Sialic acids are a family of nine-carbon acidic monosaccharides that occur naturally at the end of sugar chains attached to the surfaces of cells and soluble proteins. In the human body, the highest concentration of sialic acid (as N-acetylneuraminic acid) occurs in the brain where it participates as an integral part of ganglioside structure in synaptogenesis and neural transmission. Human milk also contains a high concentration of sialic acid attached to the terminal end of free oligosaccharides, but its metabolic fate and biological role are currently unknown. An important question is whether the sialic acid in human milk is a conditional nutrient and confers developmental advantages on breast-fed infants compared to those fed infant formula. In this review, we critically discuss the current state of knowledge of the biology and role of sialic acid in human milk and nervous tissue, and the link between sialic acid, breastfeeding and learning behaviour.
The rapid growth and development of the newborn infant puts exceptional demands on the supply of nutrients. Any deficit has profound effects on somatic growth and organ structural and functional development, especially the brain. Brain growth, including cell number, organ structural and synaptic connectivity, etc, reaches its peak at 26 weeks gestation and then continues at a high rate throughout the first year of life (Uauy and Peirano, 1999; Uauy et al, 2001). The rapid initial growth of the brain exceeds that of other body tissues. At 6th month gestation, it comprises 21% of total body weight and 15% at term (Friede, 1989). The brain weight more than doubles during the first 9 postnatal months to reach over 90% of the adult weight by the 6th year. Once the time for the critical period of brain growth has passed, it cannot be restarted. The challenge is accentuated in the premature infants, particularly in relation to nutritional support for brain growth.
Recently, neurodevelopmental research has focused attention on the role of long-chain polyunsaturated fatty acids (LCPUFAs), and particularly docosahexaenoic acid (DHA) in improving visual acuity and cognitive ability in preterm infants (Carlson et al, 1996; Gibson, 1999). There are many factors, however, that may support brain growth. One promising new candidate is sialic acid (also known as N-acetylneuraminic acid), a nine-carbon sugar that is a structural and functional component of brain gangliosides and correlates with the amounts of DHA and total long-chain polyunsaturated fatty acids in the ceramide tail of brain gangliosides. Sialic acid may be a conditionally essential nutrient in infancy, if demand outstrips the rate of endogenous synthesis.
Breastfeeding and cognition
Several studies show that children who were breast-fed as babies attain higher scores on intelligence tests than those who were bottle-fed (Rodgers, 1978; Fergusson et al, 1982; Lucas et al, 1992, 1998). On average, scores are 2–9 points higher, a difference that is considered biologically significant. The difference becomes more pronounced as the duration of breastfeeding increases (Dewey et al, 1995). Lucas et al (1992) reported that preterm infants fed human milk in the first month of life have an 8-point advantage in verbal IQ at 7–8 y of age compared with infants fed standard infant formulas. In a large cohort study of several thousand adults, sentence completion, reading ability, and vocabulary were all related to patterns of infant feeding (Richards et al, 1998). In a retrospective study, Menkes (1977) found a significantly greater incidence of bottle feeding among learning-disabled children than among controls being followed for other neurological symptoms. Rodgers (1978) described a large, stratified sample of British children. Covariates included social class, parental interest in education, material home conditions, parental education, family size and birth rank, and age at weaning. After control of confounding variables, there was a significant advantage to breast-fed children on a picture vocabulary test at 8 y of age and on nonverbal ability, mathematics, and sentence completion at 15 y. A few studies have examined reading ability or school attainment, and breast-fed children tended to do better (Ounsted et al, 1984; Rogan & Gladen, 1993; Horwood & Fergusson, 1998). The most recent follow-up study in a New Zealand cohort of 1000 children reported that breastfeeding is associated with small but detectable increases in cognitive ability and academic achievement, extending from 8 to 18 y (Horwood & Fergusson, 1998). The difference was significant after adjustment for social and family history, including maternal age, education, social economic status, marital status, smoking during pregnancy, family living conditions, and family income, and perinatal factors, including gender, birth weight, child's estimated gestational age, and birth order in the family. Not all studies yield significant differences after adjustment (Sandra et al, 1999). A meta-analysis of 20 controlled studies showed that breastfeeding was associated with a 3.2 point higher cognitive development score than formula feeding after adjusting for significant covariates. The IQ advantage increased with duration of breastfeeding, reaching a plateau at 4–6 months of age. Low birth weight infants received the greatest benefits (Anderson et al, 1999). More recently, Mortensen et al (2002) also reported that duration of breastfeeding was associated with significantly higher scores on all components of the Wechsler adult intelligence scale.
There is little direct evidence for a causative mechanism whereby breastfeeding might enhance or, conversely, bottle feeding might impair cognitive growth. Rodgers (1978) suggested possible mediating factors might be differences between breast and bottle milk osmotic load or protein and lipid concentrations or differences in the feeding situation such as infection risk and psychological effects. Menkes (1977) proposed that tyrosinemia due to increased protein levels in formula milk might produce an increased incidence of learning disabilities in bottle-fed children. Improved maternal–child interaction and reduced morbidity could account for the effect or contribute to it (Grantham-McGregor et al, 2000). Interpretation is further complicated by associations between breastfeeding, social status and education of parents. Women who make a decision to breastfeed are often better educated with positive health attitudes concerning immunisation and smoking, and may provide a more desirable environment for their young to develop intellectually (Lucas et al, 1990, 1992). Statistical adjustment for these associations may not remove the full confounding effect of all these factors.
