The free carbohydrate components of human milk consist primarily of lactose and a rich repertoire of neutral and acidic oligosaccharides. More than 100 of these free human milk oligosaccharides have been detected and/or characterized by paper chromatography (1), high-performance anion-exchange chromatography (HPAEC) (2) or matrix assisted laser desorption-ionization mass spectrometry (3). Initial interest in these carbohydrates focused on their use as enzyme substrates or inhibitors in the study of carbohydrate metabolism (1), or to determine the specificities of carbohydrate binding proteins on mammalian cell surfaces (4). Recently, certain oligosaccharides have been investigated as bioactive substances that may function as antimicrobial agents (5–8) or immunogens (9,10). However, the role of these compounds in breast-fed infants has not been clarified. For this reason, some of the most intriguing aspects of milk oligosaccharides relate to their synthesis and sole presence in human milk (5). Before their biologic activities in the breast-fed infant can be investigated, milk oligosaccharides must be identified and quantitated using methods that can be applied to large numbers of samples.
Human milk oligosaccharides are synthesized by the glycosyltransferase-catalyzed transfer of monosaccharides from sugar nucleotides to elongating carbohydrate structures. For instance, lactose (Galβ1-4Glc) results from the transfer of galactose from the sugar-nucleotide uridine-5´-diphospho-galactose (UDP-Gal) to glucose (Glc). Its synthesis results in the net export of glucose from the mother's tissues to her milk. This seems to be the key element of a strategy to provide readily available calories to the breast-fed infant. However in humans, lactose is used as the basis for the synthesis of larger and more complex structures. Further elongation of lactose with N-acetylglucosamine (GlcNAc) in a β-1,3 linkage and galactose (Gal) in β-1,3 or β-1,4 linkages results in the two basic core structures, lacto-N-tetraose (LNT, Galβ1-3GlcNAcβ1-3Galβ1-4Glc) and lacto-N-neotetraose (LNnT, Galβ1-4GlcNAcβ1-3Galβ1-4Glc). Larger structures are subsequently synthesized by lengthening or decorating these neutral tetrasaccharides. Many oligosaccharides, which are synthesized only by certain individuals, contain fucose (Fuc) residues (11–13). These carbohydrate structures, which contain antigenic determinants of the Lewis blood group (LBG) and/or secretor systems (1,14), are synthesized by fucosyltransferases that are not distributed uniformly throughout the human population. Examples of structures determined by the secretor status of the donor are 2´-fucosyllactose (2´-FL, Fucα1-2Galβ1-4Glc), and lacto-N-fucopentaose I (LNF-I, Fucα1-2Galβ1-3GlcNAcβ1-3Galβ1-4Glc). These structures are secondary gene products of the fucosyltransferase FucT-II, which is encoded by the “secretor” gene (Se). “Nonsecretor” donors contain two copies of an inactive gene (se) and therefore cannot synthesize 2´-FL or other milk oligosaccharides containing the Fucα1-2 epitope, such as difucosyllactose (DFL; Fucα1-2Galβ1-4[Fucα1-3]Glc) and LNF-I (Table 1). Figure 1 shows potential synthesis pathways for the simplest fucosylated oligosaccharides, which are derivatives of lactose. Because 2´-FL is a prominent oligosaccharide in the milk of secretor donors, its detection constitutes a solid basis for secretor status assignment. The other fucosylated derivatives of lactose are 3-FL (Galβ1-4[Fucα1-3]Glc), which can be synthesized in vitro by several fucosyltransferases, and DFL, which is a hybrid structure containing both α1-2 and α1-3 linkages. The synthesis pathway leading to DFL is tentatively proposed based on results of the transgenic expression of fucosyltransferases in mice (15).
A second fucosyltransferase termed FucT-III is encoded by the Lewis gene, (Le). This enzyme catalyzes the transfer of fucose to the 3 and 4 positions of N-acetylglucosamine. Lewis-positive individuals have at least one copy of the Le gene, whereas Lewis-negative individuals have two copies of the inactive (le) gene. Other enzymes can also transfer fucose in an α1-3 linkage to N-acetylglucosamine, but apparently only FucT-III can transfer fucose in α1-4 linkages. Figure 2 shows the resultant structures of the action of this and other enzymes on the two tetrasaccharide cores (LNT and LNnT). Two fucosylated oligosaccharides in the figure, LNF-II and lacto-N-difucohexaose-I (LNDFH-I), contain antigenic determinants of the LBG system, indicated by the abbreviations Lea and Leb. Other structures, have traditionally been assigned LBG antigen nomenclature and abbreviations such as Lex (LNF-III) and Ley. However, it should be noted that these are not true Lewis antigens because, as stated above, these structures can be synthesized by the action of fucosyltransferases that are not encoded by the Lewis gene. The so-called LBG system antigens are normally found in the carbohydrate components of glycosphingolipids and glycoproteins in red cells and other tissues and biologic fluids such as saliva and plasma (16). However, in human milk they are contained in the free oligosaccharides shown in Figure 2 and Table 1. Different geographic populations have been shown to vary in prevalence of either or both of the enzymes encoded by the Lewis and secretor genes (17–35). Additionally, some oligosaccharides contain at least one residue of the monosaccharide N-acetylneuraminic acid. Structures containing this monosaccharide, which is negatively charged at neutral pH, are termed acidic or sialyloligosaccharides. The present study examined exclusively neutral oligosaccharides.
