation, ubiquitination of the proteins, nuclear transport, or mononuclear-cell differentiation, were increased by HMOs feeding relative to those of the HMOs nonfed control animals. The cell counts of CD4+T and CD8+T in the kidney and MLNs were increased more in the HMOs-fed animals at 50 days after the start of HMOs feeding than in the controls. These results suggested that the immune modulation affected by HMOs feeding occurred without colonic microflora mediation because sterile animals were used in this study.

To determine if HMOs feeding could alleviate the inflammation, caused by pathogenic bacterial infection, some studies have been performed in vitro experiments with colonic epithelial cells or in vivo experiments that administered HMOs to the animals. When enteropathogenic E. coli (EPEC) strains invade epithelial cells, bacterial LPS induces IL-8 expression, and then subsequent inflammation is caused by a reaction related to the action of TLR4. He et al. studied the potential reduction of IL-8 expression levels by exposure to 2′-FL in vitro study with T84 cells and in vivo with mice, both of which had been infected by EPEC [120]. In the in vitro study, the IL-8 expression level was reduced by exposure to 2′-FL that depended on the culture incubation time. Although the colonic length of the mice had shortened by EPEC infection in the in vivo study, feeding with 2′-FL led to recovery to a colonic length similar to that of the uninfected mice. The expression level of CD14, which had been increased by infection with EPEC, was reduced when the mice were fed 2′-FL. These results show the potential mechanisms for the alleviation of colonic inflammation by feeding with 2′-FL, even after EPEC infection.

3.6. Enhancement of cognitive ability

It is hypothesized that the enhancement of cognitive ability by feeding HMOs, as shown by in vivo experiments with mice, rats, and piglets, is achieved through two potential mechanisms. First, Neu5Ac in SHMOs is utilized to form gangliosides or sialyl glycoproteins in the brain after absorbance in the small intestine and metabolism in the liver, and then transfer of the metabolic product to the brain after passing through the blood–brain barrier. Second, the metabolic products produced from HMOs by colonic bacteria may stimulate brain activity after the transfer to the brain by colon-brain correlation via the vagus nerve.

The data from some studies suggest that the cognitive enhancement occurs by the first mechanism. Jacobi et al. administered 2 or 4 g/L of 3′-SL or 6′-SL to piglets for 21 days. Then, after scarifying and collecting the brain, the concentrations of sialic acid in the cerebral cortex, cerebellum, corpus callosum, and hippocampus were compared with those of control animals who had not been administered 3′-SL or 6′-SL [121]. The concentration of ganglioside in the cerebellum increased in the 3′-SL-fed animals, whereas that of sialic acid increased in the corpus callosum in 3′-SL- and 6′-SL-fed animals. Mudd et al. administered a milk substitute containing 180–760 mg/L of sialyllactose (SL) to piglets and then studied brain-tissue development to measure the corpus callosum and left-hippocampus axial span areas and radial directions by magnetic resonance imaging (MRI) [122]. The results showed that these tissue areas were bigger in the SL-fed animals at the mid-level administration than in the control animals who were not fed SL. The finding that the ratio of the concentration of free sialic acid to bound sialic acid in the hippocampus was higher in the SL-fed animals than that in the control animals suggested that the increase in the left-hippocampal area was related to the higher ratio in the SL-fed animals.

Other studies have suggested that cognitive enhancement occurred by the second mechanism. Vazquez et al. administered 2′-FL to mice and rats that had been implanted with an electrical node in the hippocampus and then measured the synaptic connection strength after electrical stimulation [123]. The ability of the mice to explore and learn the location of drinking water in an intellicage was studied, whereas the rats were studied for their ability to learn how they could find the foods in a Skinner box to compare their learning scores with those of the controls that were not fed 2′-FL. The results showed that the long-term persistence (LTP) of synaptic connection strength was stronger in the 2′-FL-fed animals than in the controls, and the learning scores were higher in the 2′-FL-fed animals, too. To determine if this cognitive enhancement by 2′-FL administration was due to the effect of brain stimulation by the metabolic products of 2′-FL produced by colonic bacteria after the transfer from the colon to the brain via the vagus nerve, Vazquez et al. tested the similar learning ability of male rats at 2.5–4 months of age whose vagus nerve had been cut to compare their scores with those of the test animals that had received sham operations [124]. They found that the LTP and learning ability were enhanced in the sham-operation group by 2′-FL administration, but were not enhanced in the cut vagus group. This finding suggests that the enhancement by 2′-FL feeding was achieved by the colonic bacteria metabolic products effect.

Hauser et al. conducted a study that used St6gal1 (a gene related to the biosynthesis of 6′-SL) knockout (KO) female mice [125]. To deprive lactating offspring of 6′-SL, they cross-fostered newborn wild-type (WT) pups to KO dams who provide 6′-SL-deficient milk. To test if lactational 6′-SL-deprivation affected cognitive capabilities when the mice grew into adults, the researchers assessed attention, perseveration, and memory. The results showed that the 6′-SL-deficient milk-drinking mice had lesser abilities related to novel target exploration, attention, and spatial recognition than the mice who had drunk the WT mother's milk. To investigate the underlying molecular mechanisms, they assessed gene expression (at eye-opening and in adulthood) in two brain regions mediating executive functions and memory. The results showed that the expression levels of the genes for biosynthesis of gangliosides and poly-sialylated nerve-cell adhesion factor, which are related to the formation of the neural signaling pathway, were lower in the 6′-SL-deficient milk-drinking mice than in the controls.

Cognitive enhancement by feeding of HMOs has been shown in the study with experimental minipigs. Clouard et al. studied cognitive ability for tasks to assess the anxiety, motivation, appetite, learning, and memory by feeding of SHMOs to Gottingen minipigs in early life [126]. The animals were fed with a milk substitute in the presence or absence of 4 g/L of SHMOs from 2 weeks to 11 weeks after birth. Behavioral tasks were initiated over three periods: 1) 0–11 weeks, 2) 16–29 weeks, and 3) 39–45 weeks. The tasks involved a spatial hole board, an open-field, exposure to a novel object, a runway task, a single-feed task, and home-pen behavior observation. In the hole board, the SHMOs-fed group demonstrated improved reference memory during reversal trials from 16 to 29 weeks.

3.7. Improvement of malnourished infants

Some attempts have been performed to improve the development of malnourished infants by feeding of MOs in the developing countries. Since growth disorders caused by malnutrition sometimes accompanies dysbiosis of colonic microflora, in vivo studies have been performed to determine if developmental delays could recover via improvement of the colonic microflora by feeding of MOs. In these projects, the sialyl bovine milk Os (SBMOs), which had been separated and concentrated from cheese whey, were utilized. Although the oligosaccharide compositions had not been studied in the SBMOs fraction used in these studies, it is assumed that the oligosaccharide profile was similar to that of bovine milk. It has been speculated that the BMOs contained 3′-SL and 6′-SL, which are contained in human milk, too, and 6′-SLN (Neu5Acα2–6Galβ1–4GlcNAc) and DSL (Neu5Acα2–8Neu5Acα2–3Galβ1–4Glc) as well as neutral OSs, including galactosyllactose. [10,127,128].

Charbonneau et al. administered the Malawi's usual feeds (control group) or that containing SBMOs (SBMOs-fed group) to the gnotobiotic mice or piglets that had been implanted with the colonic bacteria, separated from the feces of malnutritional infants of Malawi, to the germ-free animals, and then measured the increasing rate of body weight and bone density [129]. By comparing the SBMOs-fed group with the control group, the researchers showed that the growth rate in the body weight as well as the bone density were higher in the SBMOs-fed group than in the control group.

Then, the mechanism to improve malnutrition by feeding of SBMOs had been studied. Cowardin et al. administered SBMOs to gnotobiotic mice, which had been implanted with colonic bacteria, separated from malnutritional infants in Bangladesh, and then collected blood to analyze several biomarkers in serum [130]. They found that the level of type 1 collagen C-terminal telopeptide, a biomarker released during the bone-separation process by growth of osteoclasts, decreased in the SBMOs-fed animals, and the counts of the osteoclasts in a tibial fragment, stained with alkaline phosphatase, were also reduced in the SBMOs-fed animals. This suggests that the dysbiosis of colonic microflora was improved by feeding of SBMOs, and then the metabolic products produced by the beneficial bacteria should inhibit the growth of the osteoclasts to enhance the bone density of the mice.