It is highly possible that any benefit from breastfeeding is due to the unique nutritional content of human milk. Since the brain is undergoing rapid development during the first few weeks or months after birth, not only in anatomic terms but also in physiological, biochemical, and psychological parameters, early nutrition may modulate nervous system development. Infant formulas are the sole source of nutrition for infants who are not breast-fed and differences between them may be important. In a randomised multicenter study (Lucas et al, 1990), preterm infants fed a standard term formula for 1 month performed more poorly at 18 months than those given a nutrient-enriched preterm formula. Those infants given the enriched formula had 15 points higher score than those infants given standard term formula. Recently, neurodevelopmental research has focused on the role of LCPUFAs, and particularly, docosahexaenoic acid (DHA) (Makrides et al, 1994, 1995; Lanting & Boersma, 1996). Clinical studies in which infant formula was supplemented with DHA suggested possible improvements in visual acuity and cognitive ability in preterm infants (Carlson et al, 1994, 1996). Infants fed with formula containing DHA had significantly better visual-evoked potential scores than infants fed with control formula without DHA at both 16 and 30 weeks (Makrides et al, 1995). DHA may be the limiting factor in milk formula because the brain, in particular the visual cortex, lays down large amounts in the first year of life. However, if diet is a key to mental development, then there are many factors besides DHA that may be important, such as enzymes, hormones, growth factors, and sialic acid, which are found in human milk but poorly represented in the milk of other species and in infant formulas (Goldman & Garza, 1987). Of these, sialic acid is most intriguing because of its simultaneous presence in large amounts in both human milk and human grey matter.
Structure and state of sialic acid in nature
It is now known that sialic acids comprise a family of 43 naturally occurring derivatives of the nine-carbon sugar neuraminic acid (5-amino-3,5-dideoxy-D-glycero-D-galacto-nonulsonic acid) (Schauer et al, 1995; Schauer & Kamerling, 1997). One branch of the sialic acid family is N-acetylated to form N-acetylneuraminic acids (Neu5Ac, NANA, Sia), which are the most widespread form of sialic acid and almost the only form found in humans (Figure 1). The other branch is based on N-glycolylneuraminic acids (Neu5Gc) which are common in many animal species (best investigated in porcine tissues), but not found in humans except in the case of a particular cancer (Schauer et al, 1995) (Figure 1). Sialic acid molecules can be substituted in more than one position. O-substitution at C4, -7, -8 and -9 (O-acetyl, O-methyl, O-sulphate, and phosphate groups) or the introduction of a double bond between C-2 and C-3 can give rise to a wide variety of possible isomers (Varki & Diaz, 1984). An unusual modification is an additional hydroxyl group present instead of the amino function at position 5 of the sugar, leading to 2-keto-3-deoxy-nonulosonic acid (Ketodeoxynonulosonic acid, Kdn). This component has been found in fish eggs (Caviar) (Varki, 1992).
Sialic acids rarely occur free in nature. They are more commonly present as components of oligosaccharide chains of mucins, glycoproteins, and glycolipids. They usually occupy terminal, nonreducing positions of oligosaccharide chains of complex carbohydrates on outer and inner membrane surfaces in various linkages, mainly to galactose, N-acetylgalactosamine, and other sialic acid moieties, where they are highly exposed and functionally important.
Cells from higher animals and various microorganisms produce sialic acid in a long pathway starting from glucose. An outline of the many reactions involving the sialic acids is shown in Figure 2. All mammals, including human beings, have the capacity to synthesise sialic acid in all tissues from simple sugar precursors. However, in young rats, the activity of the key enzyme UDP-N-acetylglucosamine-2-epimerase in the liver is initially low, rising to a maximum 14–21 days after birth (Dickson & Messer, 1978). The evidence from other animal models, such as mice and guinea pigs, also indicates that enzyme activity is correlated with age and developmental stage (Gal et al, 1997). Thus, we postulate that the human infant liver may also have a limited capacity for synthesising sialic acid during early postnatal life (Wang et al, 2001a, 2001b).
Distribution of sialic acid in nature and the human body
Sialic acids are found in all vertebrates (Mammalia, Aves, Reptilia, Amphibia, Pisces) and within each organism they are present in essentially all tissues (Corfield et al, 1982; Warren, 1994), but not generally found in plants, prokaryotes, or invertebrates (Varki et al, 1999). Certain strains of bacteria contain large amounts of sialic acid in their capsular polysaccharides. The erratic distribution of sialic acids among bacteria and possibly certain protozoa, suggests that the enzymes responsible for their synthesis and metabolism were fortuitously acquired during association with animal cells (Warren, 1994).
The mammalian central nervous system has the highest concentration of sialic acid. The majority (65%) is present in gangliosides and glycoproteins (32%) with the remaining 3% as free sialic acid (Brunngraber et al, 1972). However, gangliosides are not uniformly distributed within the human body (Rosenberg, 1995; Schauer & Kamerling, 1997). Ganglioside concentration in brain grey matter is 15 times that of large visceral organs such as liver, lung, and spleen and 500 times greater than intestinal mucosa (Table 1). Human adult cerebral cortex (particularly frontal cortex) and cerebellar grey matter have approximately 3 times more than the corresponding white matter (Svennerholm, 1980). Our previous study demonstrated that total sialic acid concentration in the human brain was almost two to four times that of eight other animal species. The rank order of decreasing brain sialic acid after humans was rat, mouse, rabbit, sheep, cow, and pig. In a 2-y-old chimpanzee, the brain cortex sialic acid concentration (Wang et al, 1998) was about one-third of the level of a human infant of the same age (Svennerholm et al, 1989). Also, the sialic acid concentration in left lobe of the brain cortex was 22% higher than that of the right lobe (Wang et al, 1998), possibly because the different brain regions perform different neurological functions (Suzuki, 1965). Brain ganglioside sialic acid has implications for evolutionary development and intellectual capacity. The sialic acid moieties of gangliosides and glycoproteins in the frontal cortex play both a structural and functional role and probably participate in a variety of cellular events, such as cell recognition, cell-to-cell contact formation, receptor binding and modulation, immunological properties, and biosignal transduction.