An initial step in understanding the importance of oligosaccharides in human milk requires an extensive assessment of their quantities and distributions. To date, studies of oligosaccharide concentrations in human milk have been limited to evaluation of pooled milk or milk from only a few donors (1,2). Recent developments in instrumentation and analytical techniques have resulted in precise quantification of many oligosaccharides. Chaturvedi et al. (36) were able to characterize the oligosaccharide content of 50 milk samples from one location using reverse phase chromatography of perbenzoylated derivatives. In the current study, refined sample preparation methods coupled with efficient HPAEC were used to analyze more than 500 human milk samples from 10 countries for nine major neutral oligosaccharides. These structures (Table 1) were chosen because they represent the most prevalent small neutral oligosaccharides found in human milk. Data were compiled and studied for trends in key variables such as the geographic location of the donor and/or the lactation stage of the sample.
MATERIALS AND METHODS
Certified carbonate-free sodium hydroxide solutions were obtained from Ricca (Arlington, TX, U.S.A.). Water was deionized to a conductance of 18.2 mΩ or less (Q System; Millipore Corp., Molsheim, France). Isopropanol (IPA) was purchased from Burdick & Jackson (Muskegon, MI, U.S.A.). Milk oligosaccharide standards were purchased from V-Labs, Inc. (Covington, LA, U.S.A.). All other chemicals were obtained from Sigma (St. Louis, MO, U.S.A.).
Collection and Storage of Human Milk Samples
Milk samples were obtained from the following sites: The Center for Pediatric Research, Eastern Virginia Medical School, (Norfolk, VA, U.S.A.); The University of Chile, Microbiology Program, ICBM, (Santiago, Chile); Department of Infectious Diseases, National Institute of Nutrition, (Mexico City, Mexico); Abbott International Division, (North Chicago, IL, U.S.A.); Abbott Laboratories, Ross Products Division, (Columbus, OH, U.S.A.); and Children's Hospital, (Columbus, OH, U.S.A.). Five hundred forty-nine milk samples were obtained from 435 donors residing in 10 countries: Chile, France, Germany, China (Hong Kong), Italy, Mexico, the Philippines, Singapore, Sweden, and the United States. Information regarding lactation stage was obtained for 492 samples (90%), which permitted them to be categorized as follows: the first 2 days after birth (n = 23), days 3 through 10 after birth (n = 77), days 11 through 30 after birth (n = 222), and more than 30 days after birth (n = 170). The samples from The Center for Pediatric Research, The National Institute of Nutrition, Abbott Labs, and Children's Hospital were collected by electric pump, whereas the Chilean samples were collected manually. No report has been made claiming an effect of sampling technique on the carbohydrate content in milk. All samples represented the entire content of one or both breasts and were frozen immediately. The international and out-of-state samples were shipped on dry ice to Ross Products Division, where they were stored frozen at −70°C until analyzed. The samples collected locally were transported to Ross on ice, where they were stored frozen.
Human milk specimens were considered to be potentially infectious substances (Class 6.2, U. S. Department of Transportation) and were handled accordingly. Milk samples were thawed at room temperature, mixed, and filtered through a 10,000-molecular-weight (MW) cutoff system (Centricon; Amicon, Inc., Beverly, MA, U.S.A.) for 2 hours at 15 °C to 18°C. The retentate, which primarily contained lipids, proteins, and any potentially infectious agents, was discarded as biohazardous material. The clear and colorless oligosaccharide-containing filtrate was stored at −70°C until analyzed. Binding of free carbohydrate to the membrane and filtering apparatus was determined to be negligible by spike and recovery studies. Measures were taken to limit the extent of sample handling and to preserve the integrity of the oligosaccharides. For instance, the use of high temperature, frequently used to treat biohazardous substances, was intentionally avoided.
To resolve the oligosaccharides by HPAEC, a portion of the lactose content was first removed from the filtrate as follows. One hundred microliters of filtrate, diluted to contain 7% IPA, was injected onto a P2 exclusion chromatography column (1 × 30 cm; BioGel P2; Bio-Rad, Inc., Hercules, CA, U.S.A.) equipped with a refractive index detector (Waters, Inc., Milford, MA, U.S.A.) and a fraction collector. The saccharides were eluted with filtered and degassed 7% IPA in water at a flow rate of 1.0 ml/min; 4-ml fractions were collected. With the exception of the void volume (fraction 1), all fractions before the bulk of the lactose were collected and dried overnight (Speed-Vac SC210A; Savant, Holbrook, NY, U.S.A.). The dried samples were then resuspended and pooled in 7% IPA for a total volume of 800 μl. Neutral tri-, tetra-and pentasaccharides were further analyzed.
A series of nine oligosaccharide standards was prepared in water at concentrations of 50, 100, and 200 mg/l: 2´-FL, 3-fucosyllactose (3-FL), DFL, LNT, LNnT, and LNF-I, -II, -III, and -V. To take into account losses incurred by the procedure, standards were subjected to the same treatment as the samples, including the P2 size exclusion chromatography step. In addition, milk samples were spiked with standards to assure that recoveries from the preparation procedure were quantitative.
Analysis by High-Performance Anion-Exchange Chromatography
Oligosaccharide fractions were analyzed by HPAEC (Bio-LC system; Dionex Corp., Sunnyvale, CA, U.S.A.) equipped with a pulsed-electrochemical detector. Two CarboPac PA-1 analytical columns (4 × 250 mm each; Dionex), connected in series by no more than 15 cm of polyetheretherketone tubing and preceded by a guard column (4 × 50mm), were used. Instrument parameters were as follows: detector sensitivity, 0.300 microcoulomb; run time, 75 minutes; peak width, 8.0 seconds; peak threshold, 25.00; peak area reject, 1000; sample injection volume, 20 μl. Oligosaccharides were eluted from the columns using a 5 to 500 mM NaOH linear gradient in 60 minutes. Oligosaccharide identification and concentrations were determined from standard curves.