3.8. Effects to improve the carbohydrate and lipid metabolic disorders

To determine if HMOs feeding could improve carbohydrate or lipid metabolic disorders, a few in vivo studies have been performed.

Xiao et al. administered 1 % of the HMOs mixture to neonatal model mice that had type 1 diabetes (TID) for 4 to 10 weeks to prevent TID [131]. The results showed that TID development was slower in the HMOs-fed mice than in the control mice who were not fed HMOs, and the serum levels of glucose, IL-17, and IFN-γ were lower in the HMOs-fed animals.

Pessentheiner et al. subcutaneously administered 400 μg/200 μl of 3′-SL twice per day to low-density lipoprotein receptor knockout mice that were fed a Western diet containing 42 % fat calories for 6 weeks from 2 weeks after beginning the administration of this diet [132]. The serum macrophage counts and the aortic lesion area were significantly lower in the 3′-SL administered mice than in the control mice. In addition, these knockout mice were orally administered with 90 mg/day of 3′-SL for 6 weeks starting from 2 weeks after the start of the gavage-administered Western diet. It was shown that the area of aortic lesions was 50 % lower in the 3′-SL-fed mice than that of the controls.

To determine if carbohydrate or lipid metabolic disorders could be improved in infants or adult human by addition of HMOs to the diet, clinical studies will be needed in the future.

Paone et al. showed that 2′-FL supplementation reduced obesity and glucose intolerance in mice caused by feeding of a high-fat diet (HFD) [133]. To decipher this mechanism, they attempted if 2′-FL metabolic effects are linked to changes in intestinal mucus production and secretion, mucin glycosylation, and degradation. These effects were accompanied by several changes in the intestinal mucus layer, including mucus production and composition, in the gene expression of secreted and transmembrane mucins, and of glycosyltransferases involved in mucus secretion, suggesting that 2′-FL supplementation should mediate the metabolism of colonic epithelial mucin, which helps protection against obesity in mice.

3.9. Cohort studies to identify the associations between HMOs functions and the concentration of each HMO in breast milk

Among the biological functions that have been shown in vitro and in vivo studies, some cohort studies have been conducted to determine if the concentrations of some HMOs are correlated with certain functions, such as prevention of NEC, better cognitive ability of breastfed infants, and anti–infection against group-B Streptococcus.

There is an interest in knowing if some HMO concentrations differ between breast milks from mothers whose babies developed NEC and other breast milks from mothers whose babies did not develop NEC. If the concentration of an HMO would be significantly lower in mothers’ milk whose breast-fed infant had NEC than in that in other mothers whose infants had not, it is possible that this HMO might be effective for preventing NEC development. In vivo rat experiments showed that DSLNT prevents NEC development [110]. Masi et al. compared the HMOs profiles and the colonic microflora between milks whose breast-fed infants had NEC (n = 33) or did not (n = 37) [134]. The concentration of DSLNT was significantly lower in the milks of the mothers whose babies had NEC than in the milks of mothers whose infants had not developed NEC, and the minimum concentration of DSLNT that effectively prevented NEC was 241 mmol/mL. Bifidobacterium longum and Enterobacter cloaceae had lower and higher abundances, respectively, among the microflora in the feces of infants who developed NEC than the abundances in the feces of healthy infants.

To determine if HMOs other than DSLNT could prevent NEC, Wejryd et al. determined the concentrations of 15 HMOs in milks from mothers, whose babies’ body weights were <1000 g at birth, in three lactation periods and determined if the concentration of each HMO was correlated with NEC development and the frequency of sepsis [135]. The milks were collected from 91 mothers who had a total 106 infants at 14 and 28 days past delivery and at 36 weeks past menstruation. The concentration of LNDFH-I was lower in the milks of the mothers whose infants had NEC than in those of other mothers whose babies had not NEC at the three time points. An in vivo animal study to see if LNDFH-I can effectively prevent NEC development is needed.

Some in vivo studies have suggested that the cognitive ability could be enhanced by feeding 2′-FL, 3′-SL, 6′-SL and other HMOs. Cohort studies have begun to see if the concentrations of these HMOs in milks are related to the cognitive ability of breast-fed infants [[136], [137], [138], [139], [140], [141], [142]]. Berger et al. studied the correlation between the concentration of 2′-FL in breast milks at 1 and 6 months after delivery and the cognitive scores of their infants at 24 months of age [136]. The score for the cognitive behavior was evaluated according to the standard of the Bayley Scales of Infant Development, Third Edition (BSID-III). The concentration of 2′-FL in breast milk at 1 month after delivery was associated with cognitive development of their infants, but was not at 6 months. Oliveros et al. determined the concentrations of 2′-FL and 6′-SL in 82 samples of breast milk at 1 month after the delivery and studied the correlation between these concentrations and cognitive development of the infants at 6 and 18 months of age, scored according to BSID-III [137]. A positive correlation between the concentration of 6′-SL in milk and the scores of cognition and motor skills of the infants at 18 months of age was shown. A positive correlation between the 2′-FL concentration and the motor score at 6 months of age also was observed.

In breast milk from mothers who had only breast-fed (n = 45) or partially breast-fed (n = 13) their infants, Willemsen et al. studied the association between the concentrations of 2′-FL, 3′-SL, 6′-SL, fucosyl HMOs, and SHMOs at 12 weeks after delivery and executive function of their toddlers at 3 years age [138]. Executive function was defined as the ability to effectively achieve a certain goal by performing planned steps in the aspect of neuropsychology. This function was scored on the basis of the answers to a questionnaire survey obtained from the mothers and their partners for two executive functions of their toddlers. The results showed that the concentrations of 2′-FL and fucosyl HMOs in milk were positively correlated with better executive function of the toddlers. Cho et al. determined the concentrations of HMOs in 183 breast milk samples collected at four study visits from 99 mother–infant dyads and studied the potential association between the concentrations of some HMOs and the infants’ recognitive development, scored according to the Muller Scale of Early Learning (MSEL) [139]. Although no significant association was observed between HMOs and the MSEL when all samples were analyzed together, the composite score for listening and spoken languages and the 3′-SL levels were positively associated in the A-tetrasaccharide [GalNAcα1–3(Fucα1–2)Galβ1–4Glc] + group when the subjects were grouped by detection or nondetection of A-tetrasaccharide in milk.

These cohort studies on the association between HMOs profiles and infant cognitive ability were performed with full-term infants except in a study. Rose et al. explored the relationship between exposure to HMOs contained in mother's milk and neurodevelopment at 2-years corrected age in 137 preterm infants born to 117 mothers [142]. The Age and Stages Questionnaire (ASQ) version 2 was used to assess 2-year neurodevelopment outcomes. The result showed that LNFP-III was significantly associated with the total ASQ score among 104 infants born to Secretor + Lewis + mothers.

Infancy is a critical period for neurodevelopment, which includes myelination, synaptogenesis, synaptic pruning, and the development of motor, socioemotional, and cognitive functions. Although the associations between the concentrations of specific HMOs and cognitive development of breast-fed infants have been shown in some studies as described above, these associations with their mechanisms of action have never been studied. Rajhans et al. investigated possible mediating associations between HMO concentrations, brain myelination, and measures of motor, language, and socioemotional development in healthy term-born breast-fed infants [143]. The development of language was evaluated according to the Bayley Scales of Infant and Toddler Development (Bayley-III), whereas socioemotional development was assessed by the Ages and Stages Questionnaire: Social-Emotional Version (ASQ-SE). Brain myelination was assessed by a myelin water MRI technique. The mean myelin water fraction (MWF) values were calculated throughout the brain in many regions of anatomical interest. The results revealed an association between the concentration of 6′-SL in milk and social skills that were mediated by myelination. The association of the concentrations of 3-FL, LNFP-II, and LNFP-V with language outcomes were observed that were not mediated by myelination.