Sialic acids also exist in many human body fluids including saliva, gastric juice, serum, urine, tears, and human milk (Table 2). Free sialic acid is found in urine, particularly in patients with the disease sialuria, where up to 7 g of sialic acid can be eliminated in 1 day (Montreuil et al, 1968). Sialic acid is a significant component in all salivary mucins. Indeed, the name sialic acid is derived from the Greek ‘sialos’ meaning saliva. The structure of the mucin consists of a protein core with varying quantities of attached oligosaccharide side chains. The oligosaccharide component may constitute up to 75% of the total mucin molecule although not every carbohydrate chain carries a terminal sialic acid residue (Corfield et al, 1982). Recently, we demonstrated that breast-fed preterm infants in the first 3 months of life had almost twice the levels of bound sialic acid in their saliva as formula milk-fed preterm infants. Feeding per se did not alter the concentration of sialic acid in saliva, ruling out contamination of saliva by milk. Furthermore, we found no differences in sodium or potassium concentration in the saliva of the two groups, which might otherwise have suggested an inherent difference between the saliva of the two groups (Wang et al, 200la). Similar results were found in 4–5 months old full-term infants (Tram et al, 1997). Sialo-compounds in the body fluids may play a role in the structural and functional protection of the mucosal surfaces. Sialic acid concentrations in body fluids may also reflect metabolic status and body tissue levels. The saliva of pregnant women increases in sialic acid concentration during the course of gestation (Salvolini et al, 1998).
Sialic acid and cognition
Neural cell membranes contain 20 times more sialic acid than other types of membranes, indicating that sialic acid has a clear role in neural structure (Schauer, 1982). Gangliosides in nervous tissue contain sialic acid in the form of glycosphingolipids, which are found in highest concentrations in the cerebral cortex of the human brain (Rosenberg, 1995). It has been proposed that sialo-compounds play a role in the structural and functional establishment of synaptic pathways (Schauer, 1982). Cell membranes containing sialic acid contribute to the negative charge of the membrane (Rosenberg, 1995). Cell aggregation may be prevented by the repulsive effects of the negatively charged sialic acid, as was observed when studying the adhesion of culture cells to their substratum (Schauer & Kamerling, 1997). Cell adhesion may be facilitated via positively charged substances or Ca2+ bridges which break the repulsive effect of sialic acids. Furthermore, binding of Ca2+ to ganglioside sialic acid is of great importance in the function of nervous tissues (Schauer, 1982; Rosenberg, 1995). Gangliosides are localised in clusters on neuronal and especially synaptic membranes in the vicinity of a membrane-bound calcium pump, thus facilitating the supply of these ions for neuronal cells. Calcium–ganglioside interactions may modulate neuronal functions, not only for the short-term process of synaptic transmission of information, but also for long-term events of neuronal adaptations including storage of information.
Brain glycoproteins are also important. The expression of brain function in learning and memory is related to the proposed higher recognition functions of the brain glycoproteins (Bogoch, 1977). The sialoglycopeptides are highly concentrated in the synaptosomal fraction (Garcia-Segura et al, 1978) and there is evidence for involvement of sialic acid containing brain glycoproteins (ependymins) in memory formation (Bogoch, 1977; Schmidt, 1989). In goldfish, ependymin synthesis is preferentially upregulated during learning tasks and treatment with antiependymin antiserum up to 24 h following training inhibits memory formation in these tasks (Schmidt, 1989). The mechanism by which these proteins mediate memory formation is not clear, but the primary structure of ependymin comprises N-glycosylation sites, giving rise to glycoproteins of variable carbohydrate content. Ependymin contains clusters of negatively charged amino acids suitable for binding of calcium ions and synchronously altering neuronal activity (Schmidt, 1989). The carbohydrates in brain protein have been correlated with operant conditioning training of pigeons, higher glycoprotein levels being associated with higher learning ability (Bogoch, 1970).
Sialic acid may be a negative modulator of the interactions mediated by the protein part of neural cell adhesion molecule (N-CAM) (Sadoul et al, 1983). Thus, sialic acid has a possible role in cell-to-cell interaction (Rosenberg, 1995) and has even been postulated to be the actual receptor for neurotransmitters in the central nervous system (von Itzstein & Thomson, 1997).
Sialic acid and infection
Human milk is a rich source of sialic acid containing oligosaccharides. Human milk oligosaccharides (HMOs) have been shown to inhibit microbial pathogens in vitro and in vivo (Newburg, 1999). Harmful bacteria, viruses, and other pathogens use cell surface carbohydrates as sites for recognition and binding to their target host cell, the first step in infection. Oligosaccharide sequences on soluble glycoconjugates such as the mucins can act as ‘decoys’ for microorganisms and parasites (Rosenberg, 1995; Newburg, 1999). Thus, pathogenic organisms attempting to gain access to mucosal membranes might first encounter their cognate oligosaccharide ligands attached to soluble mucins. Upon binding to these sequences, they are swept away by ciliary action, leaving the mucosal cell untouched. In these cases, the host may successfully turn the specificity of the pathogen receptor to its own advantage (von Itzstein & Thomson, 1997; Newburg, 1999). The potential of sialic acids as antimicrobials is clearly enormous. Sialylated oligosaccharides in human milk can act as highly specific receptors for a variety of viruses, bacteria, and parasites as reviewed by Varki (1993). Both free and bound sialylated oligosaccharides in human milk prevent the binding of rotavirus (Varki, 1993) and cholera toxin (Idota et al, 1995) associated with infant diarrhoea, as well as Escherichia coli strains associated with neonatal meningitis and sepsis (Parkkinen et al, 1983). An understanding of the role that these highly functionalised carbohydrates play in the pathogenicity of infectious microorganisms is crucial to the development of infant nutrition.
Breastfeeding has been associated with a lower incidence of morbidity and mortality, especially of enteric disease, otitis media, and respiratory infection (Newburg, 2000). One mechanism is believed to be the HMOs. In the absence of HMOs, as in formula-feeding, pathogens are able to bind to the membrane-bound receptors of their host target cell, adhere to the cell surface and consequently gain entry into the cell to cause infection. In breast-fed infants, pathogens bind to human milk receptor analogs or ‘decoys’ (HMOs) and not to the host cell, thus preventing infection. HMOs may protect the infant from harmful pathogens during the first few months of life, before the immune system is fully mature and before a stable colonic microflora has been established, while providing an environment that stimulates the growth of commensal bacteria. It is not so surprising, perhaps, that human milk contains many agents, including these carbohydrate ‘decoys’ that protect against disease.