Some oligosaccharides were not fully resolved from minor components. For example, LNT coeluted with lacto-N-neohexaose (LNnH, Galβ1-4GlcNAcβ1-3[Galβ1-4GlcNAcβ1-6]Galβ1-4Glc) and 3-FL coeluted with the hexasaccharide LNDFH-I (Fucα1-2Galβ1-3[Fucα1-4]GlcNAcβ1-3Galβ1-4Glc). LNDFH-I is produced by women who express the active secretor Fucα1,2-and Lewis Fucα1,3/4-transferases (5,37). Therefore, samples without either 2´-FL or LNF-II could not contain LNDFH-I. In 282 of the 549 samples, 3-FL could be resolved from the contaminant by introduction of brand new columns. The 3-FL concentration was reported only for those samples in which the peak was resolved clearly, either because of the column change or the absence of the interfering structure.
To further verify the position of individual fucosylated oligosaccharides, standards, and oligosaccharide filtrates from human milk samples were each digested with α1,2-fucosidase from Arthrobacter oxidans F1 (Takara–PanVera, PanVera Corp., Madison, WI, U.S.A.). Similar experiments using α1,3/4-fucosidase from Streptomyces sp. 142 (Takara–PanVera) were also performed.
Statistical analysis software (SAS/STAT: Proc Multest; The SAS Institute, Cary, NC, U.S.A.) was used to compare the percentages of the populations containing each oligosaccharide tested. Fisher's exact test option was chosen for calculating the probabilities. All pairwise comparisons of percentages were done at the 0.01 level of significance. Another computer program (SAS/STAT: Proc GLM; The SAS Institute) was used to perform the analyses of variances (ANOVAs) comparing oligosaccharide concentration means. All pairwise mean comparisons were performed at the 0.01 level of significance. A two-way ANOVA, including the factors country of origin and postpartum interval or region and postpartum interval provided the mean comparisons between countries (data not shown) or regions. Region and country were not included in the same ANOVA because the data were not balanced. Region and country may be regarded more as competing and not so much as complimentary ways of classifying data. Only those samples for which the postpartum interval was known were included in the mean calculations presented here. Calculations including those subjects for whom the postpartum interval was not known were also completed, and the effect of including these additional samples was considered negligible. It was decided that the analysis could proceed with postpartum interval as a factor in the ANOVAs (data not shown). Mean values for regions within postpartum interval and vice-versa were compared using the region and postpartum interval ANOVA. The error term for this ANOVA consisted of person-to-person differences within region–postpartum interval (38–40). The error term for the country and postpartum interval ANOVA consisted of person-to-person differences within country and postpartum interval.
Thirty-seven subjects, three from Asia, seven from Europe, and the remainder from the United States donated more than one sample. In these instances only one randomly chosen sample per person was used for statistical calculations. For 54 subjects no lactation interval was identified, and the data from these subjects could not be used to determine trends in oligosaccharide levels through different stages of lactation. Three of these women were from Latin America, eight were from Europe, and the remainder were from the United States. Also, average oligosaccharide concentrations were calculated using only those samples that contained the specific structure; zeros were not included in the average, because in most cases zeros represent the genetic inability to synthesize specific structures.
The chromatographic procedures used in this study allowed for detection and quantitation of nine neutral oligosaccharides ranging in size from tri-to pentasaccharides. The elution profile of a mixture of standards is found in Figure 3A. The HPAEC chromatogram of a milk sample is shown in Figure 3B, and the same sample after enzyme treatment with α1,2-fucosidase from Arthrobacter oxidans F1 is shown in Figure 3C. Peaks designated as 2´-FL (peak E) and LNF-I (peak F) in the human milk oligosaccharide filtrate were eliminated by treatment with fucosidase; fucose (peak J), lactose (peak K), and LNT (peak H) appeared or increased. The DFL peak (D) was diminished slightly by the enzymatic reaction in the human milk sample, a result that was confirmed by digestion of a standard with the same enzyme. The results from the incubation with α1,3/4-fucosidase from Streptomyces sp. 142 resulted in peak eliminations or reductions for 3-FL, DFL, LNF-II, and LNF-III (data not shown).
Occurrence and Quantitation of Human Milk Oligosaccharides
Table 2 shows the percentages of milk samples that contained specific oligosaccharides. Donors were classified by their national origins. Each mean value in the table is shown with a letter code, and those values that share a letter do not differ significantly. Most of these percentage comparisons were weak because of the relatively small samples sizes; no doubt larger sample sizes would have declared more of the percentage differences to be statistically different. Nonetheless, interesting trends can be observed regarding the presence or absence of certain structures. Some residual lactose was detected in all the samples, as were LNnT and LNT or structures derived from these two tetrasaccharide cores. The percentage of the populations that contained LNT-LNnH and LNnT did not differ significantly for any of the countries studied, although differences were observed for some of the fucosyloligosaccharides. For example, 2´-FL was found in 100% of the milk specimens from Mexico (n = 156) and Sweden (n = 7), whereas only 46% of the Philippine samples (n = 22) tested contained this oligosaccharide. As stated earlier, this structure, which contains a Fucα1-2 linkage, is indicative of a subject's secretor status. Overall, 85% of the population tested could be classified as secretor by their milk oligosaccharide profile. On the other hand, no statistically significant variations were found in the content of another blood-group–dependent fucosyloligosaccharide, LNF-II. Although this structure was found in 100% of German and Swedish samples and only 68% of the samples from France, these differences were not significant when subjected to pairwise comparisons. Based on the presence of the Lea antigen (LNF-II) in their milk, 86% of the donors could be classified as Le+. Other fucosylated sugars, LNF-III (containing the Lex antigen Galβ1-4[Fucα1-3]GlcNAcβ1-3-) and 3-FL, also showed no significant differences in percentages between different countries. They were present in more than 96% of all the samples tested. In comparison, LNF-V was present in the lowest percentage of the population (71% overall), although its relatively low concentration and elution position made it one of the more difficult structures to detect.