Readers will be able to obtain more information on the association between HMOs and the brain-nerve development of infants [144,145].

A few cohort studies have been performed to see the potential association between each HMO concentration in milk and the frequency of disease or disorder as well as the development status in the infants. Davis et al. studied the correlation between HMO profiles in mothers’ milk and disease/disorder frequency of their infants in a cohort study in Gambia with 33 mother–infant dyads at 4, 16, and 20 weeks after delivery [146]. The results showed that the concentration of LNFP-1 in milk was positively associated with a low disease frequency in infants, whereas that of 3′-SL was positively correlated with suitable development at these ages. Andreas et al. studied the association between each HMO concentration in mothers’ milk from 0 to 90 days after delivery and the group-B Streptococcus colonization frequency in their infants with 183 mother–infant dyads in Gambia [147]. The frequency of early-stage GBS colonization and at 60–89 days after birth was lower in the breast-fed infants given the milk of Lewis-positive donor mothers than those given the milk of Lewis-negative mothers. There was no association between secretor status and GBS colonization.

A rodent study showed that administration of 2′-FL and 6′-SL reduced symptoms of food allergy [115], suggesting that the potential association between food allergy development in breast-fed infants and the HMOs profiles in their mothers’ breast milk should be investigated. Miliku et al. studied the associations of 19 individual HMO concentrations and overall HMOs profiles with food sensitization at 1 year of age in a Canadian cohort of 421 mother-infant dyads [148]. It was found that 59/421 infants (14.0 %) were sensitized to ≥1 food allergens at 1 year of age by assessing standardized skin testing. Even though they did not observe any significant association between each concentration of 19 individual HMOs and food sensitization, the overall HMO profiles differed significantly in the milk consumed by sensitized vs nonsensitized infants. The following was shown for the breast milks which had been drunk by the infants who had a lower food allergy risk. The concentrations of fucosyl-disialyl lacto-N-hexaose (FDSLNH), LNFP-II, LNnT, LNFP-I, LST c, fucosyl lacto-N-hexaose (FLNH) were relatively higher, whereas those of LNH, LNT, 2′-FL, disialyl lacto-N-hexaose (DSLNH) were relatively lower. Because the examples of such cohort studies for the association between the sensitization with food allergens of the breast-fed infants and HMOs profiles in their mothers’ breast milk are very few, additional studies should be performed in other geographical areas.

3.10. Summary

It has been shown that HMOs have the functions that prevent the disease and disorders and stimulate healthy development of breast-fed infants by in vitro, in vivo, and cohort studies. The concentrations of MOs are much lower in cows than in humans, and the compositions of MOs are also different between cows and humans. Even though it is desirable that HMOs like Os will be incorporated into milk substitute, which are manufactured with bovine milk, it has been hard to manufacture these Os at industrial scale. Although the prebiotic Os, including galacto Os, fructo Os, and lactulose, which have been manufactured at industrial scale, are incorporated into the milk substitute, they cannot replace all functions of HMOs. However, recently, a few HMOs have been manufactured and incorporated into infant formula. In the next section, we describe the development and production of HMOs at industrial scale along with the relevant clinical studies of infant formula, containing human-identical milk oligosaccharides (Hi-MOs) produced for bottle-fed infants.

4. Industrial production of HMOs by fermentation process

4.1. Overview of HMO production by fermentation process

As described in Section 2, laboratory-scale production methods for HMOs were established by combining enzymatic, chemical, and chemoenzymatic method for the purpose of bio-functional research and analysis. However, while these methods are useful for constructing HMOs libraries, the productivities of these methods have not been enough for commercial production. In the early stage of commercial scale production of HMOs, Glycom A/S was oriented toward chemical synthesis [90], but it was later replaced by fermentation using microorganisms harboring multiple heterologous enzymes to produce HMOs, and most HMOs currently available for commercial production have been produced by fermentation technique. The fermentation method uses metabolically modified microorganisms to produce HMOs from inexpensive carbon sources such as lactose and glucose inside the microorganisms [[149], [150], [151]].

Specifically, the microorganisms are modified to express glycosyltransferase and enzymes to produce the substrates for glycosyltransferase. Sugar nucleotides such as UDP-GlcNAc, GDP-Fuc, and CMP-Neu5Ac are the common substrate for glycosyltransferases. The glycosyltransferase uses sugar nucleotides as a donor and transfer sugars moiety of the nucleotide sugar to externally added lactose. Compared to enzymatic and chemoenzymatic synthesis methods, the fermentation method is now the mainstream production process because it is easier to scale up and the cost of raw materials is cheaper compared to the other methods [151,152]. In this chapter, we describe the production method of HMOs by fermentation and the trend of clinical trials using these HMOs.

4.2. Production of fucosyllactose by fermentation

De novo synthesis of 2′-FL is achieved by glycosyltransferase which catalyzes transfer of fucose from GDP-Fuc to OH-2 position of galactose of lactose. As lactose is added externally, the supply of GDP-Fuc, an expensive nucleotide sugar, is a challenge. In the fermentation process, the most important feature is that GDP-Fuc is supplied from inexpensive carbon sources such as glucose, etc. The biosynthetic pathway of GDP-Fuc can be roughly divided into two pathways. i.e., de novo synthesis and salvage pathways (Fig. 5B). In the salvage pathway, 2′-FL is synthesized from externally added fucose and lactose. Chin et al. used this pathway and introduced fkp, the gene for fucokinase/fucose-1-phosphate guanylyltransferase, a bifunctional enzyme from Bacteroides fragilis, into E. coli. Fucokinase converted externally added fucose to fucose-1-phosphate and is further converted to GDP-Fuc by transfer reaction with GTP by the action of fucose-1-phosphate guanylyltransferase. Furthermore, by disrupting fucI-fucK gene cluster, the fucose degradation pathway in E. coli, the concentration of GDP-Fuc was increased. Furthermore, by introducing fucT2, an α-1,2-fucosyltransferase from Bacteroides fragilis, 2′-FL production reached 23.1 g/L [153]. Baumgärtner et al. constructed a plasmid-free strain by introducing fkp from Bacteroides fragilis and α-1,2-fucosyltransferase from H. pyroli all in the genome, and succeeded in producing 2′-FL at 20.28 g/L from lactose, glycerol, and fucose [154].

Fig. 5.

Example of biosynthetic pathway for production of (A) 3′-SL, 6′-SL, and (B) 2′-FL.

HMOs are produced enzymatically inside the cell from externally added glucose and lactose. Nucleotide sugars such as CMP-Neu5Ac in (A) or GDP-Fuc in (B) are biosynthesized internally from glucose and glycosyltransferases transfer the sugar moiety of the nucleotide sugar to lactose. Produced 3′-SL, 6′-SL in (A), and 2′-FL in (B) are excreted from the cell to the fermentation medium through specific sugar transporters.

The other pathway, de novo synthesis pathway, utilizes the cholanic acid synthesis pathway of E. coli and is produced by multiple enzymatic reactions starting with fructose-6-phosphate (F6P), which is obtained from glucose. First, F6P is converted to mannose-6-phosphate by mannose-6-phosphate isomerase, and then converted to mannose-1-phosphate by the action of phosphomannomutase. Mannose-1-phosphate is then converted to GDP-mannose by mannose-1-phosphate guanylyltransferase and further converted to GDP-Fuc by successive actions of GDP-d-mannose 4,6-dehydratase and GDP-Fuc synthase. When using E. coli as a host, all of these genes exist in E. coli genome and no need to use heterologous genes. The GDP-Fuc supply is thereby enhanced in E. coli through the reinforcement of the biosynthetic pathway with various strong promoters, overexpression of RscA, a positive regulator of cholanic acid production, and disruption of downstream genes involved in cholanic acid biosynthesis. The glycosyltransferases that transfer the fucose moiety of GDP-Fuc to lactose include those from Helicobacter sp. [155], H. pylori [156,157,158], the aforementioned Bacteroides fragilis [153], H. hepaticus [159], H. mustelae [160], Bacteroides vulgatus [157], Campylobacter jejuni [157]. In addition, glycosyltransferase from E. coli O128 [161], O86 [162], O127 [163] are also known. For example, Huang et al. successfully produced 9.12 g/L of 2′-FL from glucose and lactose using recombinant E. coli [156], and Chen et al. by using FucT, an α-1,2-fucosyltransferase of Azospirillum lipoferum 112.5 g/L of 2′-FL was successfully produced from lactose and glycerol [164]. 3-FL can also be produced by changing the glycosyltransferase. For example, Huang et al. produced 12.43 g/L of 3-FL from glucose and lactose by using FutA from Helicobacter pylori [157]. Lee et al. also successfully produced 1.22 g/L of difucosyllactose (DFL) simultaneously with 0.47 g/L of 2′-FL by combining Helicobacter pylori-derived FutA and Bacteroides fragilis-derived fkp using a salvage pathway [165].