Sialic acid in brain gangliosides
Gangliosides are complex glycosphingolipids, which make up 10% of the total lipid mass in the brain and contain different numbers of negatively charged sialic acid moieties. Brain tissue is unique in that the quantity of lipid-bound sialic acid exceeds that of the protein-bound fraction. Gangliosides are hybrid molecules composed of a hydrophilic sialyl oligosaccharide and a hydrophobic ceramide portion that consists of sphingosine and fatty acids (Figure 3) (Svennerholm et al, 1994; Rosenberg, 1995). The ceramide is an N-acetyl-sphingosine in which the acetyl residue is usually a saturated fatty acid with a chain length greater than 14 carbons. C14–C18 predominate in certain sources and C20–C26 in others (Cabezas & Calvo, 1984). Among 100 carbohydrates in nature, only six, that is glucose (Glc), galactose (Gal), N-acetylglucosamine (GlcNAc), N-acetylgalactosamine (GalNAc), fucose (Fuc), and sialic acid (NeuAc or NeuGc), contribute to construction of gangliosides (the same six are used in the construction of HMOs). In addition, most biological properties of glycosphingolipids appear to be more influenced by particular structures of their oligosaccharides than the ceramide composition (Wiegandt, 1994).
At least 100 different ganglioside structures have been established during the last 20 y, but in higher vertebrates 80–90% of the total gangliosides in the brain are GM1, GD1a, GD1b, and GT1b (Rosenberg, 1995). Variation of sialic acid in gangliosides is well documented. All human brain gangliosides contain the Neu5Ac form of sialic acid (Haverkamp et al, 1977). Gangliosides are widely distributed in most vertebrate tissues. For a long time, it was believed that gangliosides were localised in the nerve ending, but more recent work in different laboratories (Ledeen, 1978) suggests that gangliosides are distributed over a large part of the neuronal surface. Human brain ganglioside concentration and distribution change significantly during development. The concentration increases approximately three times from the 10th gestational week to the age of about 5 y. Ganglioside GM1 and GD1a increase 12–15 times during the same period. GT1b is the major ganglioside during the third to fifth gestational month, thereafter its concentration drops rapidly to term, and then increases again up to 50 y of age (Svennerholm et al, 1989). There are also significant regional differences in the gangliosides patterns (Suzuki, 1965; Kracun et al, 1984). Different neuronal and glial cell types contain specific and characteristic sets of gangliosides (Byrne et al, 1988).
Gangliosides and memory formation
The precise functional role of gangliosides is still poorly understood. However, gangliosides have been hypothesised to be involved in the formation of memory. It is generally accepted that the neuronal transmembraneous calcium exchange can be modulated in a complex way by external factors, such as physical parameters (lateral surface pressure and temperature) or chemical parameters such as ion milieu, enzyme activities, hormones, and drugs (Rahmann, 1989). Gangliosides are highly accumulated in a complex composition in the outer leaflet of synaptic membranes. They possess special physicochemical properties particularly in their interaction with calcium. Gangliosides bind calcium ions via electrostatic interactions between the ions and sialic acid residues (Leskawa & Rosenberg, 1981) so they may act as an extracellular storage mechanism for calcium (Romer & Rahmann, 1979). Calcium is required for synaptic transmission and probably activates second messenger pathways to induce synaptic potentiation (Castillo et al, 1994). This storage hypothesis is supported by experimental treatment of frog spinal cord neurons with neuraminidase enzyme. Neuraminidase cleaves sialic acid residues from gangliosides and results in increased postsynaptic activity in response to synapse activation not unlike that observed in long-term potentiation (LTP), presumably because liberated sialic acid residues release bound calcium, which activates transmitter release (Romer & Rahmann, 1979). The addition of calcium to ganglioside monolayers induces changes in surface potential, condensation, surface pressure, and other physical properties, possibly as a result of ganglioside oligosaccharide chains forming cross-bridges with calcium ions (Leskawa & Rosenberg, 1981; Beitinger et al, 1989). Gangliosides can modify the properties of the plasma membrane, for example to increase membrane fluidity (Esmann et al, 1988; Beitinger et al, 1989). Gangliosides also activate growth-stimulating molecules like nerve growth factor, which may mediate this effect (Rosenberg & Noble, 1994).
Neuronal gangliosides also interact with membrane-bound proteins (Esmann et al, 1988). Electron microscopic studies indicate that in the presence of calcium, gangliosides form aggregates around membrane-associated peptides. This finding supports the hypothesis that gangliosides form annuli around synaptic membrane ion channels to regulate movement of ions through the membrane (Rahmann & Rahmann, 1992). The presence of gangliosides increases the conductance of membrane-bound ion carriers, where greater conductance is associated with increased ganglioside complexity and polarity. Application of exogenous gangliosides to the extracellular space influences the excitability of retinal neurons in culture (Fernandes de Lima et al, 1997).
Gangliosides are associated with neuronal growth and development. The cell produces ‘plasma membrane elements’ which contain a high proportion of gangliosides and allow axonal growth (Rosenberg, 1995). Gangliosides are required in areas of membrane extension for axonal and dendritic growth, and applied exogenous gangliosides are inserted into the membrane, increasing both neuritogenesis and synapse formation (Rosenberg, 1995). Lack of sufficient gangliosides, such as observed in the presence of ethanol, is associated with striking loss of neurites (Miller, 1993). Devascularising neural lesions involve a loss of functional synaptic connections between neurons, a process which probably also occurs during ageing. Ganglioside-mediated protection against memory loss in these situations seems likely to be the result of formation of replacement synaptic connections (Rahmann 1995). Figure 4 summarises the mechanisms by which gangliosides could be involved in memory formation.
Exogenously applied gangliosides are taken up by neurons and integrated into the plasma membrane in the same way as endogenous gangliosides (Morgan & Winick, 1981; Dreyfus et al, 1984) and have been shown to influence the activity of a number of types of neural circuits (Agnati et al, 1983; Silva et al, 1999). Recent studies in rodents demonstrate that gangliosides given intraperitoneally can protect against or reverse age- or lesion-induced memory deficits. This effect is observed in animals with various impairments, including genetic, lesion-induced, and age-related cognitive deficits, and is observed in tasks requiring either simple recall or spatial memory from a few hours to several weeks after initial training. The findings of these studies are summarised in Table 3. Gangliosides have been used to treat specific central nervous system lesions. Serum albumin has functions in the transport of gangliosides across the blood-brain barrier (BBB), since measurements of the ability of labelled albumin (albumin adheres to gangliosides) to cross the BBB indicate that it may cross by fluid-phase endocytosis (Poduslo & Curran, 1994; Schengrund & Mummert, 1998). Normal rat pups treated with gangliosides displayed improved learning performances and increased cortex levels of acetylcholine esterase as compared to untreated animals (Karpiak & Mahadik, 1990). On the other hand, some studies showed gangliosides improve memory only in the presence of a specific impairment (Agnati et al, 1983; Fong et al, 1997).