Mean values for the nine oligosaccharides studied are shown in Table 3. Because some samples had no specific carbohydrate structure, the values in Table 3 were calculated using only those samples that contained the given structure. Also, only those samples for which the lactation stage was known were included. For this reason, the samples from Sweden were not included in any of the further calculations. As has been stated, when multiple samples from the same donor were collected at different stages of lactation, the oligosaccharide concentrations from one randomly selected sample were used to represent such a donor. As in Table 2, those mean values that share a letter are not significantly different. The regional averages, and average of all samples tested, are provided. The most abundant sugar was 2´-FL, with an average of 2.38 g/l. 2´-FL was also one of the most variable, ranging from 0.06 to 4.65 g/l. In those samples without 2´-FL, 3-FL was present at the highest concentration (2.36 g/l). The saccharides found at the lowest levels were LNF-V and LNnT, with overall averages of 0.18 and 0.28 g/l, respectively. They also had the least variable of ranges (0–0.65 g/l for LNF-V, and 0–1.05 g/l for LNnT). The variation in results within a region was much greater than the difference in the mean results between regions, although some region-to-region mean differences were statistically significant. Oligosaccharide concentrations of the different regions showed great variation and overlapped substantially. By comparison, the differences between regions were small, but were in some cases statistically significant. 3-FL showed the most interregional variability, with an average of 1.84 g/l in the United States sample set, and only 0.76 g/l in the Latin American samples. LNT-LNnH, LNF-I, LNF-III, and LNF-V showed little significant differences between regions. Table 3 compares and contrasts values by their geographical locations only; lactation stage was not considered.
In several instances, the countries that comprise a given region showed statistically significant differences in the means of certain oligosaccharides. This occurred most frequently for the Latin American samples. 2´-FL, LNF-I, LNT, and LNnT all varied between samples from Mexico and Chile. Samples from France and Italy varied in LNF-III and LNF-V contents, and samples from China and the Philippines differed significantly in LNF-II averages. The regional mean values were calculated giving each sample within a given region equal weight. Therefore, one country may have more influence over the regional mean because of a larger number of samples. This was the case in the Latin American mean values: Mexico contributed 156 samples and Chile contributed 44. Samples from Asia and Europe were distributed more equally among their respective countries.
Overall Postpartum Trends in Human Milk Oligosaccharides
Table 3 also shows a comparison of the mean results for each oligosaccharide by postpartum interval. Significant variation was found for LNF-I, LNT-LNnH, LNnT, and LNF-V. 2´-FL and 3-FL maintained stable levels throughout the first three postpartum intervals, while 2´-FL decreased in the last interval and 3-FL increased. DFL and LNF-III had their highest mean values in colostrum (0–2 days after birth), and level off afterward. LNF-II was just the opposite: its lowest concentration was found in colostrum, increasing to a higher level for the remainder of lactation. The postpartum interval data are confounded by influences that the country or region of origin may have on oligosaccharide concentrations.
Regional Trends by Postpartum Interval in Human Milk Oligosaccharides
Table 4 outlines the regional effects on the oligosaccharides studied within each postpartum interval. The first interval, 0 to 2 days after birth, is not listed, because all the regions were not represented in this interval. It is interesting to note that during days 3 to 10 after birth, all infants were consuming the same amount of 3-FL regardless of regional origin. For those infants born to mothers who are secretors, the same amount of 2´-FL, DFL, and LNF-I was present in the milk. None of the sugars tested showed great variation among the regions, with at least three regions showing no statistically significant differences for each oligosaccharide. As found in the first interval, during days 11–30 after birth, secretor mothers produce essentially the same amounts of 2´-FL, DFL, and LNF-I throughout the population tested. The same is true for the core sugars LNT-LNnH and LNnT.
The secretor sugars 2´-FL, DFL, and LNF-I, along with LNT-LNnH, are produced at essentially the same concentrations throughout the population in the last postpartum interval, 31 to 452 days. 2´-FL, DFL, LNF-I, and LNT-LNnH are again similar throughout the regions. 3-FL and LNF-II show differences when comparing samples from Latin America with those from Asia or the United States. For LNF-III, samples from Europe and the United States provided similar means, as did those from Asia and Latin America.
The averages obtained for each geographical region may hide underlying differences between values obtained from samples of different countries that comprise the region. For example, samples from China and the Philippines showed the highest number of significant differences, namely 3-FL (for days 3–10 and 11–30 after birth), DFL (days 3–10), and LNF-II (days 3–10 and 31–452). Samples from Mexico and Chile differed in 2´-FL, DFL, and LNnT contents in the last interval studied. A comparison could not be made in the first two intervals, because there were no Chilean samples obtained earlier than 31 days after birth. The 3-FL content from France was not represented in days 3 through 10 or 11 through 30 after birth because of the previously mentioned interference with this structure. LNF-III varied significantly between samples from Italy and Germany and between those of Italy and France for all three intervals studied. Samples from Singapore differed from those from the Philippines in their 2´-FL and LNF-II levels in days 31 through 452 after birth, and with samples from China in their 3-FL (days 11–30 after birth) and 2´-FL (days 31–452 after birth) contents.