4.3. Production of LNT/LNnT by fermentation method

LNT and LNnT are tetrasaccharide HMOs containing N-acetylglucosamine, which have the longer sugar chain than the trisaccharide 2′-FL and 3-FL described above. UDP-GlcNAc is utilized as a precursor for peptidoglycan biosynthesis in E. coli. UDP-GlcNAc biosyhthetic pathway also exists in E. coli., Therefore as in the case of GDP-Fuc, endogenous UDP-GlcNAc biosynthetic genes can be used for the production of this nucleotide sugar. Baumgärtner et al. successfully produced 12.72 g/L of LNT by introducing β1,3-N-acylglucosaminyltransferase (LgtA) and β1,3-galactosyltransferase (WbgO) with the addition of external galactose [166]. Zhu et al. successfully produced 22.07 g/L of LNnT by using β-1,4-galactosyltransferase from Aggregatibacter actinomycetemcomitans NUM4039 [167]. In this case, Zhu et al. also synthesized UDP-Gal from glycerol. First, fructose-6-phosphate was converted to glucose-6-phosphate by glucose-6-phosphate isomerase. Next, glucose-6-phosphate was converted to glucose-1-phosphate by phosphoglucomutase and then, converted to UDP-Glc by glucose-1-phosphate uridyltransferase. UDP-Glc is further converted to UDP-Gal by epimerization with UDP-Glc 4-epimerase, a process consisting of the conversion to UDP-Gal. In this case, fructose-6-phosphate is used for productions of two different nucleotide sugar, UDP-GlcNAc and UDP-Gal, and it is apparent that fructose-6-phosphate is also required for the growth of E. coli. Therefore, a skillful balance should be required. Sugita et al. introduced LgtA from Neisseria polysachharea and further enhanced the efflux protein (SetA) to achieve 34.2 g/L LNT II production, the precursor of LNT/LNnT [168]. In addition, Liao et al. introduced β-1,4-galactosyltransferase (HpgalT) from Helicobacter pylori and β-1,3-galactosyltransferase (SewbdO) from Salmonella enterica subsp. salamae serovar, and achieved to produce 22.07 g/L of LNnT and 48.41 g/L of LNT [169].

4.4. Production of SL by fermentation method

Sialyllactose is an acidic HMO with a sialic acid (N-acetylneuraminic acid, Neu5Ac) bonded to lactose. There are two types of sialyllactoses, 3′-SL and 6′-SL, depending on the position of the sialic acid bond. Sialyllactose is produced by transferring Neu5Ac residues of CMP-N-acetylneuraminic acid (CMP-Neu5Ac) to lactose, and like as the cases of fucosyllactose, LNT, and LNnT, it is important to provide the sugar nucleotide substrate in the bacteria (Fig. 5A). Priem et al. successfully produced >1 g/L of 3′-sialyllactose with E.coli harboring CMP-Neu5Ac synthase and α2,3-sialyltransferase from N. meningitidis by adding sialic acid and lactose externally [170]. To produce CMP-N-acetylneuraminic acid de novo, UDP-GlcNAc can be the direct precursor. UDP-GlcNAc is produced through biosynthetic pathway as described in the LNT/LNnT, and then, converted to N-acetylmannosamine (ManNAc) with N-acetylglucosamine 2-epimerase. ManNAc is then converted to Neu5Ac with sialic acid synthase, and finally, CMP-Neu5Ac is produced by CMP-Neu5Ac synthase. Fierfort and Samain used the N-acetylglucosamine-6-phosphate-epimerase, neuC, from Campylobacter jejuni, sialic acid synthase neuB, and CMP-sialic acid synthase neuA to synthesize CMP-Neu5Ac from glycerol. With α2,3-sialyltransferase from N. meningitidis, the production of 3′-SL at 25 g/L was achieved [171].

There are another biosynthetic pathway to produce CMP-Neu5Ac. Deng et al. constructed artificial pathway to produce N-acetylglucosamine [172]. Glucosamine-6-phosphate is converted to N-acetylglucosamine-6-phosphate using glucosamine-6-phosphate N-acetyltransferase from budding yeast, then to N-acetylglucosamine using a phosphatase, and then to N-acetylglucosamine epimerase to N-acetylmannosamine. Yang et al. also blocked the degradation of Neu5Ac and N-acetylglucosamine to produce sialyllactose [173].

Through this pathway, Guo et al. achieved the production of 23.1 g/L of 3′-SL with genes from Pasteurella multosida [174]. Drouillard et al. also succeeded in producing 6′-SL at 30 g/L in a similar manner by changing the glycosyltransferase to α2,6-sialyltransferase from Photobacterium [175].

4.5. Scale-up of fermentation

As mentioned above, we are now able to produce various HMOs by fermentation methods. But for industrialization, efficient scale-up technology is additionally required. So far, companies such as Glycom S/A, Jennewein Biotechnologie GmbH, DuPont de Nemours, Inc., Koninklijke Friesland Campina N.V., Kyowa Hakko Bio Co. succeeded in industrial scale production of HMOs through fermentation method. These industrial production process use large fermentation tanks (100kL to 200kL) for production. These large fermentation tanks often have a vertical shape with a small diameter to reduce the footprint and often show poor mixing in the vertical direction [[176], [177], [178]]. The production of HMOs often uses carbon sources such as glucose and glycerol through multiple biosynthetic pathways. Furthermore, since this carbon source is not only the backbone of HMOs production, but is also used to generate the energy needed for biosynthesis, as well as for the growth of the producing bacteria themselves, unbalanced carbon and oxygen sources in the fermenter can result in excess or deficient carbon sources and oxygen, changing the metabolic flow to a situation different from that in the laboratory. As a result, there are concerns that this may lead to poor growth, poor productivity, and the production of by-products. In addition, if microorganisms with high productivity are cultured to increase productivity, the oxygen supply may become rate-limiting, and cooling of the fermentation heat generated may not be sufficient. Therefore, metabolic pathways and productivity must be adjusted according to the cooling capacity of the fermenter.

4.6. Downstream process

As HMOs produced by fermentation or bacteria reaction methods contain many impurities, including bacterial cells, a purification process is essential for industrial production (Fig. 6). Bacterial cells can be separated by centrifugation or membrane separation. After the bacterial cells are separated, the supernatant is supplied to ion exchange resins for separation and purification. In some cases, desaltation by electrodialysis is used. Then finally the HMOs powder is obtained by spray drying or crystallization. Crystallization is performed by taking advantage of the phenomenon of low solubility of HMOs in organic solvents, and it is important to control the crystal form. There have been reports of successful crystallization of 2′-FL, 6′-SL, 3′-SL, etc., and it is believed that this method can be used to produce high-purity products. Unlike crystallization, spray drying does not improve purity, but it is attracting attention as a simpler and cheaper process. When spray drying is used, the purification process is important because the spray-dried raw solution must contain the target product with high purity. For the purification of acidic HMOs such as sialyllactose, the use of ion exchange resins can be considered. Neutral and basic impurities can be removed by passing, and acidic impurities can be removed by chromatographic separation.

Fig. 6.

Example of purification process for HMOs produced by fermentation process.

HMOs produced by fermentation process are present in the culture medium in a dissolved state along with many impurities. Cells and solids are first removed by membrane separation or centrifugation (1), followed by electrodialysis or ion exchange chromatography to remove ions and impurities (2). The remaining liquid containing HMOs generally contains trace amounts of colored substances, so it is decolorized using activated carbon, etc. (3). Finally, it is powdered using spray drying or crystallization to make a powder product (4).