Sialic acid in the diet
Human milk oligosaccharides (HMOs)
The complex free oligosaccharides of human milk consist of glucose, galactose, fucose, N-acetylglucosamine, and sialic acid residues. HMOs are the third largest solid component after lactose and fat (Jensen, 1995). The highest concentration can be found in colostrum (22–24 g/l), but even mature milk contains oligosaccharides at levels of 13 g/l (Coppa et al, 1993). Preterm human milk may contain more than in full-term milk: 26 g/l on day 4 and 17 g/l on day 30 after giving birth (Coppa et al, 1997).
Approximately 88 different HMOs (Newburg & Neubauer, 1995; Chaturvedi et al, 1997; Shen et al, 2000) have been isolated and characterised, with about 50% 38 of 88 being sialylated (Table 4). These have a core structure consisting of a lactose unit at the reducing end and at least one N-acetylneuraminic acid unit at the nonreducing end. Sialyllacto-N-tetraose c (LSTc), 6′-sialyllactoses (6′-SL) and disialyllacto-N-tetraose (DSLNT) are the most representative constituents among sialyl-oligosaccharides, their concentration in human milk being 1050, 590, and 800 mg/l in colostrum and 120, 240, and 630 mg/l at 90 days lactation, respectively (Coppa et al, 1999). They constitute 60–70% of the weight of all sialyl-oligosaccharides (w/w) in human milk (Coppa et al, 1999). All sialyl-oligosaccharides show a gradual decrement as lactation progresses, the concentration at day 90 being about 30–50% of that present in colostrum. However, the content of 3′-sialyllactoses (3′-SL) (Figure 5) in human milk is relatively stable (100–170 mg/l) throughout lactation.
In contrast to human milk, cow's milk contains few free oligosaccharides, with only 0.025 g/l in colostrum and 0.01 g/l in mature milk (Urashima, unpublished data). Sialyl-lactose is the major component (Kunz & Rudloff, 1993). In all, 20 different oligosaccharides in bovine colostrum have been identified so far, with almost 55% of the free oligosaccharides being sialylated (Table 5). Eight structures are the same as human milk sialyl-oligosaccharides (based on Neu5Ac), while three are different, being based on Neu5Gc.
Human milk and cow's milk: quantitative and qualitative aspects
Human milk contains significantly higher levels of sialic acid than bovine milk or any type of infant formulas. However, knowledge of the amino sugar content and distribution in human milk is still limited, partly due to the absence of standardised methods of analysis (Atkinson & Lönnerdal, 1995). We investigated the milk of 20 full-term mothers and 14 preterm mothers (mean gestational age 31±3 weeks) at four stages of lactation (colostrum, transition, 1 and 3 months) and compared with 21 different formulas. A Bio-Gel P-2 column was used to separate sialic acid from milk and a colorimetric method was used to quantitate the amount (Wang et al, 2001b). We found total sialic acid levels were highest in colostrum (3.72±0.15 mmol/l in full term and 4.27±0.15 mmol/l in preterm) and decreased over time. The downward time trend was evident in the full data set and in the 15 full-term mothers and five preterm mothers who had complete data for all timepoints. By 3 months only about 20% of the initial content of sialic acid was present in both groups. Preterm milk contained about 13–23% more total sialic acid than full-term milk during first 3 months lactation. Carlson (1985) determined the sialic acid concentration in human milk at various lactational intervals using a colorimetric method and noted an exponential decay with time. The concentration was highest at 0–2 weeks (approximately 3.56 mmol/l) and dropped to 0.44 mmol/l at 10–28 weeks lactation where it plateaued. Kawakami (1997) assayed 2700 milk samples from over 2000 Japanese mothers at 3 days–16 months postpartum using an HPLC method. Sialic acid content decreased from about 4.85 mmol/l on day 3–5 to 1.03 mmol/l on day 120–240. Using the colorimetric method, Brand Miller et al (1994) found similarly high concentrations of oligosaccharide-bound sialic acid in human milk in the first month of lactation with levels declining by 70% over 3 months. At 1 month lactation, we found a two-fold variation in the milk of full-term mothers (range 1.29–2.62 mmol/l) and a three-fold variation among preterm mothers (range 1.37–4.04 mmol/l). Our previous studies also demonstrated a three-fold difference in levels among 10 mothers (Brand Miller et al, 1994). Sialic acid therefore appears to be one of the most variable fractions of human milk. The reason for the wide range remains unknown, but may represent genetic differences in synthetic capacity or in environmental exposure to infective microorganisms.
In human milk, most sialic acid (ie about 73%) is bound to free oligosaccharides and this proportion remains fairly constant throughout lactation despite an overall decline in absolute amounts. In contrast, infant formulas contain most sialic acid bound to glycoproteins (70%) (Wang et al, 2001b). Carlson (1985) made similar observations. Human milk also contains glycolipid sialic acid at low levels (0.016 mmol/l in colostrum and 0.006 mmol/l in mature milk) (Rueda et al, 1996b; Sanchez-Diaz et al, 1997). This is two times higher than that in cow's milk and infant formulas (Pan & Izumi, 2000). Furthermore, the content of GD3 in human milk was higher in colostrum than in mature milk, and tended to be higher in preterm than in term colostrum. In contrast, the GM3 content was higher in mature milk than in colostrum, and was also higher in term than in preterm milk (Rueda et al, 1996a, 1996b). GD3 is usually detected in developing tissues, whereas GM3 is more abundant in mature tissues (Yu et al, 1989). In cow's milk, GD3 was the major ganglioside followed by GM3. Other gangliosides amounted to no more than 20% of the total ganglioside content (Pan & Izumi, 2000). Both glycoprotein and glycolipid fractions of human milk contain lower levels of sialic acid than that of the free oligosaccharide fraction.