Temporal Variation of Human Milk Oligosaccharides
The temporal variations within each region are shown in Table 5. The samples from the United States showed the most stable levels throughout lactation, with slight differences found only in 3-FL, LNF-I, and LNF-V. All other sugars remained at a constant level throughout lactation. LNnT varied throughout the interval in all three of the remaining regions, ranging from 0.1 to 0.55 g/l. LNF-III showed the most consistency, exhibiting no statistical differences in any region. LNF-II also showed a marked stability through lactation in all four regions for those persons capable of synthesizing this sugar. The secretor sugars 2´-FL and DFL remained relatively stable, although another secretor sugar LNF-I varied in samples from Asia and Latin America. 3-FL was maintained at consistent levels through lactation in samples from Europe, Latin America, and the United States, but was quite varied throughout lactation in samples from Asian mothers. Overall, samples from Asia showed the most variation among postpartum intervals. The same between-country differences were seen in this table as in Table 4.
In the present study the simplest neutral oligosaccharides from human milk were quantified in more than 500 samples obtained from different regions of the world. To accomplish this, a simple oligosaccharide extract preparation procedure and an HPAEC-based method were developed. Methods to obtain milk oligosaccharide extracts have been described by several investigators including Kobata (1), Thurl et al. (2,41), and Viverge et al. (42). Most of these procedures involve several steps intended to remove lipids, proteins, and portions of the lactose content. Because most milk proteins and all known infectious agents have molecular weights (MW) higher than 10,000, a single centrifugation-filtration step through a 10,000-MW filter was used to obtain oligosaccharide extracts. Lipids tended to become trapped in the filter, which was an additional advantage of this procedure. Solvents (isopropanol) were used only as a bacteriostatic additive and as an additional safeguard to prevent microbial growth and subsequent sample degradation. Because the oligosaccharides were detected by HPAEC/pulsed electrochemical detection (PED) using a method similar to the one reported by Thurl et al. (2), time consuming derivatization procedures were circumvented (36). Several techniques were used to verify the integrity of the samples and the validity of the analytical methods. Total hexose, monosaccharide, and in some cases, protein levels, were used as internal controls to determine whether milk samples were concentrated, dried, diluted, or degraded as a result of unreported sample handling errors (data not shown). Although the methods of sample collection varied in a few of the regions, no data have been reported showing any resultant effect on the carbohydrate composition of milk. Compared with other methods, such as gel-permeation chromatography (13,43), matrix-assisted laser desorption–ionization time-of-flight mass spectrometry (3), and reverse phase chromatography of perbenzoylated derivatives (36), HPAEC/PED provides a quick and efficient means for quantitating several oligosaccharides from a large number of samples.
The present study focused on obtaining data from a large number of individual samples collected from different donors. Previous studies were conducted using pooled milk or a limited number of specimens. One of the benefits of analyzing many individual samples is that trends of the prevalence and quantity of milk oligosaccharides can be observed. As expected, the samples analyzed from all 435 subjects contained lactose. In addition to lactose, two tetrasaccharide structures, LNT and LNnT, were present in all samples. For example, 99% of the samples contained free LNnT; however, the three samples in which the tetrasaccharide was not detected contained its fucosylated derivative LNF-III (see Figure 2 and Table 1). As discussed above, we expected that certain fucosyloligosaccharides would not be present in all milk samples because carbohydrate structures are secondary gene products resulting from the catalytic activity of glycosyltransferases, and the genes encoding most fucosyltransferases are not uniformly distributed among the human population (20 22,24,30,44–48). Results from the present study reconfirm that the Se gene may be more prevalent in some geographical regions and countries than in others. The chromatographic procedure developed in the present study allows for quick and unequivocal detection of 2´-FL and other secondary products of the Se gene, and consequently reveals the secretor or nonsecretor status of the donor. Because samples from different regions and countries of the world were analyzed, it was possible to observe distribution trends for the Se gene. From the 22 milk samples analyzed from donors from the Philippines, only 10 (46%) contained 2´-FL, which is in sharp contrast to the finding that 100% of 156 randomly collected samples from Mexican donors contained this oligosaccharide. If these human milk oligosaccharides that are secondary products of the Se gene exert positive biologic effects in the breast-fed infant (49), it is a conclusion of the present study that such benefits are not uniformly conferred to breast-fed infants. Independent of this observation, the study of a large number of human milk samples indicates that retrospective clinical studies on the health of breast-fed infants and its correlation with the secretor status of the mother can be performed as long as milk samples and clinical histories of the infants are available.
Antigenic determinants of the LBG and secretor systems vary from tissue to tissue (50–53). The LBG status of a person is normally determined by serologic or saliva tests, which have been used to determine the prevalence of specific LBG in human populations (50,54). The Lewis blood type of a given individual is stated by the Lewis abbreviation “Le” followed by a parenthesis including the letters a and b followed by + or − signs, depending on the presence or absence of Lea or Leb antigens. The classification of the milk samples by their Lewis antigenic determinants resulted in an interesting observation. In serologic testing, three groups are primarily found: Le(a−b−), Le(a+b−), and Le(a−b+). This means that if Leb is detected in a sample, then its precursor, Lea, is not (see Figure 2). Presumably, the Lea antigen is all converted to Leb, resulting in the common LBG type Le(a−b+). Indeed, the intermediate blood group Le(a+b+) has only been reported to occur in populations from Asian and Australian regions (24,45,55). In milk, all LBG antigens were detected; none of the subjects tested could be classified as Le(a−b+) based on oligosaccharide detection. There were no exceptions to this finding; if the Leb antigen, LNDFH-I, was detected in a sample, then LNF-II, the Lea antigen, was also found. Most donors of the Mexican samples were serologically typed for LBG status. Analysis of the oligosaccharide profiles of their milk accurately determined whether the donors were Le(a+b−) or Le(a−b−) in agreement with the serologic results. In each instance, if the milk profile was determined to be Le(a+b+), then the serologic result was Le(a−b+). A third Lewis antigen, Lex, can be synthesized by enzymes other than the one encoded by the Le gene (56). This is demonstrated by the fact that this antigen (see Figure 3) is present in 99% of the samples as LNF-III. This supports the finding that at least another fucosyltransferase is able to synthesize this carbohydrate (57).