4.7. Regulation

As described above, through the development of genetically engineered strains, the establishment of industrial fermentation methods, as well as the use of purification methods, industrial production of some HMOs has become possible, and products with the addition of HMOs to powdered milk are already on the market.

Even though HMOs are compounds present in breast milk and are ingested by breast-fed infants, many countries have regulations regarding the addition of industrially produced HMOs: Novel Food in Europe, GRAS in the U.S., and a similar system has recently been established in China. The first registered industrially produced HMOs were 2′-FL and LNnT, chemically synthesized by Glycom. Among HMOs made by fermentation, Jennewein successfully obtained the first GRAS registration in the U.S. for 2′-FL. 6′-SL, 3′-SL, and LNT were also registered by several companies, followed by difucosyllactose (DFL), lacto-N-fucopentaose-I (LNFP-I) have been submitted as mixtures with 2′-FL, rather than as stand-alone registrations. This is presumably because the formation of 2′-FL is inevitable in the biosynthetic pathway and its removal in the downstream process is difficult. Although the above is for addition to infant formula, a small number of GRAS/NF products are now on the market as supplements for adults.

4.8. Intervention trials using the formula including artificially manufactured HMOs

With the establishment of an industrial process for the previously unavailable HMOs and the prospect of a uniform supply of high-quality HMOs, several intervention trials with artificially made supplement-fed children are underway. We conclude this section with an overview of these intervention trials. Although an excellent review of clinical trials of industrially produced HMOs by Schönknecht et al. [46] has been published, this section is intended to review and summarize the scientific literature on HMOs published after that earlier review. The objective is to provide a comprehensive overview of the clinical trials that have been conducted to examine HMOs clinically. As shown in Table 2, a total of 34 studies have been published, including 27 reported clinical trials and 7 reported follow up studies. Eight industrially manufactured HMOs, including 2′-FL, 3-FL, DFL, LNT, LNnT, 3′-GL, 3′-SL, and 6′-SL, have been used in clinical interventional studies. Most of the publications examined the effects of 2′-FL as a standalone HMO. However, some studies also have investigated the effect of other HMOs, including LNnT, 3′-SL, and 6′-SL, as individual HMOs. The remaining studies examined a mixture of HMOs, including 2′-FL + LNnT, 2′-FL + 3′-FL + LNT + 3′-SL + 6′-SL, and 2′-FL + DFL + LNnT + 3′-SL + 6′-SL. Table 2 provides an overview of the characteristics of scientific publications reporting the use of products containing HMOs. Approximately half of the studies (18) involved healthy term infants. A total of six studies included infants with medical indications, of which five involved infants diagnosed with or suspected of having food allergies and one involved preterm infants. Two studies were identified that included children as participants. One study involved children aged between 1 and 2.5 years, whereas the other involved children aged between 6 and 12 years who were overweight or obese. Eight studies were identified that included adults (>18 years old) as participants, of which two focused on healthy individuals, four focused on adults with irritable bowel syndrome (IBS) (including one follow up study), and two focused on adults diagnosed with Helicobacter pylori infection.

Table 2.

Clinical trials investigating supplementation with HMOs.