Determination of sialic acid levels in infant formulas established that most sialic acid occurs in the glycoprotein fraction, and that the content of sialic acid in the oligosaccharide fraction is dependent on the whey: casein ratio of the formula (Neeser et al, 1991; Wang et al, 2001b). Bovine mature milk, which is currently used to produce infant formulas has a low sialic acid content of 0.05–0.62 mmol/l (Carlson, 1985; Neeser et al, 1991; Sanchez-Diaz et al, 1997; Wang et al, 2001b), equivalent to less than one-quarter of the level in mature human milk (1 month lactation) (Wang et al, 2001b). Preterm formulas contained the most sialic acid (0.63±0.12 mmol/l) followed by the ‘follow-on’ formulas (0.43±0.03 mmol/l). Soy formulas do not contain any detectable residue of sialic acid (Carlson, 1985). We found soy formulas appeared to contain some sialic acid (0.05±0.003 mmol/l) although this likely represents interfering substances such as quinic acid. Cow's milk-based formulas containing 60% whey (ie whey:casein ratio in protein component=60:40) contained more sialic acid (0.37±0.01 mmol/l) than those containing 20% whey (0.21±0.02 mmol/l). The majority of sialic acid in infant formulas is bound to protein (70%) followed by free oligosaccharides (28%), with only 1% in the free form (Wang et al, 2001b). Thus infants fed bovine milk, bovine-based formula or soybean-derived formulas receive little or no sialic acid, while breast-fed babies receive an early diet rich in sialic acid.
The large amount of sialic acid in human milk compared with formulas based on cow's milk results from major differences in both the amount and types of glycoproteins and oligosaccharides in the milk of the two species. Mature human milk has a whey:casein ratio of 60:40, compared with a ratio of approximately 20:80 in cow's milk. Because casein synthesis is not initiated in the first few days of lactation (Kunz & Lönnerdal, 1990), whey proteins make the largest contribution to total protein at the beginning of lactation, particularly secretory IgA (which contains 12% carbohydrate) and lactoferrin (6% carbohydrate) (Kunz & Lönnerdal, 1990; Rudloff & Kunz, 1997). The carbohydrate side chains of human lactoferrin consist of an N-acetyl-lactosamine glycan with sialic acid in the terminal position (Spik et al, 1982). Cow's milk, however, is low in lactoferrin and the glycan side chains show pronounced differences to human lactoferrin. The glycosylated caseins (eg κ-casein contains 40–60% carbohydrate) are a major source of protein-bound sialic acid in mature human milk (Rudloff & Kunz, 1997). The source of the protein-bound sialic acid therefore changes throughout lactation from predominantly whey proteins in the beginning to the caseins as lactation proceeds. In contrast, there is only 10% carbohydrates in cow's milk κ-casein, but the absolute amount is higher (3.3 g/l) than in mature human milk (1–3 g/l) (Rudloff & Kunz, 1997).
The structure of sialic acid in human milk consists only of Neu5Ac and the dominant N-acetylneuraminic acid is linked to the galactose via an alpha 2 → 6 bond. Cow's milk contains both Neu5Ac and Neu5Gc structures and the dominant galactose is linked by a 2 → 3 bond (Ebner & Schanbacher, 1974). Puente and Hueso (1993) showed that the Neu5Gc content in bovine milk is high in colostrum (32% of the total sialic acid content) and decreases until the end of the first month. The Neu5Gc content in bovine milk gangliosides shows a similar profile. Neu5Gc has never been detected in normal human tissue (Shaw & Schauer, 1988), although it has recently been reported to occur in malignant samples (Rosenberg, 1995). Differences in the chemical structure of sialic acid in human vs cow's milk are likely to influence its bioavailability. The amounts and structures of sialic acid in human milk and cow's milk are summarised in Table 6.
Two types of formula are currently in wide use for regular infant feeding: those based on cow's milk and those formulated from protein isolates and hydrolysates. Most of the protective glycoconjugates and oligosaccharides in human milk are not found in cow's milk or protein isolates. Furthermore, processing excludes the milk fat globule membrane from the final product. Therefore, common infant formulas would not be expected to contain the high levels of glycolipids, glycosaminoglycans, mucins, and other membrane-associated materials found in human milk.
Recently, we analysed 25 brain frontal cortex samples from infants who died of sudden infant death syndrome (Wang et al, 2001c). A total of 12 infants were breast-fed, nine formula-fed, and the rest unknown. We found there was a significant positive correlation between protein-bound sialic acid and age at death in breast-fed infants, but not formula-fed infants. After adjusting for age at death and gender as a covariate, the average concentration of ganglioside-bound and protein-bound sialic acid in breast-fed infants was 32% and 22% higher than that in formula-fed infants (P = 0.013 and 0.01, respectively). Our study therefore implies that nutrition might influence the availability and incorporation of sialic acid into sialic acid containing compounds. Interestingly, we also found that ganglioside-bound sialic acid in all 25 brain samples was significantly correlated with the proportion of DHA, total n-3 PUFA and total LCPUFA in the ceramide tail. Examining the dietary groups separately, the correlations remained significant only in those who were breast-fed, particularly with regard to DHA and AA. The correlations therefore imply that sialic acid and LCPUFA are interdependent building blocks for neural tissues involved in higher cognitive function. Conceivably, both work together to increase the fluidity and functionality of neuronal membranes.
Digestion and absorption of sialic acid
Once digested, the constituents of the glycoproteins and glycolipids provide a source of amino acids, fatty acids, and sialic acid for absorption. At present however, we know little about the mechanisms and extent of digestion of sialic acid containing compounds.