With the exception of LNT-LNnH, all other oligosaccharides were not detectable in some milk samples, and broad concentration ranges were found for most of the structures. The presence or absence of certain saccharides affected the concentration of the other structures. For example, the average 3-FL (2.36 g/l) and LNF-II (1.03 g/l) concentrations were 2.8 times more abundant in nonsecretors than in secretors. This could be explained by substrate competition or substrate usage by glycosyltransferases expressed simultaneously in a given donor. For example, nonsecretors cannot transform 3-FL into DFL because they do not express FucT-II, the enzyme encoded by the Se gene. Also, LNT is a substrate for both FucT-II and FucT-III. In the absence of FucT-II (nonsecretor donors) LNF-I, which is the structure synthesized by the action of FucT-III on LNT, cannot be further elongated. When both enzymes are present (secretor, Le+), LNF-II is a substrate for FucT-II, and LNDFH-I is synthesized. This results in a decrease of the level of LNF-II present in the milk. The consistency of these findings throughout the analyzed samples strongly suggested that the quantitation of human milk oligosaccharides was sensitive enough to allow for the recognition of competitive effects of different glycosyltransferase expression in milk donors.
Other oligosaccharide structures may be absent, or present in low amounts, because they are immediately used as substrates by glycosyltransferases. An example is the trisaccharide GlcNAcβ1-3Galβ1-4Glc, which is the precursor of LNT and LNnT. Although concerted efforts were made to detect and identify this structure, its presence could not be established. Taken together, these observations support the notion that the trisaccharide GlcNAcβ1-3Galβ1-4Glc is used as a substrate for the synthesis of LNT and LNnT as soon as it is synthesized, thus, not allowing it to accumulate to any appreciable extent.
Previous reports suggest that glycosidases or glycohydrolases (enzymes which hydrolyze specific monosaccharide linkages) are present in human milk. For example, a fucosidase activity was detected by Wiederschain and Newburg (58). In the present study, several milk samples were analyzed for monosaccharide content by directly chromatographing oligosaccharide filtrates and assessing the presence of monosaccharides at early elution times. No appreciable amounts of fucose were detected in the samples using this method. Variation in concentration could also be attributed to the rate of synthesis and concentration of lactose, which is the basic core for all of the oligosaccharides. It has been reported that the disaccharide varies throughout lactation (59), a fact that could greatly influence the concentrations of lactose-based structures. The production of certain oligosaccharides may also be affected if the glycosyltransferase used in its synthesis is also used to make glycolipids and glycoproteins in the mammary gland (5).
Temporal changes appear to effect the content of certain carbohydrates in milk from within the studied regions, more so than regional changes within a given postpartum interval (Tables 4 and 5). For example, 3-FL varies greatly throughout lactation in the Asian samples, whereas no strong differences among regions were seen in 3-FL for any given postpartum interval. With the possible exception of LNF-II, most of the oligosaccharides studied were present in comparable amounts in each region for the three intervals studied. It is possible that previous reports describing temporal changes in the concentration of oligosaccharides were affected by the small number of samples analyzed or regional variations limited to selected groups of subjects. Coppa et al. (60) and Viverge et al. (42) have published studies of oligosaccharide concentrations through lactation. In these studies several samples from individual donors were analyzed. Concentration trends for certain oligosaccharides were observed. In the present study, the trends throughout lactation were obtained by analyzing one sample from many individual donors. This may explain the discrepancies with their observations. It is important to note that the regions show a difference in the percentage of samples that contain each sugar (Table 2). For example, although the concentration levels of the secondary gene products of the Se gene (2´-FL, DFL, and LNF-I) do not vary as a result of geographic origin, the percentage of individuals expressing them in each population does.
Although the data in Table 3 suggest that overall differences within the regions exist, it must be realized that these data are in fact confounded by the postpartum intervals of each sample. The data from each region are unbalanced. For example, few if any colostrum samples (days 0–2 after birth) were collected from Latin American and U. S. mothers, whereas this interval is adequately represented in the Asian and European samples tested.
The region and postpartum mean comparisons must be regarded as giving indications of oligosaccharide differences in human milk. The estimations of postpartum interval, country, and region differences that are not confounded with one another or by other factors await studies designed to provide the necessary balance in sample collection. Such studies would presumably be more limited in their scope than the current survey, and this account of the variation in human milk oligosaccharides could support their design. This report represents a distillation of several data analysis approaches. The present study is, of necessity, exploratory and is meant to suggest possible differences or trends that exist in human milk oligosaccharide content from different parts of the world. It is hoped that these data are sufficient to lend insight into the planning of future studies designed to further describe the differences in human milk oligosaccharide content. In addition, the study demonstrates that methods and tools are currently available to conduct clinical studies to determine the effect of specific oligosaccharide structures in the breast-fed infant.
Supported in part by National Institute of Child Health and Human Development Grant HD13021 (LKP) from the National Institutes of Health, Bethesda, Maryland, U.S.A.
1. Kobata A. Isolation of oligosaccharides
from human milk
. In: Colowick S, Kaplan N, eds. Methods in Enzymology. New York: Academic Press, 1983: 262–71.