Exp. No.	Subject2)	Study Population2)	Study Design3)	Intervention Groups	Comparison	Duration of Intervention4)	Key Findings	Lit.5)
1	a	aged 6–24 months (n = 228)	r, d-b, mc, c-t	0.2 g/L LNnT (n = 115)	Control formula without OSs (n = 113)	16 weeks	Safe and well-tolerated / No effect on oropharyngeal colonization with Streptococcus pneumoniae	(1)
2	a	< 5 days at inclusion (n = 420)	r, d-b, mc, c-t	1) 0.2 g/L 2′-FL + 2.2 g/L GOS (n = 104), 2) 1.0 g/L 2′-FL + 1.4 g/L GOS (n = 109)	Control formula: 2.4 g/L GOS (n = 101)	17 weeks	Safe and well tolerated -Effect of fecal amino acid degradation and bile acid conjugation -Similar relative absorption and relative excretion of 2′-FL in intervention and BF groups at DOL 42 and 119	(2)
3	a	Follow-up analysis of #2	–	Subpopulation from intervention 1: 0.2 g/L 2′-FL + 2.2 g/L GOS (n = 76)	Subpopulation from control formula: 2.4 g/L GOS (n = 75)	–	Similar cytokine profile and ex vivo stimulation of PBMCs in intervention and BF groups at 6 weeks	(3)
4	a	< 8 days at inclusion (n = 119)	r, d-b, mc, c-t	0.2 g/L 2′-FL + 2 g/L scFOS	Control formula without OSs (n = 42)	35 days	Well tolerated	(4)
5	a	< 14 days at inclusion (n = 175)	r, d-b, mc, c-t	1.0 g/L 2′-FL + 0.5 g/L LNnT	Control formula without OSs (n = 87)	a	Safe and well tolerated -Supports normal, age-appropriate growth -Reduced parent-reported cases of bronchitis and medication use	(5)
6	a	14 ± 5 days at inclusion (n = 79)	r, d-b, mc, c-t	0.25 g/L 2′-FL + Bifidobacterium animalis ssp. lactis. Partially hydrolyzed formula (n = 39)	Control formula without OSs + Bifidobacterium animalis ssp. lactis (n = 40)	6 weeks	Safe and well tolerated well -Supports normal, age-appropriate growth -Trend of reduced number of reported infections (not statistically)	(6)
7	a	Follow-up analysis of #5	–	Subpopulation from intervention: 1.0 g/L 2′FL	Subpopulation from control formula without OSs (n = 64)	–	Gut microbiome composition shifted towards BF infants -Reduced use of antibiotics during the first 12 months of life	(7)
8	a	7 days to 2 mo. at inclusion (n = 207)	n-r, o-I, m-t	1.0 g/L 2-'FL + 0.5 g/L LNnT. Partially hydrolyzed formula	Mixed group: HMO formula + BF	8 weeks	Safe and well tolerated -Supports normal, age-appropriate growth	(8)
9	a	< 14 days at inclusion (n = 175)	r, d-b, mc, c-t	1.0 g/L 2-'FL + 7.2 g/L GOS + 0.8 g/L FOS, including 0.015 g/L 3-GL. Partially hydrolyzed formula (n = 108)	Control formula without 2′-FL and 3′-GL, with 7.2 g/L GOS + 0.8 g/L FOS (n = 107)	17 weeks	Safe and well tolerated -Supports normal, age-appropriate growth	(9)
10	a	Follow-up analysis of #5	–	–	–	–	Gut microbiome composition shifted towards BF infants -Effect on fecal biomarkers at 3 months	(10)
11	a	< 14 days at inclusion (n = 341)	r, d-b, mc, c-t	5.75 g/L HMOs: 2.99 g/L 2′-FL + 0.75 g/L 3-FL + 1.5 g/L LNT + 0.23 g/L 3′-SL + 0.28 g/L 6′-SL (n = 113)	Control formula without OSs (n = 112)	b	Safe and well tolerated -Supports normal, age-appropriate growth -Stool characteristics shifted towards BF infants	(11)
12	a	< 14 days at inclusion (n = 289)	r, d-b, mc, c-t	1.0 g/L 2′-FL + Limosilactobacillus reuteri (n = 144)	Control formula without OSs + Limosilactobacillus reuteri (n = 145)	6 months	Safe and well tolerated -Supports normal age-appropriate growth -Gut microbiome composition shifted towards BF infants	(12)
13	a	< 21 days at inclusion (n = 535)	r, d-b, mc, c-t	Int. 1: 1.5 g/L HMOs: 0.87 g/L 2′-FL + 0.10 g/L DFL + 0.29 g/L LNT + 0.11 g/L 3′-SL + 0.14 g/L 6′-SL (n = 153)	Control formula without OSs (n = 155)	6 months	Gut microbiome composition shifted towards BF infants -Effect on fecal biomarkers at 3 and 6 months	(13)
14	a	< 14 days at inclusion (n = 363)	r, d-b, mc, c-t	5.75 g/L HMOs: 3.0 g/L 2′-FL + 0.80 g/L 3-FL + 1.5 g/L LNT + 0.20 g/L 3′-SL + 0.30 g/L 6′-SL (n = 130)	Control formula without OSs (n = 129)	4 months	Safe and well tolerated -Supports normal, age-appropriate growth -Stool characteristics shifted towards BF infants	(14)
15	a	Follow-up analysis of #5	–	–	–	–	Effect on gut-microbiome fecal co-metabolite profile	(15)
16	a	< 28 days at inclusion (n = 221)	r, d-b, mc, c-t	1.0 g/L 2′-FL + FOS (not specified) (n = 66)	Control formula + GOS + FOS (not specified) (n = 66)	16 weeks	Safe -Supports normal, age-appropriate growth -Gut microbiome composition shifted towards BF infants	(16)
17	b	Infants and children between 2 month and 4 years with diagnosed CMPA (n = 67)	r, d-b, p-c fcp, (DBPCFC) o-l-ch	1.0 g/L 2′-FL + 0.5 g/L LNnT. Extensively hydrolyzed formula (n = 36 in DBPCFC, n = 62 in open-label challenge)	Control formula without OSs (n = 31, only in DBPCFC)	1 week	Safe and well tolerated - Confirmed hypo-allergenicity	(17)
18	b	Infants< 60 days at inclusion with suspected food protein allergy/sensitivity (n = 48)	n-r, mc, s-a, tr	0.2 g/L 2′-FL. Extensively hydrolyzed formula (n = 48)	–	60 days	-Safe and well tolerated	(18)
19	b	Term infants aged 0–6 months diagnosed with CMPA (n = 194)	r, d-b, mc, c-t	1.0 g/L 2′-FL + 0.5 g/L LNnT. Extensively hydrolyzed formula (n = 94)	Control formula without OSs (n = 96)	c	Safe and well tolerated -Supports normal, age-appropriate growth -No effect on allergy symptoms	(19)
20	b	Term infants aged 1–8 months diagnosed with moderate-to-severe CMPA (n = 32)	n-r, o-l, mc, s-a, tr	1.0 g/L 2′-FL + 0.5 g/L LNnT. Amino acid-based formula (n = 32)	–	d	Safe and well tolerated -Supports normal, age-appropriate growth -Stool characteristics and gut microbiome composition shifted towards BF infants	(20)
21	b	Preterm infants with very low birth weight < 1700 g (n = 86)	r, d-b, mc, c-t	0.34 g/kg/d 2′-FL + 0.034 g/kg/d LNnT (n = 43)	Glucose placebo (0.140 g/kg/d) (n = 43)	–	Safe and well tolerated -Increased mean head circumference gain at day 21 before full enteral feeding	(21)
22	c	Healthy children aged 1–2.5 years (n = 461)	r, d-b, sc, c-t	Int. 1) Formula with 3.0 g/L 2′-FL (n = 114)	Formula with no supplements (n = 114)		Safe and well tolerated -Supports normal, age-appropriate growth -Reduced number of days with ard stools	(22)
23	c	Overweight/obese children aged 6–12 years (n = 75)	r, d-b, sc, c-t	Int. 1) 4.5 g/d 2′-FL (n = 25)	4.5 g/d glucose placebo (n = 25)		Safe and well tolerated -Effect on gut microbiome composition -No effect on stool characteristics, blood, and fecal markers	(23)
24	d	Adults with diagnosed H. pylori infection (n = 6)	o-l, st	10 g/d 3′-SL (n = 6)	–	1 day	No effects on H. pylori infection -well tolerated -No effects on blood markers	(24)
25	d	Adults with diagnosed H. pylori infection (n = 65)	r, d-b, pc-t	Int. 1) 10 g/d 3′-SL (n = 17) Int. 2) 20 g/d 3′-SL (n = 22)	Placebo (not specified, (n = 21)	4 weeks	No effects on H. pylori infection -well tolerated	(25)
26	e	Healthy adults (n = 100)	r, d-b, s-c, c-t	Int. 1) 5 g/d 2′-FL (n = 10) Int. 2) 10 g/d 2′-FL (n = 10) Int. 3) 20 g/d 2′-FL (n = 10)	2 g/d glucose placebo (n = 10)	2 weeks	Safe and well tolerated -Reported gastrointestinal symptoms in some higher-dose intervention groups	(26)
27	f	Adults with IBS (n = 61)	r, d-b, s-c, c-t	Intervention 1: 5 g/d 2′-FL + LNnT at 4:1 ratio (n = 20)	5 g/d glucose placebo (n = 21)	e	Well tolerated -No worsening of IBS symptoms -No effect on gut microbiome composition	(27)
28	f	Follow-up analysis of #27	–	–	–	–	-Effect on microbiome composition in fecal and mucosal colonic biopsies sample -Effect on fecal and plasma biomarkers	(28)
29	f	Adults with IBS (n = 317)	o-l. mc, s-a, tr	4 g/d 2′-FL + 1 g/d LNnT (n = 317)		12 weeks	Safe and well tolerated -Improved IBS symptoms -Effect on stool characteristics	(29)
30	f	Adults with IBS, ulcerative colitis, Crohn's disease, or celiac disease (n = 20)	o-l, mc, s-a, pilot-tr	4 g/d 2′-FL (n = 20)		6 weeks	Improved IBS symptoms -Effect on gut microbiomes composition	(30)
31	e	Healthy adults (n = 60)	r, t-b, s-c, c-t	3 g/d 6′-SL (n = 30)	3 g/d maltodextrin placebo (n = 30)	12 weeks	Safe and well tolerated -No efect on blood markers	(31)
32	b	Follow-up analysis of #19	r, d-b, mc, c-t	1.0 g/L 2′-FL + 0.5 g/L LNnT.	Control formula without oligosaccharides (n = 97)	f	-Gut microbiome composition shifted to the enrichment of bifidobacteria -Effect of fecal amino acid degradation and bile acid conjugation	(32)
33	a	aged 8–12 month (n = 55)	r, s-b, prospect. cr, tr	Int. 2) 5.0 mg iron + 3.0 g GOS (n = 49), Int. 3) 5.0 mg iron + 2.0 g 2′-FL and 1.0 g LNnT (n = 49)	Int. 1) 5.0 mg iron (n = 49)	1 day	-No effects on iron absorption	(33)
34	a	Follow-up analysis of #2, 3	r, d-b, mc, c-t	Subpopulation of Intervention 1: 0.2 g/L 2′-FL + 2.2 g/L GOS (n = 54)	Control formula: 2.4 g/L GOS (n = 48)	17 weeks	-Increases in serum metabolites derived from microbial activity in the gastrointestinal tract	(34)