Hydrolysis of sialic acid containing carbohydrates in the intestine is possible because sialidase activity in mucosal homogenates has been shown to be high during the suckling period in several species and positively correlated to the sialic acid content of the respective milks (Dickson & Messer, 1978). Sialic acid appears to be cleaved after uptake into the tissues, probably by Neu5Ac lyase to N-acetylmannosamine+pyruvic acid. N-acetylmannosamine is then used for the resynthesis of sialic acid containing compounds (Schauer, 1982). Sialic acid always occupies the terminal position of milk oligosaccharides, and the bond may be cleaved even if the remainder of the chain resists digestion. Autohydrolysis has also been claimed because sialic acids are relatively strong acids (pKa 2.2–3.0) (Schauer & Kamerling, 1997). Rat intestinal cell walls are highly permeable to free sialic acid. Radioactively labelled forms of sialic acid and sialyl lactose were found to be well absorbed (∼90%) by rat pups, 30% being retained in the body, and 3–4% in the brain after 6 h (Nöhle & Schauer, 1981). However, Brand-Miller et al (1998) demonstrated with breath hydrogen methodology that most HMOs resist digestion in the small intestine of breast-fed infants and undergo fermentation in the colon. Studies in vitro have also confirmed that sialic acid is not released from the incubation of HMOs with pancreatic and mucosal enzyme mixtures (Engfer et al, 2000). The quantities of oligosaccharides in the urine and feces of breast-fed infants are also much higher than those in formula-fed infants (about 9:1 in urine and 62:1 in feces) (Newburg, 2000). Furthermore, oligosaccharides in urine and feces of artificially fed infants have a different pattern from that of breast-fed infants. It is possible that sialidases of bacterial origin cleave sialic acid residues from milk oligosaccharides in the colon. However, it is not known if sialic acid can be absorbed across the colonic mucosa.
The high concentration of sialyloligosaccharides in human milk is intriguing, considering the multitude of developmental changes occurring in the brain and other organs during early infancy. Human milk sialic acid is highest during early lactation, a time when the brain is taking up sialic acid most rapidly (Svennerholm et al, 1989). In addition, deficits in protein or energy in rats during gestation are associated with decreased ganglioside sialic acid in the offspring and poor learning behaviour (Morgan & Naismith, 1982).
Exogenous administration of sialic acid
The effect of repeated administration of sialic acid during periods of maximal brain ganglioside accumulation was the subject of extensive research as early as 20 y ago. Morgan and Winick (1980a) examined undernourished and well-fed rat pups injected intraperitoneally with 1 mg N-acetylneuraminic acid per 50 g body weight daily from 14 to 20 days postnatally. The results showed the administration of sialic acid was associated with an increase in cerebral and cerebellar ganglioside and glycoprotein sialic acid concentration and higher scores in the maze test compared with controls. These changes occurred without affecting brain weight, cell size, cell number and DNA, RNA or protein content. Furthermore, these effects were shown to persist into adulthood (Morgan & Winick, 1980a). Interestingly, early environmental stimulation produced the same effects (Morgan & Winick, 1980b). The authors concluded that since environmental stimulation and administration of sialic acid brought about the same changes in brain sialic acid content, it was possible that brain sialic acid played an important role in learning ability (Morgan & Winick, 1980a, 1980b).
The effect of administration of sialic acid on brain subcellular localisation was also studied by Morgan and Winick (1981). The pregnant rats were either well-fed (200 g casein/kg) or protein restricted (100 g casein/kg) throughout gestation. Each of the low and high protein groups was given their respective diet for the first 11 days of lactation. On day 12 of lactation, 2.5 μCi14 Neu5Ac/kg body weight was injected intraperitoneally into the pups. After 1 h, the brains were analysed for sialic acid in nuclear, myelin-rich, synaptosomal, mitochondrial, and microsomal fractions. The results showed that 80% of the labelled Neu5Ac incorporated into the brains was found in the synaptosomal fraction in both the well-fed and protein restricted animals. The remainder was distributed among the other subcellular fractions in proportion to their total Neu5Ac content. Well-fed animals had levels 50–60% higher incorporation than the protein-restricted group in all fractions. The results suggested that sialic acid exerts its effects on learning behaviour via the synaptic membrane.
Carlson and House (1986) investigated both oral and intraperitoneal administration of sialic acid. Beginning on day 14 of life, the rat pups were administrated sialic acid by intraperitoneal (i.p.) injection or via a feeding catheter with 1.0 mg on days 1 and 2, 1.2 mg on the remaining days, or approximately 20 mg/kg body weight per day for 8 consecutive days. The results showed administration of sialic acid by both routes resulted in significantly more cerebral and cerebellar ganglioside and glycoprotein sialic acid than the control group. Both routes were similarly effective.
Earlier studies have shown that a diet lower in essential fatty acids (EFA) (100 g hydrogenated coconut oil/kg) or deficient in protein decreased the amount of brain ganglioside and glycoprotein as measured by sialic acid concentration (Merat & Dickerson, 1973; Karlsson & Svennerholm, 1978; Morgan et al, 1981). These effects were associated with depressed activities of sialidase and CMP-Neu5Ac synthetase in the brain (Morgan et al, 1981). In addition, pregnant rats fed a diet with low protein or low EFA content produced offspring with altered brain ganglioside distribution and decreased sialic acid content (Berra et al, 1981).
Oral and intravenous administration of radioactively labelled forms of both sialic acid and sialyl-lactose (a trisaccharide) were found to be well absorbed (∼90%) by 20-day-old rat pups, 30% being retained in the body, and 3–4% in the brain after 6 h (Nöhle & Schauer, 1981). Our group demonstrated that 0.23% of a labelled dose of sialic acid was taken up into the brain of 3-day-old piglets by 2 h after intravenous administration (Downing et al, 2001). However, older animals did not show significant incorporation of labelled sialic acid after acute dosing (Nöhle & Schauer, 1981). Morgon and Winick (unpublished results) also found that when sialic acid was injected after 30 days of age (by which time the rat brain had completed its growth), neither behaviour nor brain biochemistry was affected. Because brain growth is time dependent, the effect of administrated sialic acid is likely to be time dependent (Morgan, 1990).