2. Thurl S, Muller–Werner B, Sawatzki G. Quantification of individual oligosaccharide compounds from human milk
using high-pH anion exchange chromatography. Anal Biochem 1996; 235:202–6.
3. Stahl B, Steup M, Karas M, Hillenkamp F. Analysis of neutral oligosaccharides
by matrix-assisted laser desorption/ionization mass spectrometry. Anal Chem 1991; 63:1463–6.
4. Zopf D, Ginsburg V. Carbohydrate antigens of cell surfaces. In: Biology and Chemistry of Eucaryotic Cell Surfaces. New York: Academic Press, 1974: 259–71.
5. Blanc B. Biochemical aspects of human milk
: Comparison with bovine milk. World Rev Nutr Diet 1981; 36:1–89.
6. Cundell D, Weiser J, Shen J, Young A, Tuomanen E. Relationship between colonial morphology and adherence of Streptococcus pneumoniae.
Infect Immun 1995; 63:757–61.
7. Idanpaan–Heikkila I, Simon P, Zopf D, et al. Oligosaccharides
interfere with the establishment and progression of experimental pneumococcal pneumonia. J Infect Dis 1997; 176:704–12.
8. Zopf D, Roth S. Oligosaccharide anti-infective agents. Lancet 1996; 347:1017–21.
9. Kunz C, Rudloff S. Biological functions of oligosaccharides
in human milk
. Acta Paediatr 1993; 82:903–12.
10. Dwek R. Glycobiology: More functions for oligosaccharides
. Science 1995; 269:1234–5.
11. Sabharwal H, Nilsson B, Chester M, Sjoblad S, Lundblad A. Blood group specific oligosaccharides
from faeces of a blood group A breast-fed infant. Mol Immunol 1984; 21:1105–12.
12. Ginsburg V, McGinniss M, Zopf D. Biochemical basis for some blood groups. Immunobiology of the Erythrocyte.
New York: Liss, 1980:45–53.
13. Viverge D, Grimmonprez L, Cassanas G, Bardet L, Solere M. Discriminant carbohydrate components of human milk
according to donor secretor types. J Pediatr Gastroenterol Nutr 1990; 11:365–70.
14. Ginsburg V. Biochemical basis for blood groups in man. In: Biochemistry of the Glycosidic Linkage. New York: Academic Press, 1972: 387–91.
15. Prieto P, Mukerji P, Kelder B, et al. Remodeling of mouse milk glycoconjugates by transgenic expression of a human glycosyltransferase. J Biol Chem 1995; 270:29515–9.
16. Green C. The ABO, Lewis and related blood group antigens: A review of structure and biosynthesis. FEMS Microbiol Immunol 1989; 1:321–30.
17. Welch SG, Barry JV, Dodd BE, et al. A survey of blood group, serum protein and red cell enzyme polymorphisms in the Orkney islands. Hum Hered 1973; 23:230–40.
18. Gibson N, Barclay GP. Blood groups in Zambia. Hum Hered 1973; 23:586–94.
19. Bjarnason O, Bjarnason V, Edwards JH, et al. The blood groups of Icelanders. Ann Hum Genet 1973; 36:425–58.
20. Booth PB, Saave JJ, Hornabrook RW. Lewis and secretor genes in New Guinea. Hum Biol Oceania 1973; 2:155–66.
21. Gershowitz H, Layrisse M, Layrisse Z, Neel JV, Chagnon N, Ayres M. The genetic structure of a tribal population, the Vanomama Indians. II: Eleven blood-group systems and the ABH-Le secretor traits. Ann Hum Genet 1972; 35:261–9.
22. Lincoln PJ, Dodd BE. Variation in secretor and Lewis type frequencies within the British Isles. J Med Genet 1972; 9:43–5.
23. Kornstad L. Distribution of the blood groups of the Norwegian Lapps. Am J Phys Anthropol 1972; 36:257–66.
24. Boettcher B, Kenny R. A quantitative study of Lea
, A and H antigens in salivas of Australian Caucasians and Aborigines. Hum Hered 1971; 21:334–45.
25. Chakraborty R, Das SK, Roy M. Blood group genetics of some caste groups of Southern 24 Parganas, West Bengal. Hum Hered 1975; 25:218–25.
26. Ssebabi EC. Characteristics of African blood. Crit Rev Clin Lab Sci 1975; 6:19–45.
27. Neel JV, Gershowitz H, Spielman RS, Migliazza EC, Salzano FM, Oliver WJ. Genetic studies of the Macushi and Wapishana Indians. II: Data on 12 genetic polymorphisms of the red cell and serum proteins: gene flor between the tribes. Hum Genet 1977; 37:207–19.
28. Chandanayingyong D, Bejrachandra S, Metaseta P, Pongsataporn S. Further study of Rh, Kell, Duffy, P, MN, Lewis and Gerbiech blood groups of the Thais. Southeast Asian J Trop Med Public Health 1979; 10:209–11.
29. Molthan L. Lewis phenotypes of American Caucasians, American Negroes and their children. Vox Sang 1980; 39:327–30.
30. Kobyliansky E, Micle S, Goldschmidt NM, Arensburg B, Nathan H. Lewis blood group and ABH secretor systems in some Jewish populations of Israel. Acta Anthropogenet 1982; 6:133–40.
31. Calderon R, Campillo F, Escudero MC, Gallardo L. Lewis phenotypes and secretor character in the “Castilla la Nueva”-region (Spain). ABH and Lewis antigen levels in salivary secretion. Anthropol Anz 1984; 42:31–9.