1) (abbreviation) a: healthy term infants, b: infants with medical indications. c: children, d: adults with H. pyolori infection, e: healthy adults, f: adultswith IBS, 2) (abbreviation) a: Follow-up analysis of #2, b: Healthy term infants < 8 days at inclusion (n = 119), 3) (abbreviation) r: randomized, n-r: non-randomized, s-b: single-blinded, d-b: double-blinded, sc: single-center, s-a: single-arm, o-l: open-label, p-c: placebo-controlled, mc: multicenter, o-l-st: open-label study, cr: crossover, ch: challenge, fc: food challenge, tr: trial, st: study. 4) (abbreviation) a: 6 months (HMO intervention) + follow- up to 12 months of age (no HMO intervention). b: 16 weeks (HMO intervention) + 8-week voluntary follow-up (HMO intervention). c: 4 months (HMO intervention) + follow- up to 12 months of age (no HMO intervention). d: 4 months + voluntarily up to 12 months of age (HMO intervention). e: 4 weeks supplementation + 4-week follow-up (no HMO intervention). f: 4 months (HMO intervention) + follow- up to 12 months of age (no HMO intervention), 5) (1) A.P. Prieto, Foods Food Ingred. J. Jpn. 210 (2005) 1018–1030; (2) B.J. Marriage, et al., J. Pediatr. Gastroenterol. Nutr. 61 (2015) 649–658; (3) K.C. Goehring, et al., J. Nutr. 146 (2016) 2559–2566; (4) J. Kajzer, et al., FASEB J. 30 (2016) 671.4; (5) G. Puccio, et al. J. Pediatr. Gastroenterol. Nutr. 64 (2017) 624–631; (6) H.M. Storm, et al., Glob. Pediatr. Health 6 (2019); (7) B. Berger, et al., mBio (2020) 11, e03196–19; (8) R. Riechmann, et al., Nutr. Hosp. 37 (2020,) 698–706; (9) Y. Vandenplas, et al., Nutrients 12 (2020) 3560; (10) S.K. Dogra, et al., Microorganisms 9 (2021) 1939; (11) K. Parschat, et al., Nutrients 13(8) (2021) 2871–2890; (12) P. Alliet, et al., Nutr. J. 21 (2022) 11; (13) M. Bosheva, et al., Front. Nutr. 9 (2022) 920,362; (14) J. Lasekan, et al., Nutrients 14 (2022) 2625; (15) F.P. Martin, et al., Front. Nutr. 9 (2022) 935,711; (16) J.C. Wallingford, et al., Front. Nutr. 9 (2022) 961,526. [Google Scholar]. (17) A. Nowak-Wegrzyn, et al., Nutrients 11 (2019) 1447. (18) C. Ramirez-Farias, et al., Nutrients 13 (2021) 186. (19) Y. Vandenplas, et al., Nutrients 14 (2022) 530; (20) M.S. Gold, al., Nutrients 14 (2022) 2297; (21) J.M. Hascoët, et al., Front. Pediatr. 10 (2022) 858,380; (22) T.F. Leung, et al., Pediatr. Allergy Immunol. 31 (2020) 745–754; (23) C.E. Fonvig, et al., J. Pediatr. Gastroenterol.Nutr. 73 (2021) 408–414; (24) A.R. Opekun, et al., Aliment. Pharmacol. Ther. 13 (1999) 35–42; (25) F. Parente, et al., Helicobacter 8 (2003) 252–256; (26) E. Elison, et al., Br. J. Nutr. 116 (2016) 1356–1368; (27) C. Iribarren, et al., Neurogastroenterol Motil, 32 (2020) e13920; (28) C. Iribarren, et al., Nutrients 13 (2021) 3836; (29) O.S. Palsson, et al., Clin. Transl. Gastroenterol. 11 (2020) e00276; (30) J.J. Ryan, et al., Nutrients, 13 (2021) 938; (31) J.H. Kim, et al., Regul. Toxicol. Pharmacol.129 (2022) 105,110; (32) C.L. Boulangé, et al., Int. J. Mol Sci. 24 (2023) 11,422; (33) A. Giorgetti, et al., Am. J. Clin. Nutr.117(1) (2023) 64–72; (34) D.R. Hill, et al., Nutrients. 15(10) (2023) 2339.

The results of all studies indicated that HMO supplementation was safe and well tolerated in all age groups and health conditions, irrespective of HMO structure or applicable dose. All studies in healthy term infants demonstrated age-appropriate growth following HMO supplementation, with no significant differences in growth observed between the intervention and control groups. One study in preterm infants with very low birth weights (<1700 g) reported a significantly higher length-for-age z-score in the HMO group than in the placebo group. Thus, HMO supplementation facilitated age-appropriate growth and was not associated with any serious adverse events. In addition, no safety or tolerance concerns were identified in children or adults. The high intake of much higher doses of HMO, as a 20 g/d bolus in adults, caused minor adverse events, consistent with expectations. Notably, the administration of such elevated doses of HMOs as bolus drugs is unlikely to accurately reflect the actual circumstances.

Furthermore, several studies have indicated that there are health benefits associated with consumption of HMOs. Interventional studies indicate that HMOs exert a modulating effect on the gut microbiota, which are associated with a range of additional health benefits. In infants, results, including those related to gut microbiota composition, stool characteristics, blood, and gut immune markers were comparable to those observed in breast-fed infants. In several studies, it was generally observed that HMO supplementation resulted in a microbiome composition that was more similar to that of breast-fed infants (compared with controls), with a higher relative abundance of Bifidobacterium species. Intake of HMOs reportedly promoted more frequent and softer stools, which reduced the fecal pH and increased the fecal content of short-chain fatty acids (SCFAs) along with microbiome changes. The change in the microbiome may also support the immune system, either through direct interaction with immune cells or through the production of immunomodulatory makers. The plasma levels of cytokines were analyzed following supplementation with 2′-FL, which showed a notable reduction in plasma concentrations of pro-inflammatory cytokines [IL-1α, IL-1ra (1 receptor antagonist), IL-1β, IL-6, and TNF-α] relative to those in the control group. This profile was comparable to that observed in the breast-fed group. Interestingly, the changes in the microbiome may also be associated with a reduction in infections. Infants who received HMO supplementation exhibited a lower incidence of infections, were less likely to require medical professional consultation, and were less likely to experience otitis media. A correlation has been suggested between the presence of Bifidobacterium in fecal communities, microbial metabolic pathways and a reduction in infection rates.

There was also evidence of microbiome regulation in children and adults. In children from 6 to12 years of age, supplemented 2′-FL or 2′-FL + LNnT increased the abundance of Bifidobacterium adolescentis. In adults, supplementation with 2′-FL, LNnT, or a 2′-FL + LNnT mixture generally modulated the intestinal microbiome, which increased the relative abundance of Bifidobacteria. In adults with IBS, the mixture of 2′-FL + LNnT showed an increase in the relative abundance of Bifidobacterium species (particularly B. adolescentis and B. longum) as well as in the genus Faecalibacterium and the family Lachnospiraceae in fecal samples. In adults with IBS, supplementation of 4 g/d 2′-FL over 6 weeks increased the quantity of B. longum, Faecalibacterium prausnitzii, Anaerotoruncu colihominis, and Pseudoflavonifractor species in stool samples and increased total SCFA, acetate, and butyrate levels. However, the clinical significance of this remains uncertain due to the limited research and heterogeneous study design. In addition, although not included in Table 2 in this chapter because it was not an industrial HiMO intervention study, there have also been reports of infant intervention studies in which Bovine milk-derived oligosaccharides (BMOs) containing primarily galacto-oligosaccharides with inherent concentrations of sialylated oligosaccharides was added to infant formula. As Estorninos et al. showed [179], infant formula with BMOs shifts the gut microbiota and metabolic signature closer to that of breast milk, reduces fecal pathogens, and improves the intestinal immune response, and it is hoped that sialylated lactose will also have beneficial effects on infants. As 3′-SL and 6′-SL have also become industrially produced following 2′-FL, that could lead to start more intervention studies in infants and young children on the specific functions of sialylated HMOs. Further clinical research is required to elucidate the mechanisms by which specific HMO types exert beneficial effects and to substantiate the beneficial effects of HMO supplementation in products, such as infant formula and functional foods. Nevertheless, the results of these studies and preclinical studies as summarized in this review indicate the potential for health benefits associated with HMO supplementation. The increased availability of HMO commercial samples has facilitated the rapid expansion of HMO research activities in clinical practice, which could lead to further beneficial HMO research findings in the future.

5. Concluding remarks

In a review paper published in 2021 [2], we listed approximately 170 HMOs. Since then, additional HMOs present in breast milk in trace amounts have been characterized by advanced LC-MS. Consequently, approximately 200 HMOs structures are presented in the present review. The additional oligosaccharide structures include those containing LacNAc at the reducing ends, which are not classified within the 21 core series in Table S2. As the milks of domestic farm animals, such as cows, goats, and sheep, contain this type MOs, the structural diversity of HMOs should be larger than that discussed in the earlier review.

Research on the biological functions of HMOs is currently the most active area in the glycoscience. The difference in the HMOs profiles between breast milks drunk by breast-fed infants who had NEC or were malnourished and those drunk by the healthy infants have been investigated in some cohort studies. Given that a few HMOs have been shown in some in vivo studies with experimental animals have the potential to prevent NEC or improve the low nutritional status in neonates, these advanced data will be linked to the results obtained in human cohort studies. More recent cohort studies have begun to study the correlation between HMOs profiles in breast milk and cognition and language development in breast-fed infants. Such data will be collected in much larger samples in a variety of area worldwide.