Pregnancy is associated with an increase in concentration of sialic acid in maternal saliva and plasma. Sialic acid increases from about 50 mg/l at 10 weeks gestation to over 150 mg/l at 21–40 weeks gestation in saliva corresponding to the period of rapid sialic acid accumulation in the fetal brain (Salvolini et al, 1998). Plasma increases from 507±93 mg/l at the onset of pregnancy to 637±152 mg/l at term or early postpartum in plasma (Alvi et al, 1988). Briese et al (1999) measured the sialic acid concentration in maternal, retroplacental, and cord blood samples of 126 pregnant women between 28 and 42 weeks of gestation. They showed significant correlations (P<0.01) between maternal and retroplacental blood on the one side and between maternal and the cord blood on the other side. This suggests that the mother synthesises much of the sialic acid, which crosses the placenta to contribute the fetal growth in the third trimester. The high concentration in early human milk (but not infant formula) may provide for continuing need.
Medical significance of disturbances in sialic acid metabolism
Disturbances in the metabolism of sialic acid, either due to genetic error or at the post-translational level, may impair physiological function and lead to disease. A number of rare disorders that involve sialic acid accumulation and deficiency in humans have been described. Deficiencies of sialic acid in brain glycoprotein and ganglioside are related to mental retardation, but severe deficiency has not been described, perhaps because inability to use sialic acid may be lethal (McVeagh & Brand Miller, 1997).
In studies on impairment of human brain and glycoconjugates in congenital athyroidism, protein-bound and ganglioside-bound sialic acid in grey matter was 60 and 11 % lower respectively than that of control group (death from non-neurological disease) (Annunziata et al, 1983). There was a decrease in GD1b, GT1, and GQ1, however GM2, GD3, and GM1 increased compared to controls. These results infer that some types of brain ganglioside (maybe GD1b or GQ1) play a more important role in intelligence. The decrease of glycoprotein sialic acid suggested an impairment of cell maturation (Annunziata et al, 1983).
In schizophrenia, Campbell et al (1967) found the more serious the psychosis, the lower the sialic acid content in the glycoproteins of the cerebrospinal fluid. When the schizophrenic state was treated successfully, the sialic acid content rose to normal values. Edelfors (1981) showed that chronic lithium treatment in animal models of schizophrenia caused changes in the sialic acid content of rat synaptosomes. Edelfors (1981) concluded that the increased content of sialic acid in both glycoprotein and gangliosides in the synaptic membrane may alter electric properties, leading to improvements in schizophrenic patients.
Interference with the synthesis and function of gangliosides and sialoglycoproteins during development appears to contribute to brain dysfunction in phenylketonuria (PKU) (Loo et al, 1985). In the offspring of untreated PKU mothers, there is a noticeable disruption of the normal ganglioside pattern (delayed drop in GQlb and GD3 and slower rise in GM1 and GD1 a) and a significant reduction of sialoglycoproteins.
Alzheimer's disease and older Down's syndrome typically occurs after 65 y of age, but can also affect people in their 40s and 50s (Warren, 1994). Degenerative and progressive, this disease involves the brain, resulting in loss of memory and cognitive function. Alzheimer's patients have been found to have decreased sialyltransferase activity in serum that affects the α-2,3-linked sialic acid in serum glycoproteins. Ganglioside sialic acid content in cerebral cortex was also decreased (Sorbi et al, 1987). It can only be assumed that the defects also occur in neuronal tissue and contribute to the disease process. Serum sialyltransferase may thus be an early biochemical marker of neurodegeneration (Schauer & Kamerling, 1997).
In the rare genetic disease sialuria, there is an overproduction of free sialic acid with massive excretion in the urine (5–7 g/day) (Fontaine et al, 1968; Seppala et al, 1991). In Salla disease, free sialic acid, formed by cleavage of glycoproteins and glycolipids by lysosomal neuraminidase, cannot be transported out of the lysosome where it accumulates (Renlund et al, 1986). Infantile free sialic acid storage disease produces the same defects as in Salla's disease, but is more severe and leads to early death (Mancini et al, 1992).
Adult humans do not express significant levels of Neu5Gc in normal cells, but it has been found in certain human tumors and tumor cell lines (Schauer et al, 1995; Reuter & Gabius, 1996). Spontaneously occurring antibodies to Neu5Gc occur in patients with cancer and with certain infectious diseases, as well as in chickens with Marek's disease, a malignant herpes virus infection (Varki et al, 1999). In tumour metastasis, an increase in the level of cell surface sialylation in certain cell lines has been found to be linked to their metastatic potential (Troy, 1995). The increased sialylation may contribute to the survival of the metastasising cells in the circulatory system, and to adhesion of the cells to the endothelium at a secondary site (von Itzstein & Thomson, 1997).
Sialic acid levels in blood are also used as a marker of the acute-phase response proteins. Some acidic glycoproteins, for example, fibrinogen and haptoglobin, are glycoproteins with sialic acid as the terminal sugar which increase markedly in response to cell injury (Taniuchi et al, 1981). Elevated plasma sialic acid concentration is strongly related to the presence of microvascular complications in type I diabetes (Crook et al, 2001) and cardiovascular morbidity in the general population (Lindberg et al, 1991). In type II diabetes, the circulating sialic acid concentration is elevated in comparison with nondiabetic subjects (Crook et al, 1993). Links between sialic acid and risk factors for vascular disease, such as blood lipids (Wakabayashi et al, 1992), smoking (Lindberg et al, 1991), hyperfibrinogenemia (Crook et al, 1996), and lipoprotein (Kario et al, 1994) have been reported.
The nutritional and biological roles of sialic acids in human milk and other dietary sources (eg fish eggs) are still not fully understood. The higher levels of sialic acid in human milk, together with the higher content in saliva of breast-fed compared with formula-fed infants, suggests that dietary sources are absorbed. The higher level of ganglioside and protein-bound sialic acid in brain frontal cortex of breast-fed infants has important implications. These findings support the hypothesis that an exogenous source of sialic acid in human milk may contribute significantly to greater sialydation of gangliosides and glycoproteins in body fluids, tissues and brain glycoconjugates in breast-fed infants and contribute to the observed neurological and intellectual advantages of breastfeeding over formula feeding. These also have important implications for the formulation of artificial milk for the feeding of newborn infants, particularly preterm infants.