32. Clegg EJ, Tills D, Warlow A, Wilkinson J, Marin A. Blood group variation in the Isle of Lewis. Ann Hum Biol 1985; 12:345–61.
33. Kremastinou J, Tzanakaki G, Karafoti PH, Elton RA, Weir DM, Blackwell CC. Distribution of ABO and Lewis blood groups in Greece. Gene Geogr 1996; 10:201–5.
34. Mohan TC, Koo WH, Ng HW. A study of the Lewis blood group system in the Singapore population. Ann Acad Med Singapore 1989; 18:370–4.
35. Nakajima H, Okura K, Huang MC, Saito R, Seto T. The distribution of several serological and biochemical traits in East Asia. IV: The distribution of the blood groups in the Taiwanese mountain aborigines. Jpn J Hum Genet 1971; 16:57–68.
36. Chaturvedi P, Warren C, Ruiz–Palacios G, Pickering L, Newburg D. Milk oligosaccharide profiles by reversed-phase HPLC of their perbenzoylated derivatives. Anal Biochem 1997; 251:89–97.
37. Litscher E, Juntunen K, Seppo A, et al. Oligosaccharide constructs with defined structures that inhibit binding of mouse sperm to unfertilized eggs in vitro. Biochem 1995; 34:4662–9.
38. Hogg, Craig. Introduction to Mathematical Statistics.
3rd ed. New York: Macmillan Publishing Co., 1970.
39. Milliken GA, Johnson DE. Analysis of Messy Data.
New York: Chapman and Hall, 1992.
40. Littell RC, Milliken GA, Stroup WW, Wolfinger RD. SAS System for Mixed Models.
Cary, NC: SAS Institute, Inc. 1996.
41. Thurl S, Henker J, Siegel M, Tovar K, Sawatzki G. Detection of four human milk
groups with respect to Lewis blood group dependent oligosaccharides
. Glycoconj J 1997; 14:795–9.
42. Viverge D, Grimmonprez L, Cassanas G, Bardet L, Solere M. Variations in oligosaccharides
and lactose in human milk
during the first week of lactation. J Pediatr Gastroenterol Nutr 1990; 11:361–4.
43. Thurl S, Offermans J, Mueller–Werner B, Sawatzki G. Determination of neutral oligosaccharide fractions from human milk
by gel-permeation chromatography. J Chromatogr B Biomed Appl 1991; 106:291–300.
44. Koda Y, Soejima M, Liu Y, Kimura H. Molecular basis for secretor type alpha(1,2)-fucosyltransferase
gene deficiency in a Japanese population: A fusion gene generated by unequal crossover responsible for the enzyme deficiency. Am J Hum Genet 1996; 59:343–50.
45. Broadberry RE, Lin Chu M. The Lewis blood group system among Chinese in Taiwan. Hum Hered 1991; 41:290–4.
46. Henry SM, Benny AG, Woodfield DG. Investigation of Lewis phenotypes in Polynesians: Evidence of a weak secretor phenotype. Vox Sang 1990; 58:61–6.
47. Tuomanen E, Towbin H, Rosenfelder G, et al. Receptor analogs and monoclonal antibodies that inhibit adherence of Bordetella pertussis
to human ciliated respiratory cells. J Exp Med 1988; 168:267–77.
48. Thordarson G, Bjarnason O, Lincoln PJ, Dodd BE. ABH secretor and Lewis type frequencies in an Icelandic series. J Med Genet 1972; 9:46–7.
49. Cerantes LE, Newburg D, Ruiz–Palacios G. Human milk
receptor analogs for Campylobacter
and other enteropathogens: Relationship with blood group type determinants (abstract). General Meeting of Microbial Pathogenesis 1997; 97:117.
50. Henry S, Oriol R, Samuelsson B. Lewis histo-blood group system and associated secretory phenotypes. Vox Sang 1995; 69:166–82.
51. Sakamoto J, Furukawa K, Cordon–Cardo C, et al. Expression of Lewisa
, X, and Y blood group antigens in human colonic tumors and normal tissue and in human tumor-derived cell lines. Cancer Res 1986; 46:1553–61.
52. Lloyd K. Blood group antigen expression in epithelial tumors: Influence of secretor status. In: Altered Glycosylation in Tumor Cells. New York: Alan R. Liss, Inc. 1988: 235–43.
53. Greenwell P. Blood group antigens: Molecules seeking a function? Glycoconj J 1997; 14:159–73.
54. Sakamoto J, Yin BW, Lloyd KO. Analysis of the expression of H, Lewis, X, Y and precursor blood group determinants in saliva and red cells using a panel of mouse monoclonal antibodies. Mol Immunol 1984; 21:1093–8.
55. Broadberry RE, Lin M. Comparison of the Lewis phenotypes among the different population groups of Taiwan. Trans Med 1996; 6:255–60.
56. Eppenberger–Castori S, Lotscher H, Finne J. Purification of the N
of human milk
. Glycoconj J 1989; 6:101–14.
57. Johnson P, Watkins W. Separation of an α-3-L-fucosyltransferase
from the blood-group-Le
in human milk
. Biochem Soc Trans 1982; 10:445–6.
58. Wiederschain G, Newburg D. Human milk fucosyltransferase
and α-L-fucosidase activities change during the course of lactation. Nutr Biochem 1995; 6:582–7.
59. Wack RP, Lien EL, Taft D, Roscelli JD. Electrolyte composition of human breast milk beyond the early postpartum period. Nutrition 1997; 13:774–7.
60. Coppa G, Gabrielli O, Giorgi P, et al. Preliminary study of breastfeeding and bacterial adhesion to uroepithelial cells. Lancet 1990; 335:569–71