Human mature milk contains 11 g/L of HMOs, whereas the concentration of MOs is >100 mg/L in cow's milk [128], which is used in the manufacturing of infant formulae. The attempt to synthesize several HMOs has been performed by chemical synthetic, enzymatic synthetic, and chemoenzymatic synthetic methods or by fermentation methods using recombinant bacteria. Recently, some HMOs, including 2′-FL, 3-FL, DFL, 3′-SL, 6′-SL, LNT, and LNnT have been manufactured by fermentation methods at plant scale, and infant formulas incorporating 2′-FL and LNnT are commercially available at present. Intervention trials with formula-fed infants given milk substitute with a few HMOs have been recently performed, but more trials are needed in many other areas worldwide.

Other intervention trials with a 4:1 mixture of 2′-FL/LNnT (w/w) were conducted in healthy adults or adult patients who had the irritable bowel syndrome (IBS) characterized by diarrhea or constipation [[180], [181], [182]]. The abundance of Bifidobacteria, especially Bifidobacterium adolescentis increased in the fecal microbiota in both healthy adults and IBS patients. Since a function that promotes the growth of HMOs has not been found in most strains of B. adolescentis [38], future studies should be attempted to elucidate the mechanism by which the amount of B. adolescentis is increased. Overall, the results of studies to date suggest the potential to utilize HMOs as a functional food or for treatment of IBS.

To facilitate the utilization of HMOs as a functional food for infants and adults, it is hoped that additional research studies will be planned.

Abbreviations

A-tetrasaccharide: GalNAcα1–3(Fucα1–2)Galβ1–4Glc

A hexasaccharide: GalNAcα1–3(Fucα1–2)Galβ1–3GlcNAcβ1–3Galβ1–4Glc

B tetrasaccharide: Galα1–3(Fucα1–2)Galβ1–4Glc

DFL (difucosyllactose): Fucα1–2Galβ1–4(Fucα1–3)Glc

DSL (disialyllactose): Neu5Acα2–8Neu5Acα2–3Galβ1–4Glc

DSLNT (disialyllacto-N-tetraose): Neu5Acα2–3Galβ1–3(Neu5Acα2–6)GlcNAcβ1–3Galβ1–4Glc

FDS-LNH-Ⅰ (fucosyl-disialyl-lacto-N-hexaose-Ⅰ): Neu5Acα2–3Galβ1–3(Neu5Acα2–6)GlcNAcβ1–3(Fucα1–2Galβ1–4GlcNAcβ1–6)Galβ1–4Glc

2′-FL (2′-fucosyllactose): Fucα1–2Galβ1–4Glc

3-FL (3-fucosyllactose): Galβ1–4(Fucα1–3)Glc

3′-GL (3′-galactosyllactose): Galβ1–3Galβ1–4Glc

6′-GL (6′-galactosyllactose): Galβ1–6Galβ1–4Glc

LacNAc (N-acetyllactosamine): Galβ1–4GlcNAc

LNB (lacto-N-biose-Ⅰ): Galβ1–3GlcNAc

LNDFH-Ⅰ (lacto-N-difucohexaose-Ⅰ): Fucα1–2Galβ1–3(Fucα1–4)GlcNAcβ1–3Galβ1–4Glc

LNDFH-Ⅱ (lacto-N-difucohaxaose-Ⅱ): Galβ1–3(Fucα1–4)GlcNAcβ1–3Galβ1–4(Fucα1–3)Glc

LNFP-Ⅰ (lacto-N-fucopentaose-Ⅰ): Fucα1–2Galβ1–3GlcNAcβ1–3Galβ1–4Glc

LNFP-Ⅱ (lacto-N-fucopentaose-Ⅱ): Galβ1–3(Fucα1–4)GlcNAcβ1–3Galβ1–4Glc

LNFP-Ⅲ (lacto-N-fucopentaose-Ⅲ): Galβ1–4(Fucα1–3)GlcNAcβ1–3Galβ1–4Glc

LNFP-Ⅴ (lacto-N-fucopentaose-Ⅴ): Galβ1–3GlcNAcβ1–3Galβ1–4(Fucα1–3)Glc

LNH (lacto-N-hexaose): Galβ1–3GlcNAcβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4Glc

LNnH (lacto-N-neohexaose): Galβ1–4GlcNAcβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4Glc

LNT (lacto-N-tetraose): Galβ1–3GlcNAcβ1–3Galβ1–4Glc

LNnT (lacto-N-neotetraose): Galβ1–4GlcNAcβ1–3Galβ1–4Glc

LNTri-Ⅱ (lacto-N-triose -Ⅱ): GlcNAcβ1–3Galβ1–4Glc

LST c (sialyl-lacto-N-tetraose c): Neu5Acα2–6Galβ1–4GlcNAcβ1–3Galβ1–4Glc

novo-LNP-Ⅰ(lacto-N-novopentaose Ⅰ): Galβ1–3(Galβ1–4GlcNAcβ1–6)Galβ1–4Glc

SL (sialyllactose): Neu5Acα2–3Galβ1–4Glc + Neu5Acα2–6Galβ1–4Glc

3′-SL (3′-sialyllactose): Neu5Acα2–3Galβ1–4Glc

6′-SL (6′-sialyllactose): Neu5Acα2–6Galβ1–4Glc

6′-SLN (6′-sialyllactosamine): Neu5Acα2–6Galβ1–4GlcNAc

TF-LNH-Ⅰ (trifucosyl-lacto-N-hexaose-Ⅰ): Fucα1–2Galβ1–3(Fucα1–4)GalNAcβ1–3[Galβ1–4(Fucα1–3)GlcNAcβ1–6]Galβ1–4Glc

Abbreviations of the words in text

Oligosaccharides: Os; HMOs: human milk oligosaccharides; HiMOs: human-identical milk oligosaccharides; NEC: necrotizing enterocolitis; HPLC: high performance liquid chromatography; LC-MS/MS: liquid chromatography with high-resolution tandem mass spectrometry; LC: liquid chromatography; LC-MS: liquid chromatography – mass spectrometry; RBD: receptor-binding domain; CT: Cholera toxin; CTB: Cholera toxin subunit B; IFN-γ: interferon γ; IgG: immunoglobulin G; GBS: group-B Streptococcus; MIC: minimum inhibitory concentration; TEER: transepithelial electrical resistance; milk oligosaccharides: MOs; FITC: fluorescein-isothiocyanate; FD4: fluorescein-isothiocyanate (FITC)-labeled dextran 4 kDa; TNF: tumor necrosis factor; LPS: lipopolysaccharide; DCs: dendritic cells; M1Mϕs: M1 macrophages; CXCL: C-X-C motif chemokine ligand; CCL: chemokine ligand; IL: interleukin; TLR: toll-like receptor; NF: nuclear factor; SHMOs: sialyl HMOs; NLRP: NOD-like receptor; 3′-NT: 3′-nitrotyrosine; DHE: dihydroethidium; OVA: ovalbumin; RANTES: regulated on activation, normal T cell expressed and secreted; sCD40L: soluble tumor necrosis factor-like CD 40 ligand; Ag-IgE: antigen–antibody complex; MLNs: mesenteric lymph nodes; MRI: magnetic resonance imaging; LTP: long-term persistence; SBMOs: sialyl bovine milk Os; TID: type 1 diabetes; HFD: high-fat diet; MWF: myelin water fraction; F6P: fructose-6-phosphate; IBS: irritable bowel syndrome; SCFAs: short-chain fatty acids; IL-1ra: interleukin - 1 receptor antagonist; BMOs: bovine milk-derived oligosaccharides

CRediT authorship contribution statement

Tadasu Urashima: Writing – original draft. Katsumi Ajisaka: Writing – original draft. Tetsuro Ujihara: Writing – original draft. Eri Nakazaki: Writing – original draft.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Te.Uji. and E.N. are current employees of Kyowa Hakko Bio.

Data availability

No data was used for the research described in the